1 SimGrid (3.19) NOT RELEASED YET (target: March 20 2018, 16:15:27 UTC)
4 - Fix MSG_task_get_remaining_work_ratio() to return 1.0 for tasks that have
6 - Remove parameter of MSG_process_killall().
7 Resetting the PID was bogus anyway (several actors could have the same PID).
10 - Use a graphical TOC to make it easier to find the documentation you need
11 - Revamp the MSG tutorial
14 - Execution->setHost() can be called after start() to migrate it.
15 - Comm::test_any() is now implemented.
16 - s4u::Actor now has onCreation() and onDestruction() signals.
19 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
22 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
25 - The privatization of global variables via mmap() is now significantly more
26 flexible and allows the use of daemons with SMPI and to add new processes
27 while the simulation is running. (Adding new SMPI instances after the
28 simulation has already started should also be technically doable now.)
29 - Removed index notion from SMPI -- no more "getPid() - 1"!
30 This also means that the trace files
31 now contain actors called rank-<PROCESS-ID>. These are NOT the real
32 ranks, because tracing several communicators right now is not supported.
33 - Because of this, smpi_process_index() has now been deprecated and will
36 - #245: migrating an actor does not migrate its execution
39 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
42 - Remove unused run-time parameter "tracing/onelink-only".
45 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
47 SimGrid (3.18) Released December 24 2017
49 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
51 MSG IS NOW DEPRECATED. Please use S4U for new projects.
52 - Support for MSG should not be removed from SimGrid before 2020,
53 but future evolutions will be limited to the integration of
54 user-provided patches.
57 - s4u::onDeadlock() signal, run before stopping the simulation.
58 - s4u::Exec: asynchronous executions (abstraction of background
60 - s4u::Host->getLoad() returns the achieved speed in flops/s
63 - Link::name() is deprecated and will be removed in v3.21, use
64 Link::getCname() instead.
65 - Mailbox::getName() changed to return a std::string, use
66 Mailbox::getCname() instead to get a char*.
67 - Storage::getName() changed to return a std::string, use
68 Storage::getCname() instead to get a char*.
69 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
70 to have both allocation and deallocation of the map in user space.
73 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
74 The amount of flops remaining to do is a vector, not a scalar.
75 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
76 name implies on both sequential and parallel tasks.
77 - Both changes fix GitHub's #223 using PR #237 as a basis.
78 Thanks Michael Mercier.
79 - Most examples were converted to S4U and hidden elsewhere at tests.
80 Rationale: we still want MSG to work; we want newcomers to use S4U.
83 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
84 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
87 - Switch to the faster dlopen privatization mechanism by default
88 - Documentation improvments
91 - Live migration is getting moved to a plugin. Dirty page tracking is
92 the first part of this plugin. This imply that VM migration is now
93 only possible if one this function is called :
94 - C/MSG: MSG_vm_live_migration_plugin_init()
95 - C/C++: sg_vm_live_migration_plugin_init()
96 - Java: Msg.liveMigrationInit()
97 For C and C++, "simgrid/plugins/live_migration.h" also has to be
101 - Define class simgrid::xbt::Path to manage file names.
102 - Removed unused functions:
103 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
104 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
105 xbt_graph_export_graphviz()
106 - xbt/str.h: xbt_str_join()
107 - Remove unused datatypes:
108 - xbt/heap.h: use std::priority_queue or boost::heap instead
109 - xbt/swag.h: use boost::intrusive::list instead
112 - New link_energy plugin for the consumption of the links.
113 - All of the operations on files and storage contents have been
114 packaged into a plugin (src/plugins/file_system). The current
115 public interface can be found in
116 include/simgrid/plugins/file_system.h
117 To use these functions you now have to initialize the plugin by
118 calling MSG_storage_file_system_init() just after calling
119 MSG_init() or sg_storage_file_system_init() just after creating
123 - Remove the undocumented/untested tag <include>
126 - Remove viva specific tracing as the tool is no longer maintained
129 - #248: Kill of finished processes leads to segfault
130 - #240: xbt_cond_wait_timeout should gracefully return for C apps
131 - #239: please implement signal s4u::onDeadlock()
132 - #230: segfaults when exit() without run()
133 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
134 - #223: MSG_task_get_flops_amount() not working with parallel tasks
135 - #222: Actor::kill() doesn't really kill and segfaults
136 - #221: odd LMM warning when killing an actor
137 - #120: Memory leak when the processes are forcefully killed
139 SimGrid (3.17) Released October 8 2017
141 The Drained Leaks release: (almost) no known leaks despite the tests.
143 Even very long simulations will preserve your memory: our testsuite
144 stresses SimGrid for over 45mn on fast machines for a coverage of
145 over 80%, and there is only one single known leak, of about 4kb.
149 - Comm.detach(): start and forget about asynchronous emission. A cleanup
150 handler may be given to free resources if the comm cannot be completed.
151 - this_actor::send(mailbox) is now mailbox->put()
152 - New: simgrid::s4u::Comm::wait_all()
153 - New: Host.execute() for remote executions.
156 - Unused option network/sender-gap is removed.
159 - Deprecate MSG_task_isend_with_matching(): This unused feature
160 really complicates our internals. Will be removed in v3.20.
163 - Improved context termination. It is now possible to free resources, even
164 when a process is forcibly killed.
165 - ContextBoost: add support for Boost versions above 1.61.
168 - Bring back run-time option --cfg=exception/cutpath to remove exception
170 - Removed unused functions:
171 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
172 xbt_str_rtrim(), xbt_str_trim().
173 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
176 - Removed header files obsolete since SimGrid 3.12:
177 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
179 - #3: SD_exit should be made optional
180 - #120: Memory leak when the processes are forcefully killed
181 - #159: Threading test regression in Actor refcounting
182 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
183 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
184 - #186: Actor::killAll() segfaults if some process is blocked on wait()
185 - #191: VM migration and pstate
186 - #192: Updating the energy consumptions of all hosts crashes with VMs
187 - #195: All actors have PID=0 in the logs
188 - #204: Sometimes segfault with thread contexts and mmap privatization
189 - #222: Actor::kill() doesn't really kill and segfaults
190 - #225: Actor::kill() doesn't really kill when victims are doing a join()
192 ----------------------------------------------------------------------------
194 SimGrid (3.16) Released June 22. 2017.
196 The Blooming Spring Release: developments are budding.
199 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
200 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
201 - Java JAR file should be finally fully working on Mac OSX too.
203 XML platforms: Switch to platform v4.1 format.
204 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
205 - <zone> can be used as a synonym for the now deprecated <as>
206 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
207 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
208 - <actor> can be used as a synonym for the now deprecated <process>
209 - state_file and avail_file periodicity is now easier to express
210 (check the documentation) the old behavior should still work.
211 * Storage: not backward compatible, but it seems that nobody ever used it
212 - Remove attribute 'content_type' of <storage_type>: was never used
213 - Make attribute 'model' of <storage_type> optional: for future usage
214 - Remove Bconnection model property: was never (in)validated. Replaced by
215 the maximum of the read and write bandwidth as a resource constraint.
218 - New and Backwards Compatibility break:
219 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
220 When one wants to get the list of tasks whose states have changed during a
221 simulation round, s/he has to allocate and free a dynar and use it as
222 argument to this function. The former SD_simulate (double how_long)
226 - Allow multicore VMs, with the correct sharing (unless you overcommit)
227 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
230 - New function to update the consumption of all hosts at once.
231 - Fix the model for multi-core hosts, linear in the amount of busy
232 cores with an abnormality for pIdle (see paper).
233 BUG: Still not satisfactory for ptask on mono-cores.
236 - The netzone are now available from the MSG API.
237 The old names still work, but are now deprecated.
240 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
241 - New option: smpi/keep-temps to not cleanup temp files
242 - New option : smpi/shared-malloc-blocksize . Relevant only when global shared
243 mallocs mode is used, allows to change the size of the fake file used
244 (default 1MB), to potentially limit the number of mappings for large runs.
245 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
246 - Fortran ifort and flang compilers support
247 - New RMA calls supported (experimental) :
248 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
249 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
250 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
251 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
252 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
253 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
256 - New: this_actor::isMaestro() returns whether we are in kernel mode.
257 - Behind the scene, ActivityImpl are now automatically refcounted.
258 Removing this nasty bottleneck will greatly simplify our internals.
261 - Replay: New function xbt_replay_action_get():
262 Retrieve the function previously associated to an event type.
263 - DROPPED FUNCTION: xbt_str_varsubst()
264 - DROPPED MODULE: strbuff. We don't need it anymore.
265 - DROPPED MODULE: matrix. We don't need it anymore.
266 - DROPPED MODULE: lib. We don't need it anymore.
268 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
270 SimGrid (3.15) Released March 22. 2017
272 The Spring Release: continuous integration servers become green
273 We fixed even the transient bugs on all target architectures:
274 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
275 Windows, FreeBSD, NetBSD.
277 - Rename NetCards to NetPoints.
278 This was intended to help NS3 users, but that's not a netcard.
279 That's a point in the routing algorithm, let's avoid wrong simplifications.
282 - New: MSG_process_yield(). Stop and yield to other processes.
283 - New: MSG_process_daemon(). Daemon processes are automatically killed
284 when the last non-daemon process terminates
285 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
286 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
287 to make room for the upcoming network energy plugin.
288 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
291 - Ensure that an actor can kill itself with Process::exit()
292 - Kill the obscure NativeException. Nobody want to survive the issues
293 it denotes, so use JniException that is a RuntimeException (not to
294 be caught explicitly).
295 - Partial bug fix in initialization. SimGrid flags on command line were
296 consumed at C level but stayed in the original Java String[] args.
297 This could mess users' args[i] if SG flags were not put at the end of
299 The SimGrid flags are now removed from the Java arguments. However,
300 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
301 if args.length is greater than the number of YOUR OWN ARGUMENTS.
302 It might be if you have --log or --cfg flags in the command line.
303 - Fix numerous memleaks all around the place. In particular, around VMs.
307 - s4u::onPlatformCreated: right before the simulation starts
308 - s4u::onSimulationEnd: right after the main simulation loop
309 - s4u::onTimeAdvance: right after a clock change (time is discrete)
310 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
311 event from the availability_file changes the avail speed.
312 - Links are now usable from s4u
313 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
314 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
315 - New examples: The conversion of MSG examples to S4U has begun
316 - Actors: create, kill, migration, and suspend
317 - Applications: master-worker and token-ring
318 - Action replay: communications and storage
319 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
322 - Backwards Compatibility breaks
323 - The SD_route_get_list and SD_route_get_size functions have been removed.
324 They are replaced by the unique
325 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
326 The route from src to dst is built in the links dynar whose size can
327 be obtained with xbt_dynar_length.
328 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
329 sg_host_route_bandwidth and sg_host_route_latency.
330 Macros ensure the backwards compatibility, but you should fix your code
333 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
334 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
335 - Bump our claim of support from MPI 1.1 to MPI 2.2.
336 We don't support 100% of it, but it should be enough. Contact us if not.
337 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
338 naive and probably inaccurate)
339 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
340 mapped onto a unique small file using some system magic.
341 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
342 - Fortran: cleanups, fixes, support of user-added operations
343 - RMA: MPI_Accumulate are applied in correct order.
344 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
345 - Better support for MPI_IN_PLACE
346 - Support for MPI_Win attrs and keyvals.
347 - Support MPI_Comm_group_create, MPI_Type_size_x
348 - MPI ops should be properly applied to all allowed types
349 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
350 algorithm by a non failing one from mpich.
353 - Kill the fifo data container: we don't use it anymore.
356 - A <cluster> can now be created with different speed values to represent pstates.
357 see examples/platforms/energy_cluster.xml for an example.
359 SimGrid (3.14.159) Released December 28. 2016
361 The previous release was only a crude approximation of a Pi release;
362 this one should be a bit better.
364 - Revert a last minute change that broke on Mac OSX.
365 - Fix the changelog and NEWS files.
366 - Improve the documentation, in particular of the routing module.
367 - Integrate some patches flying around in Debian.
369 SimGrid (3.14) Released December 24. 2016
372 * The whole documentation was reworked and reorganized. There is
373 still much room for improvement here, but we're on it.
376 * We now need python3 for our testing infrastructure.
377 * Model checking needs libevent
380 * Now works on FreeBSD
383 * Privatization now works on FreeBSD
385 * Privatization is now activated by default in smpirun
386 You can turn it off with -no-privatize if something goes wrong.
388 * Call-location tracing for SMPI.
389 You can add the exact location (filename / linenumber) of an MPI call to
390 your trace files and slow down or speed up the simulation between two
391 consecutive calls by using an adjustment file (see the documentation).
393 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
394 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
395 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
396 values were ignored in some cases. The timings of these functions can now
397 be significantly different.
399 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
400 smpi/running-power is renamed to smpi/host-speed
402 * smpi/grow-injected-times option to enable or disable multiplication of the
403 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
404 make simulation less precise (but also much faster).
406 * smpirun script should be (much) faster for large deployments.
408 * SMPI tracing : fixed issue with poor matching of send/receives.
410 * Replay : Fix broken waitall
412 New functions and features
413 * MSG_parallel_task_execute_with_timeout, to timeout computations.
415 Dropped / renamed functions and features
416 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
417 - MSG_mailbox_is_empty() -> Mailbox::empty()
418 - MSG_mailbox_front() -> Mailbox::front()
419 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
420 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
421 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
422 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
423 * Don't pass the free_f parameter to property related functions:
424 - MSG_host_set_property_value()
425 - MSG_as_router_set_property_value()
426 - MSG_storage_set_property_value()
427 * VM properties. Since msg_vm_t are msg_host_t, just use
428 MSG_host_get_property_value() and friends
429 * VM I/O related things:
430 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
431 - Unimplemented save/restore methods
432 * MSG_as_router_get_property_value() was redundent with
433 MSG_environment_as_get_property_value().
434 - Removed MSG_as_router_*propert*() functions
435 - Added MSG_environment_as_set_property_value() for consistency
436 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
437 Well, they are still there for now, but deprecated with a warning.
438 Please switch to xbt_dict_new_homogeneous() before this is removed
440 * Task affinity. Its intended behavior (that was very badly tested
441 and probably not really working) was deceiving what most users
442 would have hoped here.
443 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
444 * Option network/coordinates is now useless and should be dropped.
447 * Added option storage/max_file_descriptors to allow more than 1024 files opened
450 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
451 * Added Dragonfly topology support
454 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
455 EiBps, EBps, Eibps, Ebps for bandwidth.
456 They may become useful to some lucky ones.
459 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
460 * ex/app_tokenring: new example, very similar to the MSG Token Ring
461 * ex/async_waitAll: new example, on asynchronous communications
464 * Memory usage should be decreased for simulations with a large number
465 of processes. This also helps for SMPI.
467 SimGrid (3.13) stable; urgency=low
469 The Half Release, a.k.a. the Zealous Easter Trim.
471 - We removed half of the lines, that were mostly experimental cruft.
472 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
473 (not counting blanks and comments -- according to openhub.net)
474 - The internals are now compiled in C++ (and will soon be clean C++)
475 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
476 59 klines of C, etc. We added only 29 klines of C++ in replacement.
478 * Backwards Compatibility breaks
479 - Removed Lua simulation bindings (switch to C or Java for that).
480 Lua can still be used to describe platforms
481 - Removed Java kernel plug-ins.
482 Will be reintroduced after the ongoing major internals reorg.
484 - the following functions were removed.
485 They were too specific and should be reimplemented in a generic
486 way, with filter function.
487 - MSG_task_listen_from_host
488 - MSG_mailbox_get_count_host_waiting_tasks
489 - MSG_mailbox_put_with_timeout was removed.
490 Please use MSG_task_send_with_timeout instead.
492 - the SD_application_reinit function was removed. It has been a noop for a while.
493 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
494 replaced by a more flexible mechanism.
495 - The following functions thus do not exist anymore
496 - SD_workstation_get_access_mode
497 - SD_workstation_set_access_mode
498 - SD_workstation_get_current_task
499 - Basic estimation functions have been removed but can easily be replaced
500 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
501 - SD_workstation_get_computation_time => amount / sg_host_speed()
503 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
504 Use VM.getSpeed()*load/100 for the legacy behavior.
506 - option enable_tracing was removed. It was not doing anything for a while.
507 - In the ModelChecker:
508 - the model-checker now ptraces the model-checked process which means
509 you cannot use a debugger on the latter anymore (we might make this
510 optional in the feature);
511 - removed soft-dirty page tracking;
512 - remove model-checked side snapshot management,
513 MC_snapshot() and MC_compare_snapshot();
514 - keep the MC_cut() function as a stub (it was not really working
515 in the previous release).
518 * All options are consistently in kebab-case. Old names are kept as alias.
521 * Switch to platform v4 format.
522 - Rename from 'power' to 'speed' the attributes describing the amount of
523 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
524 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
525 - In <host> and <link>, attributes availability and state are gone.
526 It was redundent with state and availability traces, and with peak values.
527 - In <cluster>, attributes availability_file and state_file are gone.
528 It was too complex and unused.
529 - Kill <gpu>. Was not doing anything.
530 - The DOCTYPE points to the right URL:
531 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
532 (the file at this address now documents the changelog since its v1)
533 - A warning is emitted for unit-less values (they are still accepted).
534 - speed. Default: 'f' or 'flops'. Also defined:
535 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
536 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
537 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
538 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
539 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
540 - latency. Default: 's' second. Also defined:
541 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
543 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
544 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
547 * s4u::Host is now the preferred public interface to the Host features.
548 sg_host_* functions are C bindings to the exact same behavior
549 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
552 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
553 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
554 They were deprecated since a few years, and probably did not even compile anymore.
557 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
558 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
559 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
560 encouraged to update their codes. A list of the modified functions can be found at the end of
561 include/simgrid/simdag.h
564 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
565 the lambda or closure passed as a parameter will run in kernel mode.
566 All the callback functions should be rewritten to that interface at some point.
569 * Reorganizing and cleaning the internals all around the place.
572 * Remove old default barrier/bcast buggy algorithms (see #18407)
573 * Various bug fixes to handle more codes
574 * Remove the need for the --foreground option of smpirun (it is still
575 accepted for backward compatibility).
578 * Kill the setset data container: MC don't use it anymore.
579 * Kill the queue data container: it made more sense with GRAS.
580 * Kill the xbt_peer_t data type: it's useless without GRAS.
581 * Kill rm_cb feature of config sets: it was never useful.
582 * Kill graphxml parsing feature. It was not used.
583 * Kill the deprecated code protected by XBT_USE_DEPRECATED
585 - xbt_dynar_sort_strings(), when the content is char*
586 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
587 They throw exceptions on invalid input;
588 * C++ support for declaring CLI flags (simgrid::config::Flag);
589 * class for abstracting different signal backends (simgrid::xbt::signal).
590 with no external dependencies (we need very simple signals).
593 * refactoring and cleanup of the code;
594 * ongoing process to cleanly separate the model-checking algorithms
595 from the code model-checking support.
597 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
599 SimGrid (3.12) stable; urgency=low
601 The Facelift Release.
604 * Require g++ v4.7 at least to not speak prehistorical C++.
605 * Require Boost 1.48 (for signal2 component).
606 * Java must be version 7 at least when activated.
607 * Builds on Windows again (including Java bindings).
608 * Tracing is now always enabled (no way to turn it out)
609 * Remove GTNetS. It was not working anyway.
610 * Various cleanups in the cmake scripts.
611 * Move headers around to sort them out on installed systems:
612 - instr/instr.h -> simgrid/instr.h
613 - instr/jedule/* -> simgrid/jedule
614 - simdag/datatypes.h was removed
615 - simdag/simdag.h -> simgrid/simdag.h
616 - msg/datatypes.h was removed
617 - msg/msg.h -> simgrid/msg.h
620 * Interface improvement:
621 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
622 - Sanitize the interface in MSG_task_ module:
623 - Merge two functions that were close enough but misleading:
624 set_compute_duration(t) -> set_flops_amount(t)
625 get_remaining_computation(t) -> get_flops_amount(t)
626 - set_data_size(t) -> set_bytes_amount(t)
627 get_data_size(t) -> get_bytes_amount(t)
628 - Massive cleanups in the functions related to the energy
629 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
630 - New: MSG_host_get_pstate()
632 - msg/energy/onoff: switching hosts on and off
634 * Interface improvement:
635 - Rename Host.isAvail() to Host.isOn()
636 - Rename Process.currentProcess() to Process.getCurrentProcess()
637 - Rename Task.setDataSize() to Task.setBytesAmount()
638 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
640 - #18874: Actually allows the GC to reclaim tasks
644 - SIMIX_process_throw: raises an exception in a remote process
645 * Refactoring: Separate sub-modules
646 - libsmx: the public interface, as libc in a real system
647 - popping: the strange dance that converts a user request into a kernel handling
648 - smx_context_*: the virtualization mechanisms that embed the user code
649 - smx_*: the handling of each simcalls
651 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
652 * Rename smx_action_t into smx_synchro_t, making explicit that these
653 things are used to synchronize processes with their environment.
654 For example, a communication is a sort of synchronization involving
655 the communicating processes (that may block until the exchange) and
656 the platform. The same can be said from computations, etc.
658 - #18888: segfault when a process ends before its kill_time
662 - 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.
663 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
664 - MPI_Type_set_name, MPI_Type_get_name
665 - MPI_*_c2f and MPI_*_f2c functions
666 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
667 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
668 - Activate a lot of new tests from the mpich 3 testsuite
670 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
671 - InfiniBand network model added : Based on the works of Jerome Vienne
672 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
673 - When smpi/display_timing is set, also display global simulation time and application times
674 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
675 * Collective communications
676 - 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
677 - 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/).
678 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
680 - Replay now uses algorithms from wanted collective selector
681 - Replay can be used with SMP-aware algorithms
682 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
683 - 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.
685 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
686 - Correctly use loopback on fat-tree clusters
687 - Asynchronous small messages shouldn't trigger deadlocks anymore
688 * Energy/DVFS cleanup and improvement
689 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
690 - new: smpi_get_host_pstate()
693 - "Full" network optimization flag was broken since Surf++
694 - Better handling of precision flags in maxmin
695 - Fix bug causing sometimes "Impossible" errors
696 - Properly pass cluster properties to included hosts
697 * Improvement of the Energy plugin.
698 - Always update the consumption before returning that value
699 - New property: watt_off to denote the dissipation when the host is off
700 - New functions getWattMinAt and getWattMaxAt to retrieve the
701 dissipation of pstates that we are not currently at.
702 * Java: class NetworkLink renamed to Link
703 * New function: simcall_process_get_kill_time()
704 * Massive rename s/workstation/host/
705 - That's intrusive, but that's good for the project consistency. Sorry.
706 - Change config option "workstation/model" into "host/model"
709 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
710 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
711 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
712 - xbt_dirname and xbt_basename for non-POSIX systems
714 * The model checker now runs as a separate process.
715 * The model checker runs must now be launched with the new simgrid-mc program.
716 * Record/Replay: the MC can display a textual representation of a path in the
717 execution graph. It can then be replayed outside of the model checker.
719 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
721 SimGrid (3.11) stable; urgency=low
726 * Normalizing pointers addresses tool for better diff between logs
729 * Add cloud examples using new VMs
730 - examples/msg/cloud/two_tasks_vm.tesh
731 - examples/msg/cloud/simple_vm.tesh
732 - examples/java/cloud/cloud.tesh
733 - examples/java/cloud/migration/migration.tesh
734 * Add java surf examples:
735 - examples/java/surfPlugin/surf_plugin.tesh
736 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
737 - examples/java/surfCpuModel/surf_cpu_model.tesh
738 * Add SMPI+MSG example:
739 - examples/smpi/smpi_msg_masterslave/
744 - msg task destroy cancel test
745 - msg_host on/off test
746 * Move all tests in testsuite to teshsuite (adding tesh files)
747 * Restructure teshsuites
748 - one folder for each kind of test
749 * Restructure AddTests.cmake
751 - structure the order of tests (with sections)
754 * Add virtual machine
755 - creation of a VM on a PM
756 - migration of a VM from a PM to another PM
758 - MSG_process_join(msg_process_t process, double timeout)
759 - msg_bar_t MSG_barrier_init(unsigned int count)
760 - int MSG_barrier_wait(msg_bar_t barrier)
761 - void MSG_barrier_destroy(msg_bar_t barrier)
762 - msg_as_t MSG_environment_as_get_by_name(const char * name)
763 * New option "msg/debug_multiple_use" to help debugging when a task is used
766 - Improvements and finalization of MSG_storage, MSG_file APIs and their
768 - Increase code coverage in test suites
772 * Protect context stacks against stack overflow. The number of protected memory
773 pages allocated on the top of each stack (1 by default) can be configured
774 with the new command line option --cfg=contexts/guard_size:<value>.
775 * Simcalls are now generated by a python script that
776 - generates files included by SimGrid
777 - checks that all the functions exist, and proposes prototypes
779 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
780 host_set_data, host_get_data
782 - simcall_process_join(smx_process_t process, double timeout)
783 * Fix bug where sleeping processing could not be suspended.
786 * Translate surf models from C to C++
787 - Generic classes for all models: Model, Resource, Action
788 - A generic interface for each kind of model (CPU, Network, Storage
789 Workstation, WorkstationVM)
791 * Translate surf routings from C to C++
792 * Add callbacks using sigc++ or boost::signals2
793 - Add callback functions for resource creation/destruction
794 - Add callback functions for action state change
795 - Handle Energy as a plugin
796 * Replace swag by boost::intrusive
797 * Add new routing models for clusters. For documentation, see
798 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
799 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
800 parameters for cluster tag
801 - Fat trees, with topology="FAT_TREE" and
802 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
803 - see examples/platforms/torus_cluster.xml and
804 examples/platforms/fat_tree_cluster.xml
808 * Hostfiles support host:nb_processes construct to deploy several processes on
810 * Collective communication algorithms should not crash if used with
811 improper number of nodes and report the error.
812 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
813 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
814 MPI_Dims_create, MPI_Cart_sub are supported.
815 * New interface to use SMPI programmatically (still depends on MSG for
816 some parts, see examples/smpi/smpi_msg_masterslave) :
817 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
821 * Global variables privatization in MPI executables is now performed at runtime
822 with the option smpi/privatize_global_variables (default:no).
823 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
824 dynamic libraries loaded by the application are not privatized (static
825 linking with these libraries is advised in this case)
828 * Options defined in XML work correctly now.
831 * New cmake option, enable_lib_in_jar, to control whether native libraries are
832 copied into simgrid.jar or not (ON by default). Use this option if you want
833 to reduce the size of the installed simgrid.jar, *and* the native libraries
834 are kept installed elsewhere.
835 * Surf binding with SWIG (code generated in maintainer mode only):
836 - plugin to handle callbacks
837 - CPU model only for the moment
840 * Supernovae build mode is definitively removed. It was used to improve
841 inlining and inter-module optimizations. It is nowadays superseded by
842 link-time optimizations commonly available in compilers.
843 * Update ns-3 find lib. Bindings for ns-3 should work again now.
844 * Add boost dependency for surf++
845 * Add new macro for tests
846 - ADD_TESH(name <tesh_args>)
847 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
851 - xbt_bar_t XBT_barrier_init(unsigned int count)
852 - int XBT_barrier_wait(xbt_bar_t barrier)
853 - void XBT_barrier_destroy(xbt_bar_t barrier)
854 * Make the xbt_os_time module public
856 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
858 SimGrid (3.10) stable; urgency=low
860 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
863 * Reintegrate Java to the main archive as desynchronizing these
864 package is not acceptable anymore (Java is now considered stable)
865 * Add explicit synchronization facilities through semaphores
866 * Bug fix: Task.setDataSize() only changed the C world, not the value
867 cached in the Java world
870 * Dramatically change the way files are handled. API and internals changed, but
871 this part of MSG was not considered as production grade either.
872 * Add explicit synchronization facilities through semaphores
873 * Add a new function MSG_host_get_process_list()
874 * Preliminary DVFS support (see examples/msg/energy/ for details)
877 * SMPI is now included directly in the libsimgrid as the windows
878 linker doesn't force us on splitting it anymore.
879 * Improvements of the SMPI replay tool:
880 - Most of the collective communications are now rooted in the same process as
881 in the original application.
882 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
883 used until now). Multiple data types can now be used in a trace.
884 - The replay tool now supports traces produce either by TAU or a modified
886 - Bug Fix: the compute part of the reduce action is now taken into account.
887 - Gatherv collective is now supported
888 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
889 of smpirun outputs time independent traces for the current run. One file
890 is created per process. If too many processes are simulated, this behavior
891 can be changed to one file for all processes by using the
892 tracing/smpi/format/ti_one_file flag
893 * smpirun generates the host file if needed (with given host count and platform)
894 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
895 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
896 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
897 --cfg=smpi/coll_selector:(mpich/ompi)
898 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
900 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
901 happening inside a collective SMPI call.
902 * Fix the behavior of complex data types handling.
903 * Make MPI_Wtime another synchronization point to take computations into
905 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
906 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
907 * Add all missing Fortran bindings, SMPI should work with Fortran 90
908 (no privatization of global variables yet)
909 * Preliminary DVFS support (see examples/smpi/energy/ for details)
912 * Verification of liveness properties is now available for SMPI applications
913 (in addition to MSG applications)
914 * Bugged examples using SMPI in examples/smpi/mc/
915 * Add --cfg=model-check/visited option. Allows the verification of infinite
916 programs. Detection of loops in the execution thanks to the system state
917 comparison and reduction of the state space to explore. Can be combined with
918 DPOR for safety properties.
921 * Allow to change SimGrid configuration (see --help) within the code
922 thanks to SD_config() as it can be done in MSG.
923 * Add a new function SD_task_set_amount() upon user request.
926 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
927 * Remove rule based routing (no more PCRE dependency)
928 * Add a limiter_link option to cluster tag, to specify a maximum reachable
929 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
930 * Add a loopback_bw and loopback_lat options to cluster tag.
931 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
932 routing. Now peers are encapsulated in an AS and have their own private
933 router but this is transparent.
936 * Our own implementation of getline is renamed xbt_getline, and gets
937 used even if the OS provide a getline(). This should reduce the
938 configuration complexity by using the same code on all platforms.
939 * New type: xbt_cfg_elm_boolean.
940 * Allow to use yes/no for boolean configuration options in the command line.
941 * Allow to disable SimGrid cleanups at exit from command line option.
942 There are situations where one may want a simulation to end with an exit.
943 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
944 annoying when scripting around the simulator. Adding a
945 --cfg=clean_atexit:no allows to circumvent this issue.
948 * Lots of memory leaks were corrected in this release.
949 There is no known memory leaks anymore, in all of our 600+ tests.
950 * New command line option --version, to get SimGrid version information.
951 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
953 * Supernovae builds are deprecated, and expected to be removed in the next
956 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
958 SimGrid (3.9) stable; urgency=low
960 The Grasgory release: GRAS is really dead now.
962 * Complete overhaul of the internal host structures scheme.
965 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
966 was considered as experimental and badly maintained since then.
967 * Keeping it was thus a trap to our potential users, that could take
968 it instead of MSG or SMPI by mistake despite is pity state.
969 * GRAS seems to have very few users (if any), and no one volunteered
970 to maintain it further. It also induces a lot of XBT code (for
971 portability sake), that must be maintained too.
972 * For all these reasons, we killed GRAS. If someone wants to revive it
973 in the future, don't cry, our git history still remembers of GRAS.
976 * Major overhaul. Merge our documentation again as time proved that
977 splitting it was really not helping our users.
978 * Further improve the developer documentation to help newcomers
979 hacking on SimGrid itself. The user documentation (and in
980 particular, the beginner documentation) is still in a sorry state.
983 * Now works on Windows too!
984 * Much more extensive test suite, from MPICH
987 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
988 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
989 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
990 * Bug fix: let task be scheduled when the last dependency to be solved is
991 a control dependency.
992 * Remove SD_load_environment_script function.
993 Use the C sg_platf function if you want to declare a platform
997 * New function: MSG_process_get_number()
998 * Old function documented: MSG_config()
999 * Remove MSG_load_platform_script function
1000 Use the C sg_platf function if you want to declare a platform
1004 * Change the default value of the TCP_gamma constant (maximal size of TCP
1005 congestion window) to a more realistic 4MiB value. If you notice changes in
1006 your simulation results, you can fall back to the previous 20k tiny window
1007 by adding --cfg=network/TCP_gamma:20000 on command line.
1008 * (Hopefully) fix a bug wrt periodic availability and state traces
1009 * Bug fix: use default values at start when first event in availability/state
1010 trace is not at time 0.
1013 * remove the "new_" part of function name sg_platf_new_trace_connect
1014 (resulting in sg_platf_trace_connect), since it does not create
1018 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1019 use case where it's really mandatory, and maintaining it was a pain in
1021 * New: xbt_fifo_search(), search an item with a user-provided
1022 comparison function instead of dumb pointer comparison.
1025 * Fix the lua deployment:
1026 Use `simgrid.init_application()` before deployment instead of
1027 `simgrid.msg_register_application()` after.
1030 * Transfer the tracing files into the corresponding modules.
1032 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1034 SimGrid (3.8.1) stable; urgency=low
1036 The "we are told that some people want to also *install* the simgrid
1039 * Add missing file "tesh.1" to the archive.
1041 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1043 SimGrid (3.8) stable; urgency=low
1045 The Psssshiiiit release: SimGrid jumps into the Cloud.
1048 * Add an experimental interface to manipulate VMs. They are mainly
1049 process groups with very few intrinsic semantic, but they should
1050 allow you to build the semantic you want easily.
1051 * New function: MSG_host_set_property_value()
1052 * New function: MSG_process_on_exit(). To clean memory in all cases.
1053 * Bug fixes that made the host (and link) failures unusable.
1054 * Add a way to auto-restart process when the host in which they are
1055 executing comes back (ON_FAILURE="RESTART" on deployment file,
1056 MSG_process_auto_restart_set).
1057 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1058 please stop using the old ones, they are DEPRECATED.
1060 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1061 Please use MSG_init() instead. (reducing the amount of entry
1062 points in the library helps us).
1063 * Make it impossible to link against the wrong version of the lib
1064 * Deprecate MSG_clean(). No need to call it anymore.
1065 * Function MSG_get_host_number() is not deprecated anymore.
1068 * Split the doc into a user guide and a reference guide.
1069 * Start a developper guide to help people hacking on SimGrid.
1072 * Enable tracing by default. This modules rocks you should use it.
1073 * Remove option custom_flags. Now use environment variables CFLAGS
1075 * Use default cmake things to detect lua instead of home grown ones.
1076 * New option "enable_mallocators" to disable mallocators, for debugging
1077 purpose ("on" by default).
1080 * Bug fixes around the resource failures: don't let the processes
1081 survive the host they are running onto.
1082 * Add an interface to auto-restart processes when the host in which they are
1083 executing comes back.
1084 * Ensures that SIMIX_clean is called automatically. It's not part of
1085 the public interface anymore (bindings should be updated).
1088 * Bug fix for when SD_Simulate is called with a positive value: be careful
1089 when comparing doubles. Sometimes they are different for non significant
1091 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1092 parallel task whose initial work is distributed among host according
1093 to the Amdahl's law. Such tasks are created with a parameter alpha
1094 that corresponds to the non-parallelizable part of the computation.
1095 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1096 between two sets of workstations assuming a 1D block distribution (each
1097 workstation owns a similar share of data) on both sides.
1099 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1100 Data redistribution will be automatically scheduled once parent and child
1101 are both scheduled. The filling of computation_amount and
1102 communication_amount structures is now done seamlessly thanks to the chosen
1104 * New function SD_workstation_dump to display various information
1105 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1106 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1107 bandwidth on the route onto which the task is scheduled and the amount of
1109 To divide the nominal bandwidth by 2, the rate then has to be :
1110 rate = bandwidth/(2*amount)
1111 * Compute tasks that have failed can now be rescheduled and executed again
1112 (from their beginning)
1113 * Increasing source code coverage (src/simdag is now covered at 95.8%
1117 * Re-implement time-independent trace replay using SMPI (at the
1118 smpi_smp_* level) instead of MSG. This should replace
1119 examples/msg/actions/actions.c
1120 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1121 hindexed and structs)
1122 * Implement the exchange of non-contiguous data.
1123 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1124 * Correct behavior of smpi/sender_gap and set its default value to 0
1125 * Add option to asynchronously send small messages to allow better
1126 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1127 specifies the size in bytes under which messages will be asynchronously sent.
1128 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1129 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1132 * New C interface to define a platform: XML is now optional.
1133 For more info, please check include/simgrid/platf.h.
1134 * New interface to define random platforms from the C:
1135 For more info, please check include/simgrid/platf_generator.h and
1136 examples/msg/masterslave/masterslave_platfgen.c
1137 * Export a sg_cmdline dynar containing all the arguments we got from
1141 * Two new tracing options for adding comments to trace file so you
1142 can track your experiments (see --help-tracing for details).
1143 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1144 * Adding the SimGrid version that generated the trace file as a comment.
1145 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1146 * Fix to avoid key clashes on Paje links
1147 * Other minor fixes related to the Paje specification
1150 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1151 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1152 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1153 * Greatly improve the robustness of mmalloc to user errors (such as
1154 using an area after freeing it, or freeing it twice)
1156 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1158 SimGrid-java (3.8.1) stable; urgency=low
1160 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1161 Warning: all methods are visible, but only some of them are
1162 implemented so far. Check the source (src/jtrace.c)
1163 for further information.
1164 * New module: org.simgrid.msg.File (SimGrid File management functions)
1165 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1166 * Change the meaning of Process.restart: now restart the process from
1167 the begining, like MSG_process_restart in C.
1168 * Add Process.setAutoRestart: handling of process restart when failed
1170 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1171 you to retrieve the properties of the processes/hosts
1172 * Deprecate Msg.clean(): you can just forget about it now.
1173 * New function Process.getCount(), that only works when compiling
1174 with the not yet released version 3.9 of the C library.
1176 * New context factory based on Coroutines. It mandates a modified JVM
1177 but then, the simulations run about five times faster, and there is
1178 no limit to the amount of processes (beside of the available memory).
1180 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1182 SimGrid (3.7.1) stable; urgency=low
1185 * Restore the prototype of MSG_process_create_with_environment() to
1186 the pre-3.7 situation by removing the kill_time argument.
1187 * Add a MSG_process_set_kill_time() function instead.
1190 * Fix weird behaviors when dealing with parallel tasks.
1193 * Simgrid is now built as a dll.
1194 * Simgrid-java now works on Windows.
1195 * Simgrid-Java is now included into Windows package.
1198 * First pre-build package for MacOSX.
1201 * Fix compilation when using MSG_USE_DEPRECATED.
1202 * Fix some compilation issues on Macs and Windows.
1203 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1204 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1206 We discovered that the Lua console is broken, but we are missing the
1207 manpower to fix it right now. The problem existed in 3.7 too, so we
1208 are not blocking the release for that. Sorry if you depended on this
1209 feature, any help would be really welcome.
1211 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1214 SimGrid-java (3.7.1) stable; urgency=low
1216 The "Java aint got to be bloated and slow" release
1219 * Various internal cleanups and performance improvement
1220 Simulations are expected to run up to twice faster or so
1221 * Make Process.kill(process) an instance method, not a static one
1222 * User processes are not java.lang.Thread subclasses.
1223 This breaks the compatibility (sorry), but previous API was
1224 brain-dead, making it impossible to have non-trivial
1225 initializations in the process constructor.
1226 * Require a full constructor per Process sub-class.
1227 Kinda breaks the compatibility (sorry), but this allows a much more
1228 efficient way to launch the processes at simulation startup.
1229 * Do not embeed our version of semaphores, java 1.5 can be considered
1230 as sufficiently prevalent for us to not dupplicate its features.
1234 * Add examples for almost every part of the API
1235 We spotted and fixed a lot of bugs in the process
1237 * New module: asynchronous communication API
1238 * New function: Process.sleep()
1239 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1240 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1241 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1243 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1245 SimGrid (3.7) stable; urgency=low
1247 The "spring cleanups (before the next Big Project kicks in)" release.
1250 * We can specify the SMPI latency/bandwidth factor with command line
1251 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1252 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1253 You can also use the "config tag" from platform file by adding this line
1254 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1255 (see "example/platforms/tag_config.xml" to use "config tag").
1256 Note that the command line supersedes the platform file configuration.
1257 * Change the correction factors used in LMM model, according to
1258 the latest experiments described in INRIA RR-7821.
1259 Accuracy should be improved this way.
1260 * Use the partial invalidation optimization by default for the
1261 network too. Should produce the exact same results, only faster.
1262 * Major cleanup in surf to merge models and split some optimization
1263 mechanisms from the core of the models. As a result you can now
1264 specify which model to use (e.g., --cfg=network/model:LV08
1265 --cfg=cpu/model:Cas01) and which optimization mode to use
1266 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1267 Incompatible combinations should err at initialization. See
1268 --help-models for the list of all models and optimization modes.
1269 * The CLM03 workstation model was dropped for simplicity because it
1270 used the deprecated CM02 network model. Use default instead.
1271 * Rename the TCP_gamma configuration option to network/TCP_gamma
1272 * Rename the coordinates configuration option to
1273 network/coordinates, and document it
1274 * Use now crosstraffic keyword instead of the terribly misleading
1275 fullduplex keyword. It is activated by default now in the current
1276 default model, use --cfg=network/crosstraffic:0 to turn it off.
1277 * Ongoing refactoring the model parsing to make XML files optional
1278 See include/simgrid/platf.h for details (still to be completed)
1281 * Major overhaul of the documentation. Almost instructive now :/
1282 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1283 functions and friends. This interface was considered as
1284 deprecated since over 2 years, it's time to inform our users that it is.
1285 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1286 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1287 use these (crappy) functions in your code.
1288 These functions will be removed soon. Stop using them now.
1289 * Deprecate MSG_get_host_{table,number}
1290 Implement MSG_hosts_as_dynar() instead.
1291 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1292 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1295 * Stabilize the parallel execution mode of user contexts
1296 * Introduce configuration variables to control parallel execution:
1297 - contexts/synchro: Synchronization mode to use when running
1298 contexts in parallel (either futex, posix or busy_wait)
1299 - contexts/parallel_threshold: Minimal number of user contexts
1300 that must be part of a scheduling round to switch to parallel
1301 execution mode (raw contexts only)
1302 * Fix bugs that prevented to use suspend/resume along with
1303 synchronization structures.
1304 * Fix bugs in process termination that lead to invalid memory access
1305 in very specific conditions.
1308 * Introduce a parallel mode for the models (controlled by surf/nthreads
1309 configuration item). In our tests, running the models in parallel
1310 never lead to any speedups because they are so fast that the gain
1311 of computing each model in parallel does not amortizes the
1312 synchronization costs, even when ultra fast futexes are used.
1313 This is released anyway because YMMV.
1316 * Performance boost by using a swag internally to compute the set of
1317 tasks that are finished and should constitute the return value of
1321 * Enable it by default now that it is considered rather stable.
1324 * Documentation of the tracing functions.
1325 * Performance gains when tracing categorized/uncategorized resource
1326 utilization by avoiding calls to get route when updating resource
1327 variables. LMM constraints are being used instead.
1328 * API changed to set task categories. Use MSG_task_set_category instead
1329 of TRACE_msg_set_task_category, and SD_task_set_category instead
1330 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1331 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1332 * Examples for link user variables added (see at examples/msg/tracing/)
1333 * Deprecated function TRACE_msg_set_process_category completely removed
1334 * Trace header updated according to the latest Paje file format
1335 * Tracing network lazy updates, no longer obligate users to use full updates
1336 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1337 * Experimental: let user code declare/set/push/pop application states for hosts
1338 * API changed to allow the manual creation of graph configuration files
1339 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1342 * Improve the API of Lua MSG bindings, using the Lua spirit.
1343 * Each simulated process now lives in its own Lua world (globals are
1344 automatically duplicated). It helps writing simulators. Will allow
1345 to run Splay programs within SimGrid in the future.
1346 * Add a Chord example in Lua, equivalent to the MSG one.
1349 * Start the implementation of a solution to express temporal
1350 properties, not only local assertions. This is still an
1351 experimental work in progress, stay clear from it to be safe.
1355 - Add new runtime parameters --help-logs and --help-log-categories
1356 to display information about supported logging parameters and
1358 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1359 don't exists anymore.
1360 * Mallocators: allow value NULL for the reset function.
1362 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1363 create homogeneous dictionaries, where all the elements share the
1364 same free function. Non homogeneous dictionaries will be
1365 deprecated in the next release.
1366 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1367 - Multi-level dictionaries are deprecated.
1369 - new function xbt_dynar_search_or_negative() that is useful when
1370 you have less than 2 million elements in your dynar and don't
1371 want of the extra complexity of catching exceptions when the
1372 element is not found.
1374 - Make xbt_os_thread module (for thread portability) public.
1375 Documentation is still to come, sorry.
1377 - Cleanups and simplifications to make it maintainable again.
1378 - Exotic features (such as memalign and valloc) were removed.
1379 - The metadata were extended and improved so that the
1380 model-checker becomes able to explore and inspect the heaps.
1381 - This may induce a performance drop when enable_model-checking is
1382 ON in cmake (even if it's not used in the simulation), but it is
1383 necessary at this point to get MC working.
1385 Turn model-checking OFF if simulation performance matters to you.
1386 Not enabling it at runtime is not enough, disable it in cmake.
1388 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1391 SimGrid (3.6.2) stable; urgency=low
1393 The "Not coding new stuff allows to polish old things" release.
1396 * New bindings to the NS3 packet level simulator (experimental)
1397 * Use the raw (efficient) execution contexts instead of the sysv
1398 (portable) ones when possible.
1399 * libpcre is now mandatory in any cases since not using it led to
1400 severe performance loss and possibly other issues
1401 * Update the XML platforms:
1402 - G5K: include the latest machine in Nancy
1403 - GridPP and LCG: new platforms
1404 * Documentation was partially updated, at least (more to come)
1406 Bug fixes, cosmetics and small improvements
1407 * Free terminated processes before the end of the simulation to avoid
1408 exhausting the memory of users having very dynamic amount of
1410 * Bug fix and cosmetics about canceling non-running tasks
1411 * Bug fix about the dot loader's issues when using libcgraph
1414 * Create an installer for windows with nsis (amd64 and win32)
1415 - Add an hello world project to illustrate simgrid project creation.
1416 - Embed libpcre into the Simgrid installer to avoid
1417 its compilation burden
1418 * The raw execution contexts should work on Apple now
1419 * Port to Windows 64 bits
1420 - Sysv contexts now have an implementation for this arch
1421 - GRAS communication features now support this arch
1422 * Drop support for borland compiler on windows
1423 - this code was not maintained, and we kinda depend on gcc nowadays
1424 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1425 * Fix portability issue on unstable ubuntu: linker became picky on
1428 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1430 SimGrid (3.6.1) stable; urgency=low
1432 The "Oops, we broke Macs too" release
1435 * Fixed contexts detection so that raw ones are used when possible
1436 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1437 leads to a strange error, with user code segfaulting sometimes when
1438 the generated code is not perfectly aligned (which is not
1439 controllable from the user side, depends on the amount of code)
1442 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1444 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1446 SimGrid-java (3.6) unstable; urgency=low
1449 * Split of every thing from simgrid v3.5 into a separate package.
1451 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1453 SimGrid (3.6) stable; urgency=medium
1455 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1458 * Bindings now constitute their own package, separated from the main one.
1459 Rationale: reduce our maintenance nightmare by reducing the module coupling
1460 They will soon be released on their own on gforge.
1462 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1463 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1465 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1466 * It's not quite deprecated for now because we have no replacement,
1467 but it may soon become the case.
1470 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1471 * Fortran: New user-level cache variable to store the rank of the running
1472 process. This improves performance by an order of magnitude.
1473 * C: New coccinelle script to automatically locate and modify global and
1474 local static variables.
1475 * Improved SMPI network model with a sender-side gap to account for multiple
1479 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1480 only say if a communication is finished, no matter whether it succeeded or
1481 failed. You can call MSG_comm_get_status() to know the status of a finished
1483 * New function MSG_task_dsend() to send a task and detach it. When a
1484 communication is detached, you are never notified of its success or failure
1485 and the memory is released automatically once it is finished. This function
1486 is useful when you don't care about the end nor the success of a
1488 * Change the prototypes of action replay. Sorry for inconvenience,
1489 but this is really more efficient this way (and to adapt your code,
1490 you just have to fix the initialization, that shouldn't be too long)
1491 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1492 ever managed to do anything useful with it.
1493 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1494 takes a dynar of communications. It returns immediately and gives the
1495 index of a finished communication (if any).
1496 * New example: a basic implementation of the Chord P2P algorithm.
1499 * New model for multi-core CPUs. You can now use the core attribute to
1500 precise the number of cores of a host. This is a basic model. Every
1501 process running on the host receives at most the power provided in
1502 the DTD (throughput<=power). Total throughput of process cannot exceed
1504 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1505 router linked by an up-link and a down-link (possibly asymmetrical).
1506 This kind of pattern allows to easily build last-mile model style platforms.
1507 Aggregating such patterns in a rule-based AS is thus the technique of
1508 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1509 * New model for Vivaldi routing. We transformed the Vivaldi network model
1510 into a Vivaldi routing model (based on the rule-based model). This allows to
1511 combine Vivaldi based latencies with last-mile platforms.
1514 * Added a check for NaN of IEEE754 infinite in the double entries of
1516 * Introduce a new context factory "raw", highly inspirated from the
1517 ucontext factory, but using manually crafted functions in assembly to
1518 do the work in an efficient manner.
1519 * Allow to change the used context factory at run time, not only at
1520 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1521 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1522 contexts at runtime (only with raw contexts or ucontexts).
1523 * Completely rewrote this module to allow parallel execution of user
1524 processes. Use --cfg=contexts/nthreads:N to execute user processes
1525 with N parallel threads (the default is 1, meaning no parallelism).
1526 * Allow to decide dynamically between sequential and parallel modes.
1527 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1528 processes in parallel only when their number is greater than or equal to P
1530 * Added a check for NaN of IEEE754 infinite in the double entries of
1534 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1535 won't output the state of processes when interrupted with Ctrl-C
1536 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1537 NULL-terminated array. This may solve backward compatibility issues
1538 due to the change to return type of SD_simulate. See also:
1539 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1540 * Add new macros with variable number of arguments.
1541 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1542 - in xbt/asserts.h: xbt_assert
1543 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1544 - in xbt/ex.h: THROWF and RETHROWF.
1545 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1547 * Change xbt_die() to accept a format string with arguments, just like printf.
1548 * New data structure: xbt_lib_t, like a dict but more general and with better
1552 * New configuration options
1553 Options triva/categorized and triva/uncategorized can be used to generate
1554 graph configuration files for Triva visualization tool.
1555 * Configuration option tracing/platform is renamed to tracing/categorized
1556 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1557 * New TRACE_user_link_variable interface:
1558 User provides the name of the link and the tracing variable to attach to it
1559 * the declaration of tracing categories must be done after the environment
1561 * simpler tracing interface, just one way to declare categories
1562 TRACE_category or TRACE_category_with_color, it is up to you
1563 * links in the trace file are again identified by their names
1564 * trace contains the full platform hierarchy exactly as declared using the ASes
1565 * Options tracing/msg/[task|process]:1 groups the process by hosts
1566 for both cases, tasks and processes must have names that are unique during the simulation
1567 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1568 * The experimental option tracing/msg/volume is deprecated
1569 its functionality may be reincorporated if needed
1571 The tracing generates a trace file with unordered timestamped events,
1572 because of the way the core simulator (surf) works. A script available
1573 at the tools directory (fix-paje-trace.sh) can be used to put the events
1574 in order. We have changed the tracing so it can generate ordered timestamped
1575 events in the final trace, but depending on the simulator (and how much time
1576 is simulated) that can lead to a huge memory utilization. It is deactivated
1577 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1579 Build Infrastructure
1580 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1581 - We are trying hard to keep the API stable, but it may happen that
1582 some things change (we're a research project after all, not a
1583 nuclear plant operating system). If such things should happen, you
1584 could rely on that macro to adapt.
1585 - current value: 30600 for 3.06.00, aka 3.6
1586 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1587 a number that can be compared with SIMGRID_VERSION.
1588 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1589 assertions and verbose/debug logging events are disabled at compile time.
1591 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1593 SimGrid (3.5) stable; urgency=medium
1596 * New feature to any SimGrid-based simulator: Model-Checking
1597 Check SIN#1 for more details.
1600 * New Model SMPI (three-interval linear regression for correction factors)
1601 See RR-7426, available at http://hal.inria.fr/inria-00527150
1602 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1603 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1604 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1605 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1606 * See also src/smpi/README
1611 - Tracing API changes: TRACE_start and TRACE_end should not be called
1612 by user-code. They are automatically called by simulators created
1613 with SimDAG, MSG and SMPI if the toolkit is compiled with
1614 tracing_enabled=ON. Categories declaration and utilization remain the
1615 same for MSG and SimDag.
1616 - A function was added to the tracing API to declare categories with
1618 - TRACE_category_with_color (char *category, char *color)
1619 where color must be in the following format
1620 "%f %f %f", red, green, blue
1621 and red, green, blue are float values in the interval [0, 1]
1622 - User can specify NULL as color parameter, or continue calling
1623 TRACE_category (cat)
1624 On that case, the tracing system will define random colors
1625 - The following command-line options are supported:
1626 --cfg=tracing/filename:msg.trace
1627 --cfg=tracing:1 (activate tracing, needed to use others)
1628 --cfg=tracing/platform:1 (categorized resource use)
1629 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1630 --cfg=tracing/msg/task:1 (task creation)
1631 --cfg=tracing/msg/process:1 (process creation, migration)
1632 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1633 --cfg=tracing/smpi:1 (SMPI interface tracing)
1634 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1635 - examples of examples/msg/tracing updated
1637 - DAXLoader and DOTLoader functions can generate tasks with categories
1638 - A new function to attribute a category to SD tasks:
1639 TRACE_sd_set_task_category (SD_task_t task, char *category)
1640 * Tracing the MPI interface implemented by SMPI
1641 - Collective operations are traced with states
1642 - Point-to-Point operations are traced with states/links
1643 - Tracing activated by a parameter "-trace filename" passed
1644 to smpirun during execution (considering that simgrid
1645 is compiled with tracing enabled)
1646 - To run the simulation with gdb, the simulator
1647 accepts --cfg=tracing/smpi:1 to trace SMPI
1648 - tesh files to check if smpi tracing is ok
1649 - See examples/smpi/NAS/DT-trace
1650 * GTNetS tracing re-worked
1651 - adaptation to the tracing system of GTNets to cope
1652 with modifications regarding the fullduplex mode
1653 - new tesh files to check if gtnets tracing is ok
1656 * Asynchronous communications through the functions:
1657 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1658 * New function: MSG_load_platform_script()
1659 to make possible using a lua script instead of XML files to set up platforms
1660 * New function: MSG_set_function
1661 to associate functions to processes, used when bypassing the parser
1662 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1664 Platforms: Add some more examples in examples/platforms
1665 * Grid'5000: see www.grid5000.fr
1666 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1669 * Change the XML format. This is a very important modification. SimGrid
1670 3.5 introduces a new hierarchical format based on the notion of
1671 Autonomous Systems. Compatibility with old format is ensured
1672 through the perl script provided in the install bin directory
1673 bin/simgrid_update_xml.
1674 It is now possible to build platforms with specific routing
1675 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1676 connect several platforms together. We will try to provide soon
1677 set of realistic platforms exploiting these properties (have a
1678 look at examples/platforms/ for the moment).
1679 * Take the opportunity of the XML format change to be a good XML citizen:
1680 rename link:ctn to link_ctn and similar changes (also dealt with by
1682 * Add a new routing scheme (rule-based) using regular expressions. It
1683 enables to have an extremely low memory footprint when the
1684 underlying routing is simple and can be compactly described. You
1685 need to have libpcre4-dev (perl regular expressions) installed if
1686 you want to use this routing scheme.
1687 * Revive the cluster TAG and allow to easily and efficiently (both in
1688 term of memory and speed) connect clusters together. Have a look
1689 at teshsuite/simdag/platforms/ to see how this can be done. With
1690 this tag, you can create clusters with thousands of tasks at no
1691 cost (have a look at examples/platforms/).
1692 Note: clusters are implemented as ASes, so there is no need for an
1693 enclosing AS tag if you have only one cluster in your platform.
1694 * Add new generic functions in the public interface that allows the user
1695 to call SURF 'create_resource' methods from your code (same
1696 functionality as the XML bypass mechanism but with a much lighter
1698 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1699 that uses a piecewise linear approximation to produce better
1700 results when exchanging small messages.
1701 * Add a new parameter to handle correctly full duplex link and account
1702 for interferences between uplink and downlink communications
1703 (activate with --cfg=fullduplex:1).
1706 * Rename the SD_READY (all dependencies are satisfied and task is
1707 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1708 dependencies are satisfied) state.
1709 This prevents a confusion between the notion of "ready to schedule"
1710 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1711 simulated" (SD_RUNNABLE) used by the simulation kernel.
1712 * Change the way a task is considered as ready. Instead of removing
1713 dependencies when a task is done, a counter is decreased. This way,
1714 it is always possible to reach ancestors thanks to the
1715 SD_taks_get_parents function (even after the end of the simulation.)
1716 * Change the return type of SD_Simulate from (SD_task_t*) into
1717 xbt_dynar_t. This function was in handling a dynar internally and
1718 converted it into a NULL terminated array for historical reasons.
1719 * New function SD_dotload(char*) to load a DAG described in dot
1720 format. This loader and the corresponding examples require the
1721 installation of the graphviz library.
1722 * Fix a bug in the management of tasks of size 0 in the surf network
1723 models. This problem was only visible with SIMDAG and you should
1724 thus disregard results produced with earlier versions if you
1725 relied on this feature (some tasks were blocked because of this).
1726 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1727 even though your applications were doing only point-to-point
1728 communications and sequential computations. Now you can really use any
1729 model you want (of course, if you create real parallel tasks, which are
1730 not implemented in most models beside ptaskL07, this will abort).
1731 * Add an example that schedules a DAX on an heterogeneous platform
1732 using a Min-Min strategy.
1733 * New function SD_workstation_get_current_task() that returns the kind
1734 of task currently running on a workstation in the sequential access
1736 * Raise some warnings when unexecuted tasks remains at the end of the
1737 simulation. This is usually caused by cycles in the DAG.
1740 * New function: SIMIX_process_set_function() called by MSG_set_function
1741 * Change the underlying waiting queue in semaphores so that a process
1742 can wait on several of them simultaneously (as in waitany).
1743 * Fix the way to handle tokens in semaphores so that all access patterns
1744 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1745 * kill the dirty pimple SIMIX_message_sizes_output()
1746 Please use (proper) visualization instead
1749 * New data container: setset (set of sets of elements)
1750 * New module: mmalloc (mapped malloc, allowing to have several
1751 independent segments of malloc)
1752 * New function: xbt_dict_cursor_set_data()
1753 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1754 * New function: xbt_dynar_is_empty()
1755 * New function: xbt_fifo_get_last_item()
1756 * Fix xbt_dynar_shrink(): use the right element size.
1757 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1758 uninitialized areas during expand.
1759 * Fix semaphores: previous implementation was severely broken.
1760 * Use library init/fini functions for our initialization.
1761 - you can use logs and other feature as soon as you want in your
1762 code (even before the xbt_init / MSG_init)
1763 - xbt_exit is now a no-op and produce a warning when used.
1766 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1767 benefit from the latest improvement to the simulation kernel.
1768 * Kill measurement sockets for now. If you rely on them, sorry. This
1769 release is not for you. This feature will be reintroduced in the
1770 future, but we cannot delay the release any further.
1771 * New function: gras_msgtype_get_name().
1772 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1773 * Fix (at last) the pmm example: it should not randomly fail anymore.
1775 Build chain: bug fixes and overall polishing
1776 * Cmake is now stable enough. Hence, we killed the autotools.
1777 * Port to windows ( TM :)
1778 * Fix the 'make install' target.
1779 No need to use 'make install-simgrid' anymore
1780 * Introduce a 'make dist' target compiling a *source* archive
1781 'make package' compiles a binary archive
1782 * Compile java files only on need
1783 * Add --cd and --setenv command line options to tesh
1784 * Out of source builds are not fully supported yet, but we are close
1785 * Enable supernovae and optimization flags by default for our users
1788 * Add layer to set up environment directly from lua, without XML.
1789 * The effect of gras_stub_generator can be achieved through
1790 lua too (check examples/gras/console/ping_generator.lua)
1792 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1794 SimGrid (3.4.1) stable; urgency=low
1796 The "Polishing easter eggs is probably a good idea" release.
1797 This is a bug fixes release only.
1801 * Fix a bug preventing the tasks from begin garbage collected.
1804 * Fix a bug occuring when a host involved in a communication fails.
1805 This was not detected properly by the other peer involved in the
1806 communication. Now, it's reported as a network error.
1809 * Warn the user about loop dependencies in data flow of DAX files
1810 * Obey the control-flow dependencies of DAX files
1813 * Add option "enable_smpi" allowing to not compile SMPI.
1814 Probably useful for the (Mac) users experiencing a build error here
1815 * Improve the detection of lua5.1 and ruby1.8
1817 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1819 SimGrid (3.4) stable; urgency=low
1821 The "Easter in Cargese" release. Also known as (major changes):
1823 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1824 ~> bindings were greatly improved
1825 ~> new tracing infrastructure for better visualization introduced
1827 * the "Welcome to configury modernity" release.
1828 ~> we switched from autotools to cmake, and improved our cdash
1831 A more detailled list of changes follow (full detail in svn log).
1833 Java Bindings: Various Cleanups
1834 * (install java-gcj-compat-dev on debian-like to use them)
1835 * Remove put/get: no need to export deprecated interface in Java
1836 Use send/receive instead.
1837 * Cleanup the examples and add a README per directory
1838 * Remove example autoDestination (that's the only way to go now)
1839 * Remove example explicitDestination (was a plain copy of basic)
1840 * Make JniException a runtime exception, so that there is no need to
1841 declare the fact that you may encounter such a beast. I guess that
1842 nobody will ever want to survive such error.
1843 * Create specific errors for each MSG case of failure:
1844 host failure, transfer failure, timeout, task cancelled
1845 * Cleanup the exceptions that may get thrown by each function
1846 * Other internal cleanups in Java bindings. Performance still bad :/
1847 Ruby and Lua Bindings: create them
1848 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1849 * That's new and great, you should try them out.
1850 Same functionalities than Java bindings, only even less polished
1852 * Kill the useless "rate" argument of SD_task_get_execution_time()
1853 Everyone used to provide -1 as a value, it was not used, and the
1854 semantic of a possible use wasn't even clear.
1855 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1856 either as comm costs or compute costs to mean that there is no
1857 such thing for that specific task.
1858 * Add a SD_task_set_name() function
1859 * Fix SD_task_unschedule() on typed tasks
1860 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1861 * In DAX loader, accept useless 'level' attributes to <job> since
1862 LIGO DAGs have them (seem to be to ease graphical representation).
1864 * Add an example masterslave_mailbox.c using send/receive and not
1865 the deprecated put/get interface.
1866 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1867 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1869 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1870 (declare MSG_USE_DEPRECATED to still have the old name)
1871 * Add a MSG_task_set_data() function
1872 * About trace replay (see examples/msg/actions):
1874 - Allow to work with splitted trace files for each process
1875 Give the specific trace file as argument of each process,
1876 and call MSG_action_trace_run(NULL)
1877 You can still have one merged file for all processes.
1878 - Fix implementation of collective operations
1879 * Allow task_execute() on 0-sized tasks (closes #10063)
1881 * This is the first release of SimGrid where SMPI is not considered
1882 beta anymore (even if some corners should still be improved)
1883 * Port over the new SIMIX_network submodule (internal refactoring)
1884 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1885 * Implement more missing elements of the standard:
1887 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1889 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1890 MPI_Scan() MPI_Get_processor_name()
1891 - Added implementation of missing case for Alltoall (warning: it's
1892 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1893 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1894 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1895 SMPI_MPI_Allgatherv()
1896 * Bug fixes include:
1897 - MPI_Waitsome() was broken
1898 - Allow relative includes in smpicc
1899 - Command line cfg argument 'reference_speed' was ignored...
1900 - Some functions did not properly lead to auto-benching of user code
1901 - smpicc passes -O2 by default (just like openmpi one)
1903 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1904 * Bug fixes about timeouts during communications
1905 * add SIMIX_message_sizes_output() as a pimple to write to file the
1906 amount of messages per size. Use gnuplot to get histogram.
1907 Pimple because that's the only user-visible function of simix,
1908 defined directly in xbt.h (irk, sorry)
1910 - Add a SIMIX_sem_get_capacity() function
1911 - Fix interactions with processe resume/suspende
1912 - release_forever() was stupidly broken
1913 - Fix SIMIX_display_process_status() for processes in a semaphore
1914 - Make SIMIX_sem_block_onto() user-visible
1915 * Refactoring context stuff:
1916 - Use pseudo-OOP for better modularity
1917 - reimplement SIMIX_process_kill() without process_schedule() so
1918 that the latter can take as invariant that it is called from
1920 - Merge context_start into context_new for sake of simplicity
1922 * Add a Vivaldi network model, coded live during SUD'10 ;)
1923 * Rename configuration variables to start a hierarchy:
1924 o cpu_model -> cpu/model
1925 o network_model -> network/model
1926 o workstation_model -> workstation/model
1927 * New configuration variables:
1928 o network/bandwidth_factor: correction to bandwith
1929 o network/latency_factor: correction to latency
1930 o netwotk/weight_S: correction to the weight of competing streams
1931 * Add a long description to the models, that users can see with such
1932 argument on the command line: --cfg=cpu/model:help
1933 * --help-models display the long description of all known models
1935 * config: add the ability to set a default value after registration
1936 Does not override any previously set value (e.g. from cmd line)
1937 * dict: allow to have integer key and data.
1938 When so, you need to use the following functions
1939 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1940 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1941 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1942 In contrary to regular dicts, the key is not malloced before copy.
1943 Mixing scalar and regular elements in the same dict is not tested
1945 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1946 Tracing for Visualization:
1947 * SimGrid is now instrumented in order to generate a trace file for
1948 visualization analysis: to use it, need to compile SimGrid with the
1949 "tracing" option enabled, and instrument the program using SimGrid with
1950 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1951 (among other functions).
1952 * The instrumentation only traces the platform utilization for now
1953 * Documentation to use the tracing functions and how to analyze the
1954 traces with the Triva tool is written.
1955 * More information about: SimGrid FAQ (in the section Tracing Simulations
1958 * We moved to cmake as default build system. Autotools support will
1959 be dropped soon. Check the FAQ for more info about how to use it.
1960 * Greatly improved our cdash/ctest interactions
1961 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1962 * Added memory checking tests with valgrind; lot of memleak fixing.
1963 This may be the first release of simgrid with so few memory issues
1964 * Added code coverage tests.
1965 Our coverage is still improvable, but at least we see it on cdash.
1967 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1969 SimGrid (3.3.4) stable; urgency=low
1971 The "Desktop Grid needs love too" release (also called Xmas release).
1973 Models improvements:
1974 * Major speedup in the maxmin system solving by using lazy evaluation
1975 Instead of solving completely the maxmin system at each iteration,
1976 only invalidate (and recompute) the modified parts.
1977 This new feature is enabled in default models but you can try to
1978 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1979 * Cas01 IMproved as default CPU model
1980 This CPU model is the same Cas01 model, but it uses the
1981 maxmin-selective-update flag and a heap structure to manage
1982 actions on SURF kernel.
1983 It reduces the complexity to find the next action to finish and,
1984 consequently, it's faster than the old Cas01.
1985 This is the new default CPU model (Cas01).
1986 * Rename the old Cas01 model to Cas01_fullupdate
1987 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1988 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1989 * CpuTI (CPU Trace Integration)
1990 A new CPU model whose objective is simulate faster when using
1991 availability trace files.
1992 Instead of using a full featured, over engineered maxmin system for
1993 CPU modeling, this model does the pre-integration of traces files
1994 to calculate the amount of CPU power available, and so, executes
1995 faster than the old CPU models.
1996 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1997 * Use LV08 as default network model since it gives better accuracy
1998 for small messages and shouldn't change things for big ones.
1999 Use --cfg=network_model:CM02 to get the previous behavior.
2002 ******************************************
2003 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2004 ******************************************
2005 * The new CPU model may changes simulations!
2006 The point is that events occurring at the exact same timestamp
2007 are not scheduled in the same order with the old and new
2008 version. This may be enough to completely change the execution
2009 of simulations in some cases.
2010 * The new network model will change simulations!
2011 This new model is more realistic than the previous one, so you
2012 should consider redoing your old experiments with this model.
2013 Sorry for the inconvenience.
2016 * Introduce the supernovae compilation mode
2017 When compiled that way, the whole SimGrid (or almost) is put in a
2018 single compilation unit and compiled in one shoot.
2019 This is to help gcc which has difficulties to inline stuff from one
2021 The speedup seem to be above 15%, althrough more tests are needed on
2022 amd64 to confirm that gain.
2025 * Port of MSG's mailbox on top of SIMIX network
2026 The put/get mechanism was greatly simplified on the way.
2029 * New SIMIX network module. Provides:
2030 - Mailbox: rendez-vous mechanism to find with who you want to speak
2031 - Synchronous send/recv: easier and hopefully faster since the
2032 logic is handled in the maestro process directly now
2033 - Asynchronous send/recv: you dreamt of it? It's here now
2034 Too bad that nobody cared enough to propagate the change to MSG.
2035 * Add semaphores as SIMIX synchronization mechanism.
2038 * new function SD_daxload(char*) to load a DAX file
2039 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2040 * Introduce typed tasks. Specify its kind and cost at creation.
2041 At scheduling, just give where it should be placed, and the cost
2042 for each involved resource is automatically computed.
2043 Existing constructors so far (more to come of course):
2044 - SD_task_create_comm_e2e() for end-to-end communication
2045 - SD_task_create_comp_seq() for sequential computation
2046 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2047 * new function SD_task_dump() for debuging display
2048 * new function SD_task_dotty(task,FILE*) writing to file the info
2049 about the task in dotty format
2050 * SD_task_dependency_exists() can now cope with having one of its
2051 arguments NULL. If so, it tests whether the other argument has any
2053 * Add getters on list of preceding/following tasks:
2054 SD_task_get_parents(task) and SD_task_get_children(task)
2055 * Add getters on amount of workstations and list:
2056 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2057 * Add getter on task kind: SD_task_get_kind(task)
2058 * Update the start_time and finish_time of tasks on completion/failure
2059 * Bugfix: Remove task from state swags when destroyed
2062 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2065 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2066 dynar with one separate thread per value of the dynar.
2067 * Change the prototype of xbt_thread_create(), sorry.
2068 Added a boolean parameter indicating whether we want to join this
2069 thread (used in SG only for now)
2070 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2073 * GTNetS wrappers should now be usable again (and betterly tested too)
2074 * Fix a major regression from 3.2 where the timeout provided to
2075 MSG_task_put_with_timeout() was used as absolute time before which
2076 the comm should be done.
2077 * Start to fix the <cluster> tag.
2078 - Internal links should be good now (beside of the loopback, which
2079 use the private link instead)
2080 - paths to the external world is still rather broken
2081 - the <route:multi> tag is just broken. Actually that's brain-dead.
2082 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2083 to make it less stupid
2084 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2085 * Fix a source-level compatibility glitch from 3.2: after defining
2086 MSG_USE_DEPRECATED, you can use the old name
2087 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2088 * Allow to compile from the SVN with automake 1.11
2089 * Fix some problems when using the "start_time" tag in deployment XMLs.
2090 * Fix #8569: XBT/synchro.h has redundant declarations
2091 * Fix #8563: MSG return values and exceptions
2092 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2093 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2094 Thanks to Robson Peixoto.
2095 * Fix a vicious bug in dictionaries inducing that some elements were
2096 not freed on xbt_dict_free()
2098 Portability report of this version:
2099 * Main portability targets:
2100 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2101 on (amd64/i386/ia64)
2102 - mac leopard on i386
2103 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2104 but nothing critical.
2105 * Other platforms: windows, AIX and others were not tested for this release
2107 Timing report of this version:
2108 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2109 scenario parameters). From 8h to a few seconds in desktop grid settings.
2110 * Supernovae brings about 25% speedup on i386.
2112 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2114 SimGrid (3.3.3) stable; urgency=low
2116 The "Need for Speed" release.
2118 The timings done to validate the 3.3.2 were faulty.
2119 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2121 The problem was a conversion from a manually handled vector to
2122 xbt_dynar_t on the critical path.
2123 xbt_dynar_foreach calls functions, inducing stack management crap.
2125 We inlined these functions and xbt_dynar_foreach is now breath taking.
2126 We also inlined xbt_swag_belong on the way.
2128 Here are some approximate speedup measurements (on master/slaves
2129 simulations lasting between 10s and 20s each):
2130 3.3.1 -> 3.3.2: about same performance
2131 3.3.2 -> 3.3.3: 40% speedup
2132 3.3.1 -> 3.3.3: 40% speedup
2133 3.3.1 with inline patch -> 3.3.3: 30% speedup
2135 Our reading is that the refactoring which occurred in 3.3.2 made us
2136 suffer much more from the xbt_dynar_foreach low performance, but
2137 once we solved this, this refactoring proved to be very performance
2138 effective. From the 40% speedup, somehow, 10% are due to the
2139 inlining and 30% to the refactoring.
2141 That's a pitty that gcc cannot inline functions placed in other files
2142 alone. We have to choose between:
2143 - break the encapsulation (by putting private data structures and
2144 accessors in headers files to help gcc)
2145 - live with low performance
2146 - switch to a decent compiler such as icc (not quite possible).
2148 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2150 SimGrid (3.3.2) stable; urgency=low
2152 The "Simplicity does not preceed complexity, but follows it" release.
2154 The main contributors of this release were (lexical order):
2155 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2158 * Extract the routing logic into its own object.
2159 (was dupplicated in network.c and workstation_LV07.c;
2160 Allows to implement other ways of storing that info)
2161 => kill now useless network_card concept
2162 - Use dynar to represent routes (instead of void** + int*)
2163 - kill link_set (use surf_network_model->resource_set instead)
2164 - Add a command-line option to choose the routing schema to use
2165 - Add three new models:
2166 * Floyd (shortest path computed at initialization)
2167 * Dijikstra (shortest path recomputed all the time)
2168 * Cached Dijikstra (shortest path computed on need)
2169 All these models where contributed by Silas De Munck, and are
2170 described in his ICCS09 paper.
2172 * Simplify model declaration
2173 (less redirections, less function to write when defining a model)
2174 - Factorize stuff between models:
2177 surf_model_resource_set(model)
2178 surf_model_resource_by_name(model, name)
2179 - Unify the types of models in s_surf_model_t (using an union)
2180 - Embeed fields of common_public directly into s_surf_model_t
2181 - Rename model methods:
2182 action_free ~> action_unref
2183 action_change_state ~> action_state_set
2184 action_get_state ~> action_state_get
2185 - Change model methods into functions :
2186 (model)->common_public->action_use ~> surf_action_ref
2188 * Implement a generic resource; use it as ancestor to specific ones
2189 (allows to kill duplicated code in models)
2190 Drawback: timer command don't need no name nor properties;
2191 workstation_CLM03 don't need no properties
2192 (but I guess we can live with those few bytes wasted)
2194 * Improve the action object model
2195 - implement a constructor avoiding dupplicated code about field
2196 initialization in generic_action part.
2198 * Kill the SDP model: it has an external dependency, is deprecated
2199 in flavor of modern lmm models, and didn't compile since a while
2202 * Relocation of the context module from XBT to SIMIX.
2203 (the context were decoupled from the simix processes, duplicating a lot of code)
2204 => a lot of code was factorized
2205 - less overhead is introduced during scheduling
2206 - simpler API for the context factory
2207 - the logic for process creation,destruction and manipulation was simplified
2208 * Simplification of the s_smx_process_t data structure.
2209 => accesing the simix level data associated to a process is faster now,
2210 and the code is a lot more readable.
2213 * Implement some more MPI primitives:
2214 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2215 -implementation: Bcast: flat or 2-ary tree (default),
2216 Barrier: 4-ary tree,
2218 Allreduce: Reduce then Bcast
2219 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2220 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2221 Alltoallv: flat tree, like ompi
2223 * Add support for optimized collectives (Bcast is now binomial by default)
2224 * Port smpirun and smpicc to OS X
2227 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2228 Shout out if you used it.
2231 * Display the list of still queued messages in SG mode when existing
2235 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2236 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2237 * Add xbt_str_from_file(FILE*)
2238 * Add xbt_dict_get_key achieving a linear reverse search
2239 * Remove the context module
2241 Portability report of this version:
2242 * Main portability targets:
2243 - Linux(debian)/x86/context
2244 - Linux(debian)/x86/pthreads
2245 - Linux(debian)/amd64/context
2246 - Linux(debian)/amd64/pthreads
2247 On these, we still have the eratic breakages of gras/pmm and
2248 amok/saturate_sg reported in previous version. We still think
2249 that the tests are the cause of the fault, not the tested code.
2251 - Mac OSX Leopard/x86/context
2252 Still false negative in tesh autotesting.
2253 Smpi still fails, but this time because readlink does not accept -f
2254 Everything seems to work properly beside of that.
2257 - AIX version 5.3 (only tested contexts this time)
2258 Smpi still fails there because mktemp is not installed.
2259 Everything seems to work properly beside of that.
2261 I managed to compile it for the first time, but several breakages.
2262 Won't delay the release for this exotic platform.
2264 * Windows: it's still lagging behind. If you want to help, please
2267 Timing report of this version:
2268 This version seem to be more than 5% faster than 3.3.1 (on linux
2269 64bits with contextes). The gain is less than expected, we are
2270 investigating this for next release.
2272 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2274 SimGrid (3.3.1) stable; urgency=low
2277 * Implement a --cfg-help to show existing configuration variables
2278 * Build chain do not require doxygen in maintainer mode
2281 * fix a bug on struct sizeof computation, which prevented the
2282 exchange of arrays of structs in some conditions
2283 - added a regression test about this in datadesc_usage
2284 * Allow the exchange of 0-long dynamic vectors.
2285 - for that, use -1 as indicator of dynamic size instead of 0
2286 - This implied to change any size from unsigned long to long,
2287 reducing a bit communication abilities, but I guess that with
2288 64bits being quite common, this is more than enough.
2289 - This also induce a protocol change, thus bumping network protocol
2290 version from 0 to 1 (if we have external users, we have to get
2291 clean on that point too ;)
2292 - added two regression tests about this in datadesc_usage
2293 * Be more verbose when propagating local exceptions
2294 This helps debugging.
2295 * Display the status of simulated processes when receiving SIGINT in
2299 * Allow to control the simulation from a trace file.
2300 New functions MSG_action_register() and MSG_action_trace_run()
2301 The first one allows to associate a function execution to each
2302 kind of action while the second one parses a trace file and
2303 triggers the corresponding actions within the system.
2304 For now, only a toy example is provided in examples/msg/actions
2305 * Add an exemple of process migration in examples/msg/migration
2306 * Fix a bug in task exchange which broke MSG_task_get_sender()
2307 Add a teshsuite regression test for that.
2308 [Bug: if MSG_task_get_sender() is called after sender exit,
2310 * Fix a bug which prevented suspend/resume to work properly
2311 * Display the status of simulated processes when receiving SIGINT
2312 This fixes a regression of v3.3. due to the introduction of SIMIX
2313 * Bug fixing in failure management:
2314 - trace could not start by a failure at time 0
2315 - failure during communications were not working
2318 * Add SIMIX_process_set_name() to change the name of the current
2319 process in the log messages.
2320 * Store smx_hosts in a dict since we only retrieve them by name
2321 * Move the configuration infrastructure to surf
2324 * Move the configuration infrastructure to surf
2327 * Massive internal cleanups:
2328 - Store internal structures on processes instead of hosts (allows
2329 to have more than one process per host, in addition of being more
2331 - Cleanup the initialization/finalization process
2332 - Kill a whole bunch of unneeded synchronization:
2333 processes run in exclusive manner within the simulator
2334 - Move queues from global tables to process data fields
2336 - now accept -platform and -hostfile arguments
2337 - Pass the right rank value to processes according to the hostfile
2338 * Compile the examples by default, and use them as regression tests
2339 * Implement MPI_Wtime()
2340 * Change the reference speed to a command line option
2343 * TCP_gamma can now be specified as command line option using
2344 --cfg=TCP_gamma:10000000.0
2345 * Change the --surf-path cmd line option into --cfg=path:
2348 * Also include strbuff from xbt.h public header
2349 * xbt_ex_display(): do not free the exception after displaying
2350 This allows to do more with the given exception afterward.
2351 Users should call xbt_ex_free() themselves.
2355 Portability report of this version:
2356 * Main portability targets:
2357 - Linux(debian)/x86/context
2358 - Linux(debian)/x86/pthreads
2359 - Linux(debian)/amd64/context
2360 - Linux(debian)/amd64/pthreads
2361 These targets fail about 1/10 of times on gras/pmm, but we believe
2362 that this is because of the test, not because of simgrid.
2363 amok/saturate_sg fails even more rarely, and the test may not be
2366 - Mac OSX Leopard/x86/context
2367 The test suite still spits tons of errors because some obscure
2368 force prevents us from removing the temporary directories
2369 arguing that they still contain some metadata I've never heard of.
2370 Smpi fails because seq is not installed.
2371 Everything seems to work properly beside of that.
2374 - AIX version 5.3 (both contexts and pthread)
2375 Smpi still fails there because mktemp is not installed.
2376 XML inclusions seems rosty on AIX.
2378 * Windows: it's still lagging behind. If you want to help, please
2381 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2383 SimGrid (3.3) stable; urgency=high
2387 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2388 [Malek Cherier & Mt]
2390 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2393 Until now, GRAS were using MSG as an interface to SURF. It was
2394 quite difficult because both interface have several differences
2395 (MSG channels vs GRAS sockets were the most notable point).
2397 This also opens the gate to SMPI (which should occur soon) and speed
2398 up simulations by to 40% (even if it were not the main goal).
2400 **************************************
2401 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2402 **************************************
2403 The point is that events occuring at the exact same timestamp are
2404 not scheduled in the same order with the old and new version. This
2405 may be enough to completely change the execution of simulations in
2406 some cases. Sorry for the inconvenience.
2408 * Cleanup and upgrade the XML format to push further scalability
2409 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2411 * Improve the testing infrastructure with tesh. Now a very large part of
2412 the code is tested not only by being run but also by checking that the
2413 output match an expected output [Mt].
2415 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2416 is really less memory-demanding, which should allow you to use
2417 larger files in SimGrid [AL].
2419 * Inform valgrind about our contextes, so that it becomes usable
2420 with the default (and more effecient) version of SimGrid
2421 [contributed by Sékou Diakite, many thanks]
2424 * Introduce a listener thread in charge of receiving incoming
2425 messages from the network. It allows to overlap communication and
2426 computation but most notably, it removes some stupid deadlocks due
2427 to the fact that so far, a process could not send and receive at
2428 the same time. This made most non trivial communication schema
2430 * Convert the PIDs from long int to int to match the MSG ones (and
2431 linux ones too) [Mt]
2432 * New function: gras_agent_spawn() to launch a new process on
2433 current host. Only working in simulation for now. [Mt]
2434 * New function: gras_os_hostport() returning a constant form (ie,
2435 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2438 * Make the backtrace of exceptions more human readable [Mt]
2439 * New module: xbt/str [Mt]
2440 a ton of string utility functions (split, join, printf to a newly
2441 allocated buffer, trim, etc)
2442 * New module: xbt/hash [Mt]
2443 SHA1 hashing algorithm (more to come if needed)
2444 * New module: xbt/synchro [Mt]
2445 synchronization tools (mutex and conditions) working the same way
2446 in simulation and in real life (mainly useful for GRAS, but not
2448 * New module: xbt/queue [Mt]
2449 classical producer/consumer synchronization scheme
2450 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2451 (using the classical functions will get serialized) [Mt]
2452 * Make dictionary internal table dynamic. No need to specify its size
2453 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2455 * Make sure the log channels are organized as a tree under windows
2456 (because of ANSI C compatibility issue, any channel were child of
2457 root directly) [Mt].
2460 * Cleaned many thing in surf and fixed a few bugs [AL].
2461 * Add a nice command line configuration mechanism to compose models [AL].
2462 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2463 the previous one (KCCFLN05). It relies on something that looks like
2464 a max-min sharing mechanism but cannot be written as such. A new solver
2465 was thus designed [AL].
2466 * Add a new solver to lmm. Based on Lagrange optimization and
2467 gradient-based descent, it enables to efficiently maximise systems s.a
2469 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2471 This solver enables to propose two new network models for TCP Reno and
2472 TCP Vegas based on Low's work. These models still need to be fully
2473 tested though [Pedro Velho].
2476 * Bug fix in SD_simulate. Now the time bound given as argument is
2478 * Use the new parallel task model (ptask_L07) as default.
2479 * Use the SURF command line configuration mechanism.
2480 * 0-size tasks (for synchronization) should now work.
2482 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2484 SimGrid (3.2) stable; urgency=high
2488 We still experience issues on this platform, but we believe that at
2489 least MSG is usable.
2491 GRAS API BREAKAGE (for simplification purpose, sorry):
2492 * the gras_msgtype_by_name is not used anymore. Instead of
2493 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2494 you can write (and must)
2495 gras_msg_send(toserver, "request", &request);
2496 - If you still want to pass a gras_msgtype_t to the function (to cache
2497 the type and avoid the lookup time), use the gras_msg_send_() variant.
2498 - Impacted functions:
2499 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2500 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2501 * The callbacks are now expected to return 0 when everything went well
2502 (just like the main() function)
2504 GRAS new features and improvements:
2505 * New module mechanism where user code can use per process globals [Mt]
2506 This is similar to gras_userdata_*() functions, but for libraries. It
2507 factorize some code developped over and over in the examples and AMOK.
2508 It has still to be documented and used (only amok/peermanagement is
2510 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2511 loss when they were small enough to fit into the buffer and sent quickly
2512 enough so that they can all get received in one shoot.
2513 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2514 if not found. Use the *_or_null() variant for the old semantic.
2515 * In gras_msg_handle, do not discard messages without callback.
2516 They are probably messages to be explicitly awaited later (ie, proofs of
2517 mis-synchronization in userland since they are sent before being awaited)
2518 No big deal usually.
2519 * gras_socket_meas_send/recv: semantic changed!
2520 The numerical arguments used to be (1) the total amount of data to send
2521 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2522 messages. This was need for the fool willing to send more than MAXINT
2523 bytes on quite fat pipes.
2526 * Do really rename the hostmanagement module to peermanagement. [Mt]
2527 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2528 but this is rather new and this was documented in the module
2529 documentation (poor excuses, I admit)
2530 * Bandwidth measurement semantic changed! This follows the changes to
2531 gras_socket_meas_send/recv explained above.
2534 * A sequential mode has been added to the workstations. When a workstation
2535 is in sequential mode, it can execute only one task, and the other tasks
2536 are waiting in a FIFO. [Christophe Thiery]
2539 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2540 model for SIMDAG. [Christophe Thiery]
2541 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2542 equal to 0, which caused some bad behaviors in
2543 saturated_constraint_set_update. I now use a threshold mechanism like in
2547 * When running manually src/testall, you select specific units [Mt]
2548 testall is the result of our cunit mechanism, and should replace all
2549 the scripty thingy around since bash don't run easily on billware.
2551 * A mallocator system has been added. [Christophe Thiery]
2552 Mallocators allow you to recycle your unused objects instead of freeing them
2553 and allocating new ones.
2555 Documentation update:
2556 * FAQ reworking + New FAQs:
2557 - "Valgrind spits tons of errors!" [Mt]
2558 - "How to repport bugs" [Mt]
2559 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2560 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2561 - Communication time measurement within MSG [AL]
2562 - I experience weird communication times when I change the latency [AL]
2563 * GRAS tutorial [Mt]
2565 - an introduction to the framework and to the used communication model
2566 - an initiatic tour introducing the most proheminent features:
2568 . Lesson 0: Installing GRAS
2569 . Lesson 1: Setting up your own project
2570 o Part 2: Message passing
2571 . Lesson 2: Exchanging simple messages
2572 . Lesson 3: Passing arguments to the processes (in SG)
2573 . Lesson 4: Attaching callbacks to messages
2574 . Lesson 5: Using globals in processes
2575 . Lesson 6: Logging information properly
2576 . Lesson 7: Using internal timers
2577 . Lesson 8: Handling errors through exceptions
2578 . Lesson 9: Exchanging simple data
2579 . Lesson 10: Remote Procedure Calling (RPC)
2580 . Lesson 11: Explicitely waiting for messages
2581 . Recapping of message passing features in GRAS
2582 - A HOWTO section containing:
2583 o HOWTO design a GRAS application
2584 More are due, of course. They will come latter. In the meanwhile, you can
2585 check the examples which are still here.
2587 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2589 SimGrid (3.1) stable; urgency=high
2593 There was a stack corruption somewhere, visible only when optimizing
2594 with these versions. [Vince]
2597 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2598 module that enabled to play with Directed Acyclic Graphs. It is built
2599 directly on top of SURF and provides an API rather close to the old
2600 SG. Some old codes using SG are currently under rewrite to check that
2601 all needful functions are provided. [Christophe Thiery]
2604 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2605 extension of the classical CLM03 model that gracefully handles
2606 failures. This is now the default model for MSG and GRAS. It doesn't
2607 handle parallel tasks yet though. [AL]
2608 * Bug fix: Weights were not correctly set in the network part.
2609 WARNING: This may have resulted in incorrect results with simulations
2610 where there are more than one flow on a given link. [AL]
2613 * After a (long ?) discussion on simgrid-devel, we have decided that the
2614 convention we had on units was stupid. That is why it has been decided
2615 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2616 WARNING : This means that all previous platform files will not work as
2617 such with this version! A warning is issued to ask users to update
2619 A conversion script can be found in the contrib module of the CVS, under
2620 the name contrib/platform_generation/surfxml_update.pl [MQ]
2623 * Bug fix: Processes were started in reverse order, wrt deployment file.
2624 WARNING: if your code relies on this bug, please fix it. [AL]
2625 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2626 being executed on two different locations. [AL]
2627 * Bug fix: Failures are now better supported thanks to Derrick's tests
2628 (there was many failure situations I hadn't thought of and that weren't
2629 correctly handled). [AL]
2630 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2634 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2635 Exception occurring on server-side are propagated back to client (!).
2637 API CHANGE: the callback changed their prototype. Change:
2638 int my_handler(gras_socket_t expeditor, void *payload_data) {
2640 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2641 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2643 * New! function: gras_msg_handleall to deal with all messages arriving
2644 within a given period.
2645 * New! function: gras_socket_server_range to get a server socket in a
2646 range of port numbers (ease to avoid port number conflicts) [MQ]
2647 * New! gras processes display their backtrace when they get a SIGUSR1
2648 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2649 Sweet to debug RL processes [MQ]
2653 - Do not force experiment sizes to be expressed in kb, or it becomes
2654 impossible to measure the latency this way (needs one byte-long tests)
2655 WARNING: this changes the amok_bw_* function semantic. [MQ]
2656 - Implements the link saturation stuff. [MQ]
2657 * Peer management module:
2658 New! module factorizing code that we wrote over and over [MQ].
2661 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2662 - Test units are placed directly into the library code, they get extracted
2663 automatically and placed into the src/testall binary.
2664 - Convert most of the XBT tests to this system.
2665 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2666 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2667 live in the same C code. This required to change a little bit the API
2668 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2669 * New module: sparse graph structure with basic algorithms (this is work
2670 in progress and the API is not considered to be frozen yet). [AL]
2671 * Display more information on backtraces: source line & function names are
2672 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2673 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2674 * Reworked a little bit some #include statements to load only required
2675 headers. Some user code that relied on SimGrid to include stdlib or
2676 stdio may need to include it by themselves. [AL]
2677 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2679 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2680 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2681 but I guess that almost nobody used those parts. [MQ]
2683 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2685 SimGrid (3.0.1) stable; urgency=low
2688 * Unfortunately, I had missed 5 misnamed functions:
2689 xbt_fifo_item_t xbt_fifo_newitem(void);
2690 void xbt_fifo_freeitem(xbt_fifo_item_t);
2691 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2692 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2693 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2694 They're now deprecated. Please use their new versions:
2695 xbt_fifo_item_t xbt_fifo_new_item(void);
2696 void xbt_fifo_free_item(xbt_fifo_item_t);
2697 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2698 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2699 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2701 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2702 * Documentation: xbt_log module unmercifully reworked [MQ]
2703 * Bugfix: there was a problem with the ending of contexts with
2704 the pthread backend. It caused some weird deadlock or behavior
2705 depending on the pthread implementation. [AL]
2706 * Bugfix: get the exceptions raised in the simulator repport where
2707 and why they come from when they are not catched in time [AL, MQ]
2710 * Bugfix: Do repport the error when two non-connected hosts try to
2711 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2714 * Add additionnal checkings on communications. Assert that two
2715 communicating hosts are connected by a set of links... [AL]
2718 * Add additionnal checkings on channel values in communication [AL]
2719 * New: MSG_task_get_source to see on which host a task was generated [HC]
2720 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2721 the number of tasks waiting to be received on channel and sent
2723 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2724 waits for the first task coming from a given host.. [AL]
2726 GRAS new functionnalities: [MQ]
2727 * Enhance the parsing macro to allow the size of multidimentional objects
2728 to be given thru annotations.
2729 * New example (and documentation): Matrix Multiplication a la RPC
2730 (as when I was young!) and fix a bunch of bugs found on the way.
2732 GRAS performance improvements: [MQ]
2734 * Reduce the amount of cbps creation/destruction by making it static to
2735 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2738 * Change libdata to a set so that we can search for stuff by ID (and thus
2739 reduce the insane amount of dict lookups)
2742 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2743 RL communication on top of SG since it's so uneffective.
2744 It may also be used for inter-thread communication in RL, one day.
2745 * Use gras_datadesc_copy() to exchange messages on top of SG
2747 - improve message exchange performance on top of SG
2748 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2749 * Don't exchange on the network the size of the used part of buffer,
2750 instead, specify the possible buffer size to read().
2752 - reduces the amount of read/write calls (one pair per exchange)
2753 - reduces the amount of exchanged data (the size)
2754 - allows to retrieve all arrived data on receiver side, if we don't need
2755 it right now (subsequent read will peek the buffer)
2756 - allows the receiver to proceed with the begining of the stream before
2757 everything is arrived
2758 - make it possible to build an iov transport (using readv/writev)
2760 - take care of the data with non-stable storage (like stacked data),
2762 * If possible, TCP send uses vector I/O (when writev() is here)
2763 - Don't use it for receive since we send data sizes and data on the
2764 same stream, so we wouldn't be able to chain large amount of chunks
2765 before having to flush the stuff to read the size.
2766 * Rework the transport plugin mechanism to simplify it and reduce the
2767 amount of pointer dereferencement when searching for the right function
2770 * I guess that now, we do almost as few system calls as possible while
2771 doing as few data copy as possible.
2773 To improve it further, we could try to send all the sizes first and then
2774 all the data (to use iov on receiving size), but it's only a partial
2775 solution: when you have 2 dimensional data, the sizes of the second
2776 dimension is data of the first dimension, so you need 3 streams.
2778 I'm not sure the potential performance gains justify the coding burden.
2780 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2782 SimGrid (3.00) stable; urgency=high
2785 * New! Give the possibility to hijack the surf parser and thus bypass
2786 MSG_create_environment and MSG_launch_application. Have a look at
2787 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2789 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2791 SimGrid (2.96) unstable; urgency=low
2796 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2797 This deprecates the xbt_error_t mechanisms.
2798 It modifies (simplifies) all XBT and GRAS API.
2799 MSG API keeps unchanged (exceptions raised by XBT are catched from
2800 within MSG and masked with existing error handling facilities)
2803 * New! Add a FATPIPE model. [AL]
2804 * New! Add a parallel task model. [AL]
2805 * New! Add automatically a loopback interface (in the default
2806 network model) if none was precised.
2809 * Bugfix: MSG_process_resume now works with the current running process.
2811 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2812 * Modification of MSG_task_get_compute_duration. Once a task has been
2813 processed, the value returned by this function is now equal to 0. [AL]
2814 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2815 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2816 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2817 of a m_task. For now, MSG_task_cancel only works with computation
2819 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2820 of the processor (in Mflop/s) regardless of the current load on the
2822 * API Change: use proper naming convention for MSG_getClock and
2823 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2825 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2826 This function changes the priority of a computation task. This priority
2827 doesn't affect the transfer rate. A priority of 2 will make a task
2828 receive two times more cpu power than the other ones. This function
2829 has been added to suit the needs of Nguyen The Loc and hasn't been that
2830 much tested yet. So if it fails, please report it and send me your code.
2832 * API Change: removed all functions and types that were marked "deprecated"
2833 since many months. Renamed MSG_global_init_args to MSG_global_init.
2835 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2837 SimGrid (2.95) unstable; urgency=low
2840 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2841 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2842 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2845 * Bugfix: Allow absolute paths for platform description files [MQ]
2846 * Bugfix: do free the variables after use. Leads to drastic performance
2848 * Implement max_duration (ie, timeouts) on resources [AL]
2851 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2853 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2854 the needed support (timeouts) [AL]
2857 * Implement measurement sockets. You can now get the bandwidth between two
2858 hosts thanks to AMOK (see below). [MQ]
2859 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2860 dynar over the network (yeah) [MQ]
2861 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2862 * simple/static token-ring example. [Alexandre Colucci and MQ]
2863 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2864 in place before [MQ]
2867 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2868 type parsing. "short" and "long" modifiers were also ignored; other
2869 modifier (such as reference level) are still ignored. [MQ]
2870 * Embeed the buffer size within the buffer itself on SG. [MQ]
2871 That way, send() are atomic and cannot get intermixed anymore (at least
2872 the ones which are less than 100k; bigger messages still have the issue)
2873 * Array size pushed by the field, not by the field type (or each
2874 and every long int will push stuff to the cbps) [MQ]
2875 * use select() to sleep since it allows to portably sleep less than one
2878 GRAS (minor cleanups)
2879 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2882 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2883 * type category 'ignored' killed as it was never used and were difficult
2885 * whether a type can cycle or not is now a flag, leaving room for more
2886 flags (as "ignored", if we feel the need one day ;)
2887 * Rename raw sockets to measurement sockets since "raw" has another
2888 meaning in networking community.
2891 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2892 and offers features not belonging to GRAS but that most applications
2893 need. One day, it may be a set of plugins loadable at runtime.
2894 * New module: bandwidth
2895 bandwidth measurement between arbitrary nodes running this module. [MQ]
2897 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2899 SimGrid (2.94) unstable; urgency=low
2901 The first beta release of SimGrid 3 !
2905 * Update the main page and the FAQ. Adding references to gforge.
2908 * Add a gras_os_getpid function.
2911 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2912 * Extend the logs so that they also print PID, hostname, date, ... if
2914 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2915 and kill the old version which were in testsuite/
2916 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2920 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2921 for users. Instead, each programming environment (GRAS, MSG,...) use it
2922 when needed to provide such a feature to users.
2923 Don't shortcut the mechanism or you will also shortcut the virtualization
2924 you need on the simulator.
2928 * Cleanups in configury with regard to compile optimization/warning flags.
2929 Also add -fno-loop-optimize to any powerpc since it's the optimization
2930 killing gcc (< 3.4.0).
2931 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2932 needed by MSG examples complications
2933 * Borrow configury beautifications from PHP
2936 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2937 warning (thanks loris for stumbling into this one).
2938 * Bugfix: stop loading private headers (gras_config.h) from the public
2942 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2943 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2944 RL would imply a DNS resolver, which is *hard* to do in a portable way
2945 (and therefore delayed).
2946 * Implement a real timer mechanism and use it in timing macros. This allows
2947 to avoid rounding errors and get a 0.000005 sec precision in timing
2948 macros. While I was at it, various cleanups:
2949 - allow to declare more than one timed section per file (fix a stupid bug)
2950 - move some private declaration to the right place
2951 - merge conditional execution and timing macros into emulation module
2952 - document the module
2953 - make sure the module cleanups its mess on gras_exit
2954 * Documentation improvements:
2955 - (new) how to compile applications using GRAS
2956 - (new) emulation support (timing macros)
2958 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2960 SimGrid (2.93) unstable; urgency=low
2962 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2965 - Use Paje properly where used. Still to be sanitized properly.
2966 - Portability fix: Add an implementation of the contexts using pthread
2970 - Add xbt_procname(): returns the name of the current process.
2971 Use it to show the current process's name in all logging.
2973 - fix detection of older flex version and the reaction, since we do
2974 depend on modern ones (we use lex_destroy)
2975 - Better separation of SG and RL in the libs: remove all simulation code
2976 from libgras. As a result, this lib is now only 200k when stripped.
2977 Some of the xbt modules may also be duplicated (two sets and such) and
2978 should be cleaned/killed before SG3.
2979 - Insist on using xlC on AIX because of weird problems involving gcc there.
2980 - Cleanup the make remote stuff. This is now done by scripts
2981 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2982 mainly for our private use, but we're working on changing them to user
2985 - Bugfix: flush the socket on close only if there is some *output*.
2986 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2988 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2991 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2993 SimGrid (2.92) unstable; urgency=low
2995 Alpha 3 on the path to SimGrid 3
2999 - New! First try of benchmarking macros.
3000 - New! First try so that gras_stub_generator generate deployment and
3001 remote compilation helpers.
3003 - Bugfix: Initialization fix in msg_test.
3007 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3009 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3011 - API Change: gras_os_sleep to take the amount of seconds as a double.
3012 Accepting an int was error prone since it was the only location where
3013 seconds were coded as such. It leaded to damn rounding errors.
3014 - Bugfix: Hard to belive that timers ever worked before this.
3016 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3018 SimGrid (2.91) unstable; urgency=low
3020 Alpha 2 on the path to SimGrid 3
3024 - Bug fix in the lmm_solver.
3026 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3027 through the function MSG_paje_output.
3028 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3029 - It is possible to bound the rate of a communication in MSG with
3030 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3031 to write it in the changelog).
3032 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3037 - Various cleanups to the autotools stuff
3038 - Begin to move Gras examples to examples/gras/
3039 - Let make distcheck work again (yeah!)
3041 - documentation overhauled using doxygen.
3042 gtk-doc-tools is dead in SimGrid now.
3043 - Automatically extract all existing logging categories, and add the list
3044 to the documentation (long standing one, to say the less)
3046 - Cleanup the known architecture table. Reorder the entries to group what
3047 should be, and use a more consistent naming scheme.
3048 (some of the test dataset are still to be regenerated)
3049 - New! Allow library to register globals on each process just as userdata
3051 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3052 lookup time (for now).
3053 Use it in msg and trp.
3054 This cleans a lot the internals and helps enforcing privacy of the
3055 headers between the gras components.
3056 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3057 - Bugfix: gras_os_time was delirious in RL.
3058 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3060 - Reenable GRAS now that it works.
3062 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3064 SimGrid (2.90) unstable; urgency=low
3066 Alpha 1 on the path to SimGrid 3
3068 * It is a long time since the last release of SimGrid. I'm sorry about
3069 that but as I had told you, I was rewriting a lot of things. I apologize
3070 to those who had been reporting bugs to me and that I had not answered.
3071 If your bug is still in the new version, please tell me. Here is a
3072 summary of the main changes.
3074 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3075 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3076 portability, speed, and a lot more but you'll figure it out later.
3077 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
3078 I wanted to release the new MSG as soon as possible and I have
3079 broken GRAS, which is the reason why, for now, only MSG is fully
3080 functional. A laconic description of these projects is available
3081 in the documentation.
3083 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3084 kernel optimized for our needs (called SURF but only the developers
3085 should use it). Hence, MSG is now roughly 30 times faster and I think
3086 that by rewriting a little bit MSG, I could event speed it up a little
3087 bit more. Beside the gain in speed, it is also much easier to encode a
3088 new platform model with SURF than it was with SG. More to come...
3090 * REVOLUTION 3: I have tried to change a little as possible the API of
3091 MSG but a few things really had to disappear. The main differences
3092 with the previous version are :
3093 1) no more m_links_t and the corresponding functions. Platforms are
3094 directly read from a XML description and cannot be hard-coded
3095 anymore. The same format is used for application deployment
3096 description. The new format is described in the documentation.
3097 Have a look in tools/platform_generation. There is a tiny script
3098 that converts from the old platform format to the new one. Concerning
3099 the application deployment format, parsing the old one is tricky.
3100 I think most of you should however be able to convert your files. If
3101 it is really an issue, I can write a C code that does the conversion.
3103 2) the toolbox tbx does not exist anymore. We now have a library
3104 with much more data-structures but without the hash-tables (we have
3105 dictionaries that are much faster).
3107 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3109 *****************************************************************************
3110 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3111 * don't feel like converting the oldies. (Mt) *
3112 *****************************************************************************
3115 Version 2.90: "the long awaited one"
3116 - Finished rewriting and debugging MSG. Rewrote the documentation.
3117 - disable GRAS for now since it needs to be ported to the newest SG
3120 - Finish the port to windows (using mingw32 for cross-compile)
3123 - Main loop and datastructures of SURF. A cpu resource object is
3124 functional. Surf can thus be used to create cpu's with variable
3125 performance on which you can execute some actions.
3127 2004-11-15 Martin Quinson
3128 - Port to ARM. Simply added the alignment and size descriptions. Should
3129 work, but the ARM machines are so slow that I didn't had the opportunity
3130 to 'make check' over there yet.
3132 2004-11-15 Arnaud Legrand
3133 - Trace manager now written. It uses a heap structure and is therefore
3134 expected to be efficient. It may however be speeded up (particularly
3135 when many events occur at the same date) by using red and black
3136 trees. One day maybe...
3137 - Max-min linear system solver written. It uses a sparse matrix
3138 structure taking advantage of its expected use. Most operations are
3139 O(1) and free/calloc are called as few as possible. The computation of
3140 the minimum could however be improved by using a red and black tree
3143 2004-11-03 Arnaud Legrand
3144 - Rename every gras_* function that was in xbt/ to its xbt_
3146 - Add a heap and a doubly-linked list to xbt
3147 - Added a dichotomy to the dictionaries. make check works as well before
3148 so I assume that the patch is correct. I do not know however if things
3149 run effectively faster than before now. :)
3151 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3152 SimGrid, and the version number is bumped to 2.x
3154 2004-10-29 Martin Quinson
3155 - Introduction of the remote errors.
3156 They are the result of a RMI/RPC on the remote machine.
3157 ErrCodes being scalar values, you can't get the host on which those
3158 errors did happen. Extending the error mechanism as in Gnome is possible.
3159 No idea yet whether it is a good idea.
3161 2004-10-28 Martin Quinson
3162 - Interface revolution: the Starred Structure Eradication.
3163 I used to do typedef struct {} toto_t; and then handle *toto_t.
3164 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3165 - ???_t is a valid type (builded with typedef)
3166 - s_toto_t is a structure (access to fields with .)
3167 - s_toto is a structure needing 'struct' keyword to be used
3168 - e_toto_t is an enum
3169 - toto_t is an 'object' (struct*)
3171 typedef struct s_toto {} s_toto_t, *toto_t;
3172 typedef enum {} e_toto_t;
3173 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3174 s_toto_t) is private.
3176 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3177 so that it can set the variable to NULL. It was so for dicts and sets,
3178 it changed for dynars.
3180 - Fix a bunch of memleaks in dict_remove
3181 - Fix a bug in sg/server_socket opening: it failed all the time.
3183 2004-10-07 Martin Quinson
3184 - Speed up dynar lookup operation a bit.
3186 gras_dynar_get is dead.
3188 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3189 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3190 (which gives you the address of the stored data).
3192 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3193 copy of the data using an affectation to do the job and not a memcpy.
3195 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3196 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3198 It does not really speedup the dynar test because they are
3199 setting elements all the time (and look them seldom). But the dict does
3200 far more lookup than setting.
3202 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3204 2004-10-05 Martin Quinson
3205 - Allow to (en/dis)able the cycle detection at run time.
3207 Whether we should check for cycle or not is now a property of each
3208 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3209 datadesc_cycle_unset allow to remove this property when previously set.
3211 Note that the cycle detection is off by default since it impacts the
3212 performance. Watch the data you feed GRAS with ;)
3214 This property is hereditary. Any element embedded in a structure having it
3215 set have it set for the time of this data exchange.
3217 You should set it both on sender and receiver side. If you don't set it on
3218 sender side, it will enter an endless loop. If you forget on receiver
3219 side, the cycles won't be recreated after communication.
3221 - Header reorganization.
3222 Kill gras_private.h, each submodule must load the headers it needs.
3224 2004-10-04 Martin Quinson
3225 - Interface revolution: do not try to survive to malloc failure.
3227 Now, gras_malloc and friends call gras_abort() on failure.
3228 As a conclusion, malloc_error is not a valid error anymore, and all
3229 functions for which it was the only gras_error_t return value are
3230 changed. They now return void, or there result directly.
3231 This simplify the API a lot.
3233 2004-09-29 Martin Quinson
3234 - Re-enable raw sockets.
3235 Created by gras_socket_{client,server}_ext;
3236 Used with gras_raw_{send,recv}
3239 It should allow to kill the last bits of gras first version soon.
3241 This is not completely satisfactory yet (duplicate code with
3242 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3245 - Simplify transport plugin (internal) interface by not passing any
3246 argument to _server and _client, but embedding them in the socket
3249 2004-09-28 Martin Quinson
3250 - Finish the port to AIX.
3251 autoconf was my problem (segfault within the malloc replacement
3252 function. No idea why)
3254 2004-09-16 Martin Quinson
3255 - Fix some size_t madness on 64bit architectures.
3257 2004-09-08 Martin Quinson
3258 - Reduce the number of system headers loaded, overload some more system
3259 calls (such as malloc to cast the result of the system one, and work
3261 - Fix and reintroduce the config support
3263 2004-09-07 Martin Quinson
3264 - Source code reorganization to allow Arnaud to surf all over there.
3265 - Allow to document the logging categories.
3266 - Remove all uppercase from logging categories and useless cleanup in names.
3268 2004-08-18 Martin Quinson
3269 Version 0.6.2 (protocol not changed; API changed)
3270 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3271 gras_error_t), and NULL when not found. Functions expecting a msgtype
3272 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3273 hopefully usefull message.
3274 - Portability to prehistoric sparcs again
3276 2004-08-17 Martin Quinson
3277 Version 0.6.1 (protocol not changed; ABI not changed)
3278 - prealloc some buffers to speed things up
3280 2004-08-11 Martin Quinson
3281 Version 0.6 (protocol not changed; ABI expended)
3282 - The parsing macro can deal with the references, provided that you add
3283 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3285 2004-08-09 Martin Quinson
3286 Version 0.5 (protocol not changed; ABI changed)
3287 - Allow to off turn the cycle detection code in data exchange at
3288 compilation time. It should be at run time, but I'm short of time (and
3289 the config stuff is still broken). That way, we keep dict out of the
3290 critical path, which is good because the performance is poor:
3291 - search not dichotomial yet
3292 - dynar give no way to access their content and memcpy everytime
3293 - In composed data description (struct, ref and so on), stop foolness of
3294 keeping the subtype's ID, but store the type itself. This keeps sets out
3295 of the critical path, which is good since they rely on dynar and
3296 dictionnaries. The only loose of that is that we cannot detect the
3297 redeclaration of a structure/union with another content (but I'm not sure
3298 the code detected well this error before anyway). We still can detect
3299 the redefinition discrepancy for the other types.
3300 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3301 it breaks the code because of type-punning used all over the place).
3302 This breaks on all non-gcc architectures (for now).
3304 All those changes (plus the buffer of last time) allow me to gain 2 order
3305 of magnitude on cruel tests consisting of 800000 array of integers on two
3306 level of a hierarchical structure (200 secondes -> 4 secondes)
3309 - the selector of reference must now return the type it points to, not
3310 the ID of this type.
3312 2004-08-06 Martin Quinson
3313 Version 0.4 (protocol changed; ABI not changed)
3314 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3315 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3316 - Implement a buffer transport, and use it by default (it relies on tcp in
3317 real life and on sg in simulation).
3318 That's a bit hackish since I had a new field to the structure to store
3319 its data without interfering with the subtype ones. Inheritance
3320 is tricky in C. And that's a kind of reverse inheritance with one class
3321 derivating two classes. Or maybe a game with java interfaces. Anyway,
3322 that's damn hard in C (at least).
3323 Moreover, I got tired while trying to ensure plugin separation and
3324 genericity in SG mode. MSG wants me to do weird things, so let's go for
3325 cruel hacks (temporarily of course ;).
3326 See comment in transport_private.h:71
3327 - do not include all the _interface headers in private but in the files
3328 which really need them (to cut the compilation time when they are
3331 2004-07-26 Martin Quinson
3332 Version 0.3 (protocol not changed; ABI changed)
3333 - Major overhault of the datadesc interface to simplify it:
3334 - shorted the function names:
3335 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3336 - add a trivial way to push/pop integers into the cbps without malloc.
3337 This allows to make really generic sub_type description, which simply
3338 pop their size of the stack.
3339 - add a function gras_datadesc_ref_pop_arr() which does what users want
3340 most of the time: Declare a dynamic array (which pops its size of the
3341 stack) and declare a reference to it. Poor name, but anyway.
3342 - kill the post-send callback, add a post-receive one
3344 2004-07-23 Martin Quinson
3345 Version 0.2 (protocol changed; ABI changed)
3346 - add some testing for cpbs in the test cases, and fix some more bugs.
3347 This invalidate again the little64 data file, since I cannot regenerate
3349 - remove an awfull optimization in the logging stuff, allowing me to:
3350 - understand it again
3351 - learn gcc how to check that the argument match the provided format
3352 - fix all errors revealed by gcc after that
3353 - internal keys of dict are not \0 terminated. Deal with it properly in
3354 loggings instead of segfaulting when the user want to see the logs :-/
3356 2004-07-22 Martin Quinson
3357 - Fix some stupid bug preventing cbps (callback postit) from working
3359 2004-07-21 Martin Quinson
3360 - Some documentation cleanups
3361 - remove the useless last argument of msgtype_declare
3362 - rename the Virtu functions to fit into the 'os' namespace
3363 - move headers src/include -> src/include/gras/ and stop fooling with
3364 gras -> . symbolic link
3365 - make distcheck is now successful
3367 2004-07-19 Martin Quinson
3369 - Build shared library also
3370 - Install html doc to the right location
3371 - stop removing maintainer files in make clean
3372 - build tests only on make check
3374 2004-07-13 Martin Quinson
3376 - No major issue in previous version => change versioning schema
3377 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3378 corresponding dataset.
3380 2004-07-11 Martin Quinson
3382 - Get it working with any kind of structure (we can compute the padding
3383 bytes remotely for all the architectures I have access to)
3384 - Implement the structure parsing macro (still not quite robust/complete)
3385 - Improvement to the remote testing toysuite
3387 2004-07-10 Martin Quinson
3388 [autoconf mechanism]
3389 - get ride of a bunch of deprecated macros
3390 - actually run the test for two-compliment, not only compile it :-/
3391 - test whether the structures get packed (and bail out if yes. Damn.
3392 Alignment is a serious matter)
3393 - test whether the structures get compacted (but respecting the alignment
3394 constraints of each types)
3395 - test whether the array fields of structures can straddle alignment boundaries
3397 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3398 (took me 2 hours to find that bug, looking at the wrong place)
3399 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3400 before sending/receiving and are used to compute the offsets of fields
3401 - Given that padding size depend even on compiler options, keep track of
3402 alignment and aligned_size only for the current architecture. Not a big
3403 deal since we send structure fields one after the other (seems
3405 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3406 program, let it work on linux/gcc/little32. portability todo.
3408 2004-07-08 Martin Quinson
3409 - import and improve remote compilation support from FAST
3410 - make sure make check works on half a dozen of machines out there
3412 2004-07-07 Martin Quinson
3413 Let's say it's version 0.0.3 ;)
3414 - Implement conversions (yuhu!)
3415 - Let it work on solaris (beside conversion, of course)
3416 - Stupid me, using rand() to generate the conversion datatests in not wise.
3418 2004-07-06 Martin Quinson
3419 - Let make dist work, since I'm gonna need it to compile on remote hosts
3420 - Let Tests/datadesc_usage write the architecture on which the file was
3421 generated as first byte.
3422 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3423 architecture descriptions.
3424 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3425 on those architectures.
3426 - Optimization: send/recv array of scalar in one shoot
3428 2004-07-05 Martin Quinson
3429 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3431 - Plug a whole bunch of memleaks
3432 - each process now have to call gras_{init,exit}. One day, their log
3433 settings will be separated
3434 - Continue the code factorisation between SG, RL and common in Transport.
3436 2004-07-04 Martin Quinson
3438 - Redistribution between SG and RL.
3439 We wanna have to accept in SG, so move accepted related parts of RL in
3440 the common part. (more precisely, the dynar of all known sockets is no
3441 more a static in transport.c, but part of the process_data)
3443 [gras_stub_generator]
3444 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3446 2004-07-03 Martin Quinson
3447 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3448 them all up in one shot)
3450 - Fix the ugly case of reference to dynamic array.
3451 - New (semi-public) function gras_datadesc_size to allow the messaging
3452 layer to malloc the needed space for the buffer.
3454 - gras_socket_close now expect the socket to close (and not its address to
3455 put NULL in it after it). This is because the socket passed to handlers
3456 is one of their argument (=> not writable).
3458 - propagate the interface cleanup from last week in datadesc, ie remove a
3459 superfluous level of indirection. User pass adress of variable
3460 containing data (both when sending and receiving), and not of a variable
3461 being a pointer to the data. Let's say that I like it better ;)
3462 The price for that is constructs like "int msg=*(int*)payload" in
3463 handlers, but it's a fine price, IMHO.
3465 - Let it work in RL (yuhu)
3467 2004-06-21 Martin Quinson
3469 - porting SG plugin and SG select to new standards (works almost).
3470 - plug memleaks and fix bugs around.
3473 - cleanup the prototype of data recv and force users to specify when they
3474 want to handle references to objects. Test case working even for cycles.
3475 - plug memleaks. Valgrind is perfectly ok with this.
3477 2004-06-12 Martin Quinson
3479 - cleanup the separation between plugin and main code in plugin creation
3481 2004-06-11 Martin Quinson
3483 - Reput hook for raw sockets, needed for BW experiments
3484 - kill a few lines of dead code
3485 [Data description] Interface cleanup
3486 - gras_datadesc_by_name returns the searched type or NULL.
3487 That way, no variable is needed to use a type desc once, which makes
3489 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3490 two parameters were strings (field name, type name), leading to
3492 [Dicos] Interface cleanup
3493 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3494 This is consistant with the dynar API.
3496 2004-04-21 Martin Quinson
3498 - Porting to new standards.
3500 - interface cleanup.
3501 There is no bag anymore, no need to take extra provision to mask the
3502 pointers behind "ID".
3503 Better splitup of functions between files create/exchange/convert.
3504 This is still a bit artificial since convert and receive are so
3505 interleaved, but anyway.
3507 - add a queued message list to procdata (the ones not matching criteria
3509 - factorize some more code between SG and RL wrt procdata
3511 - use gras_exit in example to track memleaks
3512 - get rid of gs_example now that GS is properly integrated into gras
3513 - update run_test to integrate the latest tests (datadesc)
3515 - rename WARNINGn macros to WARNn since it prooved error-prone
3517 2004-04-19 Martin Quinson
3519 - register init/exit functions within gras module mechanism
3520 - send/receive function.
3521 Convertion is not implemented, but short-cutted if not needed.
3522 struct/array elements are sent one by one (instead of block-wise), but
3523 nobody really cares (yet). Get a prototype before optimizing.
3524 - tests (using a file socket) for DD send/receive on:
3525 - base types: int, float
3526 - array: fixed size, string (ie ref to dynamic string)
3527 - structure: homogeneous, heterogeneous
3528 - chained list, graph with cycle
3529 Believe it or not, valgrind is not too unhappy with the results. The
3530 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3534 - Bugfix when using a filename explicitely (instead of '-')
3536 2004-04-09 Martin Quinson
3538 - factorize more code between RL and SG in socket creation
3539 - Complete the implementation and tests of:
3541 o file (only in RL, and mainly for debugging)
3543 I lost 3 days to design a portable address resolver, and then decided
3544 that the prototype mainly have to run on my box.
3545 Addressing portability too early may be like optimizing too early :-/
3547 - use gras_init in the Tests instead of the crappy parse_log_opt
3548 (the latter function is removed)
3549 [Conditional execution]
3550 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3551 [Code reorganisation]
3552 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3553 Build examples against the RL library, since there is no way to disable
3554 its creation for now.
3556 For information, the beginning of coding on GRAS was back in june
3557 2003. I guess that every line has been rewritten at least twice since