1 SimGrid (3.19) NOT RELEASED YET (target: March 20 2018, 16:15:27 UTC)
4 - Fix MSG_task_get_remaining_work_ratio() to return 1.0 for tasks that have
6 - Remove parameter of MSG_process_killall().
7 Resetting the PID was bogus anyway (several actors could have the same PID).
10 - Use a graphical TOC to make it easier to find the documentation you need
11 - Revamp the MSG tutorial
14 - Execution->setHost() can be called after start() to migrate it.
15 - Comm::test_any() is now implemented.
16 - s4u::Actor now has onCreation() and onDestruction() signals.
17 - Install some kernel header files for the users' plugins and more.
18 Warning, their API is really not stable yet.
19 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
20 include/simgrid/kernel/routing/*Zone.hpp
23 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
26 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
29 - The privatization of global variables via mmap() is now significantly more
30 flexible and allows the use of daemons with SMPI and to add new processes
31 while the simulation is running. (Adding new SMPI instances after the
32 simulation has already started should also be technically doable now.)
33 - Removed index notion from SMPI -- no more "getPid() - 1"!
34 This also means that the trace files
35 now contain actors called rank-<PROCESS-ID>. These are NOT the real
36 ranks, because tracing several communicators right now is not supported.
37 - Because of this, smpi_process_index() has now been deprecated and will
39 - Support of MPI_Type_create_subarray
40 - SMPI will now crash and not just warn when encountering an unimplemented
41 call. Some exceptions are set (mainly for errhandler calls).
42 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
45 - SimGrid now works with intel compilers (MC still buggy)
48 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
51 - Remove unused run-time parameter "tracing/onelink-only".
54 - #245: migrating an actor does not migrate its execution
55 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
56 - #256: Modernize FindSimGrid.cmake
57 - #257: Fix (ab)use of CMake install
59 SimGrid (3.18) Released December 24 2017
61 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
63 MSG IS NOW DEPRECATED. Please use S4U for new projects.
64 - Support for MSG should not be removed from SimGrid before 2020,
65 but future evolutions will be limited to the integration of
66 user-provided patches.
69 - s4u::onDeadlock() signal, run before stopping the simulation.
70 - s4u::Exec: asynchronous executions (abstraction of background
72 - s4u::Host->getLoad() returns the achieved speed in flops/s
75 - Link::name() is deprecated and will be removed in v3.21, use
76 Link::getCname() instead.
77 - Mailbox::getName() changed to return a std::string, use
78 Mailbox::getCname() instead to get a char*.
79 - Storage::getName() changed to return a std::string, use
80 Storage::getCname() instead to get a char*.
81 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
82 to have both allocation and deallocation of the map in user space.
85 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
86 The amount of flops remaining to do is a vector, not a scalar.
87 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
88 name implies on both sequential and parallel tasks.
89 - Both changes fix GitHub's #223 using PR #237 as a basis.
90 Thanks Michael Mercier.
91 - Most examples were converted to S4U and hidden elsewhere at tests.
92 Rationale: we still want MSG to work; we want newcomers to use S4U.
95 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
96 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
99 - Switch to the faster dlopen privatization mechanism by default
100 - Documentation improvments
103 - Live migration is getting moved to a plugin. Dirty page tracking is
104 the first part of this plugin. This imply that VM migration is now
105 only possible if one this function is called :
106 - C/MSG: MSG_vm_live_migration_plugin_init()
107 - C/C++: sg_vm_live_migration_plugin_init()
108 - Java: Msg.liveMigrationInit()
109 For C and C++, "simgrid/plugins/live_migration.h" also has to be
113 - Define class simgrid::xbt::Path to manage file names.
114 - Removed unused functions:
115 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
116 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
117 xbt_graph_export_graphviz()
118 - xbt/str.h: xbt_str_join()
119 - Remove unused datatypes:
120 - xbt/heap.h: use std::priority_queue or boost::heap instead
121 - xbt/swag.h: use boost::intrusive::list instead
124 - New link_energy plugin for the consumption of the links.
125 - All of the operations on files and storage contents have been
126 packaged into a plugin (src/plugins/file_system). The current
127 public interface can be found in
128 include/simgrid/plugins/file_system.h
129 To use these functions you now have to initialize the plugin by
130 calling MSG_storage_file_system_init() just after calling
131 MSG_init() or sg_storage_file_system_init() just after creating
135 - Remove the undocumented/untested tag <include>
138 - Remove viva specific tracing as the tool is no longer maintained
141 - #248: Kill of finished processes leads to segfault
142 - #240: xbt_cond_wait_timeout should gracefully return for C apps
143 - #239: please implement signal s4u::onDeadlock()
144 - #230: segfaults when exit() without run()
145 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
146 - #223: MSG_task_get_flops_amount() not working with parallel tasks
147 - #222: Actor::kill() doesn't really kill and segfaults
148 - #221: odd LMM warning when killing an actor
149 - #120: Memory leak when the processes are forcefully killed
151 SimGrid (3.17) Released October 8 2017
153 The Drained Leaks release: (almost) no known leaks despite the tests.
155 Even very long simulations will preserve your memory: our testsuite
156 stresses SimGrid for over 45mn on fast machines for a coverage of
157 over 80%, and there is only one single known leak, of about 4kb.
161 - Comm.detach(): start and forget about asynchronous emission. A cleanup
162 handler may be given to free resources if the comm cannot be completed.
163 - this_actor::send(mailbox) is now mailbox->put()
164 - New: simgrid::s4u::Comm::wait_all()
165 - New: Host.execute() for remote executions.
168 - Unused option network/sender-gap is removed.
171 - Deprecate MSG_task_isend_with_matching(): This unused feature
172 really complicates our internals. Will be removed in v3.20.
175 - Improved context termination. It is now possible to free resources, even
176 when a process is forcibly killed.
177 - ContextBoost: add support for Boost versions above 1.61.
180 - Bring back run-time option --cfg=exception/cutpath to remove exception
182 - Removed unused functions:
183 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
184 xbt_str_rtrim(), xbt_str_trim().
185 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
188 - Removed header files obsolete since SimGrid 3.12:
189 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
191 - #3: SD_exit should be made optional
192 - #120: Memory leak when the processes are forcefully killed
193 - #159: Threading test regression in Actor refcounting
194 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
195 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
196 - #186: Actor::killAll() segfaults if some process is blocked on wait()
197 - #191: VM migration and pstate
198 - #192: Updating the energy consumptions of all hosts crashes with VMs
199 - #195: All actors have PID=0 in the logs
200 - #204: Sometimes segfault with thread contexts and mmap privatization
201 - #222: Actor::kill() doesn't really kill and segfaults
202 - #225: Actor::kill() doesn't really kill when victims are doing a join()
204 ----------------------------------------------------------------------------
206 SimGrid (3.16) Released June 22. 2017.
208 The Blooming Spring Release: developments are budding.
211 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
212 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
213 - Java JAR file should be finally fully working on Mac OSX too.
215 XML platforms: Switch to platform v4.1 format.
216 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
217 - <zone> can be used as a synonym for the now deprecated <as>
218 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
219 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
220 - <actor> can be used as a synonym for the now deprecated <process>
221 - state_file and avail_file periodicity is now easier to express
222 (check the documentation) the old behavior should still work.
223 * Storage: not backward compatible, but it seems that nobody ever used it
224 - Remove attribute 'content_type' of <storage_type>: was never used
225 - Make attribute 'model' of <storage_type> optional: for future usage
226 - Remove Bconnection model property: was never (in)validated. Replaced by
227 the maximum of the read and write bandwidth as a resource constraint.
230 - New and Backwards Compatibility break:
231 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
232 When one wants to get the list of tasks whose states have changed during a
233 simulation round, s/he has to allocate and free a dynar and use it as
234 argument to this function. The former SD_simulate (double how_long)
238 - Allow multicore VMs, with the correct sharing (unless you overcommit)
239 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
242 - New function to update the consumption of all hosts at once.
243 - Fix the model for multi-core hosts, linear in the amount of busy
244 cores with an abnormality for pIdle (see paper).
245 BUG: Still not satisfactory for ptask on mono-cores.
248 - The netzone are now available from the MSG API.
249 The old names still work, but are now deprecated.
252 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
253 - New option: smpi/keep-temps to not cleanup temp files
254 - New option : smpi/shared-malloc-blocksize . Relevant only when global shared
255 mallocs mode is used, allows to change the size of the fake file used
256 (default 1MB), to potentially limit the number of mappings for large runs.
257 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
258 - Fortran ifort and flang compilers support
259 - New RMA calls supported (experimental) :
260 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
261 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
262 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
263 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
264 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
265 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
268 - New: this_actor::isMaestro() returns whether we are in kernel mode.
269 - Behind the scene, ActivityImpl are now automatically refcounted.
270 Removing this nasty bottleneck will greatly simplify our internals.
273 - Replay: New function xbt_replay_action_get():
274 Retrieve the function previously associated to an event type.
275 - DROPPED FUNCTION: xbt_str_varsubst()
276 - DROPPED MODULE: strbuff. We don't need it anymore.
277 - DROPPED MODULE: matrix. We don't need it anymore.
278 - DROPPED MODULE: lib. We don't need it anymore.
280 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
282 SimGrid (3.15) Released March 22. 2017
284 The Spring Release: continuous integration servers become green
285 We fixed even the transient bugs on all target architectures:
286 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
287 Windows, FreeBSD, NetBSD.
289 - Rename NetCards to NetPoints.
290 This was intended to help NS3 users, but that's not a netcard.
291 That's a point in the routing algorithm, let's avoid wrong simplifications.
294 - New: MSG_process_yield(). Stop and yield to other processes.
295 - New: MSG_process_daemon(). Daemon processes are automatically killed
296 when the last non-daemon process terminates
297 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
298 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
299 to make room for the upcoming network energy plugin.
300 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
303 - Ensure that an actor can kill itself with Process::exit()
304 - Kill the obscure NativeException. Nobody want to survive the issues
305 it denotes, so use JniException that is a RuntimeException (not to
306 be caught explicitly).
307 - Partial bug fix in initialization. SimGrid flags on command line were
308 consumed at C level but stayed in the original Java String[] args.
309 This could mess users' args[i] if SG flags were not put at the end of
311 The SimGrid flags are now removed from the Java arguments. However,
312 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
313 if args.length is greater than the number of YOUR OWN ARGUMENTS.
314 It might be if you have --log or --cfg flags in the command line.
315 - Fix numerous memleaks all around the place. In particular, around VMs.
319 - s4u::onPlatformCreated: right before the simulation starts
320 - s4u::onSimulationEnd: right after the main simulation loop
321 - s4u::onTimeAdvance: right after a clock change (time is discrete)
322 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
323 event from the availability_file changes the avail speed.
324 - Links are now usable from s4u
325 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
326 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
327 - New examples: The conversion of MSG examples to S4U has begun
328 - Actors: create, kill, migration, and suspend
329 - Applications: master-worker and token-ring
330 - Action replay: communications and storage
331 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
334 - Backwards Compatibility breaks
335 - The SD_route_get_list and SD_route_get_size functions have been removed.
336 They are replaced by the unique
337 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
338 The route from src to dst is built in the links dynar whose size can
339 be obtained with xbt_dynar_length.
340 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
341 sg_host_route_bandwidth and sg_host_route_latency.
342 Macros ensure the backwards compatibility, but you should fix your code
345 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
346 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
347 - Bump our claim of support from MPI 1.1 to MPI 2.2.
348 We don't support 100% of it, but it should be enough. Contact us if not.
349 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
350 naive and probably inaccurate)
351 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
352 mapped onto a unique small file using some system magic.
353 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
354 - Fortran: cleanups, fixes, support of user-added operations
355 - RMA: MPI_Accumulate are applied in correct order.
356 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
357 - Better support for MPI_IN_PLACE
358 - Support for MPI_Win attrs and keyvals.
359 - Support MPI_Comm_group_create, MPI_Type_size_x
360 - MPI ops should be properly applied to all allowed types
361 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
362 algorithm by a non failing one from mpich.
365 - Kill the fifo data container: we don't use it anymore.
368 - A <cluster> can now be created with different speed values to represent pstates.
369 see examples/platforms/energy_cluster.xml for an example.
371 SimGrid (3.14.159) Released December 28. 2016
373 The previous release was only a crude approximation of a Pi release;
374 this one should be a bit better.
376 - Revert a last minute change that broke on Mac OSX.
377 - Fix the changelog and NEWS files.
378 - Improve the documentation, in particular of the routing module.
379 - Integrate some patches flying around in Debian.
381 SimGrid (3.14) Released December 24. 2016
384 * The whole documentation was reworked and reorganized. There is
385 still much room for improvement here, but we're on it.
388 * We now need python3 for our testing infrastructure.
389 * Model checking needs libevent
392 * Now works on FreeBSD
395 * Privatization now works on FreeBSD
397 * Privatization is now activated by default in smpirun
398 You can turn it off with -no-privatize if something goes wrong.
400 * Call-location tracing for SMPI.
401 You can add the exact location (filename / linenumber) of an MPI call to
402 your trace files and slow down or speed up the simulation between two
403 consecutive calls by using an adjustment file (see the documentation).
405 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
406 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
407 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
408 values were ignored in some cases. The timings of these functions can now
409 be significantly different.
411 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
412 smpi/running-power is renamed to smpi/host-speed
414 * smpi/grow-injected-times option to enable or disable multiplication of the
415 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
416 make simulation less precise (but also much faster).
418 * smpirun script should be (much) faster for large deployments.
420 * SMPI tracing : fixed issue with poor matching of send/receives.
422 * Replay : Fix broken waitall
424 New functions and features
425 * MSG_parallel_task_execute_with_timeout, to timeout computations.
427 Dropped / renamed functions and features
428 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
429 - MSG_mailbox_is_empty() -> Mailbox::empty()
430 - MSG_mailbox_front() -> Mailbox::front()
431 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
432 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
433 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
434 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
435 * Don't pass the free_f parameter to property related functions:
436 - MSG_host_set_property_value()
437 - MSG_as_router_set_property_value()
438 - MSG_storage_set_property_value()
439 * VM properties. Since msg_vm_t are msg_host_t, just use
440 MSG_host_get_property_value() and friends
441 * VM I/O related things:
442 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
443 - Unimplemented save/restore methods
444 * MSG_as_router_get_property_value() was redundent with
445 MSG_environment_as_get_property_value().
446 - Removed MSG_as_router_*propert*() functions
447 - Added MSG_environment_as_set_property_value() for consistency
448 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
449 Well, they are still there for now, but deprecated with a warning.
450 Please switch to xbt_dict_new_homogeneous() before this is removed
452 * Task affinity. Its intended behavior (that was very badly tested
453 and probably not really working) was deceiving what most users
454 would have hoped here.
455 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
456 * Option network/coordinates is now useless and should be dropped.
459 * Added option storage/max_file_descriptors to allow more than 1024 files opened
462 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
463 * Added Dragonfly topology support
466 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
467 EiBps, EBps, Eibps, Ebps for bandwidth.
468 They may become useful to some lucky ones.
471 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
472 * ex/app_tokenring: new example, very similar to the MSG Token Ring
473 * ex/async_waitAll: new example, on asynchronous communications
476 * Memory usage should be decreased for simulations with a large number
477 of processes. This also helps for SMPI.
479 SimGrid (3.13) stable; urgency=low
481 The Half Release, a.k.a. the Zealous Easter Trim.
483 - We removed half of the lines, that were mostly experimental cruft.
484 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
485 (not counting blanks and comments -- according to openhub.net)
486 - The internals are now compiled in C++ (and will soon be clean C++)
487 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
488 59 klines of C, etc. We added only 29 klines of C++ in replacement.
490 * Backwards Compatibility breaks
491 - Removed Lua simulation bindings (switch to C or Java for that).
492 Lua can still be used to describe platforms
493 - Removed Java kernel plug-ins.
494 Will be reintroduced after the ongoing major internals reorg.
496 - the following functions were removed.
497 They were too specific and should be reimplemented in a generic
498 way, with filter function.
499 - MSG_task_listen_from_host
500 - MSG_mailbox_get_count_host_waiting_tasks
501 - MSG_mailbox_put_with_timeout was removed.
502 Please use MSG_task_send_with_timeout instead.
504 - the SD_application_reinit function was removed. It has been a noop for a while.
505 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
506 replaced by a more flexible mechanism.
507 - The following functions thus do not exist anymore
508 - SD_workstation_get_access_mode
509 - SD_workstation_set_access_mode
510 - SD_workstation_get_current_task
511 - Basic estimation functions have been removed but can easily be replaced
512 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
513 - SD_workstation_get_computation_time => amount / sg_host_speed()
515 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
516 Use VM.getSpeed()*load/100 for the legacy behavior.
518 - option enable_tracing was removed. It was not doing anything for a while.
519 - In the ModelChecker:
520 - the model-checker now ptraces the model-checked process which means
521 you cannot use a debugger on the latter anymore (we might make this
522 optional in the feature);
523 - removed soft-dirty page tracking;
524 - remove model-checked side snapshot management,
525 MC_snapshot() and MC_compare_snapshot();
526 - keep the MC_cut() function as a stub (it was not really working
527 in the previous release).
530 * All options are consistently in kebab-case. Old names are kept as alias.
533 * Switch to platform v4 format.
534 - Rename from 'power' to 'speed' the attributes describing the amount of
535 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
536 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
537 - In <host> and <link>, attributes availability and state are gone.
538 It was redundent with state and availability traces, and with peak values.
539 - In <cluster>, attributes availability_file and state_file are gone.
540 It was too complex and unused.
541 - Kill <gpu>. Was not doing anything.
542 - The DOCTYPE points to the right URL:
543 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
544 (the file at this address now documents the changelog since its v1)
545 - A warning is emitted for unit-less values (they are still accepted).
546 - speed. Default: 'f' or 'flops'. Also defined:
547 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
548 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
549 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
550 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
551 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
552 - latency. Default: 's' second. Also defined:
553 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
555 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
556 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
559 * s4u::Host is now the preferred public interface to the Host features.
560 sg_host_* functions are C bindings to the exact same behavior
561 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
564 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
565 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
566 They were deprecated since a few years, and probably did not even compile anymore.
569 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
570 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
571 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
572 encouraged to update their codes. A list of the modified functions can be found at the end of
573 include/simgrid/simdag.h
576 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
577 the lambda or closure passed as a parameter will run in kernel mode.
578 All the callback functions should be rewritten to that interface at some point.
581 * Reorganizing and cleaning the internals all around the place.
584 * Remove old default barrier/bcast buggy algorithms (see #18407)
585 * Various bug fixes to handle more codes
586 * Remove the need for the --foreground option of smpirun (it is still
587 accepted for backward compatibility).
590 * Kill the setset data container: MC don't use it anymore.
591 * Kill the queue data container: it made more sense with GRAS.
592 * Kill the xbt_peer_t data type: it's useless without GRAS.
593 * Kill rm_cb feature of config sets: it was never useful.
594 * Kill graphxml parsing feature. It was not used.
595 * Kill the deprecated code protected by XBT_USE_DEPRECATED
597 - xbt_dynar_sort_strings(), when the content is char*
598 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
599 They throw exceptions on invalid input;
600 * C++ support for declaring CLI flags (simgrid::config::Flag);
601 * class for abstracting different signal backends (simgrid::xbt::signal).
602 with no external dependencies (we need very simple signals).
605 * refactoring and cleanup of the code;
606 * ongoing process to cleanly separate the model-checking algorithms
607 from the code model-checking support.
609 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
611 SimGrid (3.12) stable; urgency=low
613 The Facelift Release.
616 * Require g++ v4.7 at least to not speak prehistorical C++.
617 * Require Boost 1.48 (for signal2 component).
618 * Java must be version 7 at least when activated.
619 * Builds on Windows again (including Java bindings).
620 * Tracing is now always enabled (no way to turn it out)
621 * Remove GTNetS. It was not working anyway.
622 * Various cleanups in the cmake scripts.
623 * Move headers around to sort them out on installed systems:
624 - instr/instr.h -> simgrid/instr.h
625 - instr/jedule/* -> simgrid/jedule
626 - simdag/datatypes.h was removed
627 - simdag/simdag.h -> simgrid/simdag.h
628 - msg/datatypes.h was removed
629 - msg/msg.h -> simgrid/msg.h
632 * Interface improvement:
633 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
634 - Sanitize the interface in MSG_task_ module:
635 - Merge two functions that were close enough but misleading:
636 set_compute_duration(t) -> set_flops_amount(t)
637 get_remaining_computation(t) -> get_flops_amount(t)
638 - set_data_size(t) -> set_bytes_amount(t)
639 get_data_size(t) -> get_bytes_amount(t)
640 - Massive cleanups in the functions related to the energy
641 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
642 - New: MSG_host_get_pstate()
644 - msg/energy/onoff: switching hosts on and off
646 * Interface improvement:
647 - Rename Host.isAvail() to Host.isOn()
648 - Rename Process.currentProcess() to Process.getCurrentProcess()
649 - Rename Task.setDataSize() to Task.setBytesAmount()
650 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
652 - #18874: Actually allows the GC to reclaim tasks
656 - SIMIX_process_throw: raises an exception in a remote process
657 * Refactoring: Separate sub-modules
658 - libsmx: the public interface, as libc in a real system
659 - popping: the strange dance that converts a user request into a kernel handling
660 - smx_context_*: the virtualization mechanisms that embed the user code
661 - smx_*: the handling of each simcalls
663 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
664 * Rename smx_action_t into smx_synchro_t, making explicit that these
665 things are used to synchronize processes with their environment.
666 For example, a communication is a sort of synchronization involving
667 the communicating processes (that may block until the exchange) and
668 the platform. The same can be said from computations, etc.
670 - #18888: segfault when a process ends before its kill_time
674 - 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.
675 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
676 - MPI_Type_set_name, MPI_Type_get_name
677 - MPI_*_c2f and MPI_*_f2c functions
678 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
679 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
680 - Activate a lot of new tests from the mpich 3 testsuite
682 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
683 - InfiniBand network model added : Based on the works of Jerome Vienne
684 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
685 - When smpi/display_timing is set, also display global simulation time and application times
686 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
687 * Collective communications
688 - 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
689 - 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/).
690 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
692 - Replay now uses algorithms from wanted collective selector
693 - Replay can be used with SMP-aware algorithms
694 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
695 - 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.
697 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
698 - Correctly use loopback on fat-tree clusters
699 - Asynchronous small messages shouldn't trigger deadlocks anymore
700 * Energy/DVFS cleanup and improvement
701 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
702 - new: smpi_get_host_pstate()
705 - "Full" network optimization flag was broken since Surf++
706 - Better handling of precision flags in maxmin
707 - Fix bug causing sometimes "Impossible" errors
708 - Properly pass cluster properties to included hosts
709 * Improvement of the Energy plugin.
710 - Always update the consumption before returning that value
711 - New property: watt_off to denote the dissipation when the host is off
712 - New functions getWattMinAt and getWattMaxAt to retrieve the
713 dissipation of pstates that we are not currently at.
714 * Java: class NetworkLink renamed to Link
715 * New function: simcall_process_get_kill_time()
716 * Massive rename s/workstation/host/
717 - That's intrusive, but that's good for the project consistency. Sorry.
718 - Change config option "workstation/model" into "host/model"
721 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
722 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
723 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
724 - xbt_dirname and xbt_basename for non-POSIX systems
726 * The model checker now runs as a separate process.
727 * The model checker runs must now be launched with the new simgrid-mc program.
728 * Record/Replay: the MC can display a textual representation of a path in the
729 execution graph. It can then be replayed outside of the model checker.
731 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
733 SimGrid (3.11) stable; urgency=low
738 * Normalizing pointers addresses tool for better diff between logs
741 * Add cloud examples using new VMs
742 - examples/msg/cloud/two_tasks_vm.tesh
743 - examples/msg/cloud/simple_vm.tesh
744 - examples/java/cloud/cloud.tesh
745 - examples/java/cloud/migration/migration.tesh
746 * Add java surf examples:
747 - examples/java/surfPlugin/surf_plugin.tesh
748 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
749 - examples/java/surfCpuModel/surf_cpu_model.tesh
750 * Add SMPI+MSG example:
751 - examples/smpi/smpi_msg_masterslave/
756 - msg task destroy cancel test
757 - msg_host on/off test
758 * Move all tests in testsuite to teshsuite (adding tesh files)
759 * Restructure teshsuites
760 - one folder for each kind of test
761 * Restructure AddTests.cmake
763 - structure the order of tests (with sections)
766 * Add virtual machine
767 - creation of a VM on a PM
768 - migration of a VM from a PM to another PM
770 - MSG_process_join(msg_process_t process, double timeout)
771 - msg_bar_t MSG_barrier_init(unsigned int count)
772 - int MSG_barrier_wait(msg_bar_t barrier)
773 - void MSG_barrier_destroy(msg_bar_t barrier)
774 - msg_as_t MSG_environment_as_get_by_name(const char * name)
775 * New option "msg/debug_multiple_use" to help debugging when a task is used
778 - Improvements and finalization of MSG_storage, MSG_file APIs and their
780 - Increase code coverage in test suites
784 * Protect context stacks against stack overflow. The number of protected memory
785 pages allocated on the top of each stack (1 by default) can be configured
786 with the new command line option --cfg=contexts/guard_size:<value>.
787 * Simcalls are now generated by a python script that
788 - generates files included by SimGrid
789 - checks that all the functions exist, and proposes prototypes
791 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
792 host_set_data, host_get_data
794 - simcall_process_join(smx_process_t process, double timeout)
795 * Fix bug where sleeping processing could not be suspended.
798 * Translate surf models from C to C++
799 - Generic classes for all models: Model, Resource, Action
800 - A generic interface for each kind of model (CPU, Network, Storage
801 Workstation, WorkstationVM)
803 * Translate surf routings from C to C++
804 * Add callbacks using sigc++ or boost::signals2
805 - Add callback functions for resource creation/destruction
806 - Add callback functions for action state change
807 - Handle Energy as a plugin
808 * Replace swag by boost::intrusive
809 * Add new routing models for clusters. For documentation, see
810 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
811 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
812 parameters for cluster tag
813 - Fat trees, with topology="FAT_TREE" and
814 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
815 - see examples/platforms/torus_cluster.xml and
816 examples/platforms/fat_tree_cluster.xml
820 * Hostfiles support host:nb_processes construct to deploy several processes on
822 * Collective communication algorithms should not crash if used with
823 improper number of nodes and report the error.
824 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
825 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
826 MPI_Dims_create, MPI_Cart_sub are supported.
827 * New interface to use SMPI programmatically (still depends on MSG for
828 some parts, see examples/smpi/smpi_msg_masterslave) :
829 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
833 * Global variables privatization in MPI executables is now performed at runtime
834 with the option smpi/privatize_global_variables (default:no).
835 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
836 dynamic libraries loaded by the application are not privatized (static
837 linking with these libraries is advised in this case)
840 * Options defined in XML work correctly now.
843 * New cmake option, enable_lib_in_jar, to control whether native libraries are
844 copied into simgrid.jar or not (ON by default). Use this option if you want
845 to reduce the size of the installed simgrid.jar, *and* the native libraries
846 are kept installed elsewhere.
847 * Surf binding with SWIG (code generated in maintainer mode only):
848 - plugin to handle callbacks
849 - CPU model only for the moment
852 * Supernovae build mode is definitively removed. It was used to improve
853 inlining and inter-module optimizations. It is nowadays superseded by
854 link-time optimizations commonly available in compilers.
855 * Update ns-3 find lib. Bindings for ns-3 should work again now.
856 * Add boost dependency for surf++
857 * Add new macro for tests
858 - ADD_TESH(name <tesh_args>)
859 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
863 - xbt_bar_t XBT_barrier_init(unsigned int count)
864 - int XBT_barrier_wait(xbt_bar_t barrier)
865 - void XBT_barrier_destroy(xbt_bar_t barrier)
866 * Make the xbt_os_time module public
868 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
870 SimGrid (3.10) stable; urgency=low
872 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
875 * Reintegrate Java to the main archive as desynchronizing these
876 package is not acceptable anymore (Java is now considered stable)
877 * Add explicit synchronization facilities through semaphores
878 * Bug fix: Task.setDataSize() only changed the C world, not the value
879 cached in the Java world
882 * Dramatically change the way files are handled. API and internals changed, but
883 this part of MSG was not considered as production grade either.
884 * Add explicit synchronization facilities through semaphores
885 * Add a new function MSG_host_get_process_list()
886 * Preliminary DVFS support (see examples/msg/energy/ for details)
889 * SMPI is now included directly in the libsimgrid as the windows
890 linker doesn't force us on splitting it anymore.
891 * Improvements of the SMPI replay tool:
892 - Most of the collective communications are now rooted in the same process as
893 in the original application.
894 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
895 used until now). Multiple data types can now be used in a trace.
896 - The replay tool now supports traces produce either by TAU or a modified
898 - Bug Fix: the compute part of the reduce action is now taken into account.
899 - Gatherv collective is now supported
900 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
901 of smpirun outputs time independent traces for the current run. One file
902 is created per process. If too many processes are simulated, this behavior
903 can be changed to one file for all processes by using the
904 tracing/smpi/format/ti_one_file flag
905 * smpirun generates the host file if needed (with given host count and platform)
906 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
907 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
908 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
909 --cfg=smpi/coll_selector:(mpich/ompi)
910 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
912 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
913 happening inside a collective SMPI call.
914 * Fix the behavior of complex data types handling.
915 * Make MPI_Wtime another synchronization point to take computations into
917 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
918 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
919 * Add all missing Fortran bindings, SMPI should work with Fortran 90
920 (no privatization of global variables yet)
921 * Preliminary DVFS support (see examples/smpi/energy/ for details)
924 * Verification of liveness properties is now available for SMPI applications
925 (in addition to MSG applications)
926 * Bugged examples using SMPI in examples/smpi/mc/
927 * Add --cfg=model-check/visited option. Allows the verification of infinite
928 programs. Detection of loops in the execution thanks to the system state
929 comparison and reduction of the state space to explore. Can be combined with
930 DPOR for safety properties.
933 * Allow to change SimGrid configuration (see --help) within the code
934 thanks to SD_config() as it can be done in MSG.
935 * Add a new function SD_task_set_amount() upon user request.
938 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
939 * Remove rule based routing (no more PCRE dependency)
940 * Add a limiter_link option to cluster tag, to specify a maximum reachable
941 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
942 * Add a loopback_bw and loopback_lat options to cluster tag.
943 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
944 routing. Now peers are encapsulated in an AS and have their own private
945 router but this is transparent.
948 * Our own implementation of getline is renamed xbt_getline, and gets
949 used even if the OS provide a getline(). This should reduce the
950 configuration complexity by using the same code on all platforms.
951 * New type: xbt_cfg_elm_boolean.
952 * Allow to use yes/no for boolean configuration options in the command line.
953 * Allow to disable SimGrid cleanups at exit from command line option.
954 There are situations where one may want a simulation to end with an exit.
955 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
956 annoying when scripting around the simulator. Adding a
957 --cfg=clean_atexit:no allows to circumvent this issue.
960 * Lots of memory leaks were corrected in this release.
961 There is no known memory leaks anymore, in all of our 600+ tests.
962 * New command line option --version, to get SimGrid version information.
963 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
965 * Supernovae builds are deprecated, and expected to be removed in the next
968 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
970 SimGrid (3.9) stable; urgency=low
972 The Grasgory release: GRAS is really dead now.
974 * Complete overhaul of the internal host structures scheme.
977 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
978 was considered as experimental and badly maintained since then.
979 * Keeping it was thus a trap to our potential users, that could take
980 it instead of MSG or SMPI by mistake despite is pity state.
981 * GRAS seems to have very few users (if any), and no one volunteered
982 to maintain it further. It also induces a lot of XBT code (for
983 portability sake), that must be maintained too.
984 * For all these reasons, we killed GRAS. If someone wants to revive it
985 in the future, don't cry, our git history still remembers of GRAS.
988 * Major overhaul. Merge our documentation again as time proved that
989 splitting it was really not helping our users.
990 * Further improve the developer documentation to help newcomers
991 hacking on SimGrid itself. The user documentation (and in
992 particular, the beginner documentation) is still in a sorry state.
995 * Now works on Windows too!
996 * Much more extensive test suite, from MPICH
999 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1000 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1001 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1002 * Bug fix: let task be scheduled when the last dependency to be solved is
1003 a control dependency.
1004 * Remove SD_load_environment_script function.
1005 Use the C sg_platf function if you want to declare a platform
1009 * New function: MSG_process_get_number()
1010 * Old function documented: MSG_config()
1011 * Remove MSG_load_platform_script function
1012 Use the C sg_platf function if you want to declare a platform
1016 * Change the default value of the TCP_gamma constant (maximal size of TCP
1017 congestion window) to a more realistic 4MiB value. If you notice changes in
1018 your simulation results, you can fall back to the previous 20k tiny window
1019 by adding --cfg=network/TCP_gamma:20000 on command line.
1020 * (Hopefully) fix a bug wrt periodic availability and state traces
1021 * Bug fix: use default values at start when first event in availability/state
1022 trace is not at time 0.
1025 * remove the "new_" part of function name sg_platf_new_trace_connect
1026 (resulting in sg_platf_trace_connect), since it does not create
1030 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1031 use case where it's really mandatory, and maintaining it was a pain in
1033 * New: xbt_fifo_search(), search an item with a user-provided
1034 comparison function instead of dumb pointer comparison.
1037 * Fix the lua deployment:
1038 Use `simgrid.init_application()` before deployment instead of
1039 `simgrid.msg_register_application()` after.
1042 * Transfer the tracing files into the corresponding modules.
1044 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1046 SimGrid (3.8.1) stable; urgency=low
1048 The "we are told that some people want to also *install* the simgrid
1051 * Add missing file "tesh.1" to the archive.
1053 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1055 SimGrid (3.8) stable; urgency=low
1057 The Psssshiiiit release: SimGrid jumps into the Cloud.
1060 * Add an experimental interface to manipulate VMs. They are mainly
1061 process groups with very few intrinsic semantic, but they should
1062 allow you to build the semantic you want easily.
1063 * New function: MSG_host_set_property_value()
1064 * New function: MSG_process_on_exit(). To clean memory in all cases.
1065 * Bug fixes that made the host (and link) failures unusable.
1066 * Add a way to auto-restart process when the host in which they are
1067 executing comes back (ON_FAILURE="RESTART" on deployment file,
1068 MSG_process_auto_restart_set).
1069 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1070 please stop using the old ones, they are DEPRECATED.
1072 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1073 Please use MSG_init() instead. (reducing the amount of entry
1074 points in the library helps us).
1075 * Make it impossible to link against the wrong version of the lib
1076 * Deprecate MSG_clean(). No need to call it anymore.
1077 * Function MSG_get_host_number() is not deprecated anymore.
1080 * Split the doc into a user guide and a reference guide.
1081 * Start a developper guide to help people hacking on SimGrid.
1084 * Enable tracing by default. This modules rocks you should use it.
1085 * Remove option custom_flags. Now use environment variables CFLAGS
1087 * Use default cmake things to detect lua instead of home grown ones.
1088 * New option "enable_mallocators" to disable mallocators, for debugging
1089 purpose ("on" by default).
1092 * Bug fixes around the resource failures: don't let the processes
1093 survive the host they are running onto.
1094 * Add an interface to auto-restart processes when the host in which they are
1095 executing comes back.
1096 * Ensures that SIMIX_clean is called automatically. It's not part of
1097 the public interface anymore (bindings should be updated).
1100 * Bug fix for when SD_Simulate is called with a positive value: be careful
1101 when comparing doubles. Sometimes they are different for non significant
1103 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1104 parallel task whose initial work is distributed among host according
1105 to the Amdahl's law. Such tasks are created with a parameter alpha
1106 that corresponds to the non-parallelizable part of the computation.
1107 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1108 between two sets of workstations assuming a 1D block distribution (each
1109 workstation owns a similar share of data) on both sides.
1111 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1112 Data redistribution will be automatically scheduled once parent and child
1113 are both scheduled. The filling of computation_amount and
1114 communication_amount structures is now done seamlessly thanks to the chosen
1116 * New function SD_workstation_dump to display various information
1117 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1118 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1119 bandwidth on the route onto which the task is scheduled and the amount of
1121 To divide the nominal bandwidth by 2, the rate then has to be :
1122 rate = bandwidth/(2*amount)
1123 * Compute tasks that have failed can now be rescheduled and executed again
1124 (from their beginning)
1125 * Increasing source code coverage (src/simdag is now covered at 95.8%
1129 * Re-implement time-independent trace replay using SMPI (at the
1130 smpi_smp_* level) instead of MSG. This should replace
1131 examples/msg/actions/actions.c
1132 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1133 hindexed and structs)
1134 * Implement the exchange of non-contiguous data.
1135 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1136 * Correct behavior of smpi/sender_gap and set its default value to 0
1137 * Add option to asynchronously send small messages to allow better
1138 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1139 specifies the size in bytes under which messages will be asynchronously sent.
1140 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1141 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1144 * New C interface to define a platform: XML is now optional.
1145 For more info, please check include/simgrid/platf.h.
1146 * New interface to define random platforms from the C:
1147 For more info, please check include/simgrid/platf_generator.h and
1148 examples/msg/masterslave/masterslave_platfgen.c
1149 * Export a sg_cmdline dynar containing all the arguments we got from
1153 * Two new tracing options for adding comments to trace file so you
1154 can track your experiments (see --help-tracing for details).
1155 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1156 * Adding the SimGrid version that generated the trace file as a comment.
1157 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1158 * Fix to avoid key clashes on Paje links
1159 * Other minor fixes related to the Paje specification
1162 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1163 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1164 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1165 * Greatly improve the robustness of mmalloc to user errors (such as
1166 using an area after freeing it, or freeing it twice)
1168 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1170 SimGrid-java (3.8.1) stable; urgency=low
1172 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1173 Warning: all methods are visible, but only some of them are
1174 implemented so far. Check the source (src/jtrace.c)
1175 for further information.
1176 * New module: org.simgrid.msg.File (SimGrid File management functions)
1177 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1178 * Change the meaning of Process.restart: now restart the process from
1179 the begining, like MSG_process_restart in C.
1180 * Add Process.setAutoRestart: handling of process restart when failed
1182 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1183 you to retrieve the properties of the processes/hosts
1184 * Deprecate Msg.clean(): you can just forget about it now.
1185 * New function Process.getCount(), that only works when compiling
1186 with the not yet released version 3.9 of the C library.
1188 * New context factory based on Coroutines. It mandates a modified JVM
1189 but then, the simulations run about five times faster, and there is
1190 no limit to the amount of processes (beside of the available memory).
1192 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1194 SimGrid (3.7.1) stable; urgency=low
1197 * Restore the prototype of MSG_process_create_with_environment() to
1198 the pre-3.7 situation by removing the kill_time argument.
1199 * Add a MSG_process_set_kill_time() function instead.
1202 * Fix weird behaviors when dealing with parallel tasks.
1205 * Simgrid is now built as a dll.
1206 * Simgrid-java now works on Windows.
1207 * Simgrid-Java is now included into Windows package.
1210 * First pre-build package for MacOSX.
1213 * Fix compilation when using MSG_USE_DEPRECATED.
1214 * Fix some compilation issues on Macs and Windows.
1215 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1216 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1218 We discovered that the Lua console is broken, but we are missing the
1219 manpower to fix it right now. The problem existed in 3.7 too, so we
1220 are not blocking the release for that. Sorry if you depended on this
1221 feature, any help would be really welcome.
1223 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1226 SimGrid-java (3.7.1) stable; urgency=low
1228 The "Java aint got to be bloated and slow" release
1231 * Various internal cleanups and performance improvement
1232 Simulations are expected to run up to twice faster or so
1233 * Make Process.kill(process) an instance method, not a static one
1234 * User processes are not java.lang.Thread subclasses.
1235 This breaks the compatibility (sorry), but previous API was
1236 brain-dead, making it impossible to have non-trivial
1237 initializations in the process constructor.
1238 * Require a full constructor per Process sub-class.
1239 Kinda breaks the compatibility (sorry), but this allows a much more
1240 efficient way to launch the processes at simulation startup.
1241 * Do not embeed our version of semaphores, java 1.5 can be considered
1242 as sufficiently prevalent for us to not dupplicate its features.
1246 * Add examples for almost every part of the API
1247 We spotted and fixed a lot of bugs in the process
1249 * New module: asynchronous communication API
1250 * New function: Process.sleep()
1251 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1252 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1253 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1255 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1257 SimGrid (3.7) stable; urgency=low
1259 The "spring cleanups (before the next Big Project kicks in)" release.
1262 * We can specify the SMPI latency/bandwidth factor with command line
1263 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1264 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1265 You can also use the "config tag" from platform file by adding this line
1266 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1267 (see "example/platforms/tag_config.xml" to use "config tag").
1268 Note that the command line supersedes the platform file configuration.
1269 * Change the correction factors used in LMM model, according to
1270 the latest experiments described in INRIA RR-7821.
1271 Accuracy should be improved this way.
1272 * Use the partial invalidation optimization by default for the
1273 network too. Should produce the exact same results, only faster.
1274 * Major cleanup in surf to merge models and split some optimization
1275 mechanisms from the core of the models. As a result you can now
1276 specify which model to use (e.g., --cfg=network/model:LV08
1277 --cfg=cpu/model:Cas01) and which optimization mode to use
1278 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1279 Incompatible combinations should err at initialization. See
1280 --help-models for the list of all models and optimization modes.
1281 * The CLM03 workstation model was dropped for simplicity because it
1282 used the deprecated CM02 network model. Use default instead.
1283 * Rename the TCP_gamma configuration option to network/TCP_gamma
1284 * Rename the coordinates configuration option to
1285 network/coordinates, and document it
1286 * Use now crosstraffic keyword instead of the terribly misleading
1287 fullduplex keyword. It is activated by default now in the current
1288 default model, use --cfg=network/crosstraffic:0 to turn it off.
1289 * Ongoing refactoring the model parsing to make XML files optional
1290 See include/simgrid/platf.h for details (still to be completed)
1293 * Major overhaul of the documentation. Almost instructive now :/
1294 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1295 functions and friends. This interface was considered as
1296 deprecated since over 2 years, it's time to inform our users that it is.
1297 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1298 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1299 use these (crappy) functions in your code.
1300 These functions will be removed soon. Stop using them now.
1301 * Deprecate MSG_get_host_{table,number}
1302 Implement MSG_hosts_as_dynar() instead.
1303 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1304 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1307 * Stabilize the parallel execution mode of user contexts
1308 * Introduce configuration variables to control parallel execution:
1309 - contexts/synchro: Synchronization mode to use when running
1310 contexts in parallel (either futex, posix or busy_wait)
1311 - contexts/parallel_threshold: Minimal number of user contexts
1312 that must be part of a scheduling round to switch to parallel
1313 execution mode (raw contexts only)
1314 * Fix bugs that prevented to use suspend/resume along with
1315 synchronization structures.
1316 * Fix bugs in process termination that lead to invalid memory access
1317 in very specific conditions.
1320 * Introduce a parallel mode for the models (controlled by surf/nthreads
1321 configuration item). In our tests, running the models in parallel
1322 never lead to any speedups because they are so fast that the gain
1323 of computing each model in parallel does not amortizes the
1324 synchronization costs, even when ultra fast futexes are used.
1325 This is released anyway because YMMV.
1328 * Performance boost by using a swag internally to compute the set of
1329 tasks that are finished and should constitute the return value of
1333 * Enable it by default now that it is considered rather stable.
1336 * Documentation of the tracing functions.
1337 * Performance gains when tracing categorized/uncategorized resource
1338 utilization by avoiding calls to get route when updating resource
1339 variables. LMM constraints are being used instead.
1340 * API changed to set task categories. Use MSG_task_set_category instead
1341 of TRACE_msg_set_task_category, and SD_task_set_category instead
1342 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1343 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1344 * Examples for link user variables added (see at examples/msg/tracing/)
1345 * Deprecated function TRACE_msg_set_process_category completely removed
1346 * Trace header updated according to the latest Paje file format
1347 * Tracing network lazy updates, no longer obligate users to use full updates
1348 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1349 * Experimental: let user code declare/set/push/pop application states for hosts
1350 * API changed to allow the manual creation of graph configuration files
1351 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1354 * Improve the API of Lua MSG bindings, using the Lua spirit.
1355 * Each simulated process now lives in its own Lua world (globals are
1356 automatically duplicated). It helps writing simulators. Will allow
1357 to run Splay programs within SimGrid in the future.
1358 * Add a Chord example in Lua, equivalent to the MSG one.
1361 * Start the implementation of a solution to express temporal
1362 properties, not only local assertions. This is still an
1363 experimental work in progress, stay clear from it to be safe.
1367 - Add new runtime parameters --help-logs and --help-log-categories
1368 to display information about supported logging parameters and
1370 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1371 don't exists anymore.
1372 * Mallocators: allow value NULL for the reset function.
1374 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1375 create homogeneous dictionaries, where all the elements share the
1376 same free function. Non homogeneous dictionaries will be
1377 deprecated in the next release.
1378 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1379 - Multi-level dictionaries are deprecated.
1381 - new function xbt_dynar_search_or_negative() that is useful when
1382 you have less than 2 million elements in your dynar and don't
1383 want of the extra complexity of catching exceptions when the
1384 element is not found.
1386 - Make xbt_os_thread module (for thread portability) public.
1387 Documentation is still to come, sorry.
1389 - Cleanups and simplifications to make it maintainable again.
1390 - Exotic features (such as memalign and valloc) were removed.
1391 - The metadata were extended and improved so that the
1392 model-checker becomes able to explore and inspect the heaps.
1393 - This may induce a performance drop when enable_model-checking is
1394 ON in cmake (even if it's not used in the simulation), but it is
1395 necessary at this point to get MC working.
1397 Turn model-checking OFF if simulation performance matters to you.
1398 Not enabling it at runtime is not enough, disable it in cmake.
1400 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1403 SimGrid (3.6.2) stable; urgency=low
1405 The "Not coding new stuff allows to polish old things" release.
1408 * New bindings to the NS3 packet level simulator (experimental)
1409 * Use the raw (efficient) execution contexts instead of the sysv
1410 (portable) ones when possible.
1411 * libpcre is now mandatory in any cases since not using it led to
1412 severe performance loss and possibly other issues
1413 * Update the XML platforms:
1414 - G5K: include the latest machine in Nancy
1415 - GridPP and LCG: new platforms
1416 * Documentation was partially updated, at least (more to come)
1418 Bug fixes, cosmetics and small improvements
1419 * Free terminated processes before the end of the simulation to avoid
1420 exhausting the memory of users having very dynamic amount of
1422 * Bug fix and cosmetics about canceling non-running tasks
1423 * Bug fix about the dot loader's issues when using libcgraph
1426 * Create an installer for windows with nsis (amd64 and win32)
1427 - Add an hello world project to illustrate simgrid project creation.
1428 - Embed libpcre into the Simgrid installer to avoid
1429 its compilation burden
1430 * The raw execution contexts should work on Apple now
1431 * Port to Windows 64 bits
1432 - Sysv contexts now have an implementation for this arch
1433 - GRAS communication features now support this arch
1434 * Drop support for borland compiler on windows
1435 - this code was not maintained, and we kinda depend on gcc nowadays
1436 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1437 * Fix portability issue on unstable ubuntu: linker became picky on
1440 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1442 SimGrid (3.6.1) stable; urgency=low
1444 The "Oops, we broke Macs too" release
1447 * Fixed contexts detection so that raw ones are used when possible
1448 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1449 leads to a strange error, with user code segfaulting sometimes when
1450 the generated code is not perfectly aligned (which is not
1451 controllable from the user side, depends on the amount of code)
1454 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1456 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1458 SimGrid-java (3.6) unstable; urgency=low
1461 * Split of every thing from simgrid v3.5 into a separate package.
1463 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1465 SimGrid (3.6) stable; urgency=medium
1467 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1470 * Bindings now constitute their own package, separated from the main one.
1471 Rationale: reduce our maintenance nightmare by reducing the module coupling
1472 They will soon be released on their own on gforge.
1474 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1475 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1477 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1478 * It's not quite deprecated for now because we have no replacement,
1479 but it may soon become the case.
1482 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1483 * Fortran: New user-level cache variable to store the rank of the running
1484 process. This improves performance by an order of magnitude.
1485 * C: New coccinelle script to automatically locate and modify global and
1486 local static variables.
1487 * Improved SMPI network model with a sender-side gap to account for multiple
1491 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1492 only say if a communication is finished, no matter whether it succeeded or
1493 failed. You can call MSG_comm_get_status() to know the status of a finished
1495 * New function MSG_task_dsend() to send a task and detach it. When a
1496 communication is detached, you are never notified of its success or failure
1497 and the memory is released automatically once it is finished. This function
1498 is useful when you don't care about the end nor the success of a
1500 * Change the prototypes of action replay. Sorry for inconvenience,
1501 but this is really more efficient this way (and to adapt your code,
1502 you just have to fix the initialization, that shouldn't be too long)
1503 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1504 ever managed to do anything useful with it.
1505 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1506 takes a dynar of communications. It returns immediately and gives the
1507 index of a finished communication (if any).
1508 * New example: a basic implementation of the Chord P2P algorithm.
1511 * New model for multi-core CPUs. You can now use the core attribute to
1512 precise the number of cores of a host. This is a basic model. Every
1513 process running on the host receives at most the power provided in
1514 the DTD (throughput<=power). Total throughput of process cannot exceed
1516 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1517 router linked by an up-link and a down-link (possibly asymmetrical).
1518 This kind of pattern allows to easily build last-mile model style platforms.
1519 Aggregating such patterns in a rule-based AS is thus the technique of
1520 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1521 * New model for Vivaldi routing. We transformed the Vivaldi network model
1522 into a Vivaldi routing model (based on the rule-based model). This allows to
1523 combine Vivaldi based latencies with last-mile platforms.
1526 * Added a check for NaN of IEEE754 infinite in the double entries of
1528 * Introduce a new context factory "raw", highly inspirated from the
1529 ucontext factory, but using manually crafted functions in assembly to
1530 do the work in an efficient manner.
1531 * Allow to change the used context factory at run time, not only at
1532 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1533 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1534 contexts at runtime (only with raw contexts or ucontexts).
1535 * Completely rewrote this module to allow parallel execution of user
1536 processes. Use --cfg=contexts/nthreads:N to execute user processes
1537 with N parallel threads (the default is 1, meaning no parallelism).
1538 * Allow to decide dynamically between sequential and parallel modes.
1539 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1540 processes in parallel only when their number is greater than or equal to P
1542 * Added a check for NaN of IEEE754 infinite in the double entries of
1546 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1547 won't output the state of processes when interrupted with Ctrl-C
1548 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1549 NULL-terminated array. This may solve backward compatibility issues
1550 due to the change to return type of SD_simulate. See also:
1551 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1552 * Add new macros with variable number of arguments.
1553 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1554 - in xbt/asserts.h: xbt_assert
1555 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1556 - in xbt/ex.h: THROWF and RETHROWF.
1557 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1559 * Change xbt_die() to accept a format string with arguments, just like printf.
1560 * New data structure: xbt_lib_t, like a dict but more general and with better
1564 * New configuration options
1565 Options triva/categorized and triva/uncategorized can be used to generate
1566 graph configuration files for Triva visualization tool.
1567 * Configuration option tracing/platform is renamed to tracing/categorized
1568 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1569 * New TRACE_user_link_variable interface:
1570 User provides the name of the link and the tracing variable to attach to it
1571 * the declaration of tracing categories must be done after the environment
1573 * simpler tracing interface, just one way to declare categories
1574 TRACE_category or TRACE_category_with_color, it is up to you
1575 * links in the trace file are again identified by their names
1576 * trace contains the full platform hierarchy exactly as declared using the ASes
1577 * Options tracing/msg/[task|process]:1 groups the process by hosts
1578 for both cases, tasks and processes must have names that are unique during the simulation
1579 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1580 * The experimental option tracing/msg/volume is deprecated
1581 its functionality may be reincorporated if needed
1583 The tracing generates a trace file with unordered timestamped events,
1584 because of the way the core simulator (surf) works. A script available
1585 at the tools directory (fix-paje-trace.sh) can be used to put the events
1586 in order. We have changed the tracing so it can generate ordered timestamped
1587 events in the final trace, but depending on the simulator (and how much time
1588 is simulated) that can lead to a huge memory utilization. It is deactivated
1589 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1591 Build Infrastructure
1592 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1593 - We are trying hard to keep the API stable, but it may happen that
1594 some things change (we're a research project after all, not a
1595 nuclear plant operating system). If such things should happen, you
1596 could rely on that macro to adapt.
1597 - current value: 30600 for 3.06.00, aka 3.6
1598 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1599 a number that can be compared with SIMGRID_VERSION.
1600 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1601 assertions and verbose/debug logging events are disabled at compile time.
1603 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1605 SimGrid (3.5) stable; urgency=medium
1608 * New feature to any SimGrid-based simulator: Model-Checking
1609 Check SIN#1 for more details.
1612 * New Model SMPI (three-interval linear regression for correction factors)
1613 See RR-7426, available at http://hal.inria.fr/inria-00527150
1614 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1615 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1616 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1617 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1618 * See also src/smpi/README
1623 - Tracing API changes: TRACE_start and TRACE_end should not be called
1624 by user-code. They are automatically called by simulators created
1625 with SimDAG, MSG and SMPI if the toolkit is compiled with
1626 tracing_enabled=ON. Categories declaration and utilization remain the
1627 same for MSG and SimDag.
1628 - A function was added to the tracing API to declare categories with
1630 - TRACE_category_with_color (char *category, char *color)
1631 where color must be in the following format
1632 "%f %f %f", red, green, blue
1633 and red, green, blue are float values in the interval [0, 1]
1634 - User can specify NULL as color parameter, or continue calling
1635 TRACE_category (cat)
1636 On that case, the tracing system will define random colors
1637 - The following command-line options are supported:
1638 --cfg=tracing/filename:msg.trace
1639 --cfg=tracing:1 (activate tracing, needed to use others)
1640 --cfg=tracing/platform:1 (categorized resource use)
1641 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1642 --cfg=tracing/msg/task:1 (task creation)
1643 --cfg=tracing/msg/process:1 (process creation, migration)
1644 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1645 --cfg=tracing/smpi:1 (SMPI interface tracing)
1646 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1647 - examples of examples/msg/tracing updated
1649 - DAXLoader and DOTLoader functions can generate tasks with categories
1650 - A new function to attribute a category to SD tasks:
1651 TRACE_sd_set_task_category (SD_task_t task, char *category)
1652 * Tracing the MPI interface implemented by SMPI
1653 - Collective operations are traced with states
1654 - Point-to-Point operations are traced with states/links
1655 - Tracing activated by a parameter "-trace filename" passed
1656 to smpirun during execution (considering that simgrid
1657 is compiled with tracing enabled)
1658 - To run the simulation with gdb, the simulator
1659 accepts --cfg=tracing/smpi:1 to trace SMPI
1660 - tesh files to check if smpi tracing is ok
1661 - See examples/smpi/NAS/DT-trace
1662 * GTNetS tracing re-worked
1663 - adaptation to the tracing system of GTNets to cope
1664 with modifications regarding the fullduplex mode
1665 - new tesh files to check if gtnets tracing is ok
1668 * Asynchronous communications through the functions:
1669 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1670 * New function: MSG_load_platform_script()
1671 to make possible using a lua script instead of XML files to set up platforms
1672 * New function: MSG_set_function
1673 to associate functions to processes, used when bypassing the parser
1674 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1676 Platforms: Add some more examples in examples/platforms
1677 * Grid'5000: see www.grid5000.fr
1678 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1681 * Change the XML format. This is a very important modification. SimGrid
1682 3.5 introduces a new hierarchical format based on the notion of
1683 Autonomous Systems. Compatibility with old format is ensured
1684 through the perl script provided in the install bin directory
1685 bin/simgrid_update_xml.
1686 It is now possible to build platforms with specific routing
1687 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1688 connect several platforms together. We will try to provide soon
1689 set of realistic platforms exploiting these properties (have a
1690 look at examples/platforms/ for the moment).
1691 * Take the opportunity of the XML format change to be a good XML citizen:
1692 rename link:ctn to link_ctn and similar changes (also dealt with by
1694 * Add a new routing scheme (rule-based) using regular expressions. It
1695 enables to have an extremely low memory footprint when the
1696 underlying routing is simple and can be compactly described. You
1697 need to have libpcre4-dev (perl regular expressions) installed if
1698 you want to use this routing scheme.
1699 * Revive the cluster TAG and allow to easily and efficiently (both in
1700 term of memory and speed) connect clusters together. Have a look
1701 at teshsuite/simdag/platforms/ to see how this can be done. With
1702 this tag, you can create clusters with thousands of tasks at no
1703 cost (have a look at examples/platforms/).
1704 Note: clusters are implemented as ASes, so there is no need for an
1705 enclosing AS tag if you have only one cluster in your platform.
1706 * Add new generic functions in the public interface that allows the user
1707 to call SURF 'create_resource' methods from your code (same
1708 functionality as the XML bypass mechanism but with a much lighter
1710 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1711 that uses a piecewise linear approximation to produce better
1712 results when exchanging small messages.
1713 * Add a new parameter to handle correctly full duplex link and account
1714 for interferences between uplink and downlink communications
1715 (activate with --cfg=fullduplex:1).
1718 * Rename the SD_READY (all dependencies are satisfied and task is
1719 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1720 dependencies are satisfied) state.
1721 This prevents a confusion between the notion of "ready to schedule"
1722 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1723 simulated" (SD_RUNNABLE) used by the simulation kernel.
1724 * Change the way a task is considered as ready. Instead of removing
1725 dependencies when a task is done, a counter is decreased. This way,
1726 it is always possible to reach ancestors thanks to the
1727 SD_taks_get_parents function (even after the end of the simulation.)
1728 * Change the return type of SD_Simulate from (SD_task_t*) into
1729 xbt_dynar_t. This function was in handling a dynar internally and
1730 converted it into a NULL terminated array for historical reasons.
1731 * New function SD_dotload(char*) to load a DAG described in dot
1732 format. This loader and the corresponding examples require the
1733 installation of the graphviz library.
1734 * Fix a bug in the management of tasks of size 0 in the surf network
1735 models. This problem was only visible with SIMDAG and you should
1736 thus disregard results produced with earlier versions if you
1737 relied on this feature (some tasks were blocked because of this).
1738 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1739 even though your applications were doing only point-to-point
1740 communications and sequential computations. Now you can really use any
1741 model you want (of course, if you create real parallel tasks, which are
1742 not implemented in most models beside ptaskL07, this will abort).
1743 * Add an example that schedules a DAX on an heterogeneous platform
1744 using a Min-Min strategy.
1745 * New function SD_workstation_get_current_task() that returns the kind
1746 of task currently running on a workstation in the sequential access
1748 * Raise some warnings when unexecuted tasks remains at the end of the
1749 simulation. This is usually caused by cycles in the DAG.
1752 * New function: SIMIX_process_set_function() called by MSG_set_function
1753 * Change the underlying waiting queue in semaphores so that a process
1754 can wait on several of them simultaneously (as in waitany).
1755 * Fix the way to handle tokens in semaphores so that all access patterns
1756 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1757 * kill the dirty pimple SIMIX_message_sizes_output()
1758 Please use (proper) visualization instead
1761 * New data container: setset (set of sets of elements)
1762 * New module: mmalloc (mapped malloc, allowing to have several
1763 independent segments of malloc)
1764 * New function: xbt_dict_cursor_set_data()
1765 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1766 * New function: xbt_dynar_is_empty()
1767 * New function: xbt_fifo_get_last_item()
1768 * Fix xbt_dynar_shrink(): use the right element size.
1769 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1770 uninitialized areas during expand.
1771 * Fix semaphores: previous implementation was severely broken.
1772 * Use library init/fini functions for our initialization.
1773 - you can use logs and other feature as soon as you want in your
1774 code (even before the xbt_init / MSG_init)
1775 - xbt_exit is now a no-op and produce a warning when used.
1778 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1779 benefit from the latest improvement to the simulation kernel.
1780 * Kill measurement sockets for now. If you rely on them, sorry. This
1781 release is not for you. This feature will be reintroduced in the
1782 future, but we cannot delay the release any further.
1783 * New function: gras_msgtype_get_name().
1784 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1785 * Fix (at last) the pmm example: it should not randomly fail anymore.
1787 Build chain: bug fixes and overall polishing
1788 * Cmake is now stable enough. Hence, we killed the autotools.
1789 * Port to windows ( TM :)
1790 * Fix the 'make install' target.
1791 No need to use 'make install-simgrid' anymore
1792 * Introduce a 'make dist' target compiling a *source* archive
1793 'make package' compiles a binary archive
1794 * Compile java files only on need
1795 * Add --cd and --setenv command line options to tesh
1796 * Out of source builds are not fully supported yet, but we are close
1797 * Enable supernovae and optimization flags by default for our users
1800 * Add layer to set up environment directly from lua, without XML.
1801 * The effect of gras_stub_generator can be achieved through
1802 lua too (check examples/gras/console/ping_generator.lua)
1804 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1806 SimGrid (3.4.1) stable; urgency=low
1808 The "Polishing easter eggs is probably a good idea" release.
1809 This is a bug fixes release only.
1813 * Fix a bug preventing the tasks from begin garbage collected.
1816 * Fix a bug occuring when a host involved in a communication fails.
1817 This was not detected properly by the other peer involved in the
1818 communication. Now, it's reported as a network error.
1821 * Warn the user about loop dependencies in data flow of DAX files
1822 * Obey the control-flow dependencies of DAX files
1825 * Add option "enable_smpi" allowing to not compile SMPI.
1826 Probably useful for the (Mac) users experiencing a build error here
1827 * Improve the detection of lua5.1 and ruby1.8
1829 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1831 SimGrid (3.4) stable; urgency=low
1833 The "Easter in Cargese" release. Also known as (major changes):
1835 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1836 ~> bindings were greatly improved
1837 ~> new tracing infrastructure for better visualization introduced
1839 * the "Welcome to configury modernity" release.
1840 ~> we switched from autotools to cmake, and improved our cdash
1843 A more detailled list of changes follow (full detail in svn log).
1845 Java Bindings: Various Cleanups
1846 * (install java-gcj-compat-dev on debian-like to use them)
1847 * Remove put/get: no need to export deprecated interface in Java
1848 Use send/receive instead.
1849 * Cleanup the examples and add a README per directory
1850 * Remove example autoDestination (that's the only way to go now)
1851 * Remove example explicitDestination (was a plain copy of basic)
1852 * Make JniException a runtime exception, so that there is no need to
1853 declare the fact that you may encounter such a beast. I guess that
1854 nobody will ever want to survive such error.
1855 * Create specific errors for each MSG case of failure:
1856 host failure, transfer failure, timeout, task cancelled
1857 * Cleanup the exceptions that may get thrown by each function
1858 * Other internal cleanups in Java bindings. Performance still bad :/
1859 Ruby and Lua Bindings: create them
1860 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1861 * That's new and great, you should try them out.
1862 Same functionalities than Java bindings, only even less polished
1864 * Kill the useless "rate" argument of SD_task_get_execution_time()
1865 Everyone used to provide -1 as a value, it was not used, and the
1866 semantic of a possible use wasn't even clear.
1867 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1868 either as comm costs or compute costs to mean that there is no
1869 such thing for that specific task.
1870 * Add a SD_task_set_name() function
1871 * Fix SD_task_unschedule() on typed tasks
1872 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1873 * In DAX loader, accept useless 'level' attributes to <job> since
1874 LIGO DAGs have them (seem to be to ease graphical representation).
1876 * Add an example masterslave_mailbox.c using send/receive and not
1877 the deprecated put/get interface.
1878 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1879 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1881 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1882 (declare MSG_USE_DEPRECATED to still have the old name)
1883 * Add a MSG_task_set_data() function
1884 * About trace replay (see examples/msg/actions):
1886 - Allow to work with splitted trace files for each process
1887 Give the specific trace file as argument of each process,
1888 and call MSG_action_trace_run(NULL)
1889 You can still have one merged file for all processes.
1890 - Fix implementation of collective operations
1891 * Allow task_execute() on 0-sized tasks (closes #10063)
1893 * This is the first release of SimGrid where SMPI is not considered
1894 beta anymore (even if some corners should still be improved)
1895 * Port over the new SIMIX_network submodule (internal refactoring)
1896 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1897 * Implement more missing elements of the standard:
1899 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1901 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1902 MPI_Scan() MPI_Get_processor_name()
1903 - Added implementation of missing case for Alltoall (warning: it's
1904 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1905 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1906 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1907 SMPI_MPI_Allgatherv()
1908 * Bug fixes include:
1909 - MPI_Waitsome() was broken
1910 - Allow relative includes in smpicc
1911 - Command line cfg argument 'reference_speed' was ignored...
1912 - Some functions did not properly lead to auto-benching of user code
1913 - smpicc passes -O2 by default (just like openmpi one)
1915 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1916 * Bug fixes about timeouts during communications
1917 * add SIMIX_message_sizes_output() as a pimple to write to file the
1918 amount of messages per size. Use gnuplot to get histogram.
1919 Pimple because that's the only user-visible function of simix,
1920 defined directly in xbt.h (irk, sorry)
1922 - Add a SIMIX_sem_get_capacity() function
1923 - Fix interactions with processe resume/suspende
1924 - release_forever() was stupidly broken
1925 - Fix SIMIX_display_process_status() for processes in a semaphore
1926 - Make SIMIX_sem_block_onto() user-visible
1927 * Refactoring context stuff:
1928 - Use pseudo-OOP for better modularity
1929 - reimplement SIMIX_process_kill() without process_schedule() so
1930 that the latter can take as invariant that it is called from
1932 - Merge context_start into context_new for sake of simplicity
1934 * Add a Vivaldi network model, coded live during SUD'10 ;)
1935 * Rename configuration variables to start a hierarchy:
1936 o cpu_model -> cpu/model
1937 o network_model -> network/model
1938 o workstation_model -> workstation/model
1939 * New configuration variables:
1940 o network/bandwidth_factor: correction to bandwith
1941 o network/latency_factor: correction to latency
1942 o netwotk/weight_S: correction to the weight of competing streams
1943 * Add a long description to the models, that users can see with such
1944 argument on the command line: --cfg=cpu/model:help
1945 * --help-models display the long description of all known models
1947 * config: add the ability to set a default value after registration
1948 Does not override any previously set value (e.g. from cmd line)
1949 * dict: allow to have integer key and data.
1950 When so, you need to use the following functions
1951 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1952 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1953 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1954 In contrary to regular dicts, the key is not malloced before copy.
1955 Mixing scalar and regular elements in the same dict is not tested
1957 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1958 Tracing for Visualization:
1959 * SimGrid is now instrumented in order to generate a trace file for
1960 visualization analysis: to use it, need to compile SimGrid with the
1961 "tracing" option enabled, and instrument the program using SimGrid with
1962 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1963 (among other functions).
1964 * The instrumentation only traces the platform utilization for now
1965 * Documentation to use the tracing functions and how to analyze the
1966 traces with the Triva tool is written.
1967 * More information about: SimGrid FAQ (in the section Tracing Simulations
1970 * We moved to cmake as default build system. Autotools support will
1971 be dropped soon. Check the FAQ for more info about how to use it.
1972 * Greatly improved our cdash/ctest interactions
1973 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1974 * Added memory checking tests with valgrind; lot of memleak fixing.
1975 This may be the first release of simgrid with so few memory issues
1976 * Added code coverage tests.
1977 Our coverage is still improvable, but at least we see it on cdash.
1979 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1981 SimGrid (3.3.4) stable; urgency=low
1983 The "Desktop Grid needs love too" release (also called Xmas release).
1985 Models improvements:
1986 * Major speedup in the maxmin system solving by using lazy evaluation
1987 Instead of solving completely the maxmin system at each iteration,
1988 only invalidate (and recompute) the modified parts.
1989 This new feature is enabled in default models but you can try to
1990 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1991 * Cas01 IMproved as default CPU model
1992 This CPU model is the same Cas01 model, but it uses the
1993 maxmin-selective-update flag and a heap structure to manage
1994 actions on SURF kernel.
1995 It reduces the complexity to find the next action to finish and,
1996 consequently, it's faster than the old Cas01.
1997 This is the new default CPU model (Cas01).
1998 * Rename the old Cas01 model to Cas01_fullupdate
1999 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2000 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2001 * CpuTI (CPU Trace Integration)
2002 A new CPU model whose objective is simulate faster when using
2003 availability trace files.
2004 Instead of using a full featured, over engineered maxmin system for
2005 CPU modeling, this model does the pre-integration of traces files
2006 to calculate the amount of CPU power available, and so, executes
2007 faster than the old CPU models.
2008 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2009 * Use LV08 as default network model since it gives better accuracy
2010 for small messages and shouldn't change things for big ones.
2011 Use --cfg=network_model:CM02 to get the previous behavior.
2014 ******************************************
2015 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2016 ******************************************
2017 * The new CPU model may changes simulations!
2018 The point is that events occurring at the exact same timestamp
2019 are not scheduled in the same order with the old and new
2020 version. This may be enough to completely change the execution
2021 of simulations in some cases.
2022 * The new network model will change simulations!
2023 This new model is more realistic than the previous one, so you
2024 should consider redoing your old experiments with this model.
2025 Sorry for the inconvenience.
2028 * Introduce the supernovae compilation mode
2029 When compiled that way, the whole SimGrid (or almost) is put in a
2030 single compilation unit and compiled in one shoot.
2031 This is to help gcc which has difficulties to inline stuff from one
2033 The speedup seem to be above 15%, althrough more tests are needed on
2034 amd64 to confirm that gain.
2037 * Port of MSG's mailbox on top of SIMIX network
2038 The put/get mechanism was greatly simplified on the way.
2041 * New SIMIX network module. Provides:
2042 - Mailbox: rendez-vous mechanism to find with who you want to speak
2043 - Synchronous send/recv: easier and hopefully faster since the
2044 logic is handled in the maestro process directly now
2045 - Asynchronous send/recv: you dreamt of it? It's here now
2046 Too bad that nobody cared enough to propagate the change to MSG.
2047 * Add semaphores as SIMIX synchronization mechanism.
2050 * new function SD_daxload(char*) to load a DAX file
2051 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2052 * Introduce typed tasks. Specify its kind and cost at creation.
2053 At scheduling, just give where it should be placed, and the cost
2054 for each involved resource is automatically computed.
2055 Existing constructors so far (more to come of course):
2056 - SD_task_create_comm_e2e() for end-to-end communication
2057 - SD_task_create_comp_seq() for sequential computation
2058 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2059 * new function SD_task_dump() for debuging display
2060 * new function SD_task_dotty(task,FILE*) writing to file the info
2061 about the task in dotty format
2062 * SD_task_dependency_exists() can now cope with having one of its
2063 arguments NULL. If so, it tests whether the other argument has any
2065 * Add getters on list of preceding/following tasks:
2066 SD_task_get_parents(task) and SD_task_get_children(task)
2067 * Add getters on amount of workstations and list:
2068 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2069 * Add getter on task kind: SD_task_get_kind(task)
2070 * Update the start_time and finish_time of tasks on completion/failure
2071 * Bugfix: Remove task from state swags when destroyed
2074 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2077 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2078 dynar with one separate thread per value of the dynar.
2079 * Change the prototype of xbt_thread_create(), sorry.
2080 Added a boolean parameter indicating whether we want to join this
2081 thread (used in SG only for now)
2082 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2085 * GTNetS wrappers should now be usable again (and betterly tested too)
2086 * Fix a major regression from 3.2 where the timeout provided to
2087 MSG_task_put_with_timeout() was used as absolute time before which
2088 the comm should be done.
2089 * Start to fix the <cluster> tag.
2090 - Internal links should be good now (beside of the loopback, which
2091 use the private link instead)
2092 - paths to the external world is still rather broken
2093 - the <route:multi> tag is just broken. Actually that's brain-dead.
2094 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2095 to make it less stupid
2096 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2097 * Fix a source-level compatibility glitch from 3.2: after defining
2098 MSG_USE_DEPRECATED, you can use the old name
2099 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2100 * Allow to compile from the SVN with automake 1.11
2101 * Fix some problems when using the "start_time" tag in deployment XMLs.
2102 * Fix #8569: XBT/synchro.h has redundant declarations
2103 * Fix #8563: MSG return values and exceptions
2104 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2105 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2106 Thanks to Robson Peixoto.
2107 * Fix a vicious bug in dictionaries inducing that some elements were
2108 not freed on xbt_dict_free()
2110 Portability report of this version:
2111 * Main portability targets:
2112 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2113 on (amd64/i386/ia64)
2114 - mac leopard on i386
2115 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2116 but nothing critical.
2117 * Other platforms: windows, AIX and others were not tested for this release
2119 Timing report of this version:
2120 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2121 scenario parameters). From 8h to a few seconds in desktop grid settings.
2122 * Supernovae brings about 25% speedup on i386.
2124 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2126 SimGrid (3.3.3) stable; urgency=low
2128 The "Need for Speed" release.
2130 The timings done to validate the 3.3.2 were faulty.
2131 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2133 The problem was a conversion from a manually handled vector to
2134 xbt_dynar_t on the critical path.
2135 xbt_dynar_foreach calls functions, inducing stack management crap.
2137 We inlined these functions and xbt_dynar_foreach is now breath taking.
2138 We also inlined xbt_swag_belong on the way.
2140 Here are some approximate speedup measurements (on master/slaves
2141 simulations lasting between 10s and 20s each):
2142 3.3.1 -> 3.3.2: about same performance
2143 3.3.2 -> 3.3.3: 40% speedup
2144 3.3.1 -> 3.3.3: 40% speedup
2145 3.3.1 with inline patch -> 3.3.3: 30% speedup
2147 Our reading is that the refactoring which occurred in 3.3.2 made us
2148 suffer much more from the xbt_dynar_foreach low performance, but
2149 once we solved this, this refactoring proved to be very performance
2150 effective. From the 40% speedup, somehow, 10% are due to the
2151 inlining and 30% to the refactoring.
2153 That's a pitty that gcc cannot inline functions placed in other files
2154 alone. We have to choose between:
2155 - break the encapsulation (by putting private data structures and
2156 accessors in headers files to help gcc)
2157 - live with low performance
2158 - switch to a decent compiler such as icc (not quite possible).
2160 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2162 SimGrid (3.3.2) stable; urgency=low
2164 The "Simplicity does not preceed complexity, but follows it" release.
2166 The main contributors of this release were (lexical order):
2167 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2170 * Extract the routing logic into its own object.
2171 (was dupplicated in network.c and workstation_LV07.c;
2172 Allows to implement other ways of storing that info)
2173 => kill now useless network_card concept
2174 - Use dynar to represent routes (instead of void** + int*)
2175 - kill link_set (use surf_network_model->resource_set instead)
2176 - Add a command-line option to choose the routing schema to use
2177 - Add three new models:
2178 * Floyd (shortest path computed at initialization)
2179 * Dijikstra (shortest path recomputed all the time)
2180 * Cached Dijikstra (shortest path computed on need)
2181 All these models where contributed by Silas De Munck, and are
2182 described in his ICCS09 paper.
2184 * Simplify model declaration
2185 (less redirections, less function to write when defining a model)
2186 - Factorize stuff between models:
2189 surf_model_resource_set(model)
2190 surf_model_resource_by_name(model, name)
2191 - Unify the types of models in s_surf_model_t (using an union)
2192 - Embeed fields of common_public directly into s_surf_model_t
2193 - Rename model methods:
2194 action_free ~> action_unref
2195 action_change_state ~> action_state_set
2196 action_get_state ~> action_state_get
2197 - Change model methods into functions :
2198 (model)->common_public->action_use ~> surf_action_ref
2200 * Implement a generic resource; use it as ancestor to specific ones
2201 (allows to kill duplicated code in models)
2202 Drawback: timer command don't need no name nor properties;
2203 workstation_CLM03 don't need no properties
2204 (but I guess we can live with those few bytes wasted)
2206 * Improve the action object model
2207 - implement a constructor avoiding dupplicated code about field
2208 initialization in generic_action part.
2210 * Kill the SDP model: it has an external dependency, is deprecated
2211 in flavor of modern lmm models, and didn't compile since a while
2214 * Relocation of the context module from XBT to SIMIX.
2215 (the context were decoupled from the simix processes, duplicating a lot of code)
2216 => a lot of code was factorized
2217 - less overhead is introduced during scheduling
2218 - simpler API for the context factory
2219 - the logic for process creation,destruction and manipulation was simplified
2220 * Simplification of the s_smx_process_t data structure.
2221 => accesing the simix level data associated to a process is faster now,
2222 and the code is a lot more readable.
2225 * Implement some more MPI primitives:
2226 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2227 -implementation: Bcast: flat or 2-ary tree (default),
2228 Barrier: 4-ary tree,
2230 Allreduce: Reduce then Bcast
2231 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2232 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2233 Alltoallv: flat tree, like ompi
2235 * Add support for optimized collectives (Bcast is now binomial by default)
2236 * Port smpirun and smpicc to OS X
2239 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2240 Shout out if you used it.
2243 * Display the list of still queued messages in SG mode when existing
2247 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2248 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2249 * Add xbt_str_from_file(FILE*)
2250 * Add xbt_dict_get_key achieving a linear reverse search
2251 * Remove the context module
2253 Portability report of this version:
2254 * Main portability targets:
2255 - Linux(debian)/x86/context
2256 - Linux(debian)/x86/pthreads
2257 - Linux(debian)/amd64/context
2258 - Linux(debian)/amd64/pthreads
2259 On these, we still have the eratic breakages of gras/pmm and
2260 amok/saturate_sg reported in previous version. We still think
2261 that the tests are the cause of the fault, not the tested code.
2263 - Mac OSX Leopard/x86/context
2264 Still false negative in tesh autotesting.
2265 Smpi still fails, but this time because readlink does not accept -f
2266 Everything seems to work properly beside of that.
2269 - AIX version 5.3 (only tested contexts this time)
2270 Smpi still fails there because mktemp is not installed.
2271 Everything seems to work properly beside of that.
2273 I managed to compile it for the first time, but several breakages.
2274 Won't delay the release for this exotic platform.
2276 * Windows: it's still lagging behind. If you want to help, please
2279 Timing report of this version:
2280 This version seem to be more than 5% faster than 3.3.1 (on linux
2281 64bits with contextes). The gain is less than expected, we are
2282 investigating this for next release.
2284 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2286 SimGrid (3.3.1) stable; urgency=low
2289 * Implement a --cfg-help to show existing configuration variables
2290 * Build chain do not require doxygen in maintainer mode
2293 * fix a bug on struct sizeof computation, which prevented the
2294 exchange of arrays of structs in some conditions
2295 - added a regression test about this in datadesc_usage
2296 * Allow the exchange of 0-long dynamic vectors.
2297 - for that, use -1 as indicator of dynamic size instead of 0
2298 - This implied to change any size from unsigned long to long,
2299 reducing a bit communication abilities, but I guess that with
2300 64bits being quite common, this is more than enough.
2301 - This also induce a protocol change, thus bumping network protocol
2302 version from 0 to 1 (if we have external users, we have to get
2303 clean on that point too ;)
2304 - added two regression tests about this in datadesc_usage
2305 * Be more verbose when propagating local exceptions
2306 This helps debugging.
2307 * Display the status of simulated processes when receiving SIGINT in
2311 * Allow to control the simulation from a trace file.
2312 New functions MSG_action_register() and MSG_action_trace_run()
2313 The first one allows to associate a function execution to each
2314 kind of action while the second one parses a trace file and
2315 triggers the corresponding actions within the system.
2316 For now, only a toy example is provided in examples/msg/actions
2317 * Add an exemple of process migration in examples/msg/migration
2318 * Fix a bug in task exchange which broke MSG_task_get_sender()
2319 Add a teshsuite regression test for that.
2320 [Bug: if MSG_task_get_sender() is called after sender exit,
2322 * Fix a bug which prevented suspend/resume to work properly
2323 * Display the status of simulated processes when receiving SIGINT
2324 This fixes a regression of v3.3. due to the introduction of SIMIX
2325 * Bug fixing in failure management:
2326 - trace could not start by a failure at time 0
2327 - failure during communications were not working
2330 * Add SIMIX_process_set_name() to change the name of the current
2331 process in the log messages.
2332 * Store smx_hosts in a dict since we only retrieve them by name
2333 * Move the configuration infrastructure to surf
2336 * Move the configuration infrastructure to surf
2339 * Massive internal cleanups:
2340 - Store internal structures on processes instead of hosts (allows
2341 to have more than one process per host, in addition of being more
2343 - Cleanup the initialization/finalization process
2344 - Kill a whole bunch of unneeded synchronization:
2345 processes run in exclusive manner within the simulator
2346 - Move queues from global tables to process data fields
2348 - now accept -platform and -hostfile arguments
2349 - Pass the right rank value to processes according to the hostfile
2350 * Compile the examples by default, and use them as regression tests
2351 * Implement MPI_Wtime()
2352 * Change the reference speed to a command line option
2355 * TCP_gamma can now be specified as command line option using
2356 --cfg=TCP_gamma:10000000.0
2357 * Change the --surf-path cmd line option into --cfg=path:
2360 * Also include strbuff from xbt.h public header
2361 * xbt_ex_display(): do not free the exception after displaying
2362 This allows to do more with the given exception afterward.
2363 Users should call xbt_ex_free() themselves.
2367 Portability report of this version:
2368 * Main portability targets:
2369 - Linux(debian)/x86/context
2370 - Linux(debian)/x86/pthreads
2371 - Linux(debian)/amd64/context
2372 - Linux(debian)/amd64/pthreads
2373 These targets fail about 1/10 of times on gras/pmm, but we believe
2374 that this is because of the test, not because of simgrid.
2375 amok/saturate_sg fails even more rarely, and the test may not be
2378 - Mac OSX Leopard/x86/context
2379 The test suite still spits tons of errors because some obscure
2380 force prevents us from removing the temporary directories
2381 arguing that they still contain some metadata I've never heard of.
2382 Smpi fails because seq is not installed.
2383 Everything seems to work properly beside of that.
2386 - AIX version 5.3 (both contexts and pthread)
2387 Smpi still fails there because mktemp is not installed.
2388 XML inclusions seems rosty on AIX.
2390 * Windows: it's still lagging behind. If you want to help, please
2393 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2395 SimGrid (3.3) stable; urgency=high
2399 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2400 [Malek Cherier & Mt]
2402 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2405 Until now, GRAS were using MSG as an interface to SURF. It was
2406 quite difficult because both interface have several differences
2407 (MSG channels vs GRAS sockets were the most notable point).
2409 This also opens the gate to SMPI (which should occur soon) and speed
2410 up simulations by to 40% (even if it were not the main goal).
2412 **************************************
2413 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2414 **************************************
2415 The point is that events occuring at the exact same timestamp are
2416 not scheduled in the same order with the old and new version. This
2417 may be enough to completely change the execution of simulations in
2418 some cases. Sorry for the inconvenience.
2420 * Cleanup and upgrade the XML format to push further scalability
2421 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2423 * Improve the testing infrastructure with tesh. Now a very large part of
2424 the code is tested not only by being run but also by checking that the
2425 output match an expected output [Mt].
2427 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2428 is really less memory-demanding, which should allow you to use
2429 larger files in SimGrid [AL].
2431 * Inform valgrind about our contextes, so that it becomes usable
2432 with the default (and more effecient) version of SimGrid
2433 [contributed by Sékou Diakite, many thanks]
2436 * Introduce a listener thread in charge of receiving incoming
2437 messages from the network. It allows to overlap communication and
2438 computation but most notably, it removes some stupid deadlocks due
2439 to the fact that so far, a process could not send and receive at
2440 the same time. This made most non trivial communication schema
2442 * Convert the PIDs from long int to int to match the MSG ones (and
2443 linux ones too) [Mt]
2444 * New function: gras_agent_spawn() to launch a new process on
2445 current host. Only working in simulation for now. [Mt]
2446 * New function: gras_os_hostport() returning a constant form (ie,
2447 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2450 * Make the backtrace of exceptions more human readable [Mt]
2451 * New module: xbt/str [Mt]
2452 a ton of string utility functions (split, join, printf to a newly
2453 allocated buffer, trim, etc)
2454 * New module: xbt/hash [Mt]
2455 SHA1 hashing algorithm (more to come if needed)
2456 * New module: xbt/synchro [Mt]
2457 synchronization tools (mutex and conditions) working the same way
2458 in simulation and in real life (mainly useful for GRAS, but not
2460 * New module: xbt/queue [Mt]
2461 classical producer/consumer synchronization scheme
2462 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2463 (using the classical functions will get serialized) [Mt]
2464 * Make dictionary internal table dynamic. No need to specify its size
2465 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2467 * Make sure the log channels are organized as a tree under windows
2468 (because of ANSI C compatibility issue, any channel were child of
2469 root directly) [Mt].
2472 * Cleaned many thing in surf and fixed a few bugs [AL].
2473 * Add a nice command line configuration mechanism to compose models [AL].
2474 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2475 the previous one (KCCFLN05). It relies on something that looks like
2476 a max-min sharing mechanism but cannot be written as such. A new solver
2477 was thus designed [AL].
2478 * Add a new solver to lmm. Based on Lagrange optimization and
2479 gradient-based descent, it enables to efficiently maximise systems s.a
2481 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2483 This solver enables to propose two new network models for TCP Reno and
2484 TCP Vegas based on Low's work. These models still need to be fully
2485 tested though [Pedro Velho].
2488 * Bug fix in SD_simulate. Now the time bound given as argument is
2490 * Use the new parallel task model (ptask_L07) as default.
2491 * Use the SURF command line configuration mechanism.
2492 * 0-size tasks (for synchronization) should now work.
2494 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2496 SimGrid (3.2) stable; urgency=high
2500 We still experience issues on this platform, but we believe that at
2501 least MSG is usable.
2503 GRAS API BREAKAGE (for simplification purpose, sorry):
2504 * the gras_msgtype_by_name is not used anymore. Instead of
2505 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2506 you can write (and must)
2507 gras_msg_send(toserver, "request", &request);
2508 - If you still want to pass a gras_msgtype_t to the function (to cache
2509 the type and avoid the lookup time), use the gras_msg_send_() variant.
2510 - Impacted functions:
2511 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2512 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2513 * The callbacks are now expected to return 0 when everything went well
2514 (just like the main() function)
2516 GRAS new features and improvements:
2517 * New module mechanism where user code can use per process globals [Mt]
2518 This is similar to gras_userdata_*() functions, but for libraries. It
2519 factorize some code developped over and over in the examples and AMOK.
2520 It has still to be documented and used (only amok/peermanagement is
2522 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2523 loss when they were small enough to fit into the buffer and sent quickly
2524 enough so that they can all get received in one shoot.
2525 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2526 if not found. Use the *_or_null() variant for the old semantic.
2527 * In gras_msg_handle, do not discard messages without callback.
2528 They are probably messages to be explicitly awaited later (ie, proofs of
2529 mis-synchronization in userland since they are sent before being awaited)
2530 No big deal usually.
2531 * gras_socket_meas_send/recv: semantic changed!
2532 The numerical arguments used to be (1) the total amount of data to send
2533 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2534 messages. This was need for the fool willing to send more than MAXINT
2535 bytes on quite fat pipes.
2538 * Do really rename the hostmanagement module to peermanagement. [Mt]
2539 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2540 but this is rather new and this was documented in the module
2541 documentation (poor excuses, I admit)
2542 * Bandwidth measurement semantic changed! This follows the changes to
2543 gras_socket_meas_send/recv explained above.
2546 * A sequential mode has been added to the workstations. When a workstation
2547 is in sequential mode, it can execute only one task, and the other tasks
2548 are waiting in a FIFO. [Christophe Thiery]
2551 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2552 model for SIMDAG. [Christophe Thiery]
2553 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2554 equal to 0, which caused some bad behaviors in
2555 saturated_constraint_set_update. I now use a threshold mechanism like in
2559 * When running manually src/testall, you select specific units [Mt]
2560 testall is the result of our cunit mechanism, and should replace all
2561 the scripty thingy around since bash don't run easily on billware.
2563 * A mallocator system has been added. [Christophe Thiery]
2564 Mallocators allow you to recycle your unused objects instead of freeing them
2565 and allocating new ones.
2567 Documentation update:
2568 * FAQ reworking + New FAQs:
2569 - "Valgrind spits tons of errors!" [Mt]
2570 - "How to repport bugs" [Mt]
2571 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2572 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2573 - Communication time measurement within MSG [AL]
2574 - I experience weird communication times when I change the latency [AL]
2575 * GRAS tutorial [Mt]
2577 - an introduction to the framework and to the used communication model
2578 - an initiatic tour introducing the most proheminent features:
2580 . Lesson 0: Installing GRAS
2581 . Lesson 1: Setting up your own project
2582 o Part 2: Message passing
2583 . Lesson 2: Exchanging simple messages
2584 . Lesson 3: Passing arguments to the processes (in SG)
2585 . Lesson 4: Attaching callbacks to messages
2586 . Lesson 5: Using globals in processes
2587 . Lesson 6: Logging information properly
2588 . Lesson 7: Using internal timers
2589 . Lesson 8: Handling errors through exceptions
2590 . Lesson 9: Exchanging simple data
2591 . Lesson 10: Remote Procedure Calling (RPC)
2592 . Lesson 11: Explicitely waiting for messages
2593 . Recapping of message passing features in GRAS
2594 - A HOWTO section containing:
2595 o HOWTO design a GRAS application
2596 More are due, of course. They will come latter. In the meanwhile, you can
2597 check the examples which are still here.
2599 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2601 SimGrid (3.1) stable; urgency=high
2605 There was a stack corruption somewhere, visible only when optimizing
2606 with these versions. [Vince]
2609 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2610 module that enabled to play with Directed Acyclic Graphs. It is built
2611 directly on top of SURF and provides an API rather close to the old
2612 SG. Some old codes using SG are currently under rewrite to check that
2613 all needful functions are provided. [Christophe Thiery]
2616 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2617 extension of the classical CLM03 model that gracefully handles
2618 failures. This is now the default model for MSG and GRAS. It doesn't
2619 handle parallel tasks yet though. [AL]
2620 * Bug fix: Weights were not correctly set in the network part.
2621 WARNING: This may have resulted in incorrect results with simulations
2622 where there are more than one flow on a given link. [AL]
2625 * After a (long ?) discussion on simgrid-devel, we have decided that the
2626 convention we had on units was stupid. That is why it has been decided
2627 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2628 WARNING : This means that all previous platform files will not work as
2629 such with this version! A warning is issued to ask users to update
2631 A conversion script can be found in the contrib module of the CVS, under
2632 the name contrib/platform_generation/surfxml_update.pl [MQ]
2635 * Bug fix: Processes were started in reverse order, wrt deployment file.
2636 WARNING: if your code relies on this bug, please fix it. [AL]
2637 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2638 being executed on two different locations. [AL]
2639 * Bug fix: Failures are now better supported thanks to Derrick's tests
2640 (there was many failure situations I hadn't thought of and that weren't
2641 correctly handled). [AL]
2642 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2646 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2647 Exception occurring on server-side are propagated back to client (!).
2649 API CHANGE: the callback changed their prototype. Change:
2650 int my_handler(gras_socket_t expeditor, void *payload_data) {
2652 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2653 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2655 * New! function: gras_msg_handleall to deal with all messages arriving
2656 within a given period.
2657 * New! function: gras_socket_server_range to get a server socket in a
2658 range of port numbers (ease to avoid port number conflicts) [MQ]
2659 * New! gras processes display their backtrace when they get a SIGUSR1
2660 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2661 Sweet to debug RL processes [MQ]
2665 - Do not force experiment sizes to be expressed in kb, or it becomes
2666 impossible to measure the latency this way (needs one byte-long tests)
2667 WARNING: this changes the amok_bw_* function semantic. [MQ]
2668 - Implements the link saturation stuff. [MQ]
2669 * Peer management module:
2670 New! module factorizing code that we wrote over and over [MQ].
2673 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2674 - Test units are placed directly into the library code, they get extracted
2675 automatically and placed into the src/testall binary.
2676 - Convert most of the XBT tests to this system.
2677 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2678 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2679 live in the same C code. This required to change a little bit the API
2680 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2681 * New module: sparse graph structure with basic algorithms (this is work
2682 in progress and the API is not considered to be frozen yet). [AL]
2683 * Display more information on backtraces: source line & function names are
2684 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2685 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2686 * Reworked a little bit some #include statements to load only required
2687 headers. Some user code that relied on SimGrid to include stdlib or
2688 stdio may need to include it by themselves. [AL]
2689 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2691 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2692 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2693 but I guess that almost nobody used those parts. [MQ]
2695 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2697 SimGrid (3.0.1) stable; urgency=low
2700 * Unfortunately, I had missed 5 misnamed functions:
2701 xbt_fifo_item_t xbt_fifo_newitem(void);
2702 void xbt_fifo_freeitem(xbt_fifo_item_t);
2703 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2704 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2705 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2706 They're now deprecated. Please use their new versions:
2707 xbt_fifo_item_t xbt_fifo_new_item(void);
2708 void xbt_fifo_free_item(xbt_fifo_item_t);
2709 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2710 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2711 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2713 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2714 * Documentation: xbt_log module unmercifully reworked [MQ]
2715 * Bugfix: there was a problem with the ending of contexts with
2716 the pthread backend. It caused some weird deadlock or behavior
2717 depending on the pthread implementation. [AL]
2718 * Bugfix: get the exceptions raised in the simulator repport where
2719 and why they come from when they are not catched in time [AL, MQ]
2722 * Bugfix: Do repport the error when two non-connected hosts try to
2723 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2726 * Add additionnal checkings on communications. Assert that two
2727 communicating hosts are connected by a set of links... [AL]
2730 * Add additionnal checkings on channel values in communication [AL]
2731 * New: MSG_task_get_source to see on which host a task was generated [HC]
2732 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2733 the number of tasks waiting to be received on channel and sent
2735 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2736 waits for the first task coming from a given host.. [AL]
2738 GRAS new functionnalities: [MQ]
2739 * Enhance the parsing macro to allow the size of multidimentional objects
2740 to be given thru annotations.
2741 * New example (and documentation): Matrix Multiplication a la RPC
2742 (as when I was young!) and fix a bunch of bugs found on the way.
2744 GRAS performance improvements: [MQ]
2746 * Reduce the amount of cbps creation/destruction by making it static to
2747 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2750 * Change libdata to a set so that we can search for stuff by ID (and thus
2751 reduce the insane amount of dict lookups)
2754 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2755 RL communication on top of SG since it's so uneffective.
2756 It may also be used for inter-thread communication in RL, one day.
2757 * Use gras_datadesc_copy() to exchange messages on top of SG
2759 - improve message exchange performance on top of SG
2760 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2761 * Don't exchange on the network the size of the used part of buffer,
2762 instead, specify the possible buffer size to read().
2764 - reduces the amount of read/write calls (one pair per exchange)
2765 - reduces the amount of exchanged data (the size)
2766 - allows to retrieve all arrived data on receiver side, if we don't need
2767 it right now (subsequent read will peek the buffer)
2768 - allows the receiver to proceed with the begining of the stream before
2769 everything is arrived
2770 - make it possible to build an iov transport (using readv/writev)
2772 - take care of the data with non-stable storage (like stacked data),
2774 * If possible, TCP send uses vector I/O (when writev() is here)
2775 - Don't use it for receive since we send data sizes and data on the
2776 same stream, so we wouldn't be able to chain large amount of chunks
2777 before having to flush the stuff to read the size.
2778 * Rework the transport plugin mechanism to simplify it and reduce the
2779 amount of pointer dereferencement when searching for the right function
2782 * I guess that now, we do almost as few system calls as possible while
2783 doing as few data copy as possible.
2785 To improve it further, we could try to send all the sizes first and then
2786 all the data (to use iov on receiving size), but it's only a partial
2787 solution: when you have 2 dimensional data, the sizes of the second
2788 dimension is data of the first dimension, so you need 3 streams.
2790 I'm not sure the potential performance gains justify the coding burden.
2792 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2794 SimGrid (3.00) stable; urgency=high
2797 * New! Give the possibility to hijack the surf parser and thus bypass
2798 MSG_create_environment and MSG_launch_application. Have a look at
2799 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2801 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2803 SimGrid (2.96) unstable; urgency=low
2808 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2809 This deprecates the xbt_error_t mechanisms.
2810 It modifies (simplifies) all XBT and GRAS API.
2811 MSG API keeps unchanged (exceptions raised by XBT are catched from
2812 within MSG and masked with existing error handling facilities)
2815 * New! Add a FATPIPE model. [AL]
2816 * New! Add a parallel task model. [AL]
2817 * New! Add automatically a loopback interface (in the default
2818 network model) if none was precised.
2821 * Bugfix: MSG_process_resume now works with the current running process.
2823 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2824 * Modification of MSG_task_get_compute_duration. Once a task has been
2825 processed, the value returned by this function is now equal to 0. [AL]
2826 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2827 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2828 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2829 of a m_task. For now, MSG_task_cancel only works with computation
2831 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2832 of the processor (in Mflop/s) regardless of the current load on the
2834 * API Change: use proper naming convention for MSG_getClock and
2835 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2837 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2838 This function changes the priority of a computation task. This priority
2839 doesn't affect the transfer rate. A priority of 2 will make a task
2840 receive two times more cpu power than the other ones. This function
2841 has been added to suit the needs of Nguyen The Loc and hasn't been that
2842 much tested yet. So if it fails, please report it and send me your code.
2844 * API Change: removed all functions and types that were marked "deprecated"
2845 since many months. Renamed MSG_global_init_args to MSG_global_init.
2847 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2849 SimGrid (2.95) unstable; urgency=low
2852 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2853 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2854 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2857 * Bugfix: Allow absolute paths for platform description files [MQ]
2858 * Bugfix: do free the variables after use. Leads to drastic performance
2860 * Implement max_duration (ie, timeouts) on resources [AL]
2863 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2865 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2866 the needed support (timeouts) [AL]
2869 * Implement measurement sockets. You can now get the bandwidth between two
2870 hosts thanks to AMOK (see below). [MQ]
2871 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2872 dynar over the network (yeah) [MQ]
2873 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2874 * simple/static token-ring example. [Alexandre Colucci and MQ]
2875 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2876 in place before [MQ]
2879 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2880 type parsing. "short" and "long" modifiers were also ignored; other
2881 modifier (such as reference level) are still ignored. [MQ]
2882 * Embeed the buffer size within the buffer itself on SG. [MQ]
2883 That way, send() are atomic and cannot get intermixed anymore (at least
2884 the ones which are less than 100k; bigger messages still have the issue)
2885 * Array size pushed by the field, not by the field type (or each
2886 and every long int will push stuff to the cbps) [MQ]
2887 * use select() to sleep since it allows to portably sleep less than one
2890 GRAS (minor cleanups)
2891 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2894 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2895 * type category 'ignored' killed as it was never used and were difficult
2897 * whether a type can cycle or not is now a flag, leaving room for more
2898 flags (as "ignored", if we feel the need one day ;)
2899 * Rename raw sockets to measurement sockets since "raw" has another
2900 meaning in networking community.
2903 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2904 and offers features not belonging to GRAS but that most applications
2905 need. One day, it may be a set of plugins loadable at runtime.
2906 * New module: bandwidth
2907 bandwidth measurement between arbitrary nodes running this module. [MQ]
2909 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2911 SimGrid (2.94) unstable; urgency=low
2913 The first beta release of SimGrid 3 !
2917 * Update the main page and the FAQ. Adding references to gforge.
2920 * Add a gras_os_getpid function.
2923 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2924 * Extend the logs so that they also print PID, hostname, date, ... if
2926 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2927 and kill the old version which were in testsuite/
2928 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2932 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2933 for users. Instead, each programming environment (GRAS, MSG,...) use it
2934 when needed to provide such a feature to users.
2935 Don't shortcut the mechanism or you will also shortcut the virtualization
2936 you need on the simulator.
2940 * Cleanups in configury with regard to compile optimization/warning flags.
2941 Also add -fno-loop-optimize to any powerpc since it's the optimization
2942 killing gcc (< 3.4.0).
2943 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2944 needed by MSG examples complications
2945 * Borrow configury beautifications from PHP
2948 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2949 warning (thanks loris for stumbling into this one).
2950 * Bugfix: stop loading private headers (gras_config.h) from the public
2954 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2955 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2956 RL would imply a DNS resolver, which is *hard* to do in a portable way
2957 (and therefore delayed).
2958 * Implement a real timer mechanism and use it in timing macros. This allows
2959 to avoid rounding errors and get a 0.000005 sec precision in timing
2960 macros. While I was at it, various cleanups:
2961 - allow to declare more than one timed section per file (fix a stupid bug)
2962 - move some private declaration to the right place
2963 - merge conditional execution and timing macros into emulation module
2964 - document the module
2965 - make sure the module cleanups its mess on gras_exit
2966 * Documentation improvements:
2967 - (new) how to compile applications using GRAS
2968 - (new) emulation support (timing macros)
2970 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2972 SimGrid (2.93) unstable; urgency=low
2974 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2977 - Use Paje properly where used. Still to be sanitized properly.
2978 - Portability fix: Add an implementation of the contexts using pthread
2982 - Add xbt_procname(): returns the name of the current process.
2983 Use it to show the current process's name in all logging.
2985 - fix detection of older flex version and the reaction, since we do
2986 depend on modern ones (we use lex_destroy)
2987 - Better separation of SG and RL in the libs: remove all simulation code
2988 from libgras. As a result, this lib is now only 200k when stripped.
2989 Some of the xbt modules may also be duplicated (two sets and such) and
2990 should be cleaned/killed before SG3.
2991 - Insist on using xlC on AIX because of weird problems involving gcc there.
2992 - Cleanup the make remote stuff. This is now done by scripts
2993 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2994 mainly for our private use, but we're working on changing them to user
2997 - Bugfix: flush the socket on close only if there is some *output*.
2998 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3000 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3003 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3005 SimGrid (2.92) unstable; urgency=low
3007 Alpha 3 on the path to SimGrid 3
3011 - New! First try of benchmarking macros.
3012 - New! First try so that gras_stub_generator generate deployment and
3013 remote compilation helpers.
3015 - Bugfix: Initialization fix in msg_test.
3019 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3021 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3023 - API Change: gras_os_sleep to take the amount of seconds as a double.
3024 Accepting an int was error prone since it was the only location where
3025 seconds were coded as such. It leaded to damn rounding errors.
3026 - Bugfix: Hard to belive that timers ever worked before this.
3028 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3030 SimGrid (2.91) unstable; urgency=low
3032 Alpha 2 on the path to SimGrid 3
3036 - Bug fix in the lmm_solver.
3038 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3039 through the function MSG_paje_output.
3040 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3041 - It is possible to bound the rate of a communication in MSG with
3042 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3043 to write it in the changelog).
3044 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3049 - Various cleanups to the autotools stuff
3050 - Begin to move Gras examples to examples/gras/
3051 - Let make distcheck work again (yeah!)
3053 - documentation overhauled using doxygen.
3054 gtk-doc-tools is dead in SimGrid now.
3055 - Automatically extract all existing logging categories, and add the list
3056 to the documentation (long standing one, to say the less)
3058 - Cleanup the known architecture table. Reorder the entries to group what
3059 should be, and use a more consistent naming scheme.
3060 (some of the test dataset are still to be regenerated)
3061 - New! Allow library to register globals on each process just as userdata
3063 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3064 lookup time (for now).
3065 Use it in msg and trp.
3066 This cleans a lot the internals and helps enforcing privacy of the
3067 headers between the gras components.
3068 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3069 - Bugfix: gras_os_time was delirious in RL.
3070 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3072 - Reenable GRAS now that it works.
3074 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3076 SimGrid (2.90) unstable; urgency=low
3078 Alpha 1 on the path to SimGrid 3
3080 * It is a long time since the last release of SimGrid. I'm sorry about
3081 that but as I had told you, I was rewriting a lot of things. I apologize
3082 to those who had been reporting bugs to me and that I had not answered.
3083 If your bug is still in the new version, please tell me. Here is a
3084 summary of the main changes.
3086 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3087 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3088 portability, speed, and a lot more but you'll figure it out later.
3089 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
3090 I wanted to release the new MSG as soon as possible and I have
3091 broken GRAS, which is the reason why, for now, only MSG is fully
3092 functional. A laconic description of these projects is available
3093 in the documentation.
3095 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3096 kernel optimized for our needs (called SURF but only the developers
3097 should use it). Hence, MSG is now roughly 30 times faster and I think
3098 that by rewriting a little bit MSG, I could event speed it up a little
3099 bit more. Beside the gain in speed, it is also much easier to encode a
3100 new platform model with SURF than it was with SG. More to come...
3102 * REVOLUTION 3: I have tried to change a little as possible the API of
3103 MSG but a few things really had to disappear. The main differences
3104 with the previous version are :
3105 1) no more m_links_t and the corresponding functions. Platforms are
3106 directly read from a XML description and cannot be hard-coded
3107 anymore. The same format is used for application deployment
3108 description. The new format is described in the documentation.
3109 Have a look in tools/platform_generation. There is a tiny script
3110 that converts from the old platform format to the new one. Concerning
3111 the application deployment format, parsing the old one is tricky.
3112 I think most of you should however be able to convert your files. If
3113 it is really an issue, I can write a C code that does the conversion.
3115 2) the toolbox tbx does not exist anymore. We now have a library
3116 with much more data-structures but without the hash-tables (we have
3117 dictionaries that are much faster).
3119 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3121 *****************************************************************************
3122 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3123 * don't feel like converting the oldies. (Mt) *
3124 *****************************************************************************
3127 Version 2.90: "the long awaited one"
3128 - Finished rewriting and debugging MSG. Rewrote the documentation.
3129 - disable GRAS for now since it needs to be ported to the newest SG
3132 - Finish the port to windows (using mingw32 for cross-compile)
3135 - Main loop and datastructures of SURF. A cpu resource object is
3136 functional. Surf can thus be used to create cpu's with variable
3137 performance on which you can execute some actions.
3139 2004-11-15 Martin Quinson
3140 - Port to ARM. Simply added the alignment and size descriptions. Should
3141 work, but the ARM machines are so slow that I didn't had the opportunity
3142 to 'make check' over there yet.
3144 2004-11-15 Arnaud Legrand
3145 - Trace manager now written. It uses a heap structure and is therefore
3146 expected to be efficient. It may however be speeded up (particularly
3147 when many events occur at the same date) by using red and black
3148 trees. One day maybe...
3149 - Max-min linear system solver written. It uses a sparse matrix
3150 structure taking advantage of its expected use. Most operations are
3151 O(1) and free/calloc are called as few as possible. The computation of
3152 the minimum could however be improved by using a red and black tree
3155 2004-11-03 Arnaud Legrand
3156 - Rename every gras_* function that was in xbt/ to its xbt_
3158 - Add a heap and a doubly-linked list to xbt
3159 - Added a dichotomy to the dictionaries. make check works as well before
3160 so I assume that the patch is correct. I do not know however if things
3161 run effectively faster than before now. :)
3163 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3164 SimGrid, and the version number is bumped to 2.x
3166 2004-10-29 Martin Quinson
3167 - Introduction of the remote errors.
3168 They are the result of a RMI/RPC on the remote machine.
3169 ErrCodes being scalar values, you can't get the host on which those
3170 errors did happen. Extending the error mechanism as in Gnome is possible.
3171 No idea yet whether it is a good idea.
3173 2004-10-28 Martin Quinson
3174 - Interface revolution: the Starred Structure Eradication.
3175 I used to do typedef struct {} toto_t; and then handle *toto_t.
3176 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3177 - ???_t is a valid type (builded with typedef)
3178 - s_toto_t is a structure (access to fields with .)
3179 - s_toto is a structure needing 'struct' keyword to be used
3180 - e_toto_t is an enum
3181 - toto_t is an 'object' (struct*)
3183 typedef struct s_toto {} s_toto_t, *toto_t;
3184 typedef enum {} e_toto_t;
3185 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3186 s_toto_t) is private.
3188 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3189 so that it can set the variable to NULL. It was so for dicts and sets,
3190 it changed for dynars.
3192 - Fix a bunch of memleaks in dict_remove
3193 - Fix a bug in sg/server_socket opening: it failed all the time.
3195 2004-10-07 Martin Quinson
3196 - Speed up dynar lookup operation a bit.
3198 gras_dynar_get is dead.
3200 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3201 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3202 (which gives you the address of the stored data).
3204 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3205 copy of the data using an affectation to do the job and not a memcpy.
3207 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3208 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3210 It does not really speedup the dynar test because they are
3211 setting elements all the time (and look them seldom). But the dict does
3212 far more lookup than setting.
3214 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3216 2004-10-05 Martin Quinson
3217 - Allow to (en/dis)able the cycle detection at run time.
3219 Whether we should check for cycle or not is now a property of each
3220 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3221 datadesc_cycle_unset allow to remove this property when previously set.
3223 Note that the cycle detection is off by default since it impacts the
3224 performance. Watch the data you feed GRAS with ;)
3226 This property is hereditary. Any element embedded in a structure having it
3227 set have it set for the time of this data exchange.
3229 You should set it both on sender and receiver side. If you don't set it on
3230 sender side, it will enter an endless loop. If you forget on receiver
3231 side, the cycles won't be recreated after communication.
3233 - Header reorganization.
3234 Kill gras_private.h, each submodule must load the headers it needs.
3236 2004-10-04 Martin Quinson
3237 - Interface revolution: do not try to survive to malloc failure.
3239 Now, gras_malloc and friends call gras_abort() on failure.
3240 As a conclusion, malloc_error is not a valid error anymore, and all
3241 functions for which it was the only gras_error_t return value are
3242 changed. They now return void, or there result directly.
3243 This simplify the API a lot.
3245 2004-09-29 Martin Quinson
3246 - Re-enable raw sockets.
3247 Created by gras_socket_{client,server}_ext;
3248 Used with gras_raw_{send,recv}
3251 It should allow to kill the last bits of gras first version soon.
3253 This is not completely satisfactory yet (duplicate code with
3254 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3257 - Simplify transport plugin (internal) interface by not passing any
3258 argument to _server and _client, but embedding them in the socket
3261 2004-09-28 Martin Quinson
3262 - Finish the port to AIX.
3263 autoconf was my problem (segfault within the malloc replacement
3264 function. No idea why)
3266 2004-09-16 Martin Quinson
3267 - Fix some size_t madness on 64bit architectures.
3269 2004-09-08 Martin Quinson
3270 - Reduce the number of system headers loaded, overload some more system
3271 calls (such as malloc to cast the result of the system one, and work
3273 - Fix and reintroduce the config support
3275 2004-09-07 Martin Quinson
3276 - Source code reorganization to allow Arnaud to surf all over there.
3277 - Allow to document the logging categories.
3278 - Remove all uppercase from logging categories and useless cleanup in names.
3280 2004-08-18 Martin Quinson
3281 Version 0.6.2 (protocol not changed; API changed)
3282 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3283 gras_error_t), and NULL when not found. Functions expecting a msgtype
3284 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3285 hopefully usefull message.
3286 - Portability to prehistoric sparcs again
3288 2004-08-17 Martin Quinson
3289 Version 0.6.1 (protocol not changed; ABI not changed)
3290 - prealloc some buffers to speed things up
3292 2004-08-11 Martin Quinson
3293 Version 0.6 (protocol not changed; ABI expended)
3294 - The parsing macro can deal with the references, provided that you add
3295 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3297 2004-08-09 Martin Quinson
3298 Version 0.5 (protocol not changed; ABI changed)
3299 - Allow to off turn the cycle detection code in data exchange at
3300 compilation time. It should be at run time, but I'm short of time (and
3301 the config stuff is still broken). That way, we keep dict out of the
3302 critical path, which is good because the performance is poor:
3303 - search not dichotomial yet
3304 - dynar give no way to access their content and memcpy everytime
3305 - In composed data description (struct, ref and so on), stop foolness of
3306 keeping the subtype's ID, but store the type itself. This keeps sets out
3307 of the critical path, which is good since they rely on dynar and
3308 dictionnaries. The only loose of that is that we cannot detect the
3309 redeclaration of a structure/union with another content (but I'm not sure
3310 the code detected well this error before anyway). We still can detect
3311 the redefinition discrepancy for the other types.
3312 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3313 it breaks the code because of type-punning used all over the place).
3314 This breaks on all non-gcc architectures (for now).
3316 All those changes (plus the buffer of last time) allow me to gain 2 order
3317 of magnitude on cruel tests consisting of 800000 array of integers on two
3318 level of a hierarchical structure (200 secondes -> 4 secondes)
3321 - the selector of reference must now return the type it points to, not
3322 the ID of this type.
3324 2004-08-06 Martin Quinson
3325 Version 0.4 (protocol changed; ABI not changed)
3326 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3327 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3328 - Implement a buffer transport, and use it by default (it relies on tcp in
3329 real life and on sg in simulation).
3330 That's a bit hackish since I had a new field to the structure to store
3331 its data without interfering with the subtype ones. Inheritance
3332 is tricky in C. And that's a kind of reverse inheritance with one class
3333 derivating two classes. Or maybe a game with java interfaces. Anyway,
3334 that's damn hard in C (at least).
3335 Moreover, I got tired while trying to ensure plugin separation and
3336 genericity in SG mode. MSG wants me to do weird things, so let's go for
3337 cruel hacks (temporarily of course ;).
3338 See comment in transport_private.h:71
3339 - do not include all the _interface headers in private but in the files
3340 which really need them (to cut the compilation time when they are
3343 2004-07-26 Martin Quinson
3344 Version 0.3 (protocol not changed; ABI changed)
3345 - Major overhault of the datadesc interface to simplify it:
3346 - shorted the function names:
3347 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3348 - add a trivial way to push/pop integers into the cbps without malloc.
3349 This allows to make really generic sub_type description, which simply
3350 pop their size of the stack.
3351 - add a function gras_datadesc_ref_pop_arr() which does what users want
3352 most of the time: Declare a dynamic array (which pops its size of the
3353 stack) and declare a reference to it. Poor name, but anyway.
3354 - kill the post-send callback, add a post-receive one
3356 2004-07-23 Martin Quinson
3357 Version 0.2 (protocol changed; ABI changed)
3358 - add some testing for cpbs in the test cases, and fix some more bugs.
3359 This invalidate again the little64 data file, since I cannot regenerate
3361 - remove an awfull optimization in the logging stuff, allowing me to:
3362 - understand it again
3363 - learn gcc how to check that the argument match the provided format
3364 - fix all errors revealed by gcc after that
3365 - internal keys of dict are not \0 terminated. Deal with it properly in
3366 loggings instead of segfaulting when the user want to see the logs :-/
3368 2004-07-22 Martin Quinson
3369 - Fix some stupid bug preventing cbps (callback postit) from working
3371 2004-07-21 Martin Quinson
3372 - Some documentation cleanups
3373 - remove the useless last argument of msgtype_declare
3374 - rename the Virtu functions to fit into the 'os' namespace
3375 - move headers src/include -> src/include/gras/ and stop fooling with
3376 gras -> . symbolic link
3377 - make distcheck is now successful
3379 2004-07-19 Martin Quinson
3381 - Build shared library also
3382 - Install html doc to the right location
3383 - stop removing maintainer files in make clean
3384 - build tests only on make check
3386 2004-07-13 Martin Quinson
3388 - No major issue in previous version => change versioning schema
3389 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3390 corresponding dataset.
3392 2004-07-11 Martin Quinson
3394 - Get it working with any kind of structure (we can compute the padding
3395 bytes remotely for all the architectures I have access to)
3396 - Implement the structure parsing macro (still not quite robust/complete)
3397 - Improvement to the remote testing toysuite
3399 2004-07-10 Martin Quinson
3400 [autoconf mechanism]
3401 - get ride of a bunch of deprecated macros
3402 - actually run the test for two-compliment, not only compile it :-/
3403 - test whether the structures get packed (and bail out if yes. Damn.
3404 Alignment is a serious matter)
3405 - test whether the structures get compacted (but respecting the alignment
3406 constraints of each types)
3407 - test whether the array fields of structures can straddle alignment boundaries
3409 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3410 (took me 2 hours to find that bug, looking at the wrong place)
3411 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3412 before sending/receiving and are used to compute the offsets of fields
3413 - Given that padding size depend even on compiler options, keep track of
3414 alignment and aligned_size only for the current architecture. Not a big
3415 deal since we send structure fields one after the other (seems
3417 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3418 program, let it work on linux/gcc/little32. portability todo.
3420 2004-07-08 Martin Quinson
3421 - import and improve remote compilation support from FAST
3422 - make sure make check works on half a dozen of machines out there
3424 2004-07-07 Martin Quinson
3425 Let's say it's version 0.0.3 ;)
3426 - Implement conversions (yuhu!)
3427 - Let it work on solaris (beside conversion, of course)
3428 - Stupid me, using rand() to generate the conversion datatests in not wise.
3430 2004-07-06 Martin Quinson
3431 - Let make dist work, since I'm gonna need it to compile on remote hosts
3432 - Let Tests/datadesc_usage write the architecture on which the file was
3433 generated as first byte.
3434 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3435 architecture descriptions.
3436 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3437 on those architectures.
3438 - Optimization: send/recv array of scalar in one shoot
3440 2004-07-05 Martin Quinson
3441 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3443 - Plug a whole bunch of memleaks
3444 - each process now have to call gras_{init,exit}. One day, their log
3445 settings will be separated
3446 - Continue the code factorisation between SG, RL and common in Transport.
3448 2004-07-04 Martin Quinson
3450 - Redistribution between SG and RL.
3451 We wanna have to accept in SG, so move accepted related parts of RL in
3452 the common part. (more precisely, the dynar of all known sockets is no
3453 more a static in transport.c, but part of the process_data)
3455 [gras_stub_generator]
3456 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3458 2004-07-03 Martin Quinson
3459 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3460 them all up in one shot)
3462 - Fix the ugly case of reference to dynamic array.
3463 - New (semi-public) function gras_datadesc_size to allow the messaging
3464 layer to malloc the needed space for the buffer.
3466 - gras_socket_close now expect the socket to close (and not its address to
3467 put NULL in it after it). This is because the socket passed to handlers
3468 is one of their argument (=> not writable).
3470 - propagate the interface cleanup from last week in datadesc, ie remove a
3471 superfluous level of indirection. User pass adress of variable
3472 containing data (both when sending and receiving), and not of a variable
3473 being a pointer to the data. Let's say that I like it better ;)
3474 The price for that is constructs like "int msg=*(int*)payload" in
3475 handlers, but it's a fine price, IMHO.
3477 - Let it work in RL (yuhu)
3479 2004-06-21 Martin Quinson
3481 - porting SG plugin and SG select to new standards (works almost).
3482 - plug memleaks and fix bugs around.
3485 - cleanup the prototype of data recv and force users to specify when they
3486 want to handle references to objects. Test case working even for cycles.
3487 - plug memleaks. Valgrind is perfectly ok with this.
3489 2004-06-12 Martin Quinson
3491 - cleanup the separation between plugin and main code in plugin creation
3493 2004-06-11 Martin Quinson
3495 - Reput hook for raw sockets, needed for BW experiments
3496 - kill a few lines of dead code
3497 [Data description] Interface cleanup
3498 - gras_datadesc_by_name returns the searched type or NULL.
3499 That way, no variable is needed to use a type desc once, which makes
3501 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3502 two parameters were strings (field name, type name), leading to
3504 [Dicos] Interface cleanup
3505 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3506 This is consistant with the dynar API.
3508 2004-04-21 Martin Quinson
3510 - Porting to new standards.
3512 - interface cleanup.
3513 There is no bag anymore, no need to take extra provision to mask the
3514 pointers behind "ID".
3515 Better splitup of functions between files create/exchange/convert.
3516 This is still a bit artificial since convert and receive are so
3517 interleaved, but anyway.
3519 - add a queued message list to procdata (the ones not matching criteria
3521 - factorize some more code between SG and RL wrt procdata
3523 - use gras_exit in example to track memleaks
3524 - get rid of gs_example now that GS is properly integrated into gras
3525 - update run_test to integrate the latest tests (datadesc)
3527 - rename WARNINGn macros to WARNn since it prooved error-prone
3529 2004-04-19 Martin Quinson
3531 - register init/exit functions within gras module mechanism
3532 - send/receive function.
3533 Convertion is not implemented, but short-cutted if not needed.
3534 struct/array elements are sent one by one (instead of block-wise), but
3535 nobody really cares (yet). Get a prototype before optimizing.
3536 - tests (using a file socket) for DD send/receive on:
3537 - base types: int, float
3538 - array: fixed size, string (ie ref to dynamic string)
3539 - structure: homogeneous, heterogeneous
3540 - chained list, graph with cycle
3541 Believe it or not, valgrind is not too unhappy with the results. The
3542 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3546 - Bugfix when using a filename explicitely (instead of '-')
3548 2004-04-09 Martin Quinson
3550 - factorize more code between RL and SG in socket creation
3551 - Complete the implementation and tests of:
3553 o file (only in RL, and mainly for debugging)
3555 I lost 3 days to design a portable address resolver, and then decided
3556 that the prototype mainly have to run on my box.
3557 Addressing portability too early may be like optimizing too early :-/
3559 - use gras_init in the Tests instead of the crappy parse_log_opt
3560 (the latter function is removed)
3561 [Conditional execution]
3562 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3563 [Code reorganisation]
3564 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3565 Build examples against the RL library, since there is no way to disable
3566 its creation for now.
3568 For information, the beginning of coding on GRAS was back in june
3569 2003. I guess that every line has been rewritten at least twice since