1 SimGrid (3.23) NOT RELEASED (Release Target: June 21. 2019, 15:54 UTC)
4 - Stop setting random seed with srand48() at initialization.
7 - New log appenders: stdout and stderr. Use stdout for xbt_help.
11 - SMPI now reports support of MPI3.1. This does not mean SMPI supports all MPI 3 calls, but it was already the case with 2.2
12 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
13 - MPI interface is now const correct for input parameters
15 Fixed bugs (GH=GitHub; FG=FramaGit):
16 - FG#10: Can not use MSG_process_set_data from SMPI any more
17 - FG#11: Auto-restart actors forget their on_exit behavior
18 - GH#133: Java : a process can run on a VM even if its host is off
19 - GH#326: Valgrind-detected error for join() when energy plugin is activated
21 ----------------------------------------------------------------------------
23 SimGrid (3.22) April 1. 2019
25 The Easter Christmas Release (this one is somewhat late).
28 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
29 - Not complete yet: asynchronous activities (amongst others) are still missing
30 - Still ongoing: the interface may change in the future. We need more
31 testers! Please report any glitches.
32 - No new project using Java should start now. Please switch to Python.
35 - Some of the internal cleanups may lead to speed improvements:
36 - The hard limitation on the amount of simulated actors with
37 Java+Mac was removed. Now, the available memory is the only limit.
38 - Our refcounting was tidyied, leading to 10% speedups in some cases.
39 - We are still working on making our code robust to the actor kills
40 and hosts' churn. Things are improving, but it's not perfect yet.
41 - Replaced our own code to display a backtrace (that was forking addr2line)
42 with the Boost.Stacktrace library.
43 You won't see your backtraces without this optional dependency.
44 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
45 - Stop setting random seed with srand() at initialization.
48 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
49 XML file version remains 4.2 since old files are still compatible.
52 - Process termination which was broken at version 3.21 has been repaired.
53 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
54 - Hide the examples into examples/deprecated. New users should use Python.
57 - Drop MSG_process_create_from_stdfunc() from the API.
58 This C++ function was a pimple in the C API, made necessary at some
59 point by the Java bindings. This is fixed now.
60 - Hide the examples into examples/deprecated. New users should use S4U.
61 - MSG_process_create and MSG_process_attach now crash if the host used
62 to run this process is off.
63 - Fix the protype of MSG_process_on_exit()
64 Now use (int,void*) callbacks instead of (void*,void*) ones.
65 The implementation was ways too messy to actually work, I guess.
68 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
69 too often and provide more accurate timings.
70 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
71 - MPI_Alltoallw support
72 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
73 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
74 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw, MPI_Ireduce,
75 MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter,
76 MPI_Ireduce_scatter_block, with fortran bindings.
77 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
78 support, MPI_Comm_get_name, MPI_Comm_set_name
79 - Basic implementation of generalized requests (SMPI doesn't
80 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
83 - Drop sg_cmdline. Please use xbt_cmdline instead.
84 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
85 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
86 OsSemaphore is implemented in a portable way with C++11 threads.
87 This should allow much more threads to be created at the same time,
88 allowing Mac and Java users to simulate many more actors.
89 - Implement the 'thread' factory with std::thread instead of xbt ones.
90 It is not possible to set the stack size with threads anymore, but
91 -fsplit-stack is the way to go nowadays when using threads.
92 - Drop the xbt_os_thread_t module (now unused)
93 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
94 - Drop xbt_str_join_array().
95 - Drop cunit, use Catch2 instead.
98 - Many cleanups in the kernel::activity namespace. This was long
99 overdue, and shall open the path to many future endeavors.
102 - #132: Java: a process can not shut down its own host
103 - #220: S4U: detached send causes exception if sender terminates soon after sending
104 - #261: Document the parameters of parallel execution's constructor
105 - #300: [s4u] BarrierPtr is missing
106 - #314: SMPI args internal cleanup
107 - #316: Fix a bug related to the CPU utilization of multi-core VM
108 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
109 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
110 - #325: Turning off a host has different behavior on sleeping actors and computing actors
112 ----------------------------------------------------------------------------
114 SimGrid (3.21) October 3. 2018
116 The Restarting Documentation (TRD) Release.
119 - Convert the user manual to Sphinx for improved usability.
120 Unstable version now hosted on https://simgrid.frama.io/simgrid
121 Many glitches remain; stay tuned.
122 - Restrict the Doxygen reference API to the installed header files.
123 (documenting all internals this way is overkill + doxygen scales badly)
124 - New tutorial on S4U (highly inspired from the old MSG one, but with
125 a git repository to fork as a starting point, and a docker image)
126 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
127 - The Developper manual is still to be converted (not compiled until then)
128 - Some parts are still missing in this conversion (such as the
129 platform and deployment sections) while others were blindly converted
130 and would need more love (such as the configuration flags).
131 Things will certainly further improve in the future releases.
134 - Fully reimplement auto_restart mechanism that was utterly broken
135 (fix #22, #131 and #281 that were all very related)
136 - Implement semaphores (fix #303 and #291)
137 - When creating an actor from a function and its parameters,
138 move-only parameters are not allowed anymore, as it would prevent
139 the actor to be restartable if its parameters are consumed on run.
140 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
141 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
142 IO operations that can be started, waited for, or canceled as a regular
144 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
148 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
149 - New host variable: 'core_count'
152 - Remove xbt_os_thread_specific features
153 - Remove portability wrapper to condition variables
154 - Remove xbt_os_thread_yield()
157 - MPICH collective selector now mimics MPICH 3.3b
158 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
159 - MPI_Init(NULL,NULL) is now allowed
160 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
161 - Command-line arguments should now be handled properly in Fortran simulations
162 - MPI Attributes and related callbacks should work in Fortran
163 - Apps using C + Fortran should now work
164 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
165 - Support MPI_Error_String
168 - Due to an internal bug, Msg.run() must now be your last line.
169 We hope to fix it in a future release, and we are sorry for the inconvenience.
172 - #22: Process autorestart seem to only work with CAS01 cpus
173 - #56: Feature request: dynamic SMPI replay
174 - #93: simgrid should not eat --help
175 - #111: How to change the loopback link characteristics?
176 - #116: Platform cluster radical
177 - #117: Platform cluster/host order
178 - #129: Replay_multiple on steroid
179 - #131: Java : autostart does not work
180 - #143: Setting a breakpoint at a given time
181 - #235: Network energy models should be integrated
182 - #264: Add ptask L07 resource tracing
183 - #271: Dynamic and manual replay of SMPI traces
184 - #279: Breakpoints option uses time deltas instead of absolute time
185 - #281: Daemonized actor no longer daemonized after an auto-restart
186 - #282: TIT convertor erase traces when using absolute path in the trace list file
187 - #285: segfault when a s4u actor kills itself with msg process tracing activated
188 - #286: Pajé traces are not exposing the number of cores
189 - #287: Command-line parsing should stop when encountering '--'
190 - #288: MPI_Init(NULL, NULL)
191 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
192 - #290: Method to check if a mailbox has a message ready to be consumed
193 - #291: [s4u] semaphores do not seem available yet
194 - #294: Alltoall collective implementation of mpich
195 - #295: Starting a Host as OFF is not supported
196 - #296: DTD too permissive
197 - #299: Add a s4u barrier example
199 ----------------------------------------------------------------------------
201 SimGrid (3.20) Released June 24. 2018
203 The proxy snake_case() release.
206 - The whole API is now using snake_case() instead of mixing it with camelCase.
207 Compatibility wrappers in place for one year.
208 - We took the opportunity to increase the consistency of the API.
209 Please report any remaining glitches.
210 - New function simgrid::s4u::Host::get_actor_count:
211 Returns the number of actors running on a specific host.
214 - SMPI is now tested with ~45 proxy apps from various sources, with none or
215 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
216 - Replay: The replay file has been re-written in C++.
217 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
218 supported. They are stored in the trace and used when replayed.
219 - Basic support of MPI_Cancel. Robustness not guaranteed.
220 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
221 (only for MPI_COMM_TYPE_SHARED).
222 - New option: smpi/privatize-libs, to add external shared libs to be privatized
223 by SMPI. They will be copied locally and loaded separately by each process.
224 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
225 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
226 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
227 with internal index optind with multiple processes. Only works if MPI_Init has
229 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
230 syntax without preprocessing tricks.
233 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
234 behavior tracing is no longer limited to MSG
235 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
236 this means that traces now need to be updated or re-obtained; both Irecv
237 and Isend lines in your traces have now in total 5 mandatory fields:
238 <rankid> <command> <to/from rankid> <tag> <size>
239 To update your traces, it suffices to add a 0 for the tag here.
240 - TIT now also supports waiting for a distinct request via MPI_Wait.
241 Wait/Test now wait for a specific request, not just the last one that was
242 issued. This unfortunately means another update, because we need to
243 identify which request you want to wait for. We do this via the
244 triplet (sender, receiver, tag), which needs to be added:
245 <rankid> <command> <sender> <receiver> <tag>
246 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
247 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
250 - The deprecation of MSG is ongoing (but this should not impact you).
251 Many MSG functions are now simple wrappers on the C API of S4U. If
252 you wish to convert your code to S4U, find the S4U counterparts of
253 your MSG calls in src/msg/msg_legacy.cpp.
254 - MSG can still be used, but won't evolve anymore.
257 - Allow to run the Link energy plugin from the command line with
258 --cfg=plugin:link_energy
259 - Rename Energy plugin into host_energy
260 - Rename Load plugin into host_load
263 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
264 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
265 - Improve performance when handling timeouts of simix synchros.
268 - Config: the C API is now deprecated (will be removed in 3.23), and
269 the C++ API has been slightly improved.
272 - Fix several build issues on OSX.
273 - Move simgrid_config.h to simgrid/config.h (old header still working)
276 - #143: Setting a breakpoint at a given time
277 - #150: Inconsistent event names in SMPI replay
278 - #258: daemonized actors hang after all non-daemonized actors have completed
279 - #267: Linker error on unit_tmgr
280 - #269: SMPI: tracing of MPI_Wait/all/any broken
281 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
283 ----------------------------------------------------------------------------
285 SimGrid (3.19.1) Released March 22 2018
287 The Fixed ABI Release.
289 As you may know, we are currently refactoring SimGrid in deep.
290 Upcoming SimGrid4 will be really different from SimGrid3: modular,
291 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
293 Our goal is to smooth this transition, with backward compatibility
294 and automatic update paths, while still progressing toward SimGrid4.
296 SimGrid remains open during works: The last pure SimGrid3 release was
297 v3.12 while all subsequent versions are usable alpha versions of
298 SimGrid4: Existing interfaces remain unchanged, but the new S4U
299 interface is budding and the internals are deeply reorganized.
301 Since 2015, we work hard to reduce the changes to public APIs. When
302 we need to rename a public library symbol in S4U, we let your
303 compiler issue an explicative warning when you use the deprecated
304 function. These messages remain for four releases, i.e. for one full
305 year, before turning into an error. Starting with v3.15, your can
306 also adapt to API changes with the SIMGRID_VERSION macro, that is
307 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
309 Starting with this v3.19.1, our commitment to reduce the changes to
310 the public interfaces is extended from the API to the ABI: a program
311 using only MSG or SimDag and compiled against a given version of
312 simgrid can probably be used with a later version of SimGrid without
313 recompilation. We will do our best... but don't expect too much of
314 it, that's a really difficult goal during such profund refactoring.
316 The difference between v3.19 and v3.19.1 is that the former was
317 accidentally breaking the ABI of MSG, while the later is restoring
320 S4U and kernel APIs will still evolve until SimGrid4, with one-year
321 deprecation warnings as currently. In fact, cleaning up these
322 interfaces and converting them to snake_case() is one release goal of
323 v3.20. But don't worry, we are working to smooth this upgrade path.
325 In summary, new projects should start with S4U to benefit of the
326 future, but old MSG projects should still be usable with no change.
328 ----------------------------------------------------------------------------
330 SimGrid (3.19) Released March 20 2018
332 The Moscovitly-cold Spring release.
334 SMPI user-visible improvements:
335 - Support of MPI_Type_create_subarray
336 - Unimplemented MPI calls now abort (instead of issuing a warning)
337 Some harmless calls around errhandlers still only issue a warning.
338 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
339 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
340 is now possible (mmap privatization was refactored to this end).
341 Adding new SMPI instances should also be technically doable now.
343 SMPI internal cleanups:
344 - Removed index notion from SMPI -- no more "getPid() - 1"!
345 In tracing files, actors are called rank-<PROCESS-ID> but these are
346 NOT the real ranks (tracing several communicators is not supported yet).
347 - Deprecate smpi_process_index() (will be removed in 3.22)
348 - Better testing on classical proxy apps, through an external project:
349 https://github.com/simgrid/SMPI-proxy-apps/
352 - Use a graphical TOC to make it easier to find the documentation you need
353 - Revamp the MSG tutorial
356 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
357 - Remove parameter of MSG_process_killall().
358 Resetting the PID was bogus anyway (several actors could have the same PID).
361 - Execution->setHost() can be called after start(), to migrate it.
362 - Comm::test_any() is now implemented.
363 - s4u::Actor now has onCreation() and onDestruction() signals.
364 - Install some kernel header files for the users' plugins and more.
365 Warning, their API is really not stable yet.
366 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
367 include/simgrid/kernel/routing/*Zone.hpp
370 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
372 Build System and other cleanups:
373 - Remove unused run-time parameter "tracing/onelink-only".
374 - SimGrid now works with intel compilers (MC still buggy)
375 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
376 - tesh kills subprocesses on timeout (but not on windows)
379 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
382 - #194: Feature request: simgrid::s4u::Comm::test_any()
383 - #245: migrating an actor does not migrate its execution
384 - #253: Feature Request: expose clusters as objects
385 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
386 - #255: Tesh broken on Windows
387 - #256: Modernize FindSimGrid.cmake
388 - #257: Fix (ab)use of CMake install
390 ----------------------------------------------------------------------------
392 SimGrid (3.18) Released December 24 2017
394 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
396 MSG IS NOW DEPRECATED. Please use S4U for new projects.
397 - Support for MSG should not be removed from SimGrid before 2020,
398 but future evolutions will be limited to the integration of
399 user-provided patches.
402 - s4u::onDeadlock() signal, run before stopping the simulation.
403 - s4u::Exec: asynchronous executions (abstraction of background
405 - s4u::Host->getLoad() returns the achieved speed in flops/s
408 - Link::name() is deprecated and will be removed in v3.21, use
409 Link::getCname() instead.
410 - Mailbox::getName() changed to return a std::string, use
411 Mailbox::getCname() instead to get a char*.
412 - Storage::getName() changed to return a std::string, use
413 Storage::getCname() instead to get a char*.
414 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
415 to have both allocation and deallocation of the map in user space.
418 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
419 The amount of flops remaining to do is a vector, not a scalar.
420 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
421 name implies on both sequential and parallel tasks.
422 - Both changes fix GitHub's #223 using PR #237 as a basis.
423 Thanks Michael Mercier.
424 - Most examples were converted to S4U and hidden elsewhere at tests.
425 Rationale: we still want MSG to work; we want newcomers to use S4U.
428 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
429 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
432 - Switch to the faster dlopen privatization mechanism by default
433 - Documentation improvments
436 - Live migration is getting moved to a plugin. Dirty page tracking is
437 the first part of this plugin. This imply that VM migration is now
438 only possible if one this function is called:
439 - C/MSG: MSG_vm_live_migration_plugin_init()
440 - C/C++: sg_vm_live_migration_plugin_init()
441 - Java: Msg.liveMigrationInit()
442 For C and C++, "simgrid/plugins/live_migration.h" also has to be
446 - Define class simgrid::xbt::Path to manage file names.
447 - Removed unused functions:
448 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
449 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
450 xbt_graph_export_graphviz()
451 - xbt/str.h: xbt_str_join()
452 - Remove unused datatypes:
453 - xbt/heap.h: use std::priority_queue or boost::heap instead
454 - xbt/swag.h: use boost::intrusive::list instead
457 - New link_energy plugin for the consumption of the links.
458 - All of the operations on files and storage contents have been
459 packaged into a plugin (src/plugins/file_system). The current
460 public interface can be found in
461 include/simgrid/plugins/file_system.h
462 To use these functions you now have to initialize the plugin by
463 calling MSG_storage_file_system_init() just after calling
464 MSG_init() or sg_storage_file_system_init() just after creating
468 - Remove the undocumented/untested tag <include>
471 - Remove viva specific tracing as the tool is no longer maintained
474 - #248: Kill of finished processes leads to segfault
475 - #240: xbt_cond_wait_timeout should gracefully return for C apps
476 - #239: please implement signal s4u::onDeadlock()
477 - #230: segfaults when exit() without run()
478 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
479 - #223: MSG_task_get_flops_amount() not working with parallel tasks
480 - #222: Actor::kill() doesn't really kill and segfaults
481 - #221: odd LMM warning when killing an actor
482 - #120: Memory leak when the processes are forcefully killed
484 ----------------------------------------------------------------------------
486 SimGrid (3.17) Released October 8 2017
488 The Drained Leaks release: (almost) no known leaks despite the tests.
490 Even very long simulations will preserve your memory: our testsuite
491 stresses SimGrid for over 45mn on fast machines for a coverage of
492 over 80%, and there is only one single known leak, of about 4kb.
496 - Comm.detach(): start and forget about asynchronous emission. A cleanup
497 handler may be given to free resources if the comm cannot be completed.
498 - this_actor::send(mailbox) is now mailbox->put()
499 - New: simgrid::s4u::Comm::wait_all()
500 - New: Host.execute() for remote executions.
503 - Unused option network/sender-gap is removed.
506 - Deprecate MSG_task_isend_with_matching(): This unused feature
507 really complicates our internals. Will be removed in v3.20.
510 - Improved context termination. It is now possible to free resources, even
511 when a process is forcibly killed.
512 - ContextBoost: add support for Boost versions above 1.61.
515 - Bring back run-time option --cfg=exception/cutpath to remove exception
517 - Removed unused functions:
518 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
519 xbt_str_rtrim(), xbt_str_trim().
520 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
523 - Removed header files obsolete since SimGrid 3.12:
524 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
526 - #3: SD_exit should be made optional
527 - #120: Memory leak when the processes are forcefully killed
528 - #159: Threading test regression in Actor refcounting
529 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
530 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
531 - #186: Actor::killAll() segfaults if some process is blocked on wait()
532 - #191: VM migration and pstate
533 - #192: Updating the energy consumptions of all hosts crashes with VMs
534 - #195: All actors have PID=0 in the logs
535 - #204: Sometimes segfault with thread contexts and mmap privatization
536 - #222: Actor::kill() doesn't really kill and segfaults
537 - #225: Actor::kill() doesn't really kill when victims are doing a join()
539 ----------------------------------------------------------------------------
541 SimGrid (3.16) Released June 22. 2017.
543 The Blooming Spring Release: developments are budding.
546 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
547 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
548 - Java JAR file should be finally fully working on Mac OSX too.
550 XML platforms: Switch to platform v4.1 format.
551 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
552 - <zone> can be used as a synonym for the now deprecated <as>
553 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
554 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
555 - <actor> can be used as a synonym for the now deprecated <process>
556 - state_file and avail_file periodicity is now easier to express
557 (check the documentation) the old behavior should still work.
558 * Storage: not backward compatible, but it seems that nobody ever used it
559 - Remove attribute 'content_type' of <storage_type>: was never used
560 - Make attribute 'model' of <storage_type> optional: for future usage
561 - Remove Bconnection model property: was never (in)validated. Replaced by
562 the maximum of the read and write bandwidth as a resource constraint.
565 - New and Backwards Compatibility break:
566 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
567 When one wants to get the list of tasks whose states have changed during a
568 simulation round, s/he has to allocate and free a dynar and use it as
569 argument to this function. The former SD_simulate (double how_long)
573 - Allow multicore VMs, with the correct sharing (unless you overcommit)
574 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
577 - New function to update the consumption of all hosts at once.
578 - Fix the model for multi-core hosts, linear in the amount of busy
579 cores with an abnormality for pIdle (see paper).
580 BUG: Still not satisfactory for ptask on mono-cores.
583 - The netzone are now available from the MSG API.
584 The old names still work, but are now deprecated.
587 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
588 - New option: smpi/keep-temps to not cleanup temp files
589 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
590 mallocs mode is used, allows to change the size of the fake file used
591 (default 1MB), to potentially limit the number of mappings for large runs.
592 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
593 - Fortran ifort and flang compilers support
594 - New RMA calls supported (experimental):
595 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
596 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
597 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
598 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
599 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
600 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
603 - New: this_actor::isMaestro() returns whether we are in kernel mode.
604 - Behind the scene, ActivityImpl are now automatically refcounted.
605 Removing this nasty bottleneck will greatly simplify our internals.
608 - Replay: New function xbt_replay_action_get():
609 Retrieve the function previously associated to an event type.
610 - DROPPED FUNCTION: xbt_str_varsubst()
611 - DROPPED MODULE: strbuff. We don't need it anymore.
612 - DROPPED MODULE: matrix. We don't need it anymore.
613 - DROPPED MODULE: lib. We don't need it anymore.
615 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
617 SimGrid (3.15) Released March 22. 2017
619 The Spring Release: continuous integration servers become green
620 We fixed even the transient bugs on all target architectures:
621 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
622 Windows, FreeBSD, NetBSD.
624 - Rename NetCards to NetPoints.
625 This was intended to help NS3 users, but that's not a netcard.
626 That's a point in the routing algorithm, let's avoid wrong simplifications.
629 - New: MSG_process_yield(). Stop and yield to other processes.
630 - New: MSG_process_daemon(). Daemon processes are automatically killed
631 when the last non-daemon process terminates
632 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
633 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
634 to make room for the upcoming network energy plugin.
635 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
638 - Ensure that an actor can kill itself with Process::exit()
639 - Kill the obscure NativeException. Nobody want to survive the issues
640 it denotes, so use JniException that is a RuntimeException (not to
641 be caught explicitly).
642 - Partial bug fix in initialization. SimGrid flags on command line were
643 consumed at C level but stayed in the original Java String[] args.
644 This could mess users' args[i] if SG flags were not put at the end of
646 The SimGrid flags are now removed from the Java arguments. However,
647 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
648 if args.length is greater than the number of YOUR OWN ARGUMENTS.
649 It might be if you have --log or --cfg flags in the command line.
650 - Fix numerous memleaks all around the place. In particular, around VMs.
654 - s4u::onPlatformCreated: right before the simulation starts
655 - s4u::onSimulationEnd: right after the main simulation loop
656 - s4u::onTimeAdvance: right after a clock change (time is discrete)
657 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
658 event from the availability_file changes the avail speed.
659 - Links are now usable from s4u
660 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
661 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
662 - New examples: The conversion of MSG examples to S4U has begun
663 - Actors: create, kill, migration, and suspend
664 - Applications: master-worker and token-ring
665 - Action replay: communications and storage
666 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
669 - Backwards Compatibility breaks
670 - The SD_route_get_list and SD_route_get_size functions have been removed.
671 They are replaced by the unique
672 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
673 The route from src to dst is built in the links dynar whose size can
674 be obtained with xbt_dynar_length.
675 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
676 sg_host_route_bandwidth and sg_host_route_latency.
677 Macros ensure the backwards compatibility, but you should fix your code
680 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
681 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
682 - Bump our claim of support from MPI 1.1 to MPI 2.2.
683 We don't support 100% of it, but it should be enough. Contact us if not.
684 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
685 naive and probably inaccurate)
686 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
687 mapped onto a unique small file using some system magic.
688 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
689 - Fortran: cleanups, fixes, support of user-added operations
690 - RMA: MPI_Accumulate are applied in correct order.
691 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
692 - Better support for MPI_IN_PLACE
693 - Support for MPI_Win attrs and keyvals.
694 - Support MPI_Comm_group_create, MPI_Type_size_x
695 - MPI ops should be properly applied to all allowed types
696 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
697 algorithm by a non failing one from mpich.
700 - Kill the fifo data container: we don't use it anymore.
703 - A <cluster> can now be created with different speed values to represent pstates.
704 see examples/platforms/energy_cluster.xml for an example.
706 SimGrid (3.14.159) Released December 28. 2016
708 The previous release was only a crude approximation of a Pi release;
709 this one should be a bit better.
711 - Revert a last minute change that broke on Mac OSX.
712 - Fix the changelog and NEWS files.
713 - Improve the documentation, in particular of the routing module.
714 - Integrate some patches flying around in Debian.
716 SimGrid (3.14) Released December 24. 2016
719 * The whole documentation was reworked and reorganized. There is
720 still much room for improvement here, but we're on it.
723 * We now need python3 for our testing infrastructure.
724 * Model checking needs libevent
727 * Now works on FreeBSD
730 * Privatization now works on FreeBSD
732 * Privatization is now activated by default in smpirun
733 You can turn it off with -no-privatize if something goes wrong.
735 * Call-location tracing for SMPI.
736 You can add the exact location (filename / linenumber) of an MPI call to
737 your trace files and slow down or speed up the simulation between two
738 consecutive calls by using an adjustment file (see the documentation).
740 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
741 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
742 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
743 values were ignored in some cases. The timings of these functions can now
744 be significantly different.
746 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
747 smpi/running-power is renamed to smpi/host-speed
749 * smpi/grow-injected-times option to enable or disable multiplication of the
750 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
751 make simulation less precise (but also much faster).
753 * smpirun script should be (much) faster for large deployments.
755 * SMPI tracing: fixed issue with poor matching of send/receives.
757 * Replay: Fix broken waitall
759 New functions and features
760 * MSG_parallel_task_execute_with_timeout, to timeout computations.
762 Dropped / renamed functions and features
763 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
764 - MSG_mailbox_is_empty() -> Mailbox::empty()
765 - MSG_mailbox_front() -> Mailbox::front()
766 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
767 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
768 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
769 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
770 * Don't pass the free_f parameter to property related functions:
771 - MSG_host_set_property_value()
772 - MSG_as_router_set_property_value()
773 - MSG_storage_set_property_value()
774 * VM properties. Since msg_vm_t are msg_host_t, just use
775 MSG_host_get_property_value() and friends
776 * VM I/O related things:
777 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
778 - Unimplemented save/restore methods
779 * MSG_as_router_get_property_value() was redundant with
780 MSG_environment_as_get_property_value().
781 - Removed MSG_as_router_*propert*() functions
782 - Added MSG_environment_as_set_property_value() for consistency
783 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
784 Well, they are still there for now, but deprecated with a warning.
785 Please switch to xbt_dict_new_homogeneous() before this is removed
787 * Task affinity. Its intended behavior (that was very badly tested
788 and probably not really working) was deceiving what most users
789 would have hoped here.
790 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
791 * Option network/coordinates is now useless and should be dropped.
794 * Added option storage/max_file_descriptors to allow more than 1024 files opened
797 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
798 * Added Dragonfly topology support
801 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
802 EiBps, EBps, Eibps, Ebps for bandwidth.
803 They may become useful to some lucky ones.
806 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
807 * ex/app_tokenring: new example, very similar to the MSG Token Ring
808 * ex/async_waitAll: new example, on asynchronous communications
811 * Memory usage should be decreased for simulations with a large number
812 of processes. This also helps for SMPI.
814 SimGrid (3.13) stable; urgency=low
816 The Half Release, a.k.a. the Zealous Easter Trim.
818 - We removed half of the lines, that were mostly experimental cruft.
819 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
820 (not counting blanks and comments -- according to openhub.net)
821 - The internals are now compiled in C++ (and will soon be clean C++)
822 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
823 59 klines of C, etc. We added only 29 klines of C++ in replacement.
825 * Backwards Compatibility breaks
826 - Removed Lua simulation bindings (switch to C or Java for that).
827 Lua can still be used to describe platforms
828 - Removed Java kernel plug-ins.
829 Will be reintroduced after the ongoing major internals reorg.
831 - the following functions were removed.
832 They were too specific and should be reimplemented in a generic
833 way, with filter function.
834 - MSG_task_listen_from_host
835 - MSG_mailbox_get_count_host_waiting_tasks
836 - MSG_mailbox_put_with_timeout was removed.
837 Please use MSG_task_send_with_timeout instead.
839 - the SD_application_reinit function was removed. It has been a noop for a while.
840 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
841 replaced by a more flexible mechanism.
842 - The following functions thus do not exist anymore
843 - SD_workstation_get_access_mode
844 - SD_workstation_set_access_mode
845 - SD_workstation_get_current_task
846 - Basic estimation functions have been removed but can easily be replaced
847 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
848 - SD_workstation_get_computation_time => amount / sg_host_speed()
850 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
851 Use VM.getSpeed()*load/100 for the legacy behavior.
853 - option enable_tracing was removed. It was not doing anything for a while.
854 - In the ModelChecker:
855 - the model-checker now ptraces the model-checked process which means
856 you cannot use a debugger on the latter anymore (we might make this
857 optional in the feature);
858 - removed soft-dirty page tracking;
859 - remove model-checked side snapshot management,
860 MC_snapshot() and MC_compare_snapshot();
861 - keep the MC_cut() function as a stub (it was not really working
862 in the previous release).
865 * All options are consistently in kebab-case. Old names are kept as alias.
868 * Switch to platform v4 format.
869 - Rename from 'power' to 'speed' the attributes describing the amount of
870 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
871 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
872 - In <host> and <link>, attributes availability and state are gone.
873 It was redundant with state and availability traces, and with peak values.
874 - In <cluster>, attributes availability_file and state_file are gone.
875 It was too complex and unused.
876 - Kill <gpu>. Was not doing anything.
877 - The DOCTYPE points to the right URL:
878 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
879 (the file at this address now documents the changelog since its v1)
880 - A warning is emitted for unit-less values (they are still accepted).
881 - speed. Default: 'f' or 'flops'. Also defined:
882 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
883 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
884 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
885 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
886 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
887 - latency. Default: 's' second. Also defined:
888 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
890 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
891 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
894 * s4u::Host is now the preferred public interface to the Host features.
895 sg_host_* functions are C bindings to the exact same behavior
896 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
899 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
900 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
901 They were deprecated since a few years, and probably did not even compile anymore.
904 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
905 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
906 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
907 encouraged to update their codes. A list of the modified functions can be found at the end of
908 include/simgrid/simdag.h
911 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
912 the lambda or closure passed as a parameter will run in kernel mode.
913 All the callback functions should be rewritten to that interface at some point.
916 * Reorganizing and cleaning the internals all around the place.
919 * Remove old default barrier/bcast buggy algorithms (see #18407)
920 * Various bug fixes to handle more codes
921 * Remove the need for the --foreground option of smpirun (it is still
922 accepted for backward compatibility).
925 * Kill the setset data container: MC don't use it anymore.
926 * Kill the queue data container: it made more sense with GRAS.
927 * Kill the xbt_peer_t data type: it's useless without GRAS.
928 * Kill rm_cb feature of config sets: it was never useful.
929 * Kill graphxml parsing feature. It was not used.
930 * Kill the deprecated code protected by XBT_USE_DEPRECATED
932 - xbt_dynar_sort_strings(), when the content is char*
933 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
934 They throw exceptions on invalid input;
935 * C++ support for declaring CLI flags (simgrid::config::Flag);
936 * class for abstracting different signal backends (simgrid::xbt::signal).
937 with no external dependencies (we need very simple signals).
940 * refactoring and cleanup of the code;
941 * ongoing process to cleanly separate the model-checking algorithms
942 from the code model-checking support.
944 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
946 SimGrid (3.12) stable; urgency=low
948 The Facelift Release.
951 * Require g++ v4.7 at least to not speak prehistorical C++.
952 * Require Boost 1.48 (for signal2 component).
953 * Java must be version 7 at least when activated.
954 * Builds on Windows again (including Java bindings).
955 * Tracing is now always enabled (no way to turn it out)
956 * Remove GTNetS. It was not working anyway.
957 * Various cleanups in the cmake scripts.
958 * Move headers around to sort them out on installed systems:
959 - instr/instr.h -> simgrid/instr.h
960 - instr/jedule/* -> simgrid/jedule
961 - simdag/datatypes.h was removed
962 - simdag/simdag.h -> simgrid/simdag.h
963 - msg/datatypes.h was removed
964 - msg/msg.h -> simgrid/msg.h
967 * Interface improvement:
968 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
969 - Sanitize the interface in MSG_task_ module:
970 - Merge two functions that were close enough but misleading:
971 set_compute_duration(t) -> set_flops_amount(t)
972 get_remaining_computation(t) -> get_flops_amount(t)
973 - set_data_size(t) -> set_bytes_amount(t)
974 get_data_size(t) -> get_bytes_amount(t)
975 - Massive cleanups in the functions related to the energy
976 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
977 - New: MSG_host_get_pstate()
979 - msg/energy/onoff: switching hosts on and off
981 * Interface improvement:
982 - Rename Host.isAvail() to Host.isOn()
983 - Rename Process.currentProcess() to Process.getCurrentProcess()
984 - Rename Task.setDataSize() to Task.setBytesAmount()
985 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
987 - #18874: Actually allows the GC to reclaim tasks
991 - SIMIX_process_throw: raises an exception in a remote process
992 * Refactoring: Separate sub-modules
993 - libsmx: the public interface, as libc in a real system
994 - popping: the strange dance that converts a user request into a kernel handling
995 - smx_context_*: the virtualization mechanisms that embed the user code
996 - smx_*: the handling of each simcalls
998 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
999 * Rename smx_action_t into smx_synchro_t, making explicit that these
1000 things are used to synchronize processes with their environment.
1001 For example, a communication is a sort of synchronization involving
1002 the communicating processes (that may block until the exchange) and
1003 the platform. The same can be said from computations, etc.
1005 - #18888: segfault when a process ends before its kill_time
1009 - 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.
1010 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1011 - MPI_Type_set_name, MPI_Type_get_name
1012 - MPI_*_c2f and MPI_*_f2c functions
1013 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1014 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1015 - Activate a lot of new tests from the mpich 3 testsuite
1017 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1018 - InfiniBand network model added: Based on the works of Jerome Vienne
1019 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1020 - When smpi/display_timing is set, also display global simulation time and application times
1021 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1022 * Collective communications
1023 - 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
1024 - 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/).
1025 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1027 - Replay now uses algorithms from wanted collective selector
1028 - Replay can be used with SMP-aware algorithms
1029 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1030 - 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.
1032 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1033 - Correctly use loopback on fat-tree clusters
1034 - Asynchronous small messages shouldn't trigger deadlocks anymore
1035 * Energy/DVFS cleanup and improvement
1036 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1037 - new: smpi_get_host_pstate()
1040 - "Full" network optimization flag was broken since Surf++
1041 - Better handling of precision flags in maxmin
1042 - Fix bug causing sometimes "Impossible" errors
1043 - Properly pass cluster properties to included hosts
1044 * Improvement of the Energy plugin.
1045 - Always update the consumption before returning that value
1046 - New property: watt_off to denote the dissipation when the host is off
1047 - New functions getWattMinAt and getWattMaxAt to retrieve the
1048 dissipation of pstates that we are not currently at.
1049 * Java: class NetworkLink renamed to Link
1050 * New function: simcall_process_get_kill_time()
1051 * Massive rename s/workstation/host/
1052 - That's intrusive, but that's good for the project consistency. Sorry.
1053 - Change config option "workstation/model" into "host/model"
1056 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1057 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1058 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1059 - xbt_dirname and xbt_basename for non-POSIX systems
1061 * The model checker now runs as a separate process.
1062 * The model checker runs must now be launched with the new simgrid-mc program.
1063 * Record/Replay: the MC can display a textual representation of a path in the
1064 execution graph. It can then be replayed outside of the model checker.
1066 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1068 SimGrid (3.11) stable; urgency=low
1073 * Normalizing pointers addresses tool for better diff between logs
1076 * Add cloud examples using new VMs
1077 - examples/msg/cloud/two_tasks_vm.tesh
1078 - examples/msg/cloud/simple_vm.tesh
1079 - examples/java/cloud/cloud.tesh
1080 - examples/java/cloud/migration/migration.tesh
1081 * Add java surf examples:
1082 - examples/java/surfPlugin/surf_plugin.tesh
1083 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1084 - examples/java/surfCpuModel/surf_cpu_model.tesh
1085 * Add SMPI+MSG example:
1086 - examples/smpi/smpi_msg_masterslave/
1091 - msg task destroy cancel test
1092 - msg_host on/off test
1093 * Move all tests in testsuite to teshsuite (adding tesh files)
1094 * Restructure teshsuites
1095 - one folder for each kind of test
1096 * Restructure AddTests.cmake
1098 - structure the order of tests (with sections)
1101 * Add virtual machine
1102 - creation of a VM on a PM
1103 - migration of a VM from a PM to another PM
1105 - MSG_process_join(msg_process_t process, double timeout)
1106 - msg_bar_t MSG_barrier_init(unsigned int count)
1107 - int MSG_barrier_wait(msg_bar_t barrier)
1108 - void MSG_barrier_destroy(msg_bar_t barrier)
1109 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1110 * New option "msg/debug_multiple_use" to help debugging when a task is used
1113 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1115 - Increase code coverage in test suites
1119 * Protect context stacks against stack overflow. The number of protected memory
1120 pages allocated on the top of each stack (1 by default) can be configured
1121 with the new command line option --cfg=contexts/guard_size:<value>.
1122 * Simcalls are now generated by a python script that
1123 - generates files included by SimGrid
1124 - checks that all the functions exist, and proposes prototypes
1126 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1127 host_set_data, host_get_data
1129 - simcall_process_join(smx_process_t process, double timeout)
1130 * Fix bug where sleeping processing could not be suspended.
1133 * Translate surf models from C to C++
1134 - Generic classes for all models: Model, Resource, Action
1135 - A generic interface for each kind of model (CPU, Network, Storage
1136 Workstation, WorkstationVM)
1138 * Translate surf routings from C to C++
1139 * Add callbacks using sigc++ or boost::signals2
1140 - Add callback functions for resource creation/destruction
1141 - Add callback functions for action state change
1142 - Handle Energy as a plugin
1143 * Replace swag by boost::intrusive
1144 * Add new routing models for clusters. For documentation, see
1145 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1146 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1147 parameters for cluster tag
1148 - Fat trees, with topology="FAT_TREE" and
1149 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1150 - see examples/platforms/torus_cluster.xml and
1151 examples/platforms/fat_tree_cluster.xml
1152 * More documentation
1155 * Hostfiles support host:nb_processes construct to deploy several processes on
1157 * Collective communication algorithms should not crash if used with
1158 improper number of nodes and report the error.
1159 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1160 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1161 MPI_Dims_create, MPI_Cart_sub are supported.
1162 * New interface to use SMPI programmatically (still depends on MSG for
1163 some parts, see examples/smpi/smpi_msg_masterslave):
1164 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1168 * Global variables privatization in MPI executables is now performed at runtime
1169 with the option smpi/privatize_global_variables (default:no).
1170 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1171 dynamic libraries loaded by the application are not privatized (static
1172 linking with these libraries is advised in this case)
1175 * Options defined in XML work correctly now.
1178 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1179 copied into simgrid.jar or not (ON by default). Use this option if you want
1180 to reduce the size of the installed simgrid.jar, *and* the native libraries
1181 are kept installed elsewhere.
1182 * Surf binding with SWIG (code generated in maintainer mode only):
1183 - plugin to handle callbacks
1184 - CPU model only for the moment
1187 * Supernovae build mode is definitively removed. It was used to improve
1188 inlining and inter-module optimizations. It is nowadays superseded by
1189 link-time optimizations commonly available in compilers.
1190 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1191 * Add boost dependency for surf++
1192 * Add new macro for tests
1193 - ADD_TESH(name <tesh_args>)
1194 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1198 - xbt_bar_t XBT_barrier_init(unsigned int count)
1199 - int XBT_barrier_wait(xbt_bar_t barrier)
1200 - void XBT_barrier_destroy(xbt_bar_t barrier)
1201 * Make the xbt_os_time module public
1203 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1205 SimGrid (3.10) stable; urgency=low
1207 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1210 * Reintegrate Java to the main archive as desynchronizing these
1211 package is not acceptable anymore (Java is now considered stable)
1212 * Add explicit synchronization facilities through semaphores
1213 * Bug fix: Task.setDataSize() only changed the C world, not the value
1214 cached in the Java world
1217 * Dramatically change the way files are handled. API and internals changed, but
1218 this part of MSG was not considered as production grade either.
1219 * Add explicit synchronization facilities through semaphores
1220 * Add a new function MSG_host_get_process_list()
1221 * Preliminary DVFS support (see examples/msg/energy/ for details)
1224 * SMPI is now included directly in the libsimgrid as the windows
1225 linker doesn't force us on splitting it anymore.
1226 * Improvements of the SMPI replay tool:
1227 - Most of the collective communications are now rooted in the same process as
1228 in the original application.
1229 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1230 used until now). Multiple data types can now be used in a trace.
1231 - The replay tool now supports traces produce either by TAU or a modified
1233 - Bug Fix: the compute part of the reduce action is now taken into account.
1234 - Gatherv collective is now supported
1235 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1236 of smpirun outputs time independent traces for the current run. One file
1237 is created per process. If too many processes are simulated, this behavior
1238 can be changed to one file for all processes by using the
1239 tracing/smpi/format/ti_one_file flag
1240 * smpirun generates the host file if needed (with given host count and platform)
1241 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1242 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1243 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1244 --cfg=smpi/coll_selector:(mpich/ompi)
1245 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1247 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1248 happening inside a collective SMPI call.
1249 * Fix the behavior of complex data types handling.
1250 * Make MPI_Wtime another synchronization point to take computations into
1252 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1253 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1254 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1255 (no privatization of global variables yet)
1256 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1259 * Verification of liveness properties is now available for SMPI applications
1260 (in addition to MSG applications)
1261 * Bugged examples using SMPI in examples/smpi/mc/
1262 * Add --cfg=model-check/visited option. Allows the verification of infinite
1263 programs. Detection of loops in the execution thanks to the system state
1264 comparison and reduction of the state space to explore. Can be combined with
1265 DPOR for safety properties.
1268 * Allow to change SimGrid configuration (see --help) within the code
1269 thanks to SD_config() as it can be done in MSG.
1270 * Add a new function SD_task_set_amount() upon user request.
1273 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1274 * Remove rule based routing (no more PCRE dependency)
1275 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1276 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1277 * Add a loopback_bw and loopback_lat options to cluster tag.
1278 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1279 routing. Now peers are encapsulated in an AS and have their own private
1280 router but this is transparent.
1283 * Our own implementation of getline is renamed xbt_getline, and gets
1284 used even if the OS provide a getline(). This should reduce the
1285 configuration complexity by using the same code on all platforms.
1286 * New type: xbt_cfg_elm_boolean.
1287 * Allow to use yes/no for boolean configuration options in the command line.
1288 * Allow to disable SimGrid cleanups at exit from command line option.
1289 There are situations where one may want a simulation to end with an exit.
1290 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1291 annoying when scripting around the simulator. Adding a
1292 --cfg=clean_atexit:no allows to circumvent this issue.
1295 * Lots of memory leaks were corrected in this release.
1296 There is no known memory leaks anymore, in all of our 600+ tests.
1297 * New command line option --version, to get SimGrid version information.
1298 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1300 * Supernovae builds are deprecated, and expected to be removed in the next
1303 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1305 SimGrid (3.9) stable; urgency=low
1307 The Grasgory release: GRAS is really dead now.
1309 * Complete overhaul of the internal host structures scheme.
1312 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1313 was considered as experimental and badly maintained since then.
1314 * Keeping it was thus a trap to our potential users, that could take
1315 it instead of MSG or SMPI by mistake despite is pity state.
1316 * GRAS seems to have very few users (if any), and no one volunteered
1317 to maintain it further. It also induces a lot of XBT code (for
1318 portability sake), that must be maintained too.
1319 * For all these reasons, we killed GRAS. If someone wants to revive it
1320 in the future, don't cry, our git history still remembers of GRAS.
1323 * Major overhaul. Merge our documentation again as time proved that
1324 splitting it was really not helping our users.
1325 * Further improve the developer documentation to help newcomers
1326 hacking on SimGrid itself. The user documentation (and in
1327 particular, the beginner documentation) is still in a sorry state.
1330 * Now works on Windows too!
1331 * Much more extensive test suite, from MPICH
1334 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1335 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1336 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1337 * Bug fix: let task be scheduled when the last dependency to be solved is
1338 a control dependency.
1339 * Remove SD_load_environment_script function.
1340 Use the C sg_platf function if you want to declare a platform
1344 * New function: MSG_process_get_number()
1345 * Old function documented: MSG_config()
1346 * Remove MSG_load_platform_script function
1347 Use the C sg_platf function if you want to declare a platform
1351 * Change the default value of the TCP_gamma constant (maximal size of TCP
1352 congestion window) to a more realistic 4MiB value. If you notice changes in
1353 your simulation results, you can fall back to the previous 20k tiny window
1354 by adding --cfg=network/TCP_gamma:20000 on command line.
1355 * (Hopefully) fix a bug wrt periodic availability and state traces
1356 * Bug fix: use default values at start when first event in availability/state
1357 trace is not at time 0.
1360 * remove the "new_" part of function name sg_platf_new_trace_connect
1361 (resulting in sg_platf_trace_connect), since it does not create
1365 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1366 use case where it's really mandatory, and maintaining it was a pain in
1368 * New: xbt_fifo_search(), search an item with a user-provided
1369 comparison function instead of dumb pointer comparison.
1372 * Fix the lua deployment:
1373 Use `simgrid.init_application()` before deployment instead of
1374 `simgrid.msg_register_application()` after.
1377 * Transfer the tracing files into the corresponding modules.
1379 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1381 SimGrid (3.8.1) stable; urgency=low
1383 The "we are told that some people want to also *install* the simgrid
1386 * Add missing file "tesh.1" to the archive.
1388 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1390 SimGrid (3.8) stable; urgency=low
1392 The Psssshiiiit release: SimGrid jumps into the Cloud.
1395 * Add an experimental interface to manipulate VMs. They are mainly
1396 process groups with very few intrinsic semantic, but they should
1397 allow you to build the semantic you want easily.
1398 * New function: MSG_host_set_property_value()
1399 * New function: MSG_process_on_exit(). To clean memory in all cases.
1400 * Bug fixes that made the host (and link) failures unusable.
1401 * Add a way to auto-restart process when the host in which they are
1402 executing comes back (ON_FAILURE="RESTART" on deployment file,
1403 MSG_process_auto_restart_set).
1404 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1405 please stop using the old ones, they are DEPRECATED.
1407 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1408 Please use MSG_init() instead. (reducing the amount of entry
1409 points in the library helps us).
1410 * Make it impossible to link against the wrong version of the lib
1411 * Deprecate MSG_clean(). No need to call it anymore.
1412 * Function MSG_get_host_number() is not deprecated anymore.
1415 * Split the doc into a user guide and a reference guide.
1416 * Start a developper guide to help people hacking on SimGrid.
1419 * Enable tracing by default. This modules rocks you should use it.
1420 * Remove option custom_flags. Now use environment variables CFLAGS
1422 * Use default cmake things to detect lua instead of home grown ones.
1423 * New option "enable_mallocators" to disable mallocators, for debugging
1424 purpose ("on" by default).
1427 * Bug fixes around the resource failures: don't let the processes
1428 survive the host they are running onto.
1429 * Add an interface to auto-restart processes when the host in which they are
1430 executing comes back.
1431 * Ensures that SIMIX_clean is called automatically. It's not part of
1432 the public interface anymore (bindings should be updated).
1435 * Bug fix for when SD_Simulate is called with a positive value: be careful
1436 when comparing doubles. Sometimes they are different for non significant
1438 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1439 parallel task whose initial work is distributed among host according
1440 to the Amdahl's law. Such tasks are created with a parameter alpha
1441 that corresponds to the non-parallelizable part of the computation.
1442 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1443 between two sets of workstations assuming a 1D block distribution (each
1444 workstation owns a similar share of data) on both sides.
1446 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1447 Data redistribution will be automatically scheduled once parent and child
1448 are both scheduled. The filling of computation_amount and
1449 communication_amount structures is now done seamlessly thanks to the chosen
1451 * New function SD_workstation_dump to display various information
1452 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1453 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1454 bandwidth on the route onto which the task is scheduled and the amount of
1456 To divide the nominal bandwidth by 2, the rate then has to be:
1457 rate = bandwidth/(2*amount)
1458 * Compute tasks that have failed can now be rescheduled and executed again
1459 (from their beginning)
1460 * Increasing source code coverage (src/simdag is now covered at 95.8%
1464 * Re-implement time-independent trace replay using SMPI (at the
1465 smpi_smp_* level) instead of MSG. This should replace
1466 examples/msg/actions/actions.c
1467 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1468 hindexed and structs)
1469 * Implement the exchange of non-contiguous data.
1470 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1471 * Correct behavior of smpi/sender_gap and set its default value to 0
1472 * Add option to asynchronously send small messages to allow better
1473 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1474 specifies the size in bytes under which messages will be asynchronously sent.
1475 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1476 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1479 * New C interface to define a platform: XML is now optional.
1480 For more info, please check include/simgrid/platf.h.
1481 * New interface to define random platforms from the C:
1482 For more info, please check include/simgrid/platf_generator.h and
1483 examples/msg/masterslave/masterslave_platfgen.c
1484 * Export a sg_cmdline dynar containing all the arguments we got from
1488 * Two new tracing options for adding comments to trace file so you
1489 can track your experiments (see --help-tracing for details).
1490 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1491 * Adding the SimGrid version that generated the trace file as a comment.
1492 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1493 * Fix to avoid key clashes on Paje links
1494 * Other minor fixes related to the Paje specification
1497 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1498 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1499 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1500 * Greatly improve the robustness of mmalloc to user errors (such as
1501 using an area after freeing it, or freeing it twice)
1503 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1505 SimGrid-java (3.8.1) stable; urgency=low
1507 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1508 Warning: all methods are visible, but only some of them are
1509 implemented so far. Check the source (src/jtrace.c)
1510 for further information.
1511 * New module: org.simgrid.msg.File (SimGrid File management functions)
1512 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1513 * Change the meaning of Process.restart: now restart the process from
1514 the begining, like MSG_process_restart in C.
1515 * Add Process.setAutoRestart: handling of process restart when failed
1517 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1518 you to retrieve the properties of the processes/hosts
1519 * Deprecate Msg.clean(): you can just forget about it now.
1520 * New function Process.getCount(), that only works when compiling
1521 with the not yet released version 3.9 of the C library.
1523 * New context factory based on Coroutines. It mandates a modified JVM
1524 but then, the simulations run about five times faster, and there is
1525 no limit to the amount of processes (beside of the available memory).
1527 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1529 SimGrid (3.7.1) stable; urgency=low
1532 * Restore the prototype of MSG_process_create_with_environment() to
1533 the pre-3.7 situation by removing the kill_time argument.
1534 * Add a MSG_process_set_kill_time() function instead.
1537 * Fix weird behaviors when dealing with parallel tasks.
1540 * Simgrid is now built as a dll.
1541 * Simgrid-java now works on Windows.
1542 * Simgrid-Java is now included into Windows package.
1545 * First pre-build package for MacOSX.
1548 * Fix compilation when using MSG_USE_DEPRECATED.
1549 * Fix some compilation issues on Macs and Windows.
1550 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1551 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1553 We discovered that the Lua console is broken, but we are missing the
1554 manpower to fix it right now. The problem existed in 3.7 too, so we
1555 are not blocking the release for that. Sorry if you depended on this
1556 feature, any help would be really welcome.
1558 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1561 SimGrid-java (3.7.1) stable; urgency=low
1563 The "Java aint got to be bloated and slow" release
1566 * Various internal cleanups and performance improvement
1567 Simulations are expected to run up to twice faster or so
1568 * Make Process.kill(process) an instance method, not a static one
1569 * User processes are not java.lang.Thread subclasses.
1570 This breaks the compatibility (sorry), but previous API was
1571 brain-dead, making it impossible to have non-trivial
1572 initializations in the process constructor.
1573 * Require a full constructor per Process sub-class.
1574 Kinda breaks the compatibility (sorry), but this allows a much more
1575 efficient way to launch the processes at simulation startup.
1576 * Do not embeed our version of semaphores, java 1.5 can be considered
1577 as sufficiently prevalent for us to not dupplicate its features.
1581 * Add examples for almost every part of the API
1582 We spotted and fixed a lot of bugs in the process
1584 * New module: asynchronous communication API
1585 * New function: Process.sleep()
1586 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1587 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1588 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1590 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1592 SimGrid (3.7) stable; urgency=low
1594 The "spring cleanups (before the next Big Project kicks in)" release.
1597 * We can specify the SMPI latency/bandwidth factor with command line
1598 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1599 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1600 You can also use the "config tag" from platform file by adding this line
1601 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1602 (see "example/platforms/tag_config.xml" to use "config tag").
1603 Note that the command line supersedes the platform file configuration.
1604 * Change the correction factors used in LMM model, according to
1605 the latest experiments described in INRIA RR-7821.
1606 Accuracy should be improved this way.
1607 * Use the partial invalidation optimization by default for the
1608 network too. Should produce the exact same results, only faster.
1609 * Major cleanup in surf to merge models and split some optimization
1610 mechanisms from the core of the models. As a result you can now
1611 specify which model to use (e.g., --cfg=network/model:LV08
1612 --cfg=cpu/model:Cas01) and which optimization mode to use
1613 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1614 Incompatible combinations should err at initialization. See
1615 --help-models for the list of all models and optimization modes.
1616 * The CLM03 workstation model was dropped for simplicity because it
1617 used the deprecated CM02 network model. Use default instead.
1618 * Rename the TCP_gamma configuration option to network/TCP_gamma
1619 * Rename the coordinates configuration option to
1620 network/coordinates, and document it
1621 * Use now crosstraffic keyword instead of the terribly misleading
1622 fullduplex keyword. It is activated by default now in the current
1623 default model, use --cfg=network/crosstraffic:0 to turn it off.
1624 * Ongoing refactoring the model parsing to make XML files optional
1625 See include/simgrid/platf.h for details (still to be completed)
1628 * Major overhaul of the documentation. Almost instructive now :/
1629 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1630 functions and friends. This interface was considered as
1631 deprecated since over 2 years, it's time to inform our users that it is.
1632 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1633 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1634 use these (crappy) functions in your code.
1635 These functions will be removed soon. Stop using them now.
1636 * Deprecate MSG_get_host_{table,number}
1637 Implement MSG_hosts_as_dynar() instead.
1638 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1639 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1642 * Stabilize the parallel execution mode of user contexts
1643 * Introduce configuration variables to control parallel execution:
1644 - contexts/synchro: Synchronization mode to use when running
1645 contexts in parallel (either futex, posix or busy_wait)
1646 - contexts/parallel_threshold: Minimal number of user contexts
1647 that must be part of a scheduling round to switch to parallel
1648 execution mode (raw contexts only)
1649 * Fix bugs that prevented to use suspend/resume along with
1650 synchronization structures.
1651 * Fix bugs in process termination that lead to invalid memory access
1652 in very specific conditions.
1655 * Introduce a parallel mode for the models (controlled by surf/nthreads
1656 configuration item). In our tests, running the models in parallel
1657 never lead to any speedups because they are so fast that the gain
1658 of computing each model in parallel does not amortizes the
1659 synchronization costs, even when ultra fast futexes are used.
1660 This is released anyway because YMMV.
1663 * Performance boost by using a swag internally to compute the set of
1664 tasks that are finished and should constitute the return value of
1668 * Enable it by default now that it is considered rather stable.
1671 * Documentation of the tracing functions.
1672 * Performance gains when tracing categorized/uncategorized resource
1673 utilization by avoiding calls to get route when updating resource
1674 variables. LMM constraints are being used instead.
1675 * API changed to set task categories. Use MSG_task_set_category instead
1676 of TRACE_msg_set_task_category, and SD_task_set_category instead
1677 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1678 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1679 * Examples for link user variables added (see at examples/msg/tracing/)
1680 * Deprecated function TRACE_msg_set_process_category completely removed
1681 * Trace header updated according to the latest Paje file format
1682 * Tracing network lazy updates, no longer obligate users to use full updates
1683 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1684 * Experimental: let user code declare/set/push/pop application states for hosts
1685 * API changed to allow the manual creation of graph configuration files
1686 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1689 * Improve the API of Lua MSG bindings, using the Lua spirit.
1690 * Each simulated process now lives in its own Lua world (globals are
1691 automatically duplicated). It helps writing simulators. Will allow
1692 to run Splay programs within SimGrid in the future.
1693 * Add a Chord example in Lua, equivalent to the MSG one.
1696 * Start the implementation of a solution to express temporal
1697 properties, not only local assertions. This is still an
1698 experimental work in progress, stay clear from it to be safe.
1702 - Add new runtime parameters --help-logs and --help-log-categories
1703 to display information about supported logging parameters and
1705 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1706 don't exists anymore.
1707 * Mallocators: allow value NULL for the reset function.
1709 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1710 create homogeneous dictionaries, where all the elements share the
1711 same free function. Non homogeneous dictionaries will be
1712 deprecated in the next release.
1713 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1714 - Multi-level dictionaries are deprecated.
1716 - new function xbt_dynar_search_or_negative() that is useful when
1717 you have less than 2 million elements in your dynar and don't
1718 want of the extra complexity of catching exceptions when the
1719 element is not found.
1721 - Make xbt_os_thread module (for thread portability) public.
1722 Documentation is still to come, sorry.
1724 - Cleanups and simplifications to make it maintainable again.
1725 - Exotic features (such as memalign and valloc) were removed.
1726 - The metadata were extended and improved so that the
1727 model-checker becomes able to explore and inspect the heaps.
1728 - This may induce a performance drop when enable_model-checking is
1729 ON in cmake (even if it's not used in the simulation), but it is
1730 necessary at this point to get MC working.
1732 Turn model-checking OFF if simulation performance matters to you.
1733 Not enabling it at runtime is not enough, disable it in cmake.
1735 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1738 SimGrid (3.6.2) stable; urgency=low
1740 The "Not coding new stuff allows to polish old things" release.
1743 * New bindings to the NS3 packet level simulator (experimental)
1744 * Use the raw (efficient) execution contexts instead of the sysv
1745 (portable) ones when possible.
1746 * libpcre is now mandatory in any cases since not using it led to
1747 severe performance loss and possibly other issues
1748 * Update the XML platforms:
1749 - G5K: include the latest machine in Nancy
1750 - GridPP and LCG: new platforms
1751 * Documentation was partially updated, at least (more to come)
1753 Bug fixes, cosmetics and small improvements
1754 * Free terminated processes before the end of the simulation to avoid
1755 exhausting the memory of users having very dynamic amount of
1757 * Bug fix and cosmetics about canceling non-running tasks
1758 * Bug fix about the dot loader's issues when using libcgraph
1761 * Create an installer for windows with nsis (amd64 and win32)
1762 - Add an hello world project to illustrate simgrid project creation.
1763 - Embed libpcre into the Simgrid installer to avoid
1764 its compilation burden
1765 * The raw execution contexts should work on Apple now
1766 * Port to Windows 64 bits
1767 - Sysv contexts now have an implementation for this arch
1768 - GRAS communication features now support this arch
1769 * Drop support for borland compiler on windows
1770 - this code was not maintained, and we kinda depend on gcc nowadays
1771 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1772 * Fix portability issue on unstable ubuntu: linker became picky on
1775 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1777 SimGrid (3.6.1) stable; urgency=low
1779 The "Oops, we broke Macs too" release
1782 * Fixed contexts detection so that raw ones are used when possible
1783 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1784 leads to a strange error, with user code segfaulting sometimes when
1785 the generated code is not perfectly aligned (which is not
1786 controllable from the user side, depends on the amount of code)
1789 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1791 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1793 SimGrid-java (3.6) unstable; urgency=low
1796 * Split of every thing from simgrid v3.5 into a separate package.
1798 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1800 SimGrid (3.6) stable; urgency=medium
1802 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1805 * Bindings now constitute their own package, separated from the main one.
1806 Rationale: reduce our maintenance nightmare by reducing the module coupling
1807 They will soon be released on their own on gforge.
1809 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1810 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1812 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1813 * It's not quite deprecated for now because we have no replacement,
1814 but it may soon become the case.
1817 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1818 * Fortran: New user-level cache variable to store the rank of the running
1819 process. This improves performance by an order of magnitude.
1820 * C: New coccinelle script to automatically locate and modify global and
1821 local static variables.
1822 * Improved SMPI network model with a sender-side gap to account for multiple
1826 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1827 only say if a communication is finished, no matter whether it succeeded or
1828 failed. You can call MSG_comm_get_status() to know the status of a finished
1830 * New function MSG_task_dsend() to send a task and detach it. When a
1831 communication is detached, you are never notified of its success or failure
1832 and the memory is released automatically once it is finished. This function
1833 is useful when you don't care about the end nor the success of a
1835 * Change the prototypes of action replay. Sorry for inconvenience,
1836 but this is really more efficient this way (and to adapt your code,
1837 you just have to fix the initialization, that shouldn't be too long)
1838 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1839 ever managed to do anything useful with it.
1840 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1841 takes a dynar of communications. It returns immediately and gives the
1842 index of a finished communication (if any).
1843 * New example: a basic implementation of the Chord P2P algorithm.
1846 * New model for multi-core CPUs. You can now use the core attribute to
1847 precise the number of cores of a host. This is a basic model. Every
1848 process running on the host receives at most the power provided in
1849 the DTD (throughput<=power). Total throughput of process cannot exceed
1851 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1852 router linked by an up-link and a down-link (possibly asymmetrical).
1853 This kind of pattern allows to easily build last-mile model style platforms.
1854 Aggregating such patterns in a rule-based AS is thus the technique of
1855 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1856 * New model for Vivaldi routing. We transformed the Vivaldi network model
1857 into a Vivaldi routing model (based on the rule-based model). This allows to
1858 combine Vivaldi based latencies with last-mile platforms.
1861 * Added a check for NaN of IEEE754 infinite in the double entries of
1863 * Introduce a new context factory "raw", highly inspirated from the
1864 ucontext factory, but using manually crafted functions in assembly to
1865 do the work in an efficient manner.
1866 * Allow to change the used context factory at run time, not only at
1867 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1868 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1869 contexts at runtime (only with raw contexts or ucontexts).
1870 * Completely rewrote this module to allow parallel execution of user
1871 processes. Use --cfg=contexts/nthreads:N to execute user processes
1872 with N parallel threads (the default is 1, meaning no parallelism).
1873 * Allow to decide dynamically between sequential and parallel modes.
1874 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1875 processes in parallel only when their number is greater than or equal to P
1877 * Added a check for NaN of IEEE754 infinite in the double entries of
1881 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1882 won't output the state of processes when interrupted with Ctrl-C
1883 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1884 NULL-terminated array. This may solve backward compatibility issues
1885 due to the change to return type of SD_simulate. See also:
1886 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1887 * Add new macros with variable number of arguments.
1888 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1889 - in xbt/asserts.h: xbt_assert
1890 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1891 - in xbt/ex.h: THROWF and RETHROWF.
1892 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1894 * Change xbt_die() to accept a format string with arguments, just like printf.
1895 * New data structure: xbt_lib_t, like a dict but more general and with better
1899 * New configuration options
1900 Options triva/categorized and triva/uncategorized can be used to generate
1901 graph configuration files for Triva visualization tool.
1902 * Configuration option tracing/platform is renamed to tracing/categorized
1903 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1904 * New TRACE_user_link_variable interface:
1905 User provides the name of the link and the tracing variable to attach to it
1906 * the declaration of tracing categories must be done after the environment
1908 * simpler tracing interface, just one way to declare categories
1909 TRACE_category or TRACE_category_with_color, it is up to you
1910 * links in the trace file are again identified by their names
1911 * trace contains the full platform hierarchy exactly as declared using the ASes
1912 * Options tracing/msg/[task|process]:1 groups the process by hosts
1913 for both cases, tasks and processes must have names that are unique during the simulation
1914 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1915 * The experimental option tracing/msg/volume is deprecated
1916 its functionality may be reincorporated if needed
1918 The tracing generates a trace file with unordered timestamped events,
1919 because of the way the core simulator (surf) works. A script available
1920 at the tools directory (fix-paje-trace.sh) can be used to put the events
1921 in order. We have changed the tracing so it can generate ordered timestamped
1922 events in the final trace, but depending on the simulator (and how much time
1923 is simulated) that can lead to a huge memory utilization. It is deactivated
1924 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1926 Build Infrastructure
1927 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1928 - We are trying hard to keep the API stable, but it may happen that
1929 some things change (we're a research project after all, not a
1930 nuclear plant operating system). If such things should happen, you
1931 could rely on that macro to adapt.
1932 - current value: 30600 for 3.06.00, aka 3.6
1933 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1934 a number that can be compared with SIMGRID_VERSION.
1935 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1936 assertions and verbose/debug logging events are disabled at compile time.
1938 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1940 SimGrid (3.5) stable; urgency=medium
1943 * New feature to any SimGrid-based simulator: Model-Checking
1944 Check SIN#1 for more details.
1947 * New Model SMPI (three-interval linear regression for correction factors)
1948 See RR-7426, available at http://hal.inria.fr/inria-00527150
1949 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1950 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1951 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1952 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1953 * See also src/smpi/README
1958 - Tracing API changes: TRACE_start and TRACE_end should not be called
1959 by user-code. They are automatically called by simulators created
1960 with SimDAG, MSG and SMPI if the toolkit is compiled with
1961 tracing_enabled=ON. Categories declaration and utilization remain the
1962 same for MSG and SimDag.
1963 - A function was added to the tracing API to declare categories with
1965 - TRACE_category_with_color (char *category, char *color)
1966 where color must be in the following format
1967 "%f %f %f", red, green, blue
1968 and red, green, blue are float values in the interval [0, 1]
1969 - User can specify NULL as color parameter, or continue calling
1970 TRACE_category (cat)
1971 On that case, the tracing system will define random colors
1972 - The following command-line options are supported:
1973 --cfg=tracing/filename:msg.trace
1974 --cfg=tracing:1 (activate tracing, needed to use others)
1975 --cfg=tracing/platform:1 (categorized resource use)
1976 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1977 --cfg=tracing/msg/task:1 (task creation)
1978 --cfg=tracing/msg/process:1 (process creation, migration)
1979 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1980 --cfg=tracing/smpi:1 (SMPI interface tracing)
1981 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1982 - examples of examples/msg/tracing updated
1984 - DAXLoader and DOTLoader functions can generate tasks with categories
1985 - A new function to attribute a category to SD tasks:
1986 TRACE_sd_set_task_category (SD_task_t task, char *category)
1987 * Tracing the MPI interface implemented by SMPI
1988 - Collective operations are traced with states
1989 - Point-to-Point operations are traced with states/links
1990 - Tracing activated by a parameter "-trace filename" passed
1991 to smpirun during execution (considering that simgrid
1992 is compiled with tracing enabled)
1993 - To run the simulation with gdb, the simulator
1994 accepts --cfg=tracing/smpi:1 to trace SMPI
1995 - tesh files to check if smpi tracing is ok
1996 - See examples/smpi/NAS/DT-trace
1997 * GTNetS tracing re-worked
1998 - adaptation to the tracing system of GTNets to cope
1999 with modifications regarding the fullduplex mode
2000 - new tesh files to check if gtnets tracing is ok
2003 * Asynchronous communications through the functions:
2004 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2005 * New function: MSG_load_platform_script()
2006 to make possible using a lua script instead of XML files to set up platforms
2007 * New function: MSG_set_function
2008 to associate functions to processes, used when bypassing the parser
2009 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2011 Platforms: Add some more examples in examples/platforms
2012 * Grid'5000: see www.grid5000.fr
2013 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2016 * Change the XML format. This is a very important modification. SimGrid
2017 3.5 introduces a new hierarchical format based on the notion of
2018 Autonomous Systems. Compatibility with old format is ensured
2019 through the perl script provided in the install bin directory
2020 bin/simgrid_update_xml.
2021 It is now possible to build platforms with specific routing
2022 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2023 connect several platforms together. We will try to provide soon
2024 set of realistic platforms exploiting these properties (have a
2025 look at examples/platforms/ for the moment).
2026 * Take the opportunity of the XML format change to be a good XML citizen:
2027 rename link:ctn to link_ctn and similar changes (also dealt with by
2029 * Add a new routing scheme (rule-based) using regular expressions. It
2030 enables to have an extremely low memory footprint when the
2031 underlying routing is simple and can be compactly described. You
2032 need to have libpcre4-dev (perl regular expressions) installed if
2033 you want to use this routing scheme.
2034 * Revive the cluster TAG and allow to easily and efficiently (both in
2035 term of memory and speed) connect clusters together. Have a look
2036 at teshsuite/simdag/platforms/ to see how this can be done. With
2037 this tag, you can create clusters with thousands of tasks at no
2038 cost (have a look at examples/platforms/).
2039 Note: clusters are implemented as ASes, so there is no need for an
2040 enclosing AS tag if you have only one cluster in your platform.
2041 * Add new generic functions in the public interface that allows the user
2042 to call SURF 'create_resource' methods from your code (same
2043 functionality as the XML bypass mechanism but with a much lighter
2045 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2046 that uses a piecewise linear approximation to produce better
2047 results when exchanging small messages.
2048 * Add a new parameter to handle correctly full duplex link and account
2049 for interferences between uplink and downlink communications
2050 (activate with --cfg=fullduplex:1).
2053 * Rename the SD_READY (all dependencies are satisfied and task is
2054 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2055 dependencies are satisfied) state.
2056 This prevents a confusion between the notion of "ready to schedule"
2057 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2058 simulated" (SD_RUNNABLE) used by the simulation kernel.
2059 * Change the way a task is considered as ready. Instead of removing
2060 dependencies when a task is done, a counter is decreased. This way,
2061 it is always possible to reach ancestors thanks to the
2062 SD_taks_get_parents function (even after the end of the simulation.)
2063 * Change the return type of SD_Simulate from (SD_task_t*) into
2064 xbt_dynar_t. This function was in handling a dynar internally and
2065 converted it into a NULL terminated array for historical reasons.
2066 * New function SD_dotload(char*) to load a DAG described in dot
2067 format. This loader and the corresponding examples require the
2068 installation of the graphviz library.
2069 * Fix a bug in the management of tasks of size 0 in the surf network
2070 models. This problem was only visible with SIMDAG and you should
2071 thus disregard results produced with earlier versions if you
2072 relied on this feature (some tasks were blocked because of this).
2073 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2074 even though your applications were doing only point-to-point
2075 communications and sequential computations. Now you can really use any
2076 model you want (of course, if you create real parallel tasks, which are
2077 not implemented in most models beside ptaskL07, this will abort).
2078 * Add an example that schedules a DAX on an heterogeneous platform
2079 using a Min-Min strategy.
2080 * New function SD_workstation_get_current_task() that returns the kind
2081 of task currently running on a workstation in the sequential access
2083 * Raise some warnings when unexecuted tasks remains at the end of the
2084 simulation. This is usually caused by cycles in the DAG.
2087 * New function: SIMIX_process_set_function() called by MSG_set_function
2088 * Change the underlying waiting queue in semaphores so that a process
2089 can wait on several of them simultaneously (as in waitany).
2090 * Fix the way to handle tokens in semaphores so that all access patterns
2091 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2092 * kill the dirty pimple SIMIX_message_sizes_output()
2093 Please use (proper) visualization instead
2096 * New data container: setset (set of sets of elements)
2097 * New module: mmalloc (mapped malloc, allowing to have several
2098 independent segments of malloc)
2099 * New function: xbt_dict_cursor_set_data()
2100 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2101 * New function: xbt_dynar_is_empty()
2102 * New function: xbt_fifo_get_last_item()
2103 * Fix xbt_dynar_shrink(): use the right element size.
2104 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2105 uninitialized areas during expand.
2106 * Fix semaphores: previous implementation was severely broken.
2107 * Use library init/fini functions for our initialization.
2108 - you can use logs and other feature as soon as you want in your
2109 code (even before the xbt_init / MSG_init)
2110 - xbt_exit is now a no-op and produce a warning when used.
2113 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2114 benefit from the latest improvement to the simulation kernel.
2115 * Kill measurement sockets for now. If you rely on them, sorry. This
2116 release is not for you. This feature will be reintroduced in the
2117 future, but we cannot delay the release any further.
2118 * New function: gras_msgtype_get_name().
2119 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2120 * Fix (at last) the pmm example: it should not randomly fail anymore.
2122 Build chain: bug fixes and overall polishing
2123 * Cmake is now stable enough. Hence, we killed the autotools.
2124 * Port to windows ( TM :)
2125 * Fix the 'make install' target.
2126 No need to use 'make install-simgrid' anymore
2127 * Introduce a 'make dist' target compiling a *source* archive
2128 'make package' compiles a binary archive
2129 * Compile java files only on need
2130 * Add --cd and --setenv command line options to tesh
2131 * Out of source builds are not fully supported yet, but we are close
2132 * Enable supernovae and optimization flags by default for our users
2135 * Add layer to set up environment directly from lua, without XML.
2136 * The effect of gras_stub_generator can be achieved through
2137 lua too (check examples/gras/console/ping_generator.lua)
2139 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2141 SimGrid (3.4.1) stable; urgency=low
2143 The "Polishing easter eggs is probably a good idea" release.
2144 This is a bug fixes release only.
2148 * Fix a bug preventing the tasks from begin garbage collected.
2151 * Fix a bug occuring when a host involved in a communication fails.
2152 This was not detected properly by the other peer involved in the
2153 communication. Now, it's reported as a network error.
2156 * Warn the user about loop dependencies in data flow of DAX files
2157 * Obey the control-flow dependencies of DAX files
2160 * Add option "enable_smpi" allowing to not compile SMPI.
2161 Probably useful for the (Mac) users experiencing a build error here
2162 * Improve the detection of lua5.1 and ruby1.8
2164 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2166 SimGrid (3.4) stable; urgency=low
2168 The "Easter in Cargese" release. Also known as (major changes):
2170 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2171 ~> bindings were greatly improved
2172 ~> new tracing infrastructure for better visualization introduced
2174 * the "Welcome to configury modernity" release.
2175 ~> we switched from autotools to cmake, and improved our cdash
2178 A more detailled list of changes follow (full detail in svn log).
2180 Java Bindings: Various Cleanups
2181 * (install java-gcj-compat-dev on debian-like to use them)
2182 * Remove put/get: no need to export deprecated interface in Java
2183 Use send/receive instead.
2184 * Cleanup the examples and add a README per directory
2185 * Remove example autoDestination (that's the only way to go now)
2186 * Remove example explicitDestination (was a plain copy of basic)
2187 * Make JniException a runtime exception, so that there is no need to
2188 declare the fact that you may encounter such a beast. I guess that
2189 nobody will ever want to survive such error.
2190 * Create specific errors for each MSG case of failure:
2191 host failure, transfer failure, timeout, task cancelled
2192 * Cleanup the exceptions that may get thrown by each function
2193 * Other internal cleanups in Java bindings. Performance still bad :/
2194 Ruby and Lua Bindings: create them
2195 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2196 * That's new and great, you should try them out.
2197 Same functionalities than Java bindings, only even less polished
2199 * Kill the useless "rate" argument of SD_task_get_execution_time()
2200 Everyone used to provide -1 as a value, it was not used, and the
2201 semantic of a possible use wasn't even clear.
2202 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2203 either as comm costs or compute costs to mean that there is no
2204 such thing for that specific task.
2205 * Add a SD_task_set_name() function
2206 * Fix SD_task_unschedule() on typed tasks
2207 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2208 * In DAX loader, accept useless 'level' attributes to <job> since
2209 LIGO DAGs have them (seem to be to ease graphical representation).
2211 * Add an example masterslave_mailbox.c using send/receive and not
2212 the deprecated put/get interface.
2213 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2214 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2216 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2217 (declare MSG_USE_DEPRECATED to still have the old name)
2218 * Add a MSG_task_set_data() function
2219 * About trace replay (see examples/msg/actions):
2221 - Allow to work with splitted trace files for each process
2222 Give the specific trace file as argument of each process,
2223 and call MSG_action_trace_run(NULL)
2224 You can still have one merged file for all processes.
2225 - Fix implementation of collective operations
2226 * Allow task_execute() on 0-sized tasks (closes #10063)
2228 * This is the first release of SimGrid where SMPI is not considered
2229 beta anymore (even if some corners should still be improved)
2230 * Port over the new SIMIX_network submodule (internal refactoring)
2231 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2232 * Implement more missing elements of the standard:
2234 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2236 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2237 MPI_Scan() MPI_Get_processor_name()
2238 - Added implementation of missing case for Alltoall (warning: it's
2239 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2240 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2241 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2242 SMPI_MPI_Allgatherv()
2243 * Bug fixes include:
2244 - MPI_Waitsome() was broken
2245 - Allow relative includes in smpicc
2246 - Command line cfg argument 'reference_speed' was ignored...
2247 - Some functions did not properly lead to auto-benching of user code
2248 - smpicc passes -O2 by default (just like openmpi one)
2250 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2251 * Bug fixes about timeouts during communications
2252 * add SIMIX_message_sizes_output() as a pimple to write to file the
2253 amount of messages per size. Use gnuplot to get histogram.
2254 Pimple because that's the only user-visible function of simix,
2255 defined directly in xbt.h (irk, sorry)
2257 - Add a SIMIX_sem_get_capacity() function
2258 - Fix interactions with processe resume/suspende
2259 - release_forever() was stupidly broken
2260 - Fix SIMIX_display_process_status() for processes in a semaphore
2261 - Make SIMIX_sem_block_onto() user-visible
2262 * Refactoring context stuff:
2263 - Use pseudo-OOP for better modularity
2264 - reimplement SIMIX_process_kill() without process_schedule() so
2265 that the latter can take as invariant that it is called from
2267 - Merge context_start into context_new for sake of simplicity
2269 * Add a Vivaldi network model, coded live during SUD'10 ;)
2270 * Rename configuration variables to start a hierarchy:
2271 o cpu_model -> cpu/model
2272 o network_model -> network/model
2273 o workstation_model -> workstation/model
2274 * New configuration variables:
2275 o network/bandwidth_factor: correction to bandwith
2276 o network/latency_factor: correction to latency
2277 o netwotk/weight_S: correction to the weight of competing streams
2278 * Add a long description to the models, that users can see with such
2279 argument on the command line: --cfg=cpu/model:help
2280 * --help-models display the long description of all known models
2282 * config: add the ability to set a default value after registration
2283 Does not override any previously set value (e.g. from cmd line)
2284 * dict: allow to have integer key and data.
2285 When so, you need to use the following functions
2286 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2287 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2288 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2289 In contrary to regular dicts, the key is not malloced before copy.
2290 Mixing scalar and regular elements in the same dict is not tested
2292 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2293 Tracing for Visualization:
2294 * SimGrid is now instrumented in order to generate a trace file for
2295 visualization analysis: to use it, need to compile SimGrid with the
2296 "tracing" option enabled, and instrument the program using SimGrid with
2297 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2298 (among other functions).
2299 * The instrumentation only traces the platform utilization for now
2300 * Documentation to use the tracing functions and how to analyze the
2301 traces with the Triva tool is written.
2302 * More information about: SimGrid FAQ (in the section Tracing Simulations
2305 * We moved to cmake as default build system. Autotools support will
2306 be dropped soon. Check the FAQ for more info about how to use it.
2307 * Greatly improved our cdash/ctest interactions
2308 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2309 * Added memory checking tests with valgrind; lot of memleak fixing.
2310 This may be the first release of simgrid with so few memory issues
2311 * Added code coverage tests.
2312 Our coverage is still improvable, but at least we see it on cdash.
2314 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2316 SimGrid (3.3.4) stable; urgency=low
2318 The "Desktop Grid needs love too" release (also called Xmas release).
2320 Models improvements:
2321 * Major speedup in the maxmin system solving by using lazy evaluation
2322 Instead of solving completely the maxmin system at each iteration,
2323 only invalidate (and recompute) the modified parts.
2324 This new feature is enabled in default models but you can try to
2325 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2326 * Cas01 IMproved as default CPU model
2327 This CPU model is the same Cas01 model, but it uses the
2328 maxmin-selective-update flag and a heap structure to manage
2329 actions on SURF kernel.
2330 It reduces the complexity to find the next action to finish and,
2331 consequently, it's faster than the old Cas01.
2332 This is the new default CPU model (Cas01).
2333 * Rename the old Cas01 model to Cas01_fullupdate
2334 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2335 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2336 * CpuTI (CPU Trace Integration)
2337 A new CPU model whose objective is simulate faster when using
2338 availability trace files.
2339 Instead of using a full featured, over engineered maxmin system for
2340 CPU modeling, this model does the pre-integration of traces files
2341 to calculate the amount of CPU power available, and so, executes
2342 faster than the old CPU models.
2343 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2344 * Use LV08 as default network model since it gives better accuracy
2345 for small messages and shouldn't change things for big ones.
2346 Use --cfg=network_model:CM02 to get the previous behavior.
2349 ******************************************
2350 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2351 ******************************************
2352 * The new CPU model may changes simulations!
2353 The point is that events occurring at the exact same timestamp
2354 are not scheduled in the same order with the old and new
2355 version. This may be enough to completely change the execution
2356 of simulations in some cases.
2357 * The new network model will change simulations!
2358 This new model is more realistic than the previous one, so you
2359 should consider redoing your old experiments with this model.
2360 Sorry for the inconvenience.
2363 * Introduce the supernovae compilation mode
2364 When compiled that way, the whole SimGrid (or almost) is put in a
2365 single compilation unit and compiled in one shoot.
2366 This is to help gcc which has difficulties to inline stuff from one
2368 The speedup seem to be above 15%, althrough more tests are needed on
2369 amd64 to confirm that gain.
2372 * Port of MSG's mailbox on top of SIMIX network
2373 The put/get mechanism was greatly simplified on the way.
2376 * New SIMIX network module. Provides:
2377 - Mailbox: rendez-vous mechanism to find with who you want to speak
2378 - Synchronous send/recv: easier and hopefully faster since the
2379 logic is handled in the maestro process directly now
2380 - Asynchronous send/recv: you dreamt of it? It's here now
2381 Too bad that nobody cared enough to propagate the change to MSG.
2382 * Add semaphores as SIMIX synchronization mechanism.
2385 * new function SD_daxload(char*) to load a DAX file
2386 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2387 * Introduce typed tasks. Specify its kind and cost at creation.
2388 At scheduling, just give where it should be placed, and the cost
2389 for each involved resource is automatically computed.
2390 Existing constructors so far (more to come of course):
2391 - SD_task_create_comm_e2e() for end-to-end communication
2392 - SD_task_create_comp_seq() for sequential computation
2393 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2394 * new function SD_task_dump() for debuging display
2395 * new function SD_task_dotty(task,FILE*) writing to file the info
2396 about the task in dotty format
2397 * SD_task_dependency_exists() can now cope with having one of its
2398 arguments NULL. If so, it tests whether the other argument has any
2400 * Add getters on list of preceding/following tasks:
2401 SD_task_get_parents(task) and SD_task_get_children(task)
2402 * Add getters on amount of workstations and list:
2403 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2404 * Add getter on task kind: SD_task_get_kind(task)
2405 * Update the start_time and finish_time of tasks on completion/failure
2406 * Bugfix: Remove task from state swags when destroyed
2409 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2412 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2413 dynar with one separate thread per value of the dynar.
2414 * Change the prototype of xbt_thread_create(), sorry.
2415 Added a boolean parameter indicating whether we want to join this
2416 thread (used in SG only for now)
2417 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2420 * GTNetS wrappers should now be usable again (and betterly tested too)
2421 * Fix a major regression from 3.2 where the timeout provided to
2422 MSG_task_put_with_timeout() was used as absolute time before which
2423 the comm should be done.
2424 * Start to fix the <cluster> tag.
2425 - Internal links should be good now (beside of the loopback, which
2426 use the private link instead)
2427 - paths to the external world is still rather broken
2428 - the <route:multi> tag is just broken. Actually that's brain-dead.
2429 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2430 to make it less stupid
2431 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2432 * Fix a source-level compatibility glitch from 3.2: after defining
2433 MSG_USE_DEPRECATED, you can use the old name
2434 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2435 * Allow to compile from the SVN with automake 1.11
2436 * Fix some problems when using the "start_time" tag in deployment XMLs.
2437 * Fix #8569: XBT/synchro.h has redundant declarations
2438 * Fix #8563: MSG return values and exceptions
2439 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2440 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2441 Thanks to Robson Peixoto.
2442 * Fix a vicious bug in dictionaries inducing that some elements were
2443 not freed on xbt_dict_free()
2445 Portability report of this version:
2446 * Main portability targets:
2447 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2448 on (amd64/i386/ia64)
2449 - mac leopard on i386
2450 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2451 but nothing critical.
2452 * Other platforms: windows, AIX and others were not tested for this release
2454 Timing report of this version:
2455 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2456 scenario parameters). From 8h to a few seconds in desktop grid settings.
2457 * Supernovae brings about 25% speedup on i386.
2459 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2461 SimGrid (3.3.3) stable; urgency=low
2463 The "Need for Speed" release.
2465 The timings done to validate the 3.3.2 were faulty.
2466 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2468 The problem was a conversion from a manually handled vector to
2469 xbt_dynar_t on the critical path.
2470 xbt_dynar_foreach calls functions, inducing stack management crap.
2472 We inlined these functions and xbt_dynar_foreach is now breath taking.
2473 We also inlined xbt_swag_belong on the way.
2475 Here are some approximate speedup measurements (on master/slaves
2476 simulations lasting between 10s and 20s each):
2477 3.3.1 -> 3.3.2: about same performance
2478 3.3.2 -> 3.3.3: 40% speedup
2479 3.3.1 -> 3.3.3: 40% speedup
2480 3.3.1 with inline patch -> 3.3.3: 30% speedup
2482 Our reading is that the refactoring which occurred in 3.3.2 made us
2483 suffer much more from the xbt_dynar_foreach low performance, but
2484 once we solved this, this refactoring proved to be very performance
2485 effective. From the 40% speedup, somehow, 10% are due to the
2486 inlining and 30% to the refactoring.
2488 That's a pitty that gcc cannot inline functions placed in other files
2489 alone. We have to choose between:
2490 - break the encapsulation (by putting private data structures and
2491 accessors in headers files to help gcc)
2492 - live with low performance
2493 - switch to a decent compiler such as icc (not quite possible).
2495 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2497 SimGrid (3.3.2) stable; urgency=low
2499 The "Simplicity does not preceed complexity, but follows it" release.
2501 The main contributors of this release were (lexical order):
2502 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2505 * Extract the routing logic into its own object.
2506 (was dupplicated in network.c and workstation_LV07.c;
2507 Allows to implement other ways of storing that info)
2508 => kill now useless network_card concept
2509 - Use dynar to represent routes (instead of void** + int*)
2510 - kill link_set (use surf_network_model->resource_set instead)
2511 - Add a command-line option to choose the routing schema to use
2512 - Add three new models:
2513 * Floyd (shortest path computed at initialization)
2514 * Dijikstra (shortest path recomputed all the time)
2515 * Cached Dijikstra (shortest path computed on need)
2516 All these models where contributed by Silas De Munck, and are
2517 described in his ICCS09 paper.
2519 * Simplify model declaration
2520 (less redirections, less function to write when defining a model)
2521 - Factorize stuff between models:
2524 surf_model_resource_set(model)
2525 surf_model_resource_by_name(model, name)
2526 - Unify the types of models in s_surf_model_t (using an union)
2527 - Embeed fields of common_public directly into s_surf_model_t
2528 - Rename model methods:
2529 action_free ~> action_unref
2530 action_change_state ~> action_state_set
2531 action_get_state ~> action_state_get
2532 - Change model methods into functions:
2533 (model)->common_public->action_use ~> surf_action_ref
2535 * Implement a generic resource; use it as ancestor to specific ones
2536 (allows to kill duplicated code in models)
2537 Drawback: timer command don't need no name nor properties;
2538 workstation_CLM03 don't need no properties
2539 (but I guess we can live with those few bytes wasted)
2541 * Improve the action object model
2542 - implement a constructor avoiding dupplicated code about field
2543 initialization in generic_action part.
2545 * Kill the SDP model: it has an external dependency, is deprecated
2546 in flavor of modern lmm models, and didn't compile since a while
2549 * Relocation of the context module from XBT to SIMIX.
2550 (the context were decoupled from the simix processes, duplicating a lot of code)
2551 => a lot of code was factorized
2552 - less overhead is introduced during scheduling
2553 - simpler API for the context factory
2554 - the logic for process creation,destruction and manipulation was simplified
2555 * Simplification of the s_smx_process_t data structure.
2556 => accesing the simix level data associated to a process is faster now,
2557 and the code is a lot more readable.
2560 * Implement some more MPI primitives:
2561 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2562 -implementation: Bcast: flat or 2-ary tree (default),
2563 Barrier: 4-ary tree,
2565 Allreduce: Reduce then Bcast
2566 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2567 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2568 Alltoallv: flat tree, like ompi
2570 * Add support for optimized collectives (Bcast is now binomial by default)
2571 * Port smpirun and smpicc to OS X
2574 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2575 Shout out if you used it.
2578 * Display the list of still queued messages in SG mode when existing
2582 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2583 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2584 * Add xbt_str_from_file(FILE*)
2585 * Add xbt_dict_get_key achieving a linear reverse search
2586 * Remove the context module
2588 Portability report of this version:
2589 * Main portability targets:
2590 - Linux(debian)/x86/context
2591 - Linux(debian)/x86/pthreads
2592 - Linux(debian)/amd64/context
2593 - Linux(debian)/amd64/pthreads
2594 On these, we still have the eratic breakages of gras/pmm and
2595 amok/saturate_sg reported in previous version. We still think
2596 that the tests are the cause of the fault, not the tested code.
2598 - Mac OSX Leopard/x86/context
2599 Still false negative in tesh autotesting.
2600 Smpi still fails, but this time because readlink does not accept -f
2601 Everything seems to work properly beside of that.
2604 - AIX version 5.3 (only tested contexts this time)
2605 Smpi still fails there because mktemp is not installed.
2606 Everything seems to work properly beside of that.
2608 I managed to compile it for the first time, but several breakages.
2609 Won't delay the release for this exotic platform.
2611 * Windows: it's still lagging behind. If you want to help, please
2614 Timing report of this version:
2615 This version seem to be more than 5% faster than 3.3.1 (on linux
2616 64bits with contextes). The gain is less than expected, we are
2617 investigating this for next release.
2619 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2621 SimGrid (3.3.1) stable; urgency=low
2624 * Implement a --cfg-help to show existing configuration variables
2625 * Build chain do not require doxygen in maintainer mode
2628 * fix a bug on struct sizeof computation, which prevented the
2629 exchange of arrays of structs in some conditions
2630 - added a regression test about this in datadesc_usage
2631 * Allow the exchange of 0-long dynamic vectors.
2632 - for that, use -1 as indicator of dynamic size instead of 0
2633 - This implied to change any size from unsigned long to long,
2634 reducing a bit communication abilities, but I guess that with
2635 64bits being quite common, this is more than enough.
2636 - This also induce a protocol change, thus bumping network protocol
2637 version from 0 to 1 (if we have external users, we have to get
2638 clean on that point too ;)
2639 - added two regression tests about this in datadesc_usage
2640 * Be more verbose when propagating local exceptions
2641 This helps debugging.
2642 * Display the status of simulated processes when receiving SIGINT in
2646 * Allow to control the simulation from a trace file.
2647 New functions MSG_action_register() and MSG_action_trace_run()
2648 The first one allows to associate a function execution to each
2649 kind of action while the second one parses a trace file and
2650 triggers the corresponding actions within the system.
2651 For now, only a toy example is provided in examples/msg/actions
2652 * Add an exemple of process migration in examples/msg/migration
2653 * Fix a bug in task exchange which broke MSG_task_get_sender()
2654 Add a teshsuite regression test for that.
2655 [Bug: if MSG_task_get_sender() is called after sender exit,
2657 * Fix a bug which prevented suspend/resume to work properly
2658 * Display the status of simulated processes when receiving SIGINT
2659 This fixes a regression of v3.3. due to the introduction of SIMIX
2660 * Bug fixing in failure management:
2661 - trace could not start by a failure at time 0
2662 - failure during communications were not working
2665 * Add SIMIX_process_set_name() to change the name of the current
2666 process in the log messages.
2667 * Store smx_hosts in a dict since we only retrieve them by name
2668 * Move the configuration infrastructure to surf
2671 * Move the configuration infrastructure to surf
2674 * Massive internal cleanups:
2675 - Store internal structures on processes instead of hosts (allows
2676 to have more than one process per host, in addition of being more
2678 - Cleanup the initialization/finalization process
2679 - Kill a whole bunch of unneeded synchronization:
2680 processes run in exclusive manner within the simulator
2681 - Move queues from global tables to process data fields
2683 - now accept -platform and -hostfile arguments
2684 - Pass the right rank value to processes according to the hostfile
2685 * Compile the examples by default, and use them as regression tests
2686 * Implement MPI_Wtime()
2687 * Change the reference speed to a command line option
2690 * TCP_gamma can now be specified as command line option using
2691 --cfg=TCP_gamma:10000000.0
2692 * Change the --surf-path cmd line option into --cfg=path:
2695 * Also include strbuff from xbt.h public header
2696 * xbt_ex_display(): do not free the exception after displaying
2697 This allows to do more with the given exception afterward.
2698 Users should call xbt_ex_free() themselves.
2702 Portability report of this version:
2703 * Main portability targets:
2704 - Linux(debian)/x86/context
2705 - Linux(debian)/x86/pthreads
2706 - Linux(debian)/amd64/context
2707 - Linux(debian)/amd64/pthreads
2708 These targets fail about 1/10 of times on gras/pmm, but we believe
2709 that this is because of the test, not because of simgrid.
2710 amok/saturate_sg fails even more rarely, and the test may not be
2713 - Mac OSX Leopard/x86/context
2714 The test suite still spits tons of errors because some obscure
2715 force prevents us from removing the temporary directories
2716 arguing that they still contain some metadata I've never heard of.
2717 Smpi fails because seq is not installed.
2718 Everything seems to work properly beside of that.
2721 - AIX version 5.3 (both contexts and pthread)
2722 Smpi still fails there because mktemp is not installed.
2723 XML inclusions seems rosty on AIX.
2725 * Windows: it's still lagging behind. If you want to help, please
2728 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2730 SimGrid (3.3) stable; urgency=high
2734 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2735 [Malek Cherier & Mt]
2737 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2740 Until now, GRAS were using MSG as an interface to SURF. It was
2741 quite difficult because both interface have several differences
2742 (MSG channels vs GRAS sockets were the most notable point).
2744 This also opens the gate to SMPI (which should occur soon) and speed
2745 up simulations by to 40% (even if it were not the main goal).
2747 **************************************
2748 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2749 **************************************
2750 The point is that events occuring at the exact same timestamp are
2751 not scheduled in the same order with the old and new version. This
2752 may be enough to completely change the execution of simulations in
2753 some cases. Sorry for the inconvenience.
2755 * Cleanup and upgrade the XML format to push further scalability
2756 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2758 * Improve the testing infrastructure with tesh. Now a very large part of
2759 the code is tested not only by being run but also by checking that the
2760 output match an expected output [Mt].
2762 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2763 is really less memory-demanding, which should allow you to use
2764 larger files in SimGrid [AL].
2766 * Inform valgrind about our contextes, so that it becomes usable
2767 with the default (and more effecient) version of SimGrid
2768 [contributed by Sékou Diakite, many thanks]
2771 * Introduce a listener thread in charge of receiving incoming
2772 messages from the network. It allows to overlap communication and
2773 computation but most notably, it removes some stupid deadlocks due
2774 to the fact that so far, a process could not send and receive at
2775 the same time. This made most non trivial communication schema
2777 * Convert the PIDs from long int to int to match the MSG ones (and
2778 linux ones too) [Mt]
2779 * New function: gras_agent_spawn() to launch a new process on
2780 current host. Only working in simulation for now. [Mt]
2781 * New function: gras_os_hostport() returning a constant form (ie,
2782 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2785 * Make the backtrace of exceptions more human readable [Mt]
2786 * New module: xbt/str [Mt]
2787 a ton of string utility functions (split, join, printf to a newly
2788 allocated buffer, trim, etc)
2789 * New module: xbt/hash [Mt]
2790 SHA1 hashing algorithm (more to come if needed)
2791 * New module: xbt/synchro [Mt]
2792 synchronization tools (mutex and conditions) working the same way
2793 in simulation and in real life (mainly useful for GRAS, but not
2795 * New module: xbt/queue [Mt]
2796 classical producer/consumer synchronization scheme
2797 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2798 (using the classical functions will get serialized) [Mt]
2799 * Make dictionary internal table dynamic. No need to specify its size
2800 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2802 * Make sure the log channels are organized as a tree under windows
2803 (because of ANSI C compatibility issue, any channel were child of
2804 root directly) [Mt].
2807 * Cleaned many thing in surf and fixed a few bugs [AL].
2808 * Add a nice command line configuration mechanism to compose models [AL].
2809 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2810 the previous one (KCCFLN05). It relies on something that looks like
2811 a max-min sharing mechanism but cannot be written as such. A new solver
2812 was thus designed [AL].
2813 * Add a new solver to lmm. Based on Lagrange optimization and
2814 gradient-based descent, it enables to efficiently maximise systems s.a
2816 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2818 This solver enables to propose two new network models for TCP Reno and
2819 TCP Vegas based on Low's work. These models still need to be fully
2820 tested though [Pedro Velho].
2823 * Bug fix in SD_simulate. Now the time bound given as argument is
2825 * Use the new parallel task model (ptask_L07) as default.
2826 * Use the SURF command line configuration mechanism.
2827 * 0-size tasks (for synchronization) should now work.
2829 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2831 SimGrid (3.2) stable; urgency=high
2835 We still experience issues on this platform, but we believe that at
2836 least MSG is usable.
2838 GRAS API BREAKAGE (for simplification purpose, sorry):
2839 * the gras_msgtype_by_name is not used anymore. Instead of
2840 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2841 you can write (and must)
2842 gras_msg_send(toserver, "request", &request);
2843 - If you still want to pass a gras_msgtype_t to the function (to cache
2844 the type and avoid the lookup time), use the gras_msg_send_() variant.
2845 - Impacted functions:
2846 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2847 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2848 * The callbacks are now expected to return 0 when everything went well
2849 (just like the main() function)
2851 GRAS new features and improvements:
2852 * New module mechanism where user code can use per process globals [Mt]
2853 This is similar to gras_userdata_*() functions, but for libraries. It
2854 factorize some code developped over and over in the examples and AMOK.
2855 It has still to be documented and used (only amok/peermanagement is
2857 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2858 loss when they were small enough to fit into the buffer and sent quickly
2859 enough so that they can all get received in one shoot.
2860 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2861 if not found. Use the *_or_null() variant for the old semantic.
2862 * In gras_msg_handle, do not discard messages without callback.
2863 They are probably messages to be explicitly awaited later (ie, proofs of
2864 mis-synchronization in userland since they are sent before being awaited)
2865 No big deal usually.
2866 * gras_socket_meas_send/recv: semantic changed!
2867 The numerical arguments used to be (1) the total amount of data to send
2868 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2869 messages. This was need for the fool willing to send more than MAXINT
2870 bytes on quite fat pipes.
2873 * Do really rename the hostmanagement module to peermanagement. [Mt]
2874 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2875 but this is rather new and this was documented in the module
2876 documentation (poor excuses, I admit)
2877 * Bandwidth measurement semantic changed! This follows the changes to
2878 gras_socket_meas_send/recv explained above.
2881 * A sequential mode has been added to the workstations. When a workstation
2882 is in sequential mode, it can execute only one task, and the other tasks
2883 are waiting in a FIFO. [Christophe Thiery]
2886 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2887 model for SIMDAG. [Christophe Thiery]
2888 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2889 equal to 0, which caused some bad behaviors in
2890 saturated_constraint_set_update. I now use a threshold mechanism like in
2894 * When running manually src/testall, you select specific units [Mt]
2895 testall is the result of our cunit mechanism, and should replace all
2896 the scripty thingy around since bash don't run easily on billware.
2898 * A mallocator system has been added. [Christophe Thiery]
2899 Mallocators allow you to recycle your unused objects instead of freeing them
2900 and allocating new ones.
2902 Documentation update:
2903 * FAQ reworking + New FAQs:
2904 - "Valgrind spits tons of errors!" [Mt]
2905 - "How to repport bugs" [Mt]
2906 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2907 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2908 - Communication time measurement within MSG [AL]
2909 - I experience weird communication times when I change the latency [AL]
2910 * GRAS tutorial [Mt]
2912 - an introduction to the framework and to the used communication model
2913 - an initiatic tour introducing the most proheminent features:
2915 . Lesson 0: Installing GRAS
2916 . Lesson 1: Setting up your own project
2917 o Part 2: Message passing
2918 . Lesson 2: Exchanging simple messages
2919 . Lesson 3: Passing arguments to the processes (in SG)
2920 . Lesson 4: Attaching callbacks to messages
2921 . Lesson 5: Using globals in processes
2922 . Lesson 6: Logging information properly
2923 . Lesson 7: Using internal timers
2924 . Lesson 8: Handling errors through exceptions
2925 . Lesson 9: Exchanging simple data
2926 . Lesson 10: Remote Procedure Calling (RPC)
2927 . Lesson 11: Explicitely waiting for messages
2928 . Recapping of message passing features in GRAS
2929 - A HOWTO section containing:
2930 o HOWTO design a GRAS application
2931 More are due, of course. They will come latter. In the meanwhile, you can
2932 check the examples which are still here.
2934 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2936 SimGrid (3.1) stable; urgency=high
2940 There was a stack corruption somewhere, visible only when optimizing
2941 with these versions. [Vince]
2944 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2945 module that enabled to play with Directed Acyclic Graphs. It is built
2946 directly on top of SURF and provides an API rather close to the old
2947 SG. Some old codes using SG are currently under rewrite to check that
2948 all needful functions are provided. [Christophe Thiery]
2951 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2952 extension of the classical CLM03 model that gracefully handles
2953 failures. This is now the default model for MSG and GRAS. It doesn't
2954 handle parallel tasks yet though. [AL]
2955 * Bug fix: Weights were not correctly set in the network part.
2956 WARNING: This may have resulted in incorrect results with simulations
2957 where there are more than one flow on a given link. [AL]
2960 * After a (long ?) discussion on simgrid-devel, we have decided that the
2961 convention we had on units was stupid. That is why it has been decided
2962 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2963 WARNING: This means that all previous platform files will not work as
2964 such with this version! A warning is issued to ask users to update
2966 A conversion script can be found in the contrib module of the CVS, under
2967 the name contrib/platform_generation/surfxml_update.pl [MQ]
2970 * Bug fix: Processes were started in reverse order, wrt deployment file.
2971 WARNING: if your code relies on this bug, please fix it. [AL]
2972 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2973 being executed on two different locations. [AL]
2974 * Bug fix: Failures are now better supported thanks to Derrick's tests
2975 (there was many failure situations I hadn't thought of and that weren't
2976 correctly handled). [AL]
2977 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2981 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2982 Exception occurring on server-side are propagated back to client (!).
2984 API CHANGE: the callback changed their prototype. Change:
2985 int my_handler(gras_socket_t expeditor, void *payload_data) {
2987 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2988 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2990 * New! function: gras_msg_handleall to deal with all messages arriving
2991 within a given period.
2992 * New! function: gras_socket_server_range to get a server socket in a
2993 range of port numbers (ease to avoid port number conflicts) [MQ]
2994 * New! gras processes display their backtrace when they get a SIGUSR1
2995 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2996 Sweet to debug RL processes [MQ]
3000 - Do not force experiment sizes to be expressed in kb, or it becomes
3001 impossible to measure the latency this way (needs one byte-long tests)
3002 WARNING: this changes the amok_bw_* function semantic. [MQ]
3003 - Implements the link saturation stuff. [MQ]
3004 * Peer management module:
3005 New! module factorizing code that we wrote over and over [MQ].
3008 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3009 - Test units are placed directly into the library code, they get extracted
3010 automatically and placed into the src/testall binary.
3011 - Convert most of the XBT tests to this system.
3012 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3013 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3014 live in the same C code. This required to change a little bit the API
3015 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3016 * New module: sparse graph structure with basic algorithms (this is work
3017 in progress and the API is not considered to be frozen yet). [AL]
3018 * Display more information on backtraces: source line & function names are
3019 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3020 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
3021 * Reworked a little bit some #include statements to load only required
3022 headers. Some user code that relied on SimGrid to include stdlib or
3023 stdio may need to include it by themselves. [AL]
3024 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3026 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3027 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3028 but I guess that almost nobody used those parts. [MQ]
3030 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3032 SimGrid (3.0.1) stable; urgency=low
3035 * Unfortunately, I had missed 5 misnamed functions:
3036 xbt_fifo_item_t xbt_fifo_newitem(void);
3037 void xbt_fifo_freeitem(xbt_fifo_item_t);
3038 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3039 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3040 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3041 They're now deprecated. Please use their new versions:
3042 xbt_fifo_item_t xbt_fifo_new_item(void);
3043 void xbt_fifo_free_item(xbt_fifo_item_t);
3044 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3045 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3046 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3048 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3049 * Documentation: xbt_log module unmercifully reworked [MQ]
3050 * Bugfix: there was a problem with the ending of contexts with
3051 the pthread backend. It caused some weird deadlock or behavior
3052 depending on the pthread implementation. [AL]
3053 * Bugfix: get the exceptions raised in the simulator repport where
3054 and why they come from when they are not caught in time [AL, MQ]
3057 * Bugfix: Do repport the error when two non-connected hosts try to
3058 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3061 * Add additionnal checkings on communications. Assert that two
3062 communicating hosts are connected by a set of links... [AL]
3065 * Add additionnal checkings on channel values in communication [AL]
3066 * New: MSG_task_get_source to see on which host a task was generated [HC]
3067 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3068 the number of tasks waiting to be received on channel and sent
3070 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3071 waits for the first task coming from a given host.. [AL]
3073 GRAS new functionnalities: [MQ]
3074 * Enhance the parsing macro to allow the size of multidimentional objects
3075 to be given thru annotations.
3076 * New example (and documentation): Matrix Multiplication a la RPC
3077 (as when I was young!) and fix a bunch of bugs found on the way.
3079 GRAS performance improvements: [MQ]
3081 * Reduce the amount of cbps creation/destruction by making it static to
3082 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3085 * Change libdata to a set so that we can search for stuff by ID (and thus
3086 reduce the insane amount of dict lookups)
3089 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3090 RL communication on top of SG since it's so uneffective.
3091 It may also be used for inter-thread communication in RL, one day.
3092 * Use gras_datadesc_copy() to exchange messages on top of SG
3094 - improve message exchange performance on top of SG
3095 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3096 * Don't exchange on the network the size of the used part of buffer,
3097 instead, specify the possible buffer size to read().
3099 - reduces the amount of read/write calls (one pair per exchange)
3100 - reduces the amount of exchanged data (the size)
3101 - allows to retrieve all arrived data on receiver side, if we don't need
3102 it right now (subsequent read will peek the buffer)
3103 - allows the receiver to proceed with the begining of the stream before
3104 everything is arrived
3105 - make it possible to build an iov transport (using readv/writev)
3107 - take care of the data with non-stable storage (like stacked data),
3109 * If possible, TCP send uses vector I/O (when writev() is here)
3110 - Don't use it for receive since we send data sizes and data on the
3111 same stream, so we wouldn't be able to chain large amount of chunks
3112 before having to flush the stuff to read the size.
3113 * Rework the transport plugin mechanism to simplify it and reduce the
3114 amount of pointer dereferencement when searching for the right function
3117 * I guess that now, we do almost as few system calls as possible while
3118 doing as few data copy as possible.
3120 To improve it further, we could try to send all the sizes first and then
3121 all the data (to use iov on receiving size), but it's only a partial
3122 solution: when you have 2 dimensional data, the sizes of the second
3123 dimension is data of the first dimension, so you need 3 streams.
3125 I'm not sure the potential performance gains justify the coding burden.
3127 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3129 SimGrid (3.00) stable; urgency=high
3132 * New! Give the possibility to hijack the surf parser and thus bypass
3133 MSG_create_environment and MSG_launch_application. Have a look at
3134 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3136 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3138 SimGrid (2.96) unstable; urgency=low
3143 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3144 This deprecates the xbt_error_t mechanisms.
3145 It modifies (simplifies) all XBT and GRAS API.
3146 MSG API keeps unchanged (exceptions raised by XBT are caught from
3147 within MSG and masked with existing error handling facilities)
3150 * New! Add a FATPIPE model. [AL]
3151 * New! Add a parallel task model. [AL]
3152 * New! Add automatically a loopback interface (in the default
3153 network model) if none was precised.
3156 * Bugfix: MSG_process_resume now works with the current running process.
3158 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3159 * Modification of MSG_task_get_compute_duration. Once a task has been
3160 processed, the value returned by this function is now equal to 0. [AL]
3161 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3162 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3163 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3164 of a m_task. For now, MSG_task_cancel only works with computation
3166 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3167 of the processor (in Mflop/s) regardless of the current load on the
3169 * API Change: use proper naming convention for MSG_getClock and
3170 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3172 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3173 This function changes the priority of a computation task. This priority
3174 doesn't affect the transfer rate. A priority of 2 will make a task
3175 receive two times more cpu power than the other ones. This function
3176 has been added to suit the needs of Nguyen The Loc and hasn't been that
3177 much tested yet. So if it fails, please report it and send me your code.
3179 * API Change: removed all functions and types that were marked "deprecated"
3180 since many months. Renamed MSG_global_init_args to MSG_global_init.
3182 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3184 SimGrid (2.95) unstable; urgency=low
3187 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3188 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3189 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3192 * Bugfix: Allow absolute paths for platform description files [MQ]
3193 * Bugfix: do free the variables after use. Leads to drastic performance
3195 * Implement max_duration (ie, timeouts) on resources [AL]
3198 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3200 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3201 the needed support (timeouts) [AL]
3204 * Implement measurement sockets. You can now get the bandwidth between two
3205 hosts thanks to AMOK (see below). [MQ]
3206 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3207 dynar over the network (yeah) [MQ]
3208 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3209 * simple/static token-ring example. [Alexandre Colucci and MQ]
3210 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3211 in place before [MQ]
3214 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3215 type parsing. "short" and "long" modifiers were also ignored; other
3216 modifier (such as reference level) are still ignored. [MQ]
3217 * Embeed the buffer size within the buffer itself on SG. [MQ]
3218 That way, send() are atomic and cannot get intermixed anymore (at least
3219 the ones which are less than 100k; bigger messages still have the issue)
3220 * Array size pushed by the field, not by the field type (or each
3221 and every long int will push stuff to the cbps) [MQ]
3222 * use select() to sleep since it allows to portably sleep less than one
3225 GRAS (minor cleanups)
3226 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3229 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3230 * type category 'ignored' killed as it was never used and were difficult
3232 * whether a type can cycle or not is now a flag, leaving room for more
3233 flags (as "ignored", if we feel the need one day ;)
3234 * Rename raw sockets to measurement sockets since "raw" has another
3235 meaning in networking community.
3238 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3239 and offers features not belonging to GRAS but that most applications
3240 need. One day, it may be a set of plugins loadable at runtime.
3241 * New module: bandwidth
3242 bandwidth measurement between arbitrary nodes running this module. [MQ]
3244 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3246 SimGrid (2.94) unstable; urgency=low
3248 The first beta release of SimGrid 3 !
3252 * Update the main page and the FAQ. Adding references to gforge.
3255 * Add a gras_os_getpid function.
3258 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3259 * Extend the logs so that they also print PID, hostname, date, ... if
3261 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3262 and kill the old version which were in testsuite/
3263 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3267 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3268 for users. Instead, each programming environment (GRAS, MSG,...) use it
3269 when needed to provide such a feature to users.
3270 Don't shortcut the mechanism or you will also shortcut the virtualization
3271 you need on the simulator.
3275 * Cleanups in configury with regard to compile optimization/warning flags.
3276 Also add -fno-loop-optimize to any powerpc since it's the optimization
3277 killing gcc (< 3.4.0).
3278 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3279 needed by MSG examples complications
3280 * Borrow configury beautifications from PHP
3283 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3284 warning (thanks loris for stumbling into this one).
3285 * Bugfix: stop loading private headers (gras_config.h) from the public
3289 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3290 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3291 RL would imply a DNS resolver, which is *hard* to do in a portable way
3292 (and therefore delayed).
3293 * Implement a real timer mechanism and use it in timing macros. This allows
3294 to avoid rounding errors and get a 0.000005 sec precision in timing
3295 macros. While I was at it, various cleanups:
3296 - allow to declare more than one timed section per file (fix a stupid bug)
3297 - move some private declaration to the right place
3298 - merge conditional execution and timing macros into emulation module
3299 - document the module
3300 - make sure the module cleanups its mess on gras_exit
3301 * Documentation improvements:
3302 - (new) how to compile applications using GRAS
3303 - (new) emulation support (timing macros)
3305 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3307 SimGrid (2.93) unstable; urgency=low
3309 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3312 - Use Paje properly where used. Still to be sanitized properly.
3313 - Portability fix: Add an implementation of the contexts using pthread
3317 - Add xbt_procname(): returns the name of the current process.
3318 Use it to show the current process's name in all logging.
3320 - fix detection of older flex version and the reaction, since we do
3321 depend on modern ones (we use lex_destroy)
3322 - Better separation of SG and RL in the libs: remove all simulation code
3323 from libgras. As a result, this lib is now only 200k when stripped.
3324 Some of the xbt modules may also be duplicated (two sets and such) and
3325 should be cleaned/killed before SG3.
3326 - Insist on using xlC on AIX because of weird problems involving gcc there.
3327 - Cleanup the make remote stuff. This is now done by scripts
3328 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3329 mainly for our private use, but we're working on changing them to user
3332 - Bugfix: flush the socket on close only if there is some *output*.
3333 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3335 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3338 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3340 SimGrid (2.92) unstable; urgency=low
3342 Alpha 3 on the path to SimGrid 3
3346 - New! First try of benchmarking macros.
3347 - New! First try so that gras_stub_generator generate deployment and
3348 remote compilation helpers.
3350 - Bugfix: Initialization fix in msg_test.
3354 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3356 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3358 - API Change: gras_os_sleep to take the amount of seconds as a double.
3359 Accepting an int was error prone since it was the only location where
3360 seconds were coded as such. It leaded to damn rounding errors.
3361 - Bugfix: Hard to belive that timers ever worked before this.
3363 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3365 SimGrid (2.91) unstable; urgency=low
3367 Alpha 2 on the path to SimGrid 3
3371 - Bug fix in the lmm_solver.
3373 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3374 through the function MSG_paje_output.
3375 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3376 - It is possible to bound the rate of a communication in MSG with
3377 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3378 to write it in the changelog).
3379 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3384 - Various cleanups to the autotools stuff
3385 - Begin to move Gras examples to examples/gras/
3386 - Let make distcheck work again (yeah!)
3388 - documentation overhauled using doxygen.
3389 gtk-doc-tools is dead in SimGrid now.
3390 - Automatically extract all existing logging categories, and add the list
3391 to the documentation (long standing one, to say the less)
3393 - Cleanup the known architecture table. Reorder the entries to group what
3394 should be, and use a more consistent naming scheme.
3395 (some of the test dataset are still to be regenerated)
3396 - New! Allow library to register globals on each process just as userdata
3398 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3399 lookup time (for now).
3400 Use it in msg and trp.
3401 This cleans a lot the internals and helps enforcing privacy of the
3402 headers between the gras components.
3403 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3404 - Bugfix: gras_os_time was delirious in RL.
3405 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3407 - Reenable GRAS now that it works.
3409 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3411 SimGrid (2.90) unstable; urgency=low
3413 Alpha 1 on the path to SimGrid 3
3415 * It is a long time since the last release of SimGrid. I'm sorry about
3416 that but as I had told you, I was rewriting a lot of things. I apologize
3417 to those who had been reporting bugs to me and that I had not answered.
3418 If your bug is still in the new version, please tell me. Here is a
3419 summary of the main changes.
3421 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3422 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3423 portability, speed, and a lot more but you'll figure it out later.
3424 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3425 I wanted to release the new MSG as soon as possible and I have
3426 broken GRAS, which is the reason why, for now, only MSG is fully
3427 functional. A laconic description of these projects is available
3428 in the documentation.
3430 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3431 kernel optimized for our needs (called SURF but only the developers
3432 should use it). Hence, MSG is now roughly 30 times faster and I think
3433 that by rewriting a little bit MSG, I could event speed it up a little
3434 bit more. Beside the gain in speed, it is also much easier to encode a
3435 new platform model with SURF than it was with SG. More to come...
3437 * REVOLUTION 3: I have tried to change a little as possible the API of
3438 MSG but a few things really had to disappear. The main differences
3439 with the previous version are:
3440 1) no more m_links_t and the corresponding functions. Platforms are
3441 directly read from a XML description and cannot be hard-coded
3442 anymore. The same format is used for application deployment
3443 description. The new format is described in the documentation.
3444 Have a look in tools/platform_generation. There is a tiny script
3445 that converts from the old platform format to the new one. Concerning
3446 the application deployment format, parsing the old one is tricky.
3447 I think most of you should however be able to convert your files. If
3448 it is really an issue, I can write a C code that does the conversion.
3450 2) the toolbox tbx does not exist anymore. We now have a library
3451 with much more data-structures but without the hash-tables (we have
3452 dictionaries that are much faster).
3454 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3456 *****************************************************************************
3457 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3458 * don't feel like converting the oldies. (Mt) *
3459 *****************************************************************************
3462 Version 2.90: "the long awaited one"
3463 - Finished rewriting and debugging MSG. Rewrote the documentation.
3464 - disable GRAS for now since it needs to be ported to the newest SG
3467 - Finish the port to windows (using mingw32 for cross-compile)
3470 - Main loop and datastructures of SURF. A cpu resource object is
3471 functional. Surf can thus be used to create cpu's with variable
3472 performance on which you can execute some actions.
3474 2004-11-15 Martin Quinson
3475 - Port to ARM. Simply added the alignment and size descriptions. Should
3476 work, but the ARM machines are so slow that I didn't had the opportunity
3477 to 'make check' over there yet.
3479 2004-11-15 Arnaud Legrand
3480 - Trace manager now written. It uses a heap structure and is therefore
3481 expected to be efficient. It may however be speeded up (particularly
3482 when many events occur at the same date) by using red and black
3483 trees. One day maybe...
3484 - Max-min linear system solver written. It uses a sparse matrix
3485 structure taking advantage of its expected use. Most operations are
3486 O(1) and free/calloc are called as few as possible. The computation of
3487 the minimum could however be improved by using a red and black tree
3490 2004-11-03 Arnaud Legrand
3491 - Rename every gras_* function that was in xbt/ to its xbt_
3493 - Add a heap and a doubly-linked list to xbt
3494 - Added a dichotomy to the dictionaries. make check works as well before
3495 so I assume that the patch is correct. I do not know however if things
3496 run effectively faster than before now. :)
3498 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3499 SimGrid, and the version number is bumped to 2.x
3501 2004-10-29 Martin Quinson
3502 - Introduction of the remote errors.
3503 They are the result of a RMI/RPC on the remote machine.
3504 ErrCodes being scalar values, you can't get the host on which those
3505 errors did happen. Extending the error mechanism as in Gnome is possible.
3506 No idea yet whether it is a good idea.
3508 2004-10-28 Martin Quinson
3509 - Interface revolution: the Starred Structure Eradication.
3510 I used to do typedef struct {} toto_t; and then handle *toto_t.
3511 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3512 - ???_t is a valid type (builded with typedef)
3513 - s_toto_t is a structure (access to fields with .)
3514 - s_toto is a structure needing 'struct' keyword to be used
3515 - e_toto_t is an enum
3516 - toto_t is an 'object' (struct*)
3518 typedef struct s_toto {} s_toto_t, *toto_t;
3519 typedef enum {} e_toto_t;
3520 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3521 s_toto_t) is private.
3523 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3524 so that it can set the variable to NULL. It was so for dicts and sets,
3525 it changed for dynars.
3527 - Fix a bunch of memleaks in dict_remove
3528 - Fix a bug in sg/server_socket opening: it failed all the time.
3530 2004-10-07 Martin Quinson
3531 - Speed up dynar lookup operation a bit.
3533 gras_dynar_get is dead.
3535 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3536 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3537 (which gives you the address of the stored data).
3539 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3540 copy of the data using an affectation to do the job and not a memcpy.
3542 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3543 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3545 It does not really speedup the dynar test because they are
3546 setting elements all the time (and look them seldom). But the dict does
3547 far more lookup than setting.
3549 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3551 2004-10-05 Martin Quinson
3552 - Allow to (en/dis)able the cycle detection at run time.
3554 Whether we should check for cycle or not is now a property of each
3555 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3556 datadesc_cycle_unset allow to remove this property when previously set.
3558 Note that the cycle detection is off by default since it impacts the
3559 performance. Watch the data you feed GRAS with ;)
3561 This property is hereditary. Any element embedded in a structure having it
3562 set have it set for the time of this data exchange.
3564 You should set it both on sender and receiver side. If you don't set it on
3565 sender side, it will enter an endless loop. If you forget on receiver
3566 side, the cycles won't be recreated after communication.
3568 - Header reorganization.
3569 Kill gras_private.h, each submodule must load the headers it needs.
3571 2004-10-04 Martin Quinson
3572 - Interface revolution: do not try to survive to malloc failure.
3574 Now, gras_malloc and friends call gras_abort() on failure.
3575 As a conclusion, malloc_error is not a valid error anymore, and all
3576 functions for which it was the only gras_error_t return value are
3577 changed. They now return void, or there result directly.
3578 This simplify the API a lot.
3580 2004-09-29 Martin Quinson
3581 - Re-enable raw sockets.
3582 Created by gras_socket_{client,server}_ext;
3583 Used with gras_raw_{send,recv}
3586 It should allow to kill the last bits of gras first version soon.
3588 This is not completely satisfactory yet (duplicate code with
3589 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3592 - Simplify transport plugin (internal) interface by not passing any
3593 argument to _server and _client, but embedding them in the socket
3596 2004-09-28 Martin Quinson
3597 - Finish the port to AIX.
3598 autoconf was my problem (segfault within the malloc replacement
3599 function. No idea why)
3601 2004-09-16 Martin Quinson
3602 - Fix some size_t madness on 64bit architectures.
3604 2004-09-08 Martin Quinson
3605 - Reduce the number of system headers loaded, overload some more system
3606 calls (such as malloc to cast the result of the system one, and work
3608 - Fix and reintroduce the config support
3610 2004-09-07 Martin Quinson
3611 - Source code reorganization to allow Arnaud to surf all over there.
3612 - Allow to document the logging categories.
3613 - Remove all uppercase from logging categories and useless cleanup in names.
3615 2004-08-18 Martin Quinson
3616 Version 0.6.2 (protocol not changed; API changed)
3617 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3618 gras_error_t), and NULL when not found. Functions expecting a msgtype
3619 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3620 hopefully usefull message.
3621 - Portability to prehistoric sparcs again
3623 2004-08-17 Martin Quinson
3624 Version 0.6.1 (protocol not changed; ABI not changed)
3625 - prealloc some buffers to speed things up
3627 2004-08-11 Martin Quinson
3628 Version 0.6 (protocol not changed; ABI expended)
3629 - The parsing macro can deal with the references, provided that you add
3630 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3632 2004-08-09 Martin Quinson
3633 Version 0.5 (protocol not changed; ABI changed)
3634 - Allow to off turn the cycle detection code in data exchange at
3635 compilation time. It should be at run time, but I'm short of time (and
3636 the config stuff is still broken). That way, we keep dict out of the
3637 critical path, which is good because the performance is poor:
3638 - search not dichotomial yet
3639 - dynar give no way to access their content and memcpy everytime
3640 - In composed data description (struct, ref and so on), stop foolness of
3641 keeping the subtype's ID, but store the type itself. This keeps sets out
3642 of the critical path, which is good since they rely on dynar and
3643 dictionnaries. The only loose of that is that we cannot detect the
3644 redeclaration of a structure/union with another content (but I'm not sure
3645 the code detected well this error before anyway). We still can detect
3646 the redefinition discrepancy for the other types.
3647 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3648 it breaks the code because of type-punning used all over the place).
3649 This breaks on all non-gcc architectures (for now).
3651 All those changes (plus the buffer of last time) allow me to gain 2 order
3652 of magnitude on cruel tests consisting of 800000 array of integers on two
3653 level of a hierarchical structure (200 secondes -> 4 secondes)
3656 - the selector of reference must now return the type it points to, not
3657 the ID of this type.
3659 2004-08-06 Martin Quinson
3660 Version 0.4 (protocol changed; ABI not changed)
3661 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3662 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3663 - Implement a buffer transport, and use it by default (it relies on tcp in
3664 real life and on sg in simulation).
3665 That's a bit hackish since I had a new field to the structure to store
3666 its data without interfering with the subtype ones. Inheritance
3667 is tricky in C. And that's a kind of reverse inheritance with one class
3668 derivating two classes. Or maybe a game with java interfaces. Anyway,
3669 that's damn hard in C (at least).
3670 Moreover, I got tired while trying to ensure plugin separation and
3671 genericity in SG mode. MSG wants me to do weird things, so let's go for
3672 cruel hacks (temporarily of course ;).
3673 See comment in transport_private.h:71
3674 - do not include all the _interface headers in private but in the files
3675 which really need them (to cut the compilation time when they are
3678 2004-07-26 Martin Quinson
3679 Version 0.3 (protocol not changed; ABI changed)
3680 - Major overhault of the datadesc interface to simplify it:
3681 - shorted the function names:
3682 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3683 - add a trivial way to push/pop integers into the cbps without malloc.
3684 This allows to make really generic sub_type description, which simply
3685 pop their size of the stack.
3686 - add a function gras_datadesc_ref_pop_arr() which does what users want
3687 most of the time: Declare a dynamic array (which pops its size of the
3688 stack) and declare a reference to it. Poor name, but anyway.
3689 - kill the post-send callback, add a post-receive one
3691 2004-07-23 Martin Quinson
3692 Version 0.2 (protocol changed; ABI changed)
3693 - add some testing for cpbs in the test cases, and fix some more bugs.
3694 This invalidate again the little64 data file, since I cannot regenerate
3696 - remove an awfull optimization in the logging stuff, allowing me to:
3697 - understand it again
3698 - learn gcc how to check that the argument match the provided format
3699 - fix all errors revealed by gcc after that
3700 - internal keys of dict are not \0 terminated. Deal with it properly in
3701 loggings instead of segfaulting when the user want to see the logs :-/
3703 2004-07-22 Martin Quinson
3704 - Fix some stupid bug preventing cbps (callback postit) from working
3706 2004-07-21 Martin Quinson
3707 - Some documentation cleanups
3708 - remove the useless last argument of msgtype_declare
3709 - rename the Virtu functions to fit into the 'os' namespace
3710 - move headers src/include -> src/include/gras/ and stop fooling with
3711 gras -> . symbolic link
3712 - make distcheck is now successful
3714 2004-07-19 Martin Quinson
3716 - Build shared library also
3717 - Install html doc to the right location
3718 - stop removing maintainer files in make clean
3719 - build tests only on make check
3721 2004-07-13 Martin Quinson
3723 - No major issue in previous version => change versioning schema
3724 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3725 corresponding dataset.
3727 2004-07-11 Martin Quinson
3729 - Get it working with any kind of structure (we can compute the padding
3730 bytes remotely for all the architectures I have access to)
3731 - Implement the structure parsing macro (still not quite robust/complete)
3732 - Improvement to the remote testing toysuite
3734 2004-07-10 Martin Quinson
3735 [autoconf mechanism]
3736 - get ride of a bunch of deprecated macros
3737 - actually run the test for two-compliment, not only compile it :-/
3738 - test whether the structures get packed (and bail out if yes. Damn.
3739 Alignment is a serious matter)
3740 - test whether the structures get compacted (but respecting the alignment
3741 constraints of each types)
3742 - test whether the array fields of structures can straddle alignment boundaries
3744 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3745 (took me 2 hours to find that bug, looking at the wrong place)
3746 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3747 before sending/receiving and are used to compute the offsets of fields
3748 - Given that padding size depend even on compiler options, keep track of
3749 alignment and aligned_size only for the current architecture. Not a big
3750 deal since we send structure fields one after the other (seems
3752 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3753 program, let it work on linux/gcc/little32. portability todo.
3755 2004-07-08 Martin Quinson
3756 - import and improve remote compilation support from FAST
3757 - make sure make check works on half a dozen of machines out there
3759 2004-07-07 Martin Quinson
3760 Let's say it's version 0.0.3 ;)
3761 - Implement conversions (yuhu!)
3762 - Let it work on solaris (beside conversion, of course)
3763 - Stupid me, using rand() to generate the conversion datatests in not wise.
3765 2004-07-06 Martin Quinson
3766 - Let make dist work, since I'm gonna need it to compile on remote hosts
3767 - Let Tests/datadesc_usage write the architecture on which the file was
3768 generated as first byte.
3769 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3770 architecture descriptions.
3771 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3772 on those architectures.
3773 - Optimization: send/recv array of scalar in one shoot
3775 2004-07-05 Martin Quinson
3776 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3778 - Plug a whole bunch of memleaks
3779 - each process now have to call gras_{init,exit}. One day, their log
3780 settings will be separated
3781 - Continue the code factorisation between SG, RL and common in Transport.
3783 2004-07-04 Martin Quinson
3785 - Redistribution between SG and RL.
3786 We wanna have to accept in SG, so move accepted related parts of RL in
3787 the common part. (more precisely, the dynar of all known sockets is no
3788 more a static in transport.c, but part of the process_data)
3790 [gras_stub_generator]
3791 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3793 2004-07-03 Martin Quinson
3794 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3795 them all up in one shot)
3797 - Fix the ugly case of reference to dynamic array.
3798 - New (semi-public) function gras_datadesc_size to allow the messaging
3799 layer to malloc the needed space for the buffer.
3801 - gras_socket_close now expect the socket to close (and not its address to
3802 put NULL in it after it). This is because the socket passed to handlers
3803 is one of their argument (=> not writable).
3805 - propagate the interface cleanup from last week in datadesc, ie remove a
3806 superfluous level of indirection. User pass adress of variable
3807 containing data (both when sending and receiving), and not of a variable
3808 being a pointer to the data. Let's say that I like it better ;)
3809 The price for that is constructs like "int msg=*(int*)payload" in
3810 handlers, but it's a fine price, IMHO.
3812 - Let it work in RL (yuhu)
3814 2004-06-21 Martin Quinson
3816 - porting SG plugin and SG select to new standards (works almost).
3817 - plug memleaks and fix bugs around.
3820 - cleanup the prototype of data recv and force users to specify when they
3821 want to handle references to objects. Test case working even for cycles.
3822 - plug memleaks. Valgrind is perfectly ok with this.
3824 2004-06-12 Martin Quinson
3826 - cleanup the separation between plugin and main code in plugin creation
3828 2004-06-11 Martin Quinson
3830 - Reput hook for raw sockets, needed for BW experiments
3831 - kill a few lines of dead code
3832 [Data description] Interface cleanup
3833 - gras_datadesc_by_name returns the searched type or NULL.
3834 That way, no variable is needed to use a type desc once, which makes
3836 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3837 two parameters were strings (field name, type name), leading to
3839 [Dicos] Interface cleanup
3840 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3841 This is consistant with the dynar API.
3843 2004-04-21 Martin Quinson
3845 - Porting to new standards.
3847 - interface cleanup.
3848 There is no bag anymore, no need to take extra provision to mask the
3849 pointers behind "ID".
3850 Better splitup of functions between files create/exchange/convert.
3851 This is still a bit artificial since convert and receive are so
3852 interleaved, but anyway.
3854 - add a queued message list to procdata (the ones not matching criteria
3856 - factorize some more code between SG and RL wrt procdata
3858 - use gras_exit in example to track memleaks
3859 - get rid of gs_example now that GS is properly integrated into gras
3860 - update run_test to integrate the latest tests (datadesc)
3862 - rename WARNINGn macros to WARNn since it prooved error-prone
3864 2004-04-19 Martin Quinson
3866 - register init/exit functions within gras module mechanism
3867 - send/receive function.
3868 Convertion is not implemented, but short-cutted if not needed.
3869 struct/array elements are sent one by one (instead of block-wise), but
3870 nobody really cares (yet). Get a prototype before optimizing.
3871 - tests (using a file socket) for DD send/receive on:
3872 - base types: int, float
3873 - array: fixed size, string (ie ref to dynamic string)
3874 - structure: homogeneous, heterogeneous
3875 - chained list, graph with cycle
3876 Believe it or not, valgrind is not too unhappy with the results. The
3877 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3881 - Bugfix when using a filename explicitely (instead of '-')
3883 2004-04-09 Martin Quinson
3885 - factorize more code between RL and SG in socket creation
3886 - Complete the implementation and tests of:
3888 o file (only in RL, and mainly for debugging)
3890 I lost 3 days to design a portable address resolver, and then decided
3891 that the prototype mainly have to run on my box.
3892 Addressing portability too early may be like optimizing too early :-/
3894 - use gras_init in the Tests instead of the crappy parse_log_opt
3895 (the latter function is removed)
3896 [Conditional execution]
3897 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3898 [Code reorganisation]
3899 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3900 Build examples against the RL library, since there is no way to disable
3901 its creation for now.
3903 For information, the beginning of coding on GRAS was back in june
3904 2003. I guess that every line has been rewritten at least twice since