1 SimGrid (3.18) NOT RELEASED YET (target: December 24 2017)
3 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
5 MSG IS NOW DEPRECATED. Please use S4U for new projects.
6 - Support for MSG should not be removed from SimGrid before 2020,
7 but future evolutions will be limited to the integration of
11 - s4u::onDeadlock() signal, run before stopping the simulation.
12 - s4u::Exec: asynchronous executions (abstraction of background
14 - s4u::Host->getLoad() returns the achieved speed in flops/s
17 - Link::name() is deprecated and will be removed in v3.21, use
18 Link::getCname() instead.
19 - Mailbox::getName() changed to return a std::string, use
20 Mailbox::getCname() instead to get a char*.
21 - Storage::getName() changed to return a std::string, use
22 Storage::getCname() instead to get a char*.
23 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
24 to have both allocation and deallocation of the map in user space.
27 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
28 The amount of flops remaining to do is a vector, not a scalar.
29 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
30 name implies on both sequential and parallel tasks.
31 - Both changes fix GitHub's #223 using PR #237 as a basis.
32 Thanks Michael Mercier.
33 - Most examples were converted to S4U and hidden elsewhere at tests.
34 Rational: we still want MSG to work; we want newcomers to use S4U.
37 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
38 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
41 - Switch to the faster dlopen privatization mechanism by default
42 - Documentation improvments
45 - Live migration is getting moved to a plugin. Dirty page tracking is
46 the first part of this plugin. This imply that VM migration is now
47 only possible if one this function is called :
48 - C/MSG: MSG_vm_live_migration_plugin_init()
49 - C/C++: sg_vm_live_migration_plugin_init()
50 - Java: Msg.liveMigrationInit()
51 For C and C++, "simgrid/plugins/live_migration.h" also has to be
55 - Define class simgrid::xbt::Path to manage file names.
56 - Removed unused functions:
57 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
58 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
59 xbt_graph_export_graphviz()
60 - xbt/str.h: xbt_str_join()
61 - Remove unused datatypes:
62 - xbt/heap.h: use std::priority_queue or boost::heap instead
63 - xbt/swag.h: use boost::intrusive::list instead
66 - New link_energy plugin for the consumption of the links.
67 - All of the operations on files and storage contents have been
68 packaged into a plugin (src/plugins/file_system). The current
69 public interface can be found in
70 include/simgrid/plugins/file_system.h
71 To use these functions you now have to initialize the plugin by
72 calling MSG_storage_file_system_init() just after calling
73 MSG_init() or sg_storage_file_system_init() just after creating
77 - Remove the undocumented/untested tag <include>
80 - Remove viva specific tracing as the tool is no longer maintained
83 - #248: Kill of finished processes leads to segfault
84 - #240: xbt_cond_wait_timeout should gracefully return for C apps
85 - #239: please implement signal s4u::onDeadlock()
86 - #230: segfaults when exit() without run()
87 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
88 - #223: MSG_task_get_flops_amount() not working with parallel tasks
89 - #222: Actor::kill() doesn't really kill and segfaults
90 - #221: odd LMM warning when killing an actor
91 - #120: Memory leak when the processes are forcefully killed
93 SimGrid (3.17) Released October 8 2017
95 The Drained Leaks release: (almost) no known leaks despite the tests.
97 Even very long simulations will preserve your memory: our testsuite
98 stresses SimGrid for over 45mn on fast machines for a coverage of
99 over 80%, and there is only one single known leak, of about 4kb.
103 - Comm.detach(): start and forget about asynchronous emission. A cleanup
104 handler may be given to free resources if the comm cannot be completed.
105 - this_actor::send(mailbox) is now mailbox->put()
106 - New: simgrid::s4u::Comm::wait_all()
107 - New: Host.execute() for remote executions.
110 - Unused option network/sender-gap is removed.
113 - Deprecate MSG_task_isend_with_matching(): This unused feature
114 really complicates our internals. Will be removed in v3.20.
117 - Improved context termination. It is now possible to free resources, even
118 when a process is forcibly killed.
119 - ContextBoost: add support for Boost versions above 1.61.
122 - Bring back run-time option --cfg=exception/cutpath to remove exception
124 - Removed unused functions:
125 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
126 xbt_str_rtrim(), xbt_str_trim().
127 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
130 - Removed header files obsolete since SimGrid 3.12:
131 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
133 - #3: SD_exit should be made optional
134 - #120: Memory leak when the processes are forcefully killed
135 - #159: Threading test regression in Actor refcounting
136 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
137 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
138 - #186: Actor::killAll() segfaults if some process is blocked on wait()
139 - #191: VM migration and pstate
140 - #192: Updating the energy consumptions of all hosts crashes with VMs
141 - #195: All actors have PID=0 in the logs
142 - #204: Sometimes segfault with thread contexts and mmap privatization
143 - #222: Actor::kill() doesn't really kill and segfaults
144 - #225: Actor::kill() doesn't really kill when victims are doing a join()
146 ----------------------------------------------------------------------------
148 SimGrid (3.16) Released June 22. 2017.
150 The Blooming Spring Release: developments are budding.
153 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
154 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
155 - Java JAR file should be finally fully working on Mac OSX too.
157 XML platforms: Switch to platform v4.1 format.
158 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
159 - <zone> can be used as a synonym for the now deprecated <as>
160 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
161 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
162 - <actor> can be used as a synonym for the now deprecated <process>
163 - state_file and avail_file periodicity is now easier to express
164 (check the documentation) the old behavior should still work.
165 * Storage: not backward compatible, but it seems that nobody ever used it
166 - Remove attribute 'content_type' of <storage_type>: was never used
167 - Make attribute 'model' of <storage_type> optional: for future usage
168 - Remove Bconnection model property: was never (in)validated. Replaced by
169 the maximum of the read and write bandwidth as a resource constraint.
172 - New and Backwards Compatibility break:
173 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
174 When one wants to get the list of tasks whose states have changed during a
175 simulation round, s/he has to allocate and free a dynar and use it as
176 argument to this function. The former SD_simulate (double how_long)
180 - Allow multicore VMs, with the correct sharing (unless you overcommit)
181 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
184 - New function to update the consumption of all hosts at once.
185 - Fix the model for multi-core hosts, linear in the amount of busy
186 cores with an abnormality for pIdle (see paper).
187 BUG: Still not satisfactory for ptask on mono-cores.
190 - The netzone are now available from the MSG API.
191 The old names still work, but are now deprecated.
194 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
195 - New option: smpi/keep-temps to not cleanup temp files
196 - New option : smpi/shared-malloc-blocksize . Relevant only when global shared
197 mallocs mode is used, allows to change the size of the fake file used
198 (default 1MB), to potentially limit the number of mappings for large runs.
199 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
200 - Fortran ifort and flang compilers support
201 - New RMA calls supported (experimental) :
202 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
203 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
204 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
205 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
206 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
207 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
210 - New: this_actor::isMaestro() returns whether we are in kernel mode.
211 - Behind the scene, ActivityImpl are now automatically refcounted.
212 Removing this nasty bottleneck will greatly simplify our internals.
215 - Replay: New function xbt_replay_action_get():
216 Retrieve the function previously associated to an event type.
217 - DROPPED FUNCTION: xbt_str_varsubst()
218 - DROPPED MODULE: strbuff. We don't need it anymore.
219 - DROPPED MODULE: matrix. We don't need it anymore.
220 - DROPPED MODULE: lib. We don't need it anymore.
222 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
224 SimGrid (3.15) Released March 22. 2017
226 The Spring Release: continuous integration servers become green
227 We fixed even the transient bugs on all target architectures:
228 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
229 Windows, FreeBSD, NetBSD.
231 - Rename NetCards to NetPoints.
232 This was intended to help NS3 users, but that's not a netcard.
233 That's a point in the routing algorithm, let's avoid wrong simplifications.
236 - New: MSG_process_yield(). Stop and yield to other processes.
237 - New: MSG_process_daemon(). Daemon processes are automatically killed
238 when the last non-daemon process terminates
239 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
240 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
241 to make room for the upcoming network energy plugin.
242 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
245 - Ensure that an actor can kill itself with Process::exit()
246 - Kill the obscure NativeException. Nobody want to survive the issues
247 it denotes, so use JniException that is a RuntimeException (not to
248 be caught explicitly).
249 - Partial bug fix in initialization. SimGrid flags on command line were
250 consumed at C level but stayed in the original Java String[] args.
251 This could mess users' args[i] if SG flags were not put at the end of
253 The SimGrid flags are now removed from the Java arguments. However,
254 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
255 if args.length is greater than the number of YOUR OWN ARGUMENTS.
256 It might be if you have --log or --cfg flags in the command line.
257 - Fix numerous memleaks all around the place. In particular, around VMs.
261 - s4u::onPlatformCreated: right before the simulation starts
262 - s4u::onSimulationEnd: right after the main simulation loop
263 - s4u::onTimeAdvance: right after a clock change (time is discrete)
264 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
265 event from the availability_file changes the avail speed.
266 - Links are now usable from s4u
267 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
268 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
269 - New examples: The conversion of MSG examples to S4U has begun
270 - Actors: create, kill, migration, and suspend
271 - Applications: master-worker and token-ring
272 - Action replay: communications and storage
273 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
276 - Backwards Compatibility breaks
277 - The SD_route_get_list and SD_route_get_size functions have been removed.
278 They are replaced by the unique
279 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
280 The route from src to dst is built in the links dynar whose size can
281 be obtained with xbt_dynar_length.
282 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
283 sg_host_route_bandwidth and sg_host_route_latency.
284 Macros ensure the backwards compatibility, but you should fix your code
287 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
288 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
289 - Bump our claim of support from MPI 1.1 to MPI 2.2.
290 We don't support 100% of it, but it should be enough. Contact us if not.
291 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
292 naive and probably inaccurate)
293 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
294 mapped onto a unique small file using some system magic.
295 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
296 - Fortran: cleanups, fixes, support of user-added operations
297 - RMA: MPI_Accumulate are applied in correct order.
298 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
299 - Better support for MPI_IN_PLACE
300 - Support for MPI_Win attrs and keyvals.
301 - Support MPI_Comm_group_create, MPI_Type_size_x
302 - MPI ops should be properly applied to all allowed types
303 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
304 algorithm by a non failing one from mpich.
307 - Kill the fifo data container: we don't use it anymore.
310 - A <cluster> can now be created with different speed values to represent pstates.
311 see examples/platforms/energy_cluster.xml for an example.
313 SimGrid (3.14.159) Released December 28. 2016
315 The previous release was only a crude approximation of a Pi release;
316 this one should be a bit better.
318 - Revert a last minute change that broke on Mac OSX.
319 - Fix the changelog and NEWS files.
320 - Improve the documentation, in particular of the routing module.
321 - Integrate some patches flying around in Debian.
323 SimGrid (3.14) Released December 24. 2016
326 * The whole documentation was reworked and reorganized. There is
327 still much room for improvement here, but we're on it.
330 * We now need python3 for our testing infrastructure.
331 * Model checking needs libevent
334 * Now works on FreeBSD
337 * Privatization now works on FreeBSD
339 * Privatization is now activated by default in smpirun
340 You can turn it off with -no-privatize if something goes wrong.
342 * Call-location tracing for SMPI.
343 You can add the exact location (filename / linenumber) of an MPI call to
344 your trace files and slow down or speed up the simulation between two
345 consecutive calls by using an adjustment file (see the documentation).
347 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
348 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
349 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
350 values were ignored in some cases. The timings of these functions can now
351 be significantly different.
353 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
354 smpi/running-power is renamed to smpi/host-speed
356 * smpi/grow-injected-times option to enable or disable multiplication of the
357 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
358 make simulation less precise (but also much faster).
360 * smpirun script should be (much) faster for large deployments.
362 * SMPI tracing : fixed issue with poor matching of send/receives.
364 * Replay : Fix broken waitall
366 New functions and features
367 * MSG_parallel_task_execute_with_timeout, to timeout computations.
369 Dropped / renamed functions and features
370 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
371 - MSG_mailbox_is_empty() -> Mailbox::empty()
372 - MSG_mailbox_front() -> Mailbox::front()
373 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
374 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
375 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
376 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
377 * Don't pass the free_f parameter to property related functions:
378 - MSG_host_set_property_value()
379 - MSG_as_router_set_property_value()
380 - MSG_storage_set_property_value()
381 * VM properties. Since msg_vm_t are msg_host_t, just use
382 MSG_host_get_property_value() and friends
383 * VM I/O related things:
384 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
385 - Unimplemented save/restore methods
386 * MSG_as_router_get_property_value() was redundent with
387 MSG_environment_as_get_property_value().
388 - Removed MSG_as_router_*propert*() functions
389 - Added MSG_environment_as_set_property_value() for consistency
390 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
391 Well, they are still there for now, but deprecated with a warning.
392 Please switch to xbt_dict_new_homogeneous() before this is removed
394 * Task affinity. Its intended behavior (that was very badly tested
395 and probably not really working) was deceiving what most users
396 would have hoped here.
397 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
398 * Option network/coordinates is now useless and should be dropped.
401 * Added option storage/max_file_descriptors to allow more than 1024 files opened
404 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
405 * Added Dragonfly topology support
408 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
409 EiBps, EBps, Eibps, Ebps for bandwidth.
410 They may become useful to some lucky ones.
413 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
414 * ex/app_tokenring: new example, very similar to the MSG Token Ring
415 * ex/async_waitAll: new example, on asynchronous communications
418 * Memory usage should be decreased for simulations with a large number
419 of processes. This also helps for SMPI.
421 SimGrid (3.13) stable; urgency=low
423 The Half Release, a.k.a. the Zealous Easter Trim.
425 - We removed half of the lines, that were mostly experimental cruft.
426 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
427 (not counting blanks and comments -- according to openhub.net)
428 - The internals are now compiled in C++ (and will soon be clean C++)
429 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
430 59 klines of C, etc. We added only 29 klines of C++ in replacement.
432 * Backwards Compatibility breaks
433 - Removed Lua simulation bindings (switch to C or Java for that).
434 Lua can still be used to describe platforms
435 - Removed Java kernel plug-ins.
436 Will be reintroduced after the ongoing major internals reorg.
438 - the following functions were removed.
439 They were too specific and should be reimplemented in a generic
440 way, with filter function.
441 - MSG_task_listen_from_host
442 - MSG_mailbox_get_count_host_waiting_tasks
443 - MSG_mailbox_put_with_timeout was removed.
444 Please use MSG_task_send_with_timeout instead.
446 - the SD_application_reinit function was removed. It has been a noop for a while.
447 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
448 replaced by a more flexible mechanism.
449 - The following functions thus do not exist anymore
450 - SD_workstation_get_access_mode
451 - SD_workstation_set_access_mode
452 - SD_workstation_get_current_task
453 - Basic estimation functions have been removed but can easily be replaced
454 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
455 - SD_workstation_get_computation_time => amount / sg_host_speed()
457 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
458 Use VM.getSpeed()*load/100 for the legacy behavior.
460 - option enable_tracing was removed. It was not doing anything for a while.
461 - In the ModelChecker:
462 - the model-checker now ptraces the model-checked process which means
463 you cannot use a debugger on the latter anymore (we might make this
464 optional in the feature);
465 - removed soft-dirty page tracking;
466 - remove model-checked side snapshot management,
467 MC_snapshot() and MC_compare_snapshot();
468 - keep the MC_cut() function as a stub (it was not really working
469 in the previous release).
472 * All options are consistently in kebab-case. Old names are kept as alias.
475 * Switch to platform v4 format.
476 - Rename from 'power' to 'speed' the attributes describing the amount of
477 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
478 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
479 - In <host> and <link>, attributes availability and state are gone.
480 It was redundent with state and availability traces, and with peak values.
481 - In <cluster>, attributes availability_file and state_file are gone.
482 It was too complex and unused.
483 - Kill <gpu>. Was not doing anything.
484 - The DOCTYPE points to the right URL:
485 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
486 (the file at this address now documents the changelog since its v1)
487 - A warning is emitted for unit-less values (they are still accepted).
488 - speed. Default: 'f' or 'flops'. Also defined:
489 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
490 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
491 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
492 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
493 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
494 - latency. Default: 's' second. Also defined:
495 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
497 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
498 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
501 * s4u::Host is now the preferred public interface to the Host features.
502 sg_host_* functions are C bindings to the exact same behavior
503 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
506 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
507 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
508 They were deprecated since a few years, and probably did not even compile anymore.
511 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
512 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
513 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
514 encouraged to update their codes. A list of the modified functions can be found at the end of
515 include/simgrid/simdag.h
518 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
519 the lambda or closure passed as a parameter will run in kernel mode.
520 All the callback functions should be rewritten to that interface at some point.
523 * Reorganizing and cleaning the internals all around the place.
526 * Remove old default barrier/bcast buggy algorithms (see #18407)
527 * Various bug fixes to handle more codes
528 * Remove the need for the --foreground option of smpirun (it is still
529 accepted for backward compatibility).
532 * Kill the setset data container: MC don't use it anymore.
533 * Kill the queue data container: it made more sense with GRAS.
534 * Kill the xbt_peer_t data type: it's useless without GRAS.
535 * Kill rm_cb feature of config sets: it was never useful.
536 * Kill graphxml parsing feature. It was not used.
537 * Kill the deprecated code protected by XBT_USE_DEPRECATED
539 - xbt_dynar_sort_strings(), when the content is char*
540 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
541 They throw exceptions on invalid input;
542 * C++ support for declaring CLI flags (simgrid::config::Flag);
543 * class for abstracting different signal backends (simgrid::xbt::signal).
544 with no external dependencies (we need very simple signals).
547 * refactoring and cleanup of the code;
548 * ongoing process to cleanly separate the model-checking algorithms
549 from the code model-checking support.
551 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
553 SimGrid (3.12) stable; urgency=low
555 The Facelift Release.
558 * Require g++ v4.7 at least to not speak prehistorical C++.
559 * Require Boost 1.48 (for signal2 component).
560 * Java must be version 7 at least when activated.
561 * Builds on Windows again (including Java bindings).
562 * Tracing is now always enabled (no way to turn it out)
563 * Remove GTNetS. It was not working anyway.
564 * Various cleanups in the cmake scripts.
565 * Move headers around to sort them out on installed systems:
566 - instr/instr.h -> simgrid/instr.h
567 - instr/jedule/* -> simgrid/jedule
568 - simdag/datatypes.h was removed
569 - simdag/simdag.h -> simgrid/simdag.h
570 - msg/datatypes.h was removed
571 - msg/msg.h -> simgrid/msg.h
574 * Interface improvement:
575 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
576 - Sanitize the interface in MSG_task_ module:
577 - Merge two functions that were close enough but misleading:
578 set_compute_duration(t) -> set_flops_amount(t)
579 get_remaining_computation(t) -> get_flops_amount(t)
580 - set_data_size(t) -> set_bytes_amount(t)
581 get_data_size(t) -> get_bytes_amount(t)
582 - Massive cleanups in the functions related to the energy
583 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
584 - New: MSG_host_get_pstate()
586 - msg/energy/onoff: switching hosts on and off
588 * Interface improvement:
589 - Rename Host.isAvail() to Host.isOn()
590 - Rename Process.currentProcess() to Process.getCurrentProcess()
591 - Rename Task.setDataSize() to Task.setBytesAmount()
592 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
594 - #18874: Actually allows the GC to reclaim tasks
598 - SIMIX_process_throw: raises an exception in a remote process
599 * Refactoring: Separate sub-modules
600 - libsmx: the public interface, as libc in a real system
601 - popping: the strange dance that converts a user request into a kernel handling
602 - smx_context_*: the virtualization mechanisms that embed the user code
603 - smx_*: the handling of each simcalls
605 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
606 * Rename smx_action_t into smx_synchro_t, making explicit that these
607 things are used to synchronize processes with their environment.
608 For example, a communication is a sort of synchronization involving
609 the communicating processes (that may block until the exchange) and
610 the platform. The same can be said from computations, etc.
612 - #18888: segfault when a process ends before its kill_time
616 - 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.
617 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
618 - MPI_Type_set_name, MPI_Type_get_name
619 - MPI_*_c2f and MPI_*_f2c functions
620 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
621 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
622 - Activate a lot of new tests from the mpich 3 testsuite
624 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
625 - InfiniBand network model added : Based on the works of Jerome Vienne
626 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
627 - When smpi/display_timing is set, also display global simulation time and application times
628 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
629 * Collective communications
630 - 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
631 - 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/).
632 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
634 - Replay now uses algorithms from wanted collective selector
635 - Replay can be used with SMP-aware algorithms
636 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
637 - 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.
639 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
640 - Correctly use loopback on fat-tree clusters
641 - Asynchronous small messages shouldn't trigger deadlocks anymore
642 * Energy/DVFS cleanup and improvement
643 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
644 - new: smpi_get_host_pstate()
647 - "Full" network optimization flag was broken since Surf++
648 - Better handling of precision flags in maxmin
649 - Fix bug causing sometimes "Impossible" errors
650 - Properly pass cluster properties to included hosts
651 * Improvement of the Energy plugin.
652 - Always update the consumption before returning that value
653 - New property: watt_off to denote the dissipation when the host is off
654 - New functions getWattMinAt and getWattMaxAt to retrieve the
655 dissipation of pstates that we are not currently at.
656 * Java: class NetworkLink renamed to Link
657 * New function: simcall_process_get_kill_time()
658 * Massive rename s/workstation/host/
659 - That's intrusive, but that's good for the project consistency. Sorry.
660 - Change config option "workstation/model" into "host/model"
663 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
664 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
665 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
666 - xbt_dirname and xbt_basename for non-POSIX systems
668 * The model checker now runs as a separate process.
669 * The model checker runs must now be launched with the new simgrid-mc program.
670 * Record/Replay: the MC can display a textual representation of a path in the
671 execution graph. It can then be replayed outside of the model checker.
673 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
675 SimGrid (3.11) stable; urgency=low
680 * Normalizing pointers addresses tool for better diff between logs
683 * Add cloud examples using new VMs
684 - examples/msg/cloud/two_tasks_vm.tesh
685 - examples/msg/cloud/simple_vm.tesh
686 - examples/java/cloud/cloud.tesh
687 - examples/java/cloud/migration/migration.tesh
688 * Add java surf examples:
689 - examples/java/surfPlugin/surf_plugin.tesh
690 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
691 - examples/java/surfCpuModel/surf_cpu_model.tesh
692 * Add SMPI+MSG example:
693 - examples/smpi/smpi_msg_masterslave/
698 - msg task destroy cancel test
699 - msg_host on/off test
700 * Move all tests in testsuite to teshsuite (adding tesh files)
701 * Restructure teshsuites
702 - one folder for each kind of test
703 * Restructure AddTests.cmake
705 - structure the order of tests (with sections)
708 * Add virtual machine
709 - creation of a VM on a PM
710 - migration of a VM from a PM to another PM
712 - MSG_process_join(msg_process_t process, double timeout)
713 - msg_bar_t MSG_barrier_init(unsigned int count)
714 - int MSG_barrier_wait(msg_bar_t barrier)
715 - void MSG_barrier_destroy(msg_bar_t barrier)
716 - msg_as_t MSG_environment_as_get_by_name(const char * name)
717 * New option "msg/debug_multiple_use" to help debugging when a task is used
720 - Improvements and finalization of MSG_storage, MSG_file APIs and their
722 - Increase code coverage in test suites
726 * Protect context stacks against stack overflow. The number of protected memory
727 pages allocated on the top of each stack (1 by default) can be configured
728 with the new command line option --cfg=contexts/guard_size:<value>.
729 * Simcalls are now generated by a python script that
730 - generates files included by SimGrid
731 - checks that all the functions exist, and proposes prototypes
733 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
734 host_set_data, host_get_data
736 - simcall_process_join(smx_process_t process, double timeout)
737 * Fix bug where sleeping processing could not be suspended.
740 * Translate surf models from C to C++
741 - Generic classes for all models: Model, Resource, Action
742 - A generic interface for each kind of model (CPU, Network, Storage
743 Workstation, WorkstationVM)
745 * Translate surf routings from C to C++
746 * Add callbacks using sigc++ or boost::signals2
747 - Add callback functions for resource creation/destruction
748 - Add callback functions for action state change
749 - Handle Energy as a plugin
750 * Replace swag by boost::intrusive
751 * Add new routing models for clusters. For documentation, see
752 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
753 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
754 parameters for cluster tag
755 - Fat trees, with topology="FAT_TREE" and
756 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
757 - see examples/platforms/torus_cluster.xml and
758 examples/platforms/fat_tree_cluster.xml
762 * Hostfiles support host:nb_processes construct to deploy several processes on
764 * Collective communication algorithms should not crash if used with
765 improper number of nodes and report the error.
766 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
767 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
768 MPI_Dims_create, MPI_Cart_sub are supported.
769 * New interface to use SMPI programmatically (still depends on MSG for
770 some parts, see examples/smpi/smpi_msg_masterslave) :
771 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
775 * Global variables privatization in MPI executables is now performed at runtime
776 with the option smpi/privatize_global_variables (default:no).
777 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
778 dynamic libraries loaded by the application are not privatized (static
779 linking with these libraries is advised in this case)
782 * Options defined in XML work correctly now.
785 * New cmake option, enable_lib_in_jar, to control whether native libraries are
786 copied into simgrid.jar or not (ON by default). Use this option if you want
787 to reduce the size of the installed simgrid.jar, *and* the native libraries
788 are kept installed elsewhere.
789 * Surf binding with SWIG (code generated in maintainer mode only):
790 - plugin to handle callbacks
791 - CPU model only for the moment
794 * Supernovae build mode is definitively removed. It was used to improve
795 inlining and inter-module optimizations. It is nowadays superseded by
796 link-time optimizations commonly available in compilers.
797 * Update ns-3 find lib. Bindings for ns-3 should work again now.
798 * Add boost dependency for surf++
799 * Add new macro for tests
800 - ADD_TESH(name <tesh_args>)
801 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
805 - xbt_bar_t XBT_barrier_init(unsigned int count)
806 - int XBT_barrier_wait(xbt_bar_t barrier)
807 - void XBT_barrier_destroy(xbt_bar_t barrier)
808 * Make the xbt_os_time module public
810 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
812 SimGrid (3.10) stable; urgency=low
814 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
817 * Reintegrate Java to the main archive as desynchronizing these
818 package is not acceptable anymore (Java is now considered stable)
819 * Add explicit synchronization facilities through semaphores
820 * Bug fix: Task.setDataSize() only changed the C world, not the value
821 cached in the Java world
824 * Dramatically change the way files are handled. API and internals changed, but
825 this part of MSG was not considered as production grade either.
826 * Add explicit synchronization facilities through semaphores
827 * Add a new function MSG_host_get_process_list()
828 * Preliminary DVFS support (see examples/msg/energy/ for details)
831 * SMPI is now included directly in the libsimgrid as the windows
832 linker doesn't force us on splitting it anymore.
833 * Improvements of the SMPI replay tool:
834 - Most of the collective communications are now rooted in the same process as
835 in the original application.
836 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
837 used until now). Multiple data types can now be used in a trace.
838 - The replay tool now supports traces produce either by TAU or a modified
840 - Bug Fix: the compute part of the reduce action is now taken into account.
841 - Gatherv collective is now supported
842 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
843 of smpirun outputs time independent traces for the current run. One file
844 is created per process. If too many processes are simulated, this behavior
845 can be changed to one file for all processes by using the
846 tracing/smpi/format/ti_one_file flag
847 * smpirun generates the host file if needed (with given host count and platform)
848 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
849 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
850 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
851 --cfg=smpi/coll_selector:(mpich/ompi)
852 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
854 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
855 happening inside a collective SMPI call.
856 * Fix the behavior of complex data types handling.
857 * Make MPI_Wtime another synchronization point to take computations into
859 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
860 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
861 * Add all missing Fortran bindings, SMPI should work with Fortran 90
862 (no privatization of global variables yet)
863 * Preliminary DVFS support (see examples/smpi/energy/ for details)
866 * Verification of liveness properties is now available for SMPI applications
867 (in addition to MSG applications)
868 * Bugged examples using SMPI in examples/smpi/mc/
869 * Add --cfg=model-check/visited option. Allows the verification of infinite
870 programs. Detection of loops in the execution thanks to the system state
871 comparison and reduction of the state space to explore. Can be combined with
872 DPOR for safety properties.
875 * Allow to change SimGrid configuration (see --help) within the code
876 thanks to SD_config() as it can be done in MSG.
877 * Add a new function SD_task_set_amount() upon user request.
880 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
881 * Remove rule based routing (no more PCRE dependency)
882 * Add a limiter_link option to cluster tag, to specify a maximum reachable
883 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
884 * Add a loopback_bw and loopback_lat options to cluster tag.
885 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
886 routing. Now peers are encapsulated in an AS and have their own private
887 router but this is transparent.
890 * Our own implementation of getline is renamed xbt_getline, and gets
891 used even if the OS provide a getline(). This should reduce the
892 configuration complexity by using the same code on all platforms.
893 * New type: xbt_cfg_elm_boolean.
894 * Allow to use yes/no for boolean configuration options in the command line.
895 * Allow to disable SimGrid cleanups at exit from command line option.
896 There are situations where one may want a simulation to end with an exit.
897 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
898 annoying when scripting around the simulator. Adding a
899 --cfg=clean_atexit:no allows to circumvent this issue.
902 * Lots of memory leaks were corrected in this release.
903 There is no known memory leaks anymore, in all of our 600+ tests.
904 * New command line option --version, to get SimGrid version information.
905 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
907 * Supernovae builds are deprecated, and expected to be removed in the next
910 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
912 SimGrid (3.9) stable; urgency=low
914 The Grasgory release: GRAS is really dead now.
916 * Complete overhaul of the internal host structures scheme.
919 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
920 was considered as experimental and badly maintained since then.
921 * Keeping it was thus a trap to our potential users, that could take
922 it instead of MSG or SMPI by mistake despite is pity state.
923 * GRAS seems to have very few users (if any), and no one volunteered
924 to maintain it further. It also induces a lot of XBT code (for
925 portability sake), that must be maintained too.
926 * For all these reasons, we killed GRAS. If someone wants to revive it
927 in the future, don't cry, our git history still remembers of GRAS.
930 * Major overhaul. Merge our documentation again as time proved that
931 splitting it was really not helping our users.
932 * Further improve the developer documentation to help newcomers
933 hacking on SimGrid itself. The user documentation (and in
934 particular, the beginner documentation) is still in a sorry state.
937 * Now works on Windows too!
938 * Much more extensive test suite, from MPICH
941 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
942 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
943 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
944 * Bug fix: let task be scheduled when the last dependency to be solved is
945 a control dependency.
946 * Remove SD_load_environment_script function.
947 Use the C sg_platf function if you want to declare a platform
951 * New function: MSG_process_get_number()
952 * Old function documented: MSG_config()
953 * Remove MSG_load_platform_script function
954 Use the C sg_platf function if you want to declare a platform
958 * Change the default value of the TCP_gamma constant (maximal size of TCP
959 congestion window) to a more realistic 4MiB value. If you notice changes in
960 your simulation results, you can fall back to the previous 20k tiny window
961 by adding --cfg=network/TCP_gamma:20000 on command line.
962 * (Hopefully) fix a bug wrt periodic availability and state traces
963 * Bug fix: use default values at start when first event in availability/state
964 trace is not at time 0.
967 * remove the "new_" part of function name sg_platf_new_trace_connect
968 (resulting in sg_platf_trace_connect), since it does not create
972 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
973 use case where it's really mandatory, and maintaining it was a pain in
975 * New: xbt_fifo_search(), search an item with a user-provided
976 comparison function instead of dumb pointer comparison.
979 * Fix the lua deployment:
980 Use `simgrid.init_application()` before deployment instead of
981 `simgrid.msg_register_application()` after.
984 * Transfer the tracing files into the corresponding modules.
986 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
988 SimGrid (3.8.1) stable; urgency=low
990 The "we are told that some people want to also *install* the simgrid
993 * Add missing file "tesh.1" to the archive.
995 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
997 SimGrid (3.8) stable; urgency=low
999 The Psssshiiiit release: SimGrid jumps into the Cloud.
1002 * Add an experimental interface to manipulate VMs. They are mainly
1003 process groups with very few intrinsic semantic, but they should
1004 allow you to build the semantic you want easily.
1005 * New function: MSG_host_set_property_value()
1006 * New function: MSG_process_on_exit(). To clean memory in all cases.
1007 * Bug fixes that made the host (and link) failures unusable.
1008 * Add a way to auto-restart process when the host in which they are
1009 executing comes back (ON_FAILURE="RESTART" on deployment file,
1010 MSG_process_auto_restart_set).
1011 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1012 please stop using the old ones, they are DEPRECATED.
1014 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1015 Please use MSG_init() instead. (reducing the amount of entry
1016 points in the library helps us).
1017 * Make it impossible to link against the wrong version of the lib
1018 * Deprecate MSG_clean(). No need to call it anymore.
1019 * Function MSG_get_host_number() is not deprecated anymore.
1022 * Split the doc into a user guide and a reference guide.
1023 * Start a developper guide to help people hacking on SimGrid.
1026 * Enable tracing by default. This modules rocks you should use it.
1027 * Remove option custom_flags. Now use environment variables CFLAGS
1029 * Use default cmake things to detect lua instead of home grown ones.
1030 * New option "enable_mallocators" to disable mallocators, for debugging
1031 purpose ("on" by default).
1034 * Bug fixes around the resource failures: don't let the processes
1035 survive the host they are running onto.
1036 * Add an interface to auto-restart processes when the host in which they are
1037 executing comes back.
1038 * Ensures that SIMIX_clean is called automatically. It's not part of
1039 the public interface anymore (bindings should be updated).
1042 * Bug fix for when SD_Simulate is called with a positive value: be careful
1043 when comparing doubles. Sometimes they are different for non significant
1045 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1046 parallel task whose initial work is distributed among host according
1047 to the Amdahl's law. Such tasks are created with a parameter alpha
1048 that corresponds to the non-parallelizable part of the computation.
1049 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1050 between two sets of workstations assuming a 1D block distribution (each
1051 workstation owns a similar share of data) on both sides.
1053 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1054 Data redistribution will be automatically scheduled once parent and child
1055 are both scheduled. The filling of computation_amount and
1056 communication_amount structures is now done seamlessly thanks to the chosen
1058 * New function SD_workstation_dump to display various information
1059 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1060 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1061 bandwidth on the route onto which the task is scheduled and the amount of
1063 To divide the nominal bandwidth by 2, the rate then has to be :
1064 rate = bandwidth/(2*amount)
1065 * Compute tasks that have failed can now be rescheduled and executed again
1066 (from their beginning)
1067 * Increasing source code coverage (src/simdag is now covered at 95.8%
1071 * Re-implement time-independent trace replay using SMPI (at the
1072 smpi_smp_* level) instead of MSG. This should replace
1073 examples/msg/actions/actions.c
1074 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1075 hindexed and structs)
1076 * Implement the exchange of non-contiguous data.
1077 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1078 * Correct behavior of smpi/sender_gap and set its default value to 0
1079 * Add option to asynchronously send small messages to allow better
1080 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1081 specifies the size in bytes under which messages will be asynchronously sent.
1082 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1083 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1086 * New C interface to define a platform: XML is now optional.
1087 For more info, please check include/simgrid/platf.h.
1088 * New interface to define random platforms from the C:
1089 For more info, please check include/simgrid/platf_generator.h and
1090 examples/msg/masterslave/masterslave_platfgen.c
1091 * Export a sg_cmdline dynar containing all the arguments we got from
1095 * Two new tracing options for adding comments to trace file so you
1096 can track your experiments (see --help-tracing for details).
1097 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1098 * Adding the SimGrid version that generated the trace file as a comment.
1099 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1100 * Fix to avoid key clashes on Paje links
1101 * Other minor fixes related to the Paje specification
1104 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1105 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1106 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1107 * Greatly improve the robustness of mmalloc to user errors (such as
1108 using an area after freeing it, or freeing it twice)
1110 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1112 SimGrid-java (3.8.1) stable; urgency=low
1114 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1115 Warning: all methods are visible, but only some of them are
1116 implemented so far. Check the source (src/jtrace.c)
1117 for further information.
1118 * New module: org.simgrid.msg.File (SimGrid File management functions)
1119 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1120 * Change the meaning of Process.restart: now restart the process from
1121 the begining, like MSG_process_restart in C.
1122 * Add Process.setAutoRestart: handling of process restart when failed
1124 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1125 you to retrieve the properties of the processes/hosts
1126 * Deprecate Msg.clean(): you can just forget about it now.
1127 * New function Process.getCount(), that only works when compiling
1128 with the not yet released version 3.9 of the C library.
1130 * New context factory based on Coroutines. It mandates a modified JVM
1131 but then, the simulations run about five times faster, and there is
1132 no limit to the amount of processes (beside of the available memory).
1134 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1136 SimGrid (3.7.1) stable; urgency=low
1139 * Restore the prototype of MSG_process_create_with_environment() to
1140 the pre-3.7 situation by removing the kill_time argument.
1141 * Add a MSG_process_set_kill_time() function instead.
1144 * Fix weird behaviors when dealing with parallel tasks.
1147 * Simgrid is now built as a dll.
1148 * Simgrid-java now works on Windows.
1149 * Simgrid-Java is now included into Windows package.
1152 * First pre-build package for MacOSX.
1155 * Fix compilation when using MSG_USE_DEPRECATED.
1156 * Fix some compilation issues on Macs and Windows.
1157 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1158 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1160 We discovered that the Lua console is broken, but we are missing the
1161 manpower to fix it right now. The problem existed in 3.7 too, so we
1162 are not blocking the release for that. Sorry if you depended on this
1163 feature, any help would be really welcome.
1165 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1168 SimGrid-java (3.7.1) stable; urgency=low
1170 The "Java aint got to be bloated and slow" release
1173 * Various internal cleanups and performance improvement
1174 Simulations are expected to run up to twice faster or so
1175 * Make Process.kill(process) an instance method, not a static one
1176 * User processes are not java.lang.Thread subclasses.
1177 This breaks the compatibility (sorry), but previous API was
1178 brain-dead, making it impossible to have non-trivial
1179 initializations in the process constructor.
1180 * Require a full constructor per Process sub-class.
1181 Kinda breaks the compatibility (sorry), but this allows a much more
1182 efficient way to launch the processes at simulation startup.
1183 * Do not embeed our version of semaphores, java 1.5 can be considered
1184 as sufficiently prevalent for us to not dupplicate its features.
1188 * Add examples for almost every part of the API
1189 We spotted and fixed a lot of bugs in the process
1191 * New module: asynchronous communication API
1192 * New function: Process.sleep()
1193 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1194 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1195 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1197 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1199 SimGrid (3.7) stable; urgency=low
1201 The "spring cleanups (before the next Big Project kicks in)" release.
1204 * We can specify the SMPI latency/bandwidth factor with command line
1205 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1206 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1207 You can also use the "config tag" from platform file by adding this line
1208 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1209 (see "example/platforms/tag_config.xml" to use "config tag").
1210 Note that the command line supersedes the platform file configuration.
1211 * Change the correction factors used in LMM model, according to
1212 the latest experiments described in INRIA RR-7821.
1213 Accuracy should be improved this way.
1214 * Use the partial invalidation optimization by default for the
1215 network too. Should produce the exact same results, only faster.
1216 * Major cleanup in surf to merge models and split some optimization
1217 mechanisms from the core of the models. As a result you can now
1218 specify which model to use (e.g., --cfg=network/model:LV08
1219 --cfg=cpu/model:Cas01) and which optimization mode to use
1220 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1221 Incompatible combinations should err at initialization. See
1222 --help-models for the list of all models and optimization modes.
1223 * The CLM03 workstation model was dropped for simplicity because it
1224 used the deprecated CM02 network model. Use default instead.
1225 * Rename the TCP_gamma configuration option to network/TCP_gamma
1226 * Rename the coordinates configuration option to
1227 network/coordinates, and document it
1228 * Use now crosstraffic keyword instead of the terribly misleading
1229 fullduplex keyword. It is activated by default now in the current
1230 default model, use --cfg=network/crosstraffic:0 to turn it off.
1231 * Ongoing refactoring the model parsing to make XML files optional
1232 See include/simgrid/platf.h for details (still to be completed)
1235 * Major overhaul of the documentation. Almost instructive now :/
1236 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1237 functions and friends. This interface was considered as
1238 deprecated since over 2 years, it's time to inform our users that it is.
1239 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1240 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1241 use these (crappy) functions in your code.
1242 These functions will be removed soon. Stop using them now.
1243 * Deprecate MSG_get_host_{table,number}
1244 Implement MSG_hosts_as_dynar() instead.
1245 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1246 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1249 * Stabilize the parallel execution mode of user contexts
1250 * Introduce configuration variables to control parallel execution:
1251 - contexts/synchro: Synchronization mode to use when running
1252 contexts in parallel (either futex, posix or busy_wait)
1253 - contexts/parallel_threshold: Minimal number of user contexts
1254 that must be part of a scheduling round to switch to parallel
1255 execution mode (raw contexts only)
1256 * Fix bugs that prevented to use suspend/resume along with
1257 synchronization structures.
1258 * Fix bugs in process termination that lead to invalid memory access
1259 in very specific conditions.
1262 * Introduce a parallel mode for the models (controlled by surf/nthreads
1263 configuration item). In our tests, running the models in parallel
1264 never lead to any speedups because they are so fast that the gain
1265 of computing each model in parallel does not amortizes the
1266 synchronization costs, even when ultra fast futexes are used.
1267 This is released anyway because YMMV.
1270 * Performance boost by using a swag internally to compute the set of
1271 tasks that are finished and should constitute the return value of
1275 * Enable it by default now that it is considered rather stable.
1278 * Documentation of the tracing functions.
1279 * Performance gains when tracing categorized/uncategorized resource
1280 utilization by avoiding calls to get route when updating resource
1281 variables. LMM constraints are being used instead.
1282 * API changed to set task categories. Use MSG_task_set_category instead
1283 of TRACE_msg_set_task_category, and SD_task_set_category instead
1284 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1285 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1286 * Examples for link user variables added (see at examples/msg/tracing/)
1287 * Deprecated function TRACE_msg_set_process_category completely removed
1288 * Trace header updated according to the latest Paje file format
1289 * Tracing network lazy updates, no longer obligate users to use full updates
1290 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1291 * Experimental: let user code declare/set/push/pop application states for hosts
1292 * API changed to allow the manual creation of graph configuration files
1293 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1296 * Improve the API of Lua MSG bindings, using the Lua spirit.
1297 * Each simulated process now lives in its own Lua world (globals are
1298 automatically duplicated). It helps writing simulators. Will allow
1299 to run Splay programs within SimGrid in the future.
1300 * Add a Chord example in Lua, equivalent to the MSG one.
1303 * Start the implementation of a solution to express temporal
1304 properties, not only local assertions. This is still an
1305 experimental work in progress, stay clear from it to be safe.
1309 - Add new runtime parameters --help-logs and --help-log-categories
1310 to display information about supported logging parameters and
1312 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1313 don't exists anymore.
1314 * Mallocators: allow value NULL for the reset function.
1316 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1317 create homogeneous dictionaries, where all the elements share the
1318 same free function. Non homogeneous dictionaries will be
1319 deprecated in the next release.
1320 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1321 - Multi-level dictionaries are deprecated.
1323 - new function xbt_dynar_search_or_negative() that is useful when
1324 you have less than 2 million elements in your dynar and don't
1325 want of the extra complexity of catching exceptions when the
1326 element is not found.
1328 - Make xbt_os_thread module (for thread portability) public.
1329 Documentation is still to come, sorry.
1331 - Cleanups and simplifications to make it maintainable again.
1332 - Exotic features (such as memalign and valloc) were removed.
1333 - The metadata were extended and improved so that the
1334 model-checker becomes able to explore and inspect the heaps.
1335 - This may induce a performance drop when enable_model-checking is
1336 ON in cmake (even if it's not used in the simulation), but it is
1337 necessary at this point to get MC working.
1339 Turn model-checking OFF if simulation performance matters to you.
1340 Not enabling it at runtime is not enough, disable it in cmake.
1342 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1345 SimGrid (3.6.2) stable; urgency=low
1347 The "Not coding new stuff allows to polish old things" release.
1350 * New bindings to the NS3 packet level simulator (experimental)
1351 * Use the raw (efficient) execution contexts instead of the sysv
1352 (portable) ones when possible.
1353 * libpcre is now mandatory in any cases since not using it led to
1354 severe performance loss and possibly other issues
1355 * Update the XML platforms:
1356 - G5K: include the latest machine in Nancy
1357 - GridPP and LCG: new platforms
1358 * Documentation was partially updated, at least (more to come)
1360 Bug fixes, cosmetics and small improvements
1361 * Free terminated processes before the end of the simulation to avoid
1362 exhausting the memory of users having very dynamic amount of
1364 * Bug fix and cosmetics about canceling non-running tasks
1365 * Bug fix about the dot loader's issues when using libcgraph
1368 * Create an installer for windows with nsis (amd64 and win32)
1369 - Add an hello world project to illustrate simgrid project creation.
1370 - Embed libpcre into the Simgrid installer to avoid
1371 its compilation burden
1372 * The raw execution contexts should work on Apple now
1373 * Port to Windows 64 bits
1374 - Sysv contexts now have an implementation for this arch
1375 - GRAS communication features now support this arch
1376 * Drop support for borland compiler on windows
1377 - this code was not maintained, and we kinda depend on gcc nowadays
1378 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1379 * Fix portability issue on unstable ubuntu: linker became picky on
1382 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1384 SimGrid (3.6.1) stable; urgency=low
1386 The "Oops, we broke Macs too" release
1389 * Fixed contexts detection so that raw ones are used when possible
1390 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1391 leads to a strange error, with user code segfaulting sometimes when
1392 the generated code is not perfectly aligned (which is not
1393 controllable from the user side, depends on the amount of code)
1396 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1398 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1400 SimGrid-java (3.6) unstable; urgency=low
1403 * Split of every thing from simgrid v3.5 into a separate package.
1405 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1407 SimGrid (3.6) stable; urgency=medium
1409 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1412 * Bindings now constitute their own package, separated from the main one.
1413 Rationale: reduce our maintenance nightmare by reducing the module coupling
1414 They will soon be released on their own on gforge.
1416 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1417 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1419 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1420 * It's not quite deprecated for now because we have no replacement,
1421 but it may soon become the case.
1424 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1425 * Fortran: New user-level cache variable to store the rank of the running
1426 process. This improves performance by an order of magnitude.
1427 * C: New coccinelle script to automatically locate and modify global and
1428 local static variables.
1429 * Improved SMPI network model with a sender-side gap to account for multiple
1433 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1434 only say if a communication is finished, no matter whether it succeeded or
1435 failed. You can call MSG_comm_get_status() to know the status of a finished
1437 * New function MSG_task_dsend() to send a task and detach it. When a
1438 communication is detached, you are never notified of its success or failure
1439 and the memory is released automatically once it is finished. This function
1440 is useful when you don't care about the end nor the success of a
1442 * Change the prototypes of action replay. Sorry for inconvenience,
1443 but this is really more efficient this way (and to adapt your code,
1444 you just have to fix the initialization, that shouldn't be too long)
1445 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1446 ever managed to do anything useful with it.
1447 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1448 takes a dynar of communications. It returns immediately and gives the
1449 index of a finished communication (if any).
1450 * New example: a basic implementation of the Chord P2P algorithm.
1453 * New model for multi-core CPUs. You can now use the core attribute to
1454 precise the number of cores of a host. This is a basic model. Every
1455 process running on the host receives at most the power provided in
1456 the DTD (throughput<=power). Total throughput of process cannot exceed
1458 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1459 router linked by an up-link and a down-link (possibly asymmetrical).
1460 This kind of pattern allows to easily build last-mile model style platforms.
1461 Aggregating such patterns in a rule-based AS is thus the technique of
1462 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1463 * New model for Vivaldi routing. We transformed the Vivaldi network model
1464 into a Vivaldi routing model (based on the rule-based model). This allows to
1465 combine Vivaldi based latencies with last-mile platforms.
1468 * Added a check for NaN of IEEE754 infinite in the double entries of
1470 * Introduce a new context factory "raw", highly inspirated from the
1471 ucontext factory, but using manually crafted functions in assembly to
1472 do the work in an efficient manner.
1473 * Allow to change the used context factory at run time, not only at
1474 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1475 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1476 contexts at runtime (only with raw contexts or ucontexts).
1477 * Completely rewrote this module to allow parallel execution of user
1478 processes. Use --cfg=contexts/nthreads:N to execute user processes
1479 with N parallel threads (the default is 1, meaning no parallelism).
1480 * Allow to decide dynamically between sequential and parallel modes.
1481 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1482 processes in parallel only when their number is greater than or equal to P
1484 * Added a check for NaN of IEEE754 infinite in the double entries of
1488 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1489 won't output the state of processes when interrupted with Ctrl-C
1490 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1491 NULL-terminated array. This may solve backward compatibility issues
1492 due to the change to return type of SD_simulate. See also:
1493 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1494 * Add new macros with variable number of arguments.
1495 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1496 - in xbt/asserts.h: xbt_assert
1497 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1498 - in xbt/ex.h: THROWF and RETHROWF.
1499 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1501 * Change xbt_die() to accept a format string with arguments, just like printf.
1502 * New data structure: xbt_lib_t, like a dict but more general and with better
1506 * New configuration options
1507 Options triva/categorized and triva/uncategorized can be used to generate
1508 graph configuration files for Triva visualization tool.
1509 * Configuration option tracing/platform is renamed to tracing/categorized
1510 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1511 * New TRACE_user_link_variable interface:
1512 User provides the name of the link and the tracing variable to attach to it
1513 * the declaration of tracing categories must be done after the environment
1515 * simpler tracing interface, just one way to declare categories
1516 TRACE_category or TRACE_category_with_color, it is up to you
1517 * links in the trace file are again identified by their names
1518 * trace contains the full platform hierarchy exactly as declared using the ASes
1519 * Options tracing/msg/[task|process]:1 groups the process by hosts
1520 for both cases, tasks and processes must have names that are unique during the simulation
1521 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1522 * The experimental option tracing/msg/volume is deprecated
1523 its functionality may be reincorporated if needed
1525 The tracing generates a trace file with unordered timestamped events,
1526 because of the way the core simulator (surf) works. A script available
1527 at the tools directory (fix-paje-trace.sh) can be used to put the events
1528 in order. We have changed the tracing so it can generate ordered timestamped
1529 events in the final trace, but depending on the simulator (and how much time
1530 is simulated) that can lead to a huge memory utilization. It is deactivated
1531 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1533 Build Infrastructure
1534 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1535 - We are trying hard to keep the API stable, but it may happen that
1536 some things change (we're a research project after all, not a
1537 nuclear plant operating system). If such things should happen, you
1538 could rely on that macro to adapt.
1539 - current value: 30600 for 3.06.00, aka 3.6
1540 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1541 a number that can be compared with SIMGRID_VERSION.
1542 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1543 assertions and verbose/debug logging events are disabled at compile time.
1545 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1547 SimGrid (3.5) stable; urgency=medium
1550 * New feature to any SimGrid-based simulator: Model-Checking
1551 Check SIN#1 for more details.
1554 * New Model SMPI (three-interval linear regression for correction factors)
1555 See RR-7426, available at http://hal.inria.fr/inria-00527150
1556 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1557 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1558 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1559 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1560 * See also src/smpi/README
1565 - Tracing API changes: TRACE_start and TRACE_end should not be called
1566 by user-code. They are automatically called by simulators created
1567 with SimDAG, MSG and SMPI if the toolkit is compiled with
1568 tracing_enabled=ON. Categories declaration and utilization remain the
1569 same for MSG and SimDag.
1570 - A function was added to the tracing API to declare categories with
1572 - TRACE_category_with_color (char *category, char *color)
1573 where color must be in the following format
1574 "%f %f %f", red, green, blue
1575 and red, green, blue are float values in the interval [0, 1]
1576 - User can specify NULL as color parameter, or continue calling
1577 TRACE_category (cat)
1578 On that case, the tracing system will define random colors
1579 - The following command-line options are supported:
1580 --cfg=tracing/filename:msg.trace
1581 --cfg=tracing:1 (activate tracing, needed to use others)
1582 --cfg=tracing/platform:1 (categorized resource use)
1583 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1584 --cfg=tracing/msg/task:1 (task creation)
1585 --cfg=tracing/msg/process:1 (process creation, migration)
1586 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1587 --cfg=tracing/smpi:1 (SMPI interface tracing)
1588 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1589 - examples of examples/msg/tracing updated
1591 - DAXLoader and DOTLoader functions can generate tasks with categories
1592 - A new function to attribute a category to SD tasks:
1593 TRACE_sd_set_task_category (SD_task_t task, char *category)
1594 * Tracing the MPI interface implemented by SMPI
1595 - Collective operations are traced with states
1596 - Point-to-Point operations are traced with states/links
1597 - Tracing activated by a parameter "-trace filename" passed
1598 to smpirun during execution (considering that simgrid
1599 is compiled with tracing enabled)
1600 - To run the simulation with gdb, the simulator
1601 accepts --cfg=tracing/smpi:1 to trace SMPI
1602 - tesh files to check if smpi tracing is ok
1603 - See examples/smpi/NAS/DT-trace
1604 * GTNetS tracing re-worked
1605 - adaptation to the tracing system of GTNets to cope
1606 with modifications regarding the fullduplex mode
1607 - new tesh files to check if gtnets tracing is ok
1610 * Asynchronous communications through the functions:
1611 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1612 * New function: MSG_load_platform_script()
1613 to make possible using a lua script instead of XML files to set up platforms
1614 * New function: MSG_set_function
1615 to associate functions to processes, used when bypassing the parser
1616 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1618 Platforms: Add some more examples in examples/platforms
1619 * Grid'5000: see www.grid5000.fr
1620 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1623 * Change the XML format. This is a very important modification. SimGrid
1624 3.5 introduces a new hierarchical format based on the notion of
1625 Autonomous Systems. Compatibility with old format is ensured
1626 through the perl script provided in the install bin directory
1627 bin/simgrid_update_xml.
1628 It is now possible to build platforms with specific routing
1629 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1630 connect several platforms together. We will try to provide soon
1631 set of realistic platforms exploiting these properties (have a
1632 look at examples/platforms/ for the moment).
1633 * Take the opportunity of the XML format change to be a good XML citizen:
1634 rename link:ctn to link_ctn and similar changes (also dealt with by
1636 * Add a new routing scheme (rule-based) using regular expressions. It
1637 enables to have an extremely low memory footprint when the
1638 underlying routing is simple and can be compactly described. You
1639 need to have libpcre4-dev (perl regular expressions) installed if
1640 you want to use this routing scheme.
1641 * Revive the cluster TAG and allow to easily and efficiently (both in
1642 term of memory and speed) connect clusters together. Have a look
1643 at teshsuite/simdag/platforms/ to see how this can be done. With
1644 this tag, you can create clusters with thousands of tasks at no
1645 cost (have a look at examples/platforms/).
1646 Note: clusters are implemented as ASes, so there is no need for an
1647 enclosing AS tag if you have only one cluster in your platform.
1648 * Add new generic functions in the public interface that allows the user
1649 to call SURF 'create_resource' methods from your code (same
1650 functionality as the XML bypass mechanism but with a much lighter
1652 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1653 that uses a piecewise linear approximation to produce better
1654 results when exchanging small messages.
1655 * Add a new parameter to handle correctly full duplex link and account
1656 for interferences between uplink and downlink communications
1657 (activate with --cfg=fullduplex:1).
1660 * Rename the SD_READY (all dependencies are satisfied and task is
1661 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1662 dependencies are satisfied) state.
1663 This prevents a confusion between the notion of "ready to schedule"
1664 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1665 simulated" (SD_RUNNABLE) used by the simulation kernel.
1666 * Change the way a task is considered as ready. Instead of removing
1667 dependencies when a task is done, a counter is decreased. This way,
1668 it is always possible to reach ancestors thanks to the
1669 SD_taks_get_parents function (even after the end of the simulation.)
1670 * Change the return type of SD_Simulate from (SD_task_t*) into
1671 xbt_dynar_t. This function was in handling a dynar internally and
1672 converted it into a NULL terminated array for historical reasons.
1673 * New function SD_dotload(char*) to load a DAG described in dot
1674 format. This loader and the corresponding examples require the
1675 installation of the graphviz library.
1676 * Fix a bug in the management of tasks of size 0 in the surf network
1677 models. This problem was only visible with SIMDAG and you should
1678 thus disregard results produced with earlier versions if you
1679 relied on this feature (some tasks were blocked because of this).
1680 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1681 even though your applications were doing only point-to-point
1682 communications and sequential computations. Now you can really use any
1683 model you want (of course, if you create real parallel tasks, which are
1684 not implemented in most models beside ptaskL07, this will abort).
1685 * Add an example that schedules a DAX on an heterogeneous platform
1686 using a Min-Min strategy.
1687 * New function SD_workstation_get_current_task() that returns the kind
1688 of task currently running on a workstation in the sequential access
1690 * Raise some warnings when unexecuted tasks remains at the end of the
1691 simulation. This is usually caused by cycles in the DAG.
1694 * New function: SIMIX_process_set_function() called by MSG_set_function
1695 * Change the underlying waiting queue in semaphores so that a process
1696 can wait on several of them simultaneously (as in waitany).
1697 * Fix the way to handle tokens in semaphores so that all access patterns
1698 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1699 * kill the dirty pimple SIMIX_message_sizes_output()
1700 Please use (proper) visualization instead
1703 * New data container: setset (set of sets of elements)
1704 * New module: mmalloc (mapped malloc, allowing to have several
1705 independent segments of malloc)
1706 * New function: xbt_dict_cursor_set_data()
1707 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1708 * New function: xbt_dynar_is_empty()
1709 * New function: xbt_fifo_get_last_item()
1710 * Fix xbt_dynar_shrink(): use the right element size.
1711 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1712 uninitialized areas during expand.
1713 * Fix semaphores: previous implementation was severely broken.
1714 * Use library init/fini functions for our initialization.
1715 - you can use logs and other feature as soon as you want in your
1716 code (even before the xbt_init / MSG_init)
1717 - xbt_exit is now a no-op and produce a warning when used.
1720 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1721 benefit from the latest improvement to the simulation kernel.
1722 * Kill measurement sockets for now. If you rely on them, sorry. This
1723 release is not for you. This feature will be reintroduced in the
1724 future, but we cannot delay the release any further.
1725 * New function: gras_msgtype_get_name().
1726 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1727 * Fix (at last) the pmm example: it should not randomly fail anymore.
1729 Build chain: bug fixes and overall polishing
1730 * Cmake is now stable enough. Hence, we killed the autotools.
1731 * Port to windows ( TM :)
1732 * Fix the 'make install' target.
1733 No need to use 'make install-simgrid' anymore
1734 * Introduce a 'make dist' target compiling a *source* archive
1735 'make package' compiles a binary archive
1736 * Compile java files only on need
1737 * Add --cd and --setenv command line options to tesh
1738 * Out of source builds are not fully supported yet, but we are close
1739 * Enable supernovae and optimization flags by default for our users
1742 * Add layer to set up environment directly from lua, without XML.
1743 * The effect of gras_stub_generator can be achieved through
1744 lua too (check examples/gras/console/ping_generator.lua)
1746 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1748 SimGrid (3.4.1) stable; urgency=low
1750 The "Polishing easter eggs is probably a good idea" release.
1751 This is a bug fixes release only.
1755 * Fix a bug preventing the tasks from begin garbage collected.
1758 * Fix a bug occuring when a host involved in a communication fails.
1759 This was not detected properly by the other peer involved in the
1760 communication. Now, it's reported as a network error.
1763 * Warn the user about loop dependencies in data flow of DAX files
1764 * Obey the control-flow dependencies of DAX files
1767 * Add option "enable_smpi" allowing to not compile SMPI.
1768 Probably useful for the (Mac) users experiencing a build error here
1769 * Improve the detection of lua5.1 and ruby1.8
1771 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1773 SimGrid (3.4) stable; urgency=low
1775 The "Easter in Cargese" release. Also known as (major changes):
1777 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1778 ~> bindings were greatly improved
1779 ~> new tracing infrastructure for better visualization introduced
1781 * the "Welcome to configury modernity" release.
1782 ~> we switched from autotools to cmake, and improved our cdash
1785 A more detailled list of changes follow (full detail in svn log).
1787 Java Bindings: Various Cleanups
1788 * (install java-gcj-compat-dev on debian-like to use them)
1789 * Remove put/get: no need to export deprecated interface in Java
1790 Use send/receive instead.
1791 * Cleanup the examples and add a README per directory
1792 * Remove example autoDestination (that's the only way to go now)
1793 * Remove example explicitDestination (was a plain copy of basic)
1794 * Make JniException a runtime exception, so that there is no need to
1795 declare the fact that you may encounter such a beast. I guess that
1796 nobody will ever want to survive such error.
1797 * Create specific errors for each MSG case of failure:
1798 host failure, transfer failure, timeout, task cancelled
1799 * Cleanup the exceptions that may get thrown by each function
1800 * Other internal cleanups in Java bindings. Performance still bad :/
1801 Ruby and Lua Bindings: create them
1802 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1803 * That's new and great, you should try them out.
1804 Same functionalities than Java bindings, only even less polished
1806 * Kill the useless "rate" argument of SD_task_get_execution_time()
1807 Everyone used to provide -1 as a value, it was not used, and the
1808 semantic of a possible use wasn't even clear.
1809 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1810 either as comm costs or compute costs to mean that there is no
1811 such thing for that specific task.
1812 * Add a SD_task_set_name() function
1813 * Fix SD_task_unschedule() on typed tasks
1814 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1815 * In DAX loader, accept useless 'level' attributes to <job> since
1816 LIGO DAGs have them (seem to be to ease graphical representation).
1818 * Add an example masterslave_mailbox.c using send/receive and not
1819 the deprecated put/get interface.
1820 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1821 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1823 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1824 (declare MSG_USE_DEPRECATED to still have the old name)
1825 * Add a MSG_task_set_data() function
1826 * About trace replay (see examples/msg/actions):
1828 - Allow to work with splitted trace files for each process
1829 Give the specific trace file as argument of each process,
1830 and call MSG_action_trace_run(NULL)
1831 You can still have one merged file for all processes.
1832 - Fix implementation of collective operations
1833 * Allow task_execute() on 0-sized tasks (closes #10063)
1835 * This is the first release of SimGrid where SMPI is not considered
1836 beta anymore (even if some corners should still be improved)
1837 * Port over the new SIMIX_network submodule (internal refactoring)
1838 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1839 * Implement more missing elements of the standard:
1841 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1843 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1844 MPI_Scan() MPI_Get_processor_name()
1845 - Added implementation of missing case for Alltoall (warning: it's
1846 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1847 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1848 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1849 SMPI_MPI_Allgatherv()
1850 * Bug fixes include:
1851 - MPI_Waitsome() was broken
1852 - Allow relative includes in smpicc
1853 - Command line cfg argument 'reference_speed' was ignored...
1854 - Some functions did not properly lead to auto-benching of user code
1855 - smpicc passes -O2 by default (just like openmpi one)
1857 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1858 * Bug fixes about timeouts during communications
1859 * add SIMIX_message_sizes_output() as a pimple to write to file the
1860 amount of messages per size. Use gnuplot to get histogram.
1861 Pimple because that's the only user-visible function of simix,
1862 defined directly in xbt.h (irk, sorry)
1864 - Add a SIMIX_sem_get_capacity() function
1865 - Fix interactions with processe resume/suspende
1866 - release_forever() was stupidly broken
1867 - Fix SIMIX_display_process_status() for processes in a semaphore
1868 - Make SIMIX_sem_block_onto() user-visible
1869 * Refactoring context stuff:
1870 - Use pseudo-OOP for better modularity
1871 - reimplement SIMIX_process_kill() without process_schedule() so
1872 that the latter can take as invariant that it is called from
1874 - Merge context_start into context_new for sake of simplicity
1876 * Add a Vivaldi network model, coded live during SUD'10 ;)
1877 * Rename configuration variables to start a hierarchy:
1878 o cpu_model -> cpu/model
1879 o network_model -> network/model
1880 o workstation_model -> workstation/model
1881 * New configuration variables:
1882 o network/bandwidth_factor: correction to bandwith
1883 o network/latency_factor: correction to latency
1884 o netwotk/weight_S: correction to the weight of competing streams
1885 * Add a long description to the models, that users can see with such
1886 argument on the command line: --cfg=cpu/model:help
1887 * --help-models display the long description of all known models
1889 * config: add the ability to set a default value after registration
1890 Does not override any previously set value (e.g. from cmd line)
1891 * dict: allow to have integer key and data.
1892 When so, you need to use the following functions
1893 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1894 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1895 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1896 In contrary to regular dicts, the key is not malloced before copy.
1897 Mixing scalar and regular elements in the same dict is not tested
1899 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1900 Tracing for Visualization:
1901 * SimGrid is now instrumented in order to generate a trace file for
1902 visualization analysis: to use it, need to compile SimGrid with the
1903 "tracing" option enabled, and instrument the program using SimGrid with
1904 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1905 (among other functions).
1906 * The instrumentation only traces the platform utilization for now
1907 * Documentation to use the tracing functions and how to analyze the
1908 traces with the Triva tool is written.
1909 * More information about: SimGrid FAQ (in the section Tracing Simulations
1912 * We moved to cmake as default build system. Autotools support will
1913 be dropped soon. Check the FAQ for more info about how to use it.
1914 * Greatly improved our cdash/ctest interactions
1915 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1916 * Added memory checking tests with valgrind; lot of memleak fixing.
1917 This may be the first release of simgrid with so few memory issues
1918 * Added code coverage tests.
1919 Our coverage is still improvable, but at least we see it on cdash.
1921 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1923 SimGrid (3.3.4) stable; urgency=low
1925 The "Desktop Grid needs love too" release (also called Xmas release).
1927 Models improvements:
1928 * Major speedup in the maxmin system solving by using lazy evaluation
1929 Instead of solving completely the maxmin system at each iteration,
1930 only invalidate (and recompute) the modified parts.
1931 This new feature is enabled in default models but you can try to
1932 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1933 * Cas01 IMproved as default CPU model
1934 This CPU model is the same Cas01 model, but it uses the
1935 maxmin-selective-update flag and a heap structure to manage
1936 actions on SURF kernel.
1937 It reduces the complexity to find the next action to finish and,
1938 consequently, it's faster than the old Cas01.
1939 This is the new default CPU model (Cas01).
1940 * Rename the old Cas01 model to Cas01_fullupdate
1941 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1942 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1943 * CpuTI (CPU Trace Integration)
1944 A new CPU model whose objective is simulate faster when using
1945 availability trace files.
1946 Instead of using a full featured, over engineered maxmin system for
1947 CPU modeling, this model does the pre-integration of traces files
1948 to calculate the amount of CPU power available, and so, executes
1949 faster than the old CPU models.
1950 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1951 * Use LV08 as default network model since it gives better accuracy
1952 for small messages and shouldn't change things for big ones.
1953 Use --cfg=network_model:CM02 to get the previous behavior.
1956 ******************************************
1957 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1958 ******************************************
1959 * The new CPU model may changes simulations!
1960 The point is that events occurring at the exact same timestamp
1961 are not scheduled in the same order with the old and new
1962 version. This may be enough to completely change the execution
1963 of simulations in some cases.
1964 * The new network model will change simulations!
1965 This new model is more realistic than the previous one, so you
1966 should consider redoing your old experiments with this model.
1967 Sorry for the inconvenience.
1970 * Introduce the supernovae compilation mode
1971 When compiled that way, the whole SimGrid (or almost) is put in a
1972 single compilation unit and compiled in one shoot.
1973 This is to help gcc which has difficulties to inline stuff from one
1975 The speedup seem to be above 15%, althrough more tests are needed on
1976 amd64 to confirm that gain.
1979 * Port of MSG's mailbox on top of SIMIX network
1980 The put/get mechanism was greatly simplified on the way.
1983 * New SIMIX network module. Provides:
1984 - Mailbox: rendez-vous mechanism to find with who you want to speak
1985 - Synchronous send/recv: easier and hopefully faster since the
1986 logic is handled in the maestro process directly now
1987 - Asynchronous send/recv: you dreamt of it? It's here now
1988 Too bad that nobody cared enough to propagate the change to MSG.
1989 * Add semaphores as SIMIX synchronization mechanism.
1992 * new function SD_daxload(char*) to load a DAX file
1993 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1994 * Introduce typed tasks. Specify its kind and cost at creation.
1995 At scheduling, just give where it should be placed, and the cost
1996 for each involved resource is automatically computed.
1997 Existing constructors so far (more to come of course):
1998 - SD_task_create_comm_e2e() for end-to-end communication
1999 - SD_task_create_comp_seq() for sequential computation
2000 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2001 * new function SD_task_dump() for debuging display
2002 * new function SD_task_dotty(task,FILE*) writing to file the info
2003 about the task in dotty format
2004 * SD_task_dependency_exists() can now cope with having one of its
2005 arguments NULL. If so, it tests whether the other argument has any
2007 * Add getters on list of preceding/following tasks:
2008 SD_task_get_parents(task) and SD_task_get_children(task)
2009 * Add getters on amount of workstations and list:
2010 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2011 * Add getter on task kind: SD_task_get_kind(task)
2012 * Update the start_time and finish_time of tasks on completion/failure
2013 * Bugfix: Remove task from state swags when destroyed
2016 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2019 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2020 dynar with one separate thread per value of the dynar.
2021 * Change the prototype of xbt_thread_create(), sorry.
2022 Added a boolean parameter indicating whether we want to join this
2023 thread (used in SG only for now)
2024 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2027 * GTNetS wrappers should now be usable again (and betterly tested too)
2028 * Fix a major regression from 3.2 where the timeout provided to
2029 MSG_task_put_with_timeout() was used as absolute time before which
2030 the comm should be done.
2031 * Start to fix the <cluster> tag.
2032 - Internal links should be good now (beside of the loopback, which
2033 use the private link instead)
2034 - paths to the external world is still rather broken
2035 - the <route:multi> tag is just broken. Actually that's brain-dead.
2036 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2037 to make it less stupid
2038 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2039 * Fix a source-level compatibility glitch from 3.2: after defining
2040 MSG_USE_DEPRECATED, you can use the old name
2041 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2042 * Allow to compile from the SVN with automake 1.11
2043 * Fix some problems when using the "start_time" tag in deployment XMLs.
2044 * Fix #8569: XBT/synchro.h has redundant declarations
2045 * Fix #8563: MSG return values and exceptions
2046 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2047 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2048 Thanks to Robson Peixoto.
2049 * Fix a vicious bug in dictionaries inducing that some elements were
2050 not freed on xbt_dict_free()
2052 Portability report of this version:
2053 * Main portability targets:
2054 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2055 on (amd64/i386/ia64)
2056 - mac leopard on i386
2057 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2058 but nothing critical.
2059 * Other platforms: windows, AIX and others were not tested for this release
2061 Timing report of this version:
2062 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2063 scenario parameters). From 8h to a few seconds in desktop grid settings.
2064 * Supernovae brings about 25% speedup on i386.
2066 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2068 SimGrid (3.3.3) stable; urgency=low
2070 The "Need for Speed" release.
2072 The timings done to validate the 3.3.2 were faulty.
2073 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2075 The problem was a conversion from a manually handled vector to
2076 xbt_dynar_t on the critical path.
2077 xbt_dynar_foreach calls functions, inducing stack management crap.
2079 We inlined these functions and xbt_dynar_foreach is now breath taking.
2080 We also inlined xbt_swag_belong on the way.
2082 Here are some approximate speedup measurements (on master/slaves
2083 simulations lasting between 10s and 20s each):
2084 3.3.1 -> 3.3.2: about same performance
2085 3.3.2 -> 3.3.3: 40% speedup
2086 3.3.1 -> 3.3.3: 40% speedup
2087 3.3.1 with inline patch -> 3.3.3: 30% speedup
2089 Our reading is that the refactoring which occurred in 3.3.2 made us
2090 suffer much more from the xbt_dynar_foreach low performance, but
2091 once we solved this, this refactoring proved to be very performance
2092 effective. From the 40% speedup, somehow, 10% are due to the
2093 inlining and 30% to the refactoring.
2095 That's a pitty that gcc cannot inline functions placed in other files
2096 alone. We have to choose between:
2097 - break the encapsulation (by putting private data structures and
2098 accessors in headers files to help gcc)
2099 - live with low performance
2100 - switch to a decent compiler such as icc (not quite possible).
2102 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2104 SimGrid (3.3.2) stable; urgency=low
2106 The "Simplicity does not preceed complexity, but follows it" release.
2108 The main contributors of this release were (lexical order):
2109 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2112 * Extract the routing logic into its own object.
2113 (was dupplicated in network.c and workstation_LV07.c;
2114 Allows to implement other ways of storing that info)
2115 => kill now useless network_card concept
2116 - Use dynar to represent routes (instead of void** + int*)
2117 - kill link_set (use surf_network_model->resource_set instead)
2118 - Add a command-line option to choose the routing schema to use
2119 - Add three new models:
2120 * Floyd (shortest path computed at initialization)
2121 * Dijikstra (shortest path recomputed all the time)
2122 * Cached Dijikstra (shortest path computed on need)
2123 All these models where contributed by Silas De Munck, and are
2124 described in his ICCS09 paper.
2126 * Simplify model declaration
2127 (less redirections, less function to write when defining a model)
2128 - Factorize stuff between models:
2131 surf_model_resource_set(model)
2132 surf_model_resource_by_name(model, name)
2133 - Unify the types of models in s_surf_model_t (using an union)
2134 - Embeed fields of common_public directly into s_surf_model_t
2135 - Rename model methods:
2136 action_free ~> action_unref
2137 action_change_state ~> action_state_set
2138 action_get_state ~> action_state_get
2139 - Change model methods into functions :
2140 (model)->common_public->action_use ~> surf_action_ref
2142 * Implement a generic resource; use it as ancestor to specific ones
2143 (allows to kill duplicated code in models)
2144 Drawback: timer command don't need no name nor properties;
2145 workstation_CLM03 don't need no properties
2146 (but I guess we can live with those few bytes wasted)
2148 * Improve the action object model
2149 - implement a constructor avoiding dupplicated code about field
2150 initialization in generic_action part.
2152 * Kill the SDP model: it has an external dependency, is deprecated
2153 in flavor of modern lmm models, and didn't compile since a while
2156 * Relocation of the context module from XBT to SIMIX.
2157 (the context were decoupled from the simix processes, duplicating a lot of code)
2158 => a lot of code was factorized
2159 - less overhead is introduced during scheduling
2160 - simpler API for the context factory
2161 - the logic for process creation,destruction and manipulation was simplified
2162 * Simplification of the s_smx_process_t data structure.
2163 => accesing the simix level data associated to a process is faster now,
2164 and the code is a lot more readable.
2167 * Implement some more MPI primitives:
2168 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2169 -implementation: Bcast: flat or 2-ary tree (default),
2170 Barrier: 4-ary tree,
2172 Allreduce: Reduce then Bcast
2173 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2174 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2175 Alltoallv: flat tree, like ompi
2177 * Add support for optimized collectives (Bcast is now binomial by default)
2178 * Port smpirun and smpicc to OS X
2181 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2182 Shout out if you used it.
2185 * Display the list of still queued messages in SG mode when existing
2189 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2190 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2191 * Add xbt_str_from_file(FILE*)
2192 * Add xbt_dict_get_key achieving a linear reverse search
2193 * Remove the context module
2195 Portability report of this version:
2196 * Main portability targets:
2197 - Linux(debian)/x86/context
2198 - Linux(debian)/x86/pthreads
2199 - Linux(debian)/amd64/context
2200 - Linux(debian)/amd64/pthreads
2201 On these, we still have the eratic breakages of gras/pmm and
2202 amok/saturate_sg reported in previous version. We still think
2203 that the tests are the cause of the fault, not the tested code.
2205 - Mac OSX Leopard/x86/context
2206 Still false negative in tesh autotesting.
2207 Smpi still fails, but this time because readlink does not accept -f
2208 Everything seems to work properly beside of that.
2211 - AIX version 5.3 (only tested contexts this time)
2212 Smpi still fails there because mktemp is not installed.
2213 Everything seems to work properly beside of that.
2215 I managed to compile it for the first time, but several breakages.
2216 Won't delay the release for this exotic platform.
2218 * Windows: it's still lagging behind. If you want to help, please
2221 Timing report of this version:
2222 This version seem to be more than 5% faster than 3.3.1 (on linux
2223 64bits with contextes). The gain is less than expected, we are
2224 investigating this for next release.
2226 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2228 SimGrid (3.3.1) stable; urgency=low
2231 * Implement a --cfg-help to show existing configuration variables
2232 * Build chain do not require doxygen in maintainer mode
2235 * fix a bug on struct sizeof computation, which prevented the
2236 exchange of arrays of structs in some conditions
2237 - added a regression test about this in datadesc_usage
2238 * Allow the exchange of 0-long dynamic vectors.
2239 - for that, use -1 as indicator of dynamic size instead of 0
2240 - This implied to change any size from unsigned long to long,
2241 reducing a bit communication abilities, but I guess that with
2242 64bits being quite common, this is more than enough.
2243 - This also induce a protocol change, thus bumping network protocol
2244 version from 0 to 1 (if we have external users, we have to get
2245 clean on that point too ;)
2246 - added two regression tests about this in datadesc_usage
2247 * Be more verbose when propagating local exceptions
2248 This helps debugging.
2249 * Display the status of simulated processes when receiving SIGINT in
2253 * Allow to control the simulation from a trace file.
2254 New functions MSG_action_register() and MSG_action_trace_run()
2255 The first one allows to associate a function execution to each
2256 kind of action while the second one parses a trace file and
2257 triggers the corresponding actions within the system.
2258 For now, only a toy example is provided in examples/msg/actions
2259 * Add an exemple of process migration in examples/msg/migration
2260 * Fix a bug in task exchange which broke MSG_task_get_sender()
2261 Add a teshsuite regression test for that.
2262 [Bug: if MSG_task_get_sender() is called after sender exit,
2264 * Fix a bug which prevented suspend/resume to work properly
2265 * Display the status of simulated processes when receiving SIGINT
2266 This fixes a regression of v3.3. due to the introduction of SIMIX
2267 * Bug fixing in failure management:
2268 - trace could not start by a failure at time 0
2269 - failure during communications were not working
2272 * Add SIMIX_process_set_name() to change the name of the current
2273 process in the log messages.
2274 * Store smx_hosts in a dict since we only retrieve them by name
2275 * Move the configuration infrastructure to surf
2278 * Move the configuration infrastructure to surf
2281 * Massive internal cleanups:
2282 - Store internal structures on processes instead of hosts (allows
2283 to have more than one process per host, in addition of being more
2285 - Cleanup the initialization/finalization process
2286 - Kill a whole bunch of unneeded synchronization:
2287 processes run in exclusive manner within the simulator
2288 - Move queues from global tables to process data fields
2290 - now accept -platform and -hostfile arguments
2291 - Pass the right rank value to processes according to the hostfile
2292 * Compile the examples by default, and use them as regression tests
2293 * Implement MPI_Wtime()
2294 * Change the reference speed to a command line option
2297 * TCP_gamma can now be specified as command line option using
2298 --cfg=TCP_gamma:10000000.0
2299 * Change the --surf-path cmd line option into --cfg=path:
2302 * Also include strbuff from xbt.h public header
2303 * xbt_ex_display(): do not free the exception after displaying
2304 This allows to do more with the given exception afterward.
2305 Users should call xbt_ex_free() themselves.
2309 Portability report of this version:
2310 * Main portability targets:
2311 - Linux(debian)/x86/context
2312 - Linux(debian)/x86/pthreads
2313 - Linux(debian)/amd64/context
2314 - Linux(debian)/amd64/pthreads
2315 These targets fail about 1/10 of times on gras/pmm, but we believe
2316 that this is because of the test, not because of simgrid.
2317 amok/saturate_sg fails even more rarely, and the test may not be
2320 - Mac OSX Leopard/x86/context
2321 The test suite still spits tons of errors because some obscure
2322 force prevents us from removing the temporary directories
2323 arguing that they still contain some metadata I've never heard of.
2324 Smpi fails because seq is not installed.
2325 Everything seems to work properly beside of that.
2328 - AIX version 5.3 (both contexts and pthread)
2329 Smpi still fails there because mktemp is not installed.
2330 XML inclusions seems rosty on AIX.
2332 * Windows: it's still lagging behind. If you want to help, please
2335 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2337 SimGrid (3.3) stable; urgency=high
2341 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2342 [Malek Cherier & Mt]
2344 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2347 Until now, GRAS were using MSG as an interface to SURF. It was
2348 quite difficult because both interface have several differences
2349 (MSG channels vs GRAS sockets were the most notable point).
2351 This also opens the gate to SMPI (which should occur soon) and speed
2352 up simulations by to 40% (even if it were not the main goal).
2354 **************************************
2355 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2356 **************************************
2357 The point is that events occuring at the exact same timestamp are
2358 not scheduled in the same order with the old and new version. This
2359 may be enough to completely change the execution of simulations in
2360 some cases. Sorry for the inconvenience.
2362 * Cleanup and upgrade the XML format to push further scalability
2363 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2365 * Improve the testing infrastructure with tesh. Now a very large part of
2366 the code is tested not only by being run but also by checking that the
2367 output match an expected output [Mt].
2369 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2370 is really less memory-demanding, which should allow you to use
2371 larger files in SimGrid [AL].
2373 * Inform valgrind about our contextes, so that it becomes usable
2374 with the default (and more effecient) version of SimGrid
2375 [contributed by Sékou Diakite, many thanks]
2378 * Introduce a listener thread in charge of receiving incoming
2379 messages from the network. It allows to overlap communication and
2380 computation but most notably, it removes some stupid deadlocks due
2381 to the fact that so far, a process could not send and receive at
2382 the same time. This made most non trivial communication schema
2384 * Convert the PIDs from long int to int to match the MSG ones (and
2385 linux ones too) [Mt]
2386 * New function: gras_agent_spawn() to launch a new process on
2387 current host. Only working in simulation for now. [Mt]
2388 * New function: gras_os_hostport() returning a constant form (ie,
2389 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2392 * Make the backtrace of exceptions more human readable [Mt]
2393 * New module: xbt/str [Mt]
2394 a ton of string utility functions (split, join, printf to a newly
2395 allocated buffer, trim, etc)
2396 * New module: xbt/hash [Mt]
2397 SHA1 hashing algorithm (more to come if needed)
2398 * New module: xbt/synchro [Mt]
2399 synchronization tools (mutex and conditions) working the same way
2400 in simulation and in real life (mainly useful for GRAS, but not
2402 * New module: xbt/queue [Mt]
2403 classical producer/consumer synchronization scheme
2404 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2405 (using the classical functions will get serialized) [Mt]
2406 * Make dictionary internal table dynamic. No need to specify its size
2407 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2409 * Make sure the log channels are organized as a tree under windows
2410 (because of ANSI C compatibility issue, any channel were child of
2411 root directly) [Mt].
2414 * Cleaned many thing in surf and fixed a few bugs [AL].
2415 * Add a nice command line configuration mechanism to compose models [AL].
2416 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2417 the previous one (KCCFLN05). It relies on something that looks like
2418 a max-min sharing mechanism but cannot be written as such. A new solver
2419 was thus designed [AL].
2420 * Add a new solver to lmm. Based on Lagrange optimization and
2421 gradient-based descent, it enables to efficiently maximise systems s.a
2423 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2425 This solver enables to propose two new network models for TCP Reno and
2426 TCP Vegas based on Low's work. These models still need to be fully
2427 tested though [Pedro Velho].
2430 * Bug fix in SD_simulate. Now the time bound given as argument is
2432 * Use the new parallel task model (ptask_L07) as default.
2433 * Use the SURF command line configuration mechanism.
2434 * 0-size tasks (for synchronization) should now work.
2436 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2438 SimGrid (3.2) stable; urgency=high
2442 We still experience issues on this platform, but we believe that at
2443 least MSG is usable.
2445 GRAS API BREAKAGE (for simplification purpose, sorry):
2446 * the gras_msgtype_by_name is not used anymore. Instead of
2447 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2448 you can write (and must)
2449 gras_msg_send(toserver, "request", &request);
2450 - If you still want to pass a gras_msgtype_t to the function (to cache
2451 the type and avoid the lookup time), use the gras_msg_send_() variant.
2452 - Impacted functions:
2453 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2454 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2455 * The callbacks are now expected to return 0 when everything went well
2456 (just like the main() function)
2458 GRAS new features and improvements:
2459 * New module mechanism where user code can use per process globals [Mt]
2460 This is similar to gras_userdata_*() functions, but for libraries. It
2461 factorize some code developped over and over in the examples and AMOK.
2462 It has still to be documented and used (only amok/peermanagement is
2464 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2465 loss when they were small enough to fit into the buffer and sent quickly
2466 enough so that they can all get received in one shoot.
2467 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2468 if not found. Use the *_or_null() variant for the old semantic.
2469 * In gras_msg_handle, do not discard messages without callback.
2470 They are probably messages to be explicitly awaited later (ie, proofs of
2471 mis-synchronization in userland since they are sent before being awaited)
2472 No big deal usually.
2473 * gras_socket_meas_send/recv: semantic changed!
2474 The numerical arguments used to be (1) the total amount of data to send
2475 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2476 messages. This was need for the fool willing to send more than MAXINT
2477 bytes on quite fat pipes.
2480 * Do really rename the hostmanagement module to peermanagement. [Mt]
2481 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2482 but this is rather new and this was documented in the module
2483 documentation (poor excuses, I admit)
2484 * Bandwidth measurement semantic changed! This follows the changes to
2485 gras_socket_meas_send/recv explained above.
2488 * A sequential mode has been added to the workstations. When a workstation
2489 is in sequential mode, it can execute only one task, and the other tasks
2490 are waiting in a FIFO. [Christophe Thiery]
2493 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2494 model for SIMDAG. [Christophe Thiery]
2495 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2496 equal to 0, which caused some bad behaviors in
2497 saturated_constraint_set_update. I now use a threshold mechanism like in
2501 * When running manually src/testall, you select specific units [Mt]
2502 testall is the result of our cunit mechanism, and should replace all
2503 the scripty thingy around since bash don't run easily on billware.
2505 * A mallocator system has been added. [Christophe Thiery]
2506 Mallocators allow you to recycle your unused objects instead of freeing them
2507 and allocating new ones.
2509 Documentation update:
2510 * FAQ reworking + New FAQs:
2511 - "Valgrind spits tons of errors!" [Mt]
2512 - "How to repport bugs" [Mt]
2513 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2514 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2515 - Communication time measurement within MSG [AL]
2516 - I experience weird communication times when I change the latency [AL]
2517 * GRAS tutorial [Mt]
2519 - an introduction to the framework and to the used communication model
2520 - an initiatic tour introducing the most proheminent features:
2522 . Lesson 0: Installing GRAS
2523 . Lesson 1: Setting up your own project
2524 o Part 2: Message passing
2525 . Lesson 2: Exchanging simple messages
2526 . Lesson 3: Passing arguments to the processes (in SG)
2527 . Lesson 4: Attaching callbacks to messages
2528 . Lesson 5: Using globals in processes
2529 . Lesson 6: Logging information properly
2530 . Lesson 7: Using internal timers
2531 . Lesson 8: Handling errors through exceptions
2532 . Lesson 9: Exchanging simple data
2533 . Lesson 10: Remote Procedure Calling (RPC)
2534 . Lesson 11: Explicitely waiting for messages
2535 . Recapping of message passing features in GRAS
2536 - A HOWTO section containing:
2537 o HOWTO design a GRAS application
2538 More are due, of course. They will come latter. In the meanwhile, you can
2539 check the examples which are still here.
2541 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2543 SimGrid (3.1) stable; urgency=high
2547 There was a stack corruption somewhere, visible only when optimizing
2548 with these versions. [Vince]
2551 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2552 module that enabled to play with Directed Acyclic Graphs. It is built
2553 directly on top of SURF and provides an API rather close to the old
2554 SG. Some old codes using SG are currently under rewrite to check that
2555 all needful functions are provided. [Christophe Thiery]
2558 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2559 extension of the classical CLM03 model that gracefully handles
2560 failures. This is now the default model for MSG and GRAS. It doesn't
2561 handle parallel tasks yet though. [AL]
2562 * Bug fix: Weights were not correctly set in the network part.
2563 WARNING: This may have resulted in incorrect results with simulations
2564 where there are more than one flow on a given link. [AL]
2567 * After a (long ?) discussion on simgrid-devel, we have decided that the
2568 convention we had on units was stupid. That is why it has been decided
2569 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2570 WARNING : This means that all previous platform files will not work as
2571 such with this version! A warning is issued to ask users to update
2573 A conversion script can be found in the contrib module of the CVS, under
2574 the name contrib/platform_generation/surfxml_update.pl [MQ]
2577 * Bug fix: Processes were started in reverse order, wrt deployment file.
2578 WARNING: if your code relies on this bug, please fix it. [AL]
2579 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2580 being executed on two different locations. [AL]
2581 * Bug fix: Failures are now better supported thanks to Derrick's tests
2582 (there was many failure situations I hadn't thought of and that weren't
2583 correctly handled). [AL]
2584 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2588 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2589 Exception occurring on server-side are propagated back to client (!).
2591 API CHANGE: the callback changed their prototype. Change:
2592 int my_handler(gras_socket_t expeditor, void *payload_data) {
2594 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2595 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2597 * New! function: gras_msg_handleall to deal with all messages arriving
2598 within a given period.
2599 * New! function: gras_socket_server_range to get a server socket in a
2600 range of port numbers (ease to avoid port number conflicts) [MQ]
2601 * New! gras processes display their backtrace when they get a SIGUSR1
2602 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2603 Sweet to debug RL processes [MQ]
2607 - Do not force experiment sizes to be expressed in kb, or it becomes
2608 impossible to measure the latency this way (needs one byte-long tests)
2609 WARNING: this changes the amok_bw_* function semantic. [MQ]
2610 - Implements the link saturation stuff. [MQ]
2611 * Peer management module:
2612 New! module factorizing code that we wrote over and over [MQ].
2615 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2616 - Test units are placed directly into the library code, they get extracted
2617 automatically and placed into the src/testall binary.
2618 - Convert most of the XBT tests to this system.
2619 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2620 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2621 live in the same C code. This required to change a little bit the API
2622 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2623 * New module: sparse graph structure with basic algorithms (this is work
2624 in progress and the API is not considered to be frozen yet). [AL]
2625 * Display more information on backtraces: source line & function names are
2626 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2627 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2628 * Reworked a little bit some #include statements to load only required
2629 headers. Some user code that relied on SimGrid to include stdlib or
2630 stdio may need to include it by themselves. [AL]
2631 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2633 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2634 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2635 but I guess that almost nobody used those parts. [MQ]
2637 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2639 SimGrid (3.0.1) stable; urgency=low
2642 * Unfortunately, I had missed 5 misnamed functions:
2643 xbt_fifo_item_t xbt_fifo_newitem(void);
2644 void xbt_fifo_freeitem(xbt_fifo_item_t);
2645 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2646 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2647 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2648 They're now deprecated. Please use their new versions:
2649 xbt_fifo_item_t xbt_fifo_new_item(void);
2650 void xbt_fifo_free_item(xbt_fifo_item_t);
2651 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2652 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2653 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2655 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2656 * Documentation: xbt_log module unmercifully reworked [MQ]
2657 * Bugfix: there was a problem with the ending of contexts with
2658 the pthread backend. It caused some weird deadlock or behavior
2659 depending on the pthread implementation. [AL]
2660 * Bugfix: get the exceptions raised in the simulator repport where
2661 and why they come from when they are not catched in time [AL, MQ]
2664 * Bugfix: Do repport the error when two non-connected hosts try to
2665 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2668 * Add additionnal checkings on communications. Assert that two
2669 communicating hosts are connected by a set of links... [AL]
2672 * Add additionnal checkings on channel values in communication [AL]
2673 * New: MSG_task_get_source to see on which host a task was generated [HC]
2674 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2675 the number of tasks waiting to be received on channel and sent
2677 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2678 waits for the first task coming from a given host.. [AL]
2680 GRAS new functionnalities: [MQ]
2681 * Enhance the parsing macro to allow the size of multidimentional objects
2682 to be given thru annotations.
2683 * New example (and documentation): Matrix Multiplication a la RPC
2684 (as when I was young!) and fix a bunch of bugs found on the way.
2686 GRAS performance improvements: [MQ]
2688 * Reduce the amount of cbps creation/destruction by making it static to
2689 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2692 * Change libdata to a set so that we can search for stuff by ID (and thus
2693 reduce the insane amount of dict lookups)
2696 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2697 RL communication on top of SG since it's so uneffective.
2698 It may also be used for inter-thread communication in RL, one day.
2699 * Use gras_datadesc_copy() to exchange messages on top of SG
2701 - improve message exchange performance on top of SG
2702 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2703 * Don't exchange on the network the size of the used part of buffer,
2704 instead, specify the possible buffer size to read().
2706 - reduces the amount of read/write calls (one pair per exchange)
2707 - reduces the amount of exchanged data (the size)
2708 - allows to retrieve all arrived data on receiver side, if we don't need
2709 it right now (subsequent read will peek the buffer)
2710 - allows the receiver to proceed with the begining of the stream before
2711 everything is arrived
2712 - make it possible to build an iov transport (using readv/writev)
2714 - take care of the data with non-stable storage (like stacked data),
2716 * If possible, TCP send uses vector I/O (when writev() is here)
2717 - Don't use it for receive since we send data sizes and data on the
2718 same stream, so we wouldn't be able to chain large amount of chunks
2719 before having to flush the stuff to read the size.
2720 * Rework the transport plugin mechanism to simplify it and reduce the
2721 amount of pointer dereferencement when searching for the right function
2724 * I guess that now, we do almost as few system calls as possible while
2725 doing as few data copy as possible.
2727 To improve it further, we could try to send all the sizes first and then
2728 all the data (to use iov on receiving size), but it's only a partial
2729 solution: when you have 2 dimensional data, the sizes of the second
2730 dimension is data of the first dimension, so you need 3 streams.
2732 I'm not sure the potential performance gains justify the coding burden.
2734 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2736 SimGrid (3.00) stable; urgency=high
2739 * New! Give the possibility to hijack the surf parser and thus bypass
2740 MSG_create_environment and MSG_launch_application. Have a look at
2741 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2743 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2745 SimGrid (2.96) unstable; urgency=low
2750 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2751 This deprecates the xbt_error_t mechanisms.
2752 It modifies (simplifies) all XBT and GRAS API.
2753 MSG API keeps unchanged (exceptions raised by XBT are catched from
2754 within MSG and masked with existing error handling facilities)
2757 * New! Add a FATPIPE model. [AL]
2758 * New! Add a parallel task model. [AL]
2759 * New! Add automatically a loopback interface (in the default
2760 network model) if none was precised.
2763 * Bugfix: MSG_process_resume now works with the current running process.
2765 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2766 * Modification of MSG_task_get_compute_duration. Once a task has been
2767 processed, the value returned by this function is now equal to 0. [AL]
2768 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2769 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2770 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2771 of a m_task. For now, MSG_task_cancel only works with computation
2773 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2774 of the processor (in Mflop/s) regardless of the current load on the
2776 * API Change: use proper naming convention for MSG_getClock and
2777 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2779 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2780 This function changes the priority of a computation task. This priority
2781 doesn't affect the transfer rate. A priority of 2 will make a task
2782 receive two times more cpu power than the other ones. This function
2783 has been added to suit the needs of Nguyen The Loc and hasn't been that
2784 much tested yet. So if it fails, please report it and send me your code.
2786 * API Change: removed all functions and types that were marked "deprecated"
2787 since many months. Renamed MSG_global_init_args to MSG_global_init.
2789 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2791 SimGrid (2.95) unstable; urgency=low
2794 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2795 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2796 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2799 * Bugfix: Allow absolute paths for platform description files [MQ]
2800 * Bugfix: do free the variables after use. Leads to drastic performance
2802 * Implement max_duration (ie, timeouts) on resources [AL]
2805 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2807 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2808 the needed support (timeouts) [AL]
2811 * Implement measurement sockets. You can now get the bandwidth between two
2812 hosts thanks to AMOK (see below). [MQ]
2813 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2814 dynar over the network (yeah) [MQ]
2815 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2816 * simple/static token-ring example. [Alexandre Colucci and MQ]
2817 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2818 in place before [MQ]
2821 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2822 type parsing. "short" and "long" modifiers were also ignored; other
2823 modifier (such as reference level) are still ignored. [MQ]
2824 * Embeed the buffer size within the buffer itself on SG. [MQ]
2825 That way, send() are atomic and cannot get intermixed anymore (at least
2826 the ones which are less than 100k; bigger messages still have the issue)
2827 * Array size pushed by the field, not by the field type (or each
2828 and every long int will push stuff to the cbps) [MQ]
2829 * use select() to sleep since it allows to portably sleep less than one
2832 GRAS (minor cleanups)
2833 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2836 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2837 * type category 'ignored' killed as it was never used and were difficult
2839 * whether a type can cycle or not is now a flag, leaving room for more
2840 flags (as "ignored", if we feel the need one day ;)
2841 * Rename raw sockets to measurement sockets since "raw" has another
2842 meaning in networking community.
2845 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2846 and offers features not belonging to GRAS but that most applications
2847 need. One day, it may be a set of plugins loadable at runtime.
2848 * New module: bandwidth
2849 bandwidth measurement between arbitrary nodes running this module. [MQ]
2851 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2853 SimGrid (2.94) unstable; urgency=low
2855 The first beta release of SimGrid 3 !
2859 * Update the main page and the FAQ. Adding references to gforge.
2862 * Add a gras_os_getpid function.
2865 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2866 * Extend the logs so that they also print PID, hostname, date, ... if
2868 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2869 and kill the old version which were in testsuite/
2870 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2874 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2875 for users. Instead, each programming environment (GRAS, MSG,...) use it
2876 when needed to provide such a feature to users.
2877 Don't shortcut the mechanism or you will also shortcut the virtualization
2878 you need on the simulator.
2882 * Cleanups in configury with regard to compile optimization/warning flags.
2883 Also add -fno-loop-optimize to any powerpc since it's the optimization
2884 killing gcc (< 3.4.0).
2885 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2886 needed by MSG examples complications
2887 * Borrow configury beautifications from PHP
2890 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2891 warning (thanks loris for stumbling into this one).
2892 * Bugfix: stop loading private headers (gras_config.h) from the public
2896 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2897 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2898 RL would imply a DNS resolver, which is *hard* to do in a portable way
2899 (and therefore delayed).
2900 * Implement a real timer mechanism and use it in timing macros. This allows
2901 to avoid rounding errors and get a 0.000005 sec precision in timing
2902 macros. While I was at it, various cleanups:
2903 - allow to declare more than one timed section per file (fix a stupid bug)
2904 - move some private declaration to the right place
2905 - merge conditional execution and timing macros into emulation module
2906 - document the module
2907 - make sure the module cleanups its mess on gras_exit
2908 * Documentation improvements:
2909 - (new) how to compile applications using GRAS
2910 - (new) emulation support (timing macros)
2912 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2914 SimGrid (2.93) unstable; urgency=low
2916 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2919 - Use Paje properly where used. Still to be sanitized properly.
2920 - Portability fix: Add an implementation of the contexts using pthread
2924 - Add xbt_procname(): returns the name of the current process.
2925 Use it to show the current process's name in all logging.
2927 - fix detection of older flex version and the reaction, since we do
2928 depend on modern ones (we use lex_destroy)
2929 - Better separation of SG and RL in the libs: remove all simulation code
2930 from libgras. As a result, this lib is now only 200k when stripped.
2931 Some of the xbt modules may also be duplicated (two sets and such) and
2932 should be cleaned/killed before SG3.
2933 - Insist on using xlC on AIX because of weird problems involving gcc there.
2934 - Cleanup the make remote stuff. This is now done by scripts
2935 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2936 mainly for our private use, but we're working on changing them to user
2939 - Bugfix: flush the socket on close only if there is some *output*.
2940 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2942 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2945 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2947 SimGrid (2.92) unstable; urgency=low
2949 Alpha 3 on the path to SimGrid 3
2953 - New! First try of benchmarking macros.
2954 - New! First try so that gras_stub_generator generate deployment and
2955 remote compilation helpers.
2957 - Bugfix: Initialization fix in msg_test.
2961 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2963 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2965 - API Change: gras_os_sleep to take the amount of seconds as a double.
2966 Accepting an int was error prone since it was the only location where
2967 seconds were coded as such. It leaded to damn rounding errors.
2968 - Bugfix: Hard to belive that timers ever worked before this.
2970 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2972 SimGrid (2.91) unstable; urgency=low
2974 Alpha 2 on the path to SimGrid 3
2978 - Bug fix in the lmm_solver.
2980 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2981 through the function MSG_paje_output.
2982 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2983 - It is possible to bound the rate of a communication in MSG with
2984 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2985 to write it in the changelog).
2986 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2991 - Various cleanups to the autotools stuff
2992 - Begin to move Gras examples to examples/gras/
2993 - Let make distcheck work again (yeah!)
2995 - documentation overhauled using doxygen.
2996 gtk-doc-tools is dead in SimGrid now.
2997 - Automatically extract all existing logging categories, and add the list
2998 to the documentation (long standing one, to say the less)
3000 - Cleanup the known architecture table. Reorder the entries to group what
3001 should be, and use a more consistent naming scheme.
3002 (some of the test dataset are still to be regenerated)
3003 - New! Allow library to register globals on each process just as userdata
3005 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3006 lookup time (for now).
3007 Use it in msg and trp.
3008 This cleans a lot the internals and helps enforcing privacy of the
3009 headers between the gras components.
3010 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3011 - Bugfix: gras_os_time was delirious in RL.
3012 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3014 - Reenable GRAS now that it works.
3016 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3018 SimGrid (2.90) unstable; urgency=low
3020 Alpha 1 on the path to SimGrid 3
3022 * It is a long time since the last release of SimGrid. I'm sorry about
3023 that but as I had told you, I was rewriting a lot of things. I apologize
3024 to those who had been reporting bugs to me and that I had not answered.
3025 If your bug is still in the new version, please tell me. Here is a
3026 summary of the main changes.
3028 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3029 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3030 portability, speed, and a lot more but you'll figure it out later.
3031 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
3032 I wanted to release the new MSG as soon as possible and I have
3033 broken GRAS, which is the reason why, for now, only MSG is fully
3034 functional. A laconic description of these projects is available
3035 in the documentation.
3037 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3038 kernel optimized for our needs (called SURF but only the developers
3039 should use it). Hence, MSG is now roughly 30 times faster and I think
3040 that by rewriting a little bit MSG, I could event speed it up a little
3041 bit more. Beside the gain in speed, it is also much easier to encode a
3042 new platform model with SURF than it was with SG. More to come...
3044 * REVOLUTION 3: I have tried to change a little as possible the API of
3045 MSG but a few things really had to disappear. The main differences
3046 with the previous version are :
3047 1) no more m_links_t and the corresponding functions. Platforms are
3048 directly read from a XML description and cannot be hard-coded
3049 anymore. The same format is used for application deployment
3050 description. The new format is described in the documentation.
3051 Have a look in tools/platform_generation. There is a tiny script
3052 that converts from the old platform format to the new one. Concerning
3053 the application deployment format, parsing the old one is tricky.
3054 I think most of you should however be able to convert your files. If
3055 it is really an issue, I can write a C code that does the conversion.
3057 2) the toolbox tbx does not exist anymore. We now have a library
3058 with much more data-structures but without the hash-tables (we have
3059 dictionaries that are much faster).
3061 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3063 *****************************************************************************
3064 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3065 * don't feel like converting the oldies. (Mt) *
3066 *****************************************************************************
3069 Version 2.90: "the long awaited one"
3070 - Finished rewriting and debugging MSG. Rewrote the documentation.
3071 - disable GRAS for now since it needs to be ported to the newest SG
3074 - Finish the port to windows (using mingw32 for cross-compile)
3077 - Main loop and datastructures of SURF. A cpu resource object is
3078 functional. Surf can thus be used to create cpu's with variable
3079 performance on which you can execute some actions.
3081 2004-11-15 Martin Quinson
3082 - Port to ARM. Simply added the alignment and size descriptions. Should
3083 work, but the ARM machines are so slow that I didn't had the opportunity
3084 to 'make check' over there yet.
3086 2004-11-15 Arnaud Legrand
3087 - Trace manager now written. It uses a heap structure and is therefore
3088 expected to be efficient. It may however be speeded up (particularly
3089 when many events occur at the same date) by using red and black
3090 trees. One day maybe...
3091 - Max-min linear system solver written. It uses a sparse matrix
3092 structure taking advantage of its expected use. Most operations are
3093 O(1) and free/calloc are called as few as possible. The computation of
3094 the minimum could however be improved by using a red and black tree
3097 2004-11-03 Arnaud Legrand
3098 - Rename every gras_* function that was in xbt/ to its xbt_
3100 - Add a heap and a doubly-linked list to xbt
3101 - Added a dichotomy to the dictionaries. make check works as well before
3102 so I assume that the patch is correct. I do not know however if things
3103 run effectively faster than before now. :)
3105 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3106 SimGrid, and the version number is bumped to 2.x
3108 2004-10-29 Martin Quinson
3109 - Introduction of the remote errors.
3110 They are the result of a RMI/RPC on the remote machine.
3111 ErrCodes being scalar values, you can't get the host on which those
3112 errors did happen. Extending the error mechanism as in Gnome is possible.
3113 No idea yet whether it is a good idea.
3115 2004-10-28 Martin Quinson
3116 - Interface revolution: the Starred Structure Eradication.
3117 I used to do typedef struct {} toto_t; and then handle *toto_t.
3118 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3119 - ???_t is a valid type (builded with typedef)
3120 - s_toto_t is a structure (access to fields with .)
3121 - s_toto is a structure needing 'struct' keyword to be used
3122 - e_toto_t is an enum
3123 - toto_t is an 'object' (struct*)
3125 typedef struct s_toto {} s_toto_t, *toto_t;
3126 typedef enum {} e_toto_t;
3127 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3128 s_toto_t) is private.
3130 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3131 so that it can set the variable to NULL. It was so for dicts and sets,
3132 it changed for dynars.
3134 - Fix a bunch of memleaks in dict_remove
3135 - Fix a bug in sg/server_socket opening: it failed all the time.
3137 2004-10-07 Martin Quinson
3138 - Speed up dynar lookup operation a bit.
3140 gras_dynar_get is dead.
3142 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3143 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3144 (which gives you the address of the stored data).
3146 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3147 copy of the data using an affectation to do the job and not a memcpy.
3149 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3150 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3152 It does not really speedup the dynar test because they are
3153 setting elements all the time (and look them seldom). But the dict does
3154 far more lookup than setting.
3156 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3158 2004-10-05 Martin Quinson
3159 - Allow to (en/dis)able the cycle detection at run time.
3161 Whether we should check for cycle or not is now a property of each
3162 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3163 datadesc_cycle_unset allow to remove this property when previously set.
3165 Note that the cycle detection is off by default since it impacts the
3166 performance. Watch the data you feed GRAS with ;)
3168 This property is hereditary. Any element embedded in a structure having it
3169 set have it set for the time of this data exchange.
3171 You should set it both on sender and receiver side. If you don't set it on
3172 sender side, it will enter an endless loop. If you forget on receiver
3173 side, the cycles won't be recreated after communication.
3175 - Header reorganization.
3176 Kill gras_private.h, each submodule must load the headers it needs.
3178 2004-10-04 Martin Quinson
3179 - Interface revolution: do not try to survive to malloc failure.
3181 Now, gras_malloc and friends call gras_abort() on failure.
3182 As a conclusion, malloc_error is not a valid error anymore, and all
3183 functions for which it was the only gras_error_t return value are
3184 changed. They now return void, or there result directly.
3185 This simplify the API a lot.
3187 2004-09-29 Martin Quinson
3188 - Re-enable raw sockets.
3189 Created by gras_socket_{client,server}_ext;
3190 Used with gras_raw_{send,recv}
3193 It should allow to kill the last bits of gras first version soon.
3195 This is not completely satisfactory yet (duplicate code with
3196 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3199 - Simplify transport plugin (internal) interface by not passing any
3200 argument to _server and _client, but embedding them in the socket
3203 2004-09-28 Martin Quinson
3204 - Finish the port to AIX.
3205 autoconf was my problem (segfault within the malloc replacement
3206 function. No idea why)
3208 2004-09-16 Martin Quinson
3209 - Fix some size_t madness on 64bit architectures.
3211 2004-09-08 Martin Quinson
3212 - Reduce the number of system headers loaded, overload some more system
3213 calls (such as malloc to cast the result of the system one, and work
3215 - Fix and reintroduce the config support
3217 2004-09-07 Martin Quinson
3218 - Source code reorganization to allow Arnaud to surf all over there.
3219 - Allow to document the logging categories.
3220 - Remove all uppercase from logging categories and useless cleanup in names.
3222 2004-08-18 Martin Quinson
3223 Version 0.6.2 (protocol not changed; API changed)
3224 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3225 gras_error_t), and NULL when not found. Functions expecting a msgtype
3226 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3227 hopefully usefull message.
3228 - Portability to prehistoric sparcs again
3230 2004-08-17 Martin Quinson
3231 Version 0.6.1 (protocol not changed; ABI not changed)
3232 - prealloc some buffers to speed things up
3234 2004-08-11 Martin Quinson
3235 Version 0.6 (protocol not changed; ABI expended)
3236 - The parsing macro can deal with the references, provided that you add
3237 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3239 2004-08-09 Martin Quinson
3240 Version 0.5 (protocol not changed; ABI changed)
3241 - Allow to off turn the cycle detection code in data exchange at
3242 compilation time. It should be at run time, but I'm short of time (and
3243 the config stuff is still broken). That way, we keep dict out of the
3244 critical path, which is good because the performance is poor:
3245 - search not dichotomial yet
3246 - dynar give no way to access their content and memcpy everytime
3247 - In composed data description (struct, ref and so on), stop foolness of
3248 keeping the subtype's ID, but store the type itself. This keeps sets out
3249 of the critical path, which is good since they rely on dynar and
3250 dictionnaries. The only loose of that is that we cannot detect the
3251 redeclaration of a structure/union with another content (but I'm not sure
3252 the code detected well this error before anyway). We still can detect
3253 the redefinition discrepancy for the other types.
3254 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3255 it breaks the code because of type-punning used all over the place).
3256 This breaks on all non-gcc architectures (for now).
3258 All those changes (plus the buffer of last time) allow me to gain 2 order
3259 of magnitude on cruel tests consisting of 800000 array of integers on two
3260 level of a hierarchical structure (200 secondes -> 4 secondes)
3263 - the selector of reference must now return the type it points to, not
3264 the ID of this type.
3266 2004-08-06 Martin Quinson
3267 Version 0.4 (protocol changed; ABI not changed)
3268 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3269 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3270 - Implement a buffer transport, and use it by default (it relies on tcp in
3271 real life and on sg in simulation).
3272 That's a bit hackish since I had a new field to the structure to store
3273 its data without interfering with the subtype ones. Inheritance
3274 is tricky in C. And that's a kind of reverse inheritance with one class
3275 derivating two classes. Or maybe a game with java interfaces. Anyway,
3276 that's damn hard in C (at least).
3277 Moreover, I got tired while trying to ensure plugin separation and
3278 genericity in SG mode. MSG wants me to do weird things, so let's go for
3279 cruel hacks (temporarily of course ;).
3280 See comment in transport_private.h:71
3281 - do not include all the _interface headers in private but in the files
3282 which really need them (to cut the compilation time when they are
3285 2004-07-26 Martin Quinson
3286 Version 0.3 (protocol not changed; ABI changed)
3287 - Major overhault of the datadesc interface to simplify it:
3288 - shorted the function names:
3289 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3290 - add a trivial way to push/pop integers into the cbps without malloc.
3291 This allows to make really generic sub_type description, which simply
3292 pop their size of the stack.
3293 - add a function gras_datadesc_ref_pop_arr() which does what users want
3294 most of the time: Declare a dynamic array (which pops its size of the
3295 stack) and declare a reference to it. Poor name, but anyway.
3296 - kill the post-send callback, add a post-receive one
3298 2004-07-23 Martin Quinson
3299 Version 0.2 (protocol changed; ABI changed)
3300 - add some testing for cpbs in the test cases, and fix some more bugs.
3301 This invalidate again the little64 data file, since I cannot regenerate
3303 - remove an awfull optimization in the logging stuff, allowing me to:
3304 - understand it again
3305 - learn gcc how to check that the argument match the provided format
3306 - fix all errors revealed by gcc after that
3307 - internal keys of dict are not \0 terminated. Deal with it properly in
3308 loggings instead of segfaulting when the user want to see the logs :-/
3310 2004-07-22 Martin Quinson
3311 - Fix some stupid bug preventing cbps (callback postit) from working
3313 2004-07-21 Martin Quinson
3314 - Some documentation cleanups
3315 - remove the useless last argument of msgtype_declare
3316 - rename the Virtu functions to fit into the 'os' namespace
3317 - move headers src/include -> src/include/gras/ and stop fooling with
3318 gras -> . symbolic link
3319 - make distcheck is now successful
3321 2004-07-19 Martin Quinson
3323 - Build shared library also
3324 - Install html doc to the right location
3325 - stop removing maintainer files in make clean
3326 - build tests only on make check
3328 2004-07-13 Martin Quinson
3330 - No major issue in previous version => change versioning schema
3331 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3332 corresponding dataset.
3334 2004-07-11 Martin Quinson
3336 - Get it working with any kind of structure (we can compute the padding
3337 bytes remotely for all the architectures I have access to)
3338 - Implement the structure parsing macro (still not quite robust/complete)
3339 - Improvement to the remote testing toysuite
3341 2004-07-10 Martin Quinson
3342 [autoconf mechanism]
3343 - get ride of a bunch of deprecated macros
3344 - actually run the test for two-compliment, not only compile it :-/
3345 - test whether the structures get packed (and bail out if yes. Damn.
3346 Alignment is a serious matter)
3347 - test whether the structures get compacted (but respecting the alignment
3348 constraints of each types)
3349 - test whether the array fields of structures can straddle alignment boundaries
3351 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3352 (took me 2 hours to find that bug, looking at the wrong place)
3353 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3354 before sending/receiving and are used to compute the offsets of fields
3355 - Given that padding size depend even on compiler options, keep track of
3356 alignment and aligned_size only for the current architecture. Not a big
3357 deal since we send structure fields one after the other (seems
3359 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3360 program, let it work on linux/gcc/little32. portability todo.
3362 2004-07-08 Martin Quinson
3363 - import and improve remote compilation support from FAST
3364 - make sure make check works on half a dozen of machines out there
3366 2004-07-07 Martin Quinson
3367 Let's say it's version 0.0.3 ;)
3368 - Implement conversions (yuhu!)
3369 - Let it work on solaris (beside conversion, of course)
3370 - Stupid me, using rand() to generate the conversion datatests in not wise.
3372 2004-07-06 Martin Quinson
3373 - Let make dist work, since I'm gonna need it to compile on remote hosts
3374 - Let Tests/datadesc_usage write the architecture on which the file was
3375 generated as first byte.
3376 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3377 architecture descriptions.
3378 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3379 on those architectures.
3380 - Optimization: send/recv array of scalar in one shoot
3382 2004-07-05 Martin Quinson
3383 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3385 - Plug a whole bunch of memleaks
3386 - each process now have to call gras_{init,exit}. One day, their log
3387 settings will be separated
3388 - Continue the code factorisation between SG, RL and common in Transport.
3390 2004-07-04 Martin Quinson
3392 - Redistribution between SG and RL.
3393 We wanna have to accept in SG, so move accepted related parts of RL in
3394 the common part. (more precisely, the dynar of all known sockets is no
3395 more a static in transport.c, but part of the process_data)
3397 [gras_stub_generator]
3398 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3400 2004-07-03 Martin Quinson
3401 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3402 them all up in one shot)
3404 - Fix the ugly case of reference to dynamic array.
3405 - New (semi-public) function gras_datadesc_size to allow the messaging
3406 layer to malloc the needed space for the buffer.
3408 - gras_socket_close now expect the socket to close (and not its address to
3409 put NULL in it after it). This is because the socket passed to handlers
3410 is one of their argument (=> not writable).
3412 - propagate the interface cleanup from last week in datadesc, ie remove a
3413 superfluous level of indirection. User pass adress of variable
3414 containing data (both when sending and receiving), and not of a variable
3415 being a pointer to the data. Let's say that I like it better ;)
3416 The price for that is constructs like "int msg=*(int*)payload" in
3417 handlers, but it's a fine price, IMHO.
3419 - Let it work in RL (yuhu)
3421 2004-06-21 Martin Quinson
3423 - porting SG plugin and SG select to new standards (works almost).
3424 - plug memleaks and fix bugs around.
3427 - cleanup the prototype of data recv and force users to specify when they
3428 want to handle references to objects. Test case working even for cycles.
3429 - plug memleaks. Valgrind is perfectly ok with this.
3431 2004-06-12 Martin Quinson
3433 - cleanup the separation between plugin and main code in plugin creation
3435 2004-06-11 Martin Quinson
3437 - Reput hook for raw sockets, needed for BW experiments
3438 - kill a few lines of dead code
3439 [Data description] Interface cleanup
3440 - gras_datadesc_by_name returns the searched type or NULL.
3441 That way, no variable is needed to use a type desc once, which makes
3443 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3444 two parameters were strings (field name, type name), leading to
3446 [Dicos] Interface cleanup
3447 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3448 This is consistant with the dynar API.
3450 2004-04-21 Martin Quinson
3452 - Porting to new standards.
3454 - interface cleanup.
3455 There is no bag anymore, no need to take extra provision to mask the
3456 pointers behind "ID".
3457 Better splitup of functions between files create/exchange/convert.
3458 This is still a bit artificial since convert and receive are so
3459 interleaved, but anyway.
3461 - add a queued message list to procdata (the ones not matching criteria
3463 - factorize some more code between SG and RL wrt procdata
3465 - use gras_exit in example to track memleaks
3466 - get rid of gs_example now that GS is properly integrated into gras
3467 - update run_test to integrate the latest tests (datadesc)
3469 - rename WARNINGn macros to WARNn since it prooved error-prone
3471 2004-04-19 Martin Quinson
3473 - register init/exit functions within gras module mechanism
3474 - send/receive function.
3475 Convertion is not implemented, but short-cutted if not needed.
3476 struct/array elements are sent one by one (instead of block-wise), but
3477 nobody really cares (yet). Get a prototype before optimizing.
3478 - tests (using a file socket) for DD send/receive on:
3479 - base types: int, float
3480 - array: fixed size, string (ie ref to dynamic string)
3481 - structure: homogeneous, heterogeneous
3482 - chained list, graph with cycle
3483 Believe it or not, valgrind is not too unhappy with the results. The
3484 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3488 - Bugfix when using a filename explicitely (instead of '-')
3490 2004-04-09 Martin Quinson
3492 - factorize more code between RL and SG in socket creation
3493 - Complete the implementation and tests of:
3495 o file (only in RL, and mainly for debugging)
3497 I lost 3 days to design a portable address resolver, and then decided
3498 that the prototype mainly have to run on my box.
3499 Addressing portability too early may be like optimizing too early :-/
3501 - use gras_init in the Tests instead of the crappy parse_log_opt
3502 (the latter function is removed)
3503 [Conditional execution]
3504 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3505 [Code reorganisation]
3506 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3507 Build examples against the RL library, since there is no way to disable
3508 its creation for now.
3510 For information, the beginning of coding on GRAS was back in june
3511 2003. I guess that every line has been rewritten at least twice since