1 SimGrid (3.19) Released March 20 2018
3 The Moscovitly-cold Spring release.
5 SMPI user-visible improvements:
6 - Support of MPI_Type_create_subarray
7 - Unimplemented MPI calls now abort (instead of issuing a warning)
8 Some harmless calls around errhandlers still only issue a warning.
9 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
10 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
11 is now possible (mmap privatization was refactored to this end).
12 Adding new SMPI instances should also be technically doable now.
14 SMPI internal cleanups:
15 - Removed index notion from SMPI -- no more "getPid() - 1"!
16 In tracing files, actors are called rank-<PROCESS-ID> but these are
17 NOT the real ranks (tracing several communicators is not supported yet).
18 - Deprecate smpi_process_index() (will be removed in 3.22)
19 - Better testing on classical proxy apps, through an external project:
20 https://github.com/simgrid/SMPI-proxy-apps/
23 - Use a graphical TOC to make it easier to find the documentation you need
24 - Revamp the MSG tutorial
27 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
28 - Remove parameter of MSG_process_killall().
29 Resetting the PID was bogus anyway (several actors could have the same PID).
32 - Execution->setHost() can be called after start(), to migrate it.
33 - Comm::test_any() is now implemented.
34 - s4u::Actor now has onCreation() and onDestruction() signals.
35 - Install some kernel header files for the users' plugins and more.
36 Warning, their API is really not stable yet.
37 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
38 include/simgrid/kernel/routing/*Zone.hpp
41 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
43 Build System and other cleanups:
44 - Remove unused run-time parameter "tracing/onelink-only".
45 - SimGrid now works with intel compilers (MC still buggy)
46 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
49 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
52 - #194: Feature request: simgrid::s4u::Comm::test_any()
53 - #245: migrating an actor does not migrate its execution
54 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
55 - #256: Modernize FindSimGrid.cmake
56 - #257: Fix (ab)use of CMake install
58 ----------------------------------------------------------------------------
60 SimGrid (3.18) Released December 24 2017
62 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
64 MSG IS NOW DEPRECATED. Please use S4U for new projects.
65 - Support for MSG should not be removed from SimGrid before 2020,
66 but future evolutions will be limited to the integration of
67 user-provided patches.
70 - s4u::onDeadlock() signal, run before stopping the simulation.
71 - s4u::Exec: asynchronous executions (abstraction of background
73 - s4u::Host->getLoad() returns the achieved speed in flops/s
76 - Link::name() is deprecated and will be removed in v3.21, use
77 Link::getCname() instead.
78 - Mailbox::getName() changed to return a std::string, use
79 Mailbox::getCname() instead to get a char*.
80 - Storage::getName() changed to return a std::string, use
81 Storage::getCname() instead to get a char*.
82 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
83 to have both allocation and deallocation of the map in user space.
86 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
87 The amount of flops remaining to do is a vector, not a scalar.
88 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
89 name implies on both sequential and parallel tasks.
90 - Both changes fix GitHub's #223 using PR #237 as a basis.
91 Thanks Michael Mercier.
92 - Most examples were converted to S4U and hidden elsewhere at tests.
93 Rationale: we still want MSG to work; we want newcomers to use S4U.
96 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
97 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
100 - Switch to the faster dlopen privatization mechanism by default
101 - Documentation improvments
104 - Live migration is getting moved to a plugin. Dirty page tracking is
105 the first part of this plugin. This imply that VM migration is now
106 only possible if one this function is called :
107 - C/MSG: MSG_vm_live_migration_plugin_init()
108 - C/C++: sg_vm_live_migration_plugin_init()
109 - Java: Msg.liveMigrationInit()
110 For C and C++, "simgrid/plugins/live_migration.h" also has to be
114 - Define class simgrid::xbt::Path to manage file names.
115 - Removed unused functions:
116 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
117 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
118 xbt_graph_export_graphviz()
119 - xbt/str.h: xbt_str_join()
120 - Remove unused datatypes:
121 - xbt/heap.h: use std::priority_queue or boost::heap instead
122 - xbt/swag.h: use boost::intrusive::list instead
125 - New link_energy plugin for the consumption of the links.
126 - All of the operations on files and storage contents have been
127 packaged into a plugin (src/plugins/file_system). The current
128 public interface can be found in
129 include/simgrid/plugins/file_system.h
130 To use these functions you now have to initialize the plugin by
131 calling MSG_storage_file_system_init() just after calling
132 MSG_init() or sg_storage_file_system_init() just after creating
136 - Remove the undocumented/untested tag <include>
139 - Remove viva specific tracing as the tool is no longer maintained
142 - #248: Kill of finished processes leads to segfault
143 - #240: xbt_cond_wait_timeout should gracefully return for C apps
144 - #239: please implement signal s4u::onDeadlock()
145 - #230: segfaults when exit() without run()
146 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
147 - #223: MSG_task_get_flops_amount() not working with parallel tasks
148 - #222: Actor::kill() doesn't really kill and segfaults
149 - #221: odd LMM warning when killing an actor
150 - #120: Memory leak when the processes are forcefully killed
152 ----------------------------------------------------------------------------
154 SimGrid (3.17) Released October 8 2017
156 The Drained Leaks release: (almost) no known leaks despite the tests.
158 Even very long simulations will preserve your memory: our testsuite
159 stresses SimGrid for over 45mn on fast machines for a coverage of
160 over 80%, and there is only one single known leak, of about 4kb.
164 - Comm.detach(): start and forget about asynchronous emission. A cleanup
165 handler may be given to free resources if the comm cannot be completed.
166 - this_actor::send(mailbox) is now mailbox->put()
167 - New: simgrid::s4u::Comm::wait_all()
168 - New: Host.execute() for remote executions.
171 - Unused option network/sender-gap is removed.
174 - Deprecate MSG_task_isend_with_matching(): This unused feature
175 really complicates our internals. Will be removed in v3.20.
178 - Improved context termination. It is now possible to free resources, even
179 when a process is forcibly killed.
180 - ContextBoost: add support for Boost versions above 1.61.
183 - Bring back run-time option --cfg=exception/cutpath to remove exception
185 - Removed unused functions:
186 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
187 xbt_str_rtrim(), xbt_str_trim().
188 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
191 - Removed header files obsolete since SimGrid 3.12:
192 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
194 - #3: SD_exit should be made optional
195 - #120: Memory leak when the processes are forcefully killed
196 - #159: Threading test regression in Actor refcounting
197 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
198 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
199 - #186: Actor::killAll() segfaults if some process is blocked on wait()
200 - #191: VM migration and pstate
201 - #192: Updating the energy consumptions of all hosts crashes with VMs
202 - #195: All actors have PID=0 in the logs
203 - #204: Sometimes segfault with thread contexts and mmap privatization
204 - #222: Actor::kill() doesn't really kill and segfaults
205 - #225: Actor::kill() doesn't really kill when victims are doing a join()
207 ----------------------------------------------------------------------------
209 SimGrid (3.16) Released June 22. 2017.
211 The Blooming Spring Release: developments are budding.
214 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
215 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
216 - Java JAR file should be finally fully working on Mac OSX too.
218 XML platforms: Switch to platform v4.1 format.
219 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
220 - <zone> can be used as a synonym for the now deprecated <as>
221 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
222 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
223 - <actor> can be used as a synonym for the now deprecated <process>
224 - state_file and avail_file periodicity is now easier to express
225 (check the documentation) the old behavior should still work.
226 * Storage: not backward compatible, but it seems that nobody ever used it
227 - Remove attribute 'content_type' of <storage_type>: was never used
228 - Make attribute 'model' of <storage_type> optional: for future usage
229 - Remove Bconnection model property: was never (in)validated. Replaced by
230 the maximum of the read and write bandwidth as a resource constraint.
233 - New and Backwards Compatibility break:
234 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
235 When one wants to get the list of tasks whose states have changed during a
236 simulation round, s/he has to allocate and free a dynar and use it as
237 argument to this function. The former SD_simulate (double how_long)
241 - Allow multicore VMs, with the correct sharing (unless you overcommit)
242 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
245 - New function to update the consumption of all hosts at once.
246 - Fix the model for multi-core hosts, linear in the amount of busy
247 cores with an abnormality for pIdle (see paper).
248 BUG: Still not satisfactory for ptask on mono-cores.
251 - The netzone are now available from the MSG API.
252 The old names still work, but are now deprecated.
255 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
256 - New option: smpi/keep-temps to not cleanup temp files
257 - New option : smpi/shared-malloc-blocksize . Relevant only when global shared
258 mallocs mode is used, allows to change the size of the fake file used
259 (default 1MB), to potentially limit the number of mappings for large runs.
260 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
261 - Fortran ifort and flang compilers support
262 - New RMA calls supported (experimental) :
263 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
264 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
265 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
266 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
267 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
268 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
271 - New: this_actor::isMaestro() returns whether we are in kernel mode.
272 - Behind the scene, ActivityImpl are now automatically refcounted.
273 Removing this nasty bottleneck will greatly simplify our internals.
276 - Replay: New function xbt_replay_action_get():
277 Retrieve the function previously associated to an event type.
278 - DROPPED FUNCTION: xbt_str_varsubst()
279 - DROPPED MODULE: strbuff. We don't need it anymore.
280 - DROPPED MODULE: matrix. We don't need it anymore.
281 - DROPPED MODULE: lib. We don't need it anymore.
283 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
285 SimGrid (3.15) Released March 22. 2017
287 The Spring Release: continuous integration servers become green
288 We fixed even the transient bugs on all target architectures:
289 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
290 Windows, FreeBSD, NetBSD.
292 - Rename NetCards to NetPoints.
293 This was intended to help NS3 users, but that's not a netcard.
294 That's a point in the routing algorithm, let's avoid wrong simplifications.
297 - New: MSG_process_yield(). Stop and yield to other processes.
298 - New: MSG_process_daemon(). Daemon processes are automatically killed
299 when the last non-daemon process terminates
300 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
301 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
302 to make room for the upcoming network energy plugin.
303 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
306 - Ensure that an actor can kill itself with Process::exit()
307 - Kill the obscure NativeException. Nobody want to survive the issues
308 it denotes, so use JniException that is a RuntimeException (not to
309 be caught explicitly).
310 - Partial bug fix in initialization. SimGrid flags on command line were
311 consumed at C level but stayed in the original Java String[] args.
312 This could mess users' args[i] if SG flags were not put at the end of
314 The SimGrid flags are now removed from the Java arguments. However,
315 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
316 if args.length is greater than the number of YOUR OWN ARGUMENTS.
317 It might be if you have --log or --cfg flags in the command line.
318 - Fix numerous memleaks all around the place. In particular, around VMs.
322 - s4u::onPlatformCreated: right before the simulation starts
323 - s4u::onSimulationEnd: right after the main simulation loop
324 - s4u::onTimeAdvance: right after a clock change (time is discrete)
325 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
326 event from the availability_file changes the avail speed.
327 - Links are now usable from s4u
328 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
329 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
330 - New examples: The conversion of MSG examples to S4U has begun
331 - Actors: create, kill, migration, and suspend
332 - Applications: master-worker and token-ring
333 - Action replay: communications and storage
334 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
337 - Backwards Compatibility breaks
338 - The SD_route_get_list and SD_route_get_size functions have been removed.
339 They are replaced by the unique
340 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
341 The route from src to dst is built in the links dynar whose size can
342 be obtained with xbt_dynar_length.
343 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
344 sg_host_route_bandwidth and sg_host_route_latency.
345 Macros ensure the backwards compatibility, but you should fix your code
348 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
349 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
350 - Bump our claim of support from MPI 1.1 to MPI 2.2.
351 We don't support 100% of it, but it should be enough. Contact us if not.
352 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
353 naive and probably inaccurate)
354 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
355 mapped onto a unique small file using some system magic.
356 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
357 - Fortran: cleanups, fixes, support of user-added operations
358 - RMA: MPI_Accumulate are applied in correct order.
359 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
360 - Better support for MPI_IN_PLACE
361 - Support for MPI_Win attrs and keyvals.
362 - Support MPI_Comm_group_create, MPI_Type_size_x
363 - MPI ops should be properly applied to all allowed types
364 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
365 algorithm by a non failing one from mpich.
368 - Kill the fifo data container: we don't use it anymore.
371 - A <cluster> can now be created with different speed values to represent pstates.
372 see examples/platforms/energy_cluster.xml for an example.
374 SimGrid (3.14.159) Released December 28. 2016
376 The previous release was only a crude approximation of a Pi release;
377 this one should be a bit better.
379 - Revert a last minute change that broke on Mac OSX.
380 - Fix the changelog and NEWS files.
381 - Improve the documentation, in particular of the routing module.
382 - Integrate some patches flying around in Debian.
384 SimGrid (3.14) Released December 24. 2016
387 * The whole documentation was reworked and reorganized. There is
388 still much room for improvement here, but we're on it.
391 * We now need python3 for our testing infrastructure.
392 * Model checking needs libevent
395 * Now works on FreeBSD
398 * Privatization now works on FreeBSD
400 * Privatization is now activated by default in smpirun
401 You can turn it off with -no-privatize if something goes wrong.
403 * Call-location tracing for SMPI.
404 You can add the exact location (filename / linenumber) of an MPI call to
405 your trace files and slow down or speed up the simulation between two
406 consecutive calls by using an adjustment file (see the documentation).
408 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
409 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
410 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
411 values were ignored in some cases. The timings of these functions can now
412 be significantly different.
414 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
415 smpi/running-power is renamed to smpi/host-speed
417 * smpi/grow-injected-times option to enable or disable multiplication of the
418 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
419 make simulation less precise (but also much faster).
421 * smpirun script should be (much) faster for large deployments.
423 * SMPI tracing : fixed issue with poor matching of send/receives.
425 * Replay : Fix broken waitall
427 New functions and features
428 * MSG_parallel_task_execute_with_timeout, to timeout computations.
430 Dropped / renamed functions and features
431 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
432 - MSG_mailbox_is_empty() -> Mailbox::empty()
433 - MSG_mailbox_front() -> Mailbox::front()
434 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
435 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
436 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
437 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
438 * Don't pass the free_f parameter to property related functions:
439 - MSG_host_set_property_value()
440 - MSG_as_router_set_property_value()
441 - MSG_storage_set_property_value()
442 * VM properties. Since msg_vm_t are msg_host_t, just use
443 MSG_host_get_property_value() and friends
444 * VM I/O related things:
445 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
446 - Unimplemented save/restore methods
447 * MSG_as_router_get_property_value() was redundent with
448 MSG_environment_as_get_property_value().
449 - Removed MSG_as_router_*propert*() functions
450 - Added MSG_environment_as_set_property_value() for consistency
451 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
452 Well, they are still there for now, but deprecated with a warning.
453 Please switch to xbt_dict_new_homogeneous() before this is removed
455 * Task affinity. Its intended behavior (that was very badly tested
456 and probably not really working) was deceiving what most users
457 would have hoped here.
458 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
459 * Option network/coordinates is now useless and should be dropped.
462 * Added option storage/max_file_descriptors to allow more than 1024 files opened
465 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
466 * Added Dragonfly topology support
469 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
470 EiBps, EBps, Eibps, Ebps for bandwidth.
471 They may become useful to some lucky ones.
474 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
475 * ex/app_tokenring: new example, very similar to the MSG Token Ring
476 * ex/async_waitAll: new example, on asynchronous communications
479 * Memory usage should be decreased for simulations with a large number
480 of processes. This also helps for SMPI.
482 SimGrid (3.13) stable; urgency=low
484 The Half Release, a.k.a. the Zealous Easter Trim.
486 - We removed half of the lines, that were mostly experimental cruft.
487 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
488 (not counting blanks and comments -- according to openhub.net)
489 - The internals are now compiled in C++ (and will soon be clean C++)
490 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
491 59 klines of C, etc. We added only 29 klines of C++ in replacement.
493 * Backwards Compatibility breaks
494 - Removed Lua simulation bindings (switch to C or Java for that).
495 Lua can still be used to describe platforms
496 - Removed Java kernel plug-ins.
497 Will be reintroduced after the ongoing major internals reorg.
499 - the following functions were removed.
500 They were too specific and should be reimplemented in a generic
501 way, with filter function.
502 - MSG_task_listen_from_host
503 - MSG_mailbox_get_count_host_waiting_tasks
504 - MSG_mailbox_put_with_timeout was removed.
505 Please use MSG_task_send_with_timeout instead.
507 - the SD_application_reinit function was removed. It has been a noop for a while.
508 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
509 replaced by a more flexible mechanism.
510 - The following functions thus do not exist anymore
511 - SD_workstation_get_access_mode
512 - SD_workstation_set_access_mode
513 - SD_workstation_get_current_task
514 - Basic estimation functions have been removed but can easily be replaced
515 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
516 - SD_workstation_get_computation_time => amount / sg_host_speed()
518 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
519 Use VM.getSpeed()*load/100 for the legacy behavior.
521 - option enable_tracing was removed. It was not doing anything for a while.
522 - In the ModelChecker:
523 - the model-checker now ptraces the model-checked process which means
524 you cannot use a debugger on the latter anymore (we might make this
525 optional in the feature);
526 - removed soft-dirty page tracking;
527 - remove model-checked side snapshot management,
528 MC_snapshot() and MC_compare_snapshot();
529 - keep the MC_cut() function as a stub (it was not really working
530 in the previous release).
533 * All options are consistently in kebab-case. Old names are kept as alias.
536 * Switch to platform v4 format.
537 - Rename from 'power' to 'speed' the attributes describing the amount of
538 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
539 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
540 - In <host> and <link>, attributes availability and state are gone.
541 It was redundent with state and availability traces, and with peak values.
542 - In <cluster>, attributes availability_file and state_file are gone.
543 It was too complex and unused.
544 - Kill <gpu>. Was not doing anything.
545 - The DOCTYPE points to the right URL:
546 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
547 (the file at this address now documents the changelog since its v1)
548 - A warning is emitted for unit-less values (they are still accepted).
549 - speed. Default: 'f' or 'flops'. Also defined:
550 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
551 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
552 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
553 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
554 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
555 - latency. Default: 's' second. Also defined:
556 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
558 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
559 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
562 * s4u::Host is now the preferred public interface to the Host features.
563 sg_host_* functions are C bindings to the exact same behavior
564 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
567 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
568 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
569 They were deprecated since a few years, and probably did not even compile anymore.
572 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
573 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
574 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
575 encouraged to update their codes. A list of the modified functions can be found at the end of
576 include/simgrid/simdag.h
579 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
580 the lambda or closure passed as a parameter will run in kernel mode.
581 All the callback functions should be rewritten to that interface at some point.
584 * Reorganizing and cleaning the internals all around the place.
587 * Remove old default barrier/bcast buggy algorithms (see #18407)
588 * Various bug fixes to handle more codes
589 * Remove the need for the --foreground option of smpirun (it is still
590 accepted for backward compatibility).
593 * Kill the setset data container: MC don't use it anymore.
594 * Kill the queue data container: it made more sense with GRAS.
595 * Kill the xbt_peer_t data type: it's useless without GRAS.
596 * Kill rm_cb feature of config sets: it was never useful.
597 * Kill graphxml parsing feature. It was not used.
598 * Kill the deprecated code protected by XBT_USE_DEPRECATED
600 - xbt_dynar_sort_strings(), when the content is char*
601 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
602 They throw exceptions on invalid input;
603 * C++ support for declaring CLI flags (simgrid::config::Flag);
604 * class for abstracting different signal backends (simgrid::xbt::signal).
605 with no external dependencies (we need very simple signals).
608 * refactoring and cleanup of the code;
609 * ongoing process to cleanly separate the model-checking algorithms
610 from the code model-checking support.
612 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
614 SimGrid (3.12) stable; urgency=low
616 The Facelift Release.
619 * Require g++ v4.7 at least to not speak prehistorical C++.
620 * Require Boost 1.48 (for signal2 component).
621 * Java must be version 7 at least when activated.
622 * Builds on Windows again (including Java bindings).
623 * Tracing is now always enabled (no way to turn it out)
624 * Remove GTNetS. It was not working anyway.
625 * Various cleanups in the cmake scripts.
626 * Move headers around to sort them out on installed systems:
627 - instr/instr.h -> simgrid/instr.h
628 - instr/jedule/* -> simgrid/jedule
629 - simdag/datatypes.h was removed
630 - simdag/simdag.h -> simgrid/simdag.h
631 - msg/datatypes.h was removed
632 - msg/msg.h -> simgrid/msg.h
635 * Interface improvement:
636 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
637 - Sanitize the interface in MSG_task_ module:
638 - Merge two functions that were close enough but misleading:
639 set_compute_duration(t) -> set_flops_amount(t)
640 get_remaining_computation(t) -> get_flops_amount(t)
641 - set_data_size(t) -> set_bytes_amount(t)
642 get_data_size(t) -> get_bytes_amount(t)
643 - Massive cleanups in the functions related to the energy
644 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
645 - New: MSG_host_get_pstate()
647 - msg/energy/onoff: switching hosts on and off
649 * Interface improvement:
650 - Rename Host.isAvail() to Host.isOn()
651 - Rename Process.currentProcess() to Process.getCurrentProcess()
652 - Rename Task.setDataSize() to Task.setBytesAmount()
653 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
655 - #18874: Actually allows the GC to reclaim tasks
659 - SIMIX_process_throw: raises an exception in a remote process
660 * Refactoring: Separate sub-modules
661 - libsmx: the public interface, as libc in a real system
662 - popping: the strange dance that converts a user request into a kernel handling
663 - smx_context_*: the virtualization mechanisms that embed the user code
664 - smx_*: the handling of each simcalls
666 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
667 * Rename smx_action_t into smx_synchro_t, making explicit that these
668 things are used to synchronize processes with their environment.
669 For example, a communication is a sort of synchronization involving
670 the communicating processes (that may block until the exchange) and
671 the platform. The same can be said from computations, etc.
673 - #18888: segfault when a process ends before its kill_time
677 - 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.
678 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
679 - MPI_Type_set_name, MPI_Type_get_name
680 - MPI_*_c2f and MPI_*_f2c functions
681 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
682 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
683 - Activate a lot of new tests from the mpich 3 testsuite
685 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
686 - InfiniBand network model added : Based on the works of Jerome Vienne
687 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
688 - When smpi/display_timing is set, also display global simulation time and application times
689 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
690 * Collective communications
691 - 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
692 - 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/).
693 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
695 - Replay now uses algorithms from wanted collective selector
696 - Replay can be used with SMP-aware algorithms
697 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
698 - 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.
700 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
701 - Correctly use loopback on fat-tree clusters
702 - Asynchronous small messages shouldn't trigger deadlocks anymore
703 * Energy/DVFS cleanup and improvement
704 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
705 - new: smpi_get_host_pstate()
708 - "Full" network optimization flag was broken since Surf++
709 - Better handling of precision flags in maxmin
710 - Fix bug causing sometimes "Impossible" errors
711 - Properly pass cluster properties to included hosts
712 * Improvement of the Energy plugin.
713 - Always update the consumption before returning that value
714 - New property: watt_off to denote the dissipation when the host is off
715 - New functions getWattMinAt and getWattMaxAt to retrieve the
716 dissipation of pstates that we are not currently at.
717 * Java: class NetworkLink renamed to Link
718 * New function: simcall_process_get_kill_time()
719 * Massive rename s/workstation/host/
720 - That's intrusive, but that's good for the project consistency. Sorry.
721 - Change config option "workstation/model" into "host/model"
724 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
725 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
726 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
727 - xbt_dirname and xbt_basename for non-POSIX systems
729 * The model checker now runs as a separate process.
730 * The model checker runs must now be launched with the new simgrid-mc program.
731 * Record/Replay: the MC can display a textual representation of a path in the
732 execution graph. It can then be replayed outside of the model checker.
734 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
736 SimGrid (3.11) stable; urgency=low
741 * Normalizing pointers addresses tool for better diff between logs
744 * Add cloud examples using new VMs
745 - examples/msg/cloud/two_tasks_vm.tesh
746 - examples/msg/cloud/simple_vm.tesh
747 - examples/java/cloud/cloud.tesh
748 - examples/java/cloud/migration/migration.tesh
749 * Add java surf examples:
750 - examples/java/surfPlugin/surf_plugin.tesh
751 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
752 - examples/java/surfCpuModel/surf_cpu_model.tesh
753 * Add SMPI+MSG example:
754 - examples/smpi/smpi_msg_masterslave/
759 - msg task destroy cancel test
760 - msg_host on/off test
761 * Move all tests in testsuite to teshsuite (adding tesh files)
762 * Restructure teshsuites
763 - one folder for each kind of test
764 * Restructure AddTests.cmake
766 - structure the order of tests (with sections)
769 * Add virtual machine
770 - creation of a VM on a PM
771 - migration of a VM from a PM to another PM
773 - MSG_process_join(msg_process_t process, double timeout)
774 - msg_bar_t MSG_barrier_init(unsigned int count)
775 - int MSG_barrier_wait(msg_bar_t barrier)
776 - void MSG_barrier_destroy(msg_bar_t barrier)
777 - msg_as_t MSG_environment_as_get_by_name(const char * name)
778 * New option "msg/debug_multiple_use" to help debugging when a task is used
781 - Improvements and finalization of MSG_storage, MSG_file APIs and their
783 - Increase code coverage in test suites
787 * Protect context stacks against stack overflow. The number of protected memory
788 pages allocated on the top of each stack (1 by default) can be configured
789 with the new command line option --cfg=contexts/guard_size:<value>.
790 * Simcalls are now generated by a python script that
791 - generates files included by SimGrid
792 - checks that all the functions exist, and proposes prototypes
794 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
795 host_set_data, host_get_data
797 - simcall_process_join(smx_process_t process, double timeout)
798 * Fix bug where sleeping processing could not be suspended.
801 * Translate surf models from C to C++
802 - Generic classes for all models: Model, Resource, Action
803 - A generic interface for each kind of model (CPU, Network, Storage
804 Workstation, WorkstationVM)
806 * Translate surf routings from C to C++
807 * Add callbacks using sigc++ or boost::signals2
808 - Add callback functions for resource creation/destruction
809 - Add callback functions for action state change
810 - Handle Energy as a plugin
811 * Replace swag by boost::intrusive
812 * Add new routing models for clusters. For documentation, see
813 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
814 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
815 parameters for cluster tag
816 - Fat trees, with topology="FAT_TREE" and
817 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
818 - see examples/platforms/torus_cluster.xml and
819 examples/platforms/fat_tree_cluster.xml
823 * Hostfiles support host:nb_processes construct to deploy several processes on
825 * Collective communication algorithms should not crash if used with
826 improper number of nodes and report the error.
827 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
828 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
829 MPI_Dims_create, MPI_Cart_sub are supported.
830 * New interface to use SMPI programmatically (still depends on MSG for
831 some parts, see examples/smpi/smpi_msg_masterslave) :
832 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
836 * Global variables privatization in MPI executables is now performed at runtime
837 with the option smpi/privatize_global_variables (default:no).
838 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
839 dynamic libraries loaded by the application are not privatized (static
840 linking with these libraries is advised in this case)
843 * Options defined in XML work correctly now.
846 * New cmake option, enable_lib_in_jar, to control whether native libraries are
847 copied into simgrid.jar or not (ON by default). Use this option if you want
848 to reduce the size of the installed simgrid.jar, *and* the native libraries
849 are kept installed elsewhere.
850 * Surf binding with SWIG (code generated in maintainer mode only):
851 - plugin to handle callbacks
852 - CPU model only for the moment
855 * Supernovae build mode is definitively removed. It was used to improve
856 inlining and inter-module optimizations. It is nowadays superseded by
857 link-time optimizations commonly available in compilers.
858 * Update ns-3 find lib. Bindings for ns-3 should work again now.
859 * Add boost dependency for surf++
860 * Add new macro for tests
861 - ADD_TESH(name <tesh_args>)
862 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
866 - xbt_bar_t XBT_barrier_init(unsigned int count)
867 - int XBT_barrier_wait(xbt_bar_t barrier)
868 - void XBT_barrier_destroy(xbt_bar_t barrier)
869 * Make the xbt_os_time module public
871 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
873 SimGrid (3.10) stable; urgency=low
875 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
878 * Reintegrate Java to the main archive as desynchronizing these
879 package is not acceptable anymore (Java is now considered stable)
880 * Add explicit synchronization facilities through semaphores
881 * Bug fix: Task.setDataSize() only changed the C world, not the value
882 cached in the Java world
885 * Dramatically change the way files are handled. API and internals changed, but
886 this part of MSG was not considered as production grade either.
887 * Add explicit synchronization facilities through semaphores
888 * Add a new function MSG_host_get_process_list()
889 * Preliminary DVFS support (see examples/msg/energy/ for details)
892 * SMPI is now included directly in the libsimgrid as the windows
893 linker doesn't force us on splitting it anymore.
894 * Improvements of the SMPI replay tool:
895 - Most of the collective communications are now rooted in the same process as
896 in the original application.
897 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
898 used until now). Multiple data types can now be used in a trace.
899 - The replay tool now supports traces produce either by TAU or a modified
901 - Bug Fix: the compute part of the reduce action is now taken into account.
902 - Gatherv collective is now supported
903 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
904 of smpirun outputs time independent traces for the current run. One file
905 is created per process. If too many processes are simulated, this behavior
906 can be changed to one file for all processes by using the
907 tracing/smpi/format/ti_one_file flag
908 * smpirun generates the host file if needed (with given host count and platform)
909 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
910 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
911 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
912 --cfg=smpi/coll_selector:(mpich/ompi)
913 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
915 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
916 happening inside a collective SMPI call.
917 * Fix the behavior of complex data types handling.
918 * Make MPI_Wtime another synchronization point to take computations into
920 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
921 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
922 * Add all missing Fortran bindings, SMPI should work with Fortran 90
923 (no privatization of global variables yet)
924 * Preliminary DVFS support (see examples/smpi/energy/ for details)
927 * Verification of liveness properties is now available for SMPI applications
928 (in addition to MSG applications)
929 * Bugged examples using SMPI in examples/smpi/mc/
930 * Add --cfg=model-check/visited option. Allows the verification of infinite
931 programs. Detection of loops in the execution thanks to the system state
932 comparison and reduction of the state space to explore. Can be combined with
933 DPOR for safety properties.
936 * Allow to change SimGrid configuration (see --help) within the code
937 thanks to SD_config() as it can be done in MSG.
938 * Add a new function SD_task_set_amount() upon user request.
941 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
942 * Remove rule based routing (no more PCRE dependency)
943 * Add a limiter_link option to cluster tag, to specify a maximum reachable
944 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
945 * Add a loopback_bw and loopback_lat options to cluster tag.
946 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
947 routing. Now peers are encapsulated in an AS and have their own private
948 router but this is transparent.
951 * Our own implementation of getline is renamed xbt_getline, and gets
952 used even if the OS provide a getline(). This should reduce the
953 configuration complexity by using the same code on all platforms.
954 * New type: xbt_cfg_elm_boolean.
955 * Allow to use yes/no for boolean configuration options in the command line.
956 * Allow to disable SimGrid cleanups at exit from command line option.
957 There are situations where one may want a simulation to end with an exit.
958 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
959 annoying when scripting around the simulator. Adding a
960 --cfg=clean_atexit:no allows to circumvent this issue.
963 * Lots of memory leaks were corrected in this release.
964 There is no known memory leaks anymore, in all of our 600+ tests.
965 * New command line option --version, to get SimGrid version information.
966 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
968 * Supernovae builds are deprecated, and expected to be removed in the next
971 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
973 SimGrid (3.9) stable; urgency=low
975 The Grasgory release: GRAS is really dead now.
977 * Complete overhaul of the internal host structures scheme.
980 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
981 was considered as experimental and badly maintained since then.
982 * Keeping it was thus a trap to our potential users, that could take
983 it instead of MSG or SMPI by mistake despite is pity state.
984 * GRAS seems to have very few users (if any), and no one volunteered
985 to maintain it further. It also induces a lot of XBT code (for
986 portability sake), that must be maintained too.
987 * For all these reasons, we killed GRAS. If someone wants to revive it
988 in the future, don't cry, our git history still remembers of GRAS.
991 * Major overhaul. Merge our documentation again as time proved that
992 splitting it was really not helping our users.
993 * Further improve the developer documentation to help newcomers
994 hacking on SimGrid itself. The user documentation (and in
995 particular, the beginner documentation) is still in a sorry state.
998 * Now works on Windows too!
999 * Much more extensive test suite, from MPICH
1002 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1003 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1004 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1005 * Bug fix: let task be scheduled when the last dependency to be solved is
1006 a control dependency.
1007 * Remove SD_load_environment_script function.
1008 Use the C sg_platf function if you want to declare a platform
1012 * New function: MSG_process_get_number()
1013 * Old function documented: MSG_config()
1014 * Remove MSG_load_platform_script function
1015 Use the C sg_platf function if you want to declare a platform
1019 * Change the default value of the TCP_gamma constant (maximal size of TCP
1020 congestion window) to a more realistic 4MiB value. If you notice changes in
1021 your simulation results, you can fall back to the previous 20k tiny window
1022 by adding --cfg=network/TCP_gamma:20000 on command line.
1023 * (Hopefully) fix a bug wrt periodic availability and state traces
1024 * Bug fix: use default values at start when first event in availability/state
1025 trace is not at time 0.
1028 * remove the "new_" part of function name sg_platf_new_trace_connect
1029 (resulting in sg_platf_trace_connect), since it does not create
1033 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1034 use case where it's really mandatory, and maintaining it was a pain in
1036 * New: xbt_fifo_search(), search an item with a user-provided
1037 comparison function instead of dumb pointer comparison.
1040 * Fix the lua deployment:
1041 Use `simgrid.init_application()` before deployment instead of
1042 `simgrid.msg_register_application()` after.
1045 * Transfer the tracing files into the corresponding modules.
1047 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1049 SimGrid (3.8.1) stable; urgency=low
1051 The "we are told that some people want to also *install* the simgrid
1054 * Add missing file "tesh.1" to the archive.
1056 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1058 SimGrid (3.8) stable; urgency=low
1060 The Psssshiiiit release: SimGrid jumps into the Cloud.
1063 * Add an experimental interface to manipulate VMs. They are mainly
1064 process groups with very few intrinsic semantic, but they should
1065 allow you to build the semantic you want easily.
1066 * New function: MSG_host_set_property_value()
1067 * New function: MSG_process_on_exit(). To clean memory in all cases.
1068 * Bug fixes that made the host (and link) failures unusable.
1069 * Add a way to auto-restart process when the host in which they are
1070 executing comes back (ON_FAILURE="RESTART" on deployment file,
1071 MSG_process_auto_restart_set).
1072 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1073 please stop using the old ones, they are DEPRECATED.
1075 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1076 Please use MSG_init() instead. (reducing the amount of entry
1077 points in the library helps us).
1078 * Make it impossible to link against the wrong version of the lib
1079 * Deprecate MSG_clean(). No need to call it anymore.
1080 * Function MSG_get_host_number() is not deprecated anymore.
1083 * Split the doc into a user guide and a reference guide.
1084 * Start a developper guide to help people hacking on SimGrid.
1087 * Enable tracing by default. This modules rocks you should use it.
1088 * Remove option custom_flags. Now use environment variables CFLAGS
1090 * Use default cmake things to detect lua instead of home grown ones.
1091 * New option "enable_mallocators" to disable mallocators, for debugging
1092 purpose ("on" by default).
1095 * Bug fixes around the resource failures: don't let the processes
1096 survive the host they are running onto.
1097 * Add an interface to auto-restart processes when the host in which they are
1098 executing comes back.
1099 * Ensures that SIMIX_clean is called automatically. It's not part of
1100 the public interface anymore (bindings should be updated).
1103 * Bug fix for when SD_Simulate is called with a positive value: be careful
1104 when comparing doubles. Sometimes they are different for non significant
1106 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1107 parallel task whose initial work is distributed among host according
1108 to the Amdahl's law. Such tasks are created with a parameter alpha
1109 that corresponds to the non-parallelizable part of the computation.
1110 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1111 between two sets of workstations assuming a 1D block distribution (each
1112 workstation owns a similar share of data) on both sides.
1114 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1115 Data redistribution will be automatically scheduled once parent and child
1116 are both scheduled. The filling of computation_amount and
1117 communication_amount structures is now done seamlessly thanks to the chosen
1119 * New function SD_workstation_dump to display various information
1120 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1121 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1122 bandwidth on the route onto which the task is scheduled and the amount of
1124 To divide the nominal bandwidth by 2, the rate then has to be :
1125 rate = bandwidth/(2*amount)
1126 * Compute tasks that have failed can now be rescheduled and executed again
1127 (from their beginning)
1128 * Increasing source code coverage (src/simdag is now covered at 95.8%
1132 * Re-implement time-independent trace replay using SMPI (at the
1133 smpi_smp_* level) instead of MSG. This should replace
1134 examples/msg/actions/actions.c
1135 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1136 hindexed and structs)
1137 * Implement the exchange of non-contiguous data.
1138 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1139 * Correct behavior of smpi/sender_gap and set its default value to 0
1140 * Add option to asynchronously send small messages to allow better
1141 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1142 specifies the size in bytes under which messages will be asynchronously sent.
1143 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1144 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1147 * New C interface to define a platform: XML is now optional.
1148 For more info, please check include/simgrid/platf.h.
1149 * New interface to define random platforms from the C:
1150 For more info, please check include/simgrid/platf_generator.h and
1151 examples/msg/masterslave/masterslave_platfgen.c
1152 * Export a sg_cmdline dynar containing all the arguments we got from
1156 * Two new tracing options for adding comments to trace file so you
1157 can track your experiments (see --help-tracing for details).
1158 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1159 * Adding the SimGrid version that generated the trace file as a comment.
1160 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1161 * Fix to avoid key clashes on Paje links
1162 * Other minor fixes related to the Paje specification
1165 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1166 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1167 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1168 * Greatly improve the robustness of mmalloc to user errors (such as
1169 using an area after freeing it, or freeing it twice)
1171 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1173 SimGrid-java (3.8.1) stable; urgency=low
1175 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1176 Warning: all methods are visible, but only some of them are
1177 implemented so far. Check the source (src/jtrace.c)
1178 for further information.
1179 * New module: org.simgrid.msg.File (SimGrid File management functions)
1180 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1181 * Change the meaning of Process.restart: now restart the process from
1182 the begining, like MSG_process_restart in C.
1183 * Add Process.setAutoRestart: handling of process restart when failed
1185 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1186 you to retrieve the properties of the processes/hosts
1187 * Deprecate Msg.clean(): you can just forget about it now.
1188 * New function Process.getCount(), that only works when compiling
1189 with the not yet released version 3.9 of the C library.
1191 * New context factory based on Coroutines. It mandates a modified JVM
1192 but then, the simulations run about five times faster, and there is
1193 no limit to the amount of processes (beside of the available memory).
1195 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1197 SimGrid (3.7.1) stable; urgency=low
1200 * Restore the prototype of MSG_process_create_with_environment() to
1201 the pre-3.7 situation by removing the kill_time argument.
1202 * Add a MSG_process_set_kill_time() function instead.
1205 * Fix weird behaviors when dealing with parallel tasks.
1208 * Simgrid is now built as a dll.
1209 * Simgrid-java now works on Windows.
1210 * Simgrid-Java is now included into Windows package.
1213 * First pre-build package for MacOSX.
1216 * Fix compilation when using MSG_USE_DEPRECATED.
1217 * Fix some compilation issues on Macs and Windows.
1218 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1219 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1221 We discovered that the Lua console is broken, but we are missing the
1222 manpower to fix it right now. The problem existed in 3.7 too, so we
1223 are not blocking the release for that. Sorry if you depended on this
1224 feature, any help would be really welcome.
1226 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1229 SimGrid-java (3.7.1) stable; urgency=low
1231 The "Java aint got to be bloated and slow" release
1234 * Various internal cleanups and performance improvement
1235 Simulations are expected to run up to twice faster or so
1236 * Make Process.kill(process) an instance method, not a static one
1237 * User processes are not java.lang.Thread subclasses.
1238 This breaks the compatibility (sorry), but previous API was
1239 brain-dead, making it impossible to have non-trivial
1240 initializations in the process constructor.
1241 * Require a full constructor per Process sub-class.
1242 Kinda breaks the compatibility (sorry), but this allows a much more
1243 efficient way to launch the processes at simulation startup.
1244 * Do not embeed our version of semaphores, java 1.5 can be considered
1245 as sufficiently prevalent for us to not dupplicate its features.
1249 * Add examples for almost every part of the API
1250 We spotted and fixed a lot of bugs in the process
1252 * New module: asynchronous communication API
1253 * New function: Process.sleep()
1254 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1255 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1256 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1258 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1260 SimGrid (3.7) stable; urgency=low
1262 The "spring cleanups (before the next Big Project kicks in)" release.
1265 * We can specify the SMPI latency/bandwidth factor with command line
1266 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1267 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1268 You can also use the "config tag" from platform file by adding this line
1269 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1270 (see "example/platforms/tag_config.xml" to use "config tag").
1271 Note that the command line supersedes the platform file configuration.
1272 * Change the correction factors used in LMM model, according to
1273 the latest experiments described in INRIA RR-7821.
1274 Accuracy should be improved this way.
1275 * Use the partial invalidation optimization by default for the
1276 network too. Should produce the exact same results, only faster.
1277 * Major cleanup in surf to merge models and split some optimization
1278 mechanisms from the core of the models. As a result you can now
1279 specify which model to use (e.g., --cfg=network/model:LV08
1280 --cfg=cpu/model:Cas01) and which optimization mode to use
1281 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1282 Incompatible combinations should err at initialization. See
1283 --help-models for the list of all models and optimization modes.
1284 * The CLM03 workstation model was dropped for simplicity because it
1285 used the deprecated CM02 network model. Use default instead.
1286 * Rename the TCP_gamma configuration option to network/TCP_gamma
1287 * Rename the coordinates configuration option to
1288 network/coordinates, and document it
1289 * Use now crosstraffic keyword instead of the terribly misleading
1290 fullduplex keyword. It is activated by default now in the current
1291 default model, use --cfg=network/crosstraffic:0 to turn it off.
1292 * Ongoing refactoring the model parsing to make XML files optional
1293 See include/simgrid/platf.h for details (still to be completed)
1296 * Major overhaul of the documentation. Almost instructive now :/
1297 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1298 functions and friends. This interface was considered as
1299 deprecated since over 2 years, it's time to inform our users that it is.
1300 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1301 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1302 use these (crappy) functions in your code.
1303 These functions will be removed soon. Stop using them now.
1304 * Deprecate MSG_get_host_{table,number}
1305 Implement MSG_hosts_as_dynar() instead.
1306 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1307 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1310 * Stabilize the parallel execution mode of user contexts
1311 * Introduce configuration variables to control parallel execution:
1312 - contexts/synchro: Synchronization mode to use when running
1313 contexts in parallel (either futex, posix or busy_wait)
1314 - contexts/parallel_threshold: Minimal number of user contexts
1315 that must be part of a scheduling round to switch to parallel
1316 execution mode (raw contexts only)
1317 * Fix bugs that prevented to use suspend/resume along with
1318 synchronization structures.
1319 * Fix bugs in process termination that lead to invalid memory access
1320 in very specific conditions.
1323 * Introduce a parallel mode for the models (controlled by surf/nthreads
1324 configuration item). In our tests, running the models in parallel
1325 never lead to any speedups because they are so fast that the gain
1326 of computing each model in parallel does not amortizes the
1327 synchronization costs, even when ultra fast futexes are used.
1328 This is released anyway because YMMV.
1331 * Performance boost by using a swag internally to compute the set of
1332 tasks that are finished and should constitute the return value of
1336 * Enable it by default now that it is considered rather stable.
1339 * Documentation of the tracing functions.
1340 * Performance gains when tracing categorized/uncategorized resource
1341 utilization by avoiding calls to get route when updating resource
1342 variables. LMM constraints are being used instead.
1343 * API changed to set task categories. Use MSG_task_set_category instead
1344 of TRACE_msg_set_task_category, and SD_task_set_category instead
1345 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1346 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1347 * Examples for link user variables added (see at examples/msg/tracing/)
1348 * Deprecated function TRACE_msg_set_process_category completely removed
1349 * Trace header updated according to the latest Paje file format
1350 * Tracing network lazy updates, no longer obligate users to use full updates
1351 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1352 * Experimental: let user code declare/set/push/pop application states for hosts
1353 * API changed to allow the manual creation of graph configuration files
1354 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1357 * Improve the API of Lua MSG bindings, using the Lua spirit.
1358 * Each simulated process now lives in its own Lua world (globals are
1359 automatically duplicated). It helps writing simulators. Will allow
1360 to run Splay programs within SimGrid in the future.
1361 * Add a Chord example in Lua, equivalent to the MSG one.
1364 * Start the implementation of a solution to express temporal
1365 properties, not only local assertions. This is still an
1366 experimental work in progress, stay clear from it to be safe.
1370 - Add new runtime parameters --help-logs and --help-log-categories
1371 to display information about supported logging parameters and
1373 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1374 don't exists anymore.
1375 * Mallocators: allow value NULL for the reset function.
1377 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1378 create homogeneous dictionaries, where all the elements share the
1379 same free function. Non homogeneous dictionaries will be
1380 deprecated in the next release.
1381 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1382 - Multi-level dictionaries are deprecated.
1384 - new function xbt_dynar_search_or_negative() that is useful when
1385 you have less than 2 million elements in your dynar and don't
1386 want of the extra complexity of catching exceptions when the
1387 element is not found.
1389 - Make xbt_os_thread module (for thread portability) public.
1390 Documentation is still to come, sorry.
1392 - Cleanups and simplifications to make it maintainable again.
1393 - Exotic features (such as memalign and valloc) were removed.
1394 - The metadata were extended and improved so that the
1395 model-checker becomes able to explore and inspect the heaps.
1396 - This may induce a performance drop when enable_model-checking is
1397 ON in cmake (even if it's not used in the simulation), but it is
1398 necessary at this point to get MC working.
1400 Turn model-checking OFF if simulation performance matters to you.
1401 Not enabling it at runtime is not enough, disable it in cmake.
1403 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1406 SimGrid (3.6.2) stable; urgency=low
1408 The "Not coding new stuff allows to polish old things" release.
1411 * New bindings to the NS3 packet level simulator (experimental)
1412 * Use the raw (efficient) execution contexts instead of the sysv
1413 (portable) ones when possible.
1414 * libpcre is now mandatory in any cases since not using it led to
1415 severe performance loss and possibly other issues
1416 * Update the XML platforms:
1417 - G5K: include the latest machine in Nancy
1418 - GridPP and LCG: new platforms
1419 * Documentation was partially updated, at least (more to come)
1421 Bug fixes, cosmetics and small improvements
1422 * Free terminated processes before the end of the simulation to avoid
1423 exhausting the memory of users having very dynamic amount of
1425 * Bug fix and cosmetics about canceling non-running tasks
1426 * Bug fix about the dot loader's issues when using libcgraph
1429 * Create an installer for windows with nsis (amd64 and win32)
1430 - Add an hello world project to illustrate simgrid project creation.
1431 - Embed libpcre into the Simgrid installer to avoid
1432 its compilation burden
1433 * The raw execution contexts should work on Apple now
1434 * Port to Windows 64 bits
1435 - Sysv contexts now have an implementation for this arch
1436 - GRAS communication features now support this arch
1437 * Drop support for borland compiler on windows
1438 - this code was not maintained, and we kinda depend on gcc nowadays
1439 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1440 * Fix portability issue on unstable ubuntu: linker became picky on
1443 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1445 SimGrid (3.6.1) stable; urgency=low
1447 The "Oops, we broke Macs too" release
1450 * Fixed contexts detection so that raw ones are used when possible
1451 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1452 leads to a strange error, with user code segfaulting sometimes when
1453 the generated code is not perfectly aligned (which is not
1454 controllable from the user side, depends on the amount of code)
1457 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1459 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1461 SimGrid-java (3.6) unstable; urgency=low
1464 * Split of every thing from simgrid v3.5 into a separate package.
1466 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1468 SimGrid (3.6) stable; urgency=medium
1470 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1473 * Bindings now constitute their own package, separated from the main one.
1474 Rationale: reduce our maintenance nightmare by reducing the module coupling
1475 They will soon be released on their own on gforge.
1477 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1478 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1480 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1481 * It's not quite deprecated for now because we have no replacement,
1482 but it may soon become the case.
1485 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1486 * Fortran: New user-level cache variable to store the rank of the running
1487 process. This improves performance by an order of magnitude.
1488 * C: New coccinelle script to automatically locate and modify global and
1489 local static variables.
1490 * Improved SMPI network model with a sender-side gap to account for multiple
1494 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1495 only say if a communication is finished, no matter whether it succeeded or
1496 failed. You can call MSG_comm_get_status() to know the status of a finished
1498 * New function MSG_task_dsend() to send a task and detach it. When a
1499 communication is detached, you are never notified of its success or failure
1500 and the memory is released automatically once it is finished. This function
1501 is useful when you don't care about the end nor the success of a
1503 * Change the prototypes of action replay. Sorry for inconvenience,
1504 but this is really more efficient this way (and to adapt your code,
1505 you just have to fix the initialization, that shouldn't be too long)
1506 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1507 ever managed to do anything useful with it.
1508 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1509 takes a dynar of communications. It returns immediately and gives the
1510 index of a finished communication (if any).
1511 * New example: a basic implementation of the Chord P2P algorithm.
1514 * New model for multi-core CPUs. You can now use the core attribute to
1515 precise the number of cores of a host. This is a basic model. Every
1516 process running on the host receives at most the power provided in
1517 the DTD (throughput<=power). Total throughput of process cannot exceed
1519 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1520 router linked by an up-link and a down-link (possibly asymmetrical).
1521 This kind of pattern allows to easily build last-mile model style platforms.
1522 Aggregating such patterns in a rule-based AS is thus the technique of
1523 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1524 * New model for Vivaldi routing. We transformed the Vivaldi network model
1525 into a Vivaldi routing model (based on the rule-based model). This allows to
1526 combine Vivaldi based latencies with last-mile platforms.
1529 * Added a check for NaN of IEEE754 infinite in the double entries of
1531 * Introduce a new context factory "raw", highly inspirated from the
1532 ucontext factory, but using manually crafted functions in assembly to
1533 do the work in an efficient manner.
1534 * Allow to change the used context factory at run time, not only at
1535 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1536 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1537 contexts at runtime (only with raw contexts or ucontexts).
1538 * Completely rewrote this module to allow parallel execution of user
1539 processes. Use --cfg=contexts/nthreads:N to execute user processes
1540 with N parallel threads (the default is 1, meaning no parallelism).
1541 * Allow to decide dynamically between sequential and parallel modes.
1542 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1543 processes in parallel only when their number is greater than or equal to P
1545 * Added a check for NaN of IEEE754 infinite in the double entries of
1549 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1550 won't output the state of processes when interrupted with Ctrl-C
1551 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1552 NULL-terminated array. This may solve backward compatibility issues
1553 due to the change to return type of SD_simulate. See also:
1554 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1555 * Add new macros with variable number of arguments.
1556 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1557 - in xbt/asserts.h: xbt_assert
1558 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1559 - in xbt/ex.h: THROWF and RETHROWF.
1560 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1562 * Change xbt_die() to accept a format string with arguments, just like printf.
1563 * New data structure: xbt_lib_t, like a dict but more general and with better
1567 * New configuration options
1568 Options triva/categorized and triva/uncategorized can be used to generate
1569 graph configuration files for Triva visualization tool.
1570 * Configuration option tracing/platform is renamed to tracing/categorized
1571 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1572 * New TRACE_user_link_variable interface:
1573 User provides the name of the link and the tracing variable to attach to it
1574 * the declaration of tracing categories must be done after the environment
1576 * simpler tracing interface, just one way to declare categories
1577 TRACE_category or TRACE_category_with_color, it is up to you
1578 * links in the trace file are again identified by their names
1579 * trace contains the full platform hierarchy exactly as declared using the ASes
1580 * Options tracing/msg/[task|process]:1 groups the process by hosts
1581 for both cases, tasks and processes must have names that are unique during the simulation
1582 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1583 * The experimental option tracing/msg/volume is deprecated
1584 its functionality may be reincorporated if needed
1586 The tracing generates a trace file with unordered timestamped events,
1587 because of the way the core simulator (surf) works. A script available
1588 at the tools directory (fix-paje-trace.sh) can be used to put the events
1589 in order. We have changed the tracing so it can generate ordered timestamped
1590 events in the final trace, but depending on the simulator (and how much time
1591 is simulated) that can lead to a huge memory utilization. It is deactivated
1592 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1594 Build Infrastructure
1595 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1596 - We are trying hard to keep the API stable, but it may happen that
1597 some things change (we're a research project after all, not a
1598 nuclear plant operating system). If such things should happen, you
1599 could rely on that macro to adapt.
1600 - current value: 30600 for 3.06.00, aka 3.6
1601 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1602 a number that can be compared with SIMGRID_VERSION.
1603 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1604 assertions and verbose/debug logging events are disabled at compile time.
1606 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1608 SimGrid (3.5) stable; urgency=medium
1611 * New feature to any SimGrid-based simulator: Model-Checking
1612 Check SIN#1 for more details.
1615 * New Model SMPI (three-interval linear regression for correction factors)
1616 See RR-7426, available at http://hal.inria.fr/inria-00527150
1617 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1618 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1619 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1620 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1621 * See also src/smpi/README
1626 - Tracing API changes: TRACE_start and TRACE_end should not be called
1627 by user-code. They are automatically called by simulators created
1628 with SimDAG, MSG and SMPI if the toolkit is compiled with
1629 tracing_enabled=ON. Categories declaration and utilization remain the
1630 same for MSG and SimDag.
1631 - A function was added to the tracing API to declare categories with
1633 - TRACE_category_with_color (char *category, char *color)
1634 where color must be in the following format
1635 "%f %f %f", red, green, blue
1636 and red, green, blue are float values in the interval [0, 1]
1637 - User can specify NULL as color parameter, or continue calling
1638 TRACE_category (cat)
1639 On that case, the tracing system will define random colors
1640 - The following command-line options are supported:
1641 --cfg=tracing/filename:msg.trace
1642 --cfg=tracing:1 (activate tracing, needed to use others)
1643 --cfg=tracing/platform:1 (categorized resource use)
1644 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1645 --cfg=tracing/msg/task:1 (task creation)
1646 --cfg=tracing/msg/process:1 (process creation, migration)
1647 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1648 --cfg=tracing/smpi:1 (SMPI interface tracing)
1649 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1650 - examples of examples/msg/tracing updated
1652 - DAXLoader and DOTLoader functions can generate tasks with categories
1653 - A new function to attribute a category to SD tasks:
1654 TRACE_sd_set_task_category (SD_task_t task, char *category)
1655 * Tracing the MPI interface implemented by SMPI
1656 - Collective operations are traced with states
1657 - Point-to-Point operations are traced with states/links
1658 - Tracing activated by a parameter "-trace filename" passed
1659 to smpirun during execution (considering that simgrid
1660 is compiled with tracing enabled)
1661 - To run the simulation with gdb, the simulator
1662 accepts --cfg=tracing/smpi:1 to trace SMPI
1663 - tesh files to check if smpi tracing is ok
1664 - See examples/smpi/NAS/DT-trace
1665 * GTNetS tracing re-worked
1666 - adaptation to the tracing system of GTNets to cope
1667 with modifications regarding the fullduplex mode
1668 - new tesh files to check if gtnets tracing is ok
1671 * Asynchronous communications through the functions:
1672 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1673 * New function: MSG_load_platform_script()
1674 to make possible using a lua script instead of XML files to set up platforms
1675 * New function: MSG_set_function
1676 to associate functions to processes, used when bypassing the parser
1677 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1679 Platforms: Add some more examples in examples/platforms
1680 * Grid'5000: see www.grid5000.fr
1681 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1684 * Change the XML format. This is a very important modification. SimGrid
1685 3.5 introduces a new hierarchical format based on the notion of
1686 Autonomous Systems. Compatibility with old format is ensured
1687 through the perl script provided in the install bin directory
1688 bin/simgrid_update_xml.
1689 It is now possible to build platforms with specific routing
1690 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1691 connect several platforms together. We will try to provide soon
1692 set of realistic platforms exploiting these properties (have a
1693 look at examples/platforms/ for the moment).
1694 * Take the opportunity of the XML format change to be a good XML citizen:
1695 rename link:ctn to link_ctn and similar changes (also dealt with by
1697 * Add a new routing scheme (rule-based) using regular expressions. It
1698 enables to have an extremely low memory footprint when the
1699 underlying routing is simple and can be compactly described. You
1700 need to have libpcre4-dev (perl regular expressions) installed if
1701 you want to use this routing scheme.
1702 * Revive the cluster TAG and allow to easily and efficiently (both in
1703 term of memory and speed) connect clusters together. Have a look
1704 at teshsuite/simdag/platforms/ to see how this can be done. With
1705 this tag, you can create clusters with thousands of tasks at no
1706 cost (have a look at examples/platforms/).
1707 Note: clusters are implemented as ASes, so there is no need for an
1708 enclosing AS tag if you have only one cluster in your platform.
1709 * Add new generic functions in the public interface that allows the user
1710 to call SURF 'create_resource' methods from your code (same
1711 functionality as the XML bypass mechanism but with a much lighter
1713 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1714 that uses a piecewise linear approximation to produce better
1715 results when exchanging small messages.
1716 * Add a new parameter to handle correctly full duplex link and account
1717 for interferences between uplink and downlink communications
1718 (activate with --cfg=fullduplex:1).
1721 * Rename the SD_READY (all dependencies are satisfied and task is
1722 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1723 dependencies are satisfied) state.
1724 This prevents a confusion between the notion of "ready to schedule"
1725 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1726 simulated" (SD_RUNNABLE) used by the simulation kernel.
1727 * Change the way a task is considered as ready. Instead of removing
1728 dependencies when a task is done, a counter is decreased. This way,
1729 it is always possible to reach ancestors thanks to the
1730 SD_taks_get_parents function (even after the end of the simulation.)
1731 * Change the return type of SD_Simulate from (SD_task_t*) into
1732 xbt_dynar_t. This function was in handling a dynar internally and
1733 converted it into a NULL terminated array for historical reasons.
1734 * New function SD_dotload(char*) to load a DAG described in dot
1735 format. This loader and the corresponding examples require the
1736 installation of the graphviz library.
1737 * Fix a bug in the management of tasks of size 0 in the surf network
1738 models. This problem was only visible with SIMDAG and you should
1739 thus disregard results produced with earlier versions if you
1740 relied on this feature (some tasks were blocked because of this).
1741 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1742 even though your applications were doing only point-to-point
1743 communications and sequential computations. Now you can really use any
1744 model you want (of course, if you create real parallel tasks, which are
1745 not implemented in most models beside ptaskL07, this will abort).
1746 * Add an example that schedules a DAX on an heterogeneous platform
1747 using a Min-Min strategy.
1748 * New function SD_workstation_get_current_task() that returns the kind
1749 of task currently running on a workstation in the sequential access
1751 * Raise some warnings when unexecuted tasks remains at the end of the
1752 simulation. This is usually caused by cycles in the DAG.
1755 * New function: SIMIX_process_set_function() called by MSG_set_function
1756 * Change the underlying waiting queue in semaphores so that a process
1757 can wait on several of them simultaneously (as in waitany).
1758 * Fix the way to handle tokens in semaphores so that all access patterns
1759 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1760 * kill the dirty pimple SIMIX_message_sizes_output()
1761 Please use (proper) visualization instead
1764 * New data container: setset (set of sets of elements)
1765 * New module: mmalloc (mapped malloc, allowing to have several
1766 independent segments of malloc)
1767 * New function: xbt_dict_cursor_set_data()
1768 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1769 * New function: xbt_dynar_is_empty()
1770 * New function: xbt_fifo_get_last_item()
1771 * Fix xbt_dynar_shrink(): use the right element size.
1772 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1773 uninitialized areas during expand.
1774 * Fix semaphores: previous implementation was severely broken.
1775 * Use library init/fini functions for our initialization.
1776 - you can use logs and other feature as soon as you want in your
1777 code (even before the xbt_init / MSG_init)
1778 - xbt_exit is now a no-op and produce a warning when used.
1781 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1782 benefit from the latest improvement to the simulation kernel.
1783 * Kill measurement sockets for now. If you rely on them, sorry. This
1784 release is not for you. This feature will be reintroduced in the
1785 future, but we cannot delay the release any further.
1786 * New function: gras_msgtype_get_name().
1787 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1788 * Fix (at last) the pmm example: it should not randomly fail anymore.
1790 Build chain: bug fixes and overall polishing
1791 * Cmake is now stable enough. Hence, we killed the autotools.
1792 * Port to windows ( TM :)
1793 * Fix the 'make install' target.
1794 No need to use 'make install-simgrid' anymore
1795 * Introduce a 'make dist' target compiling a *source* archive
1796 'make package' compiles a binary archive
1797 * Compile java files only on need
1798 * Add --cd and --setenv command line options to tesh
1799 * Out of source builds are not fully supported yet, but we are close
1800 * Enable supernovae and optimization flags by default for our users
1803 * Add layer to set up environment directly from lua, without XML.
1804 * The effect of gras_stub_generator can be achieved through
1805 lua too (check examples/gras/console/ping_generator.lua)
1807 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1809 SimGrid (3.4.1) stable; urgency=low
1811 The "Polishing easter eggs is probably a good idea" release.
1812 This is a bug fixes release only.
1816 * Fix a bug preventing the tasks from begin garbage collected.
1819 * Fix a bug occuring when a host involved in a communication fails.
1820 This was not detected properly by the other peer involved in the
1821 communication. Now, it's reported as a network error.
1824 * Warn the user about loop dependencies in data flow of DAX files
1825 * Obey the control-flow dependencies of DAX files
1828 * Add option "enable_smpi" allowing to not compile SMPI.
1829 Probably useful for the (Mac) users experiencing a build error here
1830 * Improve the detection of lua5.1 and ruby1.8
1832 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1834 SimGrid (3.4) stable; urgency=low
1836 The "Easter in Cargese" release. Also known as (major changes):
1838 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1839 ~> bindings were greatly improved
1840 ~> new tracing infrastructure for better visualization introduced
1842 * the "Welcome to configury modernity" release.
1843 ~> we switched from autotools to cmake, and improved our cdash
1846 A more detailled list of changes follow (full detail in svn log).
1848 Java Bindings: Various Cleanups
1849 * (install java-gcj-compat-dev on debian-like to use them)
1850 * Remove put/get: no need to export deprecated interface in Java
1851 Use send/receive instead.
1852 * Cleanup the examples and add a README per directory
1853 * Remove example autoDestination (that's the only way to go now)
1854 * Remove example explicitDestination (was a plain copy of basic)
1855 * Make JniException a runtime exception, so that there is no need to
1856 declare the fact that you may encounter such a beast. I guess that
1857 nobody will ever want to survive such error.
1858 * Create specific errors for each MSG case of failure:
1859 host failure, transfer failure, timeout, task cancelled
1860 * Cleanup the exceptions that may get thrown by each function
1861 * Other internal cleanups in Java bindings. Performance still bad :/
1862 Ruby and Lua Bindings: create them
1863 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1864 * That's new and great, you should try them out.
1865 Same functionalities than Java bindings, only even less polished
1867 * Kill the useless "rate" argument of SD_task_get_execution_time()
1868 Everyone used to provide -1 as a value, it was not used, and the
1869 semantic of a possible use wasn't even clear.
1870 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1871 either as comm costs or compute costs to mean that there is no
1872 such thing for that specific task.
1873 * Add a SD_task_set_name() function
1874 * Fix SD_task_unschedule() on typed tasks
1875 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1876 * In DAX loader, accept useless 'level' attributes to <job> since
1877 LIGO DAGs have them (seem to be to ease graphical representation).
1879 * Add an example masterslave_mailbox.c using send/receive and not
1880 the deprecated put/get interface.
1881 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1882 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1884 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1885 (declare MSG_USE_DEPRECATED to still have the old name)
1886 * Add a MSG_task_set_data() function
1887 * About trace replay (see examples/msg/actions):
1889 - Allow to work with splitted trace files for each process
1890 Give the specific trace file as argument of each process,
1891 and call MSG_action_trace_run(NULL)
1892 You can still have one merged file for all processes.
1893 - Fix implementation of collective operations
1894 * Allow task_execute() on 0-sized tasks (closes #10063)
1896 * This is the first release of SimGrid where SMPI is not considered
1897 beta anymore (even if some corners should still be improved)
1898 * Port over the new SIMIX_network submodule (internal refactoring)
1899 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1900 * Implement more missing elements of the standard:
1902 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1904 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1905 MPI_Scan() MPI_Get_processor_name()
1906 - Added implementation of missing case for Alltoall (warning: it's
1907 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1908 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1909 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1910 SMPI_MPI_Allgatherv()
1911 * Bug fixes include:
1912 - MPI_Waitsome() was broken
1913 - Allow relative includes in smpicc
1914 - Command line cfg argument 'reference_speed' was ignored...
1915 - Some functions did not properly lead to auto-benching of user code
1916 - smpicc passes -O2 by default (just like openmpi one)
1918 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1919 * Bug fixes about timeouts during communications
1920 * add SIMIX_message_sizes_output() as a pimple to write to file the
1921 amount of messages per size. Use gnuplot to get histogram.
1922 Pimple because that's the only user-visible function of simix,
1923 defined directly in xbt.h (irk, sorry)
1925 - Add a SIMIX_sem_get_capacity() function
1926 - Fix interactions with processe resume/suspende
1927 - release_forever() was stupidly broken
1928 - Fix SIMIX_display_process_status() for processes in a semaphore
1929 - Make SIMIX_sem_block_onto() user-visible
1930 * Refactoring context stuff:
1931 - Use pseudo-OOP for better modularity
1932 - reimplement SIMIX_process_kill() without process_schedule() so
1933 that the latter can take as invariant that it is called from
1935 - Merge context_start into context_new for sake of simplicity
1937 * Add a Vivaldi network model, coded live during SUD'10 ;)
1938 * Rename configuration variables to start a hierarchy:
1939 o cpu_model -> cpu/model
1940 o network_model -> network/model
1941 o workstation_model -> workstation/model
1942 * New configuration variables:
1943 o network/bandwidth_factor: correction to bandwith
1944 o network/latency_factor: correction to latency
1945 o netwotk/weight_S: correction to the weight of competing streams
1946 * Add a long description to the models, that users can see with such
1947 argument on the command line: --cfg=cpu/model:help
1948 * --help-models display the long description of all known models
1950 * config: add the ability to set a default value after registration
1951 Does not override any previously set value (e.g. from cmd line)
1952 * dict: allow to have integer key and data.
1953 When so, you need to use the following functions
1954 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1955 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1956 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1957 In contrary to regular dicts, the key is not malloced before copy.
1958 Mixing scalar and regular elements in the same dict is not tested
1960 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1961 Tracing for Visualization:
1962 * SimGrid is now instrumented in order to generate a trace file for
1963 visualization analysis: to use it, need to compile SimGrid with the
1964 "tracing" option enabled, and instrument the program using SimGrid with
1965 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1966 (among other functions).
1967 * The instrumentation only traces the platform utilization for now
1968 * Documentation to use the tracing functions and how to analyze the
1969 traces with the Triva tool is written.
1970 * More information about: SimGrid FAQ (in the section Tracing Simulations
1973 * We moved to cmake as default build system. Autotools support will
1974 be dropped soon. Check the FAQ for more info about how to use it.
1975 * Greatly improved our cdash/ctest interactions
1976 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1977 * Added memory checking tests with valgrind; lot of memleak fixing.
1978 This may be the first release of simgrid with so few memory issues
1979 * Added code coverage tests.
1980 Our coverage is still improvable, but at least we see it on cdash.
1982 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1984 SimGrid (3.3.4) stable; urgency=low
1986 The "Desktop Grid needs love too" release (also called Xmas release).
1988 Models improvements:
1989 * Major speedup in the maxmin system solving by using lazy evaluation
1990 Instead of solving completely the maxmin system at each iteration,
1991 only invalidate (and recompute) the modified parts.
1992 This new feature is enabled in default models but you can try to
1993 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1994 * Cas01 IMproved as default CPU model
1995 This CPU model is the same Cas01 model, but it uses the
1996 maxmin-selective-update flag and a heap structure to manage
1997 actions on SURF kernel.
1998 It reduces the complexity to find the next action to finish and,
1999 consequently, it's faster than the old Cas01.
2000 This is the new default CPU model (Cas01).
2001 * Rename the old Cas01 model to Cas01_fullupdate
2002 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2003 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2004 * CpuTI (CPU Trace Integration)
2005 A new CPU model whose objective is simulate faster when using
2006 availability trace files.
2007 Instead of using a full featured, over engineered maxmin system for
2008 CPU modeling, this model does the pre-integration of traces files
2009 to calculate the amount of CPU power available, and so, executes
2010 faster than the old CPU models.
2011 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2012 * Use LV08 as default network model since it gives better accuracy
2013 for small messages and shouldn't change things for big ones.
2014 Use --cfg=network_model:CM02 to get the previous behavior.
2017 ******************************************
2018 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2019 ******************************************
2020 * The new CPU model may changes simulations!
2021 The point is that events occurring at the exact same timestamp
2022 are not scheduled in the same order with the old and new
2023 version. This may be enough to completely change the execution
2024 of simulations in some cases.
2025 * The new network model will change simulations!
2026 This new model is more realistic than the previous one, so you
2027 should consider redoing your old experiments with this model.
2028 Sorry for the inconvenience.
2031 * Introduce the supernovae compilation mode
2032 When compiled that way, the whole SimGrid (or almost) is put in a
2033 single compilation unit and compiled in one shoot.
2034 This is to help gcc which has difficulties to inline stuff from one
2036 The speedup seem to be above 15%, althrough more tests are needed on
2037 amd64 to confirm that gain.
2040 * Port of MSG's mailbox on top of SIMIX network
2041 The put/get mechanism was greatly simplified on the way.
2044 * New SIMIX network module. Provides:
2045 - Mailbox: rendez-vous mechanism to find with who you want to speak
2046 - Synchronous send/recv: easier and hopefully faster since the
2047 logic is handled in the maestro process directly now
2048 - Asynchronous send/recv: you dreamt of it? It's here now
2049 Too bad that nobody cared enough to propagate the change to MSG.
2050 * Add semaphores as SIMIX synchronization mechanism.
2053 * new function SD_daxload(char*) to load a DAX file
2054 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2055 * Introduce typed tasks. Specify its kind and cost at creation.
2056 At scheduling, just give where it should be placed, and the cost
2057 for each involved resource is automatically computed.
2058 Existing constructors so far (more to come of course):
2059 - SD_task_create_comm_e2e() for end-to-end communication
2060 - SD_task_create_comp_seq() for sequential computation
2061 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2062 * new function SD_task_dump() for debuging display
2063 * new function SD_task_dotty(task,FILE*) writing to file the info
2064 about the task in dotty format
2065 * SD_task_dependency_exists() can now cope with having one of its
2066 arguments NULL. If so, it tests whether the other argument has any
2068 * Add getters on list of preceding/following tasks:
2069 SD_task_get_parents(task) and SD_task_get_children(task)
2070 * Add getters on amount of workstations and list:
2071 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2072 * Add getter on task kind: SD_task_get_kind(task)
2073 * Update the start_time and finish_time of tasks on completion/failure
2074 * Bugfix: Remove task from state swags when destroyed
2077 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2080 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2081 dynar with one separate thread per value of the dynar.
2082 * Change the prototype of xbt_thread_create(), sorry.
2083 Added a boolean parameter indicating whether we want to join this
2084 thread (used in SG only for now)
2085 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2088 * GTNetS wrappers should now be usable again (and betterly tested too)
2089 * Fix a major regression from 3.2 where the timeout provided to
2090 MSG_task_put_with_timeout() was used as absolute time before which
2091 the comm should be done.
2092 * Start to fix the <cluster> tag.
2093 - Internal links should be good now (beside of the loopback, which
2094 use the private link instead)
2095 - paths to the external world is still rather broken
2096 - the <route:multi> tag is just broken. Actually that's brain-dead.
2097 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2098 to make it less stupid
2099 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2100 * Fix a source-level compatibility glitch from 3.2: after defining
2101 MSG_USE_DEPRECATED, you can use the old name
2102 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2103 * Allow to compile from the SVN with automake 1.11
2104 * Fix some problems when using the "start_time" tag in deployment XMLs.
2105 * Fix #8569: XBT/synchro.h has redundant declarations
2106 * Fix #8563: MSG return values and exceptions
2107 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2108 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2109 Thanks to Robson Peixoto.
2110 * Fix a vicious bug in dictionaries inducing that some elements were
2111 not freed on xbt_dict_free()
2113 Portability report of this version:
2114 * Main portability targets:
2115 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2116 on (amd64/i386/ia64)
2117 - mac leopard on i386
2118 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2119 but nothing critical.
2120 * Other platforms: windows, AIX and others were not tested for this release
2122 Timing report of this version:
2123 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2124 scenario parameters). From 8h to a few seconds in desktop grid settings.
2125 * Supernovae brings about 25% speedup on i386.
2127 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2129 SimGrid (3.3.3) stable; urgency=low
2131 The "Need for Speed" release.
2133 The timings done to validate the 3.3.2 were faulty.
2134 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2136 The problem was a conversion from a manually handled vector to
2137 xbt_dynar_t on the critical path.
2138 xbt_dynar_foreach calls functions, inducing stack management crap.
2140 We inlined these functions and xbt_dynar_foreach is now breath taking.
2141 We also inlined xbt_swag_belong on the way.
2143 Here are some approximate speedup measurements (on master/slaves
2144 simulations lasting between 10s and 20s each):
2145 3.3.1 -> 3.3.2: about same performance
2146 3.3.2 -> 3.3.3: 40% speedup
2147 3.3.1 -> 3.3.3: 40% speedup
2148 3.3.1 with inline patch -> 3.3.3: 30% speedup
2150 Our reading is that the refactoring which occurred in 3.3.2 made us
2151 suffer much more from the xbt_dynar_foreach low performance, but
2152 once we solved this, this refactoring proved to be very performance
2153 effective. From the 40% speedup, somehow, 10% are due to the
2154 inlining and 30% to the refactoring.
2156 That's a pitty that gcc cannot inline functions placed in other files
2157 alone. We have to choose between:
2158 - break the encapsulation (by putting private data structures and
2159 accessors in headers files to help gcc)
2160 - live with low performance
2161 - switch to a decent compiler such as icc (not quite possible).
2163 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2165 SimGrid (3.3.2) stable; urgency=low
2167 The "Simplicity does not preceed complexity, but follows it" release.
2169 The main contributors of this release were (lexical order):
2170 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2173 * Extract the routing logic into its own object.
2174 (was dupplicated in network.c and workstation_LV07.c;
2175 Allows to implement other ways of storing that info)
2176 => kill now useless network_card concept
2177 - Use dynar to represent routes (instead of void** + int*)
2178 - kill link_set (use surf_network_model->resource_set instead)
2179 - Add a command-line option to choose the routing schema to use
2180 - Add three new models:
2181 * Floyd (shortest path computed at initialization)
2182 * Dijikstra (shortest path recomputed all the time)
2183 * Cached Dijikstra (shortest path computed on need)
2184 All these models where contributed by Silas De Munck, and are
2185 described in his ICCS09 paper.
2187 * Simplify model declaration
2188 (less redirections, less function to write when defining a model)
2189 - Factorize stuff between models:
2192 surf_model_resource_set(model)
2193 surf_model_resource_by_name(model, name)
2194 - Unify the types of models in s_surf_model_t (using an union)
2195 - Embeed fields of common_public directly into s_surf_model_t
2196 - Rename model methods:
2197 action_free ~> action_unref
2198 action_change_state ~> action_state_set
2199 action_get_state ~> action_state_get
2200 - Change model methods into functions :
2201 (model)->common_public->action_use ~> surf_action_ref
2203 * Implement a generic resource; use it as ancestor to specific ones
2204 (allows to kill duplicated code in models)
2205 Drawback: timer command don't need no name nor properties;
2206 workstation_CLM03 don't need no properties
2207 (but I guess we can live with those few bytes wasted)
2209 * Improve the action object model
2210 - implement a constructor avoiding dupplicated code about field
2211 initialization in generic_action part.
2213 * Kill the SDP model: it has an external dependency, is deprecated
2214 in flavor of modern lmm models, and didn't compile since a while
2217 * Relocation of the context module from XBT to SIMIX.
2218 (the context were decoupled from the simix processes, duplicating a lot of code)
2219 => a lot of code was factorized
2220 - less overhead is introduced during scheduling
2221 - simpler API for the context factory
2222 - the logic for process creation,destruction and manipulation was simplified
2223 * Simplification of the s_smx_process_t data structure.
2224 => accesing the simix level data associated to a process is faster now,
2225 and the code is a lot more readable.
2228 * Implement some more MPI primitives:
2229 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2230 -implementation: Bcast: flat or 2-ary tree (default),
2231 Barrier: 4-ary tree,
2233 Allreduce: Reduce then Bcast
2234 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2235 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2236 Alltoallv: flat tree, like ompi
2238 * Add support for optimized collectives (Bcast is now binomial by default)
2239 * Port smpirun and smpicc to OS X
2242 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2243 Shout out if you used it.
2246 * Display the list of still queued messages in SG mode when existing
2250 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2251 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2252 * Add xbt_str_from_file(FILE*)
2253 * Add xbt_dict_get_key achieving a linear reverse search
2254 * Remove the context module
2256 Portability report of this version:
2257 * Main portability targets:
2258 - Linux(debian)/x86/context
2259 - Linux(debian)/x86/pthreads
2260 - Linux(debian)/amd64/context
2261 - Linux(debian)/amd64/pthreads
2262 On these, we still have the eratic breakages of gras/pmm and
2263 amok/saturate_sg reported in previous version. We still think
2264 that the tests are the cause of the fault, not the tested code.
2266 - Mac OSX Leopard/x86/context
2267 Still false negative in tesh autotesting.
2268 Smpi still fails, but this time because readlink does not accept -f
2269 Everything seems to work properly beside of that.
2272 - AIX version 5.3 (only tested contexts this time)
2273 Smpi still fails there because mktemp is not installed.
2274 Everything seems to work properly beside of that.
2276 I managed to compile it for the first time, but several breakages.
2277 Won't delay the release for this exotic platform.
2279 * Windows: it's still lagging behind. If you want to help, please
2282 Timing report of this version:
2283 This version seem to be more than 5% faster than 3.3.1 (on linux
2284 64bits with contextes). The gain is less than expected, we are
2285 investigating this for next release.
2287 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2289 SimGrid (3.3.1) stable; urgency=low
2292 * Implement a --cfg-help to show existing configuration variables
2293 * Build chain do not require doxygen in maintainer mode
2296 * fix a bug on struct sizeof computation, which prevented the
2297 exchange of arrays of structs in some conditions
2298 - added a regression test about this in datadesc_usage
2299 * Allow the exchange of 0-long dynamic vectors.
2300 - for that, use -1 as indicator of dynamic size instead of 0
2301 - This implied to change any size from unsigned long to long,
2302 reducing a bit communication abilities, but I guess that with
2303 64bits being quite common, this is more than enough.
2304 - This also induce a protocol change, thus bumping network protocol
2305 version from 0 to 1 (if we have external users, we have to get
2306 clean on that point too ;)
2307 - added two regression tests about this in datadesc_usage
2308 * Be more verbose when propagating local exceptions
2309 This helps debugging.
2310 * Display the status of simulated processes when receiving SIGINT in
2314 * Allow to control the simulation from a trace file.
2315 New functions MSG_action_register() and MSG_action_trace_run()
2316 The first one allows to associate a function execution to each
2317 kind of action while the second one parses a trace file and
2318 triggers the corresponding actions within the system.
2319 For now, only a toy example is provided in examples/msg/actions
2320 * Add an exemple of process migration in examples/msg/migration
2321 * Fix a bug in task exchange which broke MSG_task_get_sender()
2322 Add a teshsuite regression test for that.
2323 [Bug: if MSG_task_get_sender() is called after sender exit,
2325 * Fix a bug which prevented suspend/resume to work properly
2326 * Display the status of simulated processes when receiving SIGINT
2327 This fixes a regression of v3.3. due to the introduction of SIMIX
2328 * Bug fixing in failure management:
2329 - trace could not start by a failure at time 0
2330 - failure during communications were not working
2333 * Add SIMIX_process_set_name() to change the name of the current
2334 process in the log messages.
2335 * Store smx_hosts in a dict since we only retrieve them by name
2336 * Move the configuration infrastructure to surf
2339 * Move the configuration infrastructure to surf
2342 * Massive internal cleanups:
2343 - Store internal structures on processes instead of hosts (allows
2344 to have more than one process per host, in addition of being more
2346 - Cleanup the initialization/finalization process
2347 - Kill a whole bunch of unneeded synchronization:
2348 processes run in exclusive manner within the simulator
2349 - Move queues from global tables to process data fields
2351 - now accept -platform and -hostfile arguments
2352 - Pass the right rank value to processes according to the hostfile
2353 * Compile the examples by default, and use them as regression tests
2354 * Implement MPI_Wtime()
2355 * Change the reference speed to a command line option
2358 * TCP_gamma can now be specified as command line option using
2359 --cfg=TCP_gamma:10000000.0
2360 * Change the --surf-path cmd line option into --cfg=path:
2363 * Also include strbuff from xbt.h public header
2364 * xbt_ex_display(): do not free the exception after displaying
2365 This allows to do more with the given exception afterward.
2366 Users should call xbt_ex_free() themselves.
2370 Portability report of this version:
2371 * Main portability targets:
2372 - Linux(debian)/x86/context
2373 - Linux(debian)/x86/pthreads
2374 - Linux(debian)/amd64/context
2375 - Linux(debian)/amd64/pthreads
2376 These targets fail about 1/10 of times on gras/pmm, but we believe
2377 that this is because of the test, not because of simgrid.
2378 amok/saturate_sg fails even more rarely, and the test may not be
2381 - Mac OSX Leopard/x86/context
2382 The test suite still spits tons of errors because some obscure
2383 force prevents us from removing the temporary directories
2384 arguing that they still contain some metadata I've never heard of.
2385 Smpi fails because seq is not installed.
2386 Everything seems to work properly beside of that.
2389 - AIX version 5.3 (both contexts and pthread)
2390 Smpi still fails there because mktemp is not installed.
2391 XML inclusions seems rosty on AIX.
2393 * Windows: it's still lagging behind. If you want to help, please
2396 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2398 SimGrid (3.3) stable; urgency=high
2402 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2403 [Malek Cherier & Mt]
2405 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2408 Until now, GRAS were using MSG as an interface to SURF. It was
2409 quite difficult because both interface have several differences
2410 (MSG channels vs GRAS sockets were the most notable point).
2412 This also opens the gate to SMPI (which should occur soon) and speed
2413 up simulations by to 40% (even if it were not the main goal).
2415 **************************************
2416 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2417 **************************************
2418 The point is that events occuring at the exact same timestamp are
2419 not scheduled in the same order with the old and new version. This
2420 may be enough to completely change the execution of simulations in
2421 some cases. Sorry for the inconvenience.
2423 * Cleanup and upgrade the XML format to push further scalability
2424 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2426 * Improve the testing infrastructure with tesh. Now a very large part of
2427 the code is tested not only by being run but also by checking that the
2428 output match an expected output [Mt].
2430 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2431 is really less memory-demanding, which should allow you to use
2432 larger files in SimGrid [AL].
2434 * Inform valgrind about our contextes, so that it becomes usable
2435 with the default (and more effecient) version of SimGrid
2436 [contributed by Sékou Diakite, many thanks]
2439 * Introduce a listener thread in charge of receiving incoming
2440 messages from the network. It allows to overlap communication and
2441 computation but most notably, it removes some stupid deadlocks due
2442 to the fact that so far, a process could not send and receive at
2443 the same time. This made most non trivial communication schema
2445 * Convert the PIDs from long int to int to match the MSG ones (and
2446 linux ones too) [Mt]
2447 * New function: gras_agent_spawn() to launch a new process on
2448 current host. Only working in simulation for now. [Mt]
2449 * New function: gras_os_hostport() returning a constant form (ie,
2450 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2453 * Make the backtrace of exceptions more human readable [Mt]
2454 * New module: xbt/str [Mt]
2455 a ton of string utility functions (split, join, printf to a newly
2456 allocated buffer, trim, etc)
2457 * New module: xbt/hash [Mt]
2458 SHA1 hashing algorithm (more to come if needed)
2459 * New module: xbt/synchro [Mt]
2460 synchronization tools (mutex and conditions) working the same way
2461 in simulation and in real life (mainly useful for GRAS, but not
2463 * New module: xbt/queue [Mt]
2464 classical producer/consumer synchronization scheme
2465 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2466 (using the classical functions will get serialized) [Mt]
2467 * Make dictionary internal table dynamic. No need to specify its size
2468 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2470 * Make sure the log channels are organized as a tree under windows
2471 (because of ANSI C compatibility issue, any channel were child of
2472 root directly) [Mt].
2475 * Cleaned many thing in surf and fixed a few bugs [AL].
2476 * Add a nice command line configuration mechanism to compose models [AL].
2477 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2478 the previous one (KCCFLN05). It relies on something that looks like
2479 a max-min sharing mechanism but cannot be written as such. A new solver
2480 was thus designed [AL].
2481 * Add a new solver to lmm. Based on Lagrange optimization and
2482 gradient-based descent, it enables to efficiently maximise systems s.a
2484 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2486 This solver enables to propose two new network models for TCP Reno and
2487 TCP Vegas based on Low's work. These models still need to be fully
2488 tested though [Pedro Velho].
2491 * Bug fix in SD_simulate. Now the time bound given as argument is
2493 * Use the new parallel task model (ptask_L07) as default.
2494 * Use the SURF command line configuration mechanism.
2495 * 0-size tasks (for synchronization) should now work.
2497 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2499 SimGrid (3.2) stable; urgency=high
2503 We still experience issues on this platform, but we believe that at
2504 least MSG is usable.
2506 GRAS API BREAKAGE (for simplification purpose, sorry):
2507 * the gras_msgtype_by_name is not used anymore. Instead of
2508 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2509 you can write (and must)
2510 gras_msg_send(toserver, "request", &request);
2511 - If you still want to pass a gras_msgtype_t to the function (to cache
2512 the type and avoid the lookup time), use the gras_msg_send_() variant.
2513 - Impacted functions:
2514 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2515 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2516 * The callbacks are now expected to return 0 when everything went well
2517 (just like the main() function)
2519 GRAS new features and improvements:
2520 * New module mechanism where user code can use per process globals [Mt]
2521 This is similar to gras_userdata_*() functions, but for libraries. It
2522 factorize some code developped over and over in the examples and AMOK.
2523 It has still to be documented and used (only amok/peermanagement is
2525 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2526 loss when they were small enough to fit into the buffer and sent quickly
2527 enough so that they can all get received in one shoot.
2528 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2529 if not found. Use the *_or_null() variant for the old semantic.
2530 * In gras_msg_handle, do not discard messages without callback.
2531 They are probably messages to be explicitly awaited later (ie, proofs of
2532 mis-synchronization in userland since they are sent before being awaited)
2533 No big deal usually.
2534 * gras_socket_meas_send/recv: semantic changed!
2535 The numerical arguments used to be (1) the total amount of data to send
2536 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2537 messages. This was need for the fool willing to send more than MAXINT
2538 bytes on quite fat pipes.
2541 * Do really rename the hostmanagement module to peermanagement. [Mt]
2542 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2543 but this is rather new and this was documented in the module
2544 documentation (poor excuses, I admit)
2545 * Bandwidth measurement semantic changed! This follows the changes to
2546 gras_socket_meas_send/recv explained above.
2549 * A sequential mode has been added to the workstations. When a workstation
2550 is in sequential mode, it can execute only one task, and the other tasks
2551 are waiting in a FIFO. [Christophe Thiery]
2554 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2555 model for SIMDAG. [Christophe Thiery]
2556 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2557 equal to 0, which caused some bad behaviors in
2558 saturated_constraint_set_update. I now use a threshold mechanism like in
2562 * When running manually src/testall, you select specific units [Mt]
2563 testall is the result of our cunit mechanism, and should replace all
2564 the scripty thingy around since bash don't run easily on billware.
2566 * A mallocator system has been added. [Christophe Thiery]
2567 Mallocators allow you to recycle your unused objects instead of freeing them
2568 and allocating new ones.
2570 Documentation update:
2571 * FAQ reworking + New FAQs:
2572 - "Valgrind spits tons of errors!" [Mt]
2573 - "How to repport bugs" [Mt]
2574 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2575 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2576 - Communication time measurement within MSG [AL]
2577 - I experience weird communication times when I change the latency [AL]
2578 * GRAS tutorial [Mt]
2580 - an introduction to the framework and to the used communication model
2581 - an initiatic tour introducing the most proheminent features:
2583 . Lesson 0: Installing GRAS
2584 . Lesson 1: Setting up your own project
2585 o Part 2: Message passing
2586 . Lesson 2: Exchanging simple messages
2587 . Lesson 3: Passing arguments to the processes (in SG)
2588 . Lesson 4: Attaching callbacks to messages
2589 . Lesson 5: Using globals in processes
2590 . Lesson 6: Logging information properly
2591 . Lesson 7: Using internal timers
2592 . Lesson 8: Handling errors through exceptions
2593 . Lesson 9: Exchanging simple data
2594 . Lesson 10: Remote Procedure Calling (RPC)
2595 . Lesson 11: Explicitely waiting for messages
2596 . Recapping of message passing features in GRAS
2597 - A HOWTO section containing:
2598 o HOWTO design a GRAS application
2599 More are due, of course. They will come latter. In the meanwhile, you can
2600 check the examples which are still here.
2602 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2604 SimGrid (3.1) stable; urgency=high
2608 There was a stack corruption somewhere, visible only when optimizing
2609 with these versions. [Vince]
2612 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2613 module that enabled to play with Directed Acyclic Graphs. It is built
2614 directly on top of SURF and provides an API rather close to the old
2615 SG. Some old codes using SG are currently under rewrite to check that
2616 all needful functions are provided. [Christophe Thiery]
2619 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2620 extension of the classical CLM03 model that gracefully handles
2621 failures. This is now the default model for MSG and GRAS. It doesn't
2622 handle parallel tasks yet though. [AL]
2623 * Bug fix: Weights were not correctly set in the network part.
2624 WARNING: This may have resulted in incorrect results with simulations
2625 where there are more than one flow on a given link. [AL]
2628 * After a (long ?) discussion on simgrid-devel, we have decided that the
2629 convention we had on units was stupid. That is why it has been decided
2630 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2631 WARNING : This means that all previous platform files will not work as
2632 such with this version! A warning is issued to ask users to update
2634 A conversion script can be found in the contrib module of the CVS, under
2635 the name contrib/platform_generation/surfxml_update.pl [MQ]
2638 * Bug fix: Processes were started in reverse order, wrt deployment file.
2639 WARNING: if your code relies on this bug, please fix it. [AL]
2640 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2641 being executed on two different locations. [AL]
2642 * Bug fix: Failures are now better supported thanks to Derrick's tests
2643 (there was many failure situations I hadn't thought of and that weren't
2644 correctly handled). [AL]
2645 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2649 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2650 Exception occurring on server-side are propagated back to client (!).
2652 API CHANGE: the callback changed their prototype. Change:
2653 int my_handler(gras_socket_t expeditor, void *payload_data) {
2655 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2656 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2658 * New! function: gras_msg_handleall to deal with all messages arriving
2659 within a given period.
2660 * New! function: gras_socket_server_range to get a server socket in a
2661 range of port numbers (ease to avoid port number conflicts) [MQ]
2662 * New! gras processes display their backtrace when they get a SIGUSR1
2663 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2664 Sweet to debug RL processes [MQ]
2668 - Do not force experiment sizes to be expressed in kb, or it becomes
2669 impossible to measure the latency this way (needs one byte-long tests)
2670 WARNING: this changes the amok_bw_* function semantic. [MQ]
2671 - Implements the link saturation stuff. [MQ]
2672 * Peer management module:
2673 New! module factorizing code that we wrote over and over [MQ].
2676 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2677 - Test units are placed directly into the library code, they get extracted
2678 automatically and placed into the src/testall binary.
2679 - Convert most of the XBT tests to this system.
2680 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2681 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2682 live in the same C code. This required to change a little bit the API
2683 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2684 * New module: sparse graph structure with basic algorithms (this is work
2685 in progress and the API is not considered to be frozen yet). [AL]
2686 * Display more information on backtraces: source line & function names are
2687 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2688 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2689 * Reworked a little bit some #include statements to load only required
2690 headers. Some user code that relied on SimGrid to include stdlib or
2691 stdio may need to include it by themselves. [AL]
2692 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2694 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2695 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2696 but I guess that almost nobody used those parts. [MQ]
2698 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2700 SimGrid (3.0.1) stable; urgency=low
2703 * Unfortunately, I had missed 5 misnamed functions:
2704 xbt_fifo_item_t xbt_fifo_newitem(void);
2705 void xbt_fifo_freeitem(xbt_fifo_item_t);
2706 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2707 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2708 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2709 They're now deprecated. Please use their new versions:
2710 xbt_fifo_item_t xbt_fifo_new_item(void);
2711 void xbt_fifo_free_item(xbt_fifo_item_t);
2712 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2713 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2714 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2716 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2717 * Documentation: xbt_log module unmercifully reworked [MQ]
2718 * Bugfix: there was a problem with the ending of contexts with
2719 the pthread backend. It caused some weird deadlock or behavior
2720 depending on the pthread implementation. [AL]
2721 * Bugfix: get the exceptions raised in the simulator repport where
2722 and why they come from when they are not catched in time [AL, MQ]
2725 * Bugfix: Do repport the error when two non-connected hosts try to
2726 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2729 * Add additionnal checkings on communications. Assert that two
2730 communicating hosts are connected by a set of links... [AL]
2733 * Add additionnal checkings on channel values in communication [AL]
2734 * New: MSG_task_get_source to see on which host a task was generated [HC]
2735 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2736 the number of tasks waiting to be received on channel and sent
2738 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2739 waits for the first task coming from a given host.. [AL]
2741 GRAS new functionnalities: [MQ]
2742 * Enhance the parsing macro to allow the size of multidimentional objects
2743 to be given thru annotations.
2744 * New example (and documentation): Matrix Multiplication a la RPC
2745 (as when I was young!) and fix a bunch of bugs found on the way.
2747 GRAS performance improvements: [MQ]
2749 * Reduce the amount of cbps creation/destruction by making it static to
2750 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2753 * Change libdata to a set so that we can search for stuff by ID (and thus
2754 reduce the insane amount of dict lookups)
2757 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2758 RL communication on top of SG since it's so uneffective.
2759 It may also be used for inter-thread communication in RL, one day.
2760 * Use gras_datadesc_copy() to exchange messages on top of SG
2762 - improve message exchange performance on top of SG
2763 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2764 * Don't exchange on the network the size of the used part of buffer,
2765 instead, specify the possible buffer size to read().
2767 - reduces the amount of read/write calls (one pair per exchange)
2768 - reduces the amount of exchanged data (the size)
2769 - allows to retrieve all arrived data on receiver side, if we don't need
2770 it right now (subsequent read will peek the buffer)
2771 - allows the receiver to proceed with the begining of the stream before
2772 everything is arrived
2773 - make it possible to build an iov transport (using readv/writev)
2775 - take care of the data with non-stable storage (like stacked data),
2777 * If possible, TCP send uses vector I/O (when writev() is here)
2778 - Don't use it for receive since we send data sizes and data on the
2779 same stream, so we wouldn't be able to chain large amount of chunks
2780 before having to flush the stuff to read the size.
2781 * Rework the transport plugin mechanism to simplify it and reduce the
2782 amount of pointer dereferencement when searching for the right function
2785 * I guess that now, we do almost as few system calls as possible while
2786 doing as few data copy as possible.
2788 To improve it further, we could try to send all the sizes first and then
2789 all the data (to use iov on receiving size), but it's only a partial
2790 solution: when you have 2 dimensional data, the sizes of the second
2791 dimension is data of the first dimension, so you need 3 streams.
2793 I'm not sure the potential performance gains justify the coding burden.
2795 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2797 SimGrid (3.00) stable; urgency=high
2800 * New! Give the possibility to hijack the surf parser and thus bypass
2801 MSG_create_environment and MSG_launch_application. Have a look at
2802 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2804 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2806 SimGrid (2.96) unstable; urgency=low
2811 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2812 This deprecates the xbt_error_t mechanisms.
2813 It modifies (simplifies) all XBT and GRAS API.
2814 MSG API keeps unchanged (exceptions raised by XBT are catched from
2815 within MSG and masked with existing error handling facilities)
2818 * New! Add a FATPIPE model. [AL]
2819 * New! Add a parallel task model. [AL]
2820 * New! Add automatically a loopback interface (in the default
2821 network model) if none was precised.
2824 * Bugfix: MSG_process_resume now works with the current running process.
2826 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2827 * Modification of MSG_task_get_compute_duration. Once a task has been
2828 processed, the value returned by this function is now equal to 0. [AL]
2829 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2830 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2831 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2832 of a m_task. For now, MSG_task_cancel only works with computation
2834 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2835 of the processor (in Mflop/s) regardless of the current load on the
2837 * API Change: use proper naming convention for MSG_getClock and
2838 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2840 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2841 This function changes the priority of a computation task. This priority
2842 doesn't affect the transfer rate. A priority of 2 will make a task
2843 receive two times more cpu power than the other ones. This function
2844 has been added to suit the needs of Nguyen The Loc and hasn't been that
2845 much tested yet. So if it fails, please report it and send me your code.
2847 * API Change: removed all functions and types that were marked "deprecated"
2848 since many months. Renamed MSG_global_init_args to MSG_global_init.
2850 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2852 SimGrid (2.95) unstable; urgency=low
2855 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2856 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2857 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2860 * Bugfix: Allow absolute paths for platform description files [MQ]
2861 * Bugfix: do free the variables after use. Leads to drastic performance
2863 * Implement max_duration (ie, timeouts) on resources [AL]
2866 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2868 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2869 the needed support (timeouts) [AL]
2872 * Implement measurement sockets. You can now get the bandwidth between two
2873 hosts thanks to AMOK (see below). [MQ]
2874 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2875 dynar over the network (yeah) [MQ]
2876 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2877 * simple/static token-ring example. [Alexandre Colucci and MQ]
2878 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2879 in place before [MQ]
2882 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2883 type parsing. "short" and "long" modifiers were also ignored; other
2884 modifier (such as reference level) are still ignored. [MQ]
2885 * Embeed the buffer size within the buffer itself on SG. [MQ]
2886 That way, send() are atomic and cannot get intermixed anymore (at least
2887 the ones which are less than 100k; bigger messages still have the issue)
2888 * Array size pushed by the field, not by the field type (or each
2889 and every long int will push stuff to the cbps) [MQ]
2890 * use select() to sleep since it allows to portably sleep less than one
2893 GRAS (minor cleanups)
2894 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2897 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2898 * type category 'ignored' killed as it was never used and were difficult
2900 * whether a type can cycle or not is now a flag, leaving room for more
2901 flags (as "ignored", if we feel the need one day ;)
2902 * Rename raw sockets to measurement sockets since "raw" has another
2903 meaning in networking community.
2906 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2907 and offers features not belonging to GRAS but that most applications
2908 need. One day, it may be a set of plugins loadable at runtime.
2909 * New module: bandwidth
2910 bandwidth measurement between arbitrary nodes running this module. [MQ]
2912 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2914 SimGrid (2.94) unstable; urgency=low
2916 The first beta release of SimGrid 3 !
2920 * Update the main page and the FAQ. Adding references to gforge.
2923 * Add a gras_os_getpid function.
2926 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2927 * Extend the logs so that they also print PID, hostname, date, ... if
2929 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2930 and kill the old version which were in testsuite/
2931 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2935 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2936 for users. Instead, each programming environment (GRAS, MSG,...) use it
2937 when needed to provide such a feature to users.
2938 Don't shortcut the mechanism or you will also shortcut the virtualization
2939 you need on the simulator.
2943 * Cleanups in configury with regard to compile optimization/warning flags.
2944 Also add -fno-loop-optimize to any powerpc since it's the optimization
2945 killing gcc (< 3.4.0).
2946 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2947 needed by MSG examples complications
2948 * Borrow configury beautifications from PHP
2951 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2952 warning (thanks loris for stumbling into this one).
2953 * Bugfix: stop loading private headers (gras_config.h) from the public
2957 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2958 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2959 RL would imply a DNS resolver, which is *hard* to do in a portable way
2960 (and therefore delayed).
2961 * Implement a real timer mechanism and use it in timing macros. This allows
2962 to avoid rounding errors and get a 0.000005 sec precision in timing
2963 macros. While I was at it, various cleanups:
2964 - allow to declare more than one timed section per file (fix a stupid bug)
2965 - move some private declaration to the right place
2966 - merge conditional execution and timing macros into emulation module
2967 - document the module
2968 - make sure the module cleanups its mess on gras_exit
2969 * Documentation improvements:
2970 - (new) how to compile applications using GRAS
2971 - (new) emulation support (timing macros)
2973 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2975 SimGrid (2.93) unstable; urgency=low
2977 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2980 - Use Paje properly where used. Still to be sanitized properly.
2981 - Portability fix: Add an implementation of the contexts using pthread
2985 - Add xbt_procname(): returns the name of the current process.
2986 Use it to show the current process's name in all logging.
2988 - fix detection of older flex version and the reaction, since we do
2989 depend on modern ones (we use lex_destroy)
2990 - Better separation of SG and RL in the libs: remove all simulation code
2991 from libgras. As a result, this lib is now only 200k when stripped.
2992 Some of the xbt modules may also be duplicated (two sets and such) and
2993 should be cleaned/killed before SG3.
2994 - Insist on using xlC on AIX because of weird problems involving gcc there.
2995 - Cleanup the make remote stuff. This is now done by scripts
2996 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2997 mainly for our private use, but we're working on changing them to user
3000 - Bugfix: flush the socket on close only if there is some *output*.
3001 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3003 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3006 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3008 SimGrid (2.92) unstable; urgency=low
3010 Alpha 3 on the path to SimGrid 3
3014 - New! First try of benchmarking macros.
3015 - New! First try so that gras_stub_generator generate deployment and
3016 remote compilation helpers.
3018 - Bugfix: Initialization fix in msg_test.
3022 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3024 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3026 - API Change: gras_os_sleep to take the amount of seconds as a double.
3027 Accepting an int was error prone since it was the only location where
3028 seconds were coded as such. It leaded to damn rounding errors.
3029 - Bugfix: Hard to belive that timers ever worked before this.
3031 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3033 SimGrid (2.91) unstable; urgency=low
3035 Alpha 2 on the path to SimGrid 3
3039 - Bug fix in the lmm_solver.
3041 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3042 through the function MSG_paje_output.
3043 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3044 - It is possible to bound the rate of a communication in MSG with
3045 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3046 to write it in the changelog).
3047 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3052 - Various cleanups to the autotools stuff
3053 - Begin to move Gras examples to examples/gras/
3054 - Let make distcheck work again (yeah!)
3056 - documentation overhauled using doxygen.
3057 gtk-doc-tools is dead in SimGrid now.
3058 - Automatically extract all existing logging categories, and add the list
3059 to the documentation (long standing one, to say the less)
3061 - Cleanup the known architecture table. Reorder the entries to group what
3062 should be, and use a more consistent naming scheme.
3063 (some of the test dataset are still to be regenerated)
3064 - New! Allow library to register globals on each process just as userdata
3066 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3067 lookup time (for now).
3068 Use it in msg and trp.
3069 This cleans a lot the internals and helps enforcing privacy of the
3070 headers between the gras components.
3071 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3072 - Bugfix: gras_os_time was delirious in RL.
3073 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3075 - Reenable GRAS now that it works.
3077 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3079 SimGrid (2.90) unstable; urgency=low
3081 Alpha 1 on the path to SimGrid 3
3083 * It is a long time since the last release of SimGrid. I'm sorry about
3084 that but as I had told you, I was rewriting a lot of things. I apologize
3085 to those who had been reporting bugs to me and that I had not answered.
3086 If your bug is still in the new version, please tell me. Here is a
3087 summary of the main changes.
3089 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3090 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3091 portability, speed, and a lot more but you'll figure it out later.
3092 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
3093 I wanted to release the new MSG as soon as possible and I have
3094 broken GRAS, which is the reason why, for now, only MSG is fully
3095 functional. A laconic description of these projects is available
3096 in the documentation.
3098 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3099 kernel optimized for our needs (called SURF but only the developers
3100 should use it). Hence, MSG is now roughly 30 times faster and I think
3101 that by rewriting a little bit MSG, I could event speed it up a little
3102 bit more. Beside the gain in speed, it is also much easier to encode a
3103 new platform model with SURF than it was with SG. More to come...
3105 * REVOLUTION 3: I have tried to change a little as possible the API of
3106 MSG but a few things really had to disappear. The main differences
3107 with the previous version are :
3108 1) no more m_links_t and the corresponding functions. Platforms are
3109 directly read from a XML description and cannot be hard-coded
3110 anymore. The same format is used for application deployment
3111 description. The new format is described in the documentation.
3112 Have a look in tools/platform_generation. There is a tiny script
3113 that converts from the old platform format to the new one. Concerning
3114 the application deployment format, parsing the old one is tricky.
3115 I think most of you should however be able to convert your files. If
3116 it is really an issue, I can write a C code that does the conversion.
3118 2) the toolbox tbx does not exist anymore. We now have a library
3119 with much more data-structures but without the hash-tables (we have
3120 dictionaries that are much faster).
3122 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3124 *****************************************************************************
3125 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3126 * don't feel like converting the oldies. (Mt) *
3127 *****************************************************************************
3130 Version 2.90: "the long awaited one"
3131 - Finished rewriting and debugging MSG. Rewrote the documentation.
3132 - disable GRAS for now since it needs to be ported to the newest SG
3135 - Finish the port to windows (using mingw32 for cross-compile)
3138 - Main loop and datastructures of SURF. A cpu resource object is
3139 functional. Surf can thus be used to create cpu's with variable
3140 performance on which you can execute some actions.
3142 2004-11-15 Martin Quinson
3143 - Port to ARM. Simply added the alignment and size descriptions. Should
3144 work, but the ARM machines are so slow that I didn't had the opportunity
3145 to 'make check' over there yet.
3147 2004-11-15 Arnaud Legrand
3148 - Trace manager now written. It uses a heap structure and is therefore
3149 expected to be efficient. It may however be speeded up (particularly
3150 when many events occur at the same date) by using red and black
3151 trees. One day maybe...
3152 - Max-min linear system solver written. It uses a sparse matrix
3153 structure taking advantage of its expected use. Most operations are
3154 O(1) and free/calloc are called as few as possible. The computation of
3155 the minimum could however be improved by using a red and black tree
3158 2004-11-03 Arnaud Legrand
3159 - Rename every gras_* function that was in xbt/ to its xbt_
3161 - Add a heap and a doubly-linked list to xbt
3162 - Added a dichotomy to the dictionaries. make check works as well before
3163 so I assume that the patch is correct. I do not know however if things
3164 run effectively faster than before now. :)
3166 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3167 SimGrid, and the version number is bumped to 2.x
3169 2004-10-29 Martin Quinson
3170 - Introduction of the remote errors.
3171 They are the result of a RMI/RPC on the remote machine.
3172 ErrCodes being scalar values, you can't get the host on which those
3173 errors did happen. Extending the error mechanism as in Gnome is possible.
3174 No idea yet whether it is a good idea.
3176 2004-10-28 Martin Quinson
3177 - Interface revolution: the Starred Structure Eradication.
3178 I used to do typedef struct {} toto_t; and then handle *toto_t.
3179 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3180 - ???_t is a valid type (builded with typedef)
3181 - s_toto_t is a structure (access to fields with .)
3182 - s_toto is a structure needing 'struct' keyword to be used
3183 - e_toto_t is an enum
3184 - toto_t is an 'object' (struct*)
3186 typedef struct s_toto {} s_toto_t, *toto_t;
3187 typedef enum {} e_toto_t;
3188 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3189 s_toto_t) is private.
3191 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3192 so that it can set the variable to NULL. It was so for dicts and sets,
3193 it changed for dynars.
3195 - Fix a bunch of memleaks in dict_remove
3196 - Fix a bug in sg/server_socket opening: it failed all the time.
3198 2004-10-07 Martin Quinson
3199 - Speed up dynar lookup operation a bit.
3201 gras_dynar_get is dead.
3203 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3204 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3205 (which gives you the address of the stored data).
3207 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3208 copy of the data using an affectation to do the job and not a memcpy.
3210 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3211 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3213 It does not really speedup the dynar test because they are
3214 setting elements all the time (and look them seldom). But the dict does
3215 far more lookup than setting.
3217 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3219 2004-10-05 Martin Quinson
3220 - Allow to (en/dis)able the cycle detection at run time.
3222 Whether we should check for cycle or not is now a property of each
3223 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3224 datadesc_cycle_unset allow to remove this property when previously set.
3226 Note that the cycle detection is off by default since it impacts the
3227 performance. Watch the data you feed GRAS with ;)
3229 This property is hereditary. Any element embedded in a structure having it
3230 set have it set for the time of this data exchange.
3232 You should set it both on sender and receiver side. If you don't set it on
3233 sender side, it will enter an endless loop. If you forget on receiver
3234 side, the cycles won't be recreated after communication.
3236 - Header reorganization.
3237 Kill gras_private.h, each submodule must load the headers it needs.
3239 2004-10-04 Martin Quinson
3240 - Interface revolution: do not try to survive to malloc failure.
3242 Now, gras_malloc and friends call gras_abort() on failure.
3243 As a conclusion, malloc_error is not a valid error anymore, and all
3244 functions for which it was the only gras_error_t return value are
3245 changed. They now return void, or there result directly.
3246 This simplify the API a lot.
3248 2004-09-29 Martin Quinson
3249 - Re-enable raw sockets.
3250 Created by gras_socket_{client,server}_ext;
3251 Used with gras_raw_{send,recv}
3254 It should allow to kill the last bits of gras first version soon.
3256 This is not completely satisfactory yet (duplicate code with
3257 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3260 - Simplify transport plugin (internal) interface by not passing any
3261 argument to _server and _client, but embedding them in the socket
3264 2004-09-28 Martin Quinson
3265 - Finish the port to AIX.
3266 autoconf was my problem (segfault within the malloc replacement
3267 function. No idea why)
3269 2004-09-16 Martin Quinson
3270 - Fix some size_t madness on 64bit architectures.
3272 2004-09-08 Martin Quinson
3273 - Reduce the number of system headers loaded, overload some more system
3274 calls (such as malloc to cast the result of the system one, and work
3276 - Fix and reintroduce the config support
3278 2004-09-07 Martin Quinson
3279 - Source code reorganization to allow Arnaud to surf all over there.
3280 - Allow to document the logging categories.
3281 - Remove all uppercase from logging categories and useless cleanup in names.
3283 2004-08-18 Martin Quinson
3284 Version 0.6.2 (protocol not changed; API changed)
3285 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3286 gras_error_t), and NULL when not found. Functions expecting a msgtype
3287 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3288 hopefully usefull message.
3289 - Portability to prehistoric sparcs again
3291 2004-08-17 Martin Quinson
3292 Version 0.6.1 (protocol not changed; ABI not changed)
3293 - prealloc some buffers to speed things up
3295 2004-08-11 Martin Quinson
3296 Version 0.6 (protocol not changed; ABI expended)
3297 - The parsing macro can deal with the references, provided that you add
3298 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3300 2004-08-09 Martin Quinson
3301 Version 0.5 (protocol not changed; ABI changed)
3302 - Allow to off turn the cycle detection code in data exchange at
3303 compilation time. It should be at run time, but I'm short of time (and
3304 the config stuff is still broken). That way, we keep dict out of the
3305 critical path, which is good because the performance is poor:
3306 - search not dichotomial yet
3307 - dynar give no way to access their content and memcpy everytime
3308 - In composed data description (struct, ref and so on), stop foolness of
3309 keeping the subtype's ID, but store the type itself. This keeps sets out
3310 of the critical path, which is good since they rely on dynar and
3311 dictionnaries. The only loose of that is that we cannot detect the
3312 redeclaration of a structure/union with another content (but I'm not sure
3313 the code detected well this error before anyway). We still can detect
3314 the redefinition discrepancy for the other types.
3315 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3316 it breaks the code because of type-punning used all over the place).
3317 This breaks on all non-gcc architectures (for now).
3319 All those changes (plus the buffer of last time) allow me to gain 2 order
3320 of magnitude on cruel tests consisting of 800000 array of integers on two
3321 level of a hierarchical structure (200 secondes -> 4 secondes)
3324 - the selector of reference must now return the type it points to, not
3325 the ID of this type.
3327 2004-08-06 Martin Quinson
3328 Version 0.4 (protocol changed; ABI not changed)
3329 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3330 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3331 - Implement a buffer transport, and use it by default (it relies on tcp in
3332 real life and on sg in simulation).
3333 That's a bit hackish since I had a new field to the structure to store
3334 its data without interfering with the subtype ones. Inheritance
3335 is tricky in C. And that's a kind of reverse inheritance with one class
3336 derivating two classes. Or maybe a game with java interfaces. Anyway,
3337 that's damn hard in C (at least).
3338 Moreover, I got tired while trying to ensure plugin separation and
3339 genericity in SG mode. MSG wants me to do weird things, so let's go for
3340 cruel hacks (temporarily of course ;).
3341 See comment in transport_private.h:71
3342 - do not include all the _interface headers in private but in the files
3343 which really need them (to cut the compilation time when they are
3346 2004-07-26 Martin Quinson
3347 Version 0.3 (protocol not changed; ABI changed)
3348 - Major overhault of the datadesc interface to simplify it:
3349 - shorted the function names:
3350 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3351 - add a trivial way to push/pop integers into the cbps without malloc.
3352 This allows to make really generic sub_type description, which simply
3353 pop their size of the stack.
3354 - add a function gras_datadesc_ref_pop_arr() which does what users want
3355 most of the time: Declare a dynamic array (which pops its size of the
3356 stack) and declare a reference to it. Poor name, but anyway.
3357 - kill the post-send callback, add a post-receive one
3359 2004-07-23 Martin Quinson
3360 Version 0.2 (protocol changed; ABI changed)
3361 - add some testing for cpbs in the test cases, and fix some more bugs.
3362 This invalidate again the little64 data file, since I cannot regenerate
3364 - remove an awfull optimization in the logging stuff, allowing me to:
3365 - understand it again
3366 - learn gcc how to check that the argument match the provided format
3367 - fix all errors revealed by gcc after that
3368 - internal keys of dict are not \0 terminated. Deal with it properly in
3369 loggings instead of segfaulting when the user want to see the logs :-/
3371 2004-07-22 Martin Quinson
3372 - Fix some stupid bug preventing cbps (callback postit) from working
3374 2004-07-21 Martin Quinson
3375 - Some documentation cleanups
3376 - remove the useless last argument of msgtype_declare
3377 - rename the Virtu functions to fit into the 'os' namespace
3378 - move headers src/include -> src/include/gras/ and stop fooling with
3379 gras -> . symbolic link
3380 - make distcheck is now successful
3382 2004-07-19 Martin Quinson
3384 - Build shared library also
3385 - Install html doc to the right location
3386 - stop removing maintainer files in make clean
3387 - build tests only on make check
3389 2004-07-13 Martin Quinson
3391 - No major issue in previous version => change versioning schema
3392 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3393 corresponding dataset.
3395 2004-07-11 Martin Quinson
3397 - Get it working with any kind of structure (we can compute the padding
3398 bytes remotely for all the architectures I have access to)
3399 - Implement the structure parsing macro (still not quite robust/complete)
3400 - Improvement to the remote testing toysuite
3402 2004-07-10 Martin Quinson
3403 [autoconf mechanism]
3404 - get ride of a bunch of deprecated macros
3405 - actually run the test for two-compliment, not only compile it :-/
3406 - test whether the structures get packed (and bail out if yes. Damn.
3407 Alignment is a serious matter)
3408 - test whether the structures get compacted (but respecting the alignment
3409 constraints of each types)
3410 - test whether the array fields of structures can straddle alignment boundaries
3412 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3413 (took me 2 hours to find that bug, looking at the wrong place)
3414 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3415 before sending/receiving and are used to compute the offsets of fields
3416 - Given that padding size depend even on compiler options, keep track of
3417 alignment and aligned_size only for the current architecture. Not a big
3418 deal since we send structure fields one after the other (seems
3420 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3421 program, let it work on linux/gcc/little32. portability todo.
3423 2004-07-08 Martin Quinson
3424 - import and improve remote compilation support from FAST
3425 - make sure make check works on half a dozen of machines out there
3427 2004-07-07 Martin Quinson
3428 Let's say it's version 0.0.3 ;)
3429 - Implement conversions (yuhu!)
3430 - Let it work on solaris (beside conversion, of course)
3431 - Stupid me, using rand() to generate the conversion datatests in not wise.
3433 2004-07-06 Martin Quinson
3434 - Let make dist work, since I'm gonna need it to compile on remote hosts
3435 - Let Tests/datadesc_usage write the architecture on which the file was
3436 generated as first byte.
3437 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3438 architecture descriptions.
3439 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3440 on those architectures.
3441 - Optimization: send/recv array of scalar in one shoot
3443 2004-07-05 Martin Quinson
3444 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3446 - Plug a whole bunch of memleaks
3447 - each process now have to call gras_{init,exit}. One day, their log
3448 settings will be separated
3449 - Continue the code factorisation between SG, RL and common in Transport.
3451 2004-07-04 Martin Quinson
3453 - Redistribution between SG and RL.
3454 We wanna have to accept in SG, so move accepted related parts of RL in
3455 the common part. (more precisely, the dynar of all known sockets is no
3456 more a static in transport.c, but part of the process_data)
3458 [gras_stub_generator]
3459 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3461 2004-07-03 Martin Quinson
3462 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3463 them all up in one shot)
3465 - Fix the ugly case of reference to dynamic array.
3466 - New (semi-public) function gras_datadesc_size to allow the messaging
3467 layer to malloc the needed space for the buffer.
3469 - gras_socket_close now expect the socket to close (and not its address to
3470 put NULL in it after it). This is because the socket passed to handlers
3471 is one of their argument (=> not writable).
3473 - propagate the interface cleanup from last week in datadesc, ie remove a
3474 superfluous level of indirection. User pass adress of variable
3475 containing data (both when sending and receiving), and not of a variable
3476 being a pointer to the data. Let's say that I like it better ;)
3477 The price for that is constructs like "int msg=*(int*)payload" in
3478 handlers, but it's a fine price, IMHO.
3480 - Let it work in RL (yuhu)
3482 2004-06-21 Martin Quinson
3484 - porting SG plugin and SG select to new standards (works almost).
3485 - plug memleaks and fix bugs around.
3488 - cleanup the prototype of data recv and force users to specify when they
3489 want to handle references to objects. Test case working even for cycles.
3490 - plug memleaks. Valgrind is perfectly ok with this.
3492 2004-06-12 Martin Quinson
3494 - cleanup the separation between plugin and main code in plugin creation
3496 2004-06-11 Martin Quinson
3498 - Reput hook for raw sockets, needed for BW experiments
3499 - kill a few lines of dead code
3500 [Data description] Interface cleanup
3501 - gras_datadesc_by_name returns the searched type or NULL.
3502 That way, no variable is needed to use a type desc once, which makes
3504 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3505 two parameters were strings (field name, type name), leading to
3507 [Dicos] Interface cleanup
3508 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3509 This is consistant with the dynar API.
3511 2004-04-21 Martin Quinson
3513 - Porting to new standards.
3515 - interface cleanup.
3516 There is no bag anymore, no need to take extra provision to mask the
3517 pointers behind "ID".
3518 Better splitup of functions between files create/exchange/convert.
3519 This is still a bit artificial since convert and receive are so
3520 interleaved, but anyway.
3522 - add a queued message list to procdata (the ones not matching criteria
3524 - factorize some more code between SG and RL wrt procdata
3526 - use gras_exit in example to track memleaks
3527 - get rid of gs_example now that GS is properly integrated into gras
3528 - update run_test to integrate the latest tests (datadesc)
3530 - rename WARNINGn macros to WARNn since it prooved error-prone
3532 2004-04-19 Martin Quinson
3534 - register init/exit functions within gras module mechanism
3535 - send/receive function.
3536 Convertion is not implemented, but short-cutted if not needed.
3537 struct/array elements are sent one by one (instead of block-wise), but
3538 nobody really cares (yet). Get a prototype before optimizing.
3539 - tests (using a file socket) for DD send/receive on:
3540 - base types: int, float
3541 - array: fixed size, string (ie ref to dynamic string)
3542 - structure: homogeneous, heterogeneous
3543 - chained list, graph with cycle
3544 Believe it or not, valgrind is not too unhappy with the results. The
3545 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3549 - Bugfix when using a filename explicitely (instead of '-')
3551 2004-04-09 Martin Quinson
3553 - factorize more code between RL and SG in socket creation
3554 - Complete the implementation and tests of:
3556 o file (only in RL, and mainly for debugging)
3558 I lost 3 days to design a portable address resolver, and then decided
3559 that the prototype mainly have to run on my box.
3560 Addressing portability too early may be like optimizing too early :-/
3562 - use gras_init in the Tests instead of the crappy parse_log_opt
3563 (the latter function is removed)
3564 [Conditional execution]
3565 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3566 [Code reorganisation]
3567 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3568 Build examples against the RL library, since there is no way to disable
3569 its creation for now.
3571 For information, the beginning of coding on GRAS was back in june
3572 2003. I guess that every line has been rewritten at least twice since