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.
48 - Add -gdb and -lldb shortcuts to help debug MPI codes with smpirun
49 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
50 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
51 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv.
52 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
54 - Basic implementation of generalized requests (SMPI doesn't
55 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
58 - Drop sg_cmdline. Please use xbt_cmdline instead.
59 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
60 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
61 OsSemaphore is implemented in a portable way with C++11 threads.
62 This should allow much more threads to be created at the same time,
63 allowing Mac and Java users to simulate many more actors.
64 - Implement the 'thread' factory with std::thread instead of xbt ones.
65 It is not possible to set the stack size with threads anymore, but
66 -fsplit-stack is the way to go nowadays when using threads.
67 - Drop the xbt_os_thread_t module (now unused)
68 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
69 - Drop xbt_str_join_array().
70 - Drop cunit, use Catch2 instead.
73 - Many cleanups in the kernel::activity namespace. This was long
74 overdue, and shall open the path to many future endeavors.
77 - #132: Java: a process can not shut down its own host
78 - #220: S4U: detached send causes exception if sender terminates soon after sending
79 - #261: Document the parameters of parallel execution's constructor
80 - #300: [s4u] BarrierPtr is missing
81 - #314: SMPI args internal cleanup
82 - #316: Fix a bug related to the CPU utilization of multi-core VM
83 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
84 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
85 - #325: Turning off a host has different behavior on sleeping actors and computing actors
87 ----------------------------------------------------------------------------
89 SimGrid (3.21) October 3. 2018
91 The Restarting Documentation (TRD) Release.
94 - Convert the user manual to Sphinx for improved usability.
95 Unstable version now hosted on https://simgrid.frama.io/simgrid
96 Many glitches remain; stay tuned.
97 - Restrict the Doxygen reference API to the installed header files.
98 (documenting all internals this way is overkill + doxygen scales badly)
99 - New tutorial on S4U (highly inspired from the old MSG one, but with
100 a git repository to fork as a starting point, and a docker image)
101 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
102 - The Developper manual is still to be converted (not compiled until then)
103 - Some parts are still missing in this conversion (such as the
104 platform and deployment sections) while others were blindly converted
105 and would need more love (such as the configuration flags).
106 Things will certainly further improve in the future releases.
109 - Fully reimplement auto_restart mechanism that was utterly broken
110 (fix #22, #131 and #281 that were all very related)
111 - Implement semaphores (fix #303 and #291)
112 - When creating an actor from a function and its parameters,
113 move-only parameters are not allowed anymore, as it would prevent
114 the actor to be restartable if its parameters are consumed on run.
115 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
116 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
117 IO operations that can be started, waited for, or canceled as a regular
119 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
123 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
124 - New host variable: 'core_count'
127 - Remove xbt_os_thread_specific features
128 - Remove portability wrapper to condition variables
129 - Remove xbt_os_thread_yield()
132 - MPICH collective selector now mimics MPICH 3.3b
133 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
134 - MPI_Init(NULL,NULL) is now allowed
135 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
136 - Command-line arguments should now be handled properly in Fortran simulations
137 - MPI Attributes and related callbacks should work in Fortran
138 - Apps using C + Fortran should now work
139 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
140 - Support MPI_Error_String
143 - Due to an internal bug, Msg.run() must now be your last line.
144 We hope to fix it in a future release, and we are sorry for the inconvenience.
147 - #22: Process autorestart seem to only work with CAS01 cpus
148 - #56: Feature request: dynamic SMPI replay
149 - #93: simgrid should not eat --help
150 - #111: How to change the loopback link characteristics?
151 - #116: Platform cluster radical
152 - #117: Platform cluster/host order
153 - #129: Replay_multiple on steroid
154 - #131: Java : autostart does not work
155 - #143: Setting a breakpoint at a given time
156 - #235: Network energy models should be integrated
157 - #264: Add ptask L07 resource tracing
158 - #271: Dynamic and manual replay of SMPI traces
159 - #279: Breakpoints option uses time deltas instead of absolute time
160 - #281: Daemonized actor no longer daemonized after an auto-restart
161 - #282: TIT convertor erase traces when using absolute path in the trace list file
162 - #285: segfault when a s4u actor kills itself with msg process tracing activated
163 - #286: Pajé traces are not exposing the number of cores
164 - #287: Command-line parsing should stop when encountering '--'
165 - #288: MPI_Init(NULL, NULL)
166 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
167 - #290: Method to check if a mailbox has a message ready to be consumed
168 - #291: [s4u] semaphores do not seem available yet
169 - #294: Alltoall collective implementation of mpich
170 - #295: Starting a Host as OFF is not supported
171 - #296: DTD too permissive
172 - #299: Add a s4u barrier example
174 ----------------------------------------------------------------------------
176 SimGrid (3.20) Released June 24. 2018
178 The proxy snake_case() release.
181 - The whole API is now using snake_case() instead of mixing it with camelCase.
182 Compatibility wrappers in place for one year.
183 - We took the opportunity to increase the consistency of the API.
184 Please report any remaining glitches.
185 - New function simgrid::s4u::Host::get_actor_count:
186 Returns the number of actors running on a specific host.
189 - SMPI is now tested with ~45 proxy apps from various sources, with none or
190 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
191 - Replay: The replay file has been re-written in C++.
192 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
193 supported. They are stored in the trace and used when replayed.
194 - Basic support of MPI_Cancel. Robustness not guaranteed.
195 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
196 (only for MPI_COMM_TYPE_SHARED).
197 - New option: smpi/privatize-libs, to add external shared libs to be privatized
198 by SMPI. They will be copied locally and loaded separately by each process.
199 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
200 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
201 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
202 with internal index optind with multiple processes. Only works if MPI_Init has
204 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
205 syntax without preprocessing tricks.
208 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
209 behavior tracing is no longer limited to MSG
210 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
211 this means that traces now need to be updated or re-obtained; both Irecv
212 and Isend lines in your traces have now in total 5 mandatory fields:
213 <rankid> <command> <to/from rankid> <tag> <size>
214 To update your traces, it suffices to add a 0 for the tag here.
215 - TIT now also supports waiting for a distinct request via MPI_Wait.
216 Wait/Test now wait for a specific request, not just the last one that was
217 issued. This unfortunately means another update, because we need to
218 identify which request you want to wait for. We do this via the
219 triplet (sender, receiver, tag), which needs to be added:
220 <rankid> <command> <sender> <receiver> <tag>
221 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
222 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
225 - The deprecation of MSG is ongoing (but this should not impact you).
226 Many MSG functions are now simple wrappers on the C API of S4U. If
227 you wish to convert your code to S4U, find the S4U counterparts of
228 your MSG calls in src/msg/msg_legacy.cpp.
229 - MSG can still be used, but won't evolve anymore.
232 - Allow to run the Link energy plugin from the command line with
233 --cfg=plugin:link_energy
234 - Rename Energy plugin into host_energy
235 - Rename Load plugin into host_load
238 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
239 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
240 - Improve performance when handling timeouts of simix synchros.
243 - Config: the C API is now deprecated (will be removed in 3.23), and
244 the C++ API has been slightly improved.
247 - Fix several build issues on OSX.
248 - Move simgrid_config.h to simgrid/config.h (old header still working)
251 - #143: Setting a breakpoint at a given time
252 - #150: Inconsistent event names in SMPI replay
253 - #258: daemonized actors hang after all non-daemonized actors have completed
254 - #267: Linker error on unit_tmgr
255 - #269: SMPI: tracing of MPI_Wait/all/any broken
256 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
258 ----------------------------------------------------------------------------
260 SimGrid (3.19.1) Released March 22 2018
262 The Fixed ABI Release.
264 As you may know, we are currently refactoring SimGrid in deep.
265 Upcoming SimGrid4 will be really different from SimGrid3: modular,
266 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
268 Our goal is to smooth this transition, with backward compatibility
269 and automatic update paths, while still progressing toward SimGrid4.
271 SimGrid remains open during works: The last pure SimGrid3 release was
272 v3.12 while all subsequent versions are usable alpha versions of
273 SimGrid4: Existing interfaces remain unchanged, but the new S4U
274 interface is budding and the internals are deeply reorganized.
276 Since 2015, we work hard to reduce the changes to public APIs. When
277 we need to rename a public library symbol in S4U, we let your
278 compiler issue an explicative warning when you use the deprecated
279 function. These messages remain for four releases, i.e. for one full
280 year, before turning into an error. Starting with v3.15, your can
281 also adapt to API changes with the SIMGRID_VERSION macro, that is
282 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
284 Starting with this v3.19.1, our commitment to reduce the changes to
285 the public interfaces is extended from the API to the ABI: a program
286 using only MSG or SimDag and compiled against a given version of
287 simgrid can probably be used with a later version of SimGrid without
288 recompilation. We will do our best... but don't expect too much of
289 it, that's a really difficult goal during such profund refactoring.
291 The difference between v3.19 and v3.19.1 is that the former was
292 accidentally breaking the ABI of MSG, while the later is restoring
295 S4U and kernel APIs will still evolve until SimGrid4, with one-year
296 deprecation warnings as currently. In fact, cleaning up these
297 interfaces and converting them to snake_case() is one release goal of
298 v3.20. But don't worry, we are working to smooth this upgrade path.
300 In summary, new projects should start with S4U to benefit of the
301 future, but old MSG projects should still be usable with no change.
303 ----------------------------------------------------------------------------
305 SimGrid (3.19) Released March 20 2018
307 The Moscovitly-cold Spring release.
309 SMPI user-visible improvements:
310 - Support of MPI_Type_create_subarray
311 - Unimplemented MPI calls now abort (instead of issuing a warning)
312 Some harmless calls around errhandlers still only issue a warning.
313 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
314 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
315 is now possible (mmap privatization was refactored to this end).
316 Adding new SMPI instances should also be technically doable now.
318 SMPI internal cleanups:
319 - Removed index notion from SMPI -- no more "getPid() - 1"!
320 In tracing files, actors are called rank-<PROCESS-ID> but these are
321 NOT the real ranks (tracing several communicators is not supported yet).
322 - Deprecate smpi_process_index() (will be removed in 3.22)
323 - Better testing on classical proxy apps, through an external project:
324 https://github.com/simgrid/SMPI-proxy-apps/
327 - Use a graphical TOC to make it easier to find the documentation you need
328 - Revamp the MSG tutorial
331 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
332 - Remove parameter of MSG_process_killall().
333 Resetting the PID was bogus anyway (several actors could have the same PID).
336 - Execution->setHost() can be called after start(), to migrate it.
337 - Comm::test_any() is now implemented.
338 - s4u::Actor now has onCreation() and onDestruction() signals.
339 - Install some kernel header files for the users' plugins and more.
340 Warning, their API is really not stable yet.
341 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
342 include/simgrid/kernel/routing/*Zone.hpp
345 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
347 Build System and other cleanups:
348 - Remove unused run-time parameter "tracing/onelink-only".
349 - SimGrid now works with intel compilers (MC still buggy)
350 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
351 - tesh kills subprocesses on timeout (but not on windows)
354 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
357 - #194: Feature request: simgrid::s4u::Comm::test_any()
358 - #245: migrating an actor does not migrate its execution
359 - #253: Feature Request: expose clusters as objects
360 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
361 - #255: Tesh broken on Windows
362 - #256: Modernize FindSimGrid.cmake
363 - #257: Fix (ab)use of CMake install
365 ----------------------------------------------------------------------------
367 SimGrid (3.18) Released December 24 2017
369 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
371 MSG IS NOW DEPRECATED. Please use S4U for new projects.
372 - Support for MSG should not be removed from SimGrid before 2020,
373 but future evolutions will be limited to the integration of
374 user-provided patches.
377 - s4u::onDeadlock() signal, run before stopping the simulation.
378 - s4u::Exec: asynchronous executions (abstraction of background
380 - s4u::Host->getLoad() returns the achieved speed in flops/s
383 - Link::name() is deprecated and will be removed in v3.21, use
384 Link::getCname() instead.
385 - Mailbox::getName() changed to return a std::string, use
386 Mailbox::getCname() instead to get a char*.
387 - Storage::getName() changed to return a std::string, use
388 Storage::getCname() instead to get a char*.
389 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
390 to have both allocation and deallocation of the map in user space.
393 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
394 The amount of flops remaining to do is a vector, not a scalar.
395 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
396 name implies on both sequential and parallel tasks.
397 - Both changes fix GitHub's #223 using PR #237 as a basis.
398 Thanks Michael Mercier.
399 - Most examples were converted to S4U and hidden elsewhere at tests.
400 Rationale: we still want MSG to work; we want newcomers to use S4U.
403 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
404 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
407 - Switch to the faster dlopen privatization mechanism by default
408 - Documentation improvments
411 - Live migration is getting moved to a plugin. Dirty page tracking is
412 the first part of this plugin. This imply that VM migration is now
413 only possible if one this function is called:
414 - C/MSG: MSG_vm_live_migration_plugin_init()
415 - C/C++: sg_vm_live_migration_plugin_init()
416 - Java: Msg.liveMigrationInit()
417 For C and C++, "simgrid/plugins/live_migration.h" also has to be
421 - Define class simgrid::xbt::Path to manage file names.
422 - Removed unused functions:
423 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
424 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
425 xbt_graph_export_graphviz()
426 - xbt/str.h: xbt_str_join()
427 - Remove unused datatypes:
428 - xbt/heap.h: use std::priority_queue or boost::heap instead
429 - xbt/swag.h: use boost::intrusive::list instead
432 - New link_energy plugin for the consumption of the links.
433 - All of the operations on files and storage contents have been
434 packaged into a plugin (src/plugins/file_system). The current
435 public interface can be found in
436 include/simgrid/plugins/file_system.h
437 To use these functions you now have to initialize the plugin by
438 calling MSG_storage_file_system_init() just after calling
439 MSG_init() or sg_storage_file_system_init() just after creating
443 - Remove the undocumented/untested tag <include>
446 - Remove viva specific tracing as the tool is no longer maintained
449 - #248: Kill of finished processes leads to segfault
450 - #240: xbt_cond_wait_timeout should gracefully return for C apps
451 - #239: please implement signal s4u::onDeadlock()
452 - #230: segfaults when exit() without run()
453 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
454 - #223: MSG_task_get_flops_amount() not working with parallel tasks
455 - #222: Actor::kill() doesn't really kill and segfaults
456 - #221: odd LMM warning when killing an actor
457 - #120: Memory leak when the processes are forcefully killed
459 ----------------------------------------------------------------------------
461 SimGrid (3.17) Released October 8 2017
463 The Drained Leaks release: (almost) no known leaks despite the tests.
465 Even very long simulations will preserve your memory: our testsuite
466 stresses SimGrid for over 45mn on fast machines for a coverage of
467 over 80%, and there is only one single known leak, of about 4kb.
471 - Comm.detach(): start and forget about asynchronous emission. A cleanup
472 handler may be given to free resources if the comm cannot be completed.
473 - this_actor::send(mailbox) is now mailbox->put()
474 - New: simgrid::s4u::Comm::wait_all()
475 - New: Host.execute() for remote executions.
478 - Unused option network/sender-gap is removed.
481 - Deprecate MSG_task_isend_with_matching(): This unused feature
482 really complicates our internals. Will be removed in v3.20.
485 - Improved context termination. It is now possible to free resources, even
486 when a process is forcibly killed.
487 - ContextBoost: add support for Boost versions above 1.61.
490 - Bring back run-time option --cfg=exception/cutpath to remove exception
492 - Removed unused functions:
493 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
494 xbt_str_rtrim(), xbt_str_trim().
495 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
498 - Removed header files obsolete since SimGrid 3.12:
499 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
501 - #3: SD_exit should be made optional
502 - #120: Memory leak when the processes are forcefully killed
503 - #159: Threading test regression in Actor refcounting
504 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
505 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
506 - #186: Actor::killAll() segfaults if some process is blocked on wait()
507 - #191: VM migration and pstate
508 - #192: Updating the energy consumptions of all hosts crashes with VMs
509 - #195: All actors have PID=0 in the logs
510 - #204: Sometimes segfault with thread contexts and mmap privatization
511 - #222: Actor::kill() doesn't really kill and segfaults
512 - #225: Actor::kill() doesn't really kill when victims are doing a join()
514 ----------------------------------------------------------------------------
516 SimGrid (3.16) Released June 22. 2017.
518 The Blooming Spring Release: developments are budding.
521 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
522 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
523 - Java JAR file should be finally fully working on Mac OSX too.
525 XML platforms: Switch to platform v4.1 format.
526 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
527 - <zone> can be used as a synonym for the now deprecated <as>
528 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
529 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
530 - <actor> can be used as a synonym for the now deprecated <process>
531 - state_file and avail_file periodicity is now easier to express
532 (check the documentation) the old behavior should still work.
533 * Storage: not backward compatible, but it seems that nobody ever used it
534 - Remove attribute 'content_type' of <storage_type>: was never used
535 - Make attribute 'model' of <storage_type> optional: for future usage
536 - Remove Bconnection model property: was never (in)validated. Replaced by
537 the maximum of the read and write bandwidth as a resource constraint.
540 - New and Backwards Compatibility break:
541 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
542 When one wants to get the list of tasks whose states have changed during a
543 simulation round, s/he has to allocate and free a dynar and use it as
544 argument to this function. The former SD_simulate (double how_long)
548 - Allow multicore VMs, with the correct sharing (unless you overcommit)
549 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
552 - New function to update the consumption of all hosts at once.
553 - Fix the model for multi-core hosts, linear in the amount of busy
554 cores with an abnormality for pIdle (see paper).
555 BUG: Still not satisfactory for ptask on mono-cores.
558 - The netzone are now available from the MSG API.
559 The old names still work, but are now deprecated.
562 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
563 - New option: smpi/keep-temps to not cleanup temp files
564 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
565 mallocs mode is used, allows to change the size of the fake file used
566 (default 1MB), to potentially limit the number of mappings for large runs.
567 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
568 - Fortran ifort and flang compilers support
569 - New RMA calls supported (experimental):
570 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
571 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
572 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
573 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
574 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
575 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
578 - New: this_actor::isMaestro() returns whether we are in kernel mode.
579 - Behind the scene, ActivityImpl are now automatically refcounted.
580 Removing this nasty bottleneck will greatly simplify our internals.
583 - Replay: New function xbt_replay_action_get():
584 Retrieve the function previously associated to an event type.
585 - DROPPED FUNCTION: xbt_str_varsubst()
586 - DROPPED MODULE: strbuff. We don't need it anymore.
587 - DROPPED MODULE: matrix. We don't need it anymore.
588 - DROPPED MODULE: lib. We don't need it anymore.
590 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
592 SimGrid (3.15) Released March 22. 2017
594 The Spring Release: continuous integration servers become green
595 We fixed even the transient bugs on all target architectures:
596 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
597 Windows, FreeBSD, NetBSD.
599 - Rename NetCards to NetPoints.
600 This was intended to help NS3 users, but that's not a netcard.
601 That's a point in the routing algorithm, let's avoid wrong simplifications.
604 - New: MSG_process_yield(). Stop and yield to other processes.
605 - New: MSG_process_daemon(). Daemon processes are automatically killed
606 when the last non-daemon process terminates
607 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
608 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
609 to make room for the upcoming network energy plugin.
610 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
613 - Ensure that an actor can kill itself with Process::exit()
614 - Kill the obscure NativeException. Nobody want to survive the issues
615 it denotes, so use JniException that is a RuntimeException (not to
616 be caught explicitly).
617 - Partial bug fix in initialization. SimGrid flags on command line were
618 consumed at C level but stayed in the original Java String[] args.
619 This could mess users' args[i] if SG flags were not put at the end of
621 The SimGrid flags are now removed from the Java arguments. However,
622 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
623 if args.length is greater than the number of YOUR OWN ARGUMENTS.
624 It might be if you have --log or --cfg flags in the command line.
625 - Fix numerous memleaks all around the place. In particular, around VMs.
629 - s4u::onPlatformCreated: right before the simulation starts
630 - s4u::onSimulationEnd: right after the main simulation loop
631 - s4u::onTimeAdvance: right after a clock change (time is discrete)
632 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
633 event from the availability_file changes the avail speed.
634 - Links are now usable from s4u
635 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
636 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
637 - New examples: The conversion of MSG examples to S4U has begun
638 - Actors: create, kill, migration, and suspend
639 - Applications: master-worker and token-ring
640 - Action replay: communications and storage
641 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
644 - Backwards Compatibility breaks
645 - The SD_route_get_list and SD_route_get_size functions have been removed.
646 They are replaced by the unique
647 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
648 The route from src to dst is built in the links dynar whose size can
649 be obtained with xbt_dynar_length.
650 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
651 sg_host_route_bandwidth and sg_host_route_latency.
652 Macros ensure the backwards compatibility, but you should fix your code
655 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
656 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
657 - Bump our claim of support from MPI 1.1 to MPI 2.2.
658 We don't support 100% of it, but it should be enough. Contact us if not.
659 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
660 naive and probably inaccurate)
661 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
662 mapped onto a unique small file using some system magic.
663 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
664 - Fortran: cleanups, fixes, support of user-added operations
665 - RMA: MPI_Accumulate are applied in correct order.
666 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
667 - Better support for MPI_IN_PLACE
668 - Support for MPI_Win attrs and keyvals.
669 - Support MPI_Comm_group_create, MPI_Type_size_x
670 - MPI ops should be properly applied to all allowed types
671 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
672 algorithm by a non failing one from mpich.
675 - Kill the fifo data container: we don't use it anymore.
678 - A <cluster> can now be created with different speed values to represent pstates.
679 see examples/platforms/energy_cluster.xml for an example.
681 SimGrid (3.14.159) Released December 28. 2016
683 The previous release was only a crude approximation of a Pi release;
684 this one should be a bit better.
686 - Revert a last minute change that broke on Mac OSX.
687 - Fix the changelog and NEWS files.
688 - Improve the documentation, in particular of the routing module.
689 - Integrate some patches flying around in Debian.
691 SimGrid (3.14) Released December 24. 2016
694 * The whole documentation was reworked and reorganized. There is
695 still much room for improvement here, but we're on it.
698 * We now need python3 for our testing infrastructure.
699 * Model checking needs libevent
702 * Now works on FreeBSD
705 * Privatization now works on FreeBSD
707 * Privatization is now activated by default in smpirun
708 You can turn it off with -no-privatize if something goes wrong.
710 * Call-location tracing for SMPI.
711 You can add the exact location (filename / linenumber) of an MPI call to
712 your trace files and slow down or speed up the simulation between two
713 consecutive calls by using an adjustment file (see the documentation).
715 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
716 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
717 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
718 values were ignored in some cases. The timings of these functions can now
719 be significantly different.
721 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
722 smpi/running-power is renamed to smpi/host-speed
724 * smpi/grow-injected-times option to enable or disable multiplication of the
725 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
726 make simulation less precise (but also much faster).
728 * smpirun script should be (much) faster for large deployments.
730 * SMPI tracing: fixed issue with poor matching of send/receives.
732 * Replay: Fix broken waitall
734 New functions and features
735 * MSG_parallel_task_execute_with_timeout, to timeout computations.
737 Dropped / renamed functions and features
738 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
739 - MSG_mailbox_is_empty() -> Mailbox::empty()
740 - MSG_mailbox_front() -> Mailbox::front()
741 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
742 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
743 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
744 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
745 * Don't pass the free_f parameter to property related functions:
746 - MSG_host_set_property_value()
747 - MSG_as_router_set_property_value()
748 - MSG_storage_set_property_value()
749 * VM properties. Since msg_vm_t are msg_host_t, just use
750 MSG_host_get_property_value() and friends
751 * VM I/O related things:
752 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
753 - Unimplemented save/restore methods
754 * MSG_as_router_get_property_value() was redundant with
755 MSG_environment_as_get_property_value().
756 - Removed MSG_as_router_*propert*() functions
757 - Added MSG_environment_as_set_property_value() for consistency
758 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
759 Well, they are still there for now, but deprecated with a warning.
760 Please switch to xbt_dict_new_homogeneous() before this is removed
762 * Task affinity. Its intended behavior (that was very badly tested
763 and probably not really working) was deceiving what most users
764 would have hoped here.
765 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
766 * Option network/coordinates is now useless and should be dropped.
769 * Added option storage/max_file_descriptors to allow more than 1024 files opened
772 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
773 * Added Dragonfly topology support
776 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
777 EiBps, EBps, Eibps, Ebps for bandwidth.
778 They may become useful to some lucky ones.
781 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
782 * ex/app_tokenring: new example, very similar to the MSG Token Ring
783 * ex/async_waitAll: new example, on asynchronous communications
786 * Memory usage should be decreased for simulations with a large number
787 of processes. This also helps for SMPI.
789 SimGrid (3.13) stable; urgency=low
791 The Half Release, a.k.a. the Zealous Easter Trim.
793 - We removed half of the lines, that were mostly experimental cruft.
794 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
795 (not counting blanks and comments -- according to openhub.net)
796 - The internals are now compiled in C++ (and will soon be clean C++)
797 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
798 59 klines of C, etc. We added only 29 klines of C++ in replacement.
800 * Backwards Compatibility breaks
801 - Removed Lua simulation bindings (switch to C or Java for that).
802 Lua can still be used to describe platforms
803 - Removed Java kernel plug-ins.
804 Will be reintroduced after the ongoing major internals reorg.
806 - the following functions were removed.
807 They were too specific and should be reimplemented in a generic
808 way, with filter function.
809 - MSG_task_listen_from_host
810 - MSG_mailbox_get_count_host_waiting_tasks
811 - MSG_mailbox_put_with_timeout was removed.
812 Please use MSG_task_send_with_timeout instead.
814 - the SD_application_reinit function was removed. It has been a noop for a while.
815 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
816 replaced by a more flexible mechanism.
817 - The following functions thus do not exist anymore
818 - SD_workstation_get_access_mode
819 - SD_workstation_set_access_mode
820 - SD_workstation_get_current_task
821 - Basic estimation functions have been removed but can easily be replaced
822 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
823 - SD_workstation_get_computation_time => amount / sg_host_speed()
825 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
826 Use VM.getSpeed()*load/100 for the legacy behavior.
828 - option enable_tracing was removed. It was not doing anything for a while.
829 - In the ModelChecker:
830 - the model-checker now ptraces the model-checked process which means
831 you cannot use a debugger on the latter anymore (we might make this
832 optional in the feature);
833 - removed soft-dirty page tracking;
834 - remove model-checked side snapshot management,
835 MC_snapshot() and MC_compare_snapshot();
836 - keep the MC_cut() function as a stub (it was not really working
837 in the previous release).
840 * All options are consistently in kebab-case. Old names are kept as alias.
843 * Switch to platform v4 format.
844 - Rename from 'power' to 'speed' the attributes describing the amount of
845 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
846 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
847 - In <host> and <link>, attributes availability and state are gone.
848 It was redundant with state and availability traces, and with peak values.
849 - In <cluster>, attributes availability_file and state_file are gone.
850 It was too complex and unused.
851 - Kill <gpu>. Was not doing anything.
852 - The DOCTYPE points to the right URL:
853 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
854 (the file at this address now documents the changelog since its v1)
855 - A warning is emitted for unit-less values (they are still accepted).
856 - speed. Default: 'f' or 'flops'. Also defined:
857 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
858 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
859 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
860 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
861 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
862 - latency. Default: 's' second. Also defined:
863 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
865 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
866 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
869 * s4u::Host is now the preferred public interface to the Host features.
870 sg_host_* functions are C bindings to the exact same behavior
871 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
874 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
875 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
876 They were deprecated since a few years, and probably did not even compile anymore.
879 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
880 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
881 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
882 encouraged to update their codes. A list of the modified functions can be found at the end of
883 include/simgrid/simdag.h
886 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
887 the lambda or closure passed as a parameter will run in kernel mode.
888 All the callback functions should be rewritten to that interface at some point.
891 * Reorganizing and cleaning the internals all around the place.
894 * Remove old default barrier/bcast buggy algorithms (see #18407)
895 * Various bug fixes to handle more codes
896 * Remove the need for the --foreground option of smpirun (it is still
897 accepted for backward compatibility).
900 * Kill the setset data container: MC don't use it anymore.
901 * Kill the queue data container: it made more sense with GRAS.
902 * Kill the xbt_peer_t data type: it's useless without GRAS.
903 * Kill rm_cb feature of config sets: it was never useful.
904 * Kill graphxml parsing feature. It was not used.
905 * Kill the deprecated code protected by XBT_USE_DEPRECATED
907 - xbt_dynar_sort_strings(), when the content is char*
908 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
909 They throw exceptions on invalid input;
910 * C++ support for declaring CLI flags (simgrid::config::Flag);
911 * class for abstracting different signal backends (simgrid::xbt::signal).
912 with no external dependencies (we need very simple signals).
915 * refactoring and cleanup of the code;
916 * ongoing process to cleanly separate the model-checking algorithms
917 from the code model-checking support.
919 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
921 SimGrid (3.12) stable; urgency=low
923 The Facelift Release.
926 * Require g++ v4.7 at least to not speak prehistorical C++.
927 * Require Boost 1.48 (for signal2 component).
928 * Java must be version 7 at least when activated.
929 * Builds on Windows again (including Java bindings).
930 * Tracing is now always enabled (no way to turn it out)
931 * Remove GTNetS. It was not working anyway.
932 * Various cleanups in the cmake scripts.
933 * Move headers around to sort them out on installed systems:
934 - instr/instr.h -> simgrid/instr.h
935 - instr/jedule/* -> simgrid/jedule
936 - simdag/datatypes.h was removed
937 - simdag/simdag.h -> simgrid/simdag.h
938 - msg/datatypes.h was removed
939 - msg/msg.h -> simgrid/msg.h
942 * Interface improvement:
943 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
944 - Sanitize the interface in MSG_task_ module:
945 - Merge two functions that were close enough but misleading:
946 set_compute_duration(t) -> set_flops_amount(t)
947 get_remaining_computation(t) -> get_flops_amount(t)
948 - set_data_size(t) -> set_bytes_amount(t)
949 get_data_size(t) -> get_bytes_amount(t)
950 - Massive cleanups in the functions related to the energy
951 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
952 - New: MSG_host_get_pstate()
954 - msg/energy/onoff: switching hosts on and off
956 * Interface improvement:
957 - Rename Host.isAvail() to Host.isOn()
958 - Rename Process.currentProcess() to Process.getCurrentProcess()
959 - Rename Task.setDataSize() to Task.setBytesAmount()
960 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
962 - #18874: Actually allows the GC to reclaim tasks
966 - SIMIX_process_throw: raises an exception in a remote process
967 * Refactoring: Separate sub-modules
968 - libsmx: the public interface, as libc in a real system
969 - popping: the strange dance that converts a user request into a kernel handling
970 - smx_context_*: the virtualization mechanisms that embed the user code
971 - smx_*: the handling of each simcalls
973 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
974 * Rename smx_action_t into smx_synchro_t, making explicit that these
975 things are used to synchronize processes with their environment.
976 For example, a communication is a sort of synchronization involving
977 the communicating processes (that may block until the exchange) and
978 the platform. The same can be said from computations, etc.
980 - #18888: segfault when a process ends before its kill_time
984 - 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.
985 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
986 - MPI_Type_set_name, MPI_Type_get_name
987 - MPI_*_c2f and MPI_*_f2c functions
988 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
989 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
990 - Activate a lot of new tests from the mpich 3 testsuite
992 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
993 - InfiniBand network model added: Based on the works of Jerome Vienne
994 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
995 - When smpi/display_timing is set, also display global simulation time and application times
996 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
997 * Collective communications
998 - 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
999 - 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/).
1000 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1002 - Replay now uses algorithms from wanted collective selector
1003 - Replay can be used with SMP-aware algorithms
1004 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1005 - 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.
1007 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1008 - Correctly use loopback on fat-tree clusters
1009 - Asynchronous small messages shouldn't trigger deadlocks anymore
1010 * Energy/DVFS cleanup and improvement
1011 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1012 - new: smpi_get_host_pstate()
1015 - "Full" network optimization flag was broken since Surf++
1016 - Better handling of precision flags in maxmin
1017 - Fix bug causing sometimes "Impossible" errors
1018 - Properly pass cluster properties to included hosts
1019 * Improvement of the Energy plugin.
1020 - Always update the consumption before returning that value
1021 - New property: watt_off to denote the dissipation when the host is off
1022 - New functions getWattMinAt and getWattMaxAt to retrieve the
1023 dissipation of pstates that we are not currently at.
1024 * Java: class NetworkLink renamed to Link
1025 * New function: simcall_process_get_kill_time()
1026 * Massive rename s/workstation/host/
1027 - That's intrusive, but that's good for the project consistency. Sorry.
1028 - Change config option "workstation/model" into "host/model"
1031 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1032 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1033 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1034 - xbt_dirname and xbt_basename for non-POSIX systems
1036 * The model checker now runs as a separate process.
1037 * The model checker runs must now be launched with the new simgrid-mc program.
1038 * Record/Replay: the MC can display a textual representation of a path in the
1039 execution graph. It can then be replayed outside of the model checker.
1041 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1043 SimGrid (3.11) stable; urgency=low
1048 * Normalizing pointers addresses tool for better diff between logs
1051 * Add cloud examples using new VMs
1052 - examples/msg/cloud/two_tasks_vm.tesh
1053 - examples/msg/cloud/simple_vm.tesh
1054 - examples/java/cloud/cloud.tesh
1055 - examples/java/cloud/migration/migration.tesh
1056 * Add java surf examples:
1057 - examples/java/surfPlugin/surf_plugin.tesh
1058 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1059 - examples/java/surfCpuModel/surf_cpu_model.tesh
1060 * Add SMPI+MSG example:
1061 - examples/smpi/smpi_msg_masterslave/
1066 - msg task destroy cancel test
1067 - msg_host on/off test
1068 * Move all tests in testsuite to teshsuite (adding tesh files)
1069 * Restructure teshsuites
1070 - one folder for each kind of test
1071 * Restructure AddTests.cmake
1073 - structure the order of tests (with sections)
1076 * Add virtual machine
1077 - creation of a VM on a PM
1078 - migration of a VM from a PM to another PM
1080 - MSG_process_join(msg_process_t process, double timeout)
1081 - msg_bar_t MSG_barrier_init(unsigned int count)
1082 - int MSG_barrier_wait(msg_bar_t barrier)
1083 - void MSG_barrier_destroy(msg_bar_t barrier)
1084 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1085 * New option "msg/debug_multiple_use" to help debugging when a task is used
1088 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1090 - Increase code coverage in test suites
1094 * Protect context stacks against stack overflow. The number of protected memory
1095 pages allocated on the top of each stack (1 by default) can be configured
1096 with the new command line option --cfg=contexts/guard_size:<value>.
1097 * Simcalls are now generated by a python script that
1098 - generates files included by SimGrid
1099 - checks that all the functions exist, and proposes prototypes
1101 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1102 host_set_data, host_get_data
1104 - simcall_process_join(smx_process_t process, double timeout)
1105 * Fix bug where sleeping processing could not be suspended.
1108 * Translate surf models from C to C++
1109 - Generic classes for all models: Model, Resource, Action
1110 - A generic interface for each kind of model (CPU, Network, Storage
1111 Workstation, WorkstationVM)
1113 * Translate surf routings from C to C++
1114 * Add callbacks using sigc++ or boost::signals2
1115 - Add callback functions for resource creation/destruction
1116 - Add callback functions for action state change
1117 - Handle Energy as a plugin
1118 * Replace swag by boost::intrusive
1119 * Add new routing models for clusters. For documentation, see
1120 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1121 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1122 parameters for cluster tag
1123 - Fat trees, with topology="FAT_TREE" and
1124 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1125 - see examples/platforms/torus_cluster.xml and
1126 examples/platforms/fat_tree_cluster.xml
1127 * More documentation
1130 * Hostfiles support host:nb_processes construct to deploy several processes on
1132 * Collective communication algorithms should not crash if used with
1133 improper number of nodes and report the error.
1134 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1135 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1136 MPI_Dims_create, MPI_Cart_sub are supported.
1137 * New interface to use SMPI programmatically (still depends on MSG for
1138 some parts, see examples/smpi/smpi_msg_masterslave):
1139 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1143 * Global variables privatization in MPI executables is now performed at runtime
1144 with the option smpi/privatize_global_variables (default:no).
1145 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1146 dynamic libraries loaded by the application are not privatized (static
1147 linking with these libraries is advised in this case)
1150 * Options defined in XML work correctly now.
1153 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1154 copied into simgrid.jar or not (ON by default). Use this option if you want
1155 to reduce the size of the installed simgrid.jar, *and* the native libraries
1156 are kept installed elsewhere.
1157 * Surf binding with SWIG (code generated in maintainer mode only):
1158 - plugin to handle callbacks
1159 - CPU model only for the moment
1162 * Supernovae build mode is definitively removed. It was used to improve
1163 inlining and inter-module optimizations. It is nowadays superseded by
1164 link-time optimizations commonly available in compilers.
1165 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1166 * Add boost dependency for surf++
1167 * Add new macro for tests
1168 - ADD_TESH(name <tesh_args>)
1169 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1173 - xbt_bar_t XBT_barrier_init(unsigned int count)
1174 - int XBT_barrier_wait(xbt_bar_t barrier)
1175 - void XBT_barrier_destroy(xbt_bar_t barrier)
1176 * Make the xbt_os_time module public
1178 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1180 SimGrid (3.10) stable; urgency=low
1182 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1185 * Reintegrate Java to the main archive as desynchronizing these
1186 package is not acceptable anymore (Java is now considered stable)
1187 * Add explicit synchronization facilities through semaphores
1188 * Bug fix: Task.setDataSize() only changed the C world, not the value
1189 cached in the Java world
1192 * Dramatically change the way files are handled. API and internals changed, but
1193 this part of MSG was not considered as production grade either.
1194 * Add explicit synchronization facilities through semaphores
1195 * Add a new function MSG_host_get_process_list()
1196 * Preliminary DVFS support (see examples/msg/energy/ for details)
1199 * SMPI is now included directly in the libsimgrid as the windows
1200 linker doesn't force us on splitting it anymore.
1201 * Improvements of the SMPI replay tool:
1202 - Most of the collective communications are now rooted in the same process as
1203 in the original application.
1204 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1205 used until now). Multiple data types can now be used in a trace.
1206 - The replay tool now supports traces produce either by TAU or a modified
1208 - Bug Fix: the compute part of the reduce action is now taken into account.
1209 - Gatherv collective is now supported
1210 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1211 of smpirun outputs time independent traces for the current run. One file
1212 is created per process. If too many processes are simulated, this behavior
1213 can be changed to one file for all processes by using the
1214 tracing/smpi/format/ti_one_file flag
1215 * smpirun generates the host file if needed (with given host count and platform)
1216 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1217 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1218 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1219 --cfg=smpi/coll_selector:(mpich/ompi)
1220 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1222 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1223 happening inside a collective SMPI call.
1224 * Fix the behavior of complex data types handling.
1225 * Make MPI_Wtime another synchronization point to take computations into
1227 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1228 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1229 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1230 (no privatization of global variables yet)
1231 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1234 * Verification of liveness properties is now available for SMPI applications
1235 (in addition to MSG applications)
1236 * Bugged examples using SMPI in examples/smpi/mc/
1237 * Add --cfg=model-check/visited option. Allows the verification of infinite
1238 programs. Detection of loops in the execution thanks to the system state
1239 comparison and reduction of the state space to explore. Can be combined with
1240 DPOR for safety properties.
1243 * Allow to change SimGrid configuration (see --help) within the code
1244 thanks to SD_config() as it can be done in MSG.
1245 * Add a new function SD_task_set_amount() upon user request.
1248 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1249 * Remove rule based routing (no more PCRE dependency)
1250 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1251 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1252 * Add a loopback_bw and loopback_lat options to cluster tag.
1253 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1254 routing. Now peers are encapsulated in an AS and have their own private
1255 router but this is transparent.
1258 * Our own implementation of getline is renamed xbt_getline, and gets
1259 used even if the OS provide a getline(). This should reduce the
1260 configuration complexity by using the same code on all platforms.
1261 * New type: xbt_cfg_elm_boolean.
1262 * Allow to use yes/no for boolean configuration options in the command line.
1263 * Allow to disable SimGrid cleanups at exit from command line option.
1264 There are situations where one may want a simulation to end with an exit.
1265 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1266 annoying when scripting around the simulator. Adding a
1267 --cfg=clean_atexit:no allows to circumvent this issue.
1270 * Lots of memory leaks were corrected in this release.
1271 There is no known memory leaks anymore, in all of our 600+ tests.
1272 * New command line option --version, to get SimGrid version information.
1273 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1275 * Supernovae builds are deprecated, and expected to be removed in the next
1278 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1280 SimGrid (3.9) stable; urgency=low
1282 The Grasgory release: GRAS is really dead now.
1284 * Complete overhaul of the internal host structures scheme.
1287 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1288 was considered as experimental and badly maintained since then.
1289 * Keeping it was thus a trap to our potential users, that could take
1290 it instead of MSG or SMPI by mistake despite is pity state.
1291 * GRAS seems to have very few users (if any), and no one volunteered
1292 to maintain it further. It also induces a lot of XBT code (for
1293 portability sake), that must be maintained too.
1294 * For all these reasons, we killed GRAS. If someone wants to revive it
1295 in the future, don't cry, our git history still remembers of GRAS.
1298 * Major overhaul. Merge our documentation again as time proved that
1299 splitting it was really not helping our users.
1300 * Further improve the developer documentation to help newcomers
1301 hacking on SimGrid itself. The user documentation (and in
1302 particular, the beginner documentation) is still in a sorry state.
1305 * Now works on Windows too!
1306 * Much more extensive test suite, from MPICH
1309 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1310 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1311 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1312 * Bug fix: let task be scheduled when the last dependency to be solved is
1313 a control dependency.
1314 * Remove SD_load_environment_script function.
1315 Use the C sg_platf function if you want to declare a platform
1319 * New function: MSG_process_get_number()
1320 * Old function documented: MSG_config()
1321 * Remove MSG_load_platform_script function
1322 Use the C sg_platf function if you want to declare a platform
1326 * Change the default value of the TCP_gamma constant (maximal size of TCP
1327 congestion window) to a more realistic 4MiB value. If you notice changes in
1328 your simulation results, you can fall back to the previous 20k tiny window
1329 by adding --cfg=network/TCP_gamma:20000 on command line.
1330 * (Hopefully) fix a bug wrt periodic availability and state traces
1331 * Bug fix: use default values at start when first event in availability/state
1332 trace is not at time 0.
1335 * remove the "new_" part of function name sg_platf_new_trace_connect
1336 (resulting in sg_platf_trace_connect), since it does not create
1340 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1341 use case where it's really mandatory, and maintaining it was a pain in
1343 * New: xbt_fifo_search(), search an item with a user-provided
1344 comparison function instead of dumb pointer comparison.
1347 * Fix the lua deployment:
1348 Use `simgrid.init_application()` before deployment instead of
1349 `simgrid.msg_register_application()` after.
1352 * Transfer the tracing files into the corresponding modules.
1354 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1356 SimGrid (3.8.1) stable; urgency=low
1358 The "we are told that some people want to also *install* the simgrid
1361 * Add missing file "tesh.1" to the archive.
1363 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1365 SimGrid (3.8) stable; urgency=low
1367 The Psssshiiiit release: SimGrid jumps into the Cloud.
1370 * Add an experimental interface to manipulate VMs. They are mainly
1371 process groups with very few intrinsic semantic, but they should
1372 allow you to build the semantic you want easily.
1373 * New function: MSG_host_set_property_value()
1374 * New function: MSG_process_on_exit(). To clean memory in all cases.
1375 * Bug fixes that made the host (and link) failures unusable.
1376 * Add a way to auto-restart process when the host in which they are
1377 executing comes back (ON_FAILURE="RESTART" on deployment file,
1378 MSG_process_auto_restart_set).
1379 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1380 please stop using the old ones, they are DEPRECATED.
1382 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1383 Please use MSG_init() instead. (reducing the amount of entry
1384 points in the library helps us).
1385 * Make it impossible to link against the wrong version of the lib
1386 * Deprecate MSG_clean(). No need to call it anymore.
1387 * Function MSG_get_host_number() is not deprecated anymore.
1390 * Split the doc into a user guide and a reference guide.
1391 * Start a developper guide to help people hacking on SimGrid.
1394 * Enable tracing by default. This modules rocks you should use it.
1395 * Remove option custom_flags. Now use environment variables CFLAGS
1397 * Use default cmake things to detect lua instead of home grown ones.
1398 * New option "enable_mallocators" to disable mallocators, for debugging
1399 purpose ("on" by default).
1402 * Bug fixes around the resource failures: don't let the processes
1403 survive the host they are running onto.
1404 * Add an interface to auto-restart processes when the host in which they are
1405 executing comes back.
1406 * Ensures that SIMIX_clean is called automatically. It's not part of
1407 the public interface anymore (bindings should be updated).
1410 * Bug fix for when SD_Simulate is called with a positive value: be careful
1411 when comparing doubles. Sometimes they are different for non significant
1413 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1414 parallel task whose initial work is distributed among host according
1415 to the Amdahl's law. Such tasks are created with a parameter alpha
1416 that corresponds to the non-parallelizable part of the computation.
1417 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1418 between two sets of workstations assuming a 1D block distribution (each
1419 workstation owns a similar share of data) on both sides.
1421 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1422 Data redistribution will be automatically scheduled once parent and child
1423 are both scheduled. The filling of computation_amount and
1424 communication_amount structures is now done seamlessly thanks to the chosen
1426 * New function SD_workstation_dump to display various information
1427 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1428 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1429 bandwidth on the route onto which the task is scheduled and the amount of
1431 To divide the nominal bandwidth by 2, the rate then has to be:
1432 rate = bandwidth/(2*amount)
1433 * Compute tasks that have failed can now be rescheduled and executed again
1434 (from their beginning)
1435 * Increasing source code coverage (src/simdag is now covered at 95.8%
1439 * Re-implement time-independent trace replay using SMPI (at the
1440 smpi_smp_* level) instead of MSG. This should replace
1441 examples/msg/actions/actions.c
1442 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1443 hindexed and structs)
1444 * Implement the exchange of non-contiguous data.
1445 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1446 * Correct behavior of smpi/sender_gap and set its default value to 0
1447 * Add option to asynchronously send small messages to allow better
1448 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1449 specifies the size in bytes under which messages will be asynchronously sent.
1450 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1451 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1454 * New C interface to define a platform: XML is now optional.
1455 For more info, please check include/simgrid/platf.h.
1456 * New interface to define random platforms from the C:
1457 For more info, please check include/simgrid/platf_generator.h and
1458 examples/msg/masterslave/masterslave_platfgen.c
1459 * Export a sg_cmdline dynar containing all the arguments we got from
1463 * Two new tracing options for adding comments to trace file so you
1464 can track your experiments (see --help-tracing for details).
1465 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1466 * Adding the SimGrid version that generated the trace file as a comment.
1467 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1468 * Fix to avoid key clashes on Paje links
1469 * Other minor fixes related to the Paje specification
1472 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1473 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1474 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1475 * Greatly improve the robustness of mmalloc to user errors (such as
1476 using an area after freeing it, or freeing it twice)
1478 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1480 SimGrid-java (3.8.1) stable; urgency=low
1482 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1483 Warning: all methods are visible, but only some of them are
1484 implemented so far. Check the source (src/jtrace.c)
1485 for further information.
1486 * New module: org.simgrid.msg.File (SimGrid File management functions)
1487 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1488 * Change the meaning of Process.restart: now restart the process from
1489 the begining, like MSG_process_restart in C.
1490 * Add Process.setAutoRestart: handling of process restart when failed
1492 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1493 you to retrieve the properties of the processes/hosts
1494 * Deprecate Msg.clean(): you can just forget about it now.
1495 * New function Process.getCount(), that only works when compiling
1496 with the not yet released version 3.9 of the C library.
1498 * New context factory based on Coroutines. It mandates a modified JVM
1499 but then, the simulations run about five times faster, and there is
1500 no limit to the amount of processes (beside of the available memory).
1502 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1504 SimGrid (3.7.1) stable; urgency=low
1507 * Restore the prototype of MSG_process_create_with_environment() to
1508 the pre-3.7 situation by removing the kill_time argument.
1509 * Add a MSG_process_set_kill_time() function instead.
1512 * Fix weird behaviors when dealing with parallel tasks.
1515 * Simgrid is now built as a dll.
1516 * Simgrid-java now works on Windows.
1517 * Simgrid-Java is now included into Windows package.
1520 * First pre-build package for MacOSX.
1523 * Fix compilation when using MSG_USE_DEPRECATED.
1524 * Fix some compilation issues on Macs and Windows.
1525 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1526 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1528 We discovered that the Lua console is broken, but we are missing the
1529 manpower to fix it right now. The problem existed in 3.7 too, so we
1530 are not blocking the release for that. Sorry if you depended on this
1531 feature, any help would be really welcome.
1533 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1536 SimGrid-java (3.7.1) stable; urgency=low
1538 The "Java aint got to be bloated and slow" release
1541 * Various internal cleanups and performance improvement
1542 Simulations are expected to run up to twice faster or so
1543 * Make Process.kill(process) an instance method, not a static one
1544 * User processes are not java.lang.Thread subclasses.
1545 This breaks the compatibility (sorry), but previous API was
1546 brain-dead, making it impossible to have non-trivial
1547 initializations in the process constructor.
1548 * Require a full constructor per Process sub-class.
1549 Kinda breaks the compatibility (sorry), but this allows a much more
1550 efficient way to launch the processes at simulation startup.
1551 * Do not embeed our version of semaphores, java 1.5 can be considered
1552 as sufficiently prevalent for us to not dupplicate its features.
1556 * Add examples for almost every part of the API
1557 We spotted and fixed a lot of bugs in the process
1559 * New module: asynchronous communication API
1560 * New function: Process.sleep()
1561 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1562 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1563 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1565 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1567 SimGrid (3.7) stable; urgency=low
1569 The "spring cleanups (before the next Big Project kicks in)" release.
1572 * We can specify the SMPI latency/bandwidth factor with command line
1573 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1574 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1575 You can also use the "config tag" from platform file by adding this line
1576 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1577 (see "example/platforms/tag_config.xml" to use "config tag").
1578 Note that the command line supersedes the platform file configuration.
1579 * Change the correction factors used in LMM model, according to
1580 the latest experiments described in INRIA RR-7821.
1581 Accuracy should be improved this way.
1582 * Use the partial invalidation optimization by default for the
1583 network too. Should produce the exact same results, only faster.
1584 * Major cleanup in surf to merge models and split some optimization
1585 mechanisms from the core of the models. As a result you can now
1586 specify which model to use (e.g., --cfg=network/model:LV08
1587 --cfg=cpu/model:Cas01) and which optimization mode to use
1588 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1589 Incompatible combinations should err at initialization. See
1590 --help-models for the list of all models and optimization modes.
1591 * The CLM03 workstation model was dropped for simplicity because it
1592 used the deprecated CM02 network model. Use default instead.
1593 * Rename the TCP_gamma configuration option to network/TCP_gamma
1594 * Rename the coordinates configuration option to
1595 network/coordinates, and document it
1596 * Use now crosstraffic keyword instead of the terribly misleading
1597 fullduplex keyword. It is activated by default now in the current
1598 default model, use --cfg=network/crosstraffic:0 to turn it off.
1599 * Ongoing refactoring the model parsing to make XML files optional
1600 See include/simgrid/platf.h for details (still to be completed)
1603 * Major overhaul of the documentation. Almost instructive now :/
1604 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1605 functions and friends. This interface was considered as
1606 deprecated since over 2 years, it's time to inform our users that it is.
1607 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1608 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1609 use these (crappy) functions in your code.
1610 These functions will be removed soon. Stop using them now.
1611 * Deprecate MSG_get_host_{table,number}
1612 Implement MSG_hosts_as_dynar() instead.
1613 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1614 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1617 * Stabilize the parallel execution mode of user contexts
1618 * Introduce configuration variables to control parallel execution:
1619 - contexts/synchro: Synchronization mode to use when running
1620 contexts in parallel (either futex, posix or busy_wait)
1621 - contexts/parallel_threshold: Minimal number of user contexts
1622 that must be part of a scheduling round to switch to parallel
1623 execution mode (raw contexts only)
1624 * Fix bugs that prevented to use suspend/resume along with
1625 synchronization structures.
1626 * Fix bugs in process termination that lead to invalid memory access
1627 in very specific conditions.
1630 * Introduce a parallel mode for the models (controlled by surf/nthreads
1631 configuration item). In our tests, running the models in parallel
1632 never lead to any speedups because they are so fast that the gain
1633 of computing each model in parallel does not amortizes the
1634 synchronization costs, even when ultra fast futexes are used.
1635 This is released anyway because YMMV.
1638 * Performance boost by using a swag internally to compute the set of
1639 tasks that are finished and should constitute the return value of
1643 * Enable it by default now that it is considered rather stable.
1646 * Documentation of the tracing functions.
1647 * Performance gains when tracing categorized/uncategorized resource
1648 utilization by avoiding calls to get route when updating resource
1649 variables. LMM constraints are being used instead.
1650 * API changed to set task categories. Use MSG_task_set_category instead
1651 of TRACE_msg_set_task_category, and SD_task_set_category instead
1652 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1653 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1654 * Examples for link user variables added (see at examples/msg/tracing/)
1655 * Deprecated function TRACE_msg_set_process_category completely removed
1656 * Trace header updated according to the latest Paje file format
1657 * Tracing network lazy updates, no longer obligate users to use full updates
1658 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1659 * Experimental: let user code declare/set/push/pop application states for hosts
1660 * API changed to allow the manual creation of graph configuration files
1661 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1664 * Improve the API of Lua MSG bindings, using the Lua spirit.
1665 * Each simulated process now lives in its own Lua world (globals are
1666 automatically duplicated). It helps writing simulators. Will allow
1667 to run Splay programs within SimGrid in the future.
1668 * Add a Chord example in Lua, equivalent to the MSG one.
1671 * Start the implementation of a solution to express temporal
1672 properties, not only local assertions. This is still an
1673 experimental work in progress, stay clear from it to be safe.
1677 - Add new runtime parameters --help-logs and --help-log-categories
1678 to display information about supported logging parameters and
1680 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1681 don't exists anymore.
1682 * Mallocators: allow value NULL for the reset function.
1684 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1685 create homogeneous dictionaries, where all the elements share the
1686 same free function. Non homogeneous dictionaries will be
1687 deprecated in the next release.
1688 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1689 - Multi-level dictionaries are deprecated.
1691 - new function xbt_dynar_search_or_negative() that is useful when
1692 you have less than 2 million elements in your dynar and don't
1693 want of the extra complexity of catching exceptions when the
1694 element is not found.
1696 - Make xbt_os_thread module (for thread portability) public.
1697 Documentation is still to come, sorry.
1699 - Cleanups and simplifications to make it maintainable again.
1700 - Exotic features (such as memalign and valloc) were removed.
1701 - The metadata were extended and improved so that the
1702 model-checker becomes able to explore and inspect the heaps.
1703 - This may induce a performance drop when enable_model-checking is
1704 ON in cmake (even if it's not used in the simulation), but it is
1705 necessary at this point to get MC working.
1707 Turn model-checking OFF if simulation performance matters to you.
1708 Not enabling it at runtime is not enough, disable it in cmake.
1710 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1713 SimGrid (3.6.2) stable; urgency=low
1715 The "Not coding new stuff allows to polish old things" release.
1718 * New bindings to the NS3 packet level simulator (experimental)
1719 * Use the raw (efficient) execution contexts instead of the sysv
1720 (portable) ones when possible.
1721 * libpcre is now mandatory in any cases since not using it led to
1722 severe performance loss and possibly other issues
1723 * Update the XML platforms:
1724 - G5K: include the latest machine in Nancy
1725 - GridPP and LCG: new platforms
1726 * Documentation was partially updated, at least (more to come)
1728 Bug fixes, cosmetics and small improvements
1729 * Free terminated processes before the end of the simulation to avoid
1730 exhausting the memory of users having very dynamic amount of
1732 * Bug fix and cosmetics about canceling non-running tasks
1733 * Bug fix about the dot loader's issues when using libcgraph
1736 * Create an installer for windows with nsis (amd64 and win32)
1737 - Add an hello world project to illustrate simgrid project creation.
1738 - Embed libpcre into the Simgrid installer to avoid
1739 its compilation burden
1740 * The raw execution contexts should work on Apple now
1741 * Port to Windows 64 bits
1742 - Sysv contexts now have an implementation for this arch
1743 - GRAS communication features now support this arch
1744 * Drop support for borland compiler on windows
1745 - this code was not maintained, and we kinda depend on gcc nowadays
1746 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1747 * Fix portability issue on unstable ubuntu: linker became picky on
1750 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1752 SimGrid (3.6.1) stable; urgency=low
1754 The "Oops, we broke Macs too" release
1757 * Fixed contexts detection so that raw ones are used when possible
1758 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1759 leads to a strange error, with user code segfaulting sometimes when
1760 the generated code is not perfectly aligned (which is not
1761 controllable from the user side, depends on the amount of code)
1764 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1766 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1768 SimGrid-java (3.6) unstable; urgency=low
1771 * Split of every thing from simgrid v3.5 into a separate package.
1773 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1775 SimGrid (3.6) stable; urgency=medium
1777 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1780 * Bindings now constitute their own package, separated from the main one.
1781 Rationale: reduce our maintenance nightmare by reducing the module coupling
1782 They will soon be released on their own on gforge.
1784 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1785 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1787 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1788 * It's not quite deprecated for now because we have no replacement,
1789 but it may soon become the case.
1792 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1793 * Fortran: New user-level cache variable to store the rank of the running
1794 process. This improves performance by an order of magnitude.
1795 * C: New coccinelle script to automatically locate and modify global and
1796 local static variables.
1797 * Improved SMPI network model with a sender-side gap to account for multiple
1801 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1802 only say if a communication is finished, no matter whether it succeeded or
1803 failed. You can call MSG_comm_get_status() to know the status of a finished
1805 * New function MSG_task_dsend() to send a task and detach it. When a
1806 communication is detached, you are never notified of its success or failure
1807 and the memory is released automatically once it is finished. This function
1808 is useful when you don't care about the end nor the success of a
1810 * Change the prototypes of action replay. Sorry for inconvenience,
1811 but this is really more efficient this way (and to adapt your code,
1812 you just have to fix the initialization, that shouldn't be too long)
1813 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1814 ever managed to do anything useful with it.
1815 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1816 takes a dynar of communications. It returns immediately and gives the
1817 index of a finished communication (if any).
1818 * New example: a basic implementation of the Chord P2P algorithm.
1821 * New model for multi-core CPUs. You can now use the core attribute to
1822 precise the number of cores of a host. This is a basic model. Every
1823 process running on the host receives at most the power provided in
1824 the DTD (throughput<=power). Total throughput of process cannot exceed
1826 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1827 router linked by an up-link and a down-link (possibly asymmetrical).
1828 This kind of pattern allows to easily build last-mile model style platforms.
1829 Aggregating such patterns in a rule-based AS is thus the technique of
1830 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1831 * New model for Vivaldi routing. We transformed the Vivaldi network model
1832 into a Vivaldi routing model (based on the rule-based model). This allows to
1833 combine Vivaldi based latencies with last-mile platforms.
1836 * Added a check for NaN of IEEE754 infinite in the double entries of
1838 * Introduce a new context factory "raw", highly inspirated from the
1839 ucontext factory, but using manually crafted functions in assembly to
1840 do the work in an efficient manner.
1841 * Allow to change the used context factory at run time, not only at
1842 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1843 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1844 contexts at runtime (only with raw contexts or ucontexts).
1845 * Completely rewrote this module to allow parallel execution of user
1846 processes. Use --cfg=contexts/nthreads:N to execute user processes
1847 with N parallel threads (the default is 1, meaning no parallelism).
1848 * Allow to decide dynamically between sequential and parallel modes.
1849 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1850 processes in parallel only when their number is greater than or equal to P
1852 * Added a check for NaN of IEEE754 infinite in the double entries of
1856 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1857 won't output the state of processes when interrupted with Ctrl-C
1858 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1859 NULL-terminated array. This may solve backward compatibility issues
1860 due to the change to return type of SD_simulate. See also:
1861 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1862 * Add new macros with variable number of arguments.
1863 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1864 - in xbt/asserts.h: xbt_assert
1865 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1866 - in xbt/ex.h: THROWF and RETHROWF.
1867 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1869 * Change xbt_die() to accept a format string with arguments, just like printf.
1870 * New data structure: xbt_lib_t, like a dict but more general and with better
1874 * New configuration options
1875 Options triva/categorized and triva/uncategorized can be used to generate
1876 graph configuration files for Triva visualization tool.
1877 * Configuration option tracing/platform is renamed to tracing/categorized
1878 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1879 * New TRACE_user_link_variable interface:
1880 User provides the name of the link and the tracing variable to attach to it
1881 * the declaration of tracing categories must be done after the environment
1883 * simpler tracing interface, just one way to declare categories
1884 TRACE_category or TRACE_category_with_color, it is up to you
1885 * links in the trace file are again identified by their names
1886 * trace contains the full platform hierarchy exactly as declared using the ASes
1887 * Options tracing/msg/[task|process]:1 groups the process by hosts
1888 for both cases, tasks and processes must have names that are unique during the simulation
1889 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1890 * The experimental option tracing/msg/volume is deprecated
1891 its functionality may be reincorporated if needed
1893 The tracing generates a trace file with unordered timestamped events,
1894 because of the way the core simulator (surf) works. A script available
1895 at the tools directory (fix-paje-trace.sh) can be used to put the events
1896 in order. We have changed the tracing so it can generate ordered timestamped
1897 events in the final trace, but depending on the simulator (and how much time
1898 is simulated) that can lead to a huge memory utilization. It is deactivated
1899 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1901 Build Infrastructure
1902 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1903 - We are trying hard to keep the API stable, but it may happen that
1904 some things change (we're a research project after all, not a
1905 nuclear plant operating system). If such things should happen, you
1906 could rely on that macro to adapt.
1907 - current value: 30600 for 3.06.00, aka 3.6
1908 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1909 a number that can be compared with SIMGRID_VERSION.
1910 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1911 assertions and verbose/debug logging events are disabled at compile time.
1913 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1915 SimGrid (3.5) stable; urgency=medium
1918 * New feature to any SimGrid-based simulator: Model-Checking
1919 Check SIN#1 for more details.
1922 * New Model SMPI (three-interval linear regression for correction factors)
1923 See RR-7426, available at http://hal.inria.fr/inria-00527150
1924 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1925 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1926 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1927 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1928 * See also src/smpi/README
1933 - Tracing API changes: TRACE_start and TRACE_end should not be called
1934 by user-code. They are automatically called by simulators created
1935 with SimDAG, MSG and SMPI if the toolkit is compiled with
1936 tracing_enabled=ON. Categories declaration and utilization remain the
1937 same for MSG and SimDag.
1938 - A function was added to the tracing API to declare categories with
1940 - TRACE_category_with_color (char *category, char *color)
1941 where color must be in the following format
1942 "%f %f %f", red, green, blue
1943 and red, green, blue are float values in the interval [0, 1]
1944 - User can specify NULL as color parameter, or continue calling
1945 TRACE_category (cat)
1946 On that case, the tracing system will define random colors
1947 - The following command-line options are supported:
1948 --cfg=tracing/filename:msg.trace
1949 --cfg=tracing:1 (activate tracing, needed to use others)
1950 --cfg=tracing/platform:1 (categorized resource use)
1951 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1952 --cfg=tracing/msg/task:1 (task creation)
1953 --cfg=tracing/msg/process:1 (process creation, migration)
1954 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1955 --cfg=tracing/smpi:1 (SMPI interface tracing)
1956 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1957 - examples of examples/msg/tracing updated
1959 - DAXLoader and DOTLoader functions can generate tasks with categories
1960 - A new function to attribute a category to SD tasks:
1961 TRACE_sd_set_task_category (SD_task_t task, char *category)
1962 * Tracing the MPI interface implemented by SMPI
1963 - Collective operations are traced with states
1964 - Point-to-Point operations are traced with states/links
1965 - Tracing activated by a parameter "-trace filename" passed
1966 to smpirun during execution (considering that simgrid
1967 is compiled with tracing enabled)
1968 - To run the simulation with gdb, the simulator
1969 accepts --cfg=tracing/smpi:1 to trace SMPI
1970 - tesh files to check if smpi tracing is ok
1971 - See examples/smpi/NAS/DT-trace
1972 * GTNetS tracing re-worked
1973 - adaptation to the tracing system of GTNets to cope
1974 with modifications regarding the fullduplex mode
1975 - new tesh files to check if gtnets tracing is ok
1978 * Asynchronous communications through the functions:
1979 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1980 * New function: MSG_load_platform_script()
1981 to make possible using a lua script instead of XML files to set up platforms
1982 * New function: MSG_set_function
1983 to associate functions to processes, used when bypassing the parser
1984 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1986 Platforms: Add some more examples in examples/platforms
1987 * Grid'5000: see www.grid5000.fr
1988 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1991 * Change the XML format. This is a very important modification. SimGrid
1992 3.5 introduces a new hierarchical format based on the notion of
1993 Autonomous Systems. Compatibility with old format is ensured
1994 through the perl script provided in the install bin directory
1995 bin/simgrid_update_xml.
1996 It is now possible to build platforms with specific routing
1997 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1998 connect several platforms together. We will try to provide soon
1999 set of realistic platforms exploiting these properties (have a
2000 look at examples/platforms/ for the moment).
2001 * Take the opportunity of the XML format change to be a good XML citizen:
2002 rename link:ctn to link_ctn and similar changes (also dealt with by
2004 * Add a new routing scheme (rule-based) using regular expressions. It
2005 enables to have an extremely low memory footprint when the
2006 underlying routing is simple and can be compactly described. You
2007 need to have libpcre4-dev (perl regular expressions) installed if
2008 you want to use this routing scheme.
2009 * Revive the cluster TAG and allow to easily and efficiently (both in
2010 term of memory and speed) connect clusters together. Have a look
2011 at teshsuite/simdag/platforms/ to see how this can be done. With
2012 this tag, you can create clusters with thousands of tasks at no
2013 cost (have a look at examples/platforms/).
2014 Note: clusters are implemented as ASes, so there is no need for an
2015 enclosing AS tag if you have only one cluster in your platform.
2016 * Add new generic functions in the public interface that allows the user
2017 to call SURF 'create_resource' methods from your code (same
2018 functionality as the XML bypass mechanism but with a much lighter
2020 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2021 that uses a piecewise linear approximation to produce better
2022 results when exchanging small messages.
2023 * Add a new parameter to handle correctly full duplex link and account
2024 for interferences between uplink and downlink communications
2025 (activate with --cfg=fullduplex:1).
2028 * Rename the SD_READY (all dependencies are satisfied and task is
2029 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2030 dependencies are satisfied) state.
2031 This prevents a confusion between the notion of "ready to schedule"
2032 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2033 simulated" (SD_RUNNABLE) used by the simulation kernel.
2034 * Change the way a task is considered as ready. Instead of removing
2035 dependencies when a task is done, a counter is decreased. This way,
2036 it is always possible to reach ancestors thanks to the
2037 SD_taks_get_parents function (even after the end of the simulation.)
2038 * Change the return type of SD_Simulate from (SD_task_t*) into
2039 xbt_dynar_t. This function was in handling a dynar internally and
2040 converted it into a NULL terminated array for historical reasons.
2041 * New function SD_dotload(char*) to load a DAG described in dot
2042 format. This loader and the corresponding examples require the
2043 installation of the graphviz library.
2044 * Fix a bug in the management of tasks of size 0 in the surf network
2045 models. This problem was only visible with SIMDAG and you should
2046 thus disregard results produced with earlier versions if you
2047 relied on this feature (some tasks were blocked because of this).
2048 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2049 even though your applications were doing only point-to-point
2050 communications and sequential computations. Now you can really use any
2051 model you want (of course, if you create real parallel tasks, which are
2052 not implemented in most models beside ptaskL07, this will abort).
2053 * Add an example that schedules a DAX on an heterogeneous platform
2054 using a Min-Min strategy.
2055 * New function SD_workstation_get_current_task() that returns the kind
2056 of task currently running on a workstation in the sequential access
2058 * Raise some warnings when unexecuted tasks remains at the end of the
2059 simulation. This is usually caused by cycles in the DAG.
2062 * New function: SIMIX_process_set_function() called by MSG_set_function
2063 * Change the underlying waiting queue in semaphores so that a process
2064 can wait on several of them simultaneously (as in waitany).
2065 * Fix the way to handle tokens in semaphores so that all access patterns
2066 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2067 * kill the dirty pimple SIMIX_message_sizes_output()
2068 Please use (proper) visualization instead
2071 * New data container: setset (set of sets of elements)
2072 * New module: mmalloc (mapped malloc, allowing to have several
2073 independent segments of malloc)
2074 * New function: xbt_dict_cursor_set_data()
2075 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2076 * New function: xbt_dynar_is_empty()
2077 * New function: xbt_fifo_get_last_item()
2078 * Fix xbt_dynar_shrink(): use the right element size.
2079 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2080 uninitialized areas during expand.
2081 * Fix semaphores: previous implementation was severely broken.
2082 * Use library init/fini functions for our initialization.
2083 - you can use logs and other feature as soon as you want in your
2084 code (even before the xbt_init / MSG_init)
2085 - xbt_exit is now a no-op and produce a warning when used.
2088 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2089 benefit from the latest improvement to the simulation kernel.
2090 * Kill measurement sockets for now. If you rely on them, sorry. This
2091 release is not for you. This feature will be reintroduced in the
2092 future, but we cannot delay the release any further.
2093 * New function: gras_msgtype_get_name().
2094 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2095 * Fix (at last) the pmm example: it should not randomly fail anymore.
2097 Build chain: bug fixes and overall polishing
2098 * Cmake is now stable enough. Hence, we killed the autotools.
2099 * Port to windows ( TM :)
2100 * Fix the 'make install' target.
2101 No need to use 'make install-simgrid' anymore
2102 * Introduce a 'make dist' target compiling a *source* archive
2103 'make package' compiles a binary archive
2104 * Compile java files only on need
2105 * Add --cd and --setenv command line options to tesh
2106 * Out of source builds are not fully supported yet, but we are close
2107 * Enable supernovae and optimization flags by default for our users
2110 * Add layer to set up environment directly from lua, without XML.
2111 * The effect of gras_stub_generator can be achieved through
2112 lua too (check examples/gras/console/ping_generator.lua)
2114 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2116 SimGrid (3.4.1) stable; urgency=low
2118 The "Polishing easter eggs is probably a good idea" release.
2119 This is a bug fixes release only.
2123 * Fix a bug preventing the tasks from begin garbage collected.
2126 * Fix a bug occuring when a host involved in a communication fails.
2127 This was not detected properly by the other peer involved in the
2128 communication. Now, it's reported as a network error.
2131 * Warn the user about loop dependencies in data flow of DAX files
2132 * Obey the control-flow dependencies of DAX files
2135 * Add option "enable_smpi" allowing to not compile SMPI.
2136 Probably useful for the (Mac) users experiencing a build error here
2137 * Improve the detection of lua5.1 and ruby1.8
2139 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2141 SimGrid (3.4) stable; urgency=low
2143 The "Easter in Cargese" release. Also known as (major changes):
2145 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2146 ~> bindings were greatly improved
2147 ~> new tracing infrastructure for better visualization introduced
2149 * the "Welcome to configury modernity" release.
2150 ~> we switched from autotools to cmake, and improved our cdash
2153 A more detailled list of changes follow (full detail in svn log).
2155 Java Bindings: Various Cleanups
2156 * (install java-gcj-compat-dev on debian-like to use them)
2157 * Remove put/get: no need to export deprecated interface in Java
2158 Use send/receive instead.
2159 * Cleanup the examples and add a README per directory
2160 * Remove example autoDestination (that's the only way to go now)
2161 * Remove example explicitDestination (was a plain copy of basic)
2162 * Make JniException a runtime exception, so that there is no need to
2163 declare the fact that you may encounter such a beast. I guess that
2164 nobody will ever want to survive such error.
2165 * Create specific errors for each MSG case of failure:
2166 host failure, transfer failure, timeout, task cancelled
2167 * Cleanup the exceptions that may get thrown by each function
2168 * Other internal cleanups in Java bindings. Performance still bad :/
2169 Ruby and Lua Bindings: create them
2170 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2171 * That's new and great, you should try them out.
2172 Same functionalities than Java bindings, only even less polished
2174 * Kill the useless "rate" argument of SD_task_get_execution_time()
2175 Everyone used to provide -1 as a value, it was not used, and the
2176 semantic of a possible use wasn't even clear.
2177 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2178 either as comm costs or compute costs to mean that there is no
2179 such thing for that specific task.
2180 * Add a SD_task_set_name() function
2181 * Fix SD_task_unschedule() on typed tasks
2182 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2183 * In DAX loader, accept useless 'level' attributes to <job> since
2184 LIGO DAGs have them (seem to be to ease graphical representation).
2186 * Add an example masterslave_mailbox.c using send/receive and not
2187 the deprecated put/get interface.
2188 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2189 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2191 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2192 (declare MSG_USE_DEPRECATED to still have the old name)
2193 * Add a MSG_task_set_data() function
2194 * About trace replay (see examples/msg/actions):
2196 - Allow to work with splitted trace files for each process
2197 Give the specific trace file as argument of each process,
2198 and call MSG_action_trace_run(NULL)
2199 You can still have one merged file for all processes.
2200 - Fix implementation of collective operations
2201 * Allow task_execute() on 0-sized tasks (closes #10063)
2203 * This is the first release of SimGrid where SMPI is not considered
2204 beta anymore (even if some corners should still be improved)
2205 * Port over the new SIMIX_network submodule (internal refactoring)
2206 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2207 * Implement more missing elements of the standard:
2209 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2211 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2212 MPI_Scan() MPI_Get_processor_name()
2213 - Added implementation of missing case for Alltoall (warning: it's
2214 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2215 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2216 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2217 SMPI_MPI_Allgatherv()
2218 * Bug fixes include:
2219 - MPI_Waitsome() was broken
2220 - Allow relative includes in smpicc
2221 - Command line cfg argument 'reference_speed' was ignored...
2222 - Some functions did not properly lead to auto-benching of user code
2223 - smpicc passes -O2 by default (just like openmpi one)
2225 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2226 * Bug fixes about timeouts during communications
2227 * add SIMIX_message_sizes_output() as a pimple to write to file the
2228 amount of messages per size. Use gnuplot to get histogram.
2229 Pimple because that's the only user-visible function of simix,
2230 defined directly in xbt.h (irk, sorry)
2232 - Add a SIMIX_sem_get_capacity() function
2233 - Fix interactions with processe resume/suspende
2234 - release_forever() was stupidly broken
2235 - Fix SIMIX_display_process_status() for processes in a semaphore
2236 - Make SIMIX_sem_block_onto() user-visible
2237 * Refactoring context stuff:
2238 - Use pseudo-OOP for better modularity
2239 - reimplement SIMIX_process_kill() without process_schedule() so
2240 that the latter can take as invariant that it is called from
2242 - Merge context_start into context_new for sake of simplicity
2244 * Add a Vivaldi network model, coded live during SUD'10 ;)
2245 * Rename configuration variables to start a hierarchy:
2246 o cpu_model -> cpu/model
2247 o network_model -> network/model
2248 o workstation_model -> workstation/model
2249 * New configuration variables:
2250 o network/bandwidth_factor: correction to bandwith
2251 o network/latency_factor: correction to latency
2252 o netwotk/weight_S: correction to the weight of competing streams
2253 * Add a long description to the models, that users can see with such
2254 argument on the command line: --cfg=cpu/model:help
2255 * --help-models display the long description of all known models
2257 * config: add the ability to set a default value after registration
2258 Does not override any previously set value (e.g. from cmd line)
2259 * dict: allow to have integer key and data.
2260 When so, you need to use the following functions
2261 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2262 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2263 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2264 In contrary to regular dicts, the key is not malloced before copy.
2265 Mixing scalar and regular elements in the same dict is not tested
2267 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2268 Tracing for Visualization:
2269 * SimGrid is now instrumented in order to generate a trace file for
2270 visualization analysis: to use it, need to compile SimGrid with the
2271 "tracing" option enabled, and instrument the program using SimGrid with
2272 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2273 (among other functions).
2274 * The instrumentation only traces the platform utilization for now
2275 * Documentation to use the tracing functions and how to analyze the
2276 traces with the Triva tool is written.
2277 * More information about: SimGrid FAQ (in the section Tracing Simulations
2280 * We moved to cmake as default build system. Autotools support will
2281 be dropped soon. Check the FAQ for more info about how to use it.
2282 * Greatly improved our cdash/ctest interactions
2283 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2284 * Added memory checking tests with valgrind; lot of memleak fixing.
2285 This may be the first release of simgrid with so few memory issues
2286 * Added code coverage tests.
2287 Our coverage is still improvable, but at least we see it on cdash.
2289 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2291 SimGrid (3.3.4) stable; urgency=low
2293 The "Desktop Grid needs love too" release (also called Xmas release).
2295 Models improvements:
2296 * Major speedup in the maxmin system solving by using lazy evaluation
2297 Instead of solving completely the maxmin system at each iteration,
2298 only invalidate (and recompute) the modified parts.
2299 This new feature is enabled in default models but you can try to
2300 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2301 * Cas01 IMproved as default CPU model
2302 This CPU model is the same Cas01 model, but it uses the
2303 maxmin-selective-update flag and a heap structure to manage
2304 actions on SURF kernel.
2305 It reduces the complexity to find the next action to finish and,
2306 consequently, it's faster than the old Cas01.
2307 This is the new default CPU model (Cas01).
2308 * Rename the old Cas01 model to Cas01_fullupdate
2309 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2310 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2311 * CpuTI (CPU Trace Integration)
2312 A new CPU model whose objective is simulate faster when using
2313 availability trace files.
2314 Instead of using a full featured, over engineered maxmin system for
2315 CPU modeling, this model does the pre-integration of traces files
2316 to calculate the amount of CPU power available, and so, executes
2317 faster than the old CPU models.
2318 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2319 * Use LV08 as default network model since it gives better accuracy
2320 for small messages and shouldn't change things for big ones.
2321 Use --cfg=network_model:CM02 to get the previous behavior.
2324 ******************************************
2325 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2326 ******************************************
2327 * The new CPU model may changes simulations!
2328 The point is that events occurring at the exact same timestamp
2329 are not scheduled in the same order with the old and new
2330 version. This may be enough to completely change the execution
2331 of simulations in some cases.
2332 * The new network model will change simulations!
2333 This new model is more realistic than the previous one, so you
2334 should consider redoing your old experiments with this model.
2335 Sorry for the inconvenience.
2338 * Introduce the supernovae compilation mode
2339 When compiled that way, the whole SimGrid (or almost) is put in a
2340 single compilation unit and compiled in one shoot.
2341 This is to help gcc which has difficulties to inline stuff from one
2343 The speedup seem to be above 15%, althrough more tests are needed on
2344 amd64 to confirm that gain.
2347 * Port of MSG's mailbox on top of SIMIX network
2348 The put/get mechanism was greatly simplified on the way.
2351 * New SIMIX network module. Provides:
2352 - Mailbox: rendez-vous mechanism to find with who you want to speak
2353 - Synchronous send/recv: easier and hopefully faster since the
2354 logic is handled in the maestro process directly now
2355 - Asynchronous send/recv: you dreamt of it? It's here now
2356 Too bad that nobody cared enough to propagate the change to MSG.
2357 * Add semaphores as SIMIX synchronization mechanism.
2360 * new function SD_daxload(char*) to load a DAX file
2361 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2362 * Introduce typed tasks. Specify its kind and cost at creation.
2363 At scheduling, just give where it should be placed, and the cost
2364 for each involved resource is automatically computed.
2365 Existing constructors so far (more to come of course):
2366 - SD_task_create_comm_e2e() for end-to-end communication
2367 - SD_task_create_comp_seq() for sequential computation
2368 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2369 * new function SD_task_dump() for debuging display
2370 * new function SD_task_dotty(task,FILE*) writing to file the info
2371 about the task in dotty format
2372 * SD_task_dependency_exists() can now cope with having one of its
2373 arguments NULL. If so, it tests whether the other argument has any
2375 * Add getters on list of preceding/following tasks:
2376 SD_task_get_parents(task) and SD_task_get_children(task)
2377 * Add getters on amount of workstations and list:
2378 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2379 * Add getter on task kind: SD_task_get_kind(task)
2380 * Update the start_time and finish_time of tasks on completion/failure
2381 * Bugfix: Remove task from state swags when destroyed
2384 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2387 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2388 dynar with one separate thread per value of the dynar.
2389 * Change the prototype of xbt_thread_create(), sorry.
2390 Added a boolean parameter indicating whether we want to join this
2391 thread (used in SG only for now)
2392 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2395 * GTNetS wrappers should now be usable again (and betterly tested too)
2396 * Fix a major regression from 3.2 where the timeout provided to
2397 MSG_task_put_with_timeout() was used as absolute time before which
2398 the comm should be done.
2399 * Start to fix the <cluster> tag.
2400 - Internal links should be good now (beside of the loopback, which
2401 use the private link instead)
2402 - paths to the external world is still rather broken
2403 - the <route:multi> tag is just broken. Actually that's brain-dead.
2404 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2405 to make it less stupid
2406 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2407 * Fix a source-level compatibility glitch from 3.2: after defining
2408 MSG_USE_DEPRECATED, you can use the old name
2409 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2410 * Allow to compile from the SVN with automake 1.11
2411 * Fix some problems when using the "start_time" tag in deployment XMLs.
2412 * Fix #8569: XBT/synchro.h has redundant declarations
2413 * Fix #8563: MSG return values and exceptions
2414 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2415 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2416 Thanks to Robson Peixoto.
2417 * Fix a vicious bug in dictionaries inducing that some elements were
2418 not freed on xbt_dict_free()
2420 Portability report of this version:
2421 * Main portability targets:
2422 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2423 on (amd64/i386/ia64)
2424 - mac leopard on i386
2425 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2426 but nothing critical.
2427 * Other platforms: windows, AIX and others were not tested for this release
2429 Timing report of this version:
2430 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2431 scenario parameters). From 8h to a few seconds in desktop grid settings.
2432 * Supernovae brings about 25% speedup on i386.
2434 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2436 SimGrid (3.3.3) stable; urgency=low
2438 The "Need for Speed" release.
2440 The timings done to validate the 3.3.2 were faulty.
2441 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2443 The problem was a conversion from a manually handled vector to
2444 xbt_dynar_t on the critical path.
2445 xbt_dynar_foreach calls functions, inducing stack management crap.
2447 We inlined these functions and xbt_dynar_foreach is now breath taking.
2448 We also inlined xbt_swag_belong on the way.
2450 Here are some approximate speedup measurements (on master/slaves
2451 simulations lasting between 10s and 20s each):
2452 3.3.1 -> 3.3.2: about same performance
2453 3.3.2 -> 3.3.3: 40% speedup
2454 3.3.1 -> 3.3.3: 40% speedup
2455 3.3.1 with inline patch -> 3.3.3: 30% speedup
2457 Our reading is that the refactoring which occurred in 3.3.2 made us
2458 suffer much more from the xbt_dynar_foreach low performance, but
2459 once we solved this, this refactoring proved to be very performance
2460 effective. From the 40% speedup, somehow, 10% are due to the
2461 inlining and 30% to the refactoring.
2463 That's a pitty that gcc cannot inline functions placed in other files
2464 alone. We have to choose between:
2465 - break the encapsulation (by putting private data structures and
2466 accessors in headers files to help gcc)
2467 - live with low performance
2468 - switch to a decent compiler such as icc (not quite possible).
2470 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2472 SimGrid (3.3.2) stable; urgency=low
2474 The "Simplicity does not preceed complexity, but follows it" release.
2476 The main contributors of this release were (lexical order):
2477 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2480 * Extract the routing logic into its own object.
2481 (was dupplicated in network.c and workstation_LV07.c;
2482 Allows to implement other ways of storing that info)
2483 => kill now useless network_card concept
2484 - Use dynar to represent routes (instead of void** + int*)
2485 - kill link_set (use surf_network_model->resource_set instead)
2486 - Add a command-line option to choose the routing schema to use
2487 - Add three new models:
2488 * Floyd (shortest path computed at initialization)
2489 * Dijikstra (shortest path recomputed all the time)
2490 * Cached Dijikstra (shortest path computed on need)
2491 All these models where contributed by Silas De Munck, and are
2492 described in his ICCS09 paper.
2494 * Simplify model declaration
2495 (less redirections, less function to write when defining a model)
2496 - Factorize stuff between models:
2499 surf_model_resource_set(model)
2500 surf_model_resource_by_name(model, name)
2501 - Unify the types of models in s_surf_model_t (using an union)
2502 - Embeed fields of common_public directly into s_surf_model_t
2503 - Rename model methods:
2504 action_free ~> action_unref
2505 action_change_state ~> action_state_set
2506 action_get_state ~> action_state_get
2507 - Change model methods into functions:
2508 (model)->common_public->action_use ~> surf_action_ref
2510 * Implement a generic resource; use it as ancestor to specific ones
2511 (allows to kill duplicated code in models)
2512 Drawback: timer command don't need no name nor properties;
2513 workstation_CLM03 don't need no properties
2514 (but I guess we can live with those few bytes wasted)
2516 * Improve the action object model
2517 - implement a constructor avoiding dupplicated code about field
2518 initialization in generic_action part.
2520 * Kill the SDP model: it has an external dependency, is deprecated
2521 in flavor of modern lmm models, and didn't compile since a while
2524 * Relocation of the context module from XBT to SIMIX.
2525 (the context were decoupled from the simix processes, duplicating a lot of code)
2526 => a lot of code was factorized
2527 - less overhead is introduced during scheduling
2528 - simpler API for the context factory
2529 - the logic for process creation,destruction and manipulation was simplified
2530 * Simplification of the s_smx_process_t data structure.
2531 => accesing the simix level data associated to a process is faster now,
2532 and the code is a lot more readable.
2535 * Implement some more MPI primitives:
2536 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2537 -implementation: Bcast: flat or 2-ary tree (default),
2538 Barrier: 4-ary tree,
2540 Allreduce: Reduce then Bcast
2541 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2542 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2543 Alltoallv: flat tree, like ompi
2545 * Add support for optimized collectives (Bcast is now binomial by default)
2546 * Port smpirun and smpicc to OS X
2549 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2550 Shout out if you used it.
2553 * Display the list of still queued messages in SG mode when existing
2557 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2558 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2559 * Add xbt_str_from_file(FILE*)
2560 * Add xbt_dict_get_key achieving a linear reverse search
2561 * Remove the context module
2563 Portability report of this version:
2564 * Main portability targets:
2565 - Linux(debian)/x86/context
2566 - Linux(debian)/x86/pthreads
2567 - Linux(debian)/amd64/context
2568 - Linux(debian)/amd64/pthreads
2569 On these, we still have the eratic breakages of gras/pmm and
2570 amok/saturate_sg reported in previous version. We still think
2571 that the tests are the cause of the fault, not the tested code.
2573 - Mac OSX Leopard/x86/context
2574 Still false negative in tesh autotesting.
2575 Smpi still fails, but this time because readlink does not accept -f
2576 Everything seems to work properly beside of that.
2579 - AIX version 5.3 (only tested contexts this time)
2580 Smpi still fails there because mktemp is not installed.
2581 Everything seems to work properly beside of that.
2583 I managed to compile it for the first time, but several breakages.
2584 Won't delay the release for this exotic platform.
2586 * Windows: it's still lagging behind. If you want to help, please
2589 Timing report of this version:
2590 This version seem to be more than 5% faster than 3.3.1 (on linux
2591 64bits with contextes). The gain is less than expected, we are
2592 investigating this for next release.
2594 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2596 SimGrid (3.3.1) stable; urgency=low
2599 * Implement a --cfg-help to show existing configuration variables
2600 * Build chain do not require doxygen in maintainer mode
2603 * fix a bug on struct sizeof computation, which prevented the
2604 exchange of arrays of structs in some conditions
2605 - added a regression test about this in datadesc_usage
2606 * Allow the exchange of 0-long dynamic vectors.
2607 - for that, use -1 as indicator of dynamic size instead of 0
2608 - This implied to change any size from unsigned long to long,
2609 reducing a bit communication abilities, but I guess that with
2610 64bits being quite common, this is more than enough.
2611 - This also induce a protocol change, thus bumping network protocol
2612 version from 0 to 1 (if we have external users, we have to get
2613 clean on that point too ;)
2614 - added two regression tests about this in datadesc_usage
2615 * Be more verbose when propagating local exceptions
2616 This helps debugging.
2617 * Display the status of simulated processes when receiving SIGINT in
2621 * Allow to control the simulation from a trace file.
2622 New functions MSG_action_register() and MSG_action_trace_run()
2623 The first one allows to associate a function execution to each
2624 kind of action while the second one parses a trace file and
2625 triggers the corresponding actions within the system.
2626 For now, only a toy example is provided in examples/msg/actions
2627 * Add an exemple of process migration in examples/msg/migration
2628 * Fix a bug in task exchange which broke MSG_task_get_sender()
2629 Add a teshsuite regression test for that.
2630 [Bug: if MSG_task_get_sender() is called after sender exit,
2632 * Fix a bug which prevented suspend/resume to work properly
2633 * Display the status of simulated processes when receiving SIGINT
2634 This fixes a regression of v3.3. due to the introduction of SIMIX
2635 * Bug fixing in failure management:
2636 - trace could not start by a failure at time 0
2637 - failure during communications were not working
2640 * Add SIMIX_process_set_name() to change the name of the current
2641 process in the log messages.
2642 * Store smx_hosts in a dict since we only retrieve them by name
2643 * Move the configuration infrastructure to surf
2646 * Move the configuration infrastructure to surf
2649 * Massive internal cleanups:
2650 - Store internal structures on processes instead of hosts (allows
2651 to have more than one process per host, in addition of being more
2653 - Cleanup the initialization/finalization process
2654 - Kill a whole bunch of unneeded synchronization:
2655 processes run in exclusive manner within the simulator
2656 - Move queues from global tables to process data fields
2658 - now accept -platform and -hostfile arguments
2659 - Pass the right rank value to processes according to the hostfile
2660 * Compile the examples by default, and use them as regression tests
2661 * Implement MPI_Wtime()
2662 * Change the reference speed to a command line option
2665 * TCP_gamma can now be specified as command line option using
2666 --cfg=TCP_gamma:10000000.0
2667 * Change the --surf-path cmd line option into --cfg=path:
2670 * Also include strbuff from xbt.h public header
2671 * xbt_ex_display(): do not free the exception after displaying
2672 This allows to do more with the given exception afterward.
2673 Users should call xbt_ex_free() themselves.
2677 Portability report of this version:
2678 * Main portability targets:
2679 - Linux(debian)/x86/context
2680 - Linux(debian)/x86/pthreads
2681 - Linux(debian)/amd64/context
2682 - Linux(debian)/amd64/pthreads
2683 These targets fail about 1/10 of times on gras/pmm, but we believe
2684 that this is because of the test, not because of simgrid.
2685 amok/saturate_sg fails even more rarely, and the test may not be
2688 - Mac OSX Leopard/x86/context
2689 The test suite still spits tons of errors because some obscure
2690 force prevents us from removing the temporary directories
2691 arguing that they still contain some metadata I've never heard of.
2692 Smpi fails because seq is not installed.
2693 Everything seems to work properly beside of that.
2696 - AIX version 5.3 (both contexts and pthread)
2697 Smpi still fails there because mktemp is not installed.
2698 XML inclusions seems rosty on AIX.
2700 * Windows: it's still lagging behind. If you want to help, please
2703 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2705 SimGrid (3.3) stable; urgency=high
2709 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2710 [Malek Cherier & Mt]
2712 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2715 Until now, GRAS were using MSG as an interface to SURF. It was
2716 quite difficult because both interface have several differences
2717 (MSG channels vs GRAS sockets were the most notable point).
2719 This also opens the gate to SMPI (which should occur soon) and speed
2720 up simulations by to 40% (even if it were not the main goal).
2722 **************************************
2723 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2724 **************************************
2725 The point is that events occuring at the exact same timestamp are
2726 not scheduled in the same order with the old and new version. This
2727 may be enough to completely change the execution of simulations in
2728 some cases. Sorry for the inconvenience.
2730 * Cleanup and upgrade the XML format to push further scalability
2731 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2733 * Improve the testing infrastructure with tesh. Now a very large part of
2734 the code is tested not only by being run but also by checking that the
2735 output match an expected output [Mt].
2737 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2738 is really less memory-demanding, which should allow you to use
2739 larger files in SimGrid [AL].
2741 * Inform valgrind about our contextes, so that it becomes usable
2742 with the default (and more effecient) version of SimGrid
2743 [contributed by Sékou Diakite, many thanks]
2746 * Introduce a listener thread in charge of receiving incoming
2747 messages from the network. It allows to overlap communication and
2748 computation but most notably, it removes some stupid deadlocks due
2749 to the fact that so far, a process could not send and receive at
2750 the same time. This made most non trivial communication schema
2752 * Convert the PIDs from long int to int to match the MSG ones (and
2753 linux ones too) [Mt]
2754 * New function: gras_agent_spawn() to launch a new process on
2755 current host. Only working in simulation for now. [Mt]
2756 * New function: gras_os_hostport() returning a constant form (ie,
2757 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2760 * Make the backtrace of exceptions more human readable [Mt]
2761 * New module: xbt/str [Mt]
2762 a ton of string utility functions (split, join, printf to a newly
2763 allocated buffer, trim, etc)
2764 * New module: xbt/hash [Mt]
2765 SHA1 hashing algorithm (more to come if needed)
2766 * New module: xbt/synchro [Mt]
2767 synchronization tools (mutex and conditions) working the same way
2768 in simulation and in real life (mainly useful for GRAS, but not
2770 * New module: xbt/queue [Mt]
2771 classical producer/consumer synchronization scheme
2772 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2773 (using the classical functions will get serialized) [Mt]
2774 * Make dictionary internal table dynamic. No need to specify its size
2775 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2777 * Make sure the log channels are organized as a tree under windows
2778 (because of ANSI C compatibility issue, any channel were child of
2779 root directly) [Mt].
2782 * Cleaned many thing in surf and fixed a few bugs [AL].
2783 * Add a nice command line configuration mechanism to compose models [AL].
2784 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2785 the previous one (KCCFLN05). It relies on something that looks like
2786 a max-min sharing mechanism but cannot be written as such. A new solver
2787 was thus designed [AL].
2788 * Add a new solver to lmm. Based on Lagrange optimization and
2789 gradient-based descent, it enables to efficiently maximise systems s.a
2791 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2793 This solver enables to propose two new network models for TCP Reno and
2794 TCP Vegas based on Low's work. These models still need to be fully
2795 tested though [Pedro Velho].
2798 * Bug fix in SD_simulate. Now the time bound given as argument is
2800 * Use the new parallel task model (ptask_L07) as default.
2801 * Use the SURF command line configuration mechanism.
2802 * 0-size tasks (for synchronization) should now work.
2804 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2806 SimGrid (3.2) stable; urgency=high
2810 We still experience issues on this platform, but we believe that at
2811 least MSG is usable.
2813 GRAS API BREAKAGE (for simplification purpose, sorry):
2814 * the gras_msgtype_by_name is not used anymore. Instead of
2815 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2816 you can write (and must)
2817 gras_msg_send(toserver, "request", &request);
2818 - If you still want to pass a gras_msgtype_t to the function (to cache
2819 the type and avoid the lookup time), use the gras_msg_send_() variant.
2820 - Impacted functions:
2821 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2822 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2823 * The callbacks are now expected to return 0 when everything went well
2824 (just like the main() function)
2826 GRAS new features and improvements:
2827 * New module mechanism where user code can use per process globals [Mt]
2828 This is similar to gras_userdata_*() functions, but for libraries. It
2829 factorize some code developped over and over in the examples and AMOK.
2830 It has still to be documented and used (only amok/peermanagement is
2832 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2833 loss when they were small enough to fit into the buffer and sent quickly
2834 enough so that they can all get received in one shoot.
2835 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2836 if not found. Use the *_or_null() variant for the old semantic.
2837 * In gras_msg_handle, do not discard messages without callback.
2838 They are probably messages to be explicitly awaited later (ie, proofs of
2839 mis-synchronization in userland since they are sent before being awaited)
2840 No big deal usually.
2841 * gras_socket_meas_send/recv: semantic changed!
2842 The numerical arguments used to be (1) the total amount of data to send
2843 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2844 messages. This was need for the fool willing to send more than MAXINT
2845 bytes on quite fat pipes.
2848 * Do really rename the hostmanagement module to peermanagement. [Mt]
2849 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2850 but this is rather new and this was documented in the module
2851 documentation (poor excuses, I admit)
2852 * Bandwidth measurement semantic changed! This follows the changes to
2853 gras_socket_meas_send/recv explained above.
2856 * A sequential mode has been added to the workstations. When a workstation
2857 is in sequential mode, it can execute only one task, and the other tasks
2858 are waiting in a FIFO. [Christophe Thiery]
2861 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2862 model for SIMDAG. [Christophe Thiery]
2863 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2864 equal to 0, which caused some bad behaviors in
2865 saturated_constraint_set_update. I now use a threshold mechanism like in
2869 * When running manually src/testall, you select specific units [Mt]
2870 testall is the result of our cunit mechanism, and should replace all
2871 the scripty thingy around since bash don't run easily on billware.
2873 * A mallocator system has been added. [Christophe Thiery]
2874 Mallocators allow you to recycle your unused objects instead of freeing them
2875 and allocating new ones.
2877 Documentation update:
2878 * FAQ reworking + New FAQs:
2879 - "Valgrind spits tons of errors!" [Mt]
2880 - "How to repport bugs" [Mt]
2881 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2882 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2883 - Communication time measurement within MSG [AL]
2884 - I experience weird communication times when I change the latency [AL]
2885 * GRAS tutorial [Mt]
2887 - an introduction to the framework and to the used communication model
2888 - an initiatic tour introducing the most proheminent features:
2890 . Lesson 0: Installing GRAS
2891 . Lesson 1: Setting up your own project
2892 o Part 2: Message passing
2893 . Lesson 2: Exchanging simple messages
2894 . Lesson 3: Passing arguments to the processes (in SG)
2895 . Lesson 4: Attaching callbacks to messages
2896 . Lesson 5: Using globals in processes
2897 . Lesson 6: Logging information properly
2898 . Lesson 7: Using internal timers
2899 . Lesson 8: Handling errors through exceptions
2900 . Lesson 9: Exchanging simple data
2901 . Lesson 10: Remote Procedure Calling (RPC)
2902 . Lesson 11: Explicitely waiting for messages
2903 . Recapping of message passing features in GRAS
2904 - A HOWTO section containing:
2905 o HOWTO design a GRAS application
2906 More are due, of course. They will come latter. In the meanwhile, you can
2907 check the examples which are still here.
2909 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2911 SimGrid (3.1) stable; urgency=high
2915 There was a stack corruption somewhere, visible only when optimizing
2916 with these versions. [Vince]
2919 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2920 module that enabled to play with Directed Acyclic Graphs. It is built
2921 directly on top of SURF and provides an API rather close to the old
2922 SG. Some old codes using SG are currently under rewrite to check that
2923 all needful functions are provided. [Christophe Thiery]
2926 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2927 extension of the classical CLM03 model that gracefully handles
2928 failures. This is now the default model for MSG and GRAS. It doesn't
2929 handle parallel tasks yet though. [AL]
2930 * Bug fix: Weights were not correctly set in the network part.
2931 WARNING: This may have resulted in incorrect results with simulations
2932 where there are more than one flow on a given link. [AL]
2935 * After a (long ?) discussion on simgrid-devel, we have decided that the
2936 convention we had on units was stupid. That is why it has been decided
2937 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2938 WARNING: This means that all previous platform files will not work as
2939 such with this version! A warning is issued to ask users to update
2941 A conversion script can be found in the contrib module of the CVS, under
2942 the name contrib/platform_generation/surfxml_update.pl [MQ]
2945 * Bug fix: Processes were started in reverse order, wrt deployment file.
2946 WARNING: if your code relies on this bug, please fix it. [AL]
2947 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2948 being executed on two different locations. [AL]
2949 * Bug fix: Failures are now better supported thanks to Derrick's tests
2950 (there was many failure situations I hadn't thought of and that weren't
2951 correctly handled). [AL]
2952 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2956 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2957 Exception occurring on server-side are propagated back to client (!).
2959 API CHANGE: the callback changed their prototype. Change:
2960 int my_handler(gras_socket_t expeditor, void *payload_data) {
2962 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2963 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2965 * New! function: gras_msg_handleall to deal with all messages arriving
2966 within a given period.
2967 * New! function: gras_socket_server_range to get a server socket in a
2968 range of port numbers (ease to avoid port number conflicts) [MQ]
2969 * New! gras processes display their backtrace when they get a SIGUSR1
2970 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2971 Sweet to debug RL processes [MQ]
2975 - Do not force experiment sizes to be expressed in kb, or it becomes
2976 impossible to measure the latency this way (needs one byte-long tests)
2977 WARNING: this changes the amok_bw_* function semantic. [MQ]
2978 - Implements the link saturation stuff. [MQ]
2979 * Peer management module:
2980 New! module factorizing code that we wrote over and over [MQ].
2983 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2984 - Test units are placed directly into the library code, they get extracted
2985 automatically and placed into the src/testall binary.
2986 - Convert most of the XBT tests to this system.
2987 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2988 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2989 live in the same C code. This required to change a little bit the API
2990 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2991 * New module: sparse graph structure with basic algorithms (this is work
2992 in progress and the API is not considered to be frozen yet). [AL]
2993 * Display more information on backtraces: source line & function names are
2994 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2995 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2996 * Reworked a little bit some #include statements to load only required
2997 headers. Some user code that relied on SimGrid to include stdlib or
2998 stdio may need to include it by themselves. [AL]
2999 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3001 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3002 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3003 but I guess that almost nobody used those parts. [MQ]
3005 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3007 SimGrid (3.0.1) stable; urgency=low
3010 * Unfortunately, I had missed 5 misnamed functions:
3011 xbt_fifo_item_t xbt_fifo_newitem(void);
3012 void xbt_fifo_freeitem(xbt_fifo_item_t);
3013 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3014 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3015 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3016 They're now deprecated. Please use their new versions:
3017 xbt_fifo_item_t xbt_fifo_new_item(void);
3018 void xbt_fifo_free_item(xbt_fifo_item_t);
3019 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3020 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3021 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3023 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3024 * Documentation: xbt_log module unmercifully reworked [MQ]
3025 * Bugfix: there was a problem with the ending of contexts with
3026 the pthread backend. It caused some weird deadlock or behavior
3027 depending on the pthread implementation. [AL]
3028 * Bugfix: get the exceptions raised in the simulator repport where
3029 and why they come from when they are not catched in time [AL, MQ]
3032 * Bugfix: Do repport the error when two non-connected hosts try to
3033 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3036 * Add additionnal checkings on communications. Assert that two
3037 communicating hosts are connected by a set of links... [AL]
3040 * Add additionnal checkings on channel values in communication [AL]
3041 * New: MSG_task_get_source to see on which host a task was generated [HC]
3042 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3043 the number of tasks waiting to be received on channel and sent
3045 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3046 waits for the first task coming from a given host.. [AL]
3048 GRAS new functionnalities: [MQ]
3049 * Enhance the parsing macro to allow the size of multidimentional objects
3050 to be given thru annotations.
3051 * New example (and documentation): Matrix Multiplication a la RPC
3052 (as when I was young!) and fix a bunch of bugs found on the way.
3054 GRAS performance improvements: [MQ]
3056 * Reduce the amount of cbps creation/destruction by making it static to
3057 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3060 * Change libdata to a set so that we can search for stuff by ID (and thus
3061 reduce the insane amount of dict lookups)
3064 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3065 RL communication on top of SG since it's so uneffective.
3066 It may also be used for inter-thread communication in RL, one day.
3067 * Use gras_datadesc_copy() to exchange messages on top of SG
3069 - improve message exchange performance on top of SG
3070 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3071 * Don't exchange on the network the size of the used part of buffer,
3072 instead, specify the possible buffer size to read().
3074 - reduces the amount of read/write calls (one pair per exchange)
3075 - reduces the amount of exchanged data (the size)
3076 - allows to retrieve all arrived data on receiver side, if we don't need
3077 it right now (subsequent read will peek the buffer)
3078 - allows the receiver to proceed with the begining of the stream before
3079 everything is arrived
3080 - make it possible to build an iov transport (using readv/writev)
3082 - take care of the data with non-stable storage (like stacked data),
3084 * If possible, TCP send uses vector I/O (when writev() is here)
3085 - Don't use it for receive since we send data sizes and data on the
3086 same stream, so we wouldn't be able to chain large amount of chunks
3087 before having to flush the stuff to read the size.
3088 * Rework the transport plugin mechanism to simplify it and reduce the
3089 amount of pointer dereferencement when searching for the right function
3092 * I guess that now, we do almost as few system calls as possible while
3093 doing as few data copy as possible.
3095 To improve it further, we could try to send all the sizes first and then
3096 all the data (to use iov on receiving size), but it's only a partial
3097 solution: when you have 2 dimensional data, the sizes of the second
3098 dimension is data of the first dimension, so you need 3 streams.
3100 I'm not sure the potential performance gains justify the coding burden.
3102 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3104 SimGrid (3.00) stable; urgency=high
3107 * New! Give the possibility to hijack the surf parser and thus bypass
3108 MSG_create_environment and MSG_launch_application. Have a look at
3109 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3111 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3113 SimGrid (2.96) unstable; urgency=low
3118 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3119 This deprecates the xbt_error_t mechanisms.
3120 It modifies (simplifies) all XBT and GRAS API.
3121 MSG API keeps unchanged (exceptions raised by XBT are catched from
3122 within MSG and masked with existing error handling facilities)
3125 * New! Add a FATPIPE model. [AL]
3126 * New! Add a parallel task model. [AL]
3127 * New! Add automatically a loopback interface (in the default
3128 network model) if none was precised.
3131 * Bugfix: MSG_process_resume now works with the current running process.
3133 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3134 * Modification of MSG_task_get_compute_duration. Once a task has been
3135 processed, the value returned by this function is now equal to 0. [AL]
3136 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3137 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3138 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3139 of a m_task. For now, MSG_task_cancel only works with computation
3141 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3142 of the processor (in Mflop/s) regardless of the current load on the
3144 * API Change: use proper naming convention for MSG_getClock and
3145 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3147 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3148 This function changes the priority of a computation task. This priority
3149 doesn't affect the transfer rate. A priority of 2 will make a task
3150 receive two times more cpu power than the other ones. This function
3151 has been added to suit the needs of Nguyen The Loc and hasn't been that
3152 much tested yet. So if it fails, please report it and send me your code.
3154 * API Change: removed all functions and types that were marked "deprecated"
3155 since many months. Renamed MSG_global_init_args to MSG_global_init.
3157 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3159 SimGrid (2.95) unstable; urgency=low
3162 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3163 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3164 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3167 * Bugfix: Allow absolute paths for platform description files [MQ]
3168 * Bugfix: do free the variables after use. Leads to drastic performance
3170 * Implement max_duration (ie, timeouts) on resources [AL]
3173 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3175 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3176 the needed support (timeouts) [AL]
3179 * Implement measurement sockets. You can now get the bandwidth between two
3180 hosts thanks to AMOK (see below). [MQ]
3181 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3182 dynar over the network (yeah) [MQ]
3183 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3184 * simple/static token-ring example. [Alexandre Colucci and MQ]
3185 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3186 in place before [MQ]
3189 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3190 type parsing. "short" and "long" modifiers were also ignored; other
3191 modifier (such as reference level) are still ignored. [MQ]
3192 * Embeed the buffer size within the buffer itself on SG. [MQ]
3193 That way, send() are atomic and cannot get intermixed anymore (at least
3194 the ones which are less than 100k; bigger messages still have the issue)
3195 * Array size pushed by the field, not by the field type (or each
3196 and every long int will push stuff to the cbps) [MQ]
3197 * use select() to sleep since it allows to portably sleep less than one
3200 GRAS (minor cleanups)
3201 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3204 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3205 * type category 'ignored' killed as it was never used and were difficult
3207 * whether a type can cycle or not is now a flag, leaving room for more
3208 flags (as "ignored", if we feel the need one day ;)
3209 * Rename raw sockets to measurement sockets since "raw" has another
3210 meaning in networking community.
3213 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3214 and offers features not belonging to GRAS but that most applications
3215 need. One day, it may be a set of plugins loadable at runtime.
3216 * New module: bandwidth
3217 bandwidth measurement between arbitrary nodes running this module. [MQ]
3219 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3221 SimGrid (2.94) unstable; urgency=low
3223 The first beta release of SimGrid 3 !
3227 * Update the main page and the FAQ. Adding references to gforge.
3230 * Add a gras_os_getpid function.
3233 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3234 * Extend the logs so that they also print PID, hostname, date, ... if
3236 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3237 and kill the old version which were in testsuite/
3238 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3242 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3243 for users. Instead, each programming environment (GRAS, MSG,...) use it
3244 when needed to provide such a feature to users.
3245 Don't shortcut the mechanism or you will also shortcut the virtualization
3246 you need on the simulator.
3250 * Cleanups in configury with regard to compile optimization/warning flags.
3251 Also add -fno-loop-optimize to any powerpc since it's the optimization
3252 killing gcc (< 3.4.0).
3253 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3254 needed by MSG examples complications
3255 * Borrow configury beautifications from PHP
3258 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3259 warning (thanks loris for stumbling into this one).
3260 * Bugfix: stop loading private headers (gras_config.h) from the public
3264 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3265 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3266 RL would imply a DNS resolver, which is *hard* to do in a portable way
3267 (and therefore delayed).
3268 * Implement a real timer mechanism and use it in timing macros. This allows
3269 to avoid rounding errors and get a 0.000005 sec precision in timing
3270 macros. While I was at it, various cleanups:
3271 - allow to declare more than one timed section per file (fix a stupid bug)
3272 - move some private declaration to the right place
3273 - merge conditional execution and timing macros into emulation module
3274 - document the module
3275 - make sure the module cleanups its mess on gras_exit
3276 * Documentation improvements:
3277 - (new) how to compile applications using GRAS
3278 - (new) emulation support (timing macros)
3280 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3282 SimGrid (2.93) unstable; urgency=low
3284 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3287 - Use Paje properly where used. Still to be sanitized properly.
3288 - Portability fix: Add an implementation of the contexts using pthread
3292 - Add xbt_procname(): returns the name of the current process.
3293 Use it to show the current process's name in all logging.
3295 - fix detection of older flex version and the reaction, since we do
3296 depend on modern ones (we use lex_destroy)
3297 - Better separation of SG and RL in the libs: remove all simulation code
3298 from libgras. As a result, this lib is now only 200k when stripped.
3299 Some of the xbt modules may also be duplicated (two sets and such) and
3300 should be cleaned/killed before SG3.
3301 - Insist on using xlC on AIX because of weird problems involving gcc there.
3302 - Cleanup the make remote stuff. This is now done by scripts
3303 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3304 mainly for our private use, but we're working on changing them to user
3307 - Bugfix: flush the socket on close only if there is some *output*.
3308 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3310 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3313 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3315 SimGrid (2.92) unstable; urgency=low
3317 Alpha 3 on the path to SimGrid 3
3321 - New! First try of benchmarking macros.
3322 - New! First try so that gras_stub_generator generate deployment and
3323 remote compilation helpers.
3325 - Bugfix: Initialization fix in msg_test.
3329 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3331 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3333 - API Change: gras_os_sleep to take the amount of seconds as a double.
3334 Accepting an int was error prone since it was the only location where
3335 seconds were coded as such. It leaded to damn rounding errors.
3336 - Bugfix: Hard to belive that timers ever worked before this.
3338 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3340 SimGrid (2.91) unstable; urgency=low
3342 Alpha 2 on the path to SimGrid 3
3346 - Bug fix in the lmm_solver.
3348 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3349 through the function MSG_paje_output.
3350 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3351 - It is possible to bound the rate of a communication in MSG with
3352 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3353 to write it in the changelog).
3354 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3359 - Various cleanups to the autotools stuff
3360 - Begin to move Gras examples to examples/gras/
3361 - Let make distcheck work again (yeah!)
3363 - documentation overhauled using doxygen.
3364 gtk-doc-tools is dead in SimGrid now.
3365 - Automatically extract all existing logging categories, and add the list
3366 to the documentation (long standing one, to say the less)
3368 - Cleanup the known architecture table. Reorder the entries to group what
3369 should be, and use a more consistent naming scheme.
3370 (some of the test dataset are still to be regenerated)
3371 - New! Allow library to register globals on each process just as userdata
3373 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3374 lookup time (for now).
3375 Use it in msg and trp.
3376 This cleans a lot the internals and helps enforcing privacy of the
3377 headers between the gras components.
3378 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3379 - Bugfix: gras_os_time was delirious in RL.
3380 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3382 - Reenable GRAS now that it works.
3384 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3386 SimGrid (2.90) unstable; urgency=low
3388 Alpha 1 on the path to SimGrid 3
3390 * It is a long time since the last release of SimGrid. I'm sorry about
3391 that but as I had told you, I was rewriting a lot of things. I apologize
3392 to those who had been reporting bugs to me and that I had not answered.
3393 If your bug is still in the new version, please tell me. Here is a
3394 summary of the main changes.
3396 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3397 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3398 portability, speed, and a lot more but you'll figure it out later.
3399 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3400 I wanted to release the new MSG as soon as possible and I have
3401 broken GRAS, which is the reason why, for now, only MSG is fully
3402 functional. A laconic description of these projects is available
3403 in the documentation.
3405 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3406 kernel optimized for our needs (called SURF but only the developers
3407 should use it). Hence, MSG is now roughly 30 times faster and I think
3408 that by rewriting a little bit MSG, I could event speed it up a little
3409 bit more. Beside the gain in speed, it is also much easier to encode a
3410 new platform model with SURF than it was with SG. More to come...
3412 * REVOLUTION 3: I have tried to change a little as possible the API of
3413 MSG but a few things really had to disappear. The main differences
3414 with the previous version are:
3415 1) no more m_links_t and the corresponding functions. Platforms are
3416 directly read from a XML description and cannot be hard-coded
3417 anymore. The same format is used for application deployment
3418 description. The new format is described in the documentation.
3419 Have a look in tools/platform_generation. There is a tiny script
3420 that converts from the old platform format to the new one. Concerning
3421 the application deployment format, parsing the old one is tricky.
3422 I think most of you should however be able to convert your files. If
3423 it is really an issue, I can write a C code that does the conversion.
3425 2) the toolbox tbx does not exist anymore. We now have a library
3426 with much more data-structures but without the hash-tables (we have
3427 dictionaries that are much faster).
3429 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3431 *****************************************************************************
3432 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3433 * don't feel like converting the oldies. (Mt) *
3434 *****************************************************************************
3437 Version 2.90: "the long awaited one"
3438 - Finished rewriting and debugging MSG. Rewrote the documentation.
3439 - disable GRAS for now since it needs to be ported to the newest SG
3442 - Finish the port to windows (using mingw32 for cross-compile)
3445 - Main loop and datastructures of SURF. A cpu resource object is
3446 functional. Surf can thus be used to create cpu's with variable
3447 performance on which you can execute some actions.
3449 2004-11-15 Martin Quinson
3450 - Port to ARM. Simply added the alignment and size descriptions. Should
3451 work, but the ARM machines are so slow that I didn't had the opportunity
3452 to 'make check' over there yet.
3454 2004-11-15 Arnaud Legrand
3455 - Trace manager now written. It uses a heap structure and is therefore
3456 expected to be efficient. It may however be speeded up (particularly
3457 when many events occur at the same date) by using red and black
3458 trees. One day maybe...
3459 - Max-min linear system solver written. It uses a sparse matrix
3460 structure taking advantage of its expected use. Most operations are
3461 O(1) and free/calloc are called as few as possible. The computation of
3462 the minimum could however be improved by using a red and black tree
3465 2004-11-03 Arnaud Legrand
3466 - Rename every gras_* function that was in xbt/ to its xbt_
3468 - Add a heap and a doubly-linked list to xbt
3469 - Added a dichotomy to the dictionaries. make check works as well before
3470 so I assume that the patch is correct. I do not know however if things
3471 run effectively faster than before now. :)
3473 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3474 SimGrid, and the version number is bumped to 2.x
3476 2004-10-29 Martin Quinson
3477 - Introduction of the remote errors.
3478 They are the result of a RMI/RPC on the remote machine.
3479 ErrCodes being scalar values, you can't get the host on which those
3480 errors did happen. Extending the error mechanism as in Gnome is possible.
3481 No idea yet whether it is a good idea.
3483 2004-10-28 Martin Quinson
3484 - Interface revolution: the Starred Structure Eradication.
3485 I used to do typedef struct {} toto_t; and then handle *toto_t.
3486 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3487 - ???_t is a valid type (builded with typedef)
3488 - s_toto_t is a structure (access to fields with .)
3489 - s_toto is a structure needing 'struct' keyword to be used
3490 - e_toto_t is an enum
3491 - toto_t is an 'object' (struct*)
3493 typedef struct s_toto {} s_toto_t, *toto_t;
3494 typedef enum {} e_toto_t;
3495 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3496 s_toto_t) is private.
3498 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3499 so that it can set the variable to NULL. It was so for dicts and sets,
3500 it changed for dynars.
3502 - Fix a bunch of memleaks in dict_remove
3503 - Fix a bug in sg/server_socket opening: it failed all the time.
3505 2004-10-07 Martin Quinson
3506 - Speed up dynar lookup operation a bit.
3508 gras_dynar_get is dead.
3510 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3511 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3512 (which gives you the address of the stored data).
3514 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3515 copy of the data using an affectation to do the job and not a memcpy.
3517 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3518 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3520 It does not really speedup the dynar test because they are
3521 setting elements all the time (and look them seldom). But the dict does
3522 far more lookup than setting.
3524 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3526 2004-10-05 Martin Quinson
3527 - Allow to (en/dis)able the cycle detection at run time.
3529 Whether we should check for cycle or not is now a property of each
3530 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3531 datadesc_cycle_unset allow to remove this property when previously set.
3533 Note that the cycle detection is off by default since it impacts the
3534 performance. Watch the data you feed GRAS with ;)
3536 This property is hereditary. Any element embedded in a structure having it
3537 set have it set for the time of this data exchange.
3539 You should set it both on sender and receiver side. If you don't set it on
3540 sender side, it will enter an endless loop. If you forget on receiver
3541 side, the cycles won't be recreated after communication.
3543 - Header reorganization.
3544 Kill gras_private.h, each submodule must load the headers it needs.
3546 2004-10-04 Martin Quinson
3547 - Interface revolution: do not try to survive to malloc failure.
3549 Now, gras_malloc and friends call gras_abort() on failure.
3550 As a conclusion, malloc_error is not a valid error anymore, and all
3551 functions for which it was the only gras_error_t return value are
3552 changed. They now return void, or there result directly.
3553 This simplify the API a lot.
3555 2004-09-29 Martin Quinson
3556 - Re-enable raw sockets.
3557 Created by gras_socket_{client,server}_ext;
3558 Used with gras_raw_{send,recv}
3561 It should allow to kill the last bits of gras first version soon.
3563 This is not completely satisfactory yet (duplicate code with
3564 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3567 - Simplify transport plugin (internal) interface by not passing any
3568 argument to _server and _client, but embedding them in the socket
3571 2004-09-28 Martin Quinson
3572 - Finish the port to AIX.
3573 autoconf was my problem (segfault within the malloc replacement
3574 function. No idea why)
3576 2004-09-16 Martin Quinson
3577 - Fix some size_t madness on 64bit architectures.
3579 2004-09-08 Martin Quinson
3580 - Reduce the number of system headers loaded, overload some more system
3581 calls (such as malloc to cast the result of the system one, and work
3583 - Fix and reintroduce the config support
3585 2004-09-07 Martin Quinson
3586 - Source code reorganization to allow Arnaud to surf all over there.
3587 - Allow to document the logging categories.
3588 - Remove all uppercase from logging categories and useless cleanup in names.
3590 2004-08-18 Martin Quinson
3591 Version 0.6.2 (protocol not changed; API changed)
3592 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3593 gras_error_t), and NULL when not found. Functions expecting a msgtype
3594 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3595 hopefully usefull message.
3596 - Portability to prehistoric sparcs again
3598 2004-08-17 Martin Quinson
3599 Version 0.6.1 (protocol not changed; ABI not changed)
3600 - prealloc some buffers to speed things up
3602 2004-08-11 Martin Quinson
3603 Version 0.6 (protocol not changed; ABI expended)
3604 - The parsing macro can deal with the references, provided that you add
3605 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3607 2004-08-09 Martin Quinson
3608 Version 0.5 (protocol not changed; ABI changed)
3609 - Allow to off turn the cycle detection code in data exchange at
3610 compilation time. It should be at run time, but I'm short of time (and
3611 the config stuff is still broken). That way, we keep dict out of the
3612 critical path, which is good because the performance is poor:
3613 - search not dichotomial yet
3614 - dynar give no way to access their content and memcpy everytime
3615 - In composed data description (struct, ref and so on), stop foolness of
3616 keeping the subtype's ID, but store the type itself. This keeps sets out
3617 of the critical path, which is good since they rely on dynar and
3618 dictionnaries. The only loose of that is that we cannot detect the
3619 redeclaration of a structure/union with another content (but I'm not sure
3620 the code detected well this error before anyway). We still can detect
3621 the redefinition discrepancy for the other types.
3622 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3623 it breaks the code because of type-punning used all over the place).
3624 This breaks on all non-gcc architectures (for now).
3626 All those changes (plus the buffer of last time) allow me to gain 2 order
3627 of magnitude on cruel tests consisting of 800000 array of integers on two
3628 level of a hierarchical structure (200 secondes -> 4 secondes)
3631 - the selector of reference must now return the type it points to, not
3632 the ID of this type.
3634 2004-08-06 Martin Quinson
3635 Version 0.4 (protocol changed; ABI not changed)
3636 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3637 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3638 - Implement a buffer transport, and use it by default (it relies on tcp in
3639 real life and on sg in simulation).
3640 That's a bit hackish since I had a new field to the structure to store
3641 its data without interfering with the subtype ones. Inheritance
3642 is tricky in C. And that's a kind of reverse inheritance with one class
3643 derivating two classes. Or maybe a game with java interfaces. Anyway,
3644 that's damn hard in C (at least).
3645 Moreover, I got tired while trying to ensure plugin separation and
3646 genericity in SG mode. MSG wants me to do weird things, so let's go for
3647 cruel hacks (temporarily of course ;).
3648 See comment in transport_private.h:71
3649 - do not include all the _interface headers in private but in the files
3650 which really need them (to cut the compilation time when they are
3653 2004-07-26 Martin Quinson
3654 Version 0.3 (protocol not changed; ABI changed)
3655 - Major overhault of the datadesc interface to simplify it:
3656 - shorted the function names:
3657 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3658 - add a trivial way to push/pop integers into the cbps without malloc.
3659 This allows to make really generic sub_type description, which simply
3660 pop their size of the stack.
3661 - add a function gras_datadesc_ref_pop_arr() which does what users want
3662 most of the time: Declare a dynamic array (which pops its size of the
3663 stack) and declare a reference to it. Poor name, but anyway.
3664 - kill the post-send callback, add a post-receive one
3666 2004-07-23 Martin Quinson
3667 Version 0.2 (protocol changed; ABI changed)
3668 - add some testing for cpbs in the test cases, and fix some more bugs.
3669 This invalidate again the little64 data file, since I cannot regenerate
3671 - remove an awfull optimization in the logging stuff, allowing me to:
3672 - understand it again
3673 - learn gcc how to check that the argument match the provided format
3674 - fix all errors revealed by gcc after that
3675 - internal keys of dict are not \0 terminated. Deal with it properly in
3676 loggings instead of segfaulting when the user want to see the logs :-/
3678 2004-07-22 Martin Quinson
3679 - Fix some stupid bug preventing cbps (callback postit) from working
3681 2004-07-21 Martin Quinson
3682 - Some documentation cleanups
3683 - remove the useless last argument of msgtype_declare
3684 - rename the Virtu functions to fit into the 'os' namespace
3685 - move headers src/include -> src/include/gras/ and stop fooling with
3686 gras -> . symbolic link
3687 - make distcheck is now successful
3689 2004-07-19 Martin Quinson
3691 - Build shared library also
3692 - Install html doc to the right location
3693 - stop removing maintainer files in make clean
3694 - build tests only on make check
3696 2004-07-13 Martin Quinson
3698 - No major issue in previous version => change versioning schema
3699 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3700 corresponding dataset.
3702 2004-07-11 Martin Quinson
3704 - Get it working with any kind of structure (we can compute the padding
3705 bytes remotely for all the architectures I have access to)
3706 - Implement the structure parsing macro (still not quite robust/complete)
3707 - Improvement to the remote testing toysuite
3709 2004-07-10 Martin Quinson
3710 [autoconf mechanism]
3711 - get ride of a bunch of deprecated macros
3712 - actually run the test for two-compliment, not only compile it :-/
3713 - test whether the structures get packed (and bail out if yes. Damn.
3714 Alignment is a serious matter)
3715 - test whether the structures get compacted (but respecting the alignment
3716 constraints of each types)
3717 - test whether the array fields of structures can straddle alignment boundaries
3719 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3720 (took me 2 hours to find that bug, looking at the wrong place)
3721 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3722 before sending/receiving and are used to compute the offsets of fields
3723 - Given that padding size depend even on compiler options, keep track of
3724 alignment and aligned_size only for the current architecture. Not a big
3725 deal since we send structure fields one after the other (seems
3727 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3728 program, let it work on linux/gcc/little32. portability todo.
3730 2004-07-08 Martin Quinson
3731 - import and improve remote compilation support from FAST
3732 - make sure make check works on half a dozen of machines out there
3734 2004-07-07 Martin Quinson
3735 Let's say it's version 0.0.3 ;)
3736 - Implement conversions (yuhu!)
3737 - Let it work on solaris (beside conversion, of course)
3738 - Stupid me, using rand() to generate the conversion datatests in not wise.
3740 2004-07-06 Martin Quinson
3741 - Let make dist work, since I'm gonna need it to compile on remote hosts
3742 - Let Tests/datadesc_usage write the architecture on which the file was
3743 generated as first byte.
3744 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3745 architecture descriptions.
3746 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3747 on those architectures.
3748 - Optimization: send/recv array of scalar in one shoot
3750 2004-07-05 Martin Quinson
3751 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3753 - Plug a whole bunch of memleaks
3754 - each process now have to call gras_{init,exit}. One day, their log
3755 settings will be separated
3756 - Continue the code factorisation between SG, RL and common in Transport.
3758 2004-07-04 Martin Quinson
3760 - Redistribution between SG and RL.
3761 We wanna have to accept in SG, so move accepted related parts of RL in
3762 the common part. (more precisely, the dynar of all known sockets is no
3763 more a static in transport.c, but part of the process_data)
3765 [gras_stub_generator]
3766 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3768 2004-07-03 Martin Quinson
3769 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3770 them all up in one shot)
3772 - Fix the ugly case of reference to dynamic array.
3773 - New (semi-public) function gras_datadesc_size to allow the messaging
3774 layer to malloc the needed space for the buffer.
3776 - gras_socket_close now expect the socket to close (and not its address to
3777 put NULL in it after it). This is because the socket passed to handlers
3778 is one of their argument (=> not writable).
3780 - propagate the interface cleanup from last week in datadesc, ie remove a
3781 superfluous level of indirection. User pass adress of variable
3782 containing data (both when sending and receiving), and not of a variable
3783 being a pointer to the data. Let's say that I like it better ;)
3784 The price for that is constructs like "int msg=*(int*)payload" in
3785 handlers, but it's a fine price, IMHO.
3787 - Let it work in RL (yuhu)
3789 2004-06-21 Martin Quinson
3791 - porting SG plugin and SG select to new standards (works almost).
3792 - plug memleaks and fix bugs around.
3795 - cleanup the prototype of data recv and force users to specify when they
3796 want to handle references to objects. Test case working even for cycles.
3797 - plug memleaks. Valgrind is perfectly ok with this.
3799 2004-06-12 Martin Quinson
3801 - cleanup the separation between plugin and main code in plugin creation
3803 2004-06-11 Martin Quinson
3805 - Reput hook for raw sockets, needed for BW experiments
3806 - kill a few lines of dead code
3807 [Data description] Interface cleanup
3808 - gras_datadesc_by_name returns the searched type or NULL.
3809 That way, no variable is needed to use a type desc once, which makes
3811 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3812 two parameters were strings (field name, type name), leading to
3814 [Dicos] Interface cleanup
3815 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3816 This is consistant with the dynar API.
3818 2004-04-21 Martin Quinson
3820 - Porting to new standards.
3822 - interface cleanup.
3823 There is no bag anymore, no need to take extra provision to mask the
3824 pointers behind "ID".
3825 Better splitup of functions between files create/exchange/convert.
3826 This is still a bit artificial since convert and receive are so
3827 interleaved, but anyway.
3829 - add a queued message list to procdata (the ones not matching criteria
3831 - factorize some more code between SG and RL wrt procdata
3833 - use gras_exit in example to track memleaks
3834 - get rid of gs_example now that GS is properly integrated into gras
3835 - update run_test to integrate the latest tests (datadesc)
3837 - rename WARNINGn macros to WARNn since it prooved error-prone
3839 2004-04-19 Martin Quinson
3841 - register init/exit functions within gras module mechanism
3842 - send/receive function.
3843 Convertion is not implemented, but short-cutted if not needed.
3844 struct/array elements are sent one by one (instead of block-wise), but
3845 nobody really cares (yet). Get a prototype before optimizing.
3846 - tests (using a file socket) for DD send/receive on:
3847 - base types: int, float
3848 - array: fixed size, string (ie ref to dynamic string)
3849 - structure: homogeneous, heterogeneous
3850 - chained list, graph with cycle
3851 Believe it or not, valgrind is not too unhappy with the results. The
3852 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3856 - Bugfix when using a filename explicitely (instead of '-')
3858 2004-04-09 Martin Quinson
3860 - factorize more code between RL and SG in socket creation
3861 - Complete the implementation and tests of:
3863 o file (only in RL, and mainly for debugging)
3865 I lost 3 days to design a portable address resolver, and then decided
3866 that the prototype mainly have to run on my box.
3867 Addressing portability too early may be like optimizing too early :-/
3869 - use gras_init in the Tests instead of the crappy parse_log_opt
3870 (the latter function is removed)
3871 [Conditional execution]
3872 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3873 [Code reorganisation]
3874 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3875 Build examples against the RL library, since there is no way to disable
3876 its creation for now.
3878 For information, the beginning of coding on GRAS was back in june
3879 2003. I guess that every line has been rewritten at least twice since