1 SimGrid (3.19) NOT RELEASED YET (target: March 20 2018, 16:15:27 UTC)
4 - Fix MSG_task_get_remaining_work_ratio() to return 1.0 for tasks that have
8 - Execution->setHost() can be called after start() to migrate it.
9 - Comm::test_any() is now implemented.
10 - s4u::Actor now has an onCreation() signal.
13 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
16 - Removed index notion from SMPI -- no more "getPid() - 1"!
17 This also means that the trace files
18 now contain actors called rank-<PROCESS-ID>. These are NOT the real
19 ranks, because tracing several communicators right now is not supported.
22 - #245: migrating an actor does not migrate its execution
24 SimGrid (3.18) Released December 24 2017
26 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
28 MSG IS NOW DEPRECATED. Please use S4U for new projects.
29 - Support for MSG should not be removed from SimGrid before 2020,
30 but future evolutions will be limited to the integration of
31 user-provided patches.
34 - s4u::onDeadlock() signal, run before stopping the simulation.
35 - s4u::Exec: asynchronous executions (abstraction of background
37 - s4u::Host->getLoad() returns the achieved speed in flops/s
40 - Link::name() is deprecated and will be removed in v3.21, use
41 Link::getCname() instead.
42 - Mailbox::getName() changed to return a std::string, use
43 Mailbox::getCname() instead to get a char*.
44 - Storage::getName() changed to return a std::string, use
45 Storage::getCname() instead to get a char*.
46 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
47 to have both allocation and deallocation of the map in user space.
50 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
51 The amount of flops remaining to do is a vector, not a scalar.
52 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
53 name implies on both sequential and parallel tasks.
54 - Both changes fix GitHub's #223 using PR #237 as a basis.
55 Thanks Michael Mercier.
56 - Most examples were converted to S4U and hidden elsewhere at tests.
57 Rationale: we still want MSG to work; we want newcomers to use S4U.
60 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
61 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
64 - Switch to the faster dlopen privatization mechanism by default
65 - Documentation improvments
68 - Live migration is getting moved to a plugin. Dirty page tracking is
69 the first part of this plugin. This imply that VM migration is now
70 only possible if one this function is called :
71 - C/MSG: MSG_vm_live_migration_plugin_init()
72 - C/C++: sg_vm_live_migration_plugin_init()
73 - Java: Msg.liveMigrationInit()
74 For C and C++, "simgrid/plugins/live_migration.h" also has to be
78 - Define class simgrid::xbt::Path to manage file names.
79 - Removed unused functions:
80 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
81 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
82 xbt_graph_export_graphviz()
83 - xbt/str.h: xbt_str_join()
84 - Remove unused datatypes:
85 - xbt/heap.h: use std::priority_queue or boost::heap instead
86 - xbt/swag.h: use boost::intrusive::list instead
89 - New link_energy plugin for the consumption of the links.
90 - All of the operations on files and storage contents have been
91 packaged into a plugin (src/plugins/file_system). The current
92 public interface can be found in
93 include/simgrid/plugins/file_system.h
94 To use these functions you now have to initialize the plugin by
95 calling MSG_storage_file_system_init() just after calling
96 MSG_init() or sg_storage_file_system_init() just after creating
100 - Remove the undocumented/untested tag <include>
103 - Remove viva specific tracing as the tool is no longer maintained
106 - #248: Kill of finished processes leads to segfault
107 - #240: xbt_cond_wait_timeout should gracefully return for C apps
108 - #239: please implement signal s4u::onDeadlock()
109 - #230: segfaults when exit() without run()
110 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
111 - #223: MSG_task_get_flops_amount() not working with parallel tasks
112 - #222: Actor::kill() doesn't really kill and segfaults
113 - #221: odd LMM warning when killing an actor
114 - #120: Memory leak when the processes are forcefully killed
116 SimGrid (3.17) Released October 8 2017
118 The Drained Leaks release: (almost) no known leaks despite the tests.
120 Even very long simulations will preserve your memory: our testsuite
121 stresses SimGrid for over 45mn on fast machines for a coverage of
122 over 80%, and there is only one single known leak, of about 4kb.
126 - Comm.detach(): start and forget about asynchronous emission. A cleanup
127 handler may be given to free resources if the comm cannot be completed.
128 - this_actor::send(mailbox) is now mailbox->put()
129 - New: simgrid::s4u::Comm::wait_all()
130 - New: Host.execute() for remote executions.
133 - Unused option network/sender-gap is removed.
136 - Deprecate MSG_task_isend_with_matching(): This unused feature
137 really complicates our internals. Will be removed in v3.20.
140 - Improved context termination. It is now possible to free resources, even
141 when a process is forcibly killed.
142 - ContextBoost: add support for Boost versions above 1.61.
145 - Bring back run-time option --cfg=exception/cutpath to remove exception
147 - Removed unused functions:
148 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
149 xbt_str_rtrim(), xbt_str_trim().
150 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
153 - Removed header files obsolete since SimGrid 3.12:
154 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
156 - #3: SD_exit should be made optional
157 - #120: Memory leak when the processes are forcefully killed
158 - #159: Threading test regression in Actor refcounting
159 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
160 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
161 - #186: Actor::killAll() segfaults if some process is blocked on wait()
162 - #191: VM migration and pstate
163 - #192: Updating the energy consumptions of all hosts crashes with VMs
164 - #195: All actors have PID=0 in the logs
165 - #204: Sometimes segfault with thread contexts and mmap privatization
166 - #222: Actor::kill() doesn't really kill and segfaults
167 - #225: Actor::kill() doesn't really kill when victims are doing a join()
169 ----------------------------------------------------------------------------
171 SimGrid (3.16) Released June 22. 2017.
173 The Blooming Spring Release: developments are budding.
176 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
177 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
178 - Java JAR file should be finally fully working on Mac OSX too.
180 XML platforms: Switch to platform v4.1 format.
181 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
182 - <zone> can be used as a synonym for the now deprecated <as>
183 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
184 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
185 - <actor> can be used as a synonym for the now deprecated <process>
186 - state_file and avail_file periodicity is now easier to express
187 (check the documentation) the old behavior should still work.
188 * Storage: not backward compatible, but it seems that nobody ever used it
189 - Remove attribute 'content_type' of <storage_type>: was never used
190 - Make attribute 'model' of <storage_type> optional: for future usage
191 - Remove Bconnection model property: was never (in)validated. Replaced by
192 the maximum of the read and write bandwidth as a resource constraint.
195 - New and Backwards Compatibility break:
196 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
197 When one wants to get the list of tasks whose states have changed during a
198 simulation round, s/he has to allocate and free a dynar and use it as
199 argument to this function. The former SD_simulate (double how_long)
203 - Allow multicore VMs, with the correct sharing (unless you overcommit)
204 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
207 - New function to update the consumption of all hosts at once.
208 - Fix the model for multi-core hosts, linear in the amount of busy
209 cores with an abnormality for pIdle (see paper).
210 BUG: Still not satisfactory for ptask on mono-cores.
213 - The netzone are now available from the MSG API.
214 The old names still work, but are now deprecated.
217 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
218 - New option: smpi/keep-temps to not cleanup temp files
219 - New option : smpi/shared-malloc-blocksize . Relevant only when global shared
220 mallocs mode is used, allows to change the size of the fake file used
221 (default 1MB), to potentially limit the number of mappings for large runs.
222 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
223 - Fortran ifort and flang compilers support
224 - New RMA calls supported (experimental) :
225 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
226 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
227 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
228 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
229 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
230 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
233 - New: this_actor::isMaestro() returns whether we are in kernel mode.
234 - Behind the scene, ActivityImpl are now automatically refcounted.
235 Removing this nasty bottleneck will greatly simplify our internals.
238 - Replay: New function xbt_replay_action_get():
239 Retrieve the function previously associated to an event type.
240 - DROPPED FUNCTION: xbt_str_varsubst()
241 - DROPPED MODULE: strbuff. We don't need it anymore.
242 - DROPPED MODULE: matrix. We don't need it anymore.
243 - DROPPED MODULE: lib. We don't need it anymore.
245 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
247 SimGrid (3.15) Released March 22. 2017
249 The Spring Release: continuous integration servers become green
250 We fixed even the transient bugs on all target architectures:
251 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
252 Windows, FreeBSD, NetBSD.
254 - Rename NetCards to NetPoints.
255 This was intended to help NS3 users, but that's not a netcard.
256 That's a point in the routing algorithm, let's avoid wrong simplifications.
259 - New: MSG_process_yield(). Stop and yield to other processes.
260 - New: MSG_process_daemon(). Daemon processes are automatically killed
261 when the last non-daemon process terminates
262 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
263 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
264 to make room for the upcoming network energy plugin.
265 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
268 - Ensure that an actor can kill itself with Process::exit()
269 - Kill the obscure NativeException. Nobody want to survive the issues
270 it denotes, so use JniException that is a RuntimeException (not to
271 be caught explicitly).
272 - Partial bug fix in initialization. SimGrid flags on command line were
273 consumed at C level but stayed in the original Java String[] args.
274 This could mess users' args[i] if SG flags were not put at the end of
276 The SimGrid flags are now removed from the Java arguments. However,
277 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
278 if args.length is greater than the number of YOUR OWN ARGUMENTS.
279 It might be if you have --log or --cfg flags in the command line.
280 - Fix numerous memleaks all around the place. In particular, around VMs.
284 - s4u::onPlatformCreated: right before the simulation starts
285 - s4u::onSimulationEnd: right after the main simulation loop
286 - s4u::onTimeAdvance: right after a clock change (time is discrete)
287 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
288 event from the availability_file changes the avail speed.
289 - Links are now usable from s4u
290 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
291 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
292 - New examples: The conversion of MSG examples to S4U has begun
293 - Actors: create, kill, migration, and suspend
294 - Applications: master-worker and token-ring
295 - Action replay: communications and storage
296 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
299 - Backwards Compatibility breaks
300 - The SD_route_get_list and SD_route_get_size functions have been removed.
301 They are replaced by the unique
302 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
303 The route from src to dst is built in the links dynar whose size can
304 be obtained with xbt_dynar_length.
305 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
306 sg_host_route_bandwidth and sg_host_route_latency.
307 Macros ensure the backwards compatibility, but you should fix your code
310 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
311 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
312 - Bump our claim of support from MPI 1.1 to MPI 2.2.
313 We don't support 100% of it, but it should be enough. Contact us if not.
314 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
315 naive and probably inaccurate)
316 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
317 mapped onto a unique small file using some system magic.
318 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
319 - Fortran: cleanups, fixes, support of user-added operations
320 - RMA: MPI_Accumulate are applied in correct order.
321 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
322 - Better support for MPI_IN_PLACE
323 - Support for MPI_Win attrs and keyvals.
324 - Support MPI_Comm_group_create, MPI_Type_size_x
325 - MPI ops should be properly applied to all allowed types
326 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
327 algorithm by a non failing one from mpich.
330 - Kill the fifo data container: we don't use it anymore.
333 - A <cluster> can now be created with different speed values to represent pstates.
334 see examples/platforms/energy_cluster.xml for an example.
336 SimGrid (3.14.159) Released December 28. 2016
338 The previous release was only a crude approximation of a Pi release;
339 this one should be a bit better.
341 - Revert a last minute change that broke on Mac OSX.
342 - Fix the changelog and NEWS files.
343 - Improve the documentation, in particular of the routing module.
344 - Integrate some patches flying around in Debian.
346 SimGrid (3.14) Released December 24. 2016
349 * The whole documentation was reworked and reorganized. There is
350 still much room for improvement here, but we're on it.
353 * We now need python3 for our testing infrastructure.
354 * Model checking needs libevent
357 * Now works on FreeBSD
360 * Privatization now works on FreeBSD
362 * Privatization is now activated by default in smpirun
363 You can turn it off with -no-privatize if something goes wrong.
365 * Call-location tracing for SMPI.
366 You can add the exact location (filename / linenumber) of an MPI call to
367 your trace files and slow down or speed up the simulation between two
368 consecutive calls by using an adjustment file (see the documentation).
370 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
371 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
372 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
373 values were ignored in some cases. The timings of these functions can now
374 be significantly different.
376 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
377 smpi/running-power is renamed to smpi/host-speed
379 * smpi/grow-injected-times option to enable or disable multiplication of the
380 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
381 make simulation less precise (but also much faster).
383 * smpirun script should be (much) faster for large deployments.
385 * SMPI tracing : fixed issue with poor matching of send/receives.
387 * Replay : Fix broken waitall
389 New functions and features
390 * MSG_parallel_task_execute_with_timeout, to timeout computations.
392 Dropped / renamed functions and features
393 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
394 - MSG_mailbox_is_empty() -> Mailbox::empty()
395 - MSG_mailbox_front() -> Mailbox::front()
396 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
397 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
398 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
399 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
400 * Don't pass the free_f parameter to property related functions:
401 - MSG_host_set_property_value()
402 - MSG_as_router_set_property_value()
403 - MSG_storage_set_property_value()
404 * VM properties. Since msg_vm_t are msg_host_t, just use
405 MSG_host_get_property_value() and friends
406 * VM I/O related things:
407 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
408 - Unimplemented save/restore methods
409 * MSG_as_router_get_property_value() was redundent with
410 MSG_environment_as_get_property_value().
411 - Removed MSG_as_router_*propert*() functions
412 - Added MSG_environment_as_set_property_value() for consistency
413 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
414 Well, they are still there for now, but deprecated with a warning.
415 Please switch to xbt_dict_new_homogeneous() before this is removed
417 * Task affinity. Its intended behavior (that was very badly tested
418 and probably not really working) was deceiving what most users
419 would have hoped here.
420 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
421 * Option network/coordinates is now useless and should be dropped.
424 * Added option storage/max_file_descriptors to allow more than 1024 files opened
427 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
428 * Added Dragonfly topology support
431 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
432 EiBps, EBps, Eibps, Ebps for bandwidth.
433 They may become useful to some lucky ones.
436 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
437 * ex/app_tokenring: new example, very similar to the MSG Token Ring
438 * ex/async_waitAll: new example, on asynchronous communications
441 * Memory usage should be decreased for simulations with a large number
442 of processes. This also helps for SMPI.
444 SimGrid (3.13) stable; urgency=low
446 The Half Release, a.k.a. the Zealous Easter Trim.
448 - We removed half of the lines, that were mostly experimental cruft.
449 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
450 (not counting blanks and comments -- according to openhub.net)
451 - The internals are now compiled in C++ (and will soon be clean C++)
452 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
453 59 klines of C, etc. We added only 29 klines of C++ in replacement.
455 * Backwards Compatibility breaks
456 - Removed Lua simulation bindings (switch to C or Java for that).
457 Lua can still be used to describe platforms
458 - Removed Java kernel plug-ins.
459 Will be reintroduced after the ongoing major internals reorg.
461 - the following functions were removed.
462 They were too specific and should be reimplemented in a generic
463 way, with filter function.
464 - MSG_task_listen_from_host
465 - MSG_mailbox_get_count_host_waiting_tasks
466 - MSG_mailbox_put_with_timeout was removed.
467 Please use MSG_task_send_with_timeout instead.
469 - the SD_application_reinit function was removed. It has been a noop for a while.
470 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
471 replaced by a more flexible mechanism.
472 - The following functions thus do not exist anymore
473 - SD_workstation_get_access_mode
474 - SD_workstation_set_access_mode
475 - SD_workstation_get_current_task
476 - Basic estimation functions have been removed but can easily be replaced
477 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
478 - SD_workstation_get_computation_time => amount / sg_host_speed()
480 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
481 Use VM.getSpeed()*load/100 for the legacy behavior.
483 - option enable_tracing was removed. It was not doing anything for a while.
484 - In the ModelChecker:
485 - the model-checker now ptraces the model-checked process which means
486 you cannot use a debugger on the latter anymore (we might make this
487 optional in the feature);
488 - removed soft-dirty page tracking;
489 - remove model-checked side snapshot management,
490 MC_snapshot() and MC_compare_snapshot();
491 - keep the MC_cut() function as a stub (it was not really working
492 in the previous release).
495 * All options are consistently in kebab-case. Old names are kept as alias.
498 * Switch to platform v4 format.
499 - Rename from 'power' to 'speed' the attributes describing the amount of
500 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
501 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
502 - In <host> and <link>, attributes availability and state are gone.
503 It was redundent with state and availability traces, and with peak values.
504 - In <cluster>, attributes availability_file and state_file are gone.
505 It was too complex and unused.
506 - Kill <gpu>. Was not doing anything.
507 - The DOCTYPE points to the right URL:
508 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
509 (the file at this address now documents the changelog since its v1)
510 - A warning is emitted for unit-less values (they are still accepted).
511 - speed. Default: 'f' or 'flops'. Also defined:
512 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
513 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
514 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
515 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
516 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
517 - latency. Default: 's' second. Also defined:
518 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
520 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
521 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
524 * s4u::Host is now the preferred public interface to the Host features.
525 sg_host_* functions are C bindings to the exact same behavior
526 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
529 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
530 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
531 They were deprecated since a few years, and probably did not even compile anymore.
534 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
535 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
536 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
537 encouraged to update their codes. A list of the modified functions can be found at the end of
538 include/simgrid/simdag.h
541 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
542 the lambda or closure passed as a parameter will run in kernel mode.
543 All the callback functions should be rewritten to that interface at some point.
546 * Reorganizing and cleaning the internals all around the place.
549 * Remove old default barrier/bcast buggy algorithms (see #18407)
550 * Various bug fixes to handle more codes
551 * Remove the need for the --foreground option of smpirun (it is still
552 accepted for backward compatibility).
555 * Kill the setset data container: MC don't use it anymore.
556 * Kill the queue data container: it made more sense with GRAS.
557 * Kill the xbt_peer_t data type: it's useless without GRAS.
558 * Kill rm_cb feature of config sets: it was never useful.
559 * Kill graphxml parsing feature. It was not used.
560 * Kill the deprecated code protected by XBT_USE_DEPRECATED
562 - xbt_dynar_sort_strings(), when the content is char*
563 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
564 They throw exceptions on invalid input;
565 * C++ support for declaring CLI flags (simgrid::config::Flag);
566 * class for abstracting different signal backends (simgrid::xbt::signal).
567 with no external dependencies (we need very simple signals).
570 * refactoring and cleanup of the code;
571 * ongoing process to cleanly separate the model-checking algorithms
572 from the code model-checking support.
574 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
576 SimGrid (3.12) stable; urgency=low
578 The Facelift Release.
581 * Require g++ v4.7 at least to not speak prehistorical C++.
582 * Require Boost 1.48 (for signal2 component).
583 * Java must be version 7 at least when activated.
584 * Builds on Windows again (including Java bindings).
585 * Tracing is now always enabled (no way to turn it out)
586 * Remove GTNetS. It was not working anyway.
587 * Various cleanups in the cmake scripts.
588 * Move headers around to sort them out on installed systems:
589 - instr/instr.h -> simgrid/instr.h
590 - instr/jedule/* -> simgrid/jedule
591 - simdag/datatypes.h was removed
592 - simdag/simdag.h -> simgrid/simdag.h
593 - msg/datatypes.h was removed
594 - msg/msg.h -> simgrid/msg.h
597 * Interface improvement:
598 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
599 - Sanitize the interface in MSG_task_ module:
600 - Merge two functions that were close enough but misleading:
601 set_compute_duration(t) -> set_flops_amount(t)
602 get_remaining_computation(t) -> get_flops_amount(t)
603 - set_data_size(t) -> set_bytes_amount(t)
604 get_data_size(t) -> get_bytes_amount(t)
605 - Massive cleanups in the functions related to the energy
606 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
607 - New: MSG_host_get_pstate()
609 - msg/energy/onoff: switching hosts on and off
611 * Interface improvement:
612 - Rename Host.isAvail() to Host.isOn()
613 - Rename Process.currentProcess() to Process.getCurrentProcess()
614 - Rename Task.setDataSize() to Task.setBytesAmount()
615 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
617 - #18874: Actually allows the GC to reclaim tasks
621 - SIMIX_process_throw: raises an exception in a remote process
622 * Refactoring: Separate sub-modules
623 - libsmx: the public interface, as libc in a real system
624 - popping: the strange dance that converts a user request into a kernel handling
625 - smx_context_*: the virtualization mechanisms that embed the user code
626 - smx_*: the handling of each simcalls
628 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
629 * Rename smx_action_t into smx_synchro_t, making explicit that these
630 things are used to synchronize processes with their environment.
631 For example, a communication is a sort of synchronization involving
632 the communicating processes (that may block until the exchange) and
633 the platform. The same can be said from computations, etc.
635 - #18888: segfault when a process ends before its kill_time
639 - 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.
640 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
641 - MPI_Type_set_name, MPI_Type_get_name
642 - MPI_*_c2f and MPI_*_f2c functions
643 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
644 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
645 - Activate a lot of new tests from the mpich 3 testsuite
647 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
648 - InfiniBand network model added : Based on the works of Jerome Vienne
649 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
650 - When smpi/display_timing is set, also display global simulation time and application times
651 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
652 * Collective communications
653 - 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
654 - 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/).
655 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
657 - Replay now uses algorithms from wanted collective selector
658 - Replay can be used with SMP-aware algorithms
659 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
660 - 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.
662 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
663 - Correctly use loopback on fat-tree clusters
664 - Asynchronous small messages shouldn't trigger deadlocks anymore
665 * Energy/DVFS cleanup and improvement
666 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
667 - new: smpi_get_host_pstate()
670 - "Full" network optimization flag was broken since Surf++
671 - Better handling of precision flags in maxmin
672 - Fix bug causing sometimes "Impossible" errors
673 - Properly pass cluster properties to included hosts
674 * Improvement of the Energy plugin.
675 - Always update the consumption before returning that value
676 - New property: watt_off to denote the dissipation when the host is off
677 - New functions getWattMinAt and getWattMaxAt to retrieve the
678 dissipation of pstates that we are not currently at.
679 * Java: class NetworkLink renamed to Link
680 * New function: simcall_process_get_kill_time()
681 * Massive rename s/workstation/host/
682 - That's intrusive, but that's good for the project consistency. Sorry.
683 - Change config option "workstation/model" into "host/model"
686 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
687 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
688 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
689 - xbt_dirname and xbt_basename for non-POSIX systems
691 * The model checker now runs as a separate process.
692 * The model checker runs must now be launched with the new simgrid-mc program.
693 * Record/Replay: the MC can display a textual representation of a path in the
694 execution graph. It can then be replayed outside of the model checker.
696 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
698 SimGrid (3.11) stable; urgency=low
703 * Normalizing pointers addresses tool for better diff between logs
706 * Add cloud examples using new VMs
707 - examples/msg/cloud/two_tasks_vm.tesh
708 - examples/msg/cloud/simple_vm.tesh
709 - examples/java/cloud/cloud.tesh
710 - examples/java/cloud/migration/migration.tesh
711 * Add java surf examples:
712 - examples/java/surfPlugin/surf_plugin.tesh
713 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
714 - examples/java/surfCpuModel/surf_cpu_model.tesh
715 * Add SMPI+MSG example:
716 - examples/smpi/smpi_msg_masterslave/
721 - msg task destroy cancel test
722 - msg_host on/off test
723 * Move all tests in testsuite to teshsuite (adding tesh files)
724 * Restructure teshsuites
725 - one folder for each kind of test
726 * Restructure AddTests.cmake
728 - structure the order of tests (with sections)
731 * Add virtual machine
732 - creation of a VM on a PM
733 - migration of a VM from a PM to another PM
735 - MSG_process_join(msg_process_t process, double timeout)
736 - msg_bar_t MSG_barrier_init(unsigned int count)
737 - int MSG_barrier_wait(msg_bar_t barrier)
738 - void MSG_barrier_destroy(msg_bar_t barrier)
739 - msg_as_t MSG_environment_as_get_by_name(const char * name)
740 * New option "msg/debug_multiple_use" to help debugging when a task is used
743 - Improvements and finalization of MSG_storage, MSG_file APIs and their
745 - Increase code coverage in test suites
749 * Protect context stacks against stack overflow. The number of protected memory
750 pages allocated on the top of each stack (1 by default) can be configured
751 with the new command line option --cfg=contexts/guard_size:<value>.
752 * Simcalls are now generated by a python script that
753 - generates files included by SimGrid
754 - checks that all the functions exist, and proposes prototypes
756 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
757 host_set_data, host_get_data
759 - simcall_process_join(smx_process_t process, double timeout)
760 * Fix bug where sleeping processing could not be suspended.
763 * Translate surf models from C to C++
764 - Generic classes for all models: Model, Resource, Action
765 - A generic interface for each kind of model (CPU, Network, Storage
766 Workstation, WorkstationVM)
768 * Translate surf routings from C to C++
769 * Add callbacks using sigc++ or boost::signals2
770 - Add callback functions for resource creation/destruction
771 - Add callback functions for action state change
772 - Handle Energy as a plugin
773 * Replace swag by boost::intrusive
774 * Add new routing models for clusters. For documentation, see
775 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
776 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
777 parameters for cluster tag
778 - Fat trees, with topology="FAT_TREE" and
779 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
780 - see examples/platforms/torus_cluster.xml and
781 examples/platforms/fat_tree_cluster.xml
785 * Hostfiles support host:nb_processes construct to deploy several processes on
787 * Collective communication algorithms should not crash if used with
788 improper number of nodes and report the error.
789 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
790 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
791 MPI_Dims_create, MPI_Cart_sub are supported.
792 * New interface to use SMPI programmatically (still depends on MSG for
793 some parts, see examples/smpi/smpi_msg_masterslave) :
794 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
798 * Global variables privatization in MPI executables is now performed at runtime
799 with the option smpi/privatize_global_variables (default:no).
800 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
801 dynamic libraries loaded by the application are not privatized (static
802 linking with these libraries is advised in this case)
805 * Options defined in XML work correctly now.
808 * New cmake option, enable_lib_in_jar, to control whether native libraries are
809 copied into simgrid.jar or not (ON by default). Use this option if you want
810 to reduce the size of the installed simgrid.jar, *and* the native libraries
811 are kept installed elsewhere.
812 * Surf binding with SWIG (code generated in maintainer mode only):
813 - plugin to handle callbacks
814 - CPU model only for the moment
817 * Supernovae build mode is definitively removed. It was used to improve
818 inlining and inter-module optimizations. It is nowadays superseded by
819 link-time optimizations commonly available in compilers.
820 * Update ns-3 find lib. Bindings for ns-3 should work again now.
821 * Add boost dependency for surf++
822 * Add new macro for tests
823 - ADD_TESH(name <tesh_args>)
824 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
828 - xbt_bar_t XBT_barrier_init(unsigned int count)
829 - int XBT_barrier_wait(xbt_bar_t barrier)
830 - void XBT_barrier_destroy(xbt_bar_t barrier)
831 * Make the xbt_os_time module public
833 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
835 SimGrid (3.10) stable; urgency=low
837 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
840 * Reintegrate Java to the main archive as desynchronizing these
841 package is not acceptable anymore (Java is now considered stable)
842 * Add explicit synchronization facilities through semaphores
843 * Bug fix: Task.setDataSize() only changed the C world, not the value
844 cached in the Java world
847 * Dramatically change the way files are handled. API and internals changed, but
848 this part of MSG was not considered as production grade either.
849 * Add explicit synchronization facilities through semaphores
850 * Add a new function MSG_host_get_process_list()
851 * Preliminary DVFS support (see examples/msg/energy/ for details)
854 * SMPI is now included directly in the libsimgrid as the windows
855 linker doesn't force us on splitting it anymore.
856 * Improvements of the SMPI replay tool:
857 - Most of the collective communications are now rooted in the same process as
858 in the original application.
859 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
860 used until now). Multiple data types can now be used in a trace.
861 - The replay tool now supports traces produce either by TAU or a modified
863 - Bug Fix: the compute part of the reduce action is now taken into account.
864 - Gatherv collective is now supported
865 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
866 of smpirun outputs time independent traces for the current run. One file
867 is created per process. If too many processes are simulated, this behavior
868 can be changed to one file for all processes by using the
869 tracing/smpi/format/ti_one_file flag
870 * smpirun generates the host file if needed (with given host count and platform)
871 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
872 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
873 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
874 --cfg=smpi/coll_selector:(mpich/ompi)
875 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
877 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
878 happening inside a collective SMPI call.
879 * Fix the behavior of complex data types handling.
880 * Make MPI_Wtime another synchronization point to take computations into
882 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
883 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
884 * Add all missing Fortran bindings, SMPI should work with Fortran 90
885 (no privatization of global variables yet)
886 * Preliminary DVFS support (see examples/smpi/energy/ for details)
889 * Verification of liveness properties is now available for SMPI applications
890 (in addition to MSG applications)
891 * Bugged examples using SMPI in examples/smpi/mc/
892 * Add --cfg=model-check/visited option. Allows the verification of infinite
893 programs. Detection of loops in the execution thanks to the system state
894 comparison and reduction of the state space to explore. Can be combined with
895 DPOR for safety properties.
898 * Allow to change SimGrid configuration (see --help) within the code
899 thanks to SD_config() as it can be done in MSG.
900 * Add a new function SD_task_set_amount() upon user request.
903 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
904 * Remove rule based routing (no more PCRE dependency)
905 * Add a limiter_link option to cluster tag, to specify a maximum reachable
906 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
907 * Add a loopback_bw and loopback_lat options to cluster tag.
908 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
909 routing. Now peers are encapsulated in an AS and have their own private
910 router but this is transparent.
913 * Our own implementation of getline is renamed xbt_getline, and gets
914 used even if the OS provide a getline(). This should reduce the
915 configuration complexity by using the same code on all platforms.
916 * New type: xbt_cfg_elm_boolean.
917 * Allow to use yes/no for boolean configuration options in the command line.
918 * Allow to disable SimGrid cleanups at exit from command line option.
919 There are situations where one may want a simulation to end with an exit.
920 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
921 annoying when scripting around the simulator. Adding a
922 --cfg=clean_atexit:no allows to circumvent this issue.
925 * Lots of memory leaks were corrected in this release.
926 There is no known memory leaks anymore, in all of our 600+ tests.
927 * New command line option --version, to get SimGrid version information.
928 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
930 * Supernovae builds are deprecated, and expected to be removed in the next
933 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
935 SimGrid (3.9) stable; urgency=low
937 The Grasgory release: GRAS is really dead now.
939 * Complete overhaul of the internal host structures scheme.
942 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
943 was considered as experimental and badly maintained since then.
944 * Keeping it was thus a trap to our potential users, that could take
945 it instead of MSG or SMPI by mistake despite is pity state.
946 * GRAS seems to have very few users (if any), and no one volunteered
947 to maintain it further. It also induces a lot of XBT code (for
948 portability sake), that must be maintained too.
949 * For all these reasons, we killed GRAS. If someone wants to revive it
950 in the future, don't cry, our git history still remembers of GRAS.
953 * Major overhaul. Merge our documentation again as time proved that
954 splitting it was really not helping our users.
955 * Further improve the developer documentation to help newcomers
956 hacking on SimGrid itself. The user documentation (and in
957 particular, the beginner documentation) is still in a sorry state.
960 * Now works on Windows too!
961 * Much more extensive test suite, from MPICH
964 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
965 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
966 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
967 * Bug fix: let task be scheduled when the last dependency to be solved is
968 a control dependency.
969 * Remove SD_load_environment_script function.
970 Use the C sg_platf function if you want to declare a platform
974 * New function: MSG_process_get_number()
975 * Old function documented: MSG_config()
976 * Remove MSG_load_platform_script function
977 Use the C sg_platf function if you want to declare a platform
981 * Change the default value of the TCP_gamma constant (maximal size of TCP
982 congestion window) to a more realistic 4MiB value. If you notice changes in
983 your simulation results, you can fall back to the previous 20k tiny window
984 by adding --cfg=network/TCP_gamma:20000 on command line.
985 * (Hopefully) fix a bug wrt periodic availability and state traces
986 * Bug fix: use default values at start when first event in availability/state
987 trace is not at time 0.
990 * remove the "new_" part of function name sg_platf_new_trace_connect
991 (resulting in sg_platf_trace_connect), since it does not create
995 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
996 use case where it's really mandatory, and maintaining it was a pain in
998 * New: xbt_fifo_search(), search an item with a user-provided
999 comparison function instead of dumb pointer comparison.
1002 * Fix the lua deployment:
1003 Use `simgrid.init_application()` before deployment instead of
1004 `simgrid.msg_register_application()` after.
1007 * Transfer the tracing files into the corresponding modules.
1009 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1011 SimGrid (3.8.1) stable; urgency=low
1013 The "we are told that some people want to also *install* the simgrid
1016 * Add missing file "tesh.1" to the archive.
1018 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1020 SimGrid (3.8) stable; urgency=low
1022 The Psssshiiiit release: SimGrid jumps into the Cloud.
1025 * Add an experimental interface to manipulate VMs. They are mainly
1026 process groups with very few intrinsic semantic, but they should
1027 allow you to build the semantic you want easily.
1028 * New function: MSG_host_set_property_value()
1029 * New function: MSG_process_on_exit(). To clean memory in all cases.
1030 * Bug fixes that made the host (and link) failures unusable.
1031 * Add a way to auto-restart process when the host in which they are
1032 executing comes back (ON_FAILURE="RESTART" on deployment file,
1033 MSG_process_auto_restart_set).
1034 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1035 please stop using the old ones, they are DEPRECATED.
1037 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1038 Please use MSG_init() instead. (reducing the amount of entry
1039 points in the library helps us).
1040 * Make it impossible to link against the wrong version of the lib
1041 * Deprecate MSG_clean(). No need to call it anymore.
1042 * Function MSG_get_host_number() is not deprecated anymore.
1045 * Split the doc into a user guide and a reference guide.
1046 * Start a developper guide to help people hacking on SimGrid.
1049 * Enable tracing by default. This modules rocks you should use it.
1050 * Remove option custom_flags. Now use environment variables CFLAGS
1052 * Use default cmake things to detect lua instead of home grown ones.
1053 * New option "enable_mallocators" to disable mallocators, for debugging
1054 purpose ("on" by default).
1057 * Bug fixes around the resource failures: don't let the processes
1058 survive the host they are running onto.
1059 * Add an interface to auto-restart processes when the host in which they are
1060 executing comes back.
1061 * Ensures that SIMIX_clean is called automatically. It's not part of
1062 the public interface anymore (bindings should be updated).
1065 * Bug fix for when SD_Simulate is called with a positive value: be careful
1066 when comparing doubles. Sometimes they are different for non significant
1068 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1069 parallel task whose initial work is distributed among host according
1070 to the Amdahl's law. Such tasks are created with a parameter alpha
1071 that corresponds to the non-parallelizable part of the computation.
1072 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1073 between two sets of workstations assuming a 1D block distribution (each
1074 workstation owns a similar share of data) on both sides.
1076 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1077 Data redistribution will be automatically scheduled once parent and child
1078 are both scheduled. The filling of computation_amount and
1079 communication_amount structures is now done seamlessly thanks to the chosen
1081 * New function SD_workstation_dump to display various information
1082 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1083 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1084 bandwidth on the route onto which the task is scheduled and the amount of
1086 To divide the nominal bandwidth by 2, the rate then has to be :
1087 rate = bandwidth/(2*amount)
1088 * Compute tasks that have failed can now be rescheduled and executed again
1089 (from their beginning)
1090 * Increasing source code coverage (src/simdag is now covered at 95.8%
1094 * Re-implement time-independent trace replay using SMPI (at the
1095 smpi_smp_* level) instead of MSG. This should replace
1096 examples/msg/actions/actions.c
1097 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1098 hindexed and structs)
1099 * Implement the exchange of non-contiguous data.
1100 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1101 * Correct behavior of smpi/sender_gap and set its default value to 0
1102 * Add option to asynchronously send small messages to allow better
1103 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1104 specifies the size in bytes under which messages will be asynchronously sent.
1105 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1106 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1109 * New C interface to define a platform: XML is now optional.
1110 For more info, please check include/simgrid/platf.h.
1111 * New interface to define random platforms from the C:
1112 For more info, please check include/simgrid/platf_generator.h and
1113 examples/msg/masterslave/masterslave_platfgen.c
1114 * Export a sg_cmdline dynar containing all the arguments we got from
1118 * Two new tracing options for adding comments to trace file so you
1119 can track your experiments (see --help-tracing for details).
1120 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1121 * Adding the SimGrid version that generated the trace file as a comment.
1122 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1123 * Fix to avoid key clashes on Paje links
1124 * Other minor fixes related to the Paje specification
1127 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1128 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1129 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1130 * Greatly improve the robustness of mmalloc to user errors (such as
1131 using an area after freeing it, or freeing it twice)
1133 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1135 SimGrid-java (3.8.1) stable; urgency=low
1137 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1138 Warning: all methods are visible, but only some of them are
1139 implemented so far. Check the source (src/jtrace.c)
1140 for further information.
1141 * New module: org.simgrid.msg.File (SimGrid File management functions)
1142 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1143 * Change the meaning of Process.restart: now restart the process from
1144 the begining, like MSG_process_restart in C.
1145 * Add Process.setAutoRestart: handling of process restart when failed
1147 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1148 you to retrieve the properties of the processes/hosts
1149 * Deprecate Msg.clean(): you can just forget about it now.
1150 * New function Process.getCount(), that only works when compiling
1151 with the not yet released version 3.9 of the C library.
1153 * New context factory based on Coroutines. It mandates a modified JVM
1154 but then, the simulations run about five times faster, and there is
1155 no limit to the amount of processes (beside of the available memory).
1157 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1159 SimGrid (3.7.1) stable; urgency=low
1162 * Restore the prototype of MSG_process_create_with_environment() to
1163 the pre-3.7 situation by removing the kill_time argument.
1164 * Add a MSG_process_set_kill_time() function instead.
1167 * Fix weird behaviors when dealing with parallel tasks.
1170 * Simgrid is now built as a dll.
1171 * Simgrid-java now works on Windows.
1172 * Simgrid-Java is now included into Windows package.
1175 * First pre-build package for MacOSX.
1178 * Fix compilation when using MSG_USE_DEPRECATED.
1179 * Fix some compilation issues on Macs and Windows.
1180 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1181 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1183 We discovered that the Lua console is broken, but we are missing the
1184 manpower to fix it right now. The problem existed in 3.7 too, so we
1185 are not blocking the release for that. Sorry if you depended on this
1186 feature, any help would be really welcome.
1188 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1191 SimGrid-java (3.7.1) stable; urgency=low
1193 The "Java aint got to be bloated and slow" release
1196 * Various internal cleanups and performance improvement
1197 Simulations are expected to run up to twice faster or so
1198 * Make Process.kill(process) an instance method, not a static one
1199 * User processes are not java.lang.Thread subclasses.
1200 This breaks the compatibility (sorry), but previous API was
1201 brain-dead, making it impossible to have non-trivial
1202 initializations in the process constructor.
1203 * Require a full constructor per Process sub-class.
1204 Kinda breaks the compatibility (sorry), but this allows a much more
1205 efficient way to launch the processes at simulation startup.
1206 * Do not embeed our version of semaphores, java 1.5 can be considered
1207 as sufficiently prevalent for us to not dupplicate its features.
1211 * Add examples for almost every part of the API
1212 We spotted and fixed a lot of bugs in the process
1214 * New module: asynchronous communication API
1215 * New function: Process.sleep()
1216 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1217 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1218 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1220 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1222 SimGrid (3.7) stable; urgency=low
1224 The "spring cleanups (before the next Big Project kicks in)" release.
1227 * We can specify the SMPI latency/bandwidth factor with command line
1228 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1229 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1230 You can also use the "config tag" from platform file by adding this line
1231 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1232 (see "example/platforms/tag_config.xml" to use "config tag").
1233 Note that the command line supersedes the platform file configuration.
1234 * Change the correction factors used in LMM model, according to
1235 the latest experiments described in INRIA RR-7821.
1236 Accuracy should be improved this way.
1237 * Use the partial invalidation optimization by default for the
1238 network too. Should produce the exact same results, only faster.
1239 * Major cleanup in surf to merge models and split some optimization
1240 mechanisms from the core of the models. As a result you can now
1241 specify which model to use (e.g., --cfg=network/model:LV08
1242 --cfg=cpu/model:Cas01) and which optimization mode to use
1243 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1244 Incompatible combinations should err at initialization. See
1245 --help-models for the list of all models and optimization modes.
1246 * The CLM03 workstation model was dropped for simplicity because it
1247 used the deprecated CM02 network model. Use default instead.
1248 * Rename the TCP_gamma configuration option to network/TCP_gamma
1249 * Rename the coordinates configuration option to
1250 network/coordinates, and document it
1251 * Use now crosstraffic keyword instead of the terribly misleading
1252 fullduplex keyword. It is activated by default now in the current
1253 default model, use --cfg=network/crosstraffic:0 to turn it off.
1254 * Ongoing refactoring the model parsing to make XML files optional
1255 See include/simgrid/platf.h for details (still to be completed)
1258 * Major overhaul of the documentation. Almost instructive now :/
1259 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1260 functions and friends. This interface was considered as
1261 deprecated since over 2 years, it's time to inform our users that it is.
1262 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1263 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1264 use these (crappy) functions in your code.
1265 These functions will be removed soon. Stop using them now.
1266 * Deprecate MSG_get_host_{table,number}
1267 Implement MSG_hosts_as_dynar() instead.
1268 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1269 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1272 * Stabilize the parallel execution mode of user contexts
1273 * Introduce configuration variables to control parallel execution:
1274 - contexts/synchro: Synchronization mode to use when running
1275 contexts in parallel (either futex, posix or busy_wait)
1276 - contexts/parallel_threshold: Minimal number of user contexts
1277 that must be part of a scheduling round to switch to parallel
1278 execution mode (raw contexts only)
1279 * Fix bugs that prevented to use suspend/resume along with
1280 synchronization structures.
1281 * Fix bugs in process termination that lead to invalid memory access
1282 in very specific conditions.
1285 * Introduce a parallel mode for the models (controlled by surf/nthreads
1286 configuration item). In our tests, running the models in parallel
1287 never lead to any speedups because they are so fast that the gain
1288 of computing each model in parallel does not amortizes the
1289 synchronization costs, even when ultra fast futexes are used.
1290 This is released anyway because YMMV.
1293 * Performance boost by using a swag internally to compute the set of
1294 tasks that are finished and should constitute the return value of
1298 * Enable it by default now that it is considered rather stable.
1301 * Documentation of the tracing functions.
1302 * Performance gains when tracing categorized/uncategorized resource
1303 utilization by avoiding calls to get route when updating resource
1304 variables. LMM constraints are being used instead.
1305 * API changed to set task categories. Use MSG_task_set_category instead
1306 of TRACE_msg_set_task_category, and SD_task_set_category instead
1307 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1308 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1309 * Examples for link user variables added (see at examples/msg/tracing/)
1310 * Deprecated function TRACE_msg_set_process_category completely removed
1311 * Trace header updated according to the latest Paje file format
1312 * Tracing network lazy updates, no longer obligate users to use full updates
1313 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1314 * Experimental: let user code declare/set/push/pop application states for hosts
1315 * API changed to allow the manual creation of graph configuration files
1316 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1319 * Improve the API of Lua MSG bindings, using the Lua spirit.
1320 * Each simulated process now lives in its own Lua world (globals are
1321 automatically duplicated). It helps writing simulators. Will allow
1322 to run Splay programs within SimGrid in the future.
1323 * Add a Chord example in Lua, equivalent to the MSG one.
1326 * Start the implementation of a solution to express temporal
1327 properties, not only local assertions. This is still an
1328 experimental work in progress, stay clear from it to be safe.
1332 - Add new runtime parameters --help-logs and --help-log-categories
1333 to display information about supported logging parameters and
1335 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1336 don't exists anymore.
1337 * Mallocators: allow value NULL for the reset function.
1339 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1340 create homogeneous dictionaries, where all the elements share the
1341 same free function. Non homogeneous dictionaries will be
1342 deprecated in the next release.
1343 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1344 - Multi-level dictionaries are deprecated.
1346 - new function xbt_dynar_search_or_negative() that is useful when
1347 you have less than 2 million elements in your dynar and don't
1348 want of the extra complexity of catching exceptions when the
1349 element is not found.
1351 - Make xbt_os_thread module (for thread portability) public.
1352 Documentation is still to come, sorry.
1354 - Cleanups and simplifications to make it maintainable again.
1355 - Exotic features (such as memalign and valloc) were removed.
1356 - The metadata were extended and improved so that the
1357 model-checker becomes able to explore and inspect the heaps.
1358 - This may induce a performance drop when enable_model-checking is
1359 ON in cmake (even if it's not used in the simulation), but it is
1360 necessary at this point to get MC working.
1362 Turn model-checking OFF if simulation performance matters to you.
1363 Not enabling it at runtime is not enough, disable it in cmake.
1365 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1368 SimGrid (3.6.2) stable; urgency=low
1370 The "Not coding new stuff allows to polish old things" release.
1373 * New bindings to the NS3 packet level simulator (experimental)
1374 * Use the raw (efficient) execution contexts instead of the sysv
1375 (portable) ones when possible.
1376 * libpcre is now mandatory in any cases since not using it led to
1377 severe performance loss and possibly other issues
1378 * Update the XML platforms:
1379 - G5K: include the latest machine in Nancy
1380 - GridPP and LCG: new platforms
1381 * Documentation was partially updated, at least (more to come)
1383 Bug fixes, cosmetics and small improvements
1384 * Free terminated processes before the end of the simulation to avoid
1385 exhausting the memory of users having very dynamic amount of
1387 * Bug fix and cosmetics about canceling non-running tasks
1388 * Bug fix about the dot loader's issues when using libcgraph
1391 * Create an installer for windows with nsis (amd64 and win32)
1392 - Add an hello world project to illustrate simgrid project creation.
1393 - Embed libpcre into the Simgrid installer to avoid
1394 its compilation burden
1395 * The raw execution contexts should work on Apple now
1396 * Port to Windows 64 bits
1397 - Sysv contexts now have an implementation for this arch
1398 - GRAS communication features now support this arch
1399 * Drop support for borland compiler on windows
1400 - this code was not maintained, and we kinda depend on gcc nowadays
1401 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1402 * Fix portability issue on unstable ubuntu: linker became picky on
1405 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1407 SimGrid (3.6.1) stable; urgency=low
1409 The "Oops, we broke Macs too" release
1412 * Fixed contexts detection so that raw ones are used when possible
1413 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1414 leads to a strange error, with user code segfaulting sometimes when
1415 the generated code is not perfectly aligned (which is not
1416 controllable from the user side, depends on the amount of code)
1419 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1421 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1423 SimGrid-java (3.6) unstable; urgency=low
1426 * Split of every thing from simgrid v3.5 into a separate package.
1428 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1430 SimGrid (3.6) stable; urgency=medium
1432 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1435 * Bindings now constitute their own package, separated from the main one.
1436 Rationale: reduce our maintenance nightmare by reducing the module coupling
1437 They will soon be released on their own on gforge.
1439 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1440 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1442 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1443 * It's not quite deprecated for now because we have no replacement,
1444 but it may soon become the case.
1447 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1448 * Fortran: New user-level cache variable to store the rank of the running
1449 process. This improves performance by an order of magnitude.
1450 * C: New coccinelle script to automatically locate and modify global and
1451 local static variables.
1452 * Improved SMPI network model with a sender-side gap to account for multiple
1456 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1457 only say if a communication is finished, no matter whether it succeeded or
1458 failed. You can call MSG_comm_get_status() to know the status of a finished
1460 * New function MSG_task_dsend() to send a task and detach it. When a
1461 communication is detached, you are never notified of its success or failure
1462 and the memory is released automatically once it is finished. This function
1463 is useful when you don't care about the end nor the success of a
1465 * Change the prototypes of action replay. Sorry for inconvenience,
1466 but this is really more efficient this way (and to adapt your code,
1467 you just have to fix the initialization, that shouldn't be too long)
1468 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1469 ever managed to do anything useful with it.
1470 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1471 takes a dynar of communications. It returns immediately and gives the
1472 index of a finished communication (if any).
1473 * New example: a basic implementation of the Chord P2P algorithm.
1476 * New model for multi-core CPUs. You can now use the core attribute to
1477 precise the number of cores of a host. This is a basic model. Every
1478 process running on the host receives at most the power provided in
1479 the DTD (throughput<=power). Total throughput of process cannot exceed
1481 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1482 router linked by an up-link and a down-link (possibly asymmetrical).
1483 This kind of pattern allows to easily build last-mile model style platforms.
1484 Aggregating such patterns in a rule-based AS is thus the technique of
1485 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1486 * New model for Vivaldi routing. We transformed the Vivaldi network model
1487 into a Vivaldi routing model (based on the rule-based model). This allows to
1488 combine Vivaldi based latencies with last-mile platforms.
1491 * Added a check for NaN of IEEE754 infinite in the double entries of
1493 * Introduce a new context factory "raw", highly inspirated from the
1494 ucontext factory, but using manually crafted functions in assembly to
1495 do the work in an efficient manner.
1496 * Allow to change the used context factory at run time, not only at
1497 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1498 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1499 contexts at runtime (only with raw contexts or ucontexts).
1500 * Completely rewrote this module to allow parallel execution of user
1501 processes. Use --cfg=contexts/nthreads:N to execute user processes
1502 with N parallel threads (the default is 1, meaning no parallelism).
1503 * Allow to decide dynamically between sequential and parallel modes.
1504 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1505 processes in parallel only when their number is greater than or equal to P
1507 * Added a check for NaN of IEEE754 infinite in the double entries of
1511 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1512 won't output the state of processes when interrupted with Ctrl-C
1513 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1514 NULL-terminated array. This may solve backward compatibility issues
1515 due to the change to return type of SD_simulate. See also:
1516 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1517 * Add new macros with variable number of arguments.
1518 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1519 - in xbt/asserts.h: xbt_assert
1520 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1521 - in xbt/ex.h: THROWF and RETHROWF.
1522 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1524 * Change xbt_die() to accept a format string with arguments, just like printf.
1525 * New data structure: xbt_lib_t, like a dict but more general and with better
1529 * New configuration options
1530 Options triva/categorized and triva/uncategorized can be used to generate
1531 graph configuration files for Triva visualization tool.
1532 * Configuration option tracing/platform is renamed to tracing/categorized
1533 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1534 * New TRACE_user_link_variable interface:
1535 User provides the name of the link and the tracing variable to attach to it
1536 * the declaration of tracing categories must be done after the environment
1538 * simpler tracing interface, just one way to declare categories
1539 TRACE_category or TRACE_category_with_color, it is up to you
1540 * links in the trace file are again identified by their names
1541 * trace contains the full platform hierarchy exactly as declared using the ASes
1542 * Options tracing/msg/[task|process]:1 groups the process by hosts
1543 for both cases, tasks and processes must have names that are unique during the simulation
1544 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1545 * The experimental option tracing/msg/volume is deprecated
1546 its functionality may be reincorporated if needed
1548 The tracing generates a trace file with unordered timestamped events,
1549 because of the way the core simulator (surf) works. A script available
1550 at the tools directory (fix-paje-trace.sh) can be used to put the events
1551 in order. We have changed the tracing so it can generate ordered timestamped
1552 events in the final trace, but depending on the simulator (and how much time
1553 is simulated) that can lead to a huge memory utilization. It is deactivated
1554 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1556 Build Infrastructure
1557 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1558 - We are trying hard to keep the API stable, but it may happen that
1559 some things change (we're a research project after all, not a
1560 nuclear plant operating system). If such things should happen, you
1561 could rely on that macro to adapt.
1562 - current value: 30600 for 3.06.00, aka 3.6
1563 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1564 a number that can be compared with SIMGRID_VERSION.
1565 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1566 assertions and verbose/debug logging events are disabled at compile time.
1568 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1570 SimGrid (3.5) stable; urgency=medium
1573 * New feature to any SimGrid-based simulator: Model-Checking
1574 Check SIN#1 for more details.
1577 * New Model SMPI (three-interval linear regression for correction factors)
1578 See RR-7426, available at http://hal.inria.fr/inria-00527150
1579 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1580 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1581 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1582 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1583 * See also src/smpi/README
1588 - Tracing API changes: TRACE_start and TRACE_end should not be called
1589 by user-code. They are automatically called by simulators created
1590 with SimDAG, MSG and SMPI if the toolkit is compiled with
1591 tracing_enabled=ON. Categories declaration and utilization remain the
1592 same for MSG and SimDag.
1593 - A function was added to the tracing API to declare categories with
1595 - TRACE_category_with_color (char *category, char *color)
1596 where color must be in the following format
1597 "%f %f %f", red, green, blue
1598 and red, green, blue are float values in the interval [0, 1]
1599 - User can specify NULL as color parameter, or continue calling
1600 TRACE_category (cat)
1601 On that case, the tracing system will define random colors
1602 - The following command-line options are supported:
1603 --cfg=tracing/filename:msg.trace
1604 --cfg=tracing:1 (activate tracing, needed to use others)
1605 --cfg=tracing/platform:1 (categorized resource use)
1606 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1607 --cfg=tracing/msg/task:1 (task creation)
1608 --cfg=tracing/msg/process:1 (process creation, migration)
1609 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1610 --cfg=tracing/smpi:1 (SMPI interface tracing)
1611 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1612 - examples of examples/msg/tracing updated
1614 - DAXLoader and DOTLoader functions can generate tasks with categories
1615 - A new function to attribute a category to SD tasks:
1616 TRACE_sd_set_task_category (SD_task_t task, char *category)
1617 * Tracing the MPI interface implemented by SMPI
1618 - Collective operations are traced with states
1619 - Point-to-Point operations are traced with states/links
1620 - Tracing activated by a parameter "-trace filename" passed
1621 to smpirun during execution (considering that simgrid
1622 is compiled with tracing enabled)
1623 - To run the simulation with gdb, the simulator
1624 accepts --cfg=tracing/smpi:1 to trace SMPI
1625 - tesh files to check if smpi tracing is ok
1626 - See examples/smpi/NAS/DT-trace
1627 * GTNetS tracing re-worked
1628 - adaptation to the tracing system of GTNets to cope
1629 with modifications regarding the fullduplex mode
1630 - new tesh files to check if gtnets tracing is ok
1633 * Asynchronous communications through the functions:
1634 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1635 * New function: MSG_load_platform_script()
1636 to make possible using a lua script instead of XML files to set up platforms
1637 * New function: MSG_set_function
1638 to associate functions to processes, used when bypassing the parser
1639 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1641 Platforms: Add some more examples in examples/platforms
1642 * Grid'5000: see www.grid5000.fr
1643 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1646 * Change the XML format. This is a very important modification. SimGrid
1647 3.5 introduces a new hierarchical format based on the notion of
1648 Autonomous Systems. Compatibility with old format is ensured
1649 through the perl script provided in the install bin directory
1650 bin/simgrid_update_xml.
1651 It is now possible to build platforms with specific routing
1652 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1653 connect several platforms together. We will try to provide soon
1654 set of realistic platforms exploiting these properties (have a
1655 look at examples/platforms/ for the moment).
1656 * Take the opportunity of the XML format change to be a good XML citizen:
1657 rename link:ctn to link_ctn and similar changes (also dealt with by
1659 * Add a new routing scheme (rule-based) using regular expressions. It
1660 enables to have an extremely low memory footprint when the
1661 underlying routing is simple and can be compactly described. You
1662 need to have libpcre4-dev (perl regular expressions) installed if
1663 you want to use this routing scheme.
1664 * Revive the cluster TAG and allow to easily and efficiently (both in
1665 term of memory and speed) connect clusters together. Have a look
1666 at teshsuite/simdag/platforms/ to see how this can be done. With
1667 this tag, you can create clusters with thousands of tasks at no
1668 cost (have a look at examples/platforms/).
1669 Note: clusters are implemented as ASes, so there is no need for an
1670 enclosing AS tag if you have only one cluster in your platform.
1671 * Add new generic functions in the public interface that allows the user
1672 to call SURF 'create_resource' methods from your code (same
1673 functionality as the XML bypass mechanism but with a much lighter
1675 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1676 that uses a piecewise linear approximation to produce better
1677 results when exchanging small messages.
1678 * Add a new parameter to handle correctly full duplex link and account
1679 for interferences between uplink and downlink communications
1680 (activate with --cfg=fullduplex:1).
1683 * Rename the SD_READY (all dependencies are satisfied and task is
1684 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1685 dependencies are satisfied) state.
1686 This prevents a confusion between the notion of "ready to schedule"
1687 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1688 simulated" (SD_RUNNABLE) used by the simulation kernel.
1689 * Change the way a task is considered as ready. Instead of removing
1690 dependencies when a task is done, a counter is decreased. This way,
1691 it is always possible to reach ancestors thanks to the
1692 SD_taks_get_parents function (even after the end of the simulation.)
1693 * Change the return type of SD_Simulate from (SD_task_t*) into
1694 xbt_dynar_t. This function was in handling a dynar internally and
1695 converted it into a NULL terminated array for historical reasons.
1696 * New function SD_dotload(char*) to load a DAG described in dot
1697 format. This loader and the corresponding examples require the
1698 installation of the graphviz library.
1699 * Fix a bug in the management of tasks of size 0 in the surf network
1700 models. This problem was only visible with SIMDAG and you should
1701 thus disregard results produced with earlier versions if you
1702 relied on this feature (some tasks were blocked because of this).
1703 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1704 even though your applications were doing only point-to-point
1705 communications and sequential computations. Now you can really use any
1706 model you want (of course, if you create real parallel tasks, which are
1707 not implemented in most models beside ptaskL07, this will abort).
1708 * Add an example that schedules a DAX on an heterogeneous platform
1709 using a Min-Min strategy.
1710 * New function SD_workstation_get_current_task() that returns the kind
1711 of task currently running on a workstation in the sequential access
1713 * Raise some warnings when unexecuted tasks remains at the end of the
1714 simulation. This is usually caused by cycles in the DAG.
1717 * New function: SIMIX_process_set_function() called by MSG_set_function
1718 * Change the underlying waiting queue in semaphores so that a process
1719 can wait on several of them simultaneously (as in waitany).
1720 * Fix the way to handle tokens in semaphores so that all access patterns
1721 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1722 * kill the dirty pimple SIMIX_message_sizes_output()
1723 Please use (proper) visualization instead
1726 * New data container: setset (set of sets of elements)
1727 * New module: mmalloc (mapped malloc, allowing to have several
1728 independent segments of malloc)
1729 * New function: xbt_dict_cursor_set_data()
1730 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1731 * New function: xbt_dynar_is_empty()
1732 * New function: xbt_fifo_get_last_item()
1733 * Fix xbt_dynar_shrink(): use the right element size.
1734 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1735 uninitialized areas during expand.
1736 * Fix semaphores: previous implementation was severely broken.
1737 * Use library init/fini functions for our initialization.
1738 - you can use logs and other feature as soon as you want in your
1739 code (even before the xbt_init / MSG_init)
1740 - xbt_exit is now a no-op and produce a warning when used.
1743 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1744 benefit from the latest improvement to the simulation kernel.
1745 * Kill measurement sockets for now. If you rely on them, sorry. This
1746 release is not for you. This feature will be reintroduced in the
1747 future, but we cannot delay the release any further.
1748 * New function: gras_msgtype_get_name().
1749 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1750 * Fix (at last) the pmm example: it should not randomly fail anymore.
1752 Build chain: bug fixes and overall polishing
1753 * Cmake is now stable enough. Hence, we killed the autotools.
1754 * Port to windows ( TM :)
1755 * Fix the 'make install' target.
1756 No need to use 'make install-simgrid' anymore
1757 * Introduce a 'make dist' target compiling a *source* archive
1758 'make package' compiles a binary archive
1759 * Compile java files only on need
1760 * Add --cd and --setenv command line options to tesh
1761 * Out of source builds are not fully supported yet, but we are close
1762 * Enable supernovae and optimization flags by default for our users
1765 * Add layer to set up environment directly from lua, without XML.
1766 * The effect of gras_stub_generator can be achieved through
1767 lua too (check examples/gras/console/ping_generator.lua)
1769 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1771 SimGrid (3.4.1) stable; urgency=low
1773 The "Polishing easter eggs is probably a good idea" release.
1774 This is a bug fixes release only.
1778 * Fix a bug preventing the tasks from begin garbage collected.
1781 * Fix a bug occuring when a host involved in a communication fails.
1782 This was not detected properly by the other peer involved in the
1783 communication. Now, it's reported as a network error.
1786 * Warn the user about loop dependencies in data flow of DAX files
1787 * Obey the control-flow dependencies of DAX files
1790 * Add option "enable_smpi" allowing to not compile SMPI.
1791 Probably useful for the (Mac) users experiencing a build error here
1792 * Improve the detection of lua5.1 and ruby1.8
1794 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1796 SimGrid (3.4) stable; urgency=low
1798 The "Easter in Cargese" release. Also known as (major changes):
1800 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1801 ~> bindings were greatly improved
1802 ~> new tracing infrastructure for better visualization introduced
1804 * the "Welcome to configury modernity" release.
1805 ~> we switched from autotools to cmake, and improved our cdash
1808 A more detailled list of changes follow (full detail in svn log).
1810 Java Bindings: Various Cleanups
1811 * (install java-gcj-compat-dev on debian-like to use them)
1812 * Remove put/get: no need to export deprecated interface in Java
1813 Use send/receive instead.
1814 * Cleanup the examples and add a README per directory
1815 * Remove example autoDestination (that's the only way to go now)
1816 * Remove example explicitDestination (was a plain copy of basic)
1817 * Make JniException a runtime exception, so that there is no need to
1818 declare the fact that you may encounter such a beast. I guess that
1819 nobody will ever want to survive such error.
1820 * Create specific errors for each MSG case of failure:
1821 host failure, transfer failure, timeout, task cancelled
1822 * Cleanup the exceptions that may get thrown by each function
1823 * Other internal cleanups in Java bindings. Performance still bad :/
1824 Ruby and Lua Bindings: create them
1825 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1826 * That's new and great, you should try them out.
1827 Same functionalities than Java bindings, only even less polished
1829 * Kill the useless "rate" argument of SD_task_get_execution_time()
1830 Everyone used to provide -1 as a value, it was not used, and the
1831 semantic of a possible use wasn't even clear.
1832 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1833 either as comm costs or compute costs to mean that there is no
1834 such thing for that specific task.
1835 * Add a SD_task_set_name() function
1836 * Fix SD_task_unschedule() on typed tasks
1837 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1838 * In DAX loader, accept useless 'level' attributes to <job> since
1839 LIGO DAGs have them (seem to be to ease graphical representation).
1841 * Add an example masterslave_mailbox.c using send/receive and not
1842 the deprecated put/get interface.
1843 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1844 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1846 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1847 (declare MSG_USE_DEPRECATED to still have the old name)
1848 * Add a MSG_task_set_data() function
1849 * About trace replay (see examples/msg/actions):
1851 - Allow to work with splitted trace files for each process
1852 Give the specific trace file as argument of each process,
1853 and call MSG_action_trace_run(NULL)
1854 You can still have one merged file for all processes.
1855 - Fix implementation of collective operations
1856 * Allow task_execute() on 0-sized tasks (closes #10063)
1858 * This is the first release of SimGrid where SMPI is not considered
1859 beta anymore (even if some corners should still be improved)
1860 * Port over the new SIMIX_network submodule (internal refactoring)
1861 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1862 * Implement more missing elements of the standard:
1864 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1866 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1867 MPI_Scan() MPI_Get_processor_name()
1868 - Added implementation of missing case for Alltoall (warning: it's
1869 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1870 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1871 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1872 SMPI_MPI_Allgatherv()
1873 * Bug fixes include:
1874 - MPI_Waitsome() was broken
1875 - Allow relative includes in smpicc
1876 - Command line cfg argument 'reference_speed' was ignored...
1877 - Some functions did not properly lead to auto-benching of user code
1878 - smpicc passes -O2 by default (just like openmpi one)
1880 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1881 * Bug fixes about timeouts during communications
1882 * add SIMIX_message_sizes_output() as a pimple to write to file the
1883 amount of messages per size. Use gnuplot to get histogram.
1884 Pimple because that's the only user-visible function of simix,
1885 defined directly in xbt.h (irk, sorry)
1887 - Add a SIMIX_sem_get_capacity() function
1888 - Fix interactions with processe resume/suspende
1889 - release_forever() was stupidly broken
1890 - Fix SIMIX_display_process_status() for processes in a semaphore
1891 - Make SIMIX_sem_block_onto() user-visible
1892 * Refactoring context stuff:
1893 - Use pseudo-OOP for better modularity
1894 - reimplement SIMIX_process_kill() without process_schedule() so
1895 that the latter can take as invariant that it is called from
1897 - Merge context_start into context_new for sake of simplicity
1899 * Add a Vivaldi network model, coded live during SUD'10 ;)
1900 * Rename configuration variables to start a hierarchy:
1901 o cpu_model -> cpu/model
1902 o network_model -> network/model
1903 o workstation_model -> workstation/model
1904 * New configuration variables:
1905 o network/bandwidth_factor: correction to bandwith
1906 o network/latency_factor: correction to latency
1907 o netwotk/weight_S: correction to the weight of competing streams
1908 * Add a long description to the models, that users can see with such
1909 argument on the command line: --cfg=cpu/model:help
1910 * --help-models display the long description of all known models
1912 * config: add the ability to set a default value after registration
1913 Does not override any previously set value (e.g. from cmd line)
1914 * dict: allow to have integer key and data.
1915 When so, you need to use the following functions
1916 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1917 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1918 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1919 In contrary to regular dicts, the key is not malloced before copy.
1920 Mixing scalar and regular elements in the same dict is not tested
1922 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1923 Tracing for Visualization:
1924 * SimGrid is now instrumented in order to generate a trace file for
1925 visualization analysis: to use it, need to compile SimGrid with the
1926 "tracing" option enabled, and instrument the program using SimGrid with
1927 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1928 (among other functions).
1929 * The instrumentation only traces the platform utilization for now
1930 * Documentation to use the tracing functions and how to analyze the
1931 traces with the Triva tool is written.
1932 * More information about: SimGrid FAQ (in the section Tracing Simulations
1935 * We moved to cmake as default build system. Autotools support will
1936 be dropped soon. Check the FAQ for more info about how to use it.
1937 * Greatly improved our cdash/ctest interactions
1938 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1939 * Added memory checking tests with valgrind; lot of memleak fixing.
1940 This may be the first release of simgrid with so few memory issues
1941 * Added code coverage tests.
1942 Our coverage is still improvable, but at least we see it on cdash.
1944 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1946 SimGrid (3.3.4) stable; urgency=low
1948 The "Desktop Grid needs love too" release (also called Xmas release).
1950 Models improvements:
1951 * Major speedup in the maxmin system solving by using lazy evaluation
1952 Instead of solving completely the maxmin system at each iteration,
1953 only invalidate (and recompute) the modified parts.
1954 This new feature is enabled in default models but you can try to
1955 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1956 * Cas01 IMproved as default CPU model
1957 This CPU model is the same Cas01 model, but it uses the
1958 maxmin-selective-update flag and a heap structure to manage
1959 actions on SURF kernel.
1960 It reduces the complexity to find the next action to finish and,
1961 consequently, it's faster than the old Cas01.
1962 This is the new default CPU model (Cas01).
1963 * Rename the old Cas01 model to Cas01_fullupdate
1964 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1965 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1966 * CpuTI (CPU Trace Integration)
1967 A new CPU model whose objective is simulate faster when using
1968 availability trace files.
1969 Instead of using a full featured, over engineered maxmin system for
1970 CPU modeling, this model does the pre-integration of traces files
1971 to calculate the amount of CPU power available, and so, executes
1972 faster than the old CPU models.
1973 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1974 * Use LV08 as default network model since it gives better accuracy
1975 for small messages and shouldn't change things for big ones.
1976 Use --cfg=network_model:CM02 to get the previous behavior.
1979 ******************************************
1980 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1981 ******************************************
1982 * The new CPU model may changes simulations!
1983 The point is that events occurring at the exact same timestamp
1984 are not scheduled in the same order with the old and new
1985 version. This may be enough to completely change the execution
1986 of simulations in some cases.
1987 * The new network model will change simulations!
1988 This new model is more realistic than the previous one, so you
1989 should consider redoing your old experiments with this model.
1990 Sorry for the inconvenience.
1993 * Introduce the supernovae compilation mode
1994 When compiled that way, the whole SimGrid (or almost) is put in a
1995 single compilation unit and compiled in one shoot.
1996 This is to help gcc which has difficulties to inline stuff from one
1998 The speedup seem to be above 15%, althrough more tests are needed on
1999 amd64 to confirm that gain.
2002 * Port of MSG's mailbox on top of SIMIX network
2003 The put/get mechanism was greatly simplified on the way.
2006 * New SIMIX network module. Provides:
2007 - Mailbox: rendez-vous mechanism to find with who you want to speak
2008 - Synchronous send/recv: easier and hopefully faster since the
2009 logic is handled in the maestro process directly now
2010 - Asynchronous send/recv: you dreamt of it? It's here now
2011 Too bad that nobody cared enough to propagate the change to MSG.
2012 * Add semaphores as SIMIX synchronization mechanism.
2015 * new function SD_daxload(char*) to load a DAX file
2016 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2017 * Introduce typed tasks. Specify its kind and cost at creation.
2018 At scheduling, just give where it should be placed, and the cost
2019 for each involved resource is automatically computed.
2020 Existing constructors so far (more to come of course):
2021 - SD_task_create_comm_e2e() for end-to-end communication
2022 - SD_task_create_comp_seq() for sequential computation
2023 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2024 * new function SD_task_dump() for debuging display
2025 * new function SD_task_dotty(task,FILE*) writing to file the info
2026 about the task in dotty format
2027 * SD_task_dependency_exists() can now cope with having one of its
2028 arguments NULL. If so, it tests whether the other argument has any
2030 * Add getters on list of preceding/following tasks:
2031 SD_task_get_parents(task) and SD_task_get_children(task)
2032 * Add getters on amount of workstations and list:
2033 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2034 * Add getter on task kind: SD_task_get_kind(task)
2035 * Update the start_time and finish_time of tasks on completion/failure
2036 * Bugfix: Remove task from state swags when destroyed
2039 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2042 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2043 dynar with one separate thread per value of the dynar.
2044 * Change the prototype of xbt_thread_create(), sorry.
2045 Added a boolean parameter indicating whether we want to join this
2046 thread (used in SG only for now)
2047 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2050 * GTNetS wrappers should now be usable again (and betterly tested too)
2051 * Fix a major regression from 3.2 where the timeout provided to
2052 MSG_task_put_with_timeout() was used as absolute time before which
2053 the comm should be done.
2054 * Start to fix the <cluster> tag.
2055 - Internal links should be good now (beside of the loopback, which
2056 use the private link instead)
2057 - paths to the external world is still rather broken
2058 - the <route:multi> tag is just broken. Actually that's brain-dead.
2059 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2060 to make it less stupid
2061 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2062 * Fix a source-level compatibility glitch from 3.2: after defining
2063 MSG_USE_DEPRECATED, you can use the old name
2064 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2065 * Allow to compile from the SVN with automake 1.11
2066 * Fix some problems when using the "start_time" tag in deployment XMLs.
2067 * Fix #8569: XBT/synchro.h has redundant declarations
2068 * Fix #8563: MSG return values and exceptions
2069 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2070 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2071 Thanks to Robson Peixoto.
2072 * Fix a vicious bug in dictionaries inducing that some elements were
2073 not freed on xbt_dict_free()
2075 Portability report of this version:
2076 * Main portability targets:
2077 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2078 on (amd64/i386/ia64)
2079 - mac leopard on i386
2080 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2081 but nothing critical.
2082 * Other platforms: windows, AIX and others were not tested for this release
2084 Timing report of this version:
2085 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2086 scenario parameters). From 8h to a few seconds in desktop grid settings.
2087 * Supernovae brings about 25% speedup on i386.
2089 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2091 SimGrid (3.3.3) stable; urgency=low
2093 The "Need for Speed" release.
2095 The timings done to validate the 3.3.2 were faulty.
2096 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2098 The problem was a conversion from a manually handled vector to
2099 xbt_dynar_t on the critical path.
2100 xbt_dynar_foreach calls functions, inducing stack management crap.
2102 We inlined these functions and xbt_dynar_foreach is now breath taking.
2103 We also inlined xbt_swag_belong on the way.
2105 Here are some approximate speedup measurements (on master/slaves
2106 simulations lasting between 10s and 20s each):
2107 3.3.1 -> 3.3.2: about same performance
2108 3.3.2 -> 3.3.3: 40% speedup
2109 3.3.1 -> 3.3.3: 40% speedup
2110 3.3.1 with inline patch -> 3.3.3: 30% speedup
2112 Our reading is that the refactoring which occurred in 3.3.2 made us
2113 suffer much more from the xbt_dynar_foreach low performance, but
2114 once we solved this, this refactoring proved to be very performance
2115 effective. From the 40% speedup, somehow, 10% are due to the
2116 inlining and 30% to the refactoring.
2118 That's a pitty that gcc cannot inline functions placed in other files
2119 alone. We have to choose between:
2120 - break the encapsulation (by putting private data structures and
2121 accessors in headers files to help gcc)
2122 - live with low performance
2123 - switch to a decent compiler such as icc (not quite possible).
2125 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2127 SimGrid (3.3.2) stable; urgency=low
2129 The "Simplicity does not preceed complexity, but follows it" release.
2131 The main contributors of this release were (lexical order):
2132 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2135 * Extract the routing logic into its own object.
2136 (was dupplicated in network.c and workstation_LV07.c;
2137 Allows to implement other ways of storing that info)
2138 => kill now useless network_card concept
2139 - Use dynar to represent routes (instead of void** + int*)
2140 - kill link_set (use surf_network_model->resource_set instead)
2141 - Add a command-line option to choose the routing schema to use
2142 - Add three new models:
2143 * Floyd (shortest path computed at initialization)
2144 * Dijikstra (shortest path recomputed all the time)
2145 * Cached Dijikstra (shortest path computed on need)
2146 All these models where contributed by Silas De Munck, and are
2147 described in his ICCS09 paper.
2149 * Simplify model declaration
2150 (less redirections, less function to write when defining a model)
2151 - Factorize stuff between models:
2154 surf_model_resource_set(model)
2155 surf_model_resource_by_name(model, name)
2156 - Unify the types of models in s_surf_model_t (using an union)
2157 - Embeed fields of common_public directly into s_surf_model_t
2158 - Rename model methods:
2159 action_free ~> action_unref
2160 action_change_state ~> action_state_set
2161 action_get_state ~> action_state_get
2162 - Change model methods into functions :
2163 (model)->common_public->action_use ~> surf_action_ref
2165 * Implement a generic resource; use it as ancestor to specific ones
2166 (allows to kill duplicated code in models)
2167 Drawback: timer command don't need no name nor properties;
2168 workstation_CLM03 don't need no properties
2169 (but I guess we can live with those few bytes wasted)
2171 * Improve the action object model
2172 - implement a constructor avoiding dupplicated code about field
2173 initialization in generic_action part.
2175 * Kill the SDP model: it has an external dependency, is deprecated
2176 in flavor of modern lmm models, and didn't compile since a while
2179 * Relocation of the context module from XBT to SIMIX.
2180 (the context were decoupled from the simix processes, duplicating a lot of code)
2181 => a lot of code was factorized
2182 - less overhead is introduced during scheduling
2183 - simpler API for the context factory
2184 - the logic for process creation,destruction and manipulation was simplified
2185 * Simplification of the s_smx_process_t data structure.
2186 => accesing the simix level data associated to a process is faster now,
2187 and the code is a lot more readable.
2190 * Implement some more MPI primitives:
2191 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2192 -implementation: Bcast: flat or 2-ary tree (default),
2193 Barrier: 4-ary tree,
2195 Allreduce: Reduce then Bcast
2196 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2197 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2198 Alltoallv: flat tree, like ompi
2200 * Add support for optimized collectives (Bcast is now binomial by default)
2201 * Port smpirun and smpicc to OS X
2204 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2205 Shout out if you used it.
2208 * Display the list of still queued messages in SG mode when existing
2212 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2213 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2214 * Add xbt_str_from_file(FILE*)
2215 * Add xbt_dict_get_key achieving a linear reverse search
2216 * Remove the context module
2218 Portability report of this version:
2219 * Main portability targets:
2220 - Linux(debian)/x86/context
2221 - Linux(debian)/x86/pthreads
2222 - Linux(debian)/amd64/context
2223 - Linux(debian)/amd64/pthreads
2224 On these, we still have the eratic breakages of gras/pmm and
2225 amok/saturate_sg reported in previous version. We still think
2226 that the tests are the cause of the fault, not the tested code.
2228 - Mac OSX Leopard/x86/context
2229 Still false negative in tesh autotesting.
2230 Smpi still fails, but this time because readlink does not accept -f
2231 Everything seems to work properly beside of that.
2234 - AIX version 5.3 (only tested contexts this time)
2235 Smpi still fails there because mktemp is not installed.
2236 Everything seems to work properly beside of that.
2238 I managed to compile it for the first time, but several breakages.
2239 Won't delay the release for this exotic platform.
2241 * Windows: it's still lagging behind. If you want to help, please
2244 Timing report of this version:
2245 This version seem to be more than 5% faster than 3.3.1 (on linux
2246 64bits with contextes). The gain is less than expected, we are
2247 investigating this for next release.
2249 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2251 SimGrid (3.3.1) stable; urgency=low
2254 * Implement a --cfg-help to show existing configuration variables
2255 * Build chain do not require doxygen in maintainer mode
2258 * fix a bug on struct sizeof computation, which prevented the
2259 exchange of arrays of structs in some conditions
2260 - added a regression test about this in datadesc_usage
2261 * Allow the exchange of 0-long dynamic vectors.
2262 - for that, use -1 as indicator of dynamic size instead of 0
2263 - This implied to change any size from unsigned long to long,
2264 reducing a bit communication abilities, but I guess that with
2265 64bits being quite common, this is more than enough.
2266 - This also induce a protocol change, thus bumping network protocol
2267 version from 0 to 1 (if we have external users, we have to get
2268 clean on that point too ;)
2269 - added two regression tests about this in datadesc_usage
2270 * Be more verbose when propagating local exceptions
2271 This helps debugging.
2272 * Display the status of simulated processes when receiving SIGINT in
2276 * Allow to control the simulation from a trace file.
2277 New functions MSG_action_register() and MSG_action_trace_run()
2278 The first one allows to associate a function execution to each
2279 kind of action while the second one parses a trace file and
2280 triggers the corresponding actions within the system.
2281 For now, only a toy example is provided in examples/msg/actions
2282 * Add an exemple of process migration in examples/msg/migration
2283 * Fix a bug in task exchange which broke MSG_task_get_sender()
2284 Add a teshsuite regression test for that.
2285 [Bug: if MSG_task_get_sender() is called after sender exit,
2287 * Fix a bug which prevented suspend/resume to work properly
2288 * Display the status of simulated processes when receiving SIGINT
2289 This fixes a regression of v3.3. due to the introduction of SIMIX
2290 * Bug fixing in failure management:
2291 - trace could not start by a failure at time 0
2292 - failure during communications were not working
2295 * Add SIMIX_process_set_name() to change the name of the current
2296 process in the log messages.
2297 * Store smx_hosts in a dict since we only retrieve them by name
2298 * Move the configuration infrastructure to surf
2301 * Move the configuration infrastructure to surf
2304 * Massive internal cleanups:
2305 - Store internal structures on processes instead of hosts (allows
2306 to have more than one process per host, in addition of being more
2308 - Cleanup the initialization/finalization process
2309 - Kill a whole bunch of unneeded synchronization:
2310 processes run in exclusive manner within the simulator
2311 - Move queues from global tables to process data fields
2313 - now accept -platform and -hostfile arguments
2314 - Pass the right rank value to processes according to the hostfile
2315 * Compile the examples by default, and use them as regression tests
2316 * Implement MPI_Wtime()
2317 * Change the reference speed to a command line option
2320 * TCP_gamma can now be specified as command line option using
2321 --cfg=TCP_gamma:10000000.0
2322 * Change the --surf-path cmd line option into --cfg=path:
2325 * Also include strbuff from xbt.h public header
2326 * xbt_ex_display(): do not free the exception after displaying
2327 This allows to do more with the given exception afterward.
2328 Users should call xbt_ex_free() themselves.
2332 Portability report of this version:
2333 * Main portability targets:
2334 - Linux(debian)/x86/context
2335 - Linux(debian)/x86/pthreads
2336 - Linux(debian)/amd64/context
2337 - Linux(debian)/amd64/pthreads
2338 These targets fail about 1/10 of times on gras/pmm, but we believe
2339 that this is because of the test, not because of simgrid.
2340 amok/saturate_sg fails even more rarely, and the test may not be
2343 - Mac OSX Leopard/x86/context
2344 The test suite still spits tons of errors because some obscure
2345 force prevents us from removing the temporary directories
2346 arguing that they still contain some metadata I've never heard of.
2347 Smpi fails because seq is not installed.
2348 Everything seems to work properly beside of that.
2351 - AIX version 5.3 (both contexts and pthread)
2352 Smpi still fails there because mktemp is not installed.
2353 XML inclusions seems rosty on AIX.
2355 * Windows: it's still lagging behind. If you want to help, please
2358 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2360 SimGrid (3.3) stable; urgency=high
2364 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2365 [Malek Cherier & Mt]
2367 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2370 Until now, GRAS were using MSG as an interface to SURF. It was
2371 quite difficult because both interface have several differences
2372 (MSG channels vs GRAS sockets were the most notable point).
2374 This also opens the gate to SMPI (which should occur soon) and speed
2375 up simulations by to 40% (even if it were not the main goal).
2377 **************************************
2378 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2379 **************************************
2380 The point is that events occuring at the exact same timestamp are
2381 not scheduled in the same order with the old and new version. This
2382 may be enough to completely change the execution of simulations in
2383 some cases. Sorry for the inconvenience.
2385 * Cleanup and upgrade the XML format to push further scalability
2386 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2388 * Improve the testing infrastructure with tesh. Now a very large part of
2389 the code is tested not only by being run but also by checking that the
2390 output match an expected output [Mt].
2392 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2393 is really less memory-demanding, which should allow you to use
2394 larger files in SimGrid [AL].
2396 * Inform valgrind about our contextes, so that it becomes usable
2397 with the default (and more effecient) version of SimGrid
2398 [contributed by Sékou Diakite, many thanks]
2401 * Introduce a listener thread in charge of receiving incoming
2402 messages from the network. It allows to overlap communication and
2403 computation but most notably, it removes some stupid deadlocks due
2404 to the fact that so far, a process could not send and receive at
2405 the same time. This made most non trivial communication schema
2407 * Convert the PIDs from long int to int to match the MSG ones (and
2408 linux ones too) [Mt]
2409 * New function: gras_agent_spawn() to launch a new process on
2410 current host. Only working in simulation for now. [Mt]
2411 * New function: gras_os_hostport() returning a constant form (ie,
2412 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2415 * Make the backtrace of exceptions more human readable [Mt]
2416 * New module: xbt/str [Mt]
2417 a ton of string utility functions (split, join, printf to a newly
2418 allocated buffer, trim, etc)
2419 * New module: xbt/hash [Mt]
2420 SHA1 hashing algorithm (more to come if needed)
2421 * New module: xbt/synchro [Mt]
2422 synchronization tools (mutex and conditions) working the same way
2423 in simulation and in real life (mainly useful for GRAS, but not
2425 * New module: xbt/queue [Mt]
2426 classical producer/consumer synchronization scheme
2427 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2428 (using the classical functions will get serialized) [Mt]
2429 * Make dictionary internal table dynamic. No need to specify its size
2430 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2432 * Make sure the log channels are organized as a tree under windows
2433 (because of ANSI C compatibility issue, any channel were child of
2434 root directly) [Mt].
2437 * Cleaned many thing in surf and fixed a few bugs [AL].
2438 * Add a nice command line configuration mechanism to compose models [AL].
2439 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2440 the previous one (KCCFLN05). It relies on something that looks like
2441 a max-min sharing mechanism but cannot be written as such. A new solver
2442 was thus designed [AL].
2443 * Add a new solver to lmm. Based on Lagrange optimization and
2444 gradient-based descent, it enables to efficiently maximise systems s.a
2446 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2448 This solver enables to propose two new network models for TCP Reno and
2449 TCP Vegas based on Low's work. These models still need to be fully
2450 tested though [Pedro Velho].
2453 * Bug fix in SD_simulate. Now the time bound given as argument is
2455 * Use the new parallel task model (ptask_L07) as default.
2456 * Use the SURF command line configuration mechanism.
2457 * 0-size tasks (for synchronization) should now work.
2459 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2461 SimGrid (3.2) stable; urgency=high
2465 We still experience issues on this platform, but we believe that at
2466 least MSG is usable.
2468 GRAS API BREAKAGE (for simplification purpose, sorry):
2469 * the gras_msgtype_by_name is not used anymore. Instead of
2470 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2471 you can write (and must)
2472 gras_msg_send(toserver, "request", &request);
2473 - If you still want to pass a gras_msgtype_t to the function (to cache
2474 the type and avoid the lookup time), use the gras_msg_send_() variant.
2475 - Impacted functions:
2476 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2477 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2478 * The callbacks are now expected to return 0 when everything went well
2479 (just like the main() function)
2481 GRAS new features and improvements:
2482 * New module mechanism where user code can use per process globals [Mt]
2483 This is similar to gras_userdata_*() functions, but for libraries. It
2484 factorize some code developped over and over in the examples and AMOK.
2485 It has still to be documented and used (only amok/peermanagement is
2487 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2488 loss when they were small enough to fit into the buffer and sent quickly
2489 enough so that they can all get received in one shoot.
2490 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2491 if not found. Use the *_or_null() variant for the old semantic.
2492 * In gras_msg_handle, do not discard messages without callback.
2493 They are probably messages to be explicitly awaited later (ie, proofs of
2494 mis-synchronization in userland since they are sent before being awaited)
2495 No big deal usually.
2496 * gras_socket_meas_send/recv: semantic changed!
2497 The numerical arguments used to be (1) the total amount of data to send
2498 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2499 messages. This was need for the fool willing to send more than MAXINT
2500 bytes on quite fat pipes.
2503 * Do really rename the hostmanagement module to peermanagement. [Mt]
2504 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2505 but this is rather new and this was documented in the module
2506 documentation (poor excuses, I admit)
2507 * Bandwidth measurement semantic changed! This follows the changes to
2508 gras_socket_meas_send/recv explained above.
2511 * A sequential mode has been added to the workstations. When a workstation
2512 is in sequential mode, it can execute only one task, and the other tasks
2513 are waiting in a FIFO. [Christophe Thiery]
2516 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2517 model for SIMDAG. [Christophe Thiery]
2518 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2519 equal to 0, which caused some bad behaviors in
2520 saturated_constraint_set_update. I now use a threshold mechanism like in
2524 * When running manually src/testall, you select specific units [Mt]
2525 testall is the result of our cunit mechanism, and should replace all
2526 the scripty thingy around since bash don't run easily on billware.
2528 * A mallocator system has been added. [Christophe Thiery]
2529 Mallocators allow you to recycle your unused objects instead of freeing them
2530 and allocating new ones.
2532 Documentation update:
2533 * FAQ reworking + New FAQs:
2534 - "Valgrind spits tons of errors!" [Mt]
2535 - "How to repport bugs" [Mt]
2536 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2537 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2538 - Communication time measurement within MSG [AL]
2539 - I experience weird communication times when I change the latency [AL]
2540 * GRAS tutorial [Mt]
2542 - an introduction to the framework and to the used communication model
2543 - an initiatic tour introducing the most proheminent features:
2545 . Lesson 0: Installing GRAS
2546 . Lesson 1: Setting up your own project
2547 o Part 2: Message passing
2548 . Lesson 2: Exchanging simple messages
2549 . Lesson 3: Passing arguments to the processes (in SG)
2550 . Lesson 4: Attaching callbacks to messages
2551 . Lesson 5: Using globals in processes
2552 . Lesson 6: Logging information properly
2553 . Lesson 7: Using internal timers
2554 . Lesson 8: Handling errors through exceptions
2555 . Lesson 9: Exchanging simple data
2556 . Lesson 10: Remote Procedure Calling (RPC)
2557 . Lesson 11: Explicitely waiting for messages
2558 . Recapping of message passing features in GRAS
2559 - A HOWTO section containing:
2560 o HOWTO design a GRAS application
2561 More are due, of course. They will come latter. In the meanwhile, you can
2562 check the examples which are still here.
2564 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2566 SimGrid (3.1) stable; urgency=high
2570 There was a stack corruption somewhere, visible only when optimizing
2571 with these versions. [Vince]
2574 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2575 module that enabled to play with Directed Acyclic Graphs. It is built
2576 directly on top of SURF and provides an API rather close to the old
2577 SG. Some old codes using SG are currently under rewrite to check that
2578 all needful functions are provided. [Christophe Thiery]
2581 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2582 extension of the classical CLM03 model that gracefully handles
2583 failures. This is now the default model for MSG and GRAS. It doesn't
2584 handle parallel tasks yet though. [AL]
2585 * Bug fix: Weights were not correctly set in the network part.
2586 WARNING: This may have resulted in incorrect results with simulations
2587 where there are more than one flow on a given link. [AL]
2590 * After a (long ?) discussion on simgrid-devel, we have decided that the
2591 convention we had on units was stupid. That is why it has been decided
2592 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2593 WARNING : This means that all previous platform files will not work as
2594 such with this version! A warning is issued to ask users to update
2596 A conversion script can be found in the contrib module of the CVS, under
2597 the name contrib/platform_generation/surfxml_update.pl [MQ]
2600 * Bug fix: Processes were started in reverse order, wrt deployment file.
2601 WARNING: if your code relies on this bug, please fix it. [AL]
2602 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2603 being executed on two different locations. [AL]
2604 * Bug fix: Failures are now better supported thanks to Derrick's tests
2605 (there was many failure situations I hadn't thought of and that weren't
2606 correctly handled). [AL]
2607 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2611 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2612 Exception occurring on server-side are propagated back to client (!).
2614 API CHANGE: the callback changed their prototype. Change:
2615 int my_handler(gras_socket_t expeditor, void *payload_data) {
2617 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2618 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2620 * New! function: gras_msg_handleall to deal with all messages arriving
2621 within a given period.
2622 * New! function: gras_socket_server_range to get a server socket in a
2623 range of port numbers (ease to avoid port number conflicts) [MQ]
2624 * New! gras processes display their backtrace when they get a SIGUSR1
2625 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2626 Sweet to debug RL processes [MQ]
2630 - Do not force experiment sizes to be expressed in kb, or it becomes
2631 impossible to measure the latency this way (needs one byte-long tests)
2632 WARNING: this changes the amok_bw_* function semantic. [MQ]
2633 - Implements the link saturation stuff. [MQ]
2634 * Peer management module:
2635 New! module factorizing code that we wrote over and over [MQ].
2638 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2639 - Test units are placed directly into the library code, they get extracted
2640 automatically and placed into the src/testall binary.
2641 - Convert most of the XBT tests to this system.
2642 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2643 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2644 live in the same C code. This required to change a little bit the API
2645 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2646 * New module: sparse graph structure with basic algorithms (this is work
2647 in progress and the API is not considered to be frozen yet). [AL]
2648 * Display more information on backtraces: source line & function names are
2649 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2650 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2651 * Reworked a little bit some #include statements to load only required
2652 headers. Some user code that relied on SimGrid to include stdlib or
2653 stdio may need to include it by themselves. [AL]
2654 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2656 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2657 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2658 but I guess that almost nobody used those parts. [MQ]
2660 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2662 SimGrid (3.0.1) stable; urgency=low
2665 * Unfortunately, I had missed 5 misnamed functions:
2666 xbt_fifo_item_t xbt_fifo_newitem(void);
2667 void xbt_fifo_freeitem(xbt_fifo_item_t);
2668 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2669 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2670 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2671 They're now deprecated. Please use their new versions:
2672 xbt_fifo_item_t xbt_fifo_new_item(void);
2673 void xbt_fifo_free_item(xbt_fifo_item_t);
2674 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2675 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2676 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2678 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2679 * Documentation: xbt_log module unmercifully reworked [MQ]
2680 * Bugfix: there was a problem with the ending of contexts with
2681 the pthread backend. It caused some weird deadlock or behavior
2682 depending on the pthread implementation. [AL]
2683 * Bugfix: get the exceptions raised in the simulator repport where
2684 and why they come from when they are not catched in time [AL, MQ]
2687 * Bugfix: Do repport the error when two non-connected hosts try to
2688 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2691 * Add additionnal checkings on communications. Assert that two
2692 communicating hosts are connected by a set of links... [AL]
2695 * Add additionnal checkings on channel values in communication [AL]
2696 * New: MSG_task_get_source to see on which host a task was generated [HC]
2697 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2698 the number of tasks waiting to be received on channel and sent
2700 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2701 waits for the first task coming from a given host.. [AL]
2703 GRAS new functionnalities: [MQ]
2704 * Enhance the parsing macro to allow the size of multidimentional objects
2705 to be given thru annotations.
2706 * New example (and documentation): Matrix Multiplication a la RPC
2707 (as when I was young!) and fix a bunch of bugs found on the way.
2709 GRAS performance improvements: [MQ]
2711 * Reduce the amount of cbps creation/destruction by making it static to
2712 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2715 * Change libdata to a set so that we can search for stuff by ID (and thus
2716 reduce the insane amount of dict lookups)
2719 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2720 RL communication on top of SG since it's so uneffective.
2721 It may also be used for inter-thread communication in RL, one day.
2722 * Use gras_datadesc_copy() to exchange messages on top of SG
2724 - improve message exchange performance on top of SG
2725 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2726 * Don't exchange on the network the size of the used part of buffer,
2727 instead, specify the possible buffer size to read().
2729 - reduces the amount of read/write calls (one pair per exchange)
2730 - reduces the amount of exchanged data (the size)
2731 - allows to retrieve all arrived data on receiver side, if we don't need
2732 it right now (subsequent read will peek the buffer)
2733 - allows the receiver to proceed with the begining of the stream before
2734 everything is arrived
2735 - make it possible to build an iov transport (using readv/writev)
2737 - take care of the data with non-stable storage (like stacked data),
2739 * If possible, TCP send uses vector I/O (when writev() is here)
2740 - Don't use it for receive since we send data sizes and data on the
2741 same stream, so we wouldn't be able to chain large amount of chunks
2742 before having to flush the stuff to read the size.
2743 * Rework the transport plugin mechanism to simplify it and reduce the
2744 amount of pointer dereferencement when searching for the right function
2747 * I guess that now, we do almost as few system calls as possible while
2748 doing as few data copy as possible.
2750 To improve it further, we could try to send all the sizes first and then
2751 all the data (to use iov on receiving size), but it's only a partial
2752 solution: when you have 2 dimensional data, the sizes of the second
2753 dimension is data of the first dimension, so you need 3 streams.
2755 I'm not sure the potential performance gains justify the coding burden.
2757 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2759 SimGrid (3.00) stable; urgency=high
2762 * New! Give the possibility to hijack the surf parser and thus bypass
2763 MSG_create_environment and MSG_launch_application. Have a look at
2764 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2766 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2768 SimGrid (2.96) unstable; urgency=low
2773 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2774 This deprecates the xbt_error_t mechanisms.
2775 It modifies (simplifies) all XBT and GRAS API.
2776 MSG API keeps unchanged (exceptions raised by XBT are catched from
2777 within MSG and masked with existing error handling facilities)
2780 * New! Add a FATPIPE model. [AL]
2781 * New! Add a parallel task model. [AL]
2782 * New! Add automatically a loopback interface (in the default
2783 network model) if none was precised.
2786 * Bugfix: MSG_process_resume now works with the current running process.
2788 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2789 * Modification of MSG_task_get_compute_duration. Once a task has been
2790 processed, the value returned by this function is now equal to 0. [AL]
2791 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2792 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2793 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2794 of a m_task. For now, MSG_task_cancel only works with computation
2796 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2797 of the processor (in Mflop/s) regardless of the current load on the
2799 * API Change: use proper naming convention for MSG_getClock and
2800 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2802 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2803 This function changes the priority of a computation task. This priority
2804 doesn't affect the transfer rate. A priority of 2 will make a task
2805 receive two times more cpu power than the other ones. This function
2806 has been added to suit the needs of Nguyen The Loc and hasn't been that
2807 much tested yet. So if it fails, please report it and send me your code.
2809 * API Change: removed all functions and types that were marked "deprecated"
2810 since many months. Renamed MSG_global_init_args to MSG_global_init.
2812 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2814 SimGrid (2.95) unstable; urgency=low
2817 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2818 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2819 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2822 * Bugfix: Allow absolute paths for platform description files [MQ]
2823 * Bugfix: do free the variables after use. Leads to drastic performance
2825 * Implement max_duration (ie, timeouts) on resources [AL]
2828 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2830 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2831 the needed support (timeouts) [AL]
2834 * Implement measurement sockets. You can now get the bandwidth between two
2835 hosts thanks to AMOK (see below). [MQ]
2836 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2837 dynar over the network (yeah) [MQ]
2838 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2839 * simple/static token-ring example. [Alexandre Colucci and MQ]
2840 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2841 in place before [MQ]
2844 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2845 type parsing. "short" and "long" modifiers were also ignored; other
2846 modifier (such as reference level) are still ignored. [MQ]
2847 * Embeed the buffer size within the buffer itself on SG. [MQ]
2848 That way, send() are atomic and cannot get intermixed anymore (at least
2849 the ones which are less than 100k; bigger messages still have the issue)
2850 * Array size pushed by the field, not by the field type (or each
2851 and every long int will push stuff to the cbps) [MQ]
2852 * use select() to sleep since it allows to portably sleep less than one
2855 GRAS (minor cleanups)
2856 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2859 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2860 * type category 'ignored' killed as it was never used and were difficult
2862 * whether a type can cycle or not is now a flag, leaving room for more
2863 flags (as "ignored", if we feel the need one day ;)
2864 * Rename raw sockets to measurement sockets since "raw" has another
2865 meaning in networking community.
2868 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2869 and offers features not belonging to GRAS but that most applications
2870 need. One day, it may be a set of plugins loadable at runtime.
2871 * New module: bandwidth
2872 bandwidth measurement between arbitrary nodes running this module. [MQ]
2874 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2876 SimGrid (2.94) unstable; urgency=low
2878 The first beta release of SimGrid 3 !
2882 * Update the main page and the FAQ. Adding references to gforge.
2885 * Add a gras_os_getpid function.
2888 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2889 * Extend the logs so that they also print PID, hostname, date, ... if
2891 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2892 and kill the old version which were in testsuite/
2893 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2897 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2898 for users. Instead, each programming environment (GRAS, MSG,...) use it
2899 when needed to provide such a feature to users.
2900 Don't shortcut the mechanism or you will also shortcut the virtualization
2901 you need on the simulator.
2905 * Cleanups in configury with regard to compile optimization/warning flags.
2906 Also add -fno-loop-optimize to any powerpc since it's the optimization
2907 killing gcc (< 3.4.0).
2908 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2909 needed by MSG examples complications
2910 * Borrow configury beautifications from PHP
2913 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2914 warning (thanks loris for stumbling into this one).
2915 * Bugfix: stop loading private headers (gras_config.h) from the public
2919 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2920 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2921 RL would imply a DNS resolver, which is *hard* to do in a portable way
2922 (and therefore delayed).
2923 * Implement a real timer mechanism and use it in timing macros. This allows
2924 to avoid rounding errors and get a 0.000005 sec precision in timing
2925 macros. While I was at it, various cleanups:
2926 - allow to declare more than one timed section per file (fix a stupid bug)
2927 - move some private declaration to the right place
2928 - merge conditional execution and timing macros into emulation module
2929 - document the module
2930 - make sure the module cleanups its mess on gras_exit
2931 * Documentation improvements:
2932 - (new) how to compile applications using GRAS
2933 - (new) emulation support (timing macros)
2935 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2937 SimGrid (2.93) unstable; urgency=low
2939 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2942 - Use Paje properly where used. Still to be sanitized properly.
2943 - Portability fix: Add an implementation of the contexts using pthread
2947 - Add xbt_procname(): returns the name of the current process.
2948 Use it to show the current process's name in all logging.
2950 - fix detection of older flex version and the reaction, since we do
2951 depend on modern ones (we use lex_destroy)
2952 - Better separation of SG and RL in the libs: remove all simulation code
2953 from libgras. As a result, this lib is now only 200k when stripped.
2954 Some of the xbt modules may also be duplicated (two sets and such) and
2955 should be cleaned/killed before SG3.
2956 - Insist on using xlC on AIX because of weird problems involving gcc there.
2957 - Cleanup the make remote stuff. This is now done by scripts
2958 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2959 mainly for our private use, but we're working on changing them to user
2962 - Bugfix: flush the socket on close only if there is some *output*.
2963 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2965 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2968 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2970 SimGrid (2.92) unstable; urgency=low
2972 Alpha 3 on the path to SimGrid 3
2976 - New! First try of benchmarking macros.
2977 - New! First try so that gras_stub_generator generate deployment and
2978 remote compilation helpers.
2980 - Bugfix: Initialization fix in msg_test.
2984 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2986 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2988 - API Change: gras_os_sleep to take the amount of seconds as a double.
2989 Accepting an int was error prone since it was the only location where
2990 seconds were coded as such. It leaded to damn rounding errors.
2991 - Bugfix: Hard to belive that timers ever worked before this.
2993 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2995 SimGrid (2.91) unstable; urgency=low
2997 Alpha 2 on the path to SimGrid 3
3001 - Bug fix in the lmm_solver.
3003 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3004 through the function MSG_paje_output.
3005 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3006 - It is possible to bound the rate of a communication in MSG with
3007 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3008 to write it in the changelog).
3009 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3014 - Various cleanups to the autotools stuff
3015 - Begin to move Gras examples to examples/gras/
3016 - Let make distcheck work again (yeah!)
3018 - documentation overhauled using doxygen.
3019 gtk-doc-tools is dead in SimGrid now.
3020 - Automatically extract all existing logging categories, and add the list
3021 to the documentation (long standing one, to say the less)
3023 - Cleanup the known architecture table. Reorder the entries to group what
3024 should be, and use a more consistent naming scheme.
3025 (some of the test dataset are still to be regenerated)
3026 - New! Allow library to register globals on each process just as userdata
3028 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3029 lookup time (for now).
3030 Use it in msg and trp.
3031 This cleans a lot the internals and helps enforcing privacy of the
3032 headers between the gras components.
3033 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3034 - Bugfix: gras_os_time was delirious in RL.
3035 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3037 - Reenable GRAS now that it works.
3039 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3041 SimGrid (2.90) unstable; urgency=low
3043 Alpha 1 on the path to SimGrid 3
3045 * It is a long time since the last release of SimGrid. I'm sorry about
3046 that but as I had told you, I was rewriting a lot of things. I apologize
3047 to those who had been reporting bugs to me and that I had not answered.
3048 If your bug is still in the new version, please tell me. Here is a
3049 summary of the main changes.
3051 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3052 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3053 portability, speed, and a lot more but you'll figure it out later.
3054 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
3055 I wanted to release the new MSG as soon as possible and I have
3056 broken GRAS, which is the reason why, for now, only MSG is fully
3057 functional. A laconic description of these projects is available
3058 in the documentation.
3060 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3061 kernel optimized for our needs (called SURF but only the developers
3062 should use it). Hence, MSG is now roughly 30 times faster and I think
3063 that by rewriting a little bit MSG, I could event speed it up a little
3064 bit more. Beside the gain in speed, it is also much easier to encode a
3065 new platform model with SURF than it was with SG. More to come...
3067 * REVOLUTION 3: I have tried to change a little as possible the API of
3068 MSG but a few things really had to disappear. The main differences
3069 with the previous version are :
3070 1) no more m_links_t and the corresponding functions. Platforms are
3071 directly read from a XML description and cannot be hard-coded
3072 anymore. The same format is used for application deployment
3073 description. The new format is described in the documentation.
3074 Have a look in tools/platform_generation. There is a tiny script
3075 that converts from the old platform format to the new one. Concerning
3076 the application deployment format, parsing the old one is tricky.
3077 I think most of you should however be able to convert your files. If
3078 it is really an issue, I can write a C code that does the conversion.
3080 2) the toolbox tbx does not exist anymore. We now have a library
3081 with much more data-structures but without the hash-tables (we have
3082 dictionaries that are much faster).
3084 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3086 *****************************************************************************
3087 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3088 * don't feel like converting the oldies. (Mt) *
3089 *****************************************************************************
3092 Version 2.90: "the long awaited one"
3093 - Finished rewriting and debugging MSG. Rewrote the documentation.
3094 - disable GRAS for now since it needs to be ported to the newest SG
3097 - Finish the port to windows (using mingw32 for cross-compile)
3100 - Main loop and datastructures of SURF. A cpu resource object is
3101 functional. Surf can thus be used to create cpu's with variable
3102 performance on which you can execute some actions.
3104 2004-11-15 Martin Quinson
3105 - Port to ARM. Simply added the alignment and size descriptions. Should
3106 work, but the ARM machines are so slow that I didn't had the opportunity
3107 to 'make check' over there yet.
3109 2004-11-15 Arnaud Legrand
3110 - Trace manager now written. It uses a heap structure and is therefore
3111 expected to be efficient. It may however be speeded up (particularly
3112 when many events occur at the same date) by using red and black
3113 trees. One day maybe...
3114 - Max-min linear system solver written. It uses a sparse matrix
3115 structure taking advantage of its expected use. Most operations are
3116 O(1) and free/calloc are called as few as possible. The computation of
3117 the minimum could however be improved by using a red and black tree
3120 2004-11-03 Arnaud Legrand
3121 - Rename every gras_* function that was in xbt/ to its xbt_
3123 - Add a heap and a doubly-linked list to xbt
3124 - Added a dichotomy to the dictionaries. make check works as well before
3125 so I assume that the patch is correct. I do not know however if things
3126 run effectively faster than before now. :)
3128 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3129 SimGrid, and the version number is bumped to 2.x
3131 2004-10-29 Martin Quinson
3132 - Introduction of the remote errors.
3133 They are the result of a RMI/RPC on the remote machine.
3134 ErrCodes being scalar values, you can't get the host on which those
3135 errors did happen. Extending the error mechanism as in Gnome is possible.
3136 No idea yet whether it is a good idea.
3138 2004-10-28 Martin Quinson
3139 - Interface revolution: the Starred Structure Eradication.
3140 I used to do typedef struct {} toto_t; and then handle *toto_t.
3141 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3142 - ???_t is a valid type (builded with typedef)
3143 - s_toto_t is a structure (access to fields with .)
3144 - s_toto is a structure needing 'struct' keyword to be used
3145 - e_toto_t is an enum
3146 - toto_t is an 'object' (struct*)
3148 typedef struct s_toto {} s_toto_t, *toto_t;
3149 typedef enum {} e_toto_t;
3150 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3151 s_toto_t) is private.
3153 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3154 so that it can set the variable to NULL. It was so for dicts and sets,
3155 it changed for dynars.
3157 - Fix a bunch of memleaks in dict_remove
3158 - Fix a bug in sg/server_socket opening: it failed all the time.
3160 2004-10-07 Martin Quinson
3161 - Speed up dynar lookup operation a bit.
3163 gras_dynar_get is dead.
3165 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3166 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3167 (which gives you the address of the stored data).
3169 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3170 copy of the data using an affectation to do the job and not a memcpy.
3172 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3173 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3175 It does not really speedup the dynar test because they are
3176 setting elements all the time (and look them seldom). But the dict does
3177 far more lookup than setting.
3179 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3181 2004-10-05 Martin Quinson
3182 - Allow to (en/dis)able the cycle detection at run time.
3184 Whether we should check for cycle or not is now a property of each
3185 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3186 datadesc_cycle_unset allow to remove this property when previously set.
3188 Note that the cycle detection is off by default since it impacts the
3189 performance. Watch the data you feed GRAS with ;)
3191 This property is hereditary. Any element embedded in a structure having it
3192 set have it set for the time of this data exchange.
3194 You should set it both on sender and receiver side. If you don't set it on
3195 sender side, it will enter an endless loop. If you forget on receiver
3196 side, the cycles won't be recreated after communication.
3198 - Header reorganization.
3199 Kill gras_private.h, each submodule must load the headers it needs.
3201 2004-10-04 Martin Quinson
3202 - Interface revolution: do not try to survive to malloc failure.
3204 Now, gras_malloc and friends call gras_abort() on failure.
3205 As a conclusion, malloc_error is not a valid error anymore, and all
3206 functions for which it was the only gras_error_t return value are
3207 changed. They now return void, or there result directly.
3208 This simplify the API a lot.
3210 2004-09-29 Martin Quinson
3211 - Re-enable raw sockets.
3212 Created by gras_socket_{client,server}_ext;
3213 Used with gras_raw_{send,recv}
3216 It should allow to kill the last bits of gras first version soon.
3218 This is not completely satisfactory yet (duplicate code with
3219 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3222 - Simplify transport plugin (internal) interface by not passing any
3223 argument to _server and _client, but embedding them in the socket
3226 2004-09-28 Martin Quinson
3227 - Finish the port to AIX.
3228 autoconf was my problem (segfault within the malloc replacement
3229 function. No idea why)
3231 2004-09-16 Martin Quinson
3232 - Fix some size_t madness on 64bit architectures.
3234 2004-09-08 Martin Quinson
3235 - Reduce the number of system headers loaded, overload some more system
3236 calls (such as malloc to cast the result of the system one, and work
3238 - Fix and reintroduce the config support
3240 2004-09-07 Martin Quinson
3241 - Source code reorganization to allow Arnaud to surf all over there.
3242 - Allow to document the logging categories.
3243 - Remove all uppercase from logging categories and useless cleanup in names.
3245 2004-08-18 Martin Quinson
3246 Version 0.6.2 (protocol not changed; API changed)
3247 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3248 gras_error_t), and NULL when not found. Functions expecting a msgtype
3249 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3250 hopefully usefull message.
3251 - Portability to prehistoric sparcs again
3253 2004-08-17 Martin Quinson
3254 Version 0.6.1 (protocol not changed; ABI not changed)
3255 - prealloc some buffers to speed things up
3257 2004-08-11 Martin Quinson
3258 Version 0.6 (protocol not changed; ABI expended)
3259 - The parsing macro can deal with the references, provided that you add
3260 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3262 2004-08-09 Martin Quinson
3263 Version 0.5 (protocol not changed; ABI changed)
3264 - Allow to off turn the cycle detection code in data exchange at
3265 compilation time. It should be at run time, but I'm short of time (and
3266 the config stuff is still broken). That way, we keep dict out of the
3267 critical path, which is good because the performance is poor:
3268 - search not dichotomial yet
3269 - dynar give no way to access their content and memcpy everytime
3270 - In composed data description (struct, ref and so on), stop foolness of
3271 keeping the subtype's ID, but store the type itself. This keeps sets out
3272 of the critical path, which is good since they rely on dynar and
3273 dictionnaries. The only loose of that is that we cannot detect the
3274 redeclaration of a structure/union with another content (but I'm not sure
3275 the code detected well this error before anyway). We still can detect
3276 the redefinition discrepancy for the other types.
3277 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3278 it breaks the code because of type-punning used all over the place).
3279 This breaks on all non-gcc architectures (for now).
3281 All those changes (plus the buffer of last time) allow me to gain 2 order
3282 of magnitude on cruel tests consisting of 800000 array of integers on two
3283 level of a hierarchical structure (200 secondes -> 4 secondes)
3286 - the selector of reference must now return the type it points to, not
3287 the ID of this type.
3289 2004-08-06 Martin Quinson
3290 Version 0.4 (protocol changed; ABI not changed)
3291 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3292 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3293 - Implement a buffer transport, and use it by default (it relies on tcp in
3294 real life and on sg in simulation).
3295 That's a bit hackish since I had a new field to the structure to store
3296 its data without interfering with the subtype ones. Inheritance
3297 is tricky in C. And that's a kind of reverse inheritance with one class
3298 derivating two classes. Or maybe a game with java interfaces. Anyway,
3299 that's damn hard in C (at least).
3300 Moreover, I got tired while trying to ensure plugin separation and
3301 genericity in SG mode. MSG wants me to do weird things, so let's go for
3302 cruel hacks (temporarily of course ;).
3303 See comment in transport_private.h:71
3304 - do not include all the _interface headers in private but in the files
3305 which really need them (to cut the compilation time when they are
3308 2004-07-26 Martin Quinson
3309 Version 0.3 (protocol not changed; ABI changed)
3310 - Major overhault of the datadesc interface to simplify it:
3311 - shorted the function names:
3312 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3313 - add a trivial way to push/pop integers into the cbps without malloc.
3314 This allows to make really generic sub_type description, which simply
3315 pop their size of the stack.
3316 - add a function gras_datadesc_ref_pop_arr() which does what users want
3317 most of the time: Declare a dynamic array (which pops its size of the
3318 stack) and declare a reference to it. Poor name, but anyway.
3319 - kill the post-send callback, add a post-receive one
3321 2004-07-23 Martin Quinson
3322 Version 0.2 (protocol changed; ABI changed)
3323 - add some testing for cpbs in the test cases, and fix some more bugs.
3324 This invalidate again the little64 data file, since I cannot regenerate
3326 - remove an awfull optimization in the logging stuff, allowing me to:
3327 - understand it again
3328 - learn gcc how to check that the argument match the provided format
3329 - fix all errors revealed by gcc after that
3330 - internal keys of dict are not \0 terminated. Deal with it properly in
3331 loggings instead of segfaulting when the user want to see the logs :-/
3333 2004-07-22 Martin Quinson
3334 - Fix some stupid bug preventing cbps (callback postit) from working
3336 2004-07-21 Martin Quinson
3337 - Some documentation cleanups
3338 - remove the useless last argument of msgtype_declare
3339 - rename the Virtu functions to fit into the 'os' namespace
3340 - move headers src/include -> src/include/gras/ and stop fooling with
3341 gras -> . symbolic link
3342 - make distcheck is now successful
3344 2004-07-19 Martin Quinson
3346 - Build shared library also
3347 - Install html doc to the right location
3348 - stop removing maintainer files in make clean
3349 - build tests only on make check
3351 2004-07-13 Martin Quinson
3353 - No major issue in previous version => change versioning schema
3354 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3355 corresponding dataset.
3357 2004-07-11 Martin Quinson
3359 - Get it working with any kind of structure (we can compute the padding
3360 bytes remotely for all the architectures I have access to)
3361 - Implement the structure parsing macro (still not quite robust/complete)
3362 - Improvement to the remote testing toysuite
3364 2004-07-10 Martin Quinson
3365 [autoconf mechanism]
3366 - get ride of a bunch of deprecated macros
3367 - actually run the test for two-compliment, not only compile it :-/
3368 - test whether the structures get packed (and bail out if yes. Damn.
3369 Alignment is a serious matter)
3370 - test whether the structures get compacted (but respecting the alignment
3371 constraints of each types)
3372 - test whether the array fields of structures can straddle alignment boundaries
3374 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3375 (took me 2 hours to find that bug, looking at the wrong place)
3376 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3377 before sending/receiving and are used to compute the offsets of fields
3378 - Given that padding size depend even on compiler options, keep track of
3379 alignment and aligned_size only for the current architecture. Not a big
3380 deal since we send structure fields one after the other (seems
3382 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3383 program, let it work on linux/gcc/little32. portability todo.
3385 2004-07-08 Martin Quinson
3386 - import and improve remote compilation support from FAST
3387 - make sure make check works on half a dozen of machines out there
3389 2004-07-07 Martin Quinson
3390 Let's say it's version 0.0.3 ;)
3391 - Implement conversions (yuhu!)
3392 - Let it work on solaris (beside conversion, of course)
3393 - Stupid me, using rand() to generate the conversion datatests in not wise.
3395 2004-07-06 Martin Quinson
3396 - Let make dist work, since I'm gonna need it to compile on remote hosts
3397 - Let Tests/datadesc_usage write the architecture on which the file was
3398 generated as first byte.
3399 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3400 architecture descriptions.
3401 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3402 on those architectures.
3403 - Optimization: send/recv array of scalar in one shoot
3405 2004-07-05 Martin Quinson
3406 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3408 - Plug a whole bunch of memleaks
3409 - each process now have to call gras_{init,exit}. One day, their log
3410 settings will be separated
3411 - Continue the code factorisation between SG, RL and common in Transport.
3413 2004-07-04 Martin Quinson
3415 - Redistribution between SG and RL.
3416 We wanna have to accept in SG, so move accepted related parts of RL in
3417 the common part. (more precisely, the dynar of all known sockets is no
3418 more a static in transport.c, but part of the process_data)
3420 [gras_stub_generator]
3421 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3423 2004-07-03 Martin Quinson
3424 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3425 them all up in one shot)
3427 - Fix the ugly case of reference to dynamic array.
3428 - New (semi-public) function gras_datadesc_size to allow the messaging
3429 layer to malloc the needed space for the buffer.
3431 - gras_socket_close now expect the socket to close (and not its address to
3432 put NULL in it after it). This is because the socket passed to handlers
3433 is one of their argument (=> not writable).
3435 - propagate the interface cleanup from last week in datadesc, ie remove a
3436 superfluous level of indirection. User pass adress of variable
3437 containing data (both when sending and receiving), and not of a variable
3438 being a pointer to the data. Let's say that I like it better ;)
3439 The price for that is constructs like "int msg=*(int*)payload" in
3440 handlers, but it's a fine price, IMHO.
3442 - Let it work in RL (yuhu)
3444 2004-06-21 Martin Quinson
3446 - porting SG plugin and SG select to new standards (works almost).
3447 - plug memleaks and fix bugs around.
3450 - cleanup the prototype of data recv and force users to specify when they
3451 want to handle references to objects. Test case working even for cycles.
3452 - plug memleaks. Valgrind is perfectly ok with this.
3454 2004-06-12 Martin Quinson
3456 - cleanup the separation between plugin and main code in plugin creation
3458 2004-06-11 Martin Quinson
3460 - Reput hook for raw sockets, needed for BW experiments
3461 - kill a few lines of dead code
3462 [Data description] Interface cleanup
3463 - gras_datadesc_by_name returns the searched type or NULL.
3464 That way, no variable is needed to use a type desc once, which makes
3466 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3467 two parameters were strings (field name, type name), leading to
3469 [Dicos] Interface cleanup
3470 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3471 This is consistant with the dynar API.
3473 2004-04-21 Martin Quinson
3475 - Porting to new standards.
3477 - interface cleanup.
3478 There is no bag anymore, no need to take extra provision to mask the
3479 pointers behind "ID".
3480 Better splitup of functions between files create/exchange/convert.
3481 This is still a bit artificial since convert and receive are so
3482 interleaved, but anyway.
3484 - add a queued message list to procdata (the ones not matching criteria
3486 - factorize some more code between SG and RL wrt procdata
3488 - use gras_exit in example to track memleaks
3489 - get rid of gs_example now that GS is properly integrated into gras
3490 - update run_test to integrate the latest tests (datadesc)
3492 - rename WARNINGn macros to WARNn since it prooved error-prone
3494 2004-04-19 Martin Quinson
3496 - register init/exit functions within gras module mechanism
3497 - send/receive function.
3498 Convertion is not implemented, but short-cutted if not needed.
3499 struct/array elements are sent one by one (instead of block-wise), but
3500 nobody really cares (yet). Get a prototype before optimizing.
3501 - tests (using a file socket) for DD send/receive on:
3502 - base types: int, float
3503 - array: fixed size, string (ie ref to dynamic string)
3504 - structure: homogeneous, heterogeneous
3505 - chained list, graph with cycle
3506 Believe it or not, valgrind is not too unhappy with the results. The
3507 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3511 - Bugfix when using a filename explicitely (instead of '-')
3513 2004-04-09 Martin Quinson
3515 - factorize more code between RL and SG in socket creation
3516 - Complete the implementation and tests of:
3518 o file (only in RL, and mainly for debugging)
3520 I lost 3 days to design a portable address resolver, and then decided
3521 that the prototype mainly have to run on my box.
3522 Addressing portability too early may be like optimizing too early :-/
3524 - use gras_init in the Tests instead of the crappy parse_log_opt
3525 (the latter function is removed)
3526 [Conditional execution]
3527 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3528 [Code reorganisation]
3529 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3530 Build examples against the RL library, since there is no way to disable
3531 its creation for now.
3533 For information, the beginning of coding on GRAS was back in june
3534 2003. I guess that every line has been rewritten at least twice since