1 SimGrid (3.22) NOT RELEASED (Release Target: December 21. 2018, 22:23 UTC)
6 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
7 - Not complete yet: asynchronous activities (amongst others) are still missing
8 - Still ongoing: the interface may change in the future. We need more
9 testers! Please report any glitches.
10 - No new project using Java should start now. Please switch to Python.
13 - Some of the internal cleanups may lead to speed improvements:
14 - The hard limitation on the amount of simulated actors with
15 Java+Mac was removed. Now, the available memory is the only limit.
16 - Our refcounting was tidyied, leading to 10% speedups in some cases.
17 - We are still working on making our code robust to the actor kills
18 and hosts' churn. Things are improving, but it's not perfect yet.
19 - Replaced our own code to display a backtrace (that was forking addr2line)
20 with the Boost.Stacktrace library.
21 You won't see your backtraces without this optional dependency.
22 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
23 - Stop setting random seed with srand() at initialization.
26 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
27 XML file version remains 4.2 since old files are still compatible.
30 - Process termination which was broken at version 3.21 has been repaired.
31 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
32 - Hide the examples into examples/deprecated. New users should use Python.
35 - Drop MSG_process_create_from_stdfunc() from the API.
36 This C++ function was a pimple in the C API, made necessary at some
37 point by the Java bindings. This is fixed now.
38 - Hide the examples into examples/deprecated. New users should use S4U.
39 - MSG_process_create and MSG_process_attach now crash if the host used
40 to run this process is off.
41 - Fix the protype of MSG_process_on_exit()
42 Now use (int,void*) callbacks instead of (void*,void*) ones.
43 The implementation was ways too messy to actually work, I guess.
46 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
47 too often and provide more accurate timings.
50 - Drop sg_cmdline. Please use xbt_cmdline instead.
51 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
52 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
53 OsSemaphore is implemented in a portable way with C++11 threads.
54 This should allow much more threads to be created at the same time,
55 allowing Mac and Java users to simulate many more actors.
56 - Implement the 'thread' factory with std::thread instead of xbt ones.
57 It is not possible to set the stack size with threads anymore, but
58 -fsplit-stack is the way to go nowadays when using threads.
59 - Drop the xbt_os_thread_t module (now unused)
60 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
61 - Drop cunit, use Catch2 instead.
64 - Many cleanups in the kernel::activity namespace. This was long
65 overdue, and shall open the path to many future endeavors.
68 - #132: Java: a process can not shut down its own host
69 - #220: S4U: detached send causes exception if sender terminates soon after sending
70 - #261: Document the parameters of parallel execution's constructor
71 - #300: [s4u] BarrierPtr is missing
72 - #314: SMPI args internal cleanup
73 - #316: Fix a bug related to the CPU utilization of multi-core VM
74 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
75 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
76 - #325: Turning off a host has different behavior on sleeping actors and computing actors
78 ----------------------------------------------------------------------------
80 SimGrid (3.21) October 3. 2018
82 The Restarting Documentation (TRD) Release.
85 - Convert the user manual to Sphinx for improved usability.
86 Unstable version now hosted on https://simgrid.frama.io/simgrid
87 Many glitches remain; stay tuned.
88 - Restrict the Doxygen reference API to the installed header files.
89 (documenting all internals this way is overkill + doxygen scales badly)
90 - New tutorial on S4U (highly inspired from the old MSG one, but with
91 a git repository to fork as a starting point, and a docker image)
92 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
93 - The Developper manual is still to be converted (not compiled until then)
94 - Some parts are still missing in this conversion (such as the
95 platform and deployment sections) while others were blindly converted
96 and would need more love (such as the configuration flags).
97 Things will certainly further improve in the future releases.
100 - Fully reimplement auto_restart mechanism that was utterly broken
101 (fix #22, #131 and #281 that were all very related)
102 - Implement semaphores (fix #303 and #291)
103 - When creating an actor from a function and its parameters,
104 move-only parameters are not allowed anymore, as it would prevent
105 the actor to be restartable if its parameters are consumed on run.
106 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
107 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
108 IO operations that can be started, waited for, or canceled as a regular
110 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
114 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
115 - New host variable: 'core_count'
118 - Remove xbt_os_thread_specific features
119 - Remove portability wrapper to condition variables
120 - Remove xbt_os_thread_yield()
123 - MPICH collective selector now mimics MPICH 3.3b
124 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
125 - MPI_Init(NULL,NULL) is now allowed
126 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
127 - Command-line arguments should now be handled properly in Fortran simulations
128 - MPI Attributes and related callbacks should work in Fortran
129 - Apps using C + Fortran should now work
130 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
131 - Support MPI_Error_String
134 - Due to an internal bug, Msg.run() must now be your last line.
135 We hope to fix it in a future release, and we are sorry for the inconvenience.
138 - #22: Process autorestart seem to only work with CAS01 cpus
139 - #56: Feature request: dynamic SMPI replay
140 - #93: simgrid should not eat --help
141 - #111: How to change the loopback link characteristics?
142 - #116: Platform cluster radical
143 - #117: Platform cluster/host order
144 - #129: Replay_multiple on steroid
145 - #131: Java : autostart does not work
146 - #143: Setting a breakpoint at a given time
147 - #235: Network energy models should be integrated
148 - #264: Add ptask L07 resource tracing
149 - #271: Dynamic and manual replay of SMPI traces
150 - #279: Breakpoints option uses time deltas instead of absolute time
151 - #281: Daemonized actor no longer daemonized after an auto-restart
152 - #282: TIT convertor erase traces when using absolute path in the trace list file
153 - #285: segfault when a s4u actor kills itself with msg process tracing activated
154 - #286: Pajé traces are not exposing the number of cores
155 - #287: Command-line parsing should stop when encountering '--'
156 - #288: MPI_Init(NULL, NULL)
157 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
158 - #290: Method to check if a mailbox has a message ready to be consumed
159 - #291: [s4u] semaphores do not seem available yet
160 - #294: Alltoall collective implementation of mpich
161 - #295: Starting a Host as OFF is not supported
162 - #296: DTD too permissive
163 - #299: Add a s4u barrier example
165 ----------------------------------------------------------------------------
167 SimGrid (3.20) Released June 24. 2018
169 The proxy snake_case() release.
172 - The whole API is now using snake_case() instead of mixing it with camelCase.
173 Compatibility wrappers in place for one year.
174 - We took the opportunity to increase the consistency of the API.
175 Please report any remaining glitches.
176 - New function simgrid::s4u::Host::get_actor_count:
177 Returns the number of actors running on a specific host.
180 - SMPI is now tested with ~45 proxy apps from various sources, with none or
181 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
182 - Replay: The replay file has been re-written in C++.
183 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
184 supported. They are stored in the trace and used when replayed.
185 - Basic support of MPI_Cancel. Robustness not guaranteed.
186 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
187 (only for MPI_COMM_TYPE_SHARED).
188 - New option: smpi/privatize-libs, to add external shared libs to be privatized
189 by SMPI. They will be copied locally and loaded separately by each process.
190 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
191 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
192 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
193 with internal index optind with multiple processes. Only works if MPI_Init has
195 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
196 syntax without preprocessing tricks.
199 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
200 behavior tracing is no longer limited to MSG
201 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
202 this means that traces now need to be updated or re-obtained; both Irecv
203 and Isend lines in your traces have now in total 5 mandatory fields:
204 <rankid> <command> <to/from rankid> <tag> <size>
205 To update your traces, it suffices to add a 0 for the tag here.
206 - TIT now also supports waiting for a distinct request via MPI_Wait.
207 Wait/Test now wait for a specific request, not just the last one that was
208 issued. This unfortunately means another update, because we need to
209 identify which request you want to wait for. We do this via the
210 triplet (sender, receiver, tag), which needs to be added:
211 <rankid> <command> <sender> <receiver> <tag>
212 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
213 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
216 - The deprecation of MSG is ongoing (but this should not impact you).
217 Many MSG functions are now simple wrappers on the C API of S4U. If
218 you wish to convert your code to S4U, find the S4U counterparts of
219 your MSG calls in src/msg/msg_legacy.cpp.
220 - MSG can still be used, but won't evolve anymore.
223 - Allow to run the Link energy plugin from the command line with
224 --cfg=plugin:link_energy
225 - Rename Energy plugin into host_energy
226 - Rename Load plugin into host_load
229 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
230 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
231 - Improve performance when handling timeouts of simix synchros.
234 - Config: the C API is now deprecated (will be removed in 3.23), and
235 the C++ API has been slightly improved.
238 - Fix several build issues on OSX.
239 - Move simgrid_config.h to simgrid/config.h (old header still working)
242 - #143: Setting a breakpoint at a given time
243 - #150: Inconsistent event names in SMPI replay
244 - #258: daemonized actors hang after all non-daemonized actors have completed
245 - #267: Linker error on unit_tmgr
246 - #269: SMPI: tracing of MPI_Wait/all/any broken
247 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
249 ----------------------------------------------------------------------------
251 SimGrid (3.19.1) Released March 22 2018
253 The Fixed ABI Release.
255 As you may know, we are currently refactoring SimGrid in deep.
256 Upcoming SimGrid4 will be really different from SimGrid3: modular,
257 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
259 Our goal is to smooth this transition, with backward compatibility
260 and automatic update paths, while still progressing toward SimGrid4.
262 SimGrid remains open during works: The last pure SimGrid3 release was
263 v3.12 while all subsequent versions are usable alpha versions of
264 SimGrid4: Existing interfaces remain unchanged, but the new S4U
265 interface is budding and the internals are deeply reorganized.
267 Since 2015, we work hard to reduce the changes to public APIs. When
268 we need to rename a public library symbol in S4U, we let your
269 compiler issue an explicative warning when you use the deprecated
270 function. These messages remain for four releases, i.e. for one full
271 year, before turning into an error. Starting with v3.15, your can
272 also adapt to API changes with the SIMGRID_VERSION macro, that is
273 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
275 Starting with this v3.19.1, our commitment to reduce the changes to
276 the public interfaces is extended from the API to the ABI: a program
277 using only MSG or SimDag and compiled against a given version of
278 simgrid can probably be used with a later version of SimGrid without
279 recompilation. We will do our best... but don't expect too much of
280 it, that's a really difficult goal during such profund refactoring.
282 The difference between v3.19 and v3.19.1 is that the former was
283 accidentally breaking the ABI of MSG, while the later is restoring
286 S4U and kernel APIs will still evolve until SimGrid4, with one-year
287 deprecation warnings as currently. In fact, cleaning up these
288 interfaces and converting them to snake_case() is one release goal of
289 v3.20. But don't worry, we are working to smooth this upgrade path.
291 In summary, new projects should start with S4U to benefit of the
292 future, but old MSG projects should still be usable with no change.
294 ----------------------------------------------------------------------------
296 SimGrid (3.19) Released March 20 2018
298 The Moscovitly-cold Spring release.
300 SMPI user-visible improvements:
301 - Support of MPI_Type_create_subarray
302 - Unimplemented MPI calls now abort (instead of issuing a warning)
303 Some harmless calls around errhandlers still only issue a warning.
304 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
305 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
306 is now possible (mmap privatization was refactored to this end).
307 Adding new SMPI instances should also be technically doable now.
309 SMPI internal cleanups:
310 - Removed index notion from SMPI -- no more "getPid() - 1"!
311 In tracing files, actors are called rank-<PROCESS-ID> but these are
312 NOT the real ranks (tracing several communicators is not supported yet).
313 - Deprecate smpi_process_index() (will be removed in 3.22)
314 - Better testing on classical proxy apps, through an external project:
315 https://github.com/simgrid/SMPI-proxy-apps/
318 - Use a graphical TOC to make it easier to find the documentation you need
319 - Revamp the MSG tutorial
322 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
323 - Remove parameter of MSG_process_killall().
324 Resetting the PID was bogus anyway (several actors could have the same PID).
327 - Execution->setHost() can be called after start(), to migrate it.
328 - Comm::test_any() is now implemented.
329 - s4u::Actor now has onCreation() and onDestruction() signals.
330 - Install some kernel header files for the users' plugins and more.
331 Warning, their API is really not stable yet.
332 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
333 include/simgrid/kernel/routing/*Zone.hpp
336 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
338 Build System and other cleanups:
339 - Remove unused run-time parameter "tracing/onelink-only".
340 - SimGrid now works with intel compilers (MC still buggy)
341 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
342 - tesh kills subprocesses on timeout (but not on windows)
345 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
348 - #194: Feature request: simgrid::s4u::Comm::test_any()
349 - #245: migrating an actor does not migrate its execution
350 - #253: Feature Request: expose clusters as objects
351 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
352 - #255: Tesh broken on Windows
353 - #256: Modernize FindSimGrid.cmake
354 - #257: Fix (ab)use of CMake install
356 ----------------------------------------------------------------------------
358 SimGrid (3.18) Released December 24 2017
360 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
362 MSG IS NOW DEPRECATED. Please use S4U for new projects.
363 - Support for MSG should not be removed from SimGrid before 2020,
364 but future evolutions will be limited to the integration of
365 user-provided patches.
368 - s4u::onDeadlock() signal, run before stopping the simulation.
369 - s4u::Exec: asynchronous executions (abstraction of background
371 - s4u::Host->getLoad() returns the achieved speed in flops/s
374 - Link::name() is deprecated and will be removed in v3.21, use
375 Link::getCname() instead.
376 - Mailbox::getName() changed to return a std::string, use
377 Mailbox::getCname() instead to get a char*.
378 - Storage::getName() changed to return a std::string, use
379 Storage::getCname() instead to get a char*.
380 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
381 to have both allocation and deallocation of the map in user space.
384 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
385 The amount of flops remaining to do is a vector, not a scalar.
386 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
387 name implies on both sequential and parallel tasks.
388 - Both changes fix GitHub's #223 using PR #237 as a basis.
389 Thanks Michael Mercier.
390 - Most examples were converted to S4U and hidden elsewhere at tests.
391 Rationale: we still want MSG to work; we want newcomers to use S4U.
394 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
395 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
398 - Switch to the faster dlopen privatization mechanism by default
399 - Documentation improvments
402 - Live migration is getting moved to a plugin. Dirty page tracking is
403 the first part of this plugin. This imply that VM migration is now
404 only possible if one this function is called:
405 - C/MSG: MSG_vm_live_migration_plugin_init()
406 - C/C++: sg_vm_live_migration_plugin_init()
407 - Java: Msg.liveMigrationInit()
408 For C and C++, "simgrid/plugins/live_migration.h" also has to be
412 - Define class simgrid::xbt::Path to manage file names.
413 - Removed unused functions:
414 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
415 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
416 xbt_graph_export_graphviz()
417 - xbt/str.h: xbt_str_join()
418 - Remove unused datatypes:
419 - xbt/heap.h: use std::priority_queue or boost::heap instead
420 - xbt/swag.h: use boost::intrusive::list instead
423 - New link_energy plugin for the consumption of the links.
424 - All of the operations on files and storage contents have been
425 packaged into a plugin (src/plugins/file_system). The current
426 public interface can be found in
427 include/simgrid/plugins/file_system.h
428 To use these functions you now have to initialize the plugin by
429 calling MSG_storage_file_system_init() just after calling
430 MSG_init() or sg_storage_file_system_init() just after creating
434 - Remove the undocumented/untested tag <include>
437 - Remove viva specific tracing as the tool is no longer maintained
440 - #248: Kill of finished processes leads to segfault
441 - #240: xbt_cond_wait_timeout should gracefully return for C apps
442 - #239: please implement signal s4u::onDeadlock()
443 - #230: segfaults when exit() without run()
444 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
445 - #223: MSG_task_get_flops_amount() not working with parallel tasks
446 - #222: Actor::kill() doesn't really kill and segfaults
447 - #221: odd LMM warning when killing an actor
448 - #120: Memory leak when the processes are forcefully killed
450 ----------------------------------------------------------------------------
452 SimGrid (3.17) Released October 8 2017
454 The Drained Leaks release: (almost) no known leaks despite the tests.
456 Even very long simulations will preserve your memory: our testsuite
457 stresses SimGrid for over 45mn on fast machines for a coverage of
458 over 80%, and there is only one single known leak, of about 4kb.
462 - Comm.detach(): start and forget about asynchronous emission. A cleanup
463 handler may be given to free resources if the comm cannot be completed.
464 - this_actor::send(mailbox) is now mailbox->put()
465 - New: simgrid::s4u::Comm::wait_all()
466 - New: Host.execute() for remote executions.
469 - Unused option network/sender-gap is removed.
472 - Deprecate MSG_task_isend_with_matching(): This unused feature
473 really complicates our internals. Will be removed in v3.20.
476 - Improved context termination. It is now possible to free resources, even
477 when a process is forcibly killed.
478 - ContextBoost: add support for Boost versions above 1.61.
481 - Bring back run-time option --cfg=exception/cutpath to remove exception
483 - Removed unused functions:
484 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
485 xbt_str_rtrim(), xbt_str_trim().
486 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
489 - Removed header files obsolete since SimGrid 3.12:
490 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
492 - #3: SD_exit should be made optional
493 - #120: Memory leak when the processes are forcefully killed
494 - #159: Threading test regression in Actor refcounting
495 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
496 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
497 - #186: Actor::killAll() segfaults if some process is blocked on wait()
498 - #191: VM migration and pstate
499 - #192: Updating the energy consumptions of all hosts crashes with VMs
500 - #195: All actors have PID=0 in the logs
501 - #204: Sometimes segfault with thread contexts and mmap privatization
502 - #222: Actor::kill() doesn't really kill and segfaults
503 - #225: Actor::kill() doesn't really kill when victims are doing a join()
505 ----------------------------------------------------------------------------
507 SimGrid (3.16) Released June 22. 2017.
509 The Blooming Spring Release: developments are budding.
512 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
513 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
514 - Java JAR file should be finally fully working on Mac OSX too.
516 XML platforms: Switch to platform v4.1 format.
517 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
518 - <zone> can be used as a synonym for the now deprecated <as>
519 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
520 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
521 - <actor> can be used as a synonym for the now deprecated <process>
522 - state_file and avail_file periodicity is now easier to express
523 (check the documentation) the old behavior should still work.
524 * Storage: not backward compatible, but it seems that nobody ever used it
525 - Remove attribute 'content_type' of <storage_type>: was never used
526 - Make attribute 'model' of <storage_type> optional: for future usage
527 - Remove Bconnection model property: was never (in)validated. Replaced by
528 the maximum of the read and write bandwidth as a resource constraint.
531 - New and Backwards Compatibility break:
532 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
533 When one wants to get the list of tasks whose states have changed during a
534 simulation round, s/he has to allocate and free a dynar and use it as
535 argument to this function. The former SD_simulate (double how_long)
539 - Allow multicore VMs, with the correct sharing (unless you overcommit)
540 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
543 - New function to update the consumption of all hosts at once.
544 - Fix the model for multi-core hosts, linear in the amount of busy
545 cores with an abnormality for pIdle (see paper).
546 BUG: Still not satisfactory for ptask on mono-cores.
549 - The netzone are now available from the MSG API.
550 The old names still work, but are now deprecated.
553 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
554 - New option: smpi/keep-temps to not cleanup temp files
555 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
556 mallocs mode is used, allows to change the size of the fake file used
557 (default 1MB), to potentially limit the number of mappings for large runs.
558 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
559 - Fortran ifort and flang compilers support
560 - New RMA calls supported (experimental):
561 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
562 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
563 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
564 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
565 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
566 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
569 - New: this_actor::isMaestro() returns whether we are in kernel mode.
570 - Behind the scene, ActivityImpl are now automatically refcounted.
571 Removing this nasty bottleneck will greatly simplify our internals.
574 - Replay: New function xbt_replay_action_get():
575 Retrieve the function previously associated to an event type.
576 - DROPPED FUNCTION: xbt_str_varsubst()
577 - DROPPED MODULE: strbuff. We don't need it anymore.
578 - DROPPED MODULE: matrix. We don't need it anymore.
579 - DROPPED MODULE: lib. We don't need it anymore.
581 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
583 SimGrid (3.15) Released March 22. 2017
585 The Spring Release: continuous integration servers become green
586 We fixed even the transient bugs on all target architectures:
587 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
588 Windows, FreeBSD, NetBSD.
590 - Rename NetCards to NetPoints.
591 This was intended to help NS3 users, but that's not a netcard.
592 That's a point in the routing algorithm, let's avoid wrong simplifications.
595 - New: MSG_process_yield(). Stop and yield to other processes.
596 - New: MSG_process_daemon(). Daemon processes are automatically killed
597 when the last non-daemon process terminates
598 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
599 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
600 to make room for the upcoming network energy plugin.
601 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
604 - Ensure that an actor can kill itself with Process::exit()
605 - Kill the obscure NativeException. Nobody want to survive the issues
606 it denotes, so use JniException that is a RuntimeException (not to
607 be caught explicitly).
608 - Partial bug fix in initialization. SimGrid flags on command line were
609 consumed at C level but stayed in the original Java String[] args.
610 This could mess users' args[i] if SG flags were not put at the end of
612 The SimGrid flags are now removed from the Java arguments. However,
613 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
614 if args.length is greater than the number of YOUR OWN ARGUMENTS.
615 It might be if you have --log or --cfg flags in the command line.
616 - Fix numerous memleaks all around the place. In particular, around VMs.
620 - s4u::onPlatformCreated: right before the simulation starts
621 - s4u::onSimulationEnd: right after the main simulation loop
622 - s4u::onTimeAdvance: right after a clock change (time is discrete)
623 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
624 event from the availability_file changes the avail speed.
625 - Links are now usable from s4u
626 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
627 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
628 - New examples: The conversion of MSG examples to S4U has begun
629 - Actors: create, kill, migration, and suspend
630 - Applications: master-worker and token-ring
631 - Action replay: communications and storage
632 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
635 - Backwards Compatibility breaks
636 - The SD_route_get_list and SD_route_get_size functions have been removed.
637 They are replaced by the unique
638 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
639 The route from src to dst is built in the links dynar whose size can
640 be obtained with xbt_dynar_length.
641 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
642 sg_host_route_bandwidth and sg_host_route_latency.
643 Macros ensure the backwards compatibility, but you should fix your code
646 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
647 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
648 - Bump our claim of support from MPI 1.1 to MPI 2.2.
649 We don't support 100% of it, but it should be enough. Contact us if not.
650 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
651 naive and probably inaccurate)
652 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
653 mapped onto a unique small file using some system magic.
654 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
655 - Fortran: cleanups, fixes, support of user-added operations
656 - RMA: MPI_Accumulate are applied in correct order.
657 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
658 - Better support for MPI_IN_PLACE
659 - Support for MPI_Win attrs and keyvals.
660 - Support MPI_Comm_group_create, MPI_Type_size_x
661 - MPI ops should be properly applied to all allowed types
662 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
663 algorithm by a non failing one from mpich.
666 - Kill the fifo data container: we don't use it anymore.
669 - A <cluster> can now be created with different speed values to represent pstates.
670 see examples/platforms/energy_cluster.xml for an example.
672 SimGrid (3.14.159) Released December 28. 2016
674 The previous release was only a crude approximation of a Pi release;
675 this one should be a bit better.
677 - Revert a last minute change that broke on Mac OSX.
678 - Fix the changelog and NEWS files.
679 - Improve the documentation, in particular of the routing module.
680 - Integrate some patches flying around in Debian.
682 SimGrid (3.14) Released December 24. 2016
685 * The whole documentation was reworked and reorganized. There is
686 still much room for improvement here, but we're on it.
689 * We now need python3 for our testing infrastructure.
690 * Model checking needs libevent
693 * Now works on FreeBSD
696 * Privatization now works on FreeBSD
698 * Privatization is now activated by default in smpirun
699 You can turn it off with -no-privatize if something goes wrong.
701 * Call-location tracing for SMPI.
702 You can add the exact location (filename / linenumber) of an MPI call to
703 your trace files and slow down or speed up the simulation between two
704 consecutive calls by using an adjustment file (see the documentation).
706 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
707 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
708 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
709 values were ignored in some cases. The timings of these functions can now
710 be significantly different.
712 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
713 smpi/running-power is renamed to smpi/host-speed
715 * smpi/grow-injected-times option to enable or disable multiplication of the
716 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
717 make simulation less precise (but also much faster).
719 * smpirun script should be (much) faster for large deployments.
721 * SMPI tracing: fixed issue with poor matching of send/receives.
723 * Replay: Fix broken waitall
725 New functions and features
726 * MSG_parallel_task_execute_with_timeout, to timeout computations.
728 Dropped / renamed functions and features
729 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
730 - MSG_mailbox_is_empty() -> Mailbox::empty()
731 - MSG_mailbox_front() -> Mailbox::front()
732 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
733 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
734 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
735 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
736 * Don't pass the free_f parameter to property related functions:
737 - MSG_host_set_property_value()
738 - MSG_as_router_set_property_value()
739 - MSG_storage_set_property_value()
740 * VM properties. Since msg_vm_t are msg_host_t, just use
741 MSG_host_get_property_value() and friends
742 * VM I/O related things:
743 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
744 - Unimplemented save/restore methods
745 * MSG_as_router_get_property_value() was redundant with
746 MSG_environment_as_get_property_value().
747 - Removed MSG_as_router_*propert*() functions
748 - Added MSG_environment_as_set_property_value() for consistency
749 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
750 Well, they are still there for now, but deprecated with a warning.
751 Please switch to xbt_dict_new_homogeneous() before this is removed
753 * Task affinity. Its intended behavior (that was very badly tested
754 and probably not really working) was deceiving what most users
755 would have hoped here.
756 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
757 * Option network/coordinates is now useless and should be dropped.
760 * Added option storage/max_file_descriptors to allow more than 1024 files opened
763 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
764 * Added Dragonfly topology support
767 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
768 EiBps, EBps, Eibps, Ebps for bandwidth.
769 They may become useful to some lucky ones.
772 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
773 * ex/app_tokenring: new example, very similar to the MSG Token Ring
774 * ex/async_waitAll: new example, on asynchronous communications
777 * Memory usage should be decreased for simulations with a large number
778 of processes. This also helps for SMPI.
780 SimGrid (3.13) stable; urgency=low
782 The Half Release, a.k.a. the Zealous Easter Trim.
784 - We removed half of the lines, that were mostly experimental cruft.
785 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
786 (not counting blanks and comments -- according to openhub.net)
787 - The internals are now compiled in C++ (and will soon be clean C++)
788 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
789 59 klines of C, etc. We added only 29 klines of C++ in replacement.
791 * Backwards Compatibility breaks
792 - Removed Lua simulation bindings (switch to C or Java for that).
793 Lua can still be used to describe platforms
794 - Removed Java kernel plug-ins.
795 Will be reintroduced after the ongoing major internals reorg.
797 - the following functions were removed.
798 They were too specific and should be reimplemented in a generic
799 way, with filter function.
800 - MSG_task_listen_from_host
801 - MSG_mailbox_get_count_host_waiting_tasks
802 - MSG_mailbox_put_with_timeout was removed.
803 Please use MSG_task_send_with_timeout instead.
805 - the SD_application_reinit function was removed. It has been a noop for a while.
806 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
807 replaced by a more flexible mechanism.
808 - The following functions thus do not exist anymore
809 - SD_workstation_get_access_mode
810 - SD_workstation_set_access_mode
811 - SD_workstation_get_current_task
812 - Basic estimation functions have been removed but can easily be replaced
813 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
814 - SD_workstation_get_computation_time => amount / sg_host_speed()
816 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
817 Use VM.getSpeed()*load/100 for the legacy behavior.
819 - option enable_tracing was removed. It was not doing anything for a while.
820 - In the ModelChecker:
821 - the model-checker now ptraces the model-checked process which means
822 you cannot use a debugger on the latter anymore (we might make this
823 optional in the feature);
824 - removed soft-dirty page tracking;
825 - remove model-checked side snapshot management,
826 MC_snapshot() and MC_compare_snapshot();
827 - keep the MC_cut() function as a stub (it was not really working
828 in the previous release).
831 * All options are consistently in kebab-case. Old names are kept as alias.
834 * Switch to platform v4 format.
835 - Rename from 'power' to 'speed' the attributes describing the amount of
836 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
837 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
838 - In <host> and <link>, attributes availability and state are gone.
839 It was redundant with state and availability traces, and with peak values.
840 - In <cluster>, attributes availability_file and state_file are gone.
841 It was too complex and unused.
842 - Kill <gpu>. Was not doing anything.
843 - The DOCTYPE points to the right URL:
844 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
845 (the file at this address now documents the changelog since its v1)
846 - A warning is emitted for unit-less values (they are still accepted).
847 - speed. Default: 'f' or 'flops'. Also defined:
848 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
849 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
850 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
851 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
852 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
853 - latency. Default: 's' second. Also defined:
854 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
856 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
857 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
860 * s4u::Host is now the preferred public interface to the Host features.
861 sg_host_* functions are C bindings to the exact same behavior
862 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
865 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
866 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
867 They were deprecated since a few years, and probably did not even compile anymore.
870 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
871 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
872 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
873 encouraged to update their codes. A list of the modified functions can be found at the end of
874 include/simgrid/simdag.h
877 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
878 the lambda or closure passed as a parameter will run in kernel mode.
879 All the callback functions should be rewritten to that interface at some point.
882 * Reorganizing and cleaning the internals all around the place.
885 * Remove old default barrier/bcast buggy algorithms (see #18407)
886 * Various bug fixes to handle more codes
887 * Remove the need for the --foreground option of smpirun (it is still
888 accepted for backward compatibility).
891 * Kill the setset data container: MC don't use it anymore.
892 * Kill the queue data container: it made more sense with GRAS.
893 * Kill the xbt_peer_t data type: it's useless without GRAS.
894 * Kill rm_cb feature of config sets: it was never useful.
895 * Kill graphxml parsing feature. It was not used.
896 * Kill the deprecated code protected by XBT_USE_DEPRECATED
898 - xbt_dynar_sort_strings(), when the content is char*
899 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
900 They throw exceptions on invalid input;
901 * C++ support for declaring CLI flags (simgrid::config::Flag);
902 * class for abstracting different signal backends (simgrid::xbt::signal).
903 with no external dependencies (we need very simple signals).
906 * refactoring and cleanup of the code;
907 * ongoing process to cleanly separate the model-checking algorithms
908 from the code model-checking support.
910 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
912 SimGrid (3.12) stable; urgency=low
914 The Facelift Release.
917 * Require g++ v4.7 at least to not speak prehistorical C++.
918 * Require Boost 1.48 (for signal2 component).
919 * Java must be version 7 at least when activated.
920 * Builds on Windows again (including Java bindings).
921 * Tracing is now always enabled (no way to turn it out)
922 * Remove GTNetS. It was not working anyway.
923 * Various cleanups in the cmake scripts.
924 * Move headers around to sort them out on installed systems:
925 - instr/instr.h -> simgrid/instr.h
926 - instr/jedule/* -> simgrid/jedule
927 - simdag/datatypes.h was removed
928 - simdag/simdag.h -> simgrid/simdag.h
929 - msg/datatypes.h was removed
930 - msg/msg.h -> simgrid/msg.h
933 * Interface improvement:
934 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
935 - Sanitize the interface in MSG_task_ module:
936 - Merge two functions that were close enough but misleading:
937 set_compute_duration(t) -> set_flops_amount(t)
938 get_remaining_computation(t) -> get_flops_amount(t)
939 - set_data_size(t) -> set_bytes_amount(t)
940 get_data_size(t) -> get_bytes_amount(t)
941 - Massive cleanups in the functions related to the energy
942 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
943 - New: MSG_host_get_pstate()
945 - msg/energy/onoff: switching hosts on and off
947 * Interface improvement:
948 - Rename Host.isAvail() to Host.isOn()
949 - Rename Process.currentProcess() to Process.getCurrentProcess()
950 - Rename Task.setDataSize() to Task.setBytesAmount()
951 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
953 - #18874: Actually allows the GC to reclaim tasks
957 - SIMIX_process_throw: raises an exception in a remote process
958 * Refactoring: Separate sub-modules
959 - libsmx: the public interface, as libc in a real system
960 - popping: the strange dance that converts a user request into a kernel handling
961 - smx_context_*: the virtualization mechanisms that embed the user code
962 - smx_*: the handling of each simcalls
964 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
965 * Rename smx_action_t into smx_synchro_t, making explicit that these
966 things are used to synchronize processes with their environment.
967 For example, a communication is a sort of synchronization involving
968 the communicating processes (that may block until the exchange) and
969 the platform. The same can be said from computations, etc.
971 - #18888: segfault when a process ends before its kill_time
975 - 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.
976 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
977 - MPI_Type_set_name, MPI_Type_get_name
978 - MPI_*_c2f and MPI_*_f2c functions
979 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
980 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
981 - Activate a lot of new tests from the mpich 3 testsuite
983 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
984 - InfiniBand network model added: Based on the works of Jerome Vienne
985 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
986 - When smpi/display_timing is set, also display global simulation time and application times
987 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
988 * Collective communications
989 - 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
990 - 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/).
991 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
993 - Replay now uses algorithms from wanted collective selector
994 - Replay can be used with SMP-aware algorithms
995 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
996 - 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.
998 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
999 - Correctly use loopback on fat-tree clusters
1000 - Asynchronous small messages shouldn't trigger deadlocks anymore
1001 * Energy/DVFS cleanup and improvement
1002 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1003 - new: smpi_get_host_pstate()
1006 - "Full" network optimization flag was broken since Surf++
1007 - Better handling of precision flags in maxmin
1008 - Fix bug causing sometimes "Impossible" errors
1009 - Properly pass cluster properties to included hosts
1010 * Improvement of the Energy plugin.
1011 - Always update the consumption before returning that value
1012 - New property: watt_off to denote the dissipation when the host is off
1013 - New functions getWattMinAt and getWattMaxAt to retrieve the
1014 dissipation of pstates that we are not currently at.
1015 * Java: class NetworkLink renamed to Link
1016 * New function: simcall_process_get_kill_time()
1017 * Massive rename s/workstation/host/
1018 - That's intrusive, but that's good for the project consistency. Sorry.
1019 - Change config option "workstation/model" into "host/model"
1022 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1023 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1024 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1025 - xbt_dirname and xbt_basename for non-POSIX systems
1027 * The model checker now runs as a separate process.
1028 * The model checker runs must now be launched with the new simgrid-mc program.
1029 * Record/Replay: the MC can display a textual representation of a path in the
1030 execution graph. It can then be replayed outside of the model checker.
1032 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1034 SimGrid (3.11) stable; urgency=low
1039 * Normalizing pointers addresses tool for better diff between logs
1042 * Add cloud examples using new VMs
1043 - examples/msg/cloud/two_tasks_vm.tesh
1044 - examples/msg/cloud/simple_vm.tesh
1045 - examples/java/cloud/cloud.tesh
1046 - examples/java/cloud/migration/migration.tesh
1047 * Add java surf examples:
1048 - examples/java/surfPlugin/surf_plugin.tesh
1049 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1050 - examples/java/surfCpuModel/surf_cpu_model.tesh
1051 * Add SMPI+MSG example:
1052 - examples/smpi/smpi_msg_masterslave/
1057 - msg task destroy cancel test
1058 - msg_host on/off test
1059 * Move all tests in testsuite to teshsuite (adding tesh files)
1060 * Restructure teshsuites
1061 - one folder for each kind of test
1062 * Restructure AddTests.cmake
1064 - structure the order of tests (with sections)
1067 * Add virtual machine
1068 - creation of a VM on a PM
1069 - migration of a VM from a PM to another PM
1071 - MSG_process_join(msg_process_t process, double timeout)
1072 - msg_bar_t MSG_barrier_init(unsigned int count)
1073 - int MSG_barrier_wait(msg_bar_t barrier)
1074 - void MSG_barrier_destroy(msg_bar_t barrier)
1075 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1076 * New option "msg/debug_multiple_use" to help debugging when a task is used
1079 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1081 - Increase code coverage in test suites
1085 * Protect context stacks against stack overflow. The number of protected memory
1086 pages allocated on the top of each stack (1 by default) can be configured
1087 with the new command line option --cfg=contexts/guard_size:<value>.
1088 * Simcalls are now generated by a python script that
1089 - generates files included by SimGrid
1090 - checks that all the functions exist, and proposes prototypes
1092 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1093 host_set_data, host_get_data
1095 - simcall_process_join(smx_process_t process, double timeout)
1096 * Fix bug where sleeping processing could not be suspended.
1099 * Translate surf models from C to C++
1100 - Generic classes for all models: Model, Resource, Action
1101 - A generic interface for each kind of model (CPU, Network, Storage
1102 Workstation, WorkstationVM)
1104 * Translate surf routings from C to C++
1105 * Add callbacks using sigc++ or boost::signals2
1106 - Add callback functions for resource creation/destruction
1107 - Add callback functions for action state change
1108 - Handle Energy as a plugin
1109 * Replace swag by boost::intrusive
1110 * Add new routing models for clusters. For documentation, see
1111 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1112 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1113 parameters for cluster tag
1114 - Fat trees, with topology="FAT_TREE" and
1115 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1116 - see examples/platforms/torus_cluster.xml and
1117 examples/platforms/fat_tree_cluster.xml
1118 * More documentation
1121 * Hostfiles support host:nb_processes construct to deploy several processes on
1123 * Collective communication algorithms should not crash if used with
1124 improper number of nodes and report the error.
1125 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1126 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1127 MPI_Dims_create, MPI_Cart_sub are supported.
1128 * New interface to use SMPI programmatically (still depends on MSG for
1129 some parts, see examples/smpi/smpi_msg_masterslave):
1130 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1134 * Global variables privatization in MPI executables is now performed at runtime
1135 with the option smpi/privatize_global_variables (default:no).
1136 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1137 dynamic libraries loaded by the application are not privatized (static
1138 linking with these libraries is advised in this case)
1141 * Options defined in XML work correctly now.
1144 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1145 copied into simgrid.jar or not (ON by default). Use this option if you want
1146 to reduce the size of the installed simgrid.jar, *and* the native libraries
1147 are kept installed elsewhere.
1148 * Surf binding with SWIG (code generated in maintainer mode only):
1149 - plugin to handle callbacks
1150 - CPU model only for the moment
1153 * Supernovae build mode is definitively removed. It was used to improve
1154 inlining and inter-module optimizations. It is nowadays superseded by
1155 link-time optimizations commonly available in compilers.
1156 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1157 * Add boost dependency for surf++
1158 * Add new macro for tests
1159 - ADD_TESH(name <tesh_args>)
1160 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1164 - xbt_bar_t XBT_barrier_init(unsigned int count)
1165 - int XBT_barrier_wait(xbt_bar_t barrier)
1166 - void XBT_barrier_destroy(xbt_bar_t barrier)
1167 * Make the xbt_os_time module public
1169 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1171 SimGrid (3.10) stable; urgency=low
1173 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1176 * Reintegrate Java to the main archive as desynchronizing these
1177 package is not acceptable anymore (Java is now considered stable)
1178 * Add explicit synchronization facilities through semaphores
1179 * Bug fix: Task.setDataSize() only changed the C world, not the value
1180 cached in the Java world
1183 * Dramatically change the way files are handled. API and internals changed, but
1184 this part of MSG was not considered as production grade either.
1185 * Add explicit synchronization facilities through semaphores
1186 * Add a new function MSG_host_get_process_list()
1187 * Preliminary DVFS support (see examples/msg/energy/ for details)
1190 * SMPI is now included directly in the libsimgrid as the windows
1191 linker doesn't force us on splitting it anymore.
1192 * Improvements of the SMPI replay tool:
1193 - Most of the collective communications are now rooted in the same process as
1194 in the original application.
1195 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1196 used until now). Multiple data types can now be used in a trace.
1197 - The replay tool now supports traces produce either by TAU or a modified
1199 - Bug Fix: the compute part of the reduce action is now taken into account.
1200 - Gatherv collective is now supported
1201 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1202 of smpirun outputs time independent traces for the current run. One file
1203 is created per process. If too many processes are simulated, this behavior
1204 can be changed to one file for all processes by using the
1205 tracing/smpi/format/ti_one_file flag
1206 * smpirun generates the host file if needed (with given host count and platform)
1207 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1208 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1209 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1210 --cfg=smpi/coll_selector:(mpich/ompi)
1211 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1213 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1214 happening inside a collective SMPI call.
1215 * Fix the behavior of complex data types handling.
1216 * Make MPI_Wtime another synchronization point to take computations into
1218 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1219 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1220 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1221 (no privatization of global variables yet)
1222 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1225 * Verification of liveness properties is now available for SMPI applications
1226 (in addition to MSG applications)
1227 * Bugged examples using SMPI in examples/smpi/mc/
1228 * Add --cfg=model-check/visited option. Allows the verification of infinite
1229 programs. Detection of loops in the execution thanks to the system state
1230 comparison and reduction of the state space to explore. Can be combined with
1231 DPOR for safety properties.
1234 * Allow to change SimGrid configuration (see --help) within the code
1235 thanks to SD_config() as it can be done in MSG.
1236 * Add a new function SD_task_set_amount() upon user request.
1239 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1240 * Remove rule based routing (no more PCRE dependency)
1241 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1242 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1243 * Add a loopback_bw and loopback_lat options to cluster tag.
1244 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1245 routing. Now peers are encapsulated in an AS and have their own private
1246 router but this is transparent.
1249 * Our own implementation of getline is renamed xbt_getline, and gets
1250 used even if the OS provide a getline(). This should reduce the
1251 configuration complexity by using the same code on all platforms.
1252 * New type: xbt_cfg_elm_boolean.
1253 * Allow to use yes/no for boolean configuration options in the command line.
1254 * Allow to disable SimGrid cleanups at exit from command line option.
1255 There are situations where one may want a simulation to end with an exit.
1256 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1257 annoying when scripting around the simulator. Adding a
1258 --cfg=clean_atexit:no allows to circumvent this issue.
1261 * Lots of memory leaks were corrected in this release.
1262 There is no known memory leaks anymore, in all of our 600+ tests.
1263 * New command line option --version, to get SimGrid version information.
1264 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1266 * Supernovae builds are deprecated, and expected to be removed in the next
1269 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1271 SimGrid (3.9) stable; urgency=low
1273 The Grasgory release: GRAS is really dead now.
1275 * Complete overhaul of the internal host structures scheme.
1278 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1279 was considered as experimental and badly maintained since then.
1280 * Keeping it was thus a trap to our potential users, that could take
1281 it instead of MSG or SMPI by mistake despite is pity state.
1282 * GRAS seems to have very few users (if any), and no one volunteered
1283 to maintain it further. It also induces a lot of XBT code (for
1284 portability sake), that must be maintained too.
1285 * For all these reasons, we killed GRAS. If someone wants to revive it
1286 in the future, don't cry, our git history still remembers of GRAS.
1289 * Major overhaul. Merge our documentation again as time proved that
1290 splitting it was really not helping our users.
1291 * Further improve the developer documentation to help newcomers
1292 hacking on SimGrid itself. The user documentation (and in
1293 particular, the beginner documentation) is still in a sorry state.
1296 * Now works on Windows too!
1297 * Much more extensive test suite, from MPICH
1300 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1301 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1302 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1303 * Bug fix: let task be scheduled when the last dependency to be solved is
1304 a control dependency.
1305 * Remove SD_load_environment_script function.
1306 Use the C sg_platf function if you want to declare a platform
1310 * New function: MSG_process_get_number()
1311 * Old function documented: MSG_config()
1312 * Remove MSG_load_platform_script function
1313 Use the C sg_platf function if you want to declare a platform
1317 * Change the default value of the TCP_gamma constant (maximal size of TCP
1318 congestion window) to a more realistic 4MiB value. If you notice changes in
1319 your simulation results, you can fall back to the previous 20k tiny window
1320 by adding --cfg=network/TCP_gamma:20000 on command line.
1321 * (Hopefully) fix a bug wrt periodic availability and state traces
1322 * Bug fix: use default values at start when first event in availability/state
1323 trace is not at time 0.
1326 * remove the "new_" part of function name sg_platf_new_trace_connect
1327 (resulting in sg_platf_trace_connect), since it does not create
1331 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1332 use case where it's really mandatory, and maintaining it was a pain in
1334 * New: xbt_fifo_search(), search an item with a user-provided
1335 comparison function instead of dumb pointer comparison.
1338 * Fix the lua deployment:
1339 Use `simgrid.init_application()` before deployment instead of
1340 `simgrid.msg_register_application()` after.
1343 * Transfer the tracing files into the corresponding modules.
1345 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1347 SimGrid (3.8.1) stable; urgency=low
1349 The "we are told that some people want to also *install* the simgrid
1352 * Add missing file "tesh.1" to the archive.
1354 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1356 SimGrid (3.8) stable; urgency=low
1358 The Psssshiiiit release: SimGrid jumps into the Cloud.
1361 * Add an experimental interface to manipulate VMs. They are mainly
1362 process groups with very few intrinsic semantic, but they should
1363 allow you to build the semantic you want easily.
1364 * New function: MSG_host_set_property_value()
1365 * New function: MSG_process_on_exit(). To clean memory in all cases.
1366 * Bug fixes that made the host (and link) failures unusable.
1367 * Add a way to auto-restart process when the host in which they are
1368 executing comes back (ON_FAILURE="RESTART" on deployment file,
1369 MSG_process_auto_restart_set).
1370 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1371 please stop using the old ones, they are DEPRECATED.
1373 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1374 Please use MSG_init() instead. (reducing the amount of entry
1375 points in the library helps us).
1376 * Make it impossible to link against the wrong version of the lib
1377 * Deprecate MSG_clean(). No need to call it anymore.
1378 * Function MSG_get_host_number() is not deprecated anymore.
1381 * Split the doc into a user guide and a reference guide.
1382 * Start a developper guide to help people hacking on SimGrid.
1385 * Enable tracing by default. This modules rocks you should use it.
1386 * Remove option custom_flags. Now use environment variables CFLAGS
1388 * Use default cmake things to detect lua instead of home grown ones.
1389 * New option "enable_mallocators" to disable mallocators, for debugging
1390 purpose ("on" by default).
1393 * Bug fixes around the resource failures: don't let the processes
1394 survive the host they are running onto.
1395 * Add an interface to auto-restart processes when the host in which they are
1396 executing comes back.
1397 * Ensures that SIMIX_clean is called automatically. It's not part of
1398 the public interface anymore (bindings should be updated).
1401 * Bug fix for when SD_Simulate is called with a positive value: be careful
1402 when comparing doubles. Sometimes they are different for non significant
1404 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1405 parallel task whose initial work is distributed among host according
1406 to the Amdahl's law. Such tasks are created with a parameter alpha
1407 that corresponds to the non-parallelizable part of the computation.
1408 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1409 between two sets of workstations assuming a 1D block distribution (each
1410 workstation owns a similar share of data) on both sides.
1412 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1413 Data redistribution will be automatically scheduled once parent and child
1414 are both scheduled. The filling of computation_amount and
1415 communication_amount structures is now done seamlessly thanks to the chosen
1417 * New function SD_workstation_dump to display various information
1418 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1419 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1420 bandwidth on the route onto which the task is scheduled and the amount of
1422 To divide the nominal bandwidth by 2, the rate then has to be:
1423 rate = bandwidth/(2*amount)
1424 * Compute tasks that have failed can now be rescheduled and executed again
1425 (from their beginning)
1426 * Increasing source code coverage (src/simdag is now covered at 95.8%
1430 * Re-implement time-independent trace replay using SMPI (at the
1431 smpi_smp_* level) instead of MSG. This should replace
1432 examples/msg/actions/actions.c
1433 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1434 hindexed and structs)
1435 * Implement the exchange of non-contiguous data.
1436 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1437 * Correct behavior of smpi/sender_gap and set its default value to 0
1438 * Add option to asynchronously send small messages to allow better
1439 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1440 specifies the size in bytes under which messages will be asynchronously sent.
1441 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1442 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1445 * New C interface to define a platform: XML is now optional.
1446 For more info, please check include/simgrid/platf.h.
1447 * New interface to define random platforms from the C:
1448 For more info, please check include/simgrid/platf_generator.h and
1449 examples/msg/masterslave/masterslave_platfgen.c
1450 * Export a sg_cmdline dynar containing all the arguments we got from
1454 * Two new tracing options for adding comments to trace file so you
1455 can track your experiments (see --help-tracing for details).
1456 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1457 * Adding the SimGrid version that generated the trace file as a comment.
1458 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1459 * Fix to avoid key clashes on Paje links
1460 * Other minor fixes related to the Paje specification
1463 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1464 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1465 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1466 * Greatly improve the robustness of mmalloc to user errors (such as
1467 using an area after freeing it, or freeing it twice)
1469 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1471 SimGrid-java (3.8.1) stable; urgency=low
1473 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1474 Warning: all methods are visible, but only some of them are
1475 implemented so far. Check the source (src/jtrace.c)
1476 for further information.
1477 * New module: org.simgrid.msg.File (SimGrid File management functions)
1478 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1479 * Change the meaning of Process.restart: now restart the process from
1480 the begining, like MSG_process_restart in C.
1481 * Add Process.setAutoRestart: handling of process restart when failed
1483 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1484 you to retrieve the properties of the processes/hosts
1485 * Deprecate Msg.clean(): you can just forget about it now.
1486 * New function Process.getCount(), that only works when compiling
1487 with the not yet released version 3.9 of the C library.
1489 * New context factory based on Coroutines. It mandates a modified JVM
1490 but then, the simulations run about five times faster, and there is
1491 no limit to the amount of processes (beside of the available memory).
1493 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1495 SimGrid (3.7.1) stable; urgency=low
1498 * Restore the prototype of MSG_process_create_with_environment() to
1499 the pre-3.7 situation by removing the kill_time argument.
1500 * Add a MSG_process_set_kill_time() function instead.
1503 * Fix weird behaviors when dealing with parallel tasks.
1506 * Simgrid is now built as a dll.
1507 * Simgrid-java now works on Windows.
1508 * Simgrid-Java is now included into Windows package.
1511 * First pre-build package for MacOSX.
1514 * Fix compilation when using MSG_USE_DEPRECATED.
1515 * Fix some compilation issues on Macs and Windows.
1516 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1517 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1519 We discovered that the Lua console is broken, but we are missing the
1520 manpower to fix it right now. The problem existed in 3.7 too, so we
1521 are not blocking the release for that. Sorry if you depended on this
1522 feature, any help would be really welcome.
1524 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1527 SimGrid-java (3.7.1) stable; urgency=low
1529 The "Java aint got to be bloated and slow" release
1532 * Various internal cleanups and performance improvement
1533 Simulations are expected to run up to twice faster or so
1534 * Make Process.kill(process) an instance method, not a static one
1535 * User processes are not java.lang.Thread subclasses.
1536 This breaks the compatibility (sorry), but previous API was
1537 brain-dead, making it impossible to have non-trivial
1538 initializations in the process constructor.
1539 * Require a full constructor per Process sub-class.
1540 Kinda breaks the compatibility (sorry), but this allows a much more
1541 efficient way to launch the processes at simulation startup.
1542 * Do not embeed our version of semaphores, java 1.5 can be considered
1543 as sufficiently prevalent for us to not dupplicate its features.
1547 * Add examples for almost every part of the API
1548 We spotted and fixed a lot of bugs in the process
1550 * New module: asynchronous communication API
1551 * New function: Process.sleep()
1552 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1553 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1554 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1556 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1558 SimGrid (3.7) stable; urgency=low
1560 The "spring cleanups (before the next Big Project kicks in)" release.
1563 * We can specify the SMPI latency/bandwidth factor with command line
1564 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1565 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1566 You can also use the "config tag" from platform file by adding this line
1567 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1568 (see "example/platforms/tag_config.xml" to use "config tag").
1569 Note that the command line supersedes the platform file configuration.
1570 * Change the correction factors used in LMM model, according to
1571 the latest experiments described in INRIA RR-7821.
1572 Accuracy should be improved this way.
1573 * Use the partial invalidation optimization by default for the
1574 network too. Should produce the exact same results, only faster.
1575 * Major cleanup in surf to merge models and split some optimization
1576 mechanisms from the core of the models. As a result you can now
1577 specify which model to use (e.g., --cfg=network/model:LV08
1578 --cfg=cpu/model:Cas01) and which optimization mode to use
1579 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1580 Incompatible combinations should err at initialization. See
1581 --help-models for the list of all models and optimization modes.
1582 * The CLM03 workstation model was dropped for simplicity because it
1583 used the deprecated CM02 network model. Use default instead.
1584 * Rename the TCP_gamma configuration option to network/TCP_gamma
1585 * Rename the coordinates configuration option to
1586 network/coordinates, and document it
1587 * Use now crosstraffic keyword instead of the terribly misleading
1588 fullduplex keyword. It is activated by default now in the current
1589 default model, use --cfg=network/crosstraffic:0 to turn it off.
1590 * Ongoing refactoring the model parsing to make XML files optional
1591 See include/simgrid/platf.h for details (still to be completed)
1594 * Major overhaul of the documentation. Almost instructive now :/
1595 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1596 functions and friends. This interface was considered as
1597 deprecated since over 2 years, it's time to inform our users that it is.
1598 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1599 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1600 use these (crappy) functions in your code.
1601 These functions will be removed soon. Stop using them now.
1602 * Deprecate MSG_get_host_{table,number}
1603 Implement MSG_hosts_as_dynar() instead.
1604 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1605 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1608 * Stabilize the parallel execution mode of user contexts
1609 * Introduce configuration variables to control parallel execution:
1610 - contexts/synchro: Synchronization mode to use when running
1611 contexts in parallel (either futex, posix or busy_wait)
1612 - contexts/parallel_threshold: Minimal number of user contexts
1613 that must be part of a scheduling round to switch to parallel
1614 execution mode (raw contexts only)
1615 * Fix bugs that prevented to use suspend/resume along with
1616 synchronization structures.
1617 * Fix bugs in process termination that lead to invalid memory access
1618 in very specific conditions.
1621 * Introduce a parallel mode for the models (controlled by surf/nthreads
1622 configuration item). In our tests, running the models in parallel
1623 never lead to any speedups because they are so fast that the gain
1624 of computing each model in parallel does not amortizes the
1625 synchronization costs, even when ultra fast futexes are used.
1626 This is released anyway because YMMV.
1629 * Performance boost by using a swag internally to compute the set of
1630 tasks that are finished and should constitute the return value of
1634 * Enable it by default now that it is considered rather stable.
1637 * Documentation of the tracing functions.
1638 * Performance gains when tracing categorized/uncategorized resource
1639 utilization by avoiding calls to get route when updating resource
1640 variables. LMM constraints are being used instead.
1641 * API changed to set task categories. Use MSG_task_set_category instead
1642 of TRACE_msg_set_task_category, and SD_task_set_category instead
1643 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1644 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1645 * Examples for link user variables added (see at examples/msg/tracing/)
1646 * Deprecated function TRACE_msg_set_process_category completely removed
1647 * Trace header updated according to the latest Paje file format
1648 * Tracing network lazy updates, no longer obligate users to use full updates
1649 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1650 * Experimental: let user code declare/set/push/pop application states for hosts
1651 * API changed to allow the manual creation of graph configuration files
1652 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1655 * Improve the API of Lua MSG bindings, using the Lua spirit.
1656 * Each simulated process now lives in its own Lua world (globals are
1657 automatically duplicated). It helps writing simulators. Will allow
1658 to run Splay programs within SimGrid in the future.
1659 * Add a Chord example in Lua, equivalent to the MSG one.
1662 * Start the implementation of a solution to express temporal
1663 properties, not only local assertions. This is still an
1664 experimental work in progress, stay clear from it to be safe.
1668 - Add new runtime parameters --help-logs and --help-log-categories
1669 to display information about supported logging parameters and
1671 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1672 don't exists anymore.
1673 * Mallocators: allow value NULL for the reset function.
1675 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1676 create homogeneous dictionaries, where all the elements share the
1677 same free function. Non homogeneous dictionaries will be
1678 deprecated in the next release.
1679 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1680 - Multi-level dictionaries are deprecated.
1682 - new function xbt_dynar_search_or_negative() that is useful when
1683 you have less than 2 million elements in your dynar and don't
1684 want of the extra complexity of catching exceptions when the
1685 element is not found.
1687 - Make xbt_os_thread module (for thread portability) public.
1688 Documentation is still to come, sorry.
1690 - Cleanups and simplifications to make it maintainable again.
1691 - Exotic features (such as memalign and valloc) were removed.
1692 - The metadata were extended and improved so that the
1693 model-checker becomes able to explore and inspect the heaps.
1694 - This may induce a performance drop when enable_model-checking is
1695 ON in cmake (even if it's not used in the simulation), but it is
1696 necessary at this point to get MC working.
1698 Turn model-checking OFF if simulation performance matters to you.
1699 Not enabling it at runtime is not enough, disable it in cmake.
1701 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1704 SimGrid (3.6.2) stable; urgency=low
1706 The "Not coding new stuff allows to polish old things" release.
1709 * New bindings to the NS3 packet level simulator (experimental)
1710 * Use the raw (efficient) execution contexts instead of the sysv
1711 (portable) ones when possible.
1712 * libpcre is now mandatory in any cases since not using it led to
1713 severe performance loss and possibly other issues
1714 * Update the XML platforms:
1715 - G5K: include the latest machine in Nancy
1716 - GridPP and LCG: new platforms
1717 * Documentation was partially updated, at least (more to come)
1719 Bug fixes, cosmetics and small improvements
1720 * Free terminated processes before the end of the simulation to avoid
1721 exhausting the memory of users having very dynamic amount of
1723 * Bug fix and cosmetics about canceling non-running tasks
1724 * Bug fix about the dot loader's issues when using libcgraph
1727 * Create an installer for windows with nsis (amd64 and win32)
1728 - Add an hello world project to illustrate simgrid project creation.
1729 - Embed libpcre into the Simgrid installer to avoid
1730 its compilation burden
1731 * The raw execution contexts should work on Apple now
1732 * Port to Windows 64 bits
1733 - Sysv contexts now have an implementation for this arch
1734 - GRAS communication features now support this arch
1735 * Drop support for borland compiler on windows
1736 - this code was not maintained, and we kinda depend on gcc nowadays
1737 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1738 * Fix portability issue on unstable ubuntu: linker became picky on
1741 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1743 SimGrid (3.6.1) stable; urgency=low
1745 The "Oops, we broke Macs too" release
1748 * Fixed contexts detection so that raw ones are used when possible
1749 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1750 leads to a strange error, with user code segfaulting sometimes when
1751 the generated code is not perfectly aligned (which is not
1752 controllable from the user side, depends on the amount of code)
1755 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1757 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1759 SimGrid-java (3.6) unstable; urgency=low
1762 * Split of every thing from simgrid v3.5 into a separate package.
1764 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1766 SimGrid (3.6) stable; urgency=medium
1768 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1771 * Bindings now constitute their own package, separated from the main one.
1772 Rationale: reduce our maintenance nightmare by reducing the module coupling
1773 They will soon be released on their own on gforge.
1775 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1776 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1778 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1779 * It's not quite deprecated for now because we have no replacement,
1780 but it may soon become the case.
1783 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1784 * Fortran: New user-level cache variable to store the rank of the running
1785 process. This improves performance by an order of magnitude.
1786 * C: New coccinelle script to automatically locate and modify global and
1787 local static variables.
1788 * Improved SMPI network model with a sender-side gap to account for multiple
1792 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1793 only say if a communication is finished, no matter whether it succeeded or
1794 failed. You can call MSG_comm_get_status() to know the status of a finished
1796 * New function MSG_task_dsend() to send a task and detach it. When a
1797 communication is detached, you are never notified of its success or failure
1798 and the memory is released automatically once it is finished. This function
1799 is useful when you don't care about the end nor the success of a
1801 * Change the prototypes of action replay. Sorry for inconvenience,
1802 but this is really more efficient this way (and to adapt your code,
1803 you just have to fix the initialization, that shouldn't be too long)
1804 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1805 ever managed to do anything useful with it.
1806 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1807 takes a dynar of communications. It returns immediately and gives the
1808 index of a finished communication (if any).
1809 * New example: a basic implementation of the Chord P2P algorithm.
1812 * New model for multi-core CPUs. You can now use the core attribute to
1813 precise the number of cores of a host. This is a basic model. Every
1814 process running on the host receives at most the power provided in
1815 the DTD (throughput<=power). Total throughput of process cannot exceed
1817 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1818 router linked by an up-link and a down-link (possibly asymmetrical).
1819 This kind of pattern allows to easily build last-mile model style platforms.
1820 Aggregating such patterns in a rule-based AS is thus the technique of
1821 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1822 * New model for Vivaldi routing. We transformed the Vivaldi network model
1823 into a Vivaldi routing model (based on the rule-based model). This allows to
1824 combine Vivaldi based latencies with last-mile platforms.
1827 * Added a check for NaN of IEEE754 infinite in the double entries of
1829 * Introduce a new context factory "raw", highly inspirated from the
1830 ucontext factory, but using manually crafted functions in assembly to
1831 do the work in an efficient manner.
1832 * Allow to change the used context factory at run time, not only at
1833 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1834 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1835 contexts at runtime (only with raw contexts or ucontexts).
1836 * Completely rewrote this module to allow parallel execution of user
1837 processes. Use --cfg=contexts/nthreads:N to execute user processes
1838 with N parallel threads (the default is 1, meaning no parallelism).
1839 * Allow to decide dynamically between sequential and parallel modes.
1840 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1841 processes in parallel only when their number is greater than or equal to P
1843 * Added a check for NaN of IEEE754 infinite in the double entries of
1847 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1848 won't output the state of processes when interrupted with Ctrl-C
1849 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1850 NULL-terminated array. This may solve backward compatibility issues
1851 due to the change to return type of SD_simulate. See also:
1852 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1853 * Add new macros with variable number of arguments.
1854 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1855 - in xbt/asserts.h: xbt_assert
1856 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1857 - in xbt/ex.h: THROWF and RETHROWF.
1858 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1860 * Change xbt_die() to accept a format string with arguments, just like printf.
1861 * New data structure: xbt_lib_t, like a dict but more general and with better
1865 * New configuration options
1866 Options triva/categorized and triva/uncategorized can be used to generate
1867 graph configuration files for Triva visualization tool.
1868 * Configuration option tracing/platform is renamed to tracing/categorized
1869 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1870 * New TRACE_user_link_variable interface:
1871 User provides the name of the link and the tracing variable to attach to it
1872 * the declaration of tracing categories must be done after the environment
1874 * simpler tracing interface, just one way to declare categories
1875 TRACE_category or TRACE_category_with_color, it is up to you
1876 * links in the trace file are again identified by their names
1877 * trace contains the full platform hierarchy exactly as declared using the ASes
1878 * Options tracing/msg/[task|process]:1 groups the process by hosts
1879 for both cases, tasks and processes must have names that are unique during the simulation
1880 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1881 * The experimental option tracing/msg/volume is deprecated
1882 its functionality may be reincorporated if needed
1884 The tracing generates a trace file with unordered timestamped events,
1885 because of the way the core simulator (surf) works. A script available
1886 at the tools directory (fix-paje-trace.sh) can be used to put the events
1887 in order. We have changed the tracing so it can generate ordered timestamped
1888 events in the final trace, but depending on the simulator (and how much time
1889 is simulated) that can lead to a huge memory utilization. It is deactivated
1890 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1892 Build Infrastructure
1893 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1894 - We are trying hard to keep the API stable, but it may happen that
1895 some things change (we're a research project after all, not a
1896 nuclear plant operating system). If such things should happen, you
1897 could rely on that macro to adapt.
1898 - current value: 30600 for 3.06.00, aka 3.6
1899 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1900 a number that can be compared with SIMGRID_VERSION.
1901 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1902 assertions and verbose/debug logging events are disabled at compile time.
1904 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1906 SimGrid (3.5) stable; urgency=medium
1909 * New feature to any SimGrid-based simulator: Model-Checking
1910 Check SIN#1 for more details.
1913 * New Model SMPI (three-interval linear regression for correction factors)
1914 See RR-7426, available at http://hal.inria.fr/inria-00527150
1915 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1916 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1917 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1918 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1919 * See also src/smpi/README
1924 - Tracing API changes: TRACE_start and TRACE_end should not be called
1925 by user-code. They are automatically called by simulators created
1926 with SimDAG, MSG and SMPI if the toolkit is compiled with
1927 tracing_enabled=ON. Categories declaration and utilization remain the
1928 same for MSG and SimDag.
1929 - A function was added to the tracing API to declare categories with
1931 - TRACE_category_with_color (char *category, char *color)
1932 where color must be in the following format
1933 "%f %f %f", red, green, blue
1934 and red, green, blue are float values in the interval [0, 1]
1935 - User can specify NULL as color parameter, or continue calling
1936 TRACE_category (cat)
1937 On that case, the tracing system will define random colors
1938 - The following command-line options are supported:
1939 --cfg=tracing/filename:msg.trace
1940 --cfg=tracing:1 (activate tracing, needed to use others)
1941 --cfg=tracing/platform:1 (categorized resource use)
1942 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1943 --cfg=tracing/msg/task:1 (task creation)
1944 --cfg=tracing/msg/process:1 (process creation, migration)
1945 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1946 --cfg=tracing/smpi:1 (SMPI interface tracing)
1947 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1948 - examples of examples/msg/tracing updated
1950 - DAXLoader and DOTLoader functions can generate tasks with categories
1951 - A new function to attribute a category to SD tasks:
1952 TRACE_sd_set_task_category (SD_task_t task, char *category)
1953 * Tracing the MPI interface implemented by SMPI
1954 - Collective operations are traced with states
1955 - Point-to-Point operations are traced with states/links
1956 - Tracing activated by a parameter "-trace filename" passed
1957 to smpirun during execution (considering that simgrid
1958 is compiled with tracing enabled)
1959 - To run the simulation with gdb, the simulator
1960 accepts --cfg=tracing/smpi:1 to trace SMPI
1961 - tesh files to check if smpi tracing is ok
1962 - See examples/smpi/NAS/DT-trace
1963 * GTNetS tracing re-worked
1964 - adaptation to the tracing system of GTNets to cope
1965 with modifications regarding the fullduplex mode
1966 - new tesh files to check if gtnets tracing is ok
1969 * Asynchronous communications through the functions:
1970 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1971 * New function: MSG_load_platform_script()
1972 to make possible using a lua script instead of XML files to set up platforms
1973 * New function: MSG_set_function
1974 to associate functions to processes, used when bypassing the parser
1975 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1977 Platforms: Add some more examples in examples/platforms
1978 * Grid'5000: see www.grid5000.fr
1979 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1982 * Change the XML format. This is a very important modification. SimGrid
1983 3.5 introduces a new hierarchical format based on the notion of
1984 Autonomous Systems. Compatibility with old format is ensured
1985 through the perl script provided in the install bin directory
1986 bin/simgrid_update_xml.
1987 It is now possible to build platforms with specific routing
1988 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1989 connect several platforms together. We will try to provide soon
1990 set of realistic platforms exploiting these properties (have a
1991 look at examples/platforms/ for the moment).
1992 * Take the opportunity of the XML format change to be a good XML citizen:
1993 rename link:ctn to link_ctn and similar changes (also dealt with by
1995 * Add a new routing scheme (rule-based) using regular expressions. It
1996 enables to have an extremely low memory footprint when the
1997 underlying routing is simple and can be compactly described. You
1998 need to have libpcre4-dev (perl regular expressions) installed if
1999 you want to use this routing scheme.
2000 * Revive the cluster TAG and allow to easily and efficiently (both in
2001 term of memory and speed) connect clusters together. Have a look
2002 at teshsuite/simdag/platforms/ to see how this can be done. With
2003 this tag, you can create clusters with thousands of tasks at no
2004 cost (have a look at examples/platforms/).
2005 Note: clusters are implemented as ASes, so there is no need for an
2006 enclosing AS tag if you have only one cluster in your platform.
2007 * Add new generic functions in the public interface that allows the user
2008 to call SURF 'create_resource' methods from your code (same
2009 functionality as the XML bypass mechanism but with a much lighter
2011 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2012 that uses a piecewise linear approximation to produce better
2013 results when exchanging small messages.
2014 * Add a new parameter to handle correctly full duplex link and account
2015 for interferences between uplink and downlink communications
2016 (activate with --cfg=fullduplex:1).
2019 * Rename the SD_READY (all dependencies are satisfied and task is
2020 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2021 dependencies are satisfied) state.
2022 This prevents a confusion between the notion of "ready to schedule"
2023 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2024 simulated" (SD_RUNNABLE) used by the simulation kernel.
2025 * Change the way a task is considered as ready. Instead of removing
2026 dependencies when a task is done, a counter is decreased. This way,
2027 it is always possible to reach ancestors thanks to the
2028 SD_taks_get_parents function (even after the end of the simulation.)
2029 * Change the return type of SD_Simulate from (SD_task_t*) into
2030 xbt_dynar_t. This function was in handling a dynar internally and
2031 converted it into a NULL terminated array for historical reasons.
2032 * New function SD_dotload(char*) to load a DAG described in dot
2033 format. This loader and the corresponding examples require the
2034 installation of the graphviz library.
2035 * Fix a bug in the management of tasks of size 0 in the surf network
2036 models. This problem was only visible with SIMDAG and you should
2037 thus disregard results produced with earlier versions if you
2038 relied on this feature (some tasks were blocked because of this).
2039 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2040 even though your applications were doing only point-to-point
2041 communications and sequential computations. Now you can really use any
2042 model you want (of course, if you create real parallel tasks, which are
2043 not implemented in most models beside ptaskL07, this will abort).
2044 * Add an example that schedules a DAX on an heterogeneous platform
2045 using a Min-Min strategy.
2046 * New function SD_workstation_get_current_task() that returns the kind
2047 of task currently running on a workstation in the sequential access
2049 * Raise some warnings when unexecuted tasks remains at the end of the
2050 simulation. This is usually caused by cycles in the DAG.
2053 * New function: SIMIX_process_set_function() called by MSG_set_function
2054 * Change the underlying waiting queue in semaphores so that a process
2055 can wait on several of them simultaneously (as in waitany).
2056 * Fix the way to handle tokens in semaphores so that all access patterns
2057 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2058 * kill the dirty pimple SIMIX_message_sizes_output()
2059 Please use (proper) visualization instead
2062 * New data container: setset (set of sets of elements)
2063 * New module: mmalloc (mapped malloc, allowing to have several
2064 independent segments of malloc)
2065 * New function: xbt_dict_cursor_set_data()
2066 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2067 * New function: xbt_dynar_is_empty()
2068 * New function: xbt_fifo_get_last_item()
2069 * Fix xbt_dynar_shrink(): use the right element size.
2070 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2071 uninitialized areas during expand.
2072 * Fix semaphores: previous implementation was severely broken.
2073 * Use library init/fini functions for our initialization.
2074 - you can use logs and other feature as soon as you want in your
2075 code (even before the xbt_init / MSG_init)
2076 - xbt_exit is now a no-op and produce a warning when used.
2079 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2080 benefit from the latest improvement to the simulation kernel.
2081 * Kill measurement sockets for now. If you rely on them, sorry. This
2082 release is not for you. This feature will be reintroduced in the
2083 future, but we cannot delay the release any further.
2084 * New function: gras_msgtype_get_name().
2085 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2086 * Fix (at last) the pmm example: it should not randomly fail anymore.
2088 Build chain: bug fixes and overall polishing
2089 * Cmake is now stable enough. Hence, we killed the autotools.
2090 * Port to windows ( TM :)
2091 * Fix the 'make install' target.
2092 No need to use 'make install-simgrid' anymore
2093 * Introduce a 'make dist' target compiling a *source* archive
2094 'make package' compiles a binary archive
2095 * Compile java files only on need
2096 * Add --cd and --setenv command line options to tesh
2097 * Out of source builds are not fully supported yet, but we are close
2098 * Enable supernovae and optimization flags by default for our users
2101 * Add layer to set up environment directly from lua, without XML.
2102 * The effect of gras_stub_generator can be achieved through
2103 lua too (check examples/gras/console/ping_generator.lua)
2105 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2107 SimGrid (3.4.1) stable; urgency=low
2109 The "Polishing easter eggs is probably a good idea" release.
2110 This is a bug fixes release only.
2114 * Fix a bug preventing the tasks from begin garbage collected.
2117 * Fix a bug occuring when a host involved in a communication fails.
2118 This was not detected properly by the other peer involved in the
2119 communication. Now, it's reported as a network error.
2122 * Warn the user about loop dependencies in data flow of DAX files
2123 * Obey the control-flow dependencies of DAX files
2126 * Add option "enable_smpi" allowing to not compile SMPI.
2127 Probably useful for the (Mac) users experiencing a build error here
2128 * Improve the detection of lua5.1 and ruby1.8
2130 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2132 SimGrid (3.4) stable; urgency=low
2134 The "Easter in Cargese" release. Also known as (major changes):
2136 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2137 ~> bindings were greatly improved
2138 ~> new tracing infrastructure for better visualization introduced
2140 * the "Welcome to configury modernity" release.
2141 ~> we switched from autotools to cmake, and improved our cdash
2144 A more detailled list of changes follow (full detail in svn log).
2146 Java Bindings: Various Cleanups
2147 * (install java-gcj-compat-dev on debian-like to use them)
2148 * Remove put/get: no need to export deprecated interface in Java
2149 Use send/receive instead.
2150 * Cleanup the examples and add a README per directory
2151 * Remove example autoDestination (that's the only way to go now)
2152 * Remove example explicitDestination (was a plain copy of basic)
2153 * Make JniException a runtime exception, so that there is no need to
2154 declare the fact that you may encounter such a beast. I guess that
2155 nobody will ever want to survive such error.
2156 * Create specific errors for each MSG case of failure:
2157 host failure, transfer failure, timeout, task cancelled
2158 * Cleanup the exceptions that may get thrown by each function
2159 * Other internal cleanups in Java bindings. Performance still bad :/
2160 Ruby and Lua Bindings: create them
2161 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2162 * That's new and great, you should try them out.
2163 Same functionalities than Java bindings, only even less polished
2165 * Kill the useless "rate" argument of SD_task_get_execution_time()
2166 Everyone used to provide -1 as a value, it was not used, and the
2167 semantic of a possible use wasn't even clear.
2168 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2169 either as comm costs or compute costs to mean that there is no
2170 such thing for that specific task.
2171 * Add a SD_task_set_name() function
2172 * Fix SD_task_unschedule() on typed tasks
2173 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2174 * In DAX loader, accept useless 'level' attributes to <job> since
2175 LIGO DAGs have them (seem to be to ease graphical representation).
2177 * Add an example masterslave_mailbox.c using send/receive and not
2178 the deprecated put/get interface.
2179 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2180 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2182 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2183 (declare MSG_USE_DEPRECATED to still have the old name)
2184 * Add a MSG_task_set_data() function
2185 * About trace replay (see examples/msg/actions):
2187 - Allow to work with splitted trace files for each process
2188 Give the specific trace file as argument of each process,
2189 and call MSG_action_trace_run(NULL)
2190 You can still have one merged file for all processes.
2191 - Fix implementation of collective operations
2192 * Allow task_execute() on 0-sized tasks (closes #10063)
2194 * This is the first release of SimGrid where SMPI is not considered
2195 beta anymore (even if some corners should still be improved)
2196 * Port over the new SIMIX_network submodule (internal refactoring)
2197 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2198 * Implement more missing elements of the standard:
2200 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2202 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2203 MPI_Scan() MPI_Get_processor_name()
2204 - Added implementation of missing case for Alltoall (warning: it's
2205 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2206 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2207 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2208 SMPI_MPI_Allgatherv()
2209 * Bug fixes include:
2210 - MPI_Waitsome() was broken
2211 - Allow relative includes in smpicc
2212 - Command line cfg argument 'reference_speed' was ignored...
2213 - Some functions did not properly lead to auto-benching of user code
2214 - smpicc passes -O2 by default (just like openmpi one)
2216 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2217 * Bug fixes about timeouts during communications
2218 * add SIMIX_message_sizes_output() as a pimple to write to file the
2219 amount of messages per size. Use gnuplot to get histogram.
2220 Pimple because that's the only user-visible function of simix,
2221 defined directly in xbt.h (irk, sorry)
2223 - Add a SIMIX_sem_get_capacity() function
2224 - Fix interactions with processe resume/suspende
2225 - release_forever() was stupidly broken
2226 - Fix SIMIX_display_process_status() for processes in a semaphore
2227 - Make SIMIX_sem_block_onto() user-visible
2228 * Refactoring context stuff:
2229 - Use pseudo-OOP for better modularity
2230 - reimplement SIMIX_process_kill() without process_schedule() so
2231 that the latter can take as invariant that it is called from
2233 - Merge context_start into context_new for sake of simplicity
2235 * Add a Vivaldi network model, coded live during SUD'10 ;)
2236 * Rename configuration variables to start a hierarchy:
2237 o cpu_model -> cpu/model
2238 o network_model -> network/model
2239 o workstation_model -> workstation/model
2240 * New configuration variables:
2241 o network/bandwidth_factor: correction to bandwith
2242 o network/latency_factor: correction to latency
2243 o netwotk/weight_S: correction to the weight of competing streams
2244 * Add a long description to the models, that users can see with such
2245 argument on the command line: --cfg=cpu/model:help
2246 * --help-models display the long description of all known models
2248 * config: add the ability to set a default value after registration
2249 Does not override any previously set value (e.g. from cmd line)
2250 * dict: allow to have integer key and data.
2251 When so, you need to use the following functions
2252 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2253 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2254 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2255 In contrary to regular dicts, the key is not malloced before copy.
2256 Mixing scalar and regular elements in the same dict is not tested
2258 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2259 Tracing for Visualization:
2260 * SimGrid is now instrumented in order to generate a trace file for
2261 visualization analysis: to use it, need to compile SimGrid with the
2262 "tracing" option enabled, and instrument the program using SimGrid with
2263 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2264 (among other functions).
2265 * The instrumentation only traces the platform utilization for now
2266 * Documentation to use the tracing functions and how to analyze the
2267 traces with the Triva tool is written.
2268 * More information about: SimGrid FAQ (in the section Tracing Simulations
2271 * We moved to cmake as default build system. Autotools support will
2272 be dropped soon. Check the FAQ for more info about how to use it.
2273 * Greatly improved our cdash/ctest interactions
2274 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2275 * Added memory checking tests with valgrind; lot of memleak fixing.
2276 This may be the first release of simgrid with so few memory issues
2277 * Added code coverage tests.
2278 Our coverage is still improvable, but at least we see it on cdash.
2280 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2282 SimGrid (3.3.4) stable; urgency=low
2284 The "Desktop Grid needs love too" release (also called Xmas release).
2286 Models improvements:
2287 * Major speedup in the maxmin system solving by using lazy evaluation
2288 Instead of solving completely the maxmin system at each iteration,
2289 only invalidate (and recompute) the modified parts.
2290 This new feature is enabled in default models but you can try to
2291 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2292 * Cas01 IMproved as default CPU model
2293 This CPU model is the same Cas01 model, but it uses the
2294 maxmin-selective-update flag and a heap structure to manage
2295 actions on SURF kernel.
2296 It reduces the complexity to find the next action to finish and,
2297 consequently, it's faster than the old Cas01.
2298 This is the new default CPU model (Cas01).
2299 * Rename the old Cas01 model to Cas01_fullupdate
2300 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2301 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2302 * CpuTI (CPU Trace Integration)
2303 A new CPU model whose objective is simulate faster when using
2304 availability trace files.
2305 Instead of using a full featured, over engineered maxmin system for
2306 CPU modeling, this model does the pre-integration of traces files
2307 to calculate the amount of CPU power available, and so, executes
2308 faster than the old CPU models.
2309 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2310 * Use LV08 as default network model since it gives better accuracy
2311 for small messages and shouldn't change things for big ones.
2312 Use --cfg=network_model:CM02 to get the previous behavior.
2315 ******************************************
2316 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2317 ******************************************
2318 * The new CPU model may changes simulations!
2319 The point is that events occurring at the exact same timestamp
2320 are not scheduled in the same order with the old and new
2321 version. This may be enough to completely change the execution
2322 of simulations in some cases.
2323 * The new network model will change simulations!
2324 This new model is more realistic than the previous one, so you
2325 should consider redoing your old experiments with this model.
2326 Sorry for the inconvenience.
2329 * Introduce the supernovae compilation mode
2330 When compiled that way, the whole SimGrid (or almost) is put in a
2331 single compilation unit and compiled in one shoot.
2332 This is to help gcc which has difficulties to inline stuff from one
2334 The speedup seem to be above 15%, althrough more tests are needed on
2335 amd64 to confirm that gain.
2338 * Port of MSG's mailbox on top of SIMIX network
2339 The put/get mechanism was greatly simplified on the way.
2342 * New SIMIX network module. Provides:
2343 - Mailbox: rendez-vous mechanism to find with who you want to speak
2344 - Synchronous send/recv: easier and hopefully faster since the
2345 logic is handled in the maestro process directly now
2346 - Asynchronous send/recv: you dreamt of it? It's here now
2347 Too bad that nobody cared enough to propagate the change to MSG.
2348 * Add semaphores as SIMIX synchronization mechanism.
2351 * new function SD_daxload(char*) to load a DAX file
2352 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2353 * Introduce typed tasks. Specify its kind and cost at creation.
2354 At scheduling, just give where it should be placed, and the cost
2355 for each involved resource is automatically computed.
2356 Existing constructors so far (more to come of course):
2357 - SD_task_create_comm_e2e() for end-to-end communication
2358 - SD_task_create_comp_seq() for sequential computation
2359 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2360 * new function SD_task_dump() for debuging display
2361 * new function SD_task_dotty(task,FILE*) writing to file the info
2362 about the task in dotty format
2363 * SD_task_dependency_exists() can now cope with having one of its
2364 arguments NULL. If so, it tests whether the other argument has any
2366 * Add getters on list of preceding/following tasks:
2367 SD_task_get_parents(task) and SD_task_get_children(task)
2368 * Add getters on amount of workstations and list:
2369 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2370 * Add getter on task kind: SD_task_get_kind(task)
2371 * Update the start_time and finish_time of tasks on completion/failure
2372 * Bugfix: Remove task from state swags when destroyed
2375 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2378 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2379 dynar with one separate thread per value of the dynar.
2380 * Change the prototype of xbt_thread_create(), sorry.
2381 Added a boolean parameter indicating whether we want to join this
2382 thread (used in SG only for now)
2383 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2386 * GTNetS wrappers should now be usable again (and betterly tested too)
2387 * Fix a major regression from 3.2 where the timeout provided to
2388 MSG_task_put_with_timeout() was used as absolute time before which
2389 the comm should be done.
2390 * Start to fix the <cluster> tag.
2391 - Internal links should be good now (beside of the loopback, which
2392 use the private link instead)
2393 - paths to the external world is still rather broken
2394 - the <route:multi> tag is just broken. Actually that's brain-dead.
2395 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2396 to make it less stupid
2397 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2398 * Fix a source-level compatibility glitch from 3.2: after defining
2399 MSG_USE_DEPRECATED, you can use the old name
2400 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2401 * Allow to compile from the SVN with automake 1.11
2402 * Fix some problems when using the "start_time" tag in deployment XMLs.
2403 * Fix #8569: XBT/synchro.h has redundant declarations
2404 * Fix #8563: MSG return values and exceptions
2405 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2406 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2407 Thanks to Robson Peixoto.
2408 * Fix a vicious bug in dictionaries inducing that some elements were
2409 not freed on xbt_dict_free()
2411 Portability report of this version:
2412 * Main portability targets:
2413 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2414 on (amd64/i386/ia64)
2415 - mac leopard on i386
2416 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2417 but nothing critical.
2418 * Other platforms: windows, AIX and others were not tested for this release
2420 Timing report of this version:
2421 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2422 scenario parameters). From 8h to a few seconds in desktop grid settings.
2423 * Supernovae brings about 25% speedup on i386.
2425 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2427 SimGrid (3.3.3) stable; urgency=low
2429 The "Need for Speed" release.
2431 The timings done to validate the 3.3.2 were faulty.
2432 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2434 The problem was a conversion from a manually handled vector to
2435 xbt_dynar_t on the critical path.
2436 xbt_dynar_foreach calls functions, inducing stack management crap.
2438 We inlined these functions and xbt_dynar_foreach is now breath taking.
2439 We also inlined xbt_swag_belong on the way.
2441 Here are some approximate speedup measurements (on master/slaves
2442 simulations lasting between 10s and 20s each):
2443 3.3.1 -> 3.3.2: about same performance
2444 3.3.2 -> 3.3.3: 40% speedup
2445 3.3.1 -> 3.3.3: 40% speedup
2446 3.3.1 with inline patch -> 3.3.3: 30% speedup
2448 Our reading is that the refactoring which occurred in 3.3.2 made us
2449 suffer much more from the xbt_dynar_foreach low performance, but
2450 once we solved this, this refactoring proved to be very performance
2451 effective. From the 40% speedup, somehow, 10% are due to the
2452 inlining and 30% to the refactoring.
2454 That's a pitty that gcc cannot inline functions placed in other files
2455 alone. We have to choose between:
2456 - break the encapsulation (by putting private data structures and
2457 accessors in headers files to help gcc)
2458 - live with low performance
2459 - switch to a decent compiler such as icc (not quite possible).
2461 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2463 SimGrid (3.3.2) stable; urgency=low
2465 The "Simplicity does not preceed complexity, but follows it" release.
2467 The main contributors of this release were (lexical order):
2468 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2471 * Extract the routing logic into its own object.
2472 (was dupplicated in network.c and workstation_LV07.c;
2473 Allows to implement other ways of storing that info)
2474 => kill now useless network_card concept
2475 - Use dynar to represent routes (instead of void** + int*)
2476 - kill link_set (use surf_network_model->resource_set instead)
2477 - Add a command-line option to choose the routing schema to use
2478 - Add three new models:
2479 * Floyd (shortest path computed at initialization)
2480 * Dijikstra (shortest path recomputed all the time)
2481 * Cached Dijikstra (shortest path computed on need)
2482 All these models where contributed by Silas De Munck, and are
2483 described in his ICCS09 paper.
2485 * Simplify model declaration
2486 (less redirections, less function to write when defining a model)
2487 - Factorize stuff between models:
2490 surf_model_resource_set(model)
2491 surf_model_resource_by_name(model, name)
2492 - Unify the types of models in s_surf_model_t (using an union)
2493 - Embeed fields of common_public directly into s_surf_model_t
2494 - Rename model methods:
2495 action_free ~> action_unref
2496 action_change_state ~> action_state_set
2497 action_get_state ~> action_state_get
2498 - Change model methods into functions:
2499 (model)->common_public->action_use ~> surf_action_ref
2501 * Implement a generic resource; use it as ancestor to specific ones
2502 (allows to kill duplicated code in models)
2503 Drawback: timer command don't need no name nor properties;
2504 workstation_CLM03 don't need no properties
2505 (but I guess we can live with those few bytes wasted)
2507 * Improve the action object model
2508 - implement a constructor avoiding dupplicated code about field
2509 initialization in generic_action part.
2511 * Kill the SDP model: it has an external dependency, is deprecated
2512 in flavor of modern lmm models, and didn't compile since a while
2515 * Relocation of the context module from XBT to SIMIX.
2516 (the context were decoupled from the simix processes, duplicating a lot of code)
2517 => a lot of code was factorized
2518 - less overhead is introduced during scheduling
2519 - simpler API for the context factory
2520 - the logic for process creation,destruction and manipulation was simplified
2521 * Simplification of the s_smx_process_t data structure.
2522 => accesing the simix level data associated to a process is faster now,
2523 and the code is a lot more readable.
2526 * Implement some more MPI primitives:
2527 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2528 -implementation: Bcast: flat or 2-ary tree (default),
2529 Barrier: 4-ary tree,
2531 Allreduce: Reduce then Bcast
2532 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2533 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2534 Alltoallv: flat tree, like ompi
2536 * Add support for optimized collectives (Bcast is now binomial by default)
2537 * Port smpirun and smpicc to OS X
2540 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2541 Shout out if you used it.
2544 * Display the list of still queued messages in SG mode when existing
2548 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2549 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2550 * Add xbt_str_from_file(FILE*)
2551 * Add xbt_dict_get_key achieving a linear reverse search
2552 * Remove the context module
2554 Portability report of this version:
2555 * Main portability targets:
2556 - Linux(debian)/x86/context
2557 - Linux(debian)/x86/pthreads
2558 - Linux(debian)/amd64/context
2559 - Linux(debian)/amd64/pthreads
2560 On these, we still have the eratic breakages of gras/pmm and
2561 amok/saturate_sg reported in previous version. We still think
2562 that the tests are the cause of the fault, not the tested code.
2564 - Mac OSX Leopard/x86/context
2565 Still false negative in tesh autotesting.
2566 Smpi still fails, but this time because readlink does not accept -f
2567 Everything seems to work properly beside of that.
2570 - AIX version 5.3 (only tested contexts this time)
2571 Smpi still fails there because mktemp is not installed.
2572 Everything seems to work properly beside of that.
2574 I managed to compile it for the first time, but several breakages.
2575 Won't delay the release for this exotic platform.
2577 * Windows: it's still lagging behind. If you want to help, please
2580 Timing report of this version:
2581 This version seem to be more than 5% faster than 3.3.1 (on linux
2582 64bits with contextes). The gain is less than expected, we are
2583 investigating this for next release.
2585 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2587 SimGrid (3.3.1) stable; urgency=low
2590 * Implement a --cfg-help to show existing configuration variables
2591 * Build chain do not require doxygen in maintainer mode
2594 * fix a bug on struct sizeof computation, which prevented the
2595 exchange of arrays of structs in some conditions
2596 - added a regression test about this in datadesc_usage
2597 * Allow the exchange of 0-long dynamic vectors.
2598 - for that, use -1 as indicator of dynamic size instead of 0
2599 - This implied to change any size from unsigned long to long,
2600 reducing a bit communication abilities, but I guess that with
2601 64bits being quite common, this is more than enough.
2602 - This also induce a protocol change, thus bumping network protocol
2603 version from 0 to 1 (if we have external users, we have to get
2604 clean on that point too ;)
2605 - added two regression tests about this in datadesc_usage
2606 * Be more verbose when propagating local exceptions
2607 This helps debugging.
2608 * Display the status of simulated processes when receiving SIGINT in
2612 * Allow to control the simulation from a trace file.
2613 New functions MSG_action_register() and MSG_action_trace_run()
2614 The first one allows to associate a function execution to each
2615 kind of action while the second one parses a trace file and
2616 triggers the corresponding actions within the system.
2617 For now, only a toy example is provided in examples/msg/actions
2618 * Add an exemple of process migration in examples/msg/migration
2619 * Fix a bug in task exchange which broke MSG_task_get_sender()
2620 Add a teshsuite regression test for that.
2621 [Bug: if MSG_task_get_sender() is called after sender exit,
2623 * Fix a bug which prevented suspend/resume to work properly
2624 * Display the status of simulated processes when receiving SIGINT
2625 This fixes a regression of v3.3. due to the introduction of SIMIX
2626 * Bug fixing in failure management:
2627 - trace could not start by a failure at time 0
2628 - failure during communications were not working
2631 * Add SIMIX_process_set_name() to change the name of the current
2632 process in the log messages.
2633 * Store smx_hosts in a dict since we only retrieve them by name
2634 * Move the configuration infrastructure to surf
2637 * Move the configuration infrastructure to surf
2640 * Massive internal cleanups:
2641 - Store internal structures on processes instead of hosts (allows
2642 to have more than one process per host, in addition of being more
2644 - Cleanup the initialization/finalization process
2645 - Kill a whole bunch of unneeded synchronization:
2646 processes run in exclusive manner within the simulator
2647 - Move queues from global tables to process data fields
2649 - now accept -platform and -hostfile arguments
2650 - Pass the right rank value to processes according to the hostfile
2651 * Compile the examples by default, and use them as regression tests
2652 * Implement MPI_Wtime()
2653 * Change the reference speed to a command line option
2656 * TCP_gamma can now be specified as command line option using
2657 --cfg=TCP_gamma:10000000.0
2658 * Change the --surf-path cmd line option into --cfg=path:
2661 * Also include strbuff from xbt.h public header
2662 * xbt_ex_display(): do not free the exception after displaying
2663 This allows to do more with the given exception afterward.
2664 Users should call xbt_ex_free() themselves.
2668 Portability report of this version:
2669 * Main portability targets:
2670 - Linux(debian)/x86/context
2671 - Linux(debian)/x86/pthreads
2672 - Linux(debian)/amd64/context
2673 - Linux(debian)/amd64/pthreads
2674 These targets fail about 1/10 of times on gras/pmm, but we believe
2675 that this is because of the test, not because of simgrid.
2676 amok/saturate_sg fails even more rarely, and the test may not be
2679 - Mac OSX Leopard/x86/context
2680 The test suite still spits tons of errors because some obscure
2681 force prevents us from removing the temporary directories
2682 arguing that they still contain some metadata I've never heard of.
2683 Smpi fails because seq is not installed.
2684 Everything seems to work properly beside of that.
2687 - AIX version 5.3 (both contexts and pthread)
2688 Smpi still fails there because mktemp is not installed.
2689 XML inclusions seems rosty on AIX.
2691 * Windows: it's still lagging behind. If you want to help, please
2694 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2696 SimGrid (3.3) stable; urgency=high
2700 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2701 [Malek Cherier & Mt]
2703 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2706 Until now, GRAS were using MSG as an interface to SURF. It was
2707 quite difficult because both interface have several differences
2708 (MSG channels vs GRAS sockets were the most notable point).
2710 This also opens the gate to SMPI (which should occur soon) and speed
2711 up simulations by to 40% (even if it were not the main goal).
2713 **************************************
2714 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2715 **************************************
2716 The point is that events occuring at the exact same timestamp are
2717 not scheduled in the same order with the old and new version. This
2718 may be enough to completely change the execution of simulations in
2719 some cases. Sorry for the inconvenience.
2721 * Cleanup and upgrade the XML format to push further scalability
2722 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2724 * Improve the testing infrastructure with tesh. Now a very large part of
2725 the code is tested not only by being run but also by checking that the
2726 output match an expected output [Mt].
2728 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2729 is really less memory-demanding, which should allow you to use
2730 larger files in SimGrid [AL].
2732 * Inform valgrind about our contextes, so that it becomes usable
2733 with the default (and more effecient) version of SimGrid
2734 [contributed by Sékou Diakite, many thanks]
2737 * Introduce a listener thread in charge of receiving incoming
2738 messages from the network. It allows to overlap communication and
2739 computation but most notably, it removes some stupid deadlocks due
2740 to the fact that so far, a process could not send and receive at
2741 the same time. This made most non trivial communication schema
2743 * Convert the PIDs from long int to int to match the MSG ones (and
2744 linux ones too) [Mt]
2745 * New function: gras_agent_spawn() to launch a new process on
2746 current host. Only working in simulation for now. [Mt]
2747 * New function: gras_os_hostport() returning a constant form (ie,
2748 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2751 * Make the backtrace of exceptions more human readable [Mt]
2752 * New module: xbt/str [Mt]
2753 a ton of string utility functions (split, join, printf to a newly
2754 allocated buffer, trim, etc)
2755 * New module: xbt/hash [Mt]
2756 SHA1 hashing algorithm (more to come if needed)
2757 * New module: xbt/synchro [Mt]
2758 synchronization tools (mutex and conditions) working the same way
2759 in simulation and in real life (mainly useful for GRAS, but not
2761 * New module: xbt/queue [Mt]
2762 classical producer/consumer synchronization scheme
2763 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2764 (using the classical functions will get serialized) [Mt]
2765 * Make dictionary internal table dynamic. No need to specify its size
2766 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2768 * Make sure the log channels are organized as a tree under windows
2769 (because of ANSI C compatibility issue, any channel were child of
2770 root directly) [Mt].
2773 * Cleaned many thing in surf and fixed a few bugs [AL].
2774 * Add a nice command line configuration mechanism to compose models [AL].
2775 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2776 the previous one (KCCFLN05). It relies on something that looks like
2777 a max-min sharing mechanism but cannot be written as such. A new solver
2778 was thus designed [AL].
2779 * Add a new solver to lmm. Based on Lagrange optimization and
2780 gradient-based descent, it enables to efficiently maximise systems s.a
2782 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2784 This solver enables to propose two new network models for TCP Reno and
2785 TCP Vegas based on Low's work. These models still need to be fully
2786 tested though [Pedro Velho].
2789 * Bug fix in SD_simulate. Now the time bound given as argument is
2791 * Use the new parallel task model (ptask_L07) as default.
2792 * Use the SURF command line configuration mechanism.
2793 * 0-size tasks (for synchronization) should now work.
2795 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2797 SimGrid (3.2) stable; urgency=high
2801 We still experience issues on this platform, but we believe that at
2802 least MSG is usable.
2804 GRAS API BREAKAGE (for simplification purpose, sorry):
2805 * the gras_msgtype_by_name is not used anymore. Instead of
2806 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2807 you can write (and must)
2808 gras_msg_send(toserver, "request", &request);
2809 - If you still want to pass a gras_msgtype_t to the function (to cache
2810 the type and avoid the lookup time), use the gras_msg_send_() variant.
2811 - Impacted functions:
2812 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2813 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2814 * The callbacks are now expected to return 0 when everything went well
2815 (just like the main() function)
2817 GRAS new features and improvements:
2818 * New module mechanism where user code can use per process globals [Mt]
2819 This is similar to gras_userdata_*() functions, but for libraries. It
2820 factorize some code developped over and over in the examples and AMOK.
2821 It has still to be documented and used (only amok/peermanagement is
2823 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2824 loss when they were small enough to fit into the buffer and sent quickly
2825 enough so that they can all get received in one shoot.
2826 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2827 if not found. Use the *_or_null() variant for the old semantic.
2828 * In gras_msg_handle, do not discard messages without callback.
2829 They are probably messages to be explicitly awaited later (ie, proofs of
2830 mis-synchronization in userland since they are sent before being awaited)
2831 No big deal usually.
2832 * gras_socket_meas_send/recv: semantic changed!
2833 The numerical arguments used to be (1) the total amount of data to send
2834 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2835 messages. This was need for the fool willing to send more than MAXINT
2836 bytes on quite fat pipes.
2839 * Do really rename the hostmanagement module to peermanagement. [Mt]
2840 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2841 but this is rather new and this was documented in the module
2842 documentation (poor excuses, I admit)
2843 * Bandwidth measurement semantic changed! This follows the changes to
2844 gras_socket_meas_send/recv explained above.
2847 * A sequential mode has been added to the workstations. When a workstation
2848 is in sequential mode, it can execute only one task, and the other tasks
2849 are waiting in a FIFO. [Christophe Thiery]
2852 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2853 model for SIMDAG. [Christophe Thiery]
2854 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2855 equal to 0, which caused some bad behaviors in
2856 saturated_constraint_set_update. I now use a threshold mechanism like in
2860 * When running manually src/testall, you select specific units [Mt]
2861 testall is the result of our cunit mechanism, and should replace all
2862 the scripty thingy around since bash don't run easily on billware.
2864 * A mallocator system has been added. [Christophe Thiery]
2865 Mallocators allow you to recycle your unused objects instead of freeing them
2866 and allocating new ones.
2868 Documentation update:
2869 * FAQ reworking + New FAQs:
2870 - "Valgrind spits tons of errors!" [Mt]
2871 - "How to repport bugs" [Mt]
2872 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2873 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2874 - Communication time measurement within MSG [AL]
2875 - I experience weird communication times when I change the latency [AL]
2876 * GRAS tutorial [Mt]
2878 - an introduction to the framework and to the used communication model
2879 - an initiatic tour introducing the most proheminent features:
2881 . Lesson 0: Installing GRAS
2882 . Lesson 1: Setting up your own project
2883 o Part 2: Message passing
2884 . Lesson 2: Exchanging simple messages
2885 . Lesson 3: Passing arguments to the processes (in SG)
2886 . Lesson 4: Attaching callbacks to messages
2887 . Lesson 5: Using globals in processes
2888 . Lesson 6: Logging information properly
2889 . Lesson 7: Using internal timers
2890 . Lesson 8: Handling errors through exceptions
2891 . Lesson 9: Exchanging simple data
2892 . Lesson 10: Remote Procedure Calling (RPC)
2893 . Lesson 11: Explicitely waiting for messages
2894 . Recapping of message passing features in GRAS
2895 - A HOWTO section containing:
2896 o HOWTO design a GRAS application
2897 More are due, of course. They will come latter. In the meanwhile, you can
2898 check the examples which are still here.
2900 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2902 SimGrid (3.1) stable; urgency=high
2906 There was a stack corruption somewhere, visible only when optimizing
2907 with these versions. [Vince]
2910 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2911 module that enabled to play with Directed Acyclic Graphs. It is built
2912 directly on top of SURF and provides an API rather close to the old
2913 SG. Some old codes using SG are currently under rewrite to check that
2914 all needful functions are provided. [Christophe Thiery]
2917 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2918 extension of the classical CLM03 model that gracefully handles
2919 failures. This is now the default model for MSG and GRAS. It doesn't
2920 handle parallel tasks yet though. [AL]
2921 * Bug fix: Weights were not correctly set in the network part.
2922 WARNING: This may have resulted in incorrect results with simulations
2923 where there are more than one flow on a given link. [AL]
2926 * After a (long ?) discussion on simgrid-devel, we have decided that the
2927 convention we had on units was stupid. That is why it has been decided
2928 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2929 WARNING: This means that all previous platform files will not work as
2930 such with this version! A warning is issued to ask users to update
2932 A conversion script can be found in the contrib module of the CVS, under
2933 the name contrib/platform_generation/surfxml_update.pl [MQ]
2936 * Bug fix: Processes were started in reverse order, wrt deployment file.
2937 WARNING: if your code relies on this bug, please fix it. [AL]
2938 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2939 being executed on two different locations. [AL]
2940 * Bug fix: Failures are now better supported thanks to Derrick's tests
2941 (there was many failure situations I hadn't thought of and that weren't
2942 correctly handled). [AL]
2943 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2947 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2948 Exception occurring on server-side are propagated back to client (!).
2950 API CHANGE: the callback changed their prototype. Change:
2951 int my_handler(gras_socket_t expeditor, void *payload_data) {
2953 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2954 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2956 * New! function: gras_msg_handleall to deal with all messages arriving
2957 within a given period.
2958 * New! function: gras_socket_server_range to get a server socket in a
2959 range of port numbers (ease to avoid port number conflicts) [MQ]
2960 * New! gras processes display their backtrace when they get a SIGUSR1
2961 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2962 Sweet to debug RL processes [MQ]
2966 - Do not force experiment sizes to be expressed in kb, or it becomes
2967 impossible to measure the latency this way (needs one byte-long tests)
2968 WARNING: this changes the amok_bw_* function semantic. [MQ]
2969 - Implements the link saturation stuff. [MQ]
2970 * Peer management module:
2971 New! module factorizing code that we wrote over and over [MQ].
2974 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2975 - Test units are placed directly into the library code, they get extracted
2976 automatically and placed into the src/testall binary.
2977 - Convert most of the XBT tests to this system.
2978 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2979 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2980 live in the same C code. This required to change a little bit the API
2981 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2982 * New module: sparse graph structure with basic algorithms (this is work
2983 in progress and the API is not considered to be frozen yet). [AL]
2984 * Display more information on backtraces: source line & function names are
2985 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2986 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2987 * Reworked a little bit some #include statements to load only required
2988 headers. Some user code that relied on SimGrid to include stdlib or
2989 stdio may need to include it by themselves. [AL]
2990 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2992 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2993 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2994 but I guess that almost nobody used those parts. [MQ]
2996 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2998 SimGrid (3.0.1) stable; urgency=low
3001 * Unfortunately, I had missed 5 misnamed functions:
3002 xbt_fifo_item_t xbt_fifo_newitem(void);
3003 void xbt_fifo_freeitem(xbt_fifo_item_t);
3004 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3005 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3006 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3007 They're now deprecated. Please use their new versions:
3008 xbt_fifo_item_t xbt_fifo_new_item(void);
3009 void xbt_fifo_free_item(xbt_fifo_item_t);
3010 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3011 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3012 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3014 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3015 * Documentation: xbt_log module unmercifully reworked [MQ]
3016 * Bugfix: there was a problem with the ending of contexts with
3017 the pthread backend. It caused some weird deadlock or behavior
3018 depending on the pthread implementation. [AL]
3019 * Bugfix: get the exceptions raised in the simulator repport where
3020 and why they come from when they are not catched in time [AL, MQ]
3023 * Bugfix: Do repport the error when two non-connected hosts try to
3024 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3027 * Add additionnal checkings on communications. Assert that two
3028 communicating hosts are connected by a set of links... [AL]
3031 * Add additionnal checkings on channel values in communication [AL]
3032 * New: MSG_task_get_source to see on which host a task was generated [HC]
3033 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3034 the number of tasks waiting to be received on channel and sent
3036 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3037 waits for the first task coming from a given host.. [AL]
3039 GRAS new functionnalities: [MQ]
3040 * Enhance the parsing macro to allow the size of multidimentional objects
3041 to be given thru annotations.
3042 * New example (and documentation): Matrix Multiplication a la RPC
3043 (as when I was young!) and fix a bunch of bugs found on the way.
3045 GRAS performance improvements: [MQ]
3047 * Reduce the amount of cbps creation/destruction by making it static to
3048 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3051 * Change libdata to a set so that we can search for stuff by ID (and thus
3052 reduce the insane amount of dict lookups)
3055 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3056 RL communication on top of SG since it's so uneffective.
3057 It may also be used for inter-thread communication in RL, one day.
3058 * Use gras_datadesc_copy() to exchange messages on top of SG
3060 - improve message exchange performance on top of SG
3061 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3062 * Don't exchange on the network the size of the used part of buffer,
3063 instead, specify the possible buffer size to read().
3065 - reduces the amount of read/write calls (one pair per exchange)
3066 - reduces the amount of exchanged data (the size)
3067 - allows to retrieve all arrived data on receiver side, if we don't need
3068 it right now (subsequent read will peek the buffer)
3069 - allows the receiver to proceed with the begining of the stream before
3070 everything is arrived
3071 - make it possible to build an iov transport (using readv/writev)
3073 - take care of the data with non-stable storage (like stacked data),
3075 * If possible, TCP send uses vector I/O (when writev() is here)
3076 - Don't use it for receive since we send data sizes and data on the
3077 same stream, so we wouldn't be able to chain large amount of chunks
3078 before having to flush the stuff to read the size.
3079 * Rework the transport plugin mechanism to simplify it and reduce the
3080 amount of pointer dereferencement when searching for the right function
3083 * I guess that now, we do almost as few system calls as possible while
3084 doing as few data copy as possible.
3086 To improve it further, we could try to send all the sizes first and then
3087 all the data (to use iov on receiving size), but it's only a partial
3088 solution: when you have 2 dimensional data, the sizes of the second
3089 dimension is data of the first dimension, so you need 3 streams.
3091 I'm not sure the potential performance gains justify the coding burden.
3093 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3095 SimGrid (3.00) stable; urgency=high
3098 * New! Give the possibility to hijack the surf parser and thus bypass
3099 MSG_create_environment and MSG_launch_application. Have a look at
3100 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3102 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3104 SimGrid (2.96) unstable; urgency=low
3109 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3110 This deprecates the xbt_error_t mechanisms.
3111 It modifies (simplifies) all XBT and GRAS API.
3112 MSG API keeps unchanged (exceptions raised by XBT are catched from
3113 within MSG and masked with existing error handling facilities)
3116 * New! Add a FATPIPE model. [AL]
3117 * New! Add a parallel task model. [AL]
3118 * New! Add automatically a loopback interface (in the default
3119 network model) if none was precised.
3122 * Bugfix: MSG_process_resume now works with the current running process.
3124 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3125 * Modification of MSG_task_get_compute_duration. Once a task has been
3126 processed, the value returned by this function is now equal to 0. [AL]
3127 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3128 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3129 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3130 of a m_task. For now, MSG_task_cancel only works with computation
3132 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3133 of the processor (in Mflop/s) regardless of the current load on the
3135 * API Change: use proper naming convention for MSG_getClock and
3136 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3138 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3139 This function changes the priority of a computation task. This priority
3140 doesn't affect the transfer rate. A priority of 2 will make a task
3141 receive two times more cpu power than the other ones. This function
3142 has been added to suit the needs of Nguyen The Loc and hasn't been that
3143 much tested yet. So if it fails, please report it and send me your code.
3145 * API Change: removed all functions and types that were marked "deprecated"
3146 since many months. Renamed MSG_global_init_args to MSG_global_init.
3148 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3150 SimGrid (2.95) unstable; urgency=low
3153 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3154 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3155 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3158 * Bugfix: Allow absolute paths for platform description files [MQ]
3159 * Bugfix: do free the variables after use. Leads to drastic performance
3161 * Implement max_duration (ie, timeouts) on resources [AL]
3164 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3166 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3167 the needed support (timeouts) [AL]
3170 * Implement measurement sockets. You can now get the bandwidth between two
3171 hosts thanks to AMOK (see below). [MQ]
3172 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3173 dynar over the network (yeah) [MQ]
3174 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3175 * simple/static token-ring example. [Alexandre Colucci and MQ]
3176 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3177 in place before [MQ]
3180 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3181 type parsing. "short" and "long" modifiers were also ignored; other
3182 modifier (such as reference level) are still ignored. [MQ]
3183 * Embeed the buffer size within the buffer itself on SG. [MQ]
3184 That way, send() are atomic and cannot get intermixed anymore (at least
3185 the ones which are less than 100k; bigger messages still have the issue)
3186 * Array size pushed by the field, not by the field type (or each
3187 and every long int will push stuff to the cbps) [MQ]
3188 * use select() to sleep since it allows to portably sleep less than one
3191 GRAS (minor cleanups)
3192 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3195 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3196 * type category 'ignored' killed as it was never used and were difficult
3198 * whether a type can cycle or not is now a flag, leaving room for more
3199 flags (as "ignored", if we feel the need one day ;)
3200 * Rename raw sockets to measurement sockets since "raw" has another
3201 meaning in networking community.
3204 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3205 and offers features not belonging to GRAS but that most applications
3206 need. One day, it may be a set of plugins loadable at runtime.
3207 * New module: bandwidth
3208 bandwidth measurement between arbitrary nodes running this module. [MQ]
3210 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3212 SimGrid (2.94) unstable; urgency=low
3214 The first beta release of SimGrid 3 !
3218 * Update the main page and the FAQ. Adding references to gforge.
3221 * Add a gras_os_getpid function.
3224 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3225 * Extend the logs so that they also print PID, hostname, date, ... if
3227 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3228 and kill the old version which were in testsuite/
3229 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3233 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3234 for users. Instead, each programming environment (GRAS, MSG,...) use it
3235 when needed to provide such a feature to users.
3236 Don't shortcut the mechanism or you will also shortcut the virtualization
3237 you need on the simulator.
3241 * Cleanups in configury with regard to compile optimization/warning flags.
3242 Also add -fno-loop-optimize to any powerpc since it's the optimization
3243 killing gcc (< 3.4.0).
3244 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3245 needed by MSG examples complications
3246 * Borrow configury beautifications from PHP
3249 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3250 warning (thanks loris for stumbling into this one).
3251 * Bugfix: stop loading private headers (gras_config.h) from the public
3255 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3256 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3257 RL would imply a DNS resolver, which is *hard* to do in a portable way
3258 (and therefore delayed).
3259 * Implement a real timer mechanism and use it in timing macros. This allows
3260 to avoid rounding errors and get a 0.000005 sec precision in timing
3261 macros. While I was at it, various cleanups:
3262 - allow to declare more than one timed section per file (fix a stupid bug)
3263 - move some private declaration to the right place
3264 - merge conditional execution and timing macros into emulation module
3265 - document the module
3266 - make sure the module cleanups its mess on gras_exit
3267 * Documentation improvements:
3268 - (new) how to compile applications using GRAS
3269 - (new) emulation support (timing macros)
3271 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3273 SimGrid (2.93) unstable; urgency=low
3275 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3278 - Use Paje properly where used. Still to be sanitized properly.
3279 - Portability fix: Add an implementation of the contexts using pthread
3283 - Add xbt_procname(): returns the name of the current process.
3284 Use it to show the current process's name in all logging.
3286 - fix detection of older flex version and the reaction, since we do
3287 depend on modern ones (we use lex_destroy)
3288 - Better separation of SG and RL in the libs: remove all simulation code
3289 from libgras. As a result, this lib is now only 200k when stripped.
3290 Some of the xbt modules may also be duplicated (two sets and such) and
3291 should be cleaned/killed before SG3.
3292 - Insist on using xlC on AIX because of weird problems involving gcc there.
3293 - Cleanup the make remote stuff. This is now done by scripts
3294 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3295 mainly for our private use, but we're working on changing them to user
3298 - Bugfix: flush the socket on close only if there is some *output*.
3299 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3301 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3304 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3306 SimGrid (2.92) unstable; urgency=low
3308 Alpha 3 on the path to SimGrid 3
3312 - New! First try of benchmarking macros.
3313 - New! First try so that gras_stub_generator generate deployment and
3314 remote compilation helpers.
3316 - Bugfix: Initialization fix in msg_test.
3320 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3322 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3324 - API Change: gras_os_sleep to take the amount of seconds as a double.
3325 Accepting an int was error prone since it was the only location where
3326 seconds were coded as such. It leaded to damn rounding errors.
3327 - Bugfix: Hard to belive that timers ever worked before this.
3329 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3331 SimGrid (2.91) unstable; urgency=low
3333 Alpha 2 on the path to SimGrid 3
3337 - Bug fix in the lmm_solver.
3339 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3340 through the function MSG_paje_output.
3341 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3342 - It is possible to bound the rate of a communication in MSG with
3343 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3344 to write it in the changelog).
3345 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3350 - Various cleanups to the autotools stuff
3351 - Begin to move Gras examples to examples/gras/
3352 - Let make distcheck work again (yeah!)
3354 - documentation overhauled using doxygen.
3355 gtk-doc-tools is dead in SimGrid now.
3356 - Automatically extract all existing logging categories, and add the list
3357 to the documentation (long standing one, to say the less)
3359 - Cleanup the known architecture table. Reorder the entries to group what
3360 should be, and use a more consistent naming scheme.
3361 (some of the test dataset are still to be regenerated)
3362 - New! Allow library to register globals on each process just as userdata
3364 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3365 lookup time (for now).
3366 Use it in msg and trp.
3367 This cleans a lot the internals and helps enforcing privacy of the
3368 headers between the gras components.
3369 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3370 - Bugfix: gras_os_time was delirious in RL.
3371 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3373 - Reenable GRAS now that it works.
3375 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3377 SimGrid (2.90) unstable; urgency=low
3379 Alpha 1 on the path to SimGrid 3
3381 * It is a long time since the last release of SimGrid. I'm sorry about
3382 that but as I had told you, I was rewriting a lot of things. I apologize
3383 to those who had been reporting bugs to me and that I had not answered.
3384 If your bug is still in the new version, please tell me. Here is a
3385 summary of the main changes.
3387 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3388 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3389 portability, speed, and a lot more but you'll figure it out later.
3390 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3391 I wanted to release the new MSG as soon as possible and I have
3392 broken GRAS, which is the reason why, for now, only MSG is fully
3393 functional. A laconic description of these projects is available
3394 in the documentation.
3396 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3397 kernel optimized for our needs (called SURF but only the developers
3398 should use it). Hence, MSG is now roughly 30 times faster and I think
3399 that by rewriting a little bit MSG, I could event speed it up a little
3400 bit more. Beside the gain in speed, it is also much easier to encode a
3401 new platform model with SURF than it was with SG. More to come...
3403 * REVOLUTION 3: I have tried to change a little as possible the API of
3404 MSG but a few things really had to disappear. The main differences
3405 with the previous version are:
3406 1) no more m_links_t and the corresponding functions. Platforms are
3407 directly read from a XML description and cannot be hard-coded
3408 anymore. The same format is used for application deployment
3409 description. The new format is described in the documentation.
3410 Have a look in tools/platform_generation. There is a tiny script
3411 that converts from the old platform format to the new one. Concerning
3412 the application deployment format, parsing the old one is tricky.
3413 I think most of you should however be able to convert your files. If
3414 it is really an issue, I can write a C code that does the conversion.
3416 2) the toolbox tbx does not exist anymore. We now have a library
3417 with much more data-structures but without the hash-tables (we have
3418 dictionaries that are much faster).
3420 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3422 *****************************************************************************
3423 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3424 * don't feel like converting the oldies. (Mt) *
3425 *****************************************************************************
3428 Version 2.90: "the long awaited one"
3429 - Finished rewriting and debugging MSG. Rewrote the documentation.
3430 - disable GRAS for now since it needs to be ported to the newest SG
3433 - Finish the port to windows (using mingw32 for cross-compile)
3436 - Main loop and datastructures of SURF. A cpu resource object is
3437 functional. Surf can thus be used to create cpu's with variable
3438 performance on which you can execute some actions.
3440 2004-11-15 Martin Quinson
3441 - Port to ARM. Simply added the alignment and size descriptions. Should
3442 work, but the ARM machines are so slow that I didn't had the opportunity
3443 to 'make check' over there yet.
3445 2004-11-15 Arnaud Legrand
3446 - Trace manager now written. It uses a heap structure and is therefore
3447 expected to be efficient. It may however be speeded up (particularly
3448 when many events occur at the same date) by using red and black
3449 trees. One day maybe...
3450 - Max-min linear system solver written. It uses a sparse matrix
3451 structure taking advantage of its expected use. Most operations are
3452 O(1) and free/calloc are called as few as possible. The computation of
3453 the minimum could however be improved by using a red and black tree
3456 2004-11-03 Arnaud Legrand
3457 - Rename every gras_* function that was in xbt/ to its xbt_
3459 - Add a heap and a doubly-linked list to xbt
3460 - Added a dichotomy to the dictionaries. make check works as well before
3461 so I assume that the patch is correct. I do not know however if things
3462 run effectively faster than before now. :)
3464 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3465 SimGrid, and the version number is bumped to 2.x
3467 2004-10-29 Martin Quinson
3468 - Introduction of the remote errors.
3469 They are the result of a RMI/RPC on the remote machine.
3470 ErrCodes being scalar values, you can't get the host on which those
3471 errors did happen. Extending the error mechanism as in Gnome is possible.
3472 No idea yet whether it is a good idea.
3474 2004-10-28 Martin Quinson
3475 - Interface revolution: the Starred Structure Eradication.
3476 I used to do typedef struct {} toto_t; and then handle *toto_t.
3477 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3478 - ???_t is a valid type (builded with typedef)
3479 - s_toto_t is a structure (access to fields with .)
3480 - s_toto is a structure needing 'struct' keyword to be used
3481 - e_toto_t is an enum
3482 - toto_t is an 'object' (struct*)
3484 typedef struct s_toto {} s_toto_t, *toto_t;
3485 typedef enum {} e_toto_t;
3486 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3487 s_toto_t) is private.
3489 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3490 so that it can set the variable to NULL. It was so for dicts and sets,
3491 it changed for dynars.
3493 - Fix a bunch of memleaks in dict_remove
3494 - Fix a bug in sg/server_socket opening: it failed all the time.
3496 2004-10-07 Martin Quinson
3497 - Speed up dynar lookup operation a bit.
3499 gras_dynar_get is dead.
3501 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3502 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3503 (which gives you the address of the stored data).
3505 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3506 copy of the data using an affectation to do the job and not a memcpy.
3508 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3509 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3511 It does not really speedup the dynar test because they are
3512 setting elements all the time (and look them seldom). But the dict does
3513 far more lookup than setting.
3515 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3517 2004-10-05 Martin Quinson
3518 - Allow to (en/dis)able the cycle detection at run time.
3520 Whether we should check for cycle or not is now a property of each
3521 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3522 datadesc_cycle_unset allow to remove this property when previously set.
3524 Note that the cycle detection is off by default since it impacts the
3525 performance. Watch the data you feed GRAS with ;)
3527 This property is hereditary. Any element embedded in a structure having it
3528 set have it set for the time of this data exchange.
3530 You should set it both on sender and receiver side. If you don't set it on
3531 sender side, it will enter an endless loop. If you forget on receiver
3532 side, the cycles won't be recreated after communication.
3534 - Header reorganization.
3535 Kill gras_private.h, each submodule must load the headers it needs.
3537 2004-10-04 Martin Quinson
3538 - Interface revolution: do not try to survive to malloc failure.
3540 Now, gras_malloc and friends call gras_abort() on failure.
3541 As a conclusion, malloc_error is not a valid error anymore, and all
3542 functions for which it was the only gras_error_t return value are
3543 changed. They now return void, or there result directly.
3544 This simplify the API a lot.
3546 2004-09-29 Martin Quinson
3547 - Re-enable raw sockets.
3548 Created by gras_socket_{client,server}_ext;
3549 Used with gras_raw_{send,recv}
3552 It should allow to kill the last bits of gras first version soon.
3554 This is not completely satisfactory yet (duplicate code with
3555 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3558 - Simplify transport plugin (internal) interface by not passing any
3559 argument to _server and _client, but embedding them in the socket
3562 2004-09-28 Martin Quinson
3563 - Finish the port to AIX.
3564 autoconf was my problem (segfault within the malloc replacement
3565 function. No idea why)
3567 2004-09-16 Martin Quinson
3568 - Fix some size_t madness on 64bit architectures.
3570 2004-09-08 Martin Quinson
3571 - Reduce the number of system headers loaded, overload some more system
3572 calls (such as malloc to cast the result of the system one, and work
3574 - Fix and reintroduce the config support
3576 2004-09-07 Martin Quinson
3577 - Source code reorganization to allow Arnaud to surf all over there.
3578 - Allow to document the logging categories.
3579 - Remove all uppercase from logging categories and useless cleanup in names.
3581 2004-08-18 Martin Quinson
3582 Version 0.6.2 (protocol not changed; API changed)
3583 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3584 gras_error_t), and NULL when not found. Functions expecting a msgtype
3585 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3586 hopefully usefull message.
3587 - Portability to prehistoric sparcs again
3589 2004-08-17 Martin Quinson
3590 Version 0.6.1 (protocol not changed; ABI not changed)
3591 - prealloc some buffers to speed things up
3593 2004-08-11 Martin Quinson
3594 Version 0.6 (protocol not changed; ABI expended)
3595 - The parsing macro can deal with the references, provided that you add
3596 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3598 2004-08-09 Martin Quinson
3599 Version 0.5 (protocol not changed; ABI changed)
3600 - Allow to off turn the cycle detection code in data exchange at
3601 compilation time. It should be at run time, but I'm short of time (and
3602 the config stuff is still broken). That way, we keep dict out of the
3603 critical path, which is good because the performance is poor:
3604 - search not dichotomial yet
3605 - dynar give no way to access their content and memcpy everytime
3606 - In composed data description (struct, ref and so on), stop foolness of
3607 keeping the subtype's ID, but store the type itself. This keeps sets out
3608 of the critical path, which is good since they rely on dynar and
3609 dictionnaries. The only loose of that is that we cannot detect the
3610 redeclaration of a structure/union with another content (but I'm not sure
3611 the code detected well this error before anyway). We still can detect
3612 the redefinition discrepancy for the other types.
3613 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3614 it breaks the code because of type-punning used all over the place).
3615 This breaks on all non-gcc architectures (for now).
3617 All those changes (plus the buffer of last time) allow me to gain 2 order
3618 of magnitude on cruel tests consisting of 800000 array of integers on two
3619 level of a hierarchical structure (200 secondes -> 4 secondes)
3622 - the selector of reference must now return the type it points to, not
3623 the ID of this type.
3625 2004-08-06 Martin Quinson
3626 Version 0.4 (protocol changed; ABI not changed)
3627 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3628 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3629 - Implement a buffer transport, and use it by default (it relies on tcp in
3630 real life and on sg in simulation).
3631 That's a bit hackish since I had a new field to the structure to store
3632 its data without interfering with the subtype ones. Inheritance
3633 is tricky in C. And that's a kind of reverse inheritance with one class
3634 derivating two classes. Or maybe a game with java interfaces. Anyway,
3635 that's damn hard in C (at least).
3636 Moreover, I got tired while trying to ensure plugin separation and
3637 genericity in SG mode. MSG wants me to do weird things, so let's go for
3638 cruel hacks (temporarily of course ;).
3639 See comment in transport_private.h:71
3640 - do not include all the _interface headers in private but in the files
3641 which really need them (to cut the compilation time when they are
3644 2004-07-26 Martin Quinson
3645 Version 0.3 (protocol not changed; ABI changed)
3646 - Major overhault of the datadesc interface to simplify it:
3647 - shorted the function names:
3648 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3649 - add a trivial way to push/pop integers into the cbps without malloc.
3650 This allows to make really generic sub_type description, which simply
3651 pop their size of the stack.
3652 - add a function gras_datadesc_ref_pop_arr() which does what users want
3653 most of the time: Declare a dynamic array (which pops its size of the
3654 stack) and declare a reference to it. Poor name, but anyway.
3655 - kill the post-send callback, add a post-receive one
3657 2004-07-23 Martin Quinson
3658 Version 0.2 (protocol changed; ABI changed)
3659 - add some testing for cpbs in the test cases, and fix some more bugs.
3660 This invalidate again the little64 data file, since I cannot regenerate
3662 - remove an awfull optimization in the logging stuff, allowing me to:
3663 - understand it again
3664 - learn gcc how to check that the argument match the provided format
3665 - fix all errors revealed by gcc after that
3666 - internal keys of dict are not \0 terminated. Deal with it properly in
3667 loggings instead of segfaulting when the user want to see the logs :-/
3669 2004-07-22 Martin Quinson
3670 - Fix some stupid bug preventing cbps (callback postit) from working
3672 2004-07-21 Martin Quinson
3673 - Some documentation cleanups
3674 - remove the useless last argument of msgtype_declare
3675 - rename the Virtu functions to fit into the 'os' namespace
3676 - move headers src/include -> src/include/gras/ and stop fooling with
3677 gras -> . symbolic link
3678 - make distcheck is now successful
3680 2004-07-19 Martin Quinson
3682 - Build shared library also
3683 - Install html doc to the right location
3684 - stop removing maintainer files in make clean
3685 - build tests only on make check
3687 2004-07-13 Martin Quinson
3689 - No major issue in previous version => change versioning schema
3690 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3691 corresponding dataset.
3693 2004-07-11 Martin Quinson
3695 - Get it working with any kind of structure (we can compute the padding
3696 bytes remotely for all the architectures I have access to)
3697 - Implement the structure parsing macro (still not quite robust/complete)
3698 - Improvement to the remote testing toysuite
3700 2004-07-10 Martin Quinson
3701 [autoconf mechanism]
3702 - get ride of a bunch of deprecated macros
3703 - actually run the test for two-compliment, not only compile it :-/
3704 - test whether the structures get packed (and bail out if yes. Damn.
3705 Alignment is a serious matter)
3706 - test whether the structures get compacted (but respecting the alignment
3707 constraints of each types)
3708 - test whether the array fields of structures can straddle alignment boundaries
3710 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3711 (took me 2 hours to find that bug, looking at the wrong place)
3712 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3713 before sending/receiving and are used to compute the offsets of fields
3714 - Given that padding size depend even on compiler options, keep track of
3715 alignment and aligned_size only for the current architecture. Not a big
3716 deal since we send structure fields one after the other (seems
3718 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3719 program, let it work on linux/gcc/little32. portability todo.
3721 2004-07-08 Martin Quinson
3722 - import and improve remote compilation support from FAST
3723 - make sure make check works on half a dozen of machines out there
3725 2004-07-07 Martin Quinson
3726 Let's say it's version 0.0.3 ;)
3727 - Implement conversions (yuhu!)
3728 - Let it work on solaris (beside conversion, of course)
3729 - Stupid me, using rand() to generate the conversion datatests in not wise.
3731 2004-07-06 Martin Quinson
3732 - Let make dist work, since I'm gonna need it to compile on remote hosts
3733 - Let Tests/datadesc_usage write the architecture on which the file was
3734 generated as first byte.
3735 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3736 architecture descriptions.
3737 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3738 on those architectures.
3739 - Optimization: send/recv array of scalar in one shoot
3741 2004-07-05 Martin Quinson
3742 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3744 - Plug a whole bunch of memleaks
3745 - each process now have to call gras_{init,exit}. One day, their log
3746 settings will be separated
3747 - Continue the code factorisation between SG, RL and common in Transport.
3749 2004-07-04 Martin Quinson
3751 - Redistribution between SG and RL.
3752 We wanna have to accept in SG, so move accepted related parts of RL in
3753 the common part. (more precisely, the dynar of all known sockets is no
3754 more a static in transport.c, but part of the process_data)
3756 [gras_stub_generator]
3757 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3759 2004-07-03 Martin Quinson
3760 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3761 them all up in one shot)
3763 - Fix the ugly case of reference to dynamic array.
3764 - New (semi-public) function gras_datadesc_size to allow the messaging
3765 layer to malloc the needed space for the buffer.
3767 - gras_socket_close now expect the socket to close (and not its address to
3768 put NULL in it after it). This is because the socket passed to handlers
3769 is one of their argument (=> not writable).
3771 - propagate the interface cleanup from last week in datadesc, ie remove a
3772 superfluous level of indirection. User pass adress of variable
3773 containing data (both when sending and receiving), and not of a variable
3774 being a pointer to the data. Let's say that I like it better ;)
3775 The price for that is constructs like "int msg=*(int*)payload" in
3776 handlers, but it's a fine price, IMHO.
3778 - Let it work in RL (yuhu)
3780 2004-06-21 Martin Quinson
3782 - porting SG plugin and SG select to new standards (works almost).
3783 - plug memleaks and fix bugs around.
3786 - cleanup the prototype of data recv and force users to specify when they
3787 want to handle references to objects. Test case working even for cycles.
3788 - plug memleaks. Valgrind is perfectly ok with this.
3790 2004-06-12 Martin Quinson
3792 - cleanup the separation between plugin and main code in plugin creation
3794 2004-06-11 Martin Quinson
3796 - Reput hook for raw sockets, needed for BW experiments
3797 - kill a few lines of dead code
3798 [Data description] Interface cleanup
3799 - gras_datadesc_by_name returns the searched type or NULL.
3800 That way, no variable is needed to use a type desc once, which makes
3802 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3803 two parameters were strings (field name, type name), leading to
3805 [Dicos] Interface cleanup
3806 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3807 This is consistant with the dynar API.
3809 2004-04-21 Martin Quinson
3811 - Porting to new standards.
3813 - interface cleanup.
3814 There is no bag anymore, no need to take extra provision to mask the
3815 pointers behind "ID".
3816 Better splitup of functions between files create/exchange/convert.
3817 This is still a bit artificial since convert and receive are so
3818 interleaved, but anyway.
3820 - add a queued message list to procdata (the ones not matching criteria
3822 - factorize some more code between SG and RL wrt procdata
3824 - use gras_exit in example to track memleaks
3825 - get rid of gs_example now that GS is properly integrated into gras
3826 - update run_test to integrate the latest tests (datadesc)
3828 - rename WARNINGn macros to WARNn since it prooved error-prone
3830 2004-04-19 Martin Quinson
3832 - register init/exit functions within gras module mechanism
3833 - send/receive function.
3834 Convertion is not implemented, but short-cutted if not needed.
3835 struct/array elements are sent one by one (instead of block-wise), but
3836 nobody really cares (yet). Get a prototype before optimizing.
3837 - tests (using a file socket) for DD send/receive on:
3838 - base types: int, float
3839 - array: fixed size, string (ie ref to dynamic string)
3840 - structure: homogeneous, heterogeneous
3841 - chained list, graph with cycle
3842 Believe it or not, valgrind is not too unhappy with the results. The
3843 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3847 - Bugfix when using a filename explicitely (instead of '-')
3849 2004-04-09 Martin Quinson
3851 - factorize more code between RL and SG in socket creation
3852 - Complete the implementation and tests of:
3854 o file (only in RL, and mainly for debugging)
3856 I lost 3 days to design a portable address resolver, and then decided
3857 that the prototype mainly have to run on my box.
3858 Addressing portability too early may be like optimizing too early :-/
3860 - use gras_init in the Tests instead of the crappy parse_log_opt
3861 (the latter function is removed)
3862 [Conditional execution]
3863 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3864 [Code reorganisation]
3865 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3866 Build examples against the RL library, since there is no way to disable
3867 its creation for now.
3869 For information, the beginning of coding on GRAS was back in june
3870 2003. I guess that every line has been rewritten at least twice since