1 SimGrid (3.23) NOT RELEASED (Release Target: June 21. 2019, 15:54 UTC)
4 - Stop setting random seed with srand48() at initialization.
5 - Use addr2line as a fallback for stacktraces when backtrace is not available.
8 - New log appenders: stdout and stderr. Use stdout for xbt_help.
12 - 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
13 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
14 - MPI interface is now const correct for input parameters
17 - Remove option 'model-check/record': Paths are recorded in any cases now.
19 Fixed bugs (GH=GitHub; FG=FramaGit):
20 - FG#10: Can not use MSG_process_set_data from SMPI any more
21 - FG#11: Auto-restart actors forget their on_exit behavior
22 - GH#133: Java : a process can run on a VM even if its host is off
23 - GH#326: Valgrind-detected error for join() when energy plugin is activated
25 ----------------------------------------------------------------------------
27 SimGrid (3.22) April 1. 2019
29 The Easter Christmas Release (this one is somewhat late).
32 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
33 - Not complete yet: asynchronous activities (amongst others) are still missing
34 - Still ongoing: the interface may change in the future. We need more
35 testers! Please report any glitches.
36 - No new project using Java should start now. Please switch to Python.
39 - Some of the internal cleanups may lead to speed improvements:
40 - The hard limitation on the amount of simulated actors with
41 Java+Mac was removed. Now, the available memory is the only limit.
42 - Our refcounting was tidyied, leading to 10% speedups in some cases.
43 - We are still working on making our code robust to the actor kills
44 and hosts' churn. Things are improving, but it's not perfect yet.
45 - Replaced our own code to display a backtrace (that was forking addr2line)
46 with the Boost.Stacktrace library.
47 You won't see your backtraces without this optional dependency.
48 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
49 - Stop setting random seed with srand() at initialization.
52 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
53 XML file version remains 4.2 since old files are still compatible.
56 - Process termination which was broken at version 3.21 has been repaired.
57 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
58 - Hide the examples into examples/deprecated. New users should use Python.
61 - Drop MSG_process_create_from_stdfunc() from the API.
62 This C++ function was a pimple in the C API, made necessary at some
63 point by the Java bindings. This is fixed now.
64 - Hide the examples into examples/deprecated. New users should use S4U.
65 - MSG_process_create and MSG_process_attach now crash if the host used
66 to run this process is off.
67 - Fix the protype of MSG_process_on_exit()
68 Now use (int,void*) callbacks instead of (void*,void*) ones.
69 The implementation was ways too messy to actually work, I guess.
72 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
73 too often and provide more accurate timings.
74 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
75 - MPI_Alltoallw support
76 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
77 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
78 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw, MPI_Ireduce,
79 MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter,
80 MPI_Ireduce_scatter_block, with fortran bindings.
81 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
82 support, MPI_Comm_get_name, MPI_Comm_set_name
83 - Basic implementation of generalized requests (SMPI doesn't
84 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
87 - Drop sg_cmdline. Please use xbt_cmdline instead.
88 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
89 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
90 OsSemaphore is implemented in a portable way with C++11 threads.
91 This should allow much more threads to be created at the same time,
92 allowing Mac and Java users to simulate many more actors.
93 - Implement the 'thread' factory with std::thread instead of xbt ones.
94 It is not possible to set the stack size with threads anymore, but
95 -fsplit-stack is the way to go nowadays when using threads.
96 - Drop the xbt_os_thread_t module (now unused)
97 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
98 - Drop xbt_str_join_array().
99 - Drop cunit, use Catch2 instead.
102 - Many cleanups in the kernel::activity namespace. This was long
103 overdue, and shall open the path to many future endeavors.
106 - #132: Java: a process can not shut down its own host
107 - #220: S4U: detached send causes exception if sender terminates soon after sending
108 - #261: Document the parameters of parallel execution's constructor
109 - #300: [s4u] BarrierPtr is missing
110 - #314: SMPI args internal cleanup
111 - #316: Fix a bug related to the CPU utilization of multi-core VM
112 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
113 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
114 - #325: Turning off a host has different behavior on sleeping actors and computing actors
116 ----------------------------------------------------------------------------
118 SimGrid (3.21) October 3. 2018
120 The Restarting Documentation (TRD) Release.
123 - Convert the user manual to Sphinx for improved usability.
124 Unstable version now hosted on https://simgrid.frama.io/simgrid
125 Many glitches remain; stay tuned.
126 - Restrict the Doxygen reference API to the installed header files.
127 (documenting all internals this way is overkill + doxygen scales badly)
128 - New tutorial on S4U (highly inspired from the old MSG one, but with
129 a git repository to fork as a starting point, and a docker image)
130 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
131 - The Developper manual is still to be converted (not compiled until then)
132 - Some parts are still missing in this conversion (such as the
133 platform and deployment sections) while others were blindly converted
134 and would need more love (such as the configuration flags).
135 Things will certainly further improve in the future releases.
138 - Fully reimplement auto_restart mechanism that was utterly broken
139 (fix #22, #131 and #281 that were all very related)
140 - Implement semaphores (fix #303 and #291)
141 - When creating an actor from a function and its parameters,
142 move-only parameters are not allowed anymore, as it would prevent
143 the actor to be restartable if its parameters are consumed on run.
144 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
145 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
146 IO operations that can be started, waited for, or canceled as a regular
148 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
152 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
153 - New host variable: 'core_count'
156 - Remove xbt_os_thread_specific features
157 - Remove portability wrapper to condition variables
158 - Remove xbt_os_thread_yield()
161 - MPICH collective selector now mimics MPICH 3.3b
162 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
163 - MPI_Init(NULL,NULL) is now allowed
164 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
165 - Command-line arguments should now be handled properly in Fortran simulations
166 - MPI Attributes and related callbacks should work in Fortran
167 - Apps using C + Fortran should now work
168 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
169 - Support MPI_Error_String
172 - Due to an internal bug, Msg.run() must now be your last line.
173 We hope to fix it in a future release, and we are sorry for the inconvenience.
176 - #22: Process autorestart seem to only work with CAS01 cpus
177 - #56: Feature request: dynamic SMPI replay
178 - #93: simgrid should not eat --help
179 - #111: How to change the loopback link characteristics?
180 - #116: Platform cluster radical
181 - #117: Platform cluster/host order
182 - #129: Replay_multiple on steroid
183 - #131: Java : autostart does not work
184 - #143: Setting a breakpoint at a given time
185 - #235: Network energy models should be integrated
186 - #264: Add ptask L07 resource tracing
187 - #271: Dynamic and manual replay of SMPI traces
188 - #279: Breakpoints option uses time deltas instead of absolute time
189 - #281: Daemonized actor no longer daemonized after an auto-restart
190 - #282: TIT convertor erase traces when using absolute path in the trace list file
191 - #285: segfault when a s4u actor kills itself with msg process tracing activated
192 - #286: Pajé traces are not exposing the number of cores
193 - #287: Command-line parsing should stop when encountering '--'
194 - #288: MPI_Init(NULL, NULL)
195 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
196 - #290: Method to check if a mailbox has a message ready to be consumed
197 - #291: [s4u] semaphores do not seem available yet
198 - #294: Alltoall collective implementation of mpich
199 - #295: Starting a Host as OFF is not supported
200 - #296: DTD too permissive
201 - #299: Add a s4u barrier example
203 ----------------------------------------------------------------------------
205 SimGrid (3.20) Released June 24. 2018
207 The proxy snake_case() release.
210 - The whole API is now using snake_case() instead of mixing it with camelCase.
211 Compatibility wrappers in place for one year.
212 - We took the opportunity to increase the consistency of the API.
213 Please report any remaining glitches.
214 - New function simgrid::s4u::Host::get_actor_count:
215 Returns the number of actors running on a specific host.
218 - SMPI is now tested with ~45 proxy apps from various sources, with none or
219 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
220 - Replay: The replay file has been re-written in C++.
221 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
222 supported. They are stored in the trace and used when replayed.
223 - Basic support of MPI_Cancel. Robustness not guaranteed.
224 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
225 (only for MPI_COMM_TYPE_SHARED).
226 - New option: smpi/privatize-libs, to add external shared libs to be privatized
227 by SMPI. They will be copied locally and loaded separately by each process.
228 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
229 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
230 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
231 with internal index optind with multiple processes. Only works if MPI_Init has
233 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
234 syntax without preprocessing tricks.
237 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
238 behavior tracing is no longer limited to MSG
239 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
240 this means that traces now need to be updated or re-obtained; both Irecv
241 and Isend lines in your traces have now in total 5 mandatory fields:
242 <rankid> <command> <to/from rankid> <tag> <size>
243 To update your traces, it suffices to add a 0 for the tag here.
244 - TIT now also supports waiting for a distinct request via MPI_Wait.
245 Wait/Test now wait for a specific request, not just the last one that was
246 issued. This unfortunately means another update, because we need to
247 identify which request you want to wait for. We do this via the
248 triplet (sender, receiver, tag), which needs to be added:
249 <rankid> <command> <sender> <receiver> <tag>
250 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
251 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
254 - The deprecation of MSG is ongoing (but this should not impact you).
255 Many MSG functions are now simple wrappers on the C API of S4U. If
256 you wish to convert your code to S4U, find the S4U counterparts of
257 your MSG calls in src/msg/msg_legacy.cpp.
258 - MSG can still be used, but won't evolve anymore.
261 - Allow to run the Link energy plugin from the command line with
262 --cfg=plugin:link_energy
263 - Rename Energy plugin into host_energy
264 - Rename Load plugin into host_load
267 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
268 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
269 - Improve performance when handling timeouts of simix synchros.
272 - Config: the C API is now deprecated (will be removed in 3.23), and
273 the C++ API has been slightly improved.
276 - Fix several build issues on OSX.
277 - Move simgrid_config.h to simgrid/config.h (old header still working)
280 - #143: Setting a breakpoint at a given time
281 - #150: Inconsistent event names in SMPI replay
282 - #258: daemonized actors hang after all non-daemonized actors have completed
283 - #267: Linker error on unit_tmgr
284 - #269: SMPI: tracing of MPI_Wait/all/any broken
285 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
287 ----------------------------------------------------------------------------
289 SimGrid (3.19.1) Released March 22 2018
291 The Fixed ABI Release.
293 As you may know, we are currently refactoring SimGrid in deep.
294 Upcoming SimGrid4 will be really different from SimGrid3: modular,
295 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
297 Our goal is to smooth this transition, with backward compatibility
298 and automatic update paths, while still progressing toward SimGrid4.
300 SimGrid remains open during works: The last pure SimGrid3 release was
301 v3.12 while all subsequent versions are usable alpha versions of
302 SimGrid4: Existing interfaces remain unchanged, but the new S4U
303 interface is budding and the internals are deeply reorganized.
305 Since 2015, we work hard to reduce the changes to public APIs. When
306 we need to rename a public library symbol in S4U, we let your
307 compiler issue an explicative warning when you use the deprecated
308 function. These messages remain for four releases, i.e. for one full
309 year, before turning into an error. Starting with v3.15, your can
310 also adapt to API changes with the SIMGRID_VERSION macro, that is
311 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
313 Starting with this v3.19.1, our commitment to reduce the changes to
314 the public interfaces is extended from the API to the ABI: a program
315 using only MSG or SimDag and compiled against a given version of
316 simgrid can probably be used with a later version of SimGrid without
317 recompilation. We will do our best... but don't expect too much of
318 it, that's a really difficult goal during such profund refactoring.
320 The difference between v3.19 and v3.19.1 is that the former was
321 accidentally breaking the ABI of MSG, while the later is restoring
324 S4U and kernel APIs will still evolve until SimGrid4, with one-year
325 deprecation warnings as currently. In fact, cleaning up these
326 interfaces and converting them to snake_case() is one release goal of
327 v3.20. But don't worry, we are working to smooth this upgrade path.
329 In summary, new projects should start with S4U to benefit of the
330 future, but old MSG projects should still be usable with no change.
332 ----------------------------------------------------------------------------
334 SimGrid (3.19) Released March 20 2018
336 The Moscovitly-cold Spring release.
338 SMPI user-visible improvements:
339 - Support of MPI_Type_create_subarray
340 - Unimplemented MPI calls now abort (instead of issuing a warning)
341 Some harmless calls around errhandlers still only issue a warning.
342 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
343 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
344 is now possible (mmap privatization was refactored to this end).
345 Adding new SMPI instances should also be technically doable now.
347 SMPI internal cleanups:
348 - Removed index notion from SMPI -- no more "getPid() - 1"!
349 In tracing files, actors are called rank-<PROCESS-ID> but these are
350 NOT the real ranks (tracing several communicators is not supported yet).
351 - Deprecate smpi_process_index() (will be removed in 3.22)
352 - Better testing on classical proxy apps, through an external project:
353 https://github.com/simgrid/SMPI-proxy-apps/
356 - Use a graphical TOC to make it easier to find the documentation you need
357 - Revamp the MSG tutorial
360 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
361 - Remove parameter of MSG_process_killall().
362 Resetting the PID was bogus anyway (several actors could have the same PID).
365 - Execution->setHost() can be called after start(), to migrate it.
366 - Comm::test_any() is now implemented.
367 - s4u::Actor now has onCreation() and onDestruction() signals.
368 - Install some kernel header files for the users' plugins and more.
369 Warning, their API is really not stable yet.
370 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
371 include/simgrid/kernel/routing/*Zone.hpp
374 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
376 Build System and other cleanups:
377 - Remove unused run-time parameter "tracing/onelink-only".
378 - SimGrid now works with intel compilers (MC still buggy)
379 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
380 - tesh kills subprocesses on timeout (but not on windows)
383 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
386 - #194: Feature request: simgrid::s4u::Comm::test_any()
387 - #245: migrating an actor does not migrate its execution
388 - #253: Feature Request: expose clusters as objects
389 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
390 - #255: Tesh broken on Windows
391 - #256: Modernize FindSimGrid.cmake
392 - #257: Fix (ab)use of CMake install
394 ----------------------------------------------------------------------------
396 SimGrid (3.18) Released December 24 2017
398 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
400 MSG IS NOW DEPRECATED. Please use S4U for new projects.
401 - Support for MSG should not be removed from SimGrid before 2020,
402 but future evolutions will be limited to the integration of
403 user-provided patches.
406 - s4u::onDeadlock() signal, run before stopping the simulation.
407 - s4u::Exec: asynchronous executions (abstraction of background
409 - s4u::Host->getLoad() returns the achieved speed in flops/s
412 - Link::name() is deprecated and will be removed in v3.21, use
413 Link::getCname() instead.
414 - Mailbox::getName() changed to return a std::string, use
415 Mailbox::getCname() instead to get a char*.
416 - Storage::getName() changed to return a std::string, use
417 Storage::getCname() instead to get a char*.
418 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
419 to have both allocation and deallocation of the map in user space.
422 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
423 The amount of flops remaining to do is a vector, not a scalar.
424 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
425 name implies on both sequential and parallel tasks.
426 - Both changes fix GitHub's #223 using PR #237 as a basis.
427 Thanks Michael Mercier.
428 - Most examples were converted to S4U and hidden elsewhere at tests.
429 Rationale: we still want MSG to work; we want newcomers to use S4U.
432 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
433 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
436 - Switch to the faster dlopen privatization mechanism by default
437 - Documentation improvments
440 - Live migration is getting moved to a plugin. Dirty page tracking is
441 the first part of this plugin. This imply that VM migration is now
442 only possible if one this function is called:
443 - C/MSG: MSG_vm_live_migration_plugin_init()
444 - C/C++: sg_vm_live_migration_plugin_init()
445 - Java: Msg.liveMigrationInit()
446 For C and C++, "simgrid/plugins/live_migration.h" also has to be
450 - Define class simgrid::xbt::Path to manage file names.
451 - Removed unused functions:
452 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
453 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
454 xbt_graph_export_graphviz()
455 - xbt/str.h: xbt_str_join()
456 - Remove unused datatypes:
457 - xbt/heap.h: use std::priority_queue or boost::heap instead
458 - xbt/swag.h: use boost::intrusive::list instead
461 - New link_energy plugin for the consumption of the links.
462 - All of the operations on files and storage contents have been
463 packaged into a plugin (src/plugins/file_system). The current
464 public interface can be found in
465 include/simgrid/plugins/file_system.h
466 To use these functions you now have to initialize the plugin by
467 calling MSG_storage_file_system_init() just after calling
468 MSG_init() or sg_storage_file_system_init() just after creating
472 - Remove the undocumented/untested tag <include>
475 - Remove viva specific tracing as the tool is no longer maintained
478 - #248: Kill of finished processes leads to segfault
479 - #240: xbt_cond_wait_timeout should gracefully return for C apps
480 - #239: please implement signal s4u::onDeadlock()
481 - #230: segfaults when exit() without run()
482 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
483 - #223: MSG_task_get_flops_amount() not working with parallel tasks
484 - #222: Actor::kill() doesn't really kill and segfaults
485 - #221: odd LMM warning when killing an actor
486 - #120: Memory leak when the processes are forcefully killed
488 ----------------------------------------------------------------------------
490 SimGrid (3.17) Released October 8 2017
492 The Drained Leaks release: (almost) no known leaks despite the tests.
494 Even very long simulations will preserve your memory: our testsuite
495 stresses SimGrid for over 45mn on fast machines for a coverage of
496 over 80%, and there is only one single known leak, of about 4kb.
500 - Comm.detach(): start and forget about asynchronous emission. A cleanup
501 handler may be given to free resources if the comm cannot be completed.
502 - this_actor::send(mailbox) is now mailbox->put()
503 - New: simgrid::s4u::Comm::wait_all()
504 - New: Host.execute() for remote executions.
507 - Unused option network/sender-gap is removed.
510 - Deprecate MSG_task_isend_with_matching(): This unused feature
511 really complicates our internals. Will be removed in v3.20.
514 - Improved context termination. It is now possible to free resources, even
515 when a process is forcibly killed.
516 - ContextBoost: add support for Boost versions above 1.61.
519 - Bring back run-time option --cfg=exception/cutpath to remove exception
521 - Removed unused functions:
522 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
523 xbt_str_rtrim(), xbt_str_trim().
524 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
527 - Removed header files obsolete since SimGrid 3.12:
528 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
530 - #3: SD_exit should be made optional
531 - #120: Memory leak when the processes are forcefully killed
532 - #159: Threading test regression in Actor refcounting
533 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
534 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
535 - #186: Actor::killAll() segfaults if some process is blocked on wait()
536 - #191: VM migration and pstate
537 - #192: Updating the energy consumptions of all hosts crashes with VMs
538 - #195: All actors have PID=0 in the logs
539 - #204: Sometimes segfault with thread contexts and mmap privatization
540 - #222: Actor::kill() doesn't really kill and segfaults
541 - #225: Actor::kill() doesn't really kill when victims are doing a join()
543 ----------------------------------------------------------------------------
545 SimGrid (3.16) Released June 22. 2017.
547 The Blooming Spring Release: developments are budding.
550 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
551 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
552 - Java JAR file should be finally fully working on Mac OSX too.
554 XML platforms: Switch to platform v4.1 format.
555 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
556 - <zone> can be used as a synonym for the now deprecated <as>
557 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
558 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
559 - <actor> can be used as a synonym for the now deprecated <process>
560 - state_file and avail_file periodicity is now easier to express
561 (check the documentation) the old behavior should still work.
562 * Storage: not backward compatible, but it seems that nobody ever used it
563 - Remove attribute 'content_type' of <storage_type>: was never used
564 - Make attribute 'model' of <storage_type> optional: for future usage
565 - Remove Bconnection model property: was never (in)validated. Replaced by
566 the maximum of the read and write bandwidth as a resource constraint.
569 - New and Backwards Compatibility break:
570 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
571 When one wants to get the list of tasks whose states have changed during a
572 simulation round, s/he has to allocate and free a dynar and use it as
573 argument to this function. The former SD_simulate (double how_long)
577 - Allow multicore VMs, with the correct sharing (unless you overcommit)
578 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
581 - New function to update the consumption of all hosts at once.
582 - Fix the model for multi-core hosts, linear in the amount of busy
583 cores with an abnormality for pIdle (see paper).
584 BUG: Still not satisfactory for ptask on mono-cores.
587 - The netzone are now available from the MSG API.
588 The old names still work, but are now deprecated.
591 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
592 - New option: smpi/keep-temps to not cleanup temp files
593 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
594 mallocs mode is used, allows to change the size of the fake file used
595 (default 1MB), to potentially limit the number of mappings for large runs.
596 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
597 - Fortran ifort and flang compilers support
598 - New RMA calls supported (experimental):
599 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
600 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
601 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
602 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
603 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
604 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
607 - New: this_actor::isMaestro() returns whether we are in kernel mode.
608 - Behind the scene, ActivityImpl are now automatically refcounted.
609 Removing this nasty bottleneck will greatly simplify our internals.
612 - Replay: New function xbt_replay_action_get():
613 Retrieve the function previously associated to an event type.
614 - DROPPED FUNCTION: xbt_str_varsubst()
615 - DROPPED MODULE: strbuff. We don't need it anymore.
616 - DROPPED MODULE: matrix. We don't need it anymore.
617 - DROPPED MODULE: lib. We don't need it anymore.
619 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
621 SimGrid (3.15) Released March 22. 2017
623 The Spring Release: continuous integration servers become green
624 We fixed even the transient bugs on all target architectures:
625 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
626 Windows, FreeBSD, NetBSD.
628 - Rename NetCards to NetPoints.
629 This was intended to help NS3 users, but that's not a netcard.
630 That's a point in the routing algorithm, let's avoid wrong simplifications.
633 - New: MSG_process_yield(). Stop and yield to other processes.
634 - New: MSG_process_daemon(). Daemon processes are automatically killed
635 when the last non-daemon process terminates
636 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
637 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
638 to make room for the upcoming network energy plugin.
639 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
642 - Ensure that an actor can kill itself with Process::exit()
643 - Kill the obscure NativeException. Nobody want to survive the issues
644 it denotes, so use JniException that is a RuntimeException (not to
645 be caught explicitly).
646 - Partial bug fix in initialization. SimGrid flags on command line were
647 consumed at C level but stayed in the original Java String[] args.
648 This could mess users' args[i] if SG flags were not put at the end of
650 The SimGrid flags are now removed from the Java arguments. However,
651 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
652 if args.length is greater than the number of YOUR OWN ARGUMENTS.
653 It might be if you have --log or --cfg flags in the command line.
654 - Fix numerous memleaks all around the place. In particular, around VMs.
658 - s4u::onPlatformCreated: right before the simulation starts
659 - s4u::onSimulationEnd: right after the main simulation loop
660 - s4u::onTimeAdvance: right after a clock change (time is discrete)
661 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
662 event from the availability_file changes the avail speed.
663 - Links are now usable from s4u
664 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
665 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
666 - New examples: The conversion of MSG examples to S4U has begun
667 - Actors: create, kill, migration, and suspend
668 - Applications: master-worker and token-ring
669 - Action replay: communications and storage
670 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
673 - Backwards Compatibility breaks
674 - The SD_route_get_list and SD_route_get_size functions have been removed.
675 They are replaced by the unique
676 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
677 The route from src to dst is built in the links dynar whose size can
678 be obtained with xbt_dynar_length.
679 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
680 sg_host_route_bandwidth and sg_host_route_latency.
681 Macros ensure the backwards compatibility, but you should fix your code
684 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
685 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
686 - Bump our claim of support from MPI 1.1 to MPI 2.2.
687 We don't support 100% of it, but it should be enough. Contact us if not.
688 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
689 naive and probably inaccurate)
690 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
691 mapped onto a unique small file using some system magic.
692 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
693 - Fortran: cleanups, fixes, support of user-added operations
694 - RMA: MPI_Accumulate are applied in correct order.
695 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
696 - Better support for MPI_IN_PLACE
697 - Support for MPI_Win attrs and keyvals.
698 - Support MPI_Comm_group_create, MPI_Type_size_x
699 - MPI ops should be properly applied to all allowed types
700 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
701 algorithm by a non failing one from mpich.
704 - Kill the fifo data container: we don't use it anymore.
707 - A <cluster> can now be created with different speed values to represent pstates.
708 see examples/platforms/energy_cluster.xml for an example.
710 SimGrid (3.14.159) Released December 28. 2016
712 The previous release was only a crude approximation of a Pi release;
713 this one should be a bit better.
715 - Revert a last minute change that broke on Mac OSX.
716 - Fix the changelog and NEWS files.
717 - Improve the documentation, in particular of the routing module.
718 - Integrate some patches flying around in Debian.
720 SimGrid (3.14) Released December 24. 2016
723 * The whole documentation was reworked and reorganized. There is
724 still much room for improvement here, but we're on it.
727 * We now need python3 for our testing infrastructure.
728 * Model checking needs libevent
731 * Now works on FreeBSD
734 * Privatization now works on FreeBSD
736 * Privatization is now activated by default in smpirun
737 You can turn it off with -no-privatize if something goes wrong.
739 * Call-location tracing for SMPI.
740 You can add the exact location (filename / linenumber) of an MPI call to
741 your trace files and slow down or speed up the simulation between two
742 consecutive calls by using an adjustment file (see the documentation).
744 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
745 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
746 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
747 values were ignored in some cases. The timings of these functions can now
748 be significantly different.
750 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
751 smpi/running-power is renamed to smpi/host-speed
753 * smpi/grow-injected-times option to enable or disable multiplication of the
754 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
755 make simulation less precise (but also much faster).
757 * smpirun script should be (much) faster for large deployments.
759 * SMPI tracing: fixed issue with poor matching of send/receives.
761 * Replay: Fix broken waitall
763 New functions and features
764 * MSG_parallel_task_execute_with_timeout, to timeout computations.
766 Dropped / renamed functions and features
767 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
768 - MSG_mailbox_is_empty() -> Mailbox::empty()
769 - MSG_mailbox_front() -> Mailbox::front()
770 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
771 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
772 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
773 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
774 * Don't pass the free_f parameter to property related functions:
775 - MSG_host_set_property_value()
776 - MSG_as_router_set_property_value()
777 - MSG_storage_set_property_value()
778 * VM properties. Since msg_vm_t are msg_host_t, just use
779 MSG_host_get_property_value() and friends
780 * VM I/O related things:
781 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
782 - Unimplemented save/restore methods
783 * MSG_as_router_get_property_value() was redundant with
784 MSG_environment_as_get_property_value().
785 - Removed MSG_as_router_*propert*() functions
786 - Added MSG_environment_as_set_property_value() for consistency
787 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
788 Well, they are still there for now, but deprecated with a warning.
789 Please switch to xbt_dict_new_homogeneous() before this is removed
791 * Task affinity. Its intended behavior (that was very badly tested
792 and probably not really working) was deceiving what most users
793 would have hoped here.
794 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
795 * Option network/coordinates is now useless and should be dropped.
798 * Added option storage/max_file_descriptors to allow more than 1024 files opened
801 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
802 * Added Dragonfly topology support
805 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
806 EiBps, EBps, Eibps, Ebps for bandwidth.
807 They may become useful to some lucky ones.
810 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
811 * ex/app_tokenring: new example, very similar to the MSG Token Ring
812 * ex/async_waitAll: new example, on asynchronous communications
815 * Memory usage should be decreased for simulations with a large number
816 of processes. This also helps for SMPI.
818 SimGrid (3.13) stable; urgency=low
820 The Half Release, a.k.a. the Zealous Easter Trim.
822 - We removed half of the lines, that were mostly experimental cruft.
823 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
824 (not counting blanks and comments -- according to openhub.net)
825 - The internals are now compiled in C++ (and will soon be clean C++)
826 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
827 59 klines of C, etc. We added only 29 klines of C++ in replacement.
829 * Backwards Compatibility breaks
830 - Removed Lua simulation bindings (switch to C or Java for that).
831 Lua can still be used to describe platforms
832 - Removed Java kernel plug-ins.
833 Will be reintroduced after the ongoing major internals reorg.
835 - the following functions were removed.
836 They were too specific and should be reimplemented in a generic
837 way, with filter function.
838 - MSG_task_listen_from_host
839 - MSG_mailbox_get_count_host_waiting_tasks
840 - MSG_mailbox_put_with_timeout was removed.
841 Please use MSG_task_send_with_timeout instead.
843 - the SD_application_reinit function was removed. It has been a noop for a while.
844 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
845 replaced by a more flexible mechanism.
846 - The following functions thus do not exist anymore
847 - SD_workstation_get_access_mode
848 - SD_workstation_set_access_mode
849 - SD_workstation_get_current_task
850 - Basic estimation functions have been removed but can easily be replaced
851 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
852 - SD_workstation_get_computation_time => amount / sg_host_speed()
854 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
855 Use VM.getSpeed()*load/100 for the legacy behavior.
857 - option enable_tracing was removed. It was not doing anything for a while.
858 - In the ModelChecker:
859 - the model-checker now ptraces the model-checked process which means
860 you cannot use a debugger on the latter anymore (we might make this
861 optional in the feature);
862 - removed soft-dirty page tracking;
863 - remove model-checked side snapshot management,
864 MC_snapshot() and MC_compare_snapshot();
865 - keep the MC_cut() function as a stub (it was not really working
866 in the previous release).
869 * All options are consistently in kebab-case. Old names are kept as alias.
872 * Switch to platform v4 format.
873 - Rename from 'power' to 'speed' the attributes describing the amount of
874 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
875 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
876 - In <host> and <link>, attributes availability and state are gone.
877 It was redundant with state and availability traces, and with peak values.
878 - In <cluster>, attributes availability_file and state_file are gone.
879 It was too complex and unused.
880 - Kill <gpu>. Was not doing anything.
881 - The DOCTYPE points to the right URL:
882 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
883 (the file at this address now documents the changelog since its v1)
884 - A warning is emitted for unit-less values (they are still accepted).
885 - speed. Default: 'f' or 'flops'. Also defined:
886 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
887 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
888 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
889 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
890 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
891 - latency. Default: 's' second. Also defined:
892 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
894 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
895 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
898 * s4u::Host is now the preferred public interface to the Host features.
899 sg_host_* functions are C bindings to the exact same behavior
900 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
903 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
904 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
905 They were deprecated since a few years, and probably did not even compile anymore.
908 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
909 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
910 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
911 encouraged to update their codes. A list of the modified functions can be found at the end of
912 include/simgrid/simdag.h
915 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
916 the lambda or closure passed as a parameter will run in kernel mode.
917 All the callback functions should be rewritten to that interface at some point.
920 * Reorganizing and cleaning the internals all around the place.
923 * Remove old default barrier/bcast buggy algorithms (see #18407)
924 * Various bug fixes to handle more codes
925 * Remove the need for the --foreground option of smpirun (it is still
926 accepted for backward compatibility).
929 * Kill the setset data container: MC don't use it anymore.
930 * Kill the queue data container: it made more sense with GRAS.
931 * Kill the xbt_peer_t data type: it's useless without GRAS.
932 * Kill rm_cb feature of config sets: it was never useful.
933 * Kill graphxml parsing feature. It was not used.
934 * Kill the deprecated code protected by XBT_USE_DEPRECATED
936 - xbt_dynar_sort_strings(), when the content is char*
937 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
938 They throw exceptions on invalid input;
939 * C++ support for declaring CLI flags (simgrid::config::Flag);
940 * class for abstracting different signal backends (simgrid::xbt::signal).
941 with no external dependencies (we need very simple signals).
944 * refactoring and cleanup of the code;
945 * ongoing process to cleanly separate the model-checking algorithms
946 from the code model-checking support.
948 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
950 SimGrid (3.12) stable; urgency=low
952 The Facelift Release.
955 * Require g++ v4.7 at least to not speak prehistorical C++.
956 * Require Boost 1.48 (for signal2 component).
957 * Java must be version 7 at least when activated.
958 * Builds on Windows again (including Java bindings).
959 * Tracing is now always enabled (no way to turn it out)
960 * Remove GTNetS. It was not working anyway.
961 * Various cleanups in the cmake scripts.
962 * Move headers around to sort them out on installed systems:
963 - instr/instr.h -> simgrid/instr.h
964 - instr/jedule/* -> simgrid/jedule
965 - simdag/datatypes.h was removed
966 - simdag/simdag.h -> simgrid/simdag.h
967 - msg/datatypes.h was removed
968 - msg/msg.h -> simgrid/msg.h
971 * Interface improvement:
972 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
973 - Sanitize the interface in MSG_task_ module:
974 - Merge two functions that were close enough but misleading:
975 set_compute_duration(t) -> set_flops_amount(t)
976 get_remaining_computation(t) -> get_flops_amount(t)
977 - set_data_size(t) -> set_bytes_amount(t)
978 get_data_size(t) -> get_bytes_amount(t)
979 - Massive cleanups in the functions related to the energy
980 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
981 - New: MSG_host_get_pstate()
983 - msg/energy/onoff: switching hosts on and off
985 * Interface improvement:
986 - Rename Host.isAvail() to Host.isOn()
987 - Rename Process.currentProcess() to Process.getCurrentProcess()
988 - Rename Task.setDataSize() to Task.setBytesAmount()
989 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
991 - #18874: Actually allows the GC to reclaim tasks
995 - SIMIX_process_throw: raises an exception in a remote process
996 * Refactoring: Separate sub-modules
997 - libsmx: the public interface, as libc in a real system
998 - popping: the strange dance that converts a user request into a kernel handling
999 - smx_context_*: the virtualization mechanisms that embed the user code
1000 - smx_*: the handling of each simcalls
1001 * Interface cleanup:
1002 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1003 * Rename smx_action_t into smx_synchro_t, making explicit that these
1004 things are used to synchronize processes with their environment.
1005 For example, a communication is a sort of synchronization involving
1006 the communicating processes (that may block until the exchange) and
1007 the platform. The same can be said from computations, etc.
1009 - #18888: segfault when a process ends before its kill_time
1013 - 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.
1014 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1015 - MPI_Type_set_name, MPI_Type_get_name
1016 - MPI_*_c2f and MPI_*_f2c functions
1017 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1018 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1019 - Activate a lot of new tests from the mpich 3 testsuite
1021 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1022 - InfiniBand network model added: Based on the works of Jerome Vienne
1023 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1024 - When smpi/display_timing is set, also display global simulation time and application times
1025 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1026 * Collective communications
1027 - 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
1028 - 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/).
1029 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1031 - Replay now uses algorithms from wanted collective selector
1032 - Replay can be used with SMP-aware algorithms
1033 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1034 - 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.
1036 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1037 - Correctly use loopback on fat-tree clusters
1038 - Asynchronous small messages shouldn't trigger deadlocks anymore
1039 * Energy/DVFS cleanup and improvement
1040 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1041 - new: smpi_get_host_pstate()
1044 - "Full" network optimization flag was broken since Surf++
1045 - Better handling of precision flags in maxmin
1046 - Fix bug causing sometimes "Impossible" errors
1047 - Properly pass cluster properties to included hosts
1048 * Improvement of the Energy plugin.
1049 - Always update the consumption before returning that value
1050 - New property: watt_off to denote the dissipation when the host is off
1051 - New functions getWattMinAt and getWattMaxAt to retrieve the
1052 dissipation of pstates that we are not currently at.
1053 * Java: class NetworkLink renamed to Link
1054 * New function: simcall_process_get_kill_time()
1055 * Massive rename s/workstation/host/
1056 - That's intrusive, but that's good for the project consistency. Sorry.
1057 - Change config option "workstation/model" into "host/model"
1060 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1061 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1062 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1063 - xbt_dirname and xbt_basename for non-POSIX systems
1065 * The model checker now runs as a separate process.
1066 * The model checker runs must now be launched with the new simgrid-mc program.
1067 * Record/Replay: the MC can display a textual representation of a path in the
1068 execution graph. It can then be replayed outside of the model checker.
1070 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1072 SimGrid (3.11) stable; urgency=low
1077 * Normalizing pointers addresses tool for better diff between logs
1080 * Add cloud examples using new VMs
1081 - examples/msg/cloud/two_tasks_vm.tesh
1082 - examples/msg/cloud/simple_vm.tesh
1083 - examples/java/cloud/cloud.tesh
1084 - examples/java/cloud/migration/migration.tesh
1085 * Add java surf examples:
1086 - examples/java/surfPlugin/surf_plugin.tesh
1087 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1088 - examples/java/surfCpuModel/surf_cpu_model.tesh
1089 * Add SMPI+MSG example:
1090 - examples/smpi/smpi_msg_masterslave/
1095 - msg task destroy cancel test
1096 - msg_host on/off test
1097 * Move all tests in testsuite to teshsuite (adding tesh files)
1098 * Restructure teshsuites
1099 - one folder for each kind of test
1100 * Restructure AddTests.cmake
1102 - structure the order of tests (with sections)
1105 * Add virtual machine
1106 - creation of a VM on a PM
1107 - migration of a VM from a PM to another PM
1109 - MSG_process_join(msg_process_t process, double timeout)
1110 - msg_bar_t MSG_barrier_init(unsigned int count)
1111 - int MSG_barrier_wait(msg_bar_t barrier)
1112 - void MSG_barrier_destroy(msg_bar_t barrier)
1113 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1114 * New option "msg/debug_multiple_use" to help debugging when a task is used
1117 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1119 - Increase code coverage in test suites
1123 * Protect context stacks against stack overflow. The number of protected memory
1124 pages allocated on the top of each stack (1 by default) can be configured
1125 with the new command line option --cfg=contexts/guard_size:<value>.
1126 * Simcalls are now generated by a python script that
1127 - generates files included by SimGrid
1128 - checks that all the functions exist, and proposes prototypes
1130 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1131 host_set_data, host_get_data
1133 - simcall_process_join(smx_process_t process, double timeout)
1134 * Fix bug where sleeping processing could not be suspended.
1137 * Translate surf models from C to C++
1138 - Generic classes for all models: Model, Resource, Action
1139 - A generic interface for each kind of model (CPU, Network, Storage
1140 Workstation, WorkstationVM)
1142 * Translate surf routings from C to C++
1143 * Add callbacks using sigc++ or boost::signals2
1144 - Add callback functions for resource creation/destruction
1145 - Add callback functions for action state change
1146 - Handle Energy as a plugin
1147 * Replace swag by boost::intrusive
1148 * Add new routing models for clusters. For documentation, see
1149 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1150 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1151 parameters for cluster tag
1152 - Fat trees, with topology="FAT_TREE" and
1153 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1154 - see examples/platforms/torus_cluster.xml and
1155 examples/platforms/fat_tree_cluster.xml
1156 * More documentation
1159 * Hostfiles support host:nb_processes construct to deploy several processes on
1161 * Collective communication algorithms should not crash if used with
1162 improper number of nodes and report the error.
1163 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1164 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1165 MPI_Dims_create, MPI_Cart_sub are supported.
1166 * New interface to use SMPI programmatically (still depends on MSG for
1167 some parts, see examples/smpi/smpi_msg_masterslave):
1168 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1172 * Global variables privatization in MPI executables is now performed at runtime
1173 with the option smpi/privatize_global_variables (default:no).
1174 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1175 dynamic libraries loaded by the application are not privatized (static
1176 linking with these libraries is advised in this case)
1179 * Options defined in XML work correctly now.
1182 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1183 copied into simgrid.jar or not (ON by default). Use this option if you want
1184 to reduce the size of the installed simgrid.jar, *and* the native libraries
1185 are kept installed elsewhere.
1186 * Surf binding with SWIG (code generated in maintainer mode only):
1187 - plugin to handle callbacks
1188 - CPU model only for the moment
1191 * Supernovae build mode is definitively removed. It was used to improve
1192 inlining and inter-module optimizations. It is nowadays superseded by
1193 link-time optimizations commonly available in compilers.
1194 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1195 * Add boost dependency for surf++
1196 * Add new macro for tests
1197 - ADD_TESH(name <tesh_args>)
1198 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1202 - xbt_bar_t XBT_barrier_init(unsigned int count)
1203 - int XBT_barrier_wait(xbt_bar_t barrier)
1204 - void XBT_barrier_destroy(xbt_bar_t barrier)
1205 * Make the xbt_os_time module public
1207 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1209 SimGrid (3.10) stable; urgency=low
1211 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1214 * Reintegrate Java to the main archive as desynchronizing these
1215 package is not acceptable anymore (Java is now considered stable)
1216 * Add explicit synchronization facilities through semaphores
1217 * Bug fix: Task.setDataSize() only changed the C world, not the value
1218 cached in the Java world
1221 * Dramatically change the way files are handled. API and internals changed, but
1222 this part of MSG was not considered as production grade either.
1223 * Add explicit synchronization facilities through semaphores
1224 * Add a new function MSG_host_get_process_list()
1225 * Preliminary DVFS support (see examples/msg/energy/ for details)
1228 * SMPI is now included directly in the libsimgrid as the windows
1229 linker doesn't force us on splitting it anymore.
1230 * Improvements of the SMPI replay tool:
1231 - Most of the collective communications are now rooted in the same process as
1232 in the original application.
1233 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1234 used until now). Multiple data types can now be used in a trace.
1235 - The replay tool now supports traces produce either by TAU or a modified
1237 - Bug Fix: the compute part of the reduce action is now taken into account.
1238 - Gatherv collective is now supported
1239 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1240 of smpirun outputs time independent traces for the current run. One file
1241 is created per process. If too many processes are simulated, this behavior
1242 can be changed to one file for all processes by using the
1243 tracing/smpi/format/ti_one_file flag
1244 * smpirun generates the host file if needed (with given host count and platform)
1245 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1246 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1247 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1248 --cfg=smpi/coll_selector:(mpich/ompi)
1249 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1251 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1252 happening inside a collective SMPI call.
1253 * Fix the behavior of complex data types handling.
1254 * Make MPI_Wtime another synchronization point to take computations into
1256 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1257 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1258 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1259 (no privatization of global variables yet)
1260 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1263 * Verification of liveness properties is now available for SMPI applications
1264 (in addition to MSG applications)
1265 * Bugged examples using SMPI in examples/smpi/mc/
1266 * Add --cfg=model-check/visited option. Allows the verification of infinite
1267 programs. Detection of loops in the execution thanks to the system state
1268 comparison and reduction of the state space to explore. Can be combined with
1269 DPOR for safety properties.
1272 * Allow to change SimGrid configuration (see --help) within the code
1273 thanks to SD_config() as it can be done in MSG.
1274 * Add a new function SD_task_set_amount() upon user request.
1277 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1278 * Remove rule based routing (no more PCRE dependency)
1279 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1280 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1281 * Add a loopback_bw and loopback_lat options to cluster tag.
1282 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1283 routing. Now peers are encapsulated in an AS and have their own private
1284 router but this is transparent.
1287 * Our own implementation of getline is renamed xbt_getline, and gets
1288 used even if the OS provide a getline(). This should reduce the
1289 configuration complexity by using the same code on all platforms.
1290 * New type: xbt_cfg_elm_boolean.
1291 * Allow to use yes/no for boolean configuration options in the command line.
1292 * Allow to disable SimGrid cleanups at exit from command line option.
1293 There are situations where one may want a simulation to end with an exit.
1294 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1295 annoying when scripting around the simulator. Adding a
1296 --cfg=clean_atexit:no allows to circumvent this issue.
1299 * Lots of memory leaks were corrected in this release.
1300 There is no known memory leaks anymore, in all of our 600+ tests.
1301 * New command line option --version, to get SimGrid version information.
1302 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1304 * Supernovae builds are deprecated, and expected to be removed in the next
1307 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1309 SimGrid (3.9) stable; urgency=low
1311 The Grasgory release: GRAS is really dead now.
1313 * Complete overhaul of the internal host structures scheme.
1316 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1317 was considered as experimental and badly maintained since then.
1318 * Keeping it was thus a trap to our potential users, that could take
1319 it instead of MSG or SMPI by mistake despite is pity state.
1320 * GRAS seems to have very few users (if any), and no one volunteered
1321 to maintain it further. It also induces a lot of XBT code (for
1322 portability sake), that must be maintained too.
1323 * For all these reasons, we killed GRAS. If someone wants to revive it
1324 in the future, don't cry, our git history still remembers of GRAS.
1327 * Major overhaul. Merge our documentation again as time proved that
1328 splitting it was really not helping our users.
1329 * Further improve the developer documentation to help newcomers
1330 hacking on SimGrid itself. The user documentation (and in
1331 particular, the beginner documentation) is still in a sorry state.
1334 * Now works on Windows too!
1335 * Much more extensive test suite, from MPICH
1338 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1339 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1340 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1341 * Bug fix: let task be scheduled when the last dependency to be solved is
1342 a control dependency.
1343 * Remove SD_load_environment_script function.
1344 Use the C sg_platf function if you want to declare a platform
1348 * New function: MSG_process_get_number()
1349 * Old function documented: MSG_config()
1350 * Remove MSG_load_platform_script function
1351 Use the C sg_platf function if you want to declare a platform
1355 * Change the default value of the TCP_gamma constant (maximal size of TCP
1356 congestion window) to a more realistic 4MiB value. If you notice changes in
1357 your simulation results, you can fall back to the previous 20k tiny window
1358 by adding --cfg=network/TCP_gamma:20000 on command line.
1359 * (Hopefully) fix a bug wrt periodic availability and state traces
1360 * Bug fix: use default values at start when first event in availability/state
1361 trace is not at time 0.
1364 * remove the "new_" part of function name sg_platf_new_trace_connect
1365 (resulting in sg_platf_trace_connect), since it does not create
1369 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1370 use case where it's really mandatory, and maintaining it was a pain in
1372 * New: xbt_fifo_search(), search an item with a user-provided
1373 comparison function instead of dumb pointer comparison.
1376 * Fix the lua deployment:
1377 Use `simgrid.init_application()` before deployment instead of
1378 `simgrid.msg_register_application()` after.
1381 * Transfer the tracing files into the corresponding modules.
1383 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1385 SimGrid (3.8.1) stable; urgency=low
1387 The "we are told that some people want to also *install* the simgrid
1390 * Add missing file "tesh.1" to the archive.
1392 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1394 SimGrid (3.8) stable; urgency=low
1396 The Psssshiiiit release: SimGrid jumps into the Cloud.
1399 * Add an experimental interface to manipulate VMs. They are mainly
1400 process groups with very few intrinsic semantic, but they should
1401 allow you to build the semantic you want easily.
1402 * New function: MSG_host_set_property_value()
1403 * New function: MSG_process_on_exit(). To clean memory in all cases.
1404 * Bug fixes that made the host (and link) failures unusable.
1405 * Add a way to auto-restart process when the host in which they are
1406 executing comes back (ON_FAILURE="RESTART" on deployment file,
1407 MSG_process_auto_restart_set).
1408 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1409 please stop using the old ones, they are DEPRECATED.
1411 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1412 Please use MSG_init() instead. (reducing the amount of entry
1413 points in the library helps us).
1414 * Make it impossible to link against the wrong version of the lib
1415 * Deprecate MSG_clean(). No need to call it anymore.
1416 * Function MSG_get_host_number() is not deprecated anymore.
1419 * Split the doc into a user guide and a reference guide.
1420 * Start a developper guide to help people hacking on SimGrid.
1423 * Enable tracing by default. This modules rocks you should use it.
1424 * Remove option custom_flags. Now use environment variables CFLAGS
1426 * Use default cmake things to detect lua instead of home grown ones.
1427 * New option "enable_mallocators" to disable mallocators, for debugging
1428 purpose ("on" by default).
1431 * Bug fixes around the resource failures: don't let the processes
1432 survive the host they are running onto.
1433 * Add an interface to auto-restart processes when the host in which they are
1434 executing comes back.
1435 * Ensures that SIMIX_clean is called automatically. It's not part of
1436 the public interface anymore (bindings should be updated).
1439 * Bug fix for when SD_Simulate is called with a positive value: be careful
1440 when comparing doubles. Sometimes they are different for non significant
1442 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1443 parallel task whose initial work is distributed among host according
1444 to the Amdahl's law. Such tasks are created with a parameter alpha
1445 that corresponds to the non-parallelizable part of the computation.
1446 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1447 between two sets of workstations assuming a 1D block distribution (each
1448 workstation owns a similar share of data) on both sides.
1450 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1451 Data redistribution will be automatically scheduled once parent and child
1452 are both scheduled. The filling of computation_amount and
1453 communication_amount structures is now done seamlessly thanks to the chosen
1455 * New function SD_workstation_dump to display various information
1456 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1457 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1458 bandwidth on the route onto which the task is scheduled and the amount of
1460 To divide the nominal bandwidth by 2, the rate then has to be:
1461 rate = bandwidth/(2*amount)
1462 * Compute tasks that have failed can now be rescheduled and executed again
1463 (from their beginning)
1464 * Increasing source code coverage (src/simdag is now covered at 95.8%
1468 * Re-implement time-independent trace replay using SMPI (at the
1469 smpi_smp_* level) instead of MSG. This should replace
1470 examples/msg/actions/actions.c
1471 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1472 hindexed and structs)
1473 * Implement the exchange of non-contiguous data.
1474 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1475 * Correct behavior of smpi/sender_gap and set its default value to 0
1476 * Add option to asynchronously send small messages to allow better
1477 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1478 specifies the size in bytes under which messages will be asynchronously sent.
1479 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1480 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1483 * New C interface to define a platform: XML is now optional.
1484 For more info, please check include/simgrid/platf.h.
1485 * New interface to define random platforms from the C:
1486 For more info, please check include/simgrid/platf_generator.h and
1487 examples/msg/masterslave/masterslave_platfgen.c
1488 * Export a sg_cmdline dynar containing all the arguments we got from
1492 * Two new tracing options for adding comments to trace file so you
1493 can track your experiments (see --help-tracing for details).
1494 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1495 * Adding the SimGrid version that generated the trace file as a comment.
1496 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1497 * Fix to avoid key clashes on Paje links
1498 * Other minor fixes related to the Paje specification
1501 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1502 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1503 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1504 * Greatly improve the robustness of mmalloc to user errors (such as
1505 using an area after freeing it, or freeing it twice)
1507 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1509 SimGrid-java (3.8.1) stable; urgency=low
1511 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1512 Warning: all methods are visible, but only some of them are
1513 implemented so far. Check the source (src/jtrace.c)
1514 for further information.
1515 * New module: org.simgrid.msg.File (SimGrid File management functions)
1516 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1517 * Change the meaning of Process.restart: now restart the process from
1518 the begining, like MSG_process_restart in C.
1519 * Add Process.setAutoRestart: handling of process restart when failed
1521 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1522 you to retrieve the properties of the processes/hosts
1523 * Deprecate Msg.clean(): you can just forget about it now.
1524 * New function Process.getCount(), that only works when compiling
1525 with the not yet released version 3.9 of the C library.
1527 * New context factory based on Coroutines. It mandates a modified JVM
1528 but then, the simulations run about five times faster, and there is
1529 no limit to the amount of processes (beside of the available memory).
1531 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1533 SimGrid (3.7.1) stable; urgency=low
1536 * Restore the prototype of MSG_process_create_with_environment() to
1537 the pre-3.7 situation by removing the kill_time argument.
1538 * Add a MSG_process_set_kill_time() function instead.
1541 * Fix weird behaviors when dealing with parallel tasks.
1544 * Simgrid is now built as a dll.
1545 * Simgrid-java now works on Windows.
1546 * Simgrid-Java is now included into Windows package.
1549 * First pre-build package for MacOSX.
1552 * Fix compilation when using MSG_USE_DEPRECATED.
1553 * Fix some compilation issues on Macs and Windows.
1554 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1555 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1557 We discovered that the Lua console is broken, but we are missing the
1558 manpower to fix it right now. The problem existed in 3.7 too, so we
1559 are not blocking the release for that. Sorry if you depended on this
1560 feature, any help would be really welcome.
1562 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1565 SimGrid-java (3.7.1) stable; urgency=low
1567 The "Java aint got to be bloated and slow" release
1570 * Various internal cleanups and performance improvement
1571 Simulations are expected to run up to twice faster or so
1572 * Make Process.kill(process) an instance method, not a static one
1573 * User processes are not java.lang.Thread subclasses.
1574 This breaks the compatibility (sorry), but previous API was
1575 brain-dead, making it impossible to have non-trivial
1576 initializations in the process constructor.
1577 * Require a full constructor per Process sub-class.
1578 Kinda breaks the compatibility (sorry), but this allows a much more
1579 efficient way to launch the processes at simulation startup.
1580 * Do not embeed our version of semaphores, java 1.5 can be considered
1581 as sufficiently prevalent for us to not dupplicate its features.
1585 * Add examples for almost every part of the API
1586 We spotted and fixed a lot of bugs in the process
1588 * New module: asynchronous communication API
1589 * New function: Process.sleep()
1590 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1591 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1592 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1594 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1596 SimGrid (3.7) stable; urgency=low
1598 The "spring cleanups (before the next Big Project kicks in)" release.
1601 * We can specify the SMPI latency/bandwidth factor with command line
1602 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1603 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1604 You can also use the "config tag" from platform file by adding this line
1605 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1606 (see "example/platforms/tag_config.xml" to use "config tag").
1607 Note that the command line supersedes the platform file configuration.
1608 * Change the correction factors used in LMM model, according to
1609 the latest experiments described in INRIA RR-7821.
1610 Accuracy should be improved this way.
1611 * Use the partial invalidation optimization by default for the
1612 network too. Should produce the exact same results, only faster.
1613 * Major cleanup in surf to merge models and split some optimization
1614 mechanisms from the core of the models. As a result you can now
1615 specify which model to use (e.g., --cfg=network/model:LV08
1616 --cfg=cpu/model:Cas01) and which optimization mode to use
1617 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1618 Incompatible combinations should err at initialization. See
1619 --help-models for the list of all models and optimization modes.
1620 * The CLM03 workstation model was dropped for simplicity because it
1621 used the deprecated CM02 network model. Use default instead.
1622 * Rename the TCP_gamma configuration option to network/TCP_gamma
1623 * Rename the coordinates configuration option to
1624 network/coordinates, and document it
1625 * Use now crosstraffic keyword instead of the terribly misleading
1626 fullduplex keyword. It is activated by default now in the current
1627 default model, use --cfg=network/crosstraffic:0 to turn it off.
1628 * Ongoing refactoring the model parsing to make XML files optional
1629 See include/simgrid/platf.h for details (still to be completed)
1632 * Major overhaul of the documentation. Almost instructive now :/
1633 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1634 functions and friends. This interface was considered as
1635 deprecated since over 2 years, it's time to inform our users that it is.
1636 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1637 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1638 use these (crappy) functions in your code.
1639 These functions will be removed soon. Stop using them now.
1640 * Deprecate MSG_get_host_{table,number}
1641 Implement MSG_hosts_as_dynar() instead.
1642 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1643 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1646 * Stabilize the parallel execution mode of user contexts
1647 * Introduce configuration variables to control parallel execution:
1648 - contexts/synchro: Synchronization mode to use when running
1649 contexts in parallel (either futex, posix or busy_wait)
1650 - contexts/parallel_threshold: Minimal number of user contexts
1651 that must be part of a scheduling round to switch to parallel
1652 execution mode (raw contexts only)
1653 * Fix bugs that prevented to use suspend/resume along with
1654 synchronization structures.
1655 * Fix bugs in process termination that lead to invalid memory access
1656 in very specific conditions.
1659 * Introduce a parallel mode for the models (controlled by surf/nthreads
1660 configuration item). In our tests, running the models in parallel
1661 never lead to any speedups because they are so fast that the gain
1662 of computing each model in parallel does not amortizes the
1663 synchronization costs, even when ultra fast futexes are used.
1664 This is released anyway because YMMV.
1667 * Performance boost by using a swag internally to compute the set of
1668 tasks that are finished and should constitute the return value of
1672 * Enable it by default now that it is considered rather stable.
1675 * Documentation of the tracing functions.
1676 * Performance gains when tracing categorized/uncategorized resource
1677 utilization by avoiding calls to get route when updating resource
1678 variables. LMM constraints are being used instead.
1679 * API changed to set task categories. Use MSG_task_set_category instead
1680 of TRACE_msg_set_task_category, and SD_task_set_category instead
1681 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1682 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1683 * Examples for link user variables added (see at examples/msg/tracing/)
1684 * Deprecated function TRACE_msg_set_process_category completely removed
1685 * Trace header updated according to the latest Paje file format
1686 * Tracing network lazy updates, no longer obligate users to use full updates
1687 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1688 * Experimental: let user code declare/set/push/pop application states for hosts
1689 * API changed to allow the manual creation of graph configuration files
1690 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1693 * Improve the API of Lua MSG bindings, using the Lua spirit.
1694 * Each simulated process now lives in its own Lua world (globals are
1695 automatically duplicated). It helps writing simulators. Will allow
1696 to run Splay programs within SimGrid in the future.
1697 * Add a Chord example in Lua, equivalent to the MSG one.
1700 * Start the implementation of a solution to express temporal
1701 properties, not only local assertions. This is still an
1702 experimental work in progress, stay clear from it to be safe.
1706 - Add new runtime parameters --help-logs and --help-log-categories
1707 to display information about supported logging parameters and
1709 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1710 don't exists anymore.
1711 * Mallocators: allow value NULL for the reset function.
1713 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1714 create homogeneous dictionaries, where all the elements share the
1715 same free function. Non homogeneous dictionaries will be
1716 deprecated in the next release.
1717 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1718 - Multi-level dictionaries are deprecated.
1720 - new function xbt_dynar_search_or_negative() that is useful when
1721 you have less than 2 million elements in your dynar and don't
1722 want of the extra complexity of catching exceptions when the
1723 element is not found.
1725 - Make xbt_os_thread module (for thread portability) public.
1726 Documentation is still to come, sorry.
1728 - Cleanups and simplifications to make it maintainable again.
1729 - Exotic features (such as memalign and valloc) were removed.
1730 - The metadata were extended and improved so that the
1731 model-checker becomes able to explore and inspect the heaps.
1732 - This may induce a performance drop when enable_model-checking is
1733 ON in cmake (even if it's not used in the simulation), but it is
1734 necessary at this point to get MC working.
1736 Turn model-checking OFF if simulation performance matters to you.
1737 Not enabling it at runtime is not enough, disable it in cmake.
1739 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1742 SimGrid (3.6.2) stable; urgency=low
1744 The "Not coding new stuff allows to polish old things" release.
1747 * New bindings to the NS3 packet level simulator (experimental)
1748 * Use the raw (efficient) execution contexts instead of the sysv
1749 (portable) ones when possible.
1750 * libpcre is now mandatory in any cases since not using it led to
1751 severe performance loss and possibly other issues
1752 * Update the XML platforms:
1753 - G5K: include the latest machine in Nancy
1754 - GridPP and LCG: new platforms
1755 * Documentation was partially updated, at least (more to come)
1757 Bug fixes, cosmetics and small improvements
1758 * Free terminated processes before the end of the simulation to avoid
1759 exhausting the memory of users having very dynamic amount of
1761 * Bug fix and cosmetics about canceling non-running tasks
1762 * Bug fix about the dot loader's issues when using libcgraph
1765 * Create an installer for windows with nsis (amd64 and win32)
1766 - Add an hello world project to illustrate simgrid project creation.
1767 - Embed libpcre into the Simgrid installer to avoid
1768 its compilation burden
1769 * The raw execution contexts should work on Apple now
1770 * Port to Windows 64 bits
1771 - Sysv contexts now have an implementation for this arch
1772 - GRAS communication features now support this arch
1773 * Drop support for borland compiler on windows
1774 - this code was not maintained, and we kinda depend on gcc nowadays
1775 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1776 * Fix portability issue on unstable ubuntu: linker became picky on
1779 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1781 SimGrid (3.6.1) stable; urgency=low
1783 The "Oops, we broke Macs too" release
1786 * Fixed contexts detection so that raw ones are used when possible
1787 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1788 leads to a strange error, with user code segfaulting sometimes when
1789 the generated code is not perfectly aligned (which is not
1790 controllable from the user side, depends on the amount of code)
1793 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1795 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1797 SimGrid-java (3.6) unstable; urgency=low
1800 * Split of every thing from simgrid v3.5 into a separate package.
1802 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1804 SimGrid (3.6) stable; urgency=medium
1806 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1809 * Bindings now constitute their own package, separated from the main one.
1810 Rationale: reduce our maintenance nightmare by reducing the module coupling
1811 They will soon be released on their own on gforge.
1813 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1814 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1816 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1817 * It's not quite deprecated for now because we have no replacement,
1818 but it may soon become the case.
1821 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1822 * Fortran: New user-level cache variable to store the rank of the running
1823 process. This improves performance by an order of magnitude.
1824 * C: New coccinelle script to automatically locate and modify global and
1825 local static variables.
1826 * Improved SMPI network model with a sender-side gap to account for multiple
1830 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1831 only say if a communication is finished, no matter whether it succeeded or
1832 failed. You can call MSG_comm_get_status() to know the status of a finished
1834 * New function MSG_task_dsend() to send a task and detach it. When a
1835 communication is detached, you are never notified of its success or failure
1836 and the memory is released automatically once it is finished. This function
1837 is useful when you don't care about the end nor the success of a
1839 * Change the prototypes of action replay. Sorry for inconvenience,
1840 but this is really more efficient this way (and to adapt your code,
1841 you just have to fix the initialization, that shouldn't be too long)
1842 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1843 ever managed to do anything useful with it.
1844 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1845 takes a dynar of communications. It returns immediately and gives the
1846 index of a finished communication (if any).
1847 * New example: a basic implementation of the Chord P2P algorithm.
1850 * New model for multi-core CPUs. You can now use the core attribute to
1851 precise the number of cores of a host. This is a basic model. Every
1852 process running on the host receives at most the power provided in
1853 the DTD (throughput<=power). Total throughput of process cannot exceed
1855 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1856 router linked by an up-link and a down-link (possibly asymmetrical).
1857 This kind of pattern allows to easily build last-mile model style platforms.
1858 Aggregating such patterns in a rule-based AS is thus the technique of
1859 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1860 * New model for Vivaldi routing. We transformed the Vivaldi network model
1861 into a Vivaldi routing model (based on the rule-based model). This allows to
1862 combine Vivaldi based latencies with last-mile platforms.
1865 * Added a check for NaN of IEEE754 infinite in the double entries of
1867 * Introduce a new context factory "raw", highly inspirated from the
1868 ucontext factory, but using manually crafted functions in assembly to
1869 do the work in an efficient manner.
1870 * Allow to change the used context factory at run time, not only at
1871 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1872 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1873 contexts at runtime (only with raw contexts or ucontexts).
1874 * Completely rewrote this module to allow parallel execution of user
1875 processes. Use --cfg=contexts/nthreads:N to execute user processes
1876 with N parallel threads (the default is 1, meaning no parallelism).
1877 * Allow to decide dynamically between sequential and parallel modes.
1878 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1879 processes in parallel only when their number is greater than or equal to P
1881 * Added a check for NaN of IEEE754 infinite in the double entries of
1885 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1886 won't output the state of processes when interrupted with Ctrl-C
1887 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1888 NULL-terminated array. This may solve backward compatibility issues
1889 due to the change to return type of SD_simulate. See also:
1890 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1891 * Add new macros with variable number of arguments.
1892 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1893 - in xbt/asserts.h: xbt_assert
1894 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1895 - in xbt/ex.h: THROWF and RETHROWF.
1896 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1898 * Change xbt_die() to accept a format string with arguments, just like printf.
1899 * New data structure: xbt_lib_t, like a dict but more general and with better
1903 * New configuration options
1904 Options triva/categorized and triva/uncategorized can be used to generate
1905 graph configuration files for Triva visualization tool.
1906 * Configuration option tracing/platform is renamed to tracing/categorized
1907 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1908 * New TRACE_user_link_variable interface:
1909 User provides the name of the link and the tracing variable to attach to it
1910 * the declaration of tracing categories must be done after the environment
1912 * simpler tracing interface, just one way to declare categories
1913 TRACE_category or TRACE_category_with_color, it is up to you
1914 * links in the trace file are again identified by their names
1915 * trace contains the full platform hierarchy exactly as declared using the ASes
1916 * Options tracing/msg/[task|process]:1 groups the process by hosts
1917 for both cases, tasks and processes must have names that are unique during the simulation
1918 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1919 * The experimental option tracing/msg/volume is deprecated
1920 its functionality may be reincorporated if needed
1922 The tracing generates a trace file with unordered timestamped events,
1923 because of the way the core simulator (surf) works. A script available
1924 at the tools directory (fix-paje-trace.sh) can be used to put the events
1925 in order. We have changed the tracing so it can generate ordered timestamped
1926 events in the final trace, but depending on the simulator (and how much time
1927 is simulated) that can lead to a huge memory utilization. It is deactivated
1928 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1930 Build Infrastructure
1931 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1932 - We are trying hard to keep the API stable, but it may happen that
1933 some things change (we're a research project after all, not a
1934 nuclear plant operating system). If such things should happen, you
1935 could rely on that macro to adapt.
1936 - current value: 30600 for 3.06.00, aka 3.6
1937 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1938 a number that can be compared with SIMGRID_VERSION.
1939 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1940 assertions and verbose/debug logging events are disabled at compile time.
1942 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1944 SimGrid (3.5) stable; urgency=medium
1947 * New feature to any SimGrid-based simulator: Model-Checking
1948 Check SIN#1 for more details.
1951 * New Model SMPI (three-interval linear regression for correction factors)
1952 See RR-7426, available at http://hal.inria.fr/inria-00527150
1953 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1954 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1955 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1956 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1957 * See also src/smpi/README
1962 - Tracing API changes: TRACE_start and TRACE_end should not be called
1963 by user-code. They are automatically called by simulators created
1964 with SimDAG, MSG and SMPI if the toolkit is compiled with
1965 tracing_enabled=ON. Categories declaration and utilization remain the
1966 same for MSG and SimDag.
1967 - A function was added to the tracing API to declare categories with
1969 - TRACE_category_with_color (char *category, char *color)
1970 where color must be in the following format
1971 "%f %f %f", red, green, blue
1972 and red, green, blue are float values in the interval [0, 1]
1973 - User can specify NULL as color parameter, or continue calling
1974 TRACE_category (cat)
1975 On that case, the tracing system will define random colors
1976 - The following command-line options are supported:
1977 --cfg=tracing/filename:msg.trace
1978 --cfg=tracing:1 (activate tracing, needed to use others)
1979 --cfg=tracing/platform:1 (categorized resource use)
1980 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1981 --cfg=tracing/msg/task:1 (task creation)
1982 --cfg=tracing/msg/process:1 (process creation, migration)
1983 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1984 --cfg=tracing/smpi:1 (SMPI interface tracing)
1985 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1986 - examples of examples/msg/tracing updated
1988 - DAXLoader and DOTLoader functions can generate tasks with categories
1989 - A new function to attribute a category to SD tasks:
1990 TRACE_sd_set_task_category (SD_task_t task, char *category)
1991 * Tracing the MPI interface implemented by SMPI
1992 - Collective operations are traced with states
1993 - Point-to-Point operations are traced with states/links
1994 - Tracing activated by a parameter "-trace filename" passed
1995 to smpirun during execution (considering that simgrid
1996 is compiled with tracing enabled)
1997 - To run the simulation with gdb, the simulator
1998 accepts --cfg=tracing/smpi:1 to trace SMPI
1999 - tesh files to check if smpi tracing is ok
2000 - See examples/smpi/NAS/DT-trace
2001 * GTNetS tracing re-worked
2002 - adaptation to the tracing system of GTNets to cope
2003 with modifications regarding the fullduplex mode
2004 - new tesh files to check if gtnets tracing is ok
2007 * Asynchronous communications through the functions:
2008 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2009 * New function: MSG_load_platform_script()
2010 to make possible using a lua script instead of XML files to set up platforms
2011 * New function: MSG_set_function
2012 to associate functions to processes, used when bypassing the parser
2013 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2015 Platforms: Add some more examples in examples/platforms
2016 * Grid'5000: see www.grid5000.fr
2017 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2020 * Change the XML format. This is a very important modification. SimGrid
2021 3.5 introduces a new hierarchical format based on the notion of
2022 Autonomous Systems. Compatibility with old format is ensured
2023 through the perl script provided in the install bin directory
2024 bin/simgrid_update_xml.
2025 It is now possible to build platforms with specific routing
2026 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2027 connect several platforms together. We will try to provide soon
2028 set of realistic platforms exploiting these properties (have a
2029 look at examples/platforms/ for the moment).
2030 * Take the opportunity of the XML format change to be a good XML citizen:
2031 rename link:ctn to link_ctn and similar changes (also dealt with by
2033 * Add a new routing scheme (rule-based) using regular expressions. It
2034 enables to have an extremely low memory footprint when the
2035 underlying routing is simple and can be compactly described. You
2036 need to have libpcre4-dev (perl regular expressions) installed if
2037 you want to use this routing scheme.
2038 * Revive the cluster TAG and allow to easily and efficiently (both in
2039 term of memory and speed) connect clusters together. Have a look
2040 at teshsuite/simdag/platforms/ to see how this can be done. With
2041 this tag, you can create clusters with thousands of tasks at no
2042 cost (have a look at examples/platforms/).
2043 Note: clusters are implemented as ASes, so there is no need for an
2044 enclosing AS tag if you have only one cluster in your platform.
2045 * Add new generic functions in the public interface that allows the user
2046 to call SURF 'create_resource' methods from your code (same
2047 functionality as the XML bypass mechanism but with a much lighter
2049 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2050 that uses a piecewise linear approximation to produce better
2051 results when exchanging small messages.
2052 * Add a new parameter to handle correctly full duplex link and account
2053 for interferences between uplink and downlink communications
2054 (activate with --cfg=fullduplex:1).
2057 * Rename the SD_READY (all dependencies are satisfied and task is
2058 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2059 dependencies are satisfied) state.
2060 This prevents a confusion between the notion of "ready to schedule"
2061 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2062 simulated" (SD_RUNNABLE) used by the simulation kernel.
2063 * Change the way a task is considered as ready. Instead of removing
2064 dependencies when a task is done, a counter is decreased. This way,
2065 it is always possible to reach ancestors thanks to the
2066 SD_taks_get_parents function (even after the end of the simulation.)
2067 * Change the return type of SD_Simulate from (SD_task_t*) into
2068 xbt_dynar_t. This function was in handling a dynar internally and
2069 converted it into a NULL terminated array for historical reasons.
2070 * New function SD_dotload(char*) to load a DAG described in dot
2071 format. This loader and the corresponding examples require the
2072 installation of the graphviz library.
2073 * Fix a bug in the management of tasks of size 0 in the surf network
2074 models. This problem was only visible with SIMDAG and you should
2075 thus disregard results produced with earlier versions if you
2076 relied on this feature (some tasks were blocked because of this).
2077 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2078 even though your applications were doing only point-to-point
2079 communications and sequential computations. Now you can really use any
2080 model you want (of course, if you create real parallel tasks, which are
2081 not implemented in most models beside ptaskL07, this will abort).
2082 * Add an example that schedules a DAX on an heterogeneous platform
2083 using a Min-Min strategy.
2084 * New function SD_workstation_get_current_task() that returns the kind
2085 of task currently running on a workstation in the sequential access
2087 * Raise some warnings when unexecuted tasks remains at the end of the
2088 simulation. This is usually caused by cycles in the DAG.
2091 * New function: SIMIX_process_set_function() called by MSG_set_function
2092 * Change the underlying waiting queue in semaphores so that a process
2093 can wait on several of them simultaneously (as in waitany).
2094 * Fix the way to handle tokens in semaphores so that all access patterns
2095 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2096 * kill the dirty pimple SIMIX_message_sizes_output()
2097 Please use (proper) visualization instead
2100 * New data container: setset (set of sets of elements)
2101 * New module: mmalloc (mapped malloc, allowing to have several
2102 independent segments of malloc)
2103 * New function: xbt_dict_cursor_set_data()
2104 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2105 * New function: xbt_dynar_is_empty()
2106 * New function: xbt_fifo_get_last_item()
2107 * Fix xbt_dynar_shrink(): use the right element size.
2108 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2109 uninitialized areas during expand.
2110 * Fix semaphores: previous implementation was severely broken.
2111 * Use library init/fini functions for our initialization.
2112 - you can use logs and other feature as soon as you want in your
2113 code (even before the xbt_init / MSG_init)
2114 - xbt_exit is now a no-op and produce a warning when used.
2117 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2118 benefit from the latest improvement to the simulation kernel.
2119 * Kill measurement sockets for now. If you rely on them, sorry. This
2120 release is not for you. This feature will be reintroduced in the
2121 future, but we cannot delay the release any further.
2122 * New function: gras_msgtype_get_name().
2123 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2124 * Fix (at last) the pmm example: it should not randomly fail anymore.
2126 Build chain: bug fixes and overall polishing
2127 * Cmake is now stable enough. Hence, we killed the autotools.
2128 * Port to windows ( TM :)
2129 * Fix the 'make install' target.
2130 No need to use 'make install-simgrid' anymore
2131 * Introduce a 'make dist' target compiling a *source* archive
2132 'make package' compiles a binary archive
2133 * Compile java files only on need
2134 * Add --cd and --setenv command line options to tesh
2135 * Out of source builds are not fully supported yet, but we are close
2136 * Enable supernovae and optimization flags by default for our users
2139 * Add layer to set up environment directly from lua, without XML.
2140 * The effect of gras_stub_generator can be achieved through
2141 lua too (check examples/gras/console/ping_generator.lua)
2143 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2145 SimGrid (3.4.1) stable; urgency=low
2147 The "Polishing easter eggs is probably a good idea" release.
2148 This is a bug fixes release only.
2152 * Fix a bug preventing the tasks from begin garbage collected.
2155 * Fix a bug occuring when a host involved in a communication fails.
2156 This was not detected properly by the other peer involved in the
2157 communication. Now, it's reported as a network error.
2160 * Warn the user about loop dependencies in data flow of DAX files
2161 * Obey the control-flow dependencies of DAX files
2164 * Add option "enable_smpi" allowing to not compile SMPI.
2165 Probably useful for the (Mac) users experiencing a build error here
2166 * Improve the detection of lua5.1 and ruby1.8
2168 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2170 SimGrid (3.4) stable; urgency=low
2172 The "Easter in Cargese" release. Also known as (major changes):
2174 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2175 ~> bindings were greatly improved
2176 ~> new tracing infrastructure for better visualization introduced
2178 * the "Welcome to configury modernity" release.
2179 ~> we switched from autotools to cmake, and improved our cdash
2182 A more detailled list of changes follow (full detail in svn log).
2184 Java Bindings: Various Cleanups
2185 * (install java-gcj-compat-dev on debian-like to use them)
2186 * Remove put/get: no need to export deprecated interface in Java
2187 Use send/receive instead.
2188 * Cleanup the examples and add a README per directory
2189 * Remove example autoDestination (that's the only way to go now)
2190 * Remove example explicitDestination (was a plain copy of basic)
2191 * Make JniException a runtime exception, so that there is no need to
2192 declare the fact that you may encounter such a beast. I guess that
2193 nobody will ever want to survive such error.
2194 * Create specific errors for each MSG case of failure:
2195 host failure, transfer failure, timeout, task cancelled
2196 * Cleanup the exceptions that may get thrown by each function
2197 * Other internal cleanups in Java bindings. Performance still bad :/
2198 Ruby and Lua Bindings: create them
2199 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2200 * That's new and great, you should try them out.
2201 Same functionalities than Java bindings, only even less polished
2203 * Kill the useless "rate" argument of SD_task_get_execution_time()
2204 Everyone used to provide -1 as a value, it was not used, and the
2205 semantic of a possible use wasn't even clear.
2206 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2207 either as comm costs or compute costs to mean that there is no
2208 such thing for that specific task.
2209 * Add a SD_task_set_name() function
2210 * Fix SD_task_unschedule() on typed tasks
2211 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2212 * In DAX loader, accept useless 'level' attributes to <job> since
2213 LIGO DAGs have them (seem to be to ease graphical representation).
2215 * Add an example masterslave_mailbox.c using send/receive and not
2216 the deprecated put/get interface.
2217 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2218 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2220 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2221 (declare MSG_USE_DEPRECATED to still have the old name)
2222 * Add a MSG_task_set_data() function
2223 * About trace replay (see examples/msg/actions):
2225 - Allow to work with splitted trace files for each process
2226 Give the specific trace file as argument of each process,
2227 and call MSG_action_trace_run(NULL)
2228 You can still have one merged file for all processes.
2229 - Fix implementation of collective operations
2230 * Allow task_execute() on 0-sized tasks (closes #10063)
2232 * This is the first release of SimGrid where SMPI is not considered
2233 beta anymore (even if some corners should still be improved)
2234 * Port over the new SIMIX_network submodule (internal refactoring)
2235 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2236 * Implement more missing elements of the standard:
2238 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2240 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2241 MPI_Scan() MPI_Get_processor_name()
2242 - Added implementation of missing case for Alltoall (warning: it's
2243 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2244 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2245 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2246 SMPI_MPI_Allgatherv()
2247 * Bug fixes include:
2248 - MPI_Waitsome() was broken
2249 - Allow relative includes in smpicc
2250 - Command line cfg argument 'reference_speed' was ignored...
2251 - Some functions did not properly lead to auto-benching of user code
2252 - smpicc passes -O2 by default (just like openmpi one)
2254 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2255 * Bug fixes about timeouts during communications
2256 * add SIMIX_message_sizes_output() as a pimple to write to file the
2257 amount of messages per size. Use gnuplot to get histogram.
2258 Pimple because that's the only user-visible function of simix,
2259 defined directly in xbt.h (irk, sorry)
2261 - Add a SIMIX_sem_get_capacity() function
2262 - Fix interactions with processe resume/suspende
2263 - release_forever() was stupidly broken
2264 - Fix SIMIX_display_process_status() for processes in a semaphore
2265 - Make SIMIX_sem_block_onto() user-visible
2266 * Refactoring context stuff:
2267 - Use pseudo-OOP for better modularity
2268 - reimplement SIMIX_process_kill() without process_schedule() so
2269 that the latter can take as invariant that it is called from
2271 - Merge context_start into context_new for sake of simplicity
2273 * Add a Vivaldi network model, coded live during SUD'10 ;)
2274 * Rename configuration variables to start a hierarchy:
2275 o cpu_model -> cpu/model
2276 o network_model -> network/model
2277 o workstation_model -> workstation/model
2278 * New configuration variables:
2279 o network/bandwidth_factor: correction to bandwith
2280 o network/latency_factor: correction to latency
2281 o netwotk/weight_S: correction to the weight of competing streams
2282 * Add a long description to the models, that users can see with such
2283 argument on the command line: --cfg=cpu/model:help
2284 * --help-models display the long description of all known models
2286 * config: add the ability to set a default value after registration
2287 Does not override any previously set value (e.g. from cmd line)
2288 * dict: allow to have integer key and data.
2289 When so, you need to use the following functions
2290 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2291 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2292 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2293 In contrary to regular dicts, the key is not malloced before copy.
2294 Mixing scalar and regular elements in the same dict is not tested
2296 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2297 Tracing for Visualization:
2298 * SimGrid is now instrumented in order to generate a trace file for
2299 visualization analysis: to use it, need to compile SimGrid with the
2300 "tracing" option enabled, and instrument the program using SimGrid with
2301 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2302 (among other functions).
2303 * The instrumentation only traces the platform utilization for now
2304 * Documentation to use the tracing functions and how to analyze the
2305 traces with the Triva tool is written.
2306 * More information about: SimGrid FAQ (in the section Tracing Simulations
2309 * We moved to cmake as default build system. Autotools support will
2310 be dropped soon. Check the FAQ for more info about how to use it.
2311 * Greatly improved our cdash/ctest interactions
2312 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2313 * Added memory checking tests with valgrind; lot of memleak fixing.
2314 This may be the first release of simgrid with so few memory issues
2315 * Added code coverage tests.
2316 Our coverage is still improvable, but at least we see it on cdash.
2318 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2320 SimGrid (3.3.4) stable; urgency=low
2322 The "Desktop Grid needs love too" release (also called Xmas release).
2324 Models improvements:
2325 * Major speedup in the maxmin system solving by using lazy evaluation
2326 Instead of solving completely the maxmin system at each iteration,
2327 only invalidate (and recompute) the modified parts.
2328 This new feature is enabled in default models but you can try to
2329 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2330 * Cas01 IMproved as default CPU model
2331 This CPU model is the same Cas01 model, but it uses the
2332 maxmin-selective-update flag and a heap structure to manage
2333 actions on SURF kernel.
2334 It reduces the complexity to find the next action to finish and,
2335 consequently, it's faster than the old Cas01.
2336 This is the new default CPU model (Cas01).
2337 * Rename the old Cas01 model to Cas01_fullupdate
2338 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2339 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2340 * CpuTI (CPU Trace Integration)
2341 A new CPU model whose objective is simulate faster when using
2342 availability trace files.
2343 Instead of using a full featured, over engineered maxmin system for
2344 CPU modeling, this model does the pre-integration of traces files
2345 to calculate the amount of CPU power available, and so, executes
2346 faster than the old CPU models.
2347 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2348 * Use LV08 as default network model since it gives better accuracy
2349 for small messages and shouldn't change things for big ones.
2350 Use --cfg=network_model:CM02 to get the previous behavior.
2353 ******************************************
2354 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2355 ******************************************
2356 * The new CPU model may changes simulations!
2357 The point is that events occurring at the exact same timestamp
2358 are not scheduled in the same order with the old and new
2359 version. This may be enough to completely change the execution
2360 of simulations in some cases.
2361 * The new network model will change simulations!
2362 This new model is more realistic than the previous one, so you
2363 should consider redoing your old experiments with this model.
2364 Sorry for the inconvenience.
2367 * Introduce the supernovae compilation mode
2368 When compiled that way, the whole SimGrid (or almost) is put in a
2369 single compilation unit and compiled in one shoot.
2370 This is to help gcc which has difficulties to inline stuff from one
2372 The speedup seem to be above 15%, althrough more tests are needed on
2373 amd64 to confirm that gain.
2376 * Port of MSG's mailbox on top of SIMIX network
2377 The put/get mechanism was greatly simplified on the way.
2380 * New SIMIX network module. Provides:
2381 - Mailbox: rendez-vous mechanism to find with who you want to speak
2382 - Synchronous send/recv: easier and hopefully faster since the
2383 logic is handled in the maestro process directly now
2384 - Asynchronous send/recv: you dreamt of it? It's here now
2385 Too bad that nobody cared enough to propagate the change to MSG.
2386 * Add semaphores as SIMIX synchronization mechanism.
2389 * new function SD_daxload(char*) to load a DAX file
2390 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2391 * Introduce typed tasks. Specify its kind and cost at creation.
2392 At scheduling, just give where it should be placed, and the cost
2393 for each involved resource is automatically computed.
2394 Existing constructors so far (more to come of course):
2395 - SD_task_create_comm_e2e() for end-to-end communication
2396 - SD_task_create_comp_seq() for sequential computation
2397 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2398 * new function SD_task_dump() for debuging display
2399 * new function SD_task_dotty(task,FILE*) writing to file the info
2400 about the task in dotty format
2401 * SD_task_dependency_exists() can now cope with having one of its
2402 arguments NULL. If so, it tests whether the other argument has any
2404 * Add getters on list of preceding/following tasks:
2405 SD_task_get_parents(task) and SD_task_get_children(task)
2406 * Add getters on amount of workstations and list:
2407 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2408 * Add getter on task kind: SD_task_get_kind(task)
2409 * Update the start_time and finish_time of tasks on completion/failure
2410 * Bugfix: Remove task from state swags when destroyed
2413 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2416 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2417 dynar with one separate thread per value of the dynar.
2418 * Change the prototype of xbt_thread_create(), sorry.
2419 Added a boolean parameter indicating whether we want to join this
2420 thread (used in SG only for now)
2421 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2424 * GTNetS wrappers should now be usable again (and betterly tested too)
2425 * Fix a major regression from 3.2 where the timeout provided to
2426 MSG_task_put_with_timeout() was used as absolute time before which
2427 the comm should be done.
2428 * Start to fix the <cluster> tag.
2429 - Internal links should be good now (beside of the loopback, which
2430 use the private link instead)
2431 - paths to the external world is still rather broken
2432 - the <route:multi> tag is just broken. Actually that's brain-dead.
2433 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2434 to make it less stupid
2435 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2436 * Fix a source-level compatibility glitch from 3.2: after defining
2437 MSG_USE_DEPRECATED, you can use the old name
2438 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2439 * Allow to compile from the SVN with automake 1.11
2440 * Fix some problems when using the "start_time" tag in deployment XMLs.
2441 * Fix #8569: XBT/synchro.h has redundant declarations
2442 * Fix #8563: MSG return values and exceptions
2443 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2444 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2445 Thanks to Robson Peixoto.
2446 * Fix a vicious bug in dictionaries inducing that some elements were
2447 not freed on xbt_dict_free()
2449 Portability report of this version:
2450 * Main portability targets:
2451 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2452 on (amd64/i386/ia64)
2453 - mac leopard on i386
2454 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2455 but nothing critical.
2456 * Other platforms: windows, AIX and others were not tested for this release
2458 Timing report of this version:
2459 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2460 scenario parameters). From 8h to a few seconds in desktop grid settings.
2461 * Supernovae brings about 25% speedup on i386.
2463 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2465 SimGrid (3.3.3) stable; urgency=low
2467 The "Need for Speed" release.
2469 The timings done to validate the 3.3.2 were faulty.
2470 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2472 The problem was a conversion from a manually handled vector to
2473 xbt_dynar_t on the critical path.
2474 xbt_dynar_foreach calls functions, inducing stack management crap.
2476 We inlined these functions and xbt_dynar_foreach is now breath taking.
2477 We also inlined xbt_swag_belong on the way.
2479 Here are some approximate speedup measurements (on master/slaves
2480 simulations lasting between 10s and 20s each):
2481 3.3.1 -> 3.3.2: about same performance
2482 3.3.2 -> 3.3.3: 40% speedup
2483 3.3.1 -> 3.3.3: 40% speedup
2484 3.3.1 with inline patch -> 3.3.3: 30% speedup
2486 Our reading is that the refactoring which occurred in 3.3.2 made us
2487 suffer much more from the xbt_dynar_foreach low performance, but
2488 once we solved this, this refactoring proved to be very performance
2489 effective. From the 40% speedup, somehow, 10% are due to the
2490 inlining and 30% to the refactoring.
2492 That's a pitty that gcc cannot inline functions placed in other files
2493 alone. We have to choose between:
2494 - break the encapsulation (by putting private data structures and
2495 accessors in headers files to help gcc)
2496 - live with low performance
2497 - switch to a decent compiler such as icc (not quite possible).
2499 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2501 SimGrid (3.3.2) stable; urgency=low
2503 The "Simplicity does not preceed complexity, but follows it" release.
2505 The main contributors of this release were (lexical order):
2506 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2509 * Extract the routing logic into its own object.
2510 (was dupplicated in network.c and workstation_LV07.c;
2511 Allows to implement other ways of storing that info)
2512 => kill now useless network_card concept
2513 - Use dynar to represent routes (instead of void** + int*)
2514 - kill link_set (use surf_network_model->resource_set instead)
2515 - Add a command-line option to choose the routing schema to use
2516 - Add three new models:
2517 * Floyd (shortest path computed at initialization)
2518 * Dijikstra (shortest path recomputed all the time)
2519 * Cached Dijikstra (shortest path computed on need)
2520 All these models where contributed by Silas De Munck, and are
2521 described in his ICCS09 paper.
2523 * Simplify model declaration
2524 (less redirections, less function to write when defining a model)
2525 - Factorize stuff between models:
2528 surf_model_resource_set(model)
2529 surf_model_resource_by_name(model, name)
2530 - Unify the types of models in s_surf_model_t (using an union)
2531 - Embeed fields of common_public directly into s_surf_model_t
2532 - Rename model methods:
2533 action_free ~> action_unref
2534 action_change_state ~> action_state_set
2535 action_get_state ~> action_state_get
2536 - Change model methods into functions:
2537 (model)->common_public->action_use ~> surf_action_ref
2539 * Implement a generic resource; use it as ancestor to specific ones
2540 (allows to kill duplicated code in models)
2541 Drawback: timer command don't need no name nor properties;
2542 workstation_CLM03 don't need no properties
2543 (but I guess we can live with those few bytes wasted)
2545 * Improve the action object model
2546 - implement a constructor avoiding dupplicated code about field
2547 initialization in generic_action part.
2549 * Kill the SDP model: it has an external dependency, is deprecated
2550 in flavor of modern lmm models, and didn't compile since a while
2553 * Relocation of the context module from XBT to SIMIX.
2554 (the context were decoupled from the simix processes, duplicating a lot of code)
2555 => a lot of code was factorized
2556 - less overhead is introduced during scheduling
2557 - simpler API for the context factory
2558 - the logic for process creation,destruction and manipulation was simplified
2559 * Simplification of the s_smx_process_t data structure.
2560 => accesing the simix level data associated to a process is faster now,
2561 and the code is a lot more readable.
2564 * Implement some more MPI primitives:
2565 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2566 -implementation: Bcast: flat or 2-ary tree (default),
2567 Barrier: 4-ary tree,
2569 Allreduce: Reduce then Bcast
2570 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2571 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2572 Alltoallv: flat tree, like ompi
2574 * Add support for optimized collectives (Bcast is now binomial by default)
2575 * Port smpirun and smpicc to OS X
2578 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2579 Shout out if you used it.
2582 * Display the list of still queued messages in SG mode when existing
2586 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2587 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2588 * Add xbt_str_from_file(FILE*)
2589 * Add xbt_dict_get_key achieving a linear reverse search
2590 * Remove the context module
2592 Portability report of this version:
2593 * Main portability targets:
2594 - Linux(debian)/x86/context
2595 - Linux(debian)/x86/pthreads
2596 - Linux(debian)/amd64/context
2597 - Linux(debian)/amd64/pthreads
2598 On these, we still have the eratic breakages of gras/pmm and
2599 amok/saturate_sg reported in previous version. We still think
2600 that the tests are the cause of the fault, not the tested code.
2602 - Mac OSX Leopard/x86/context
2603 Still false negative in tesh autotesting.
2604 Smpi still fails, but this time because readlink does not accept -f
2605 Everything seems to work properly beside of that.
2608 - AIX version 5.3 (only tested contexts this time)
2609 Smpi still fails there because mktemp is not installed.
2610 Everything seems to work properly beside of that.
2612 I managed to compile it for the first time, but several breakages.
2613 Won't delay the release for this exotic platform.
2615 * Windows: it's still lagging behind. If you want to help, please
2618 Timing report of this version:
2619 This version seem to be more than 5% faster than 3.3.1 (on linux
2620 64bits with contextes). The gain is less than expected, we are
2621 investigating this for next release.
2623 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2625 SimGrid (3.3.1) stable; urgency=low
2628 * Implement a --cfg-help to show existing configuration variables
2629 * Build chain do not require doxygen in maintainer mode
2632 * fix a bug on struct sizeof computation, which prevented the
2633 exchange of arrays of structs in some conditions
2634 - added a regression test about this in datadesc_usage
2635 * Allow the exchange of 0-long dynamic vectors.
2636 - for that, use -1 as indicator of dynamic size instead of 0
2637 - This implied to change any size from unsigned long to long,
2638 reducing a bit communication abilities, but I guess that with
2639 64bits being quite common, this is more than enough.
2640 - This also induce a protocol change, thus bumping network protocol
2641 version from 0 to 1 (if we have external users, we have to get
2642 clean on that point too ;)
2643 - added two regression tests about this in datadesc_usage
2644 * Be more verbose when propagating local exceptions
2645 This helps debugging.
2646 * Display the status of simulated processes when receiving SIGINT in
2650 * Allow to control the simulation from a trace file.
2651 New functions MSG_action_register() and MSG_action_trace_run()
2652 The first one allows to associate a function execution to each
2653 kind of action while the second one parses a trace file and
2654 triggers the corresponding actions within the system.
2655 For now, only a toy example is provided in examples/msg/actions
2656 * Add an exemple of process migration in examples/msg/migration
2657 * Fix a bug in task exchange which broke MSG_task_get_sender()
2658 Add a teshsuite regression test for that.
2659 [Bug: if MSG_task_get_sender() is called after sender exit,
2661 * Fix a bug which prevented suspend/resume to work properly
2662 * Display the status of simulated processes when receiving SIGINT
2663 This fixes a regression of v3.3. due to the introduction of SIMIX
2664 * Bug fixing in failure management:
2665 - trace could not start by a failure at time 0
2666 - failure during communications were not working
2669 * Add SIMIX_process_set_name() to change the name of the current
2670 process in the log messages.
2671 * Store smx_hosts in a dict since we only retrieve them by name
2672 * Move the configuration infrastructure to surf
2675 * Move the configuration infrastructure to surf
2678 * Massive internal cleanups:
2679 - Store internal structures on processes instead of hosts (allows
2680 to have more than one process per host, in addition of being more
2682 - Cleanup the initialization/finalization process
2683 - Kill a whole bunch of unneeded synchronization:
2684 processes run in exclusive manner within the simulator
2685 - Move queues from global tables to process data fields
2687 - now accept -platform and -hostfile arguments
2688 - Pass the right rank value to processes according to the hostfile
2689 * Compile the examples by default, and use them as regression tests
2690 * Implement MPI_Wtime()
2691 * Change the reference speed to a command line option
2694 * TCP_gamma can now be specified as command line option using
2695 --cfg=TCP_gamma:10000000.0
2696 * Change the --surf-path cmd line option into --cfg=path:
2699 * Also include strbuff from xbt.h public header
2700 * xbt_ex_display(): do not free the exception after displaying
2701 This allows to do more with the given exception afterward.
2702 Users should call xbt_ex_free() themselves.
2706 Portability report of this version:
2707 * Main portability targets:
2708 - Linux(debian)/x86/context
2709 - Linux(debian)/x86/pthreads
2710 - Linux(debian)/amd64/context
2711 - Linux(debian)/amd64/pthreads
2712 These targets fail about 1/10 of times on gras/pmm, but we believe
2713 that this is because of the test, not because of simgrid.
2714 amok/saturate_sg fails even more rarely, and the test may not be
2717 - Mac OSX Leopard/x86/context
2718 The test suite still spits tons of errors because some obscure
2719 force prevents us from removing the temporary directories
2720 arguing that they still contain some metadata I've never heard of.
2721 Smpi fails because seq is not installed.
2722 Everything seems to work properly beside of that.
2725 - AIX version 5.3 (both contexts and pthread)
2726 Smpi still fails there because mktemp is not installed.
2727 XML inclusions seems rosty on AIX.
2729 * Windows: it's still lagging behind. If you want to help, please
2732 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2734 SimGrid (3.3) stable; urgency=high
2738 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2739 [Malek Cherier & Mt]
2741 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2744 Until now, GRAS were using MSG as an interface to SURF. It was
2745 quite difficult because both interface have several differences
2746 (MSG channels vs GRAS sockets were the most notable point).
2748 This also opens the gate to SMPI (which should occur soon) and speed
2749 up simulations by to 40% (even if it were not the main goal).
2751 **************************************
2752 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2753 **************************************
2754 The point is that events occuring at the exact same timestamp are
2755 not scheduled in the same order with the old and new version. This
2756 may be enough to completely change the execution of simulations in
2757 some cases. Sorry for the inconvenience.
2759 * Cleanup and upgrade the XML format to push further scalability
2760 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2762 * Improve the testing infrastructure with tesh. Now a very large part of
2763 the code is tested not only by being run but also by checking that the
2764 output match an expected output [Mt].
2766 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2767 is really less memory-demanding, which should allow you to use
2768 larger files in SimGrid [AL].
2770 * Inform valgrind about our contextes, so that it becomes usable
2771 with the default (and more effecient) version of SimGrid
2772 [contributed by Sékou Diakite, many thanks]
2775 * Introduce a listener thread in charge of receiving incoming
2776 messages from the network. It allows to overlap communication and
2777 computation but most notably, it removes some stupid deadlocks due
2778 to the fact that so far, a process could not send and receive at
2779 the same time. This made most non trivial communication schema
2781 * Convert the PIDs from long int to int to match the MSG ones (and
2782 linux ones too) [Mt]
2783 * New function: gras_agent_spawn() to launch a new process on
2784 current host. Only working in simulation for now. [Mt]
2785 * New function: gras_os_hostport() returning a constant form (ie,
2786 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2789 * Make the backtrace of exceptions more human readable [Mt]
2790 * New module: xbt/str [Mt]
2791 a ton of string utility functions (split, join, printf to a newly
2792 allocated buffer, trim, etc)
2793 * New module: xbt/hash [Mt]
2794 SHA1 hashing algorithm (more to come if needed)
2795 * New module: xbt/synchro [Mt]
2796 synchronization tools (mutex and conditions) working the same way
2797 in simulation and in real life (mainly useful for GRAS, but not
2799 * New module: xbt/queue [Mt]
2800 classical producer/consumer synchronization scheme
2801 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2802 (using the classical functions will get serialized) [Mt]
2803 * Make dictionary internal table dynamic. No need to specify its size
2804 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2806 * Make sure the log channels are organized as a tree under windows
2807 (because of ANSI C compatibility issue, any channel were child of
2808 root directly) [Mt].
2811 * Cleaned many thing in surf and fixed a few bugs [AL].
2812 * Add a nice command line configuration mechanism to compose models [AL].
2813 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2814 the previous one (KCCFLN05). It relies on something that looks like
2815 a max-min sharing mechanism but cannot be written as such. A new solver
2816 was thus designed [AL].
2817 * Add a new solver to lmm. Based on Lagrange optimization and
2818 gradient-based descent, it enables to efficiently maximise systems s.a
2820 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2822 This solver enables to propose two new network models for TCP Reno and
2823 TCP Vegas based on Low's work. These models still need to be fully
2824 tested though [Pedro Velho].
2827 * Bug fix in SD_simulate. Now the time bound given as argument is
2829 * Use the new parallel task model (ptask_L07) as default.
2830 * Use the SURF command line configuration mechanism.
2831 * 0-size tasks (for synchronization) should now work.
2833 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2835 SimGrid (3.2) stable; urgency=high
2839 We still experience issues on this platform, but we believe that at
2840 least MSG is usable.
2842 GRAS API BREAKAGE (for simplification purpose, sorry):
2843 * the gras_msgtype_by_name is not used anymore. Instead of
2844 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2845 you can write (and must)
2846 gras_msg_send(toserver, "request", &request);
2847 - If you still want to pass a gras_msgtype_t to the function (to cache
2848 the type and avoid the lookup time), use the gras_msg_send_() variant.
2849 - Impacted functions:
2850 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2851 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2852 * The callbacks are now expected to return 0 when everything went well
2853 (just like the main() function)
2855 GRAS new features and improvements:
2856 * New module mechanism where user code can use per process globals [Mt]
2857 This is similar to gras_userdata_*() functions, but for libraries. It
2858 factorize some code developped over and over in the examples and AMOK.
2859 It has still to be documented and used (only amok/peermanagement is
2861 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2862 loss when they were small enough to fit into the buffer and sent quickly
2863 enough so that they can all get received in one shoot.
2864 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2865 if not found. Use the *_or_null() variant for the old semantic.
2866 * In gras_msg_handle, do not discard messages without callback.
2867 They are probably messages to be explicitly awaited later (ie, proofs of
2868 mis-synchronization in userland since they are sent before being awaited)
2869 No big deal usually.
2870 * gras_socket_meas_send/recv: semantic changed!
2871 The numerical arguments used to be (1) the total amount of data to send
2872 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2873 messages. This was need for the fool willing to send more than MAXINT
2874 bytes on quite fat pipes.
2877 * Do really rename the hostmanagement module to peermanagement. [Mt]
2878 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2879 but this is rather new and this was documented in the module
2880 documentation (poor excuses, I admit)
2881 * Bandwidth measurement semantic changed! This follows the changes to
2882 gras_socket_meas_send/recv explained above.
2885 * A sequential mode has been added to the workstations. When a workstation
2886 is in sequential mode, it can execute only one task, and the other tasks
2887 are waiting in a FIFO. [Christophe Thiery]
2890 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2891 model for SIMDAG. [Christophe Thiery]
2892 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2893 equal to 0, which caused some bad behaviors in
2894 saturated_constraint_set_update. I now use a threshold mechanism like in
2898 * When running manually src/testall, you select specific units [Mt]
2899 testall is the result of our cunit mechanism, and should replace all
2900 the scripty thingy around since bash don't run easily on billware.
2902 * A mallocator system has been added. [Christophe Thiery]
2903 Mallocators allow you to recycle your unused objects instead of freeing them
2904 and allocating new ones.
2906 Documentation update:
2907 * FAQ reworking + New FAQs:
2908 - "Valgrind spits tons of errors!" [Mt]
2909 - "How to repport bugs" [Mt]
2910 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2911 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2912 - Communication time measurement within MSG [AL]
2913 - I experience weird communication times when I change the latency [AL]
2914 * GRAS tutorial [Mt]
2916 - an introduction to the framework and to the used communication model
2917 - an initiatic tour introducing the most proheminent features:
2919 . Lesson 0: Installing GRAS
2920 . Lesson 1: Setting up your own project
2921 o Part 2: Message passing
2922 . Lesson 2: Exchanging simple messages
2923 . Lesson 3: Passing arguments to the processes (in SG)
2924 . Lesson 4: Attaching callbacks to messages
2925 . Lesson 5: Using globals in processes
2926 . Lesson 6: Logging information properly
2927 . Lesson 7: Using internal timers
2928 . Lesson 8: Handling errors through exceptions
2929 . Lesson 9: Exchanging simple data
2930 . Lesson 10: Remote Procedure Calling (RPC)
2931 . Lesson 11: Explicitely waiting for messages
2932 . Recapping of message passing features in GRAS
2933 - A HOWTO section containing:
2934 o HOWTO design a GRAS application
2935 More are due, of course. They will come latter. In the meanwhile, you can
2936 check the examples which are still here.
2938 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2940 SimGrid (3.1) stable; urgency=high
2944 There was a stack corruption somewhere, visible only when optimizing
2945 with these versions. [Vince]
2948 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2949 module that enabled to play with Directed Acyclic Graphs. It is built
2950 directly on top of SURF and provides an API rather close to the old
2951 SG. Some old codes using SG are currently under rewrite to check that
2952 all needful functions are provided. [Christophe Thiery]
2955 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2956 extension of the classical CLM03 model that gracefully handles
2957 failures. This is now the default model for MSG and GRAS. It doesn't
2958 handle parallel tasks yet though. [AL]
2959 * Bug fix: Weights were not correctly set in the network part.
2960 WARNING: This may have resulted in incorrect results with simulations
2961 where there are more than one flow on a given link. [AL]
2964 * After a (long ?) discussion on simgrid-devel, we have decided that the
2965 convention we had on units was stupid. That is why it has been decided
2966 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2967 WARNING: This means that all previous platform files will not work as
2968 such with this version! A warning is issued to ask users to update
2970 A conversion script can be found in the contrib module of the CVS, under
2971 the name contrib/platform_generation/surfxml_update.pl [MQ]
2974 * Bug fix: Processes were started in reverse order, wrt deployment file.
2975 WARNING: if your code relies on this bug, please fix it. [AL]
2976 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2977 being executed on two different locations. [AL]
2978 * Bug fix: Failures are now better supported thanks to Derrick's tests
2979 (there was many failure situations I hadn't thought of and that weren't
2980 correctly handled). [AL]
2981 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2985 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2986 Exception occurring on server-side are propagated back to client (!).
2988 API CHANGE: the callback changed their prototype. Change:
2989 int my_handler(gras_socket_t expeditor, void *payload_data) {
2991 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2992 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2994 * New! function: gras_msg_handleall to deal with all messages arriving
2995 within a given period.
2996 * New! function: gras_socket_server_range to get a server socket in a
2997 range of port numbers (ease to avoid port number conflicts) [MQ]
2998 * New! gras processes display their backtrace when they get a SIGUSR1
2999 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3000 Sweet to debug RL processes [MQ]
3004 - Do not force experiment sizes to be expressed in kb, or it becomes
3005 impossible to measure the latency this way (needs one byte-long tests)
3006 WARNING: this changes the amok_bw_* function semantic. [MQ]
3007 - Implements the link saturation stuff. [MQ]
3008 * Peer management module:
3009 New! module factorizing code that we wrote over and over [MQ].
3012 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3013 - Test units are placed directly into the library code, they get extracted
3014 automatically and placed into the src/testall binary.
3015 - Convert most of the XBT tests to this system.
3016 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3017 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3018 live in the same C code. This required to change a little bit the API
3019 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3020 * New module: sparse graph structure with basic algorithms (this is work
3021 in progress and the API is not considered to be frozen yet). [AL]
3022 * Display more information on backtraces: source line & function names are
3023 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3024 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
3025 * Reworked a little bit some #include statements to load only required
3026 headers. Some user code that relied on SimGrid to include stdlib or
3027 stdio may need to include it by themselves. [AL]
3028 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3030 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3031 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3032 but I guess that almost nobody used those parts. [MQ]
3034 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3036 SimGrid (3.0.1) stable; urgency=low
3039 * Unfortunately, I had missed 5 misnamed functions:
3040 xbt_fifo_item_t xbt_fifo_newitem(void);
3041 void xbt_fifo_freeitem(xbt_fifo_item_t);
3042 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3043 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3044 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3045 They're now deprecated. Please use their new versions:
3046 xbt_fifo_item_t xbt_fifo_new_item(void);
3047 void xbt_fifo_free_item(xbt_fifo_item_t);
3048 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3049 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3050 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3052 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3053 * Documentation: xbt_log module unmercifully reworked [MQ]
3054 * Bugfix: there was a problem with the ending of contexts with
3055 the pthread backend. It caused some weird deadlock or behavior
3056 depending on the pthread implementation. [AL]
3057 * Bugfix: get the exceptions raised in the simulator repport where
3058 and why they come from when they are not caught in time [AL, MQ]
3061 * Bugfix: Do repport the error when two non-connected hosts try to
3062 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3065 * Add additionnal checkings on communications. Assert that two
3066 communicating hosts are connected by a set of links... [AL]
3069 * Add additionnal checkings on channel values in communication [AL]
3070 * New: MSG_task_get_source to see on which host a task was generated [HC]
3071 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3072 the number of tasks waiting to be received on channel and sent
3074 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3075 waits for the first task coming from a given host.. [AL]
3077 GRAS new functionnalities: [MQ]
3078 * Enhance the parsing macro to allow the size of multidimentional objects
3079 to be given thru annotations.
3080 * New example (and documentation): Matrix Multiplication a la RPC
3081 (as when I was young!) and fix a bunch of bugs found on the way.
3083 GRAS performance improvements: [MQ]
3085 * Reduce the amount of cbps creation/destruction by making it static to
3086 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3089 * Change libdata to a set so that we can search for stuff by ID (and thus
3090 reduce the insane amount of dict lookups)
3093 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3094 RL communication on top of SG since it's so uneffective.
3095 It may also be used for inter-thread communication in RL, one day.
3096 * Use gras_datadesc_copy() to exchange messages on top of SG
3098 - improve message exchange performance on top of SG
3099 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3100 * Don't exchange on the network the size of the used part of buffer,
3101 instead, specify the possible buffer size to read().
3103 - reduces the amount of read/write calls (one pair per exchange)
3104 - reduces the amount of exchanged data (the size)
3105 - allows to retrieve all arrived data on receiver side, if we don't need
3106 it right now (subsequent read will peek the buffer)
3107 - allows the receiver to proceed with the begining of the stream before
3108 everything is arrived
3109 - make it possible to build an iov transport (using readv/writev)
3111 - take care of the data with non-stable storage (like stacked data),
3113 * If possible, TCP send uses vector I/O (when writev() is here)
3114 - Don't use it for receive since we send data sizes and data on the
3115 same stream, so we wouldn't be able to chain large amount of chunks
3116 before having to flush the stuff to read the size.
3117 * Rework the transport plugin mechanism to simplify it and reduce the
3118 amount of pointer dereferencement when searching for the right function
3121 * I guess that now, we do almost as few system calls as possible while
3122 doing as few data copy as possible.
3124 To improve it further, we could try to send all the sizes first and then
3125 all the data (to use iov on receiving size), but it's only a partial
3126 solution: when you have 2 dimensional data, the sizes of the second
3127 dimension is data of the first dimension, so you need 3 streams.
3129 I'm not sure the potential performance gains justify the coding burden.
3131 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3133 SimGrid (3.00) stable; urgency=high
3136 * New! Give the possibility to hijack the surf parser and thus bypass
3137 MSG_create_environment and MSG_launch_application. Have a look at
3138 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3140 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3142 SimGrid (2.96) unstable; urgency=low
3147 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3148 This deprecates the xbt_error_t mechanisms.
3149 It modifies (simplifies) all XBT and GRAS API.
3150 MSG API keeps unchanged (exceptions raised by XBT are caught from
3151 within MSG and masked with existing error handling facilities)
3154 * New! Add a FATPIPE model. [AL]
3155 * New! Add a parallel task model. [AL]
3156 * New! Add automatically a loopback interface (in the default
3157 network model) if none was precised.
3160 * Bugfix: MSG_process_resume now works with the current running process.
3162 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3163 * Modification of MSG_task_get_compute_duration. Once a task has been
3164 processed, the value returned by this function is now equal to 0. [AL]
3165 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3166 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3167 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3168 of a m_task. For now, MSG_task_cancel only works with computation
3170 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3171 of the processor (in Mflop/s) regardless of the current load on the
3173 * API Change: use proper naming convention for MSG_getClock and
3174 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3176 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3177 This function changes the priority of a computation task. This priority
3178 doesn't affect the transfer rate. A priority of 2 will make a task
3179 receive two times more cpu power than the other ones. This function
3180 has been added to suit the needs of Nguyen The Loc and hasn't been that
3181 much tested yet. So if it fails, please report it and send me your code.
3183 * API Change: removed all functions and types that were marked "deprecated"
3184 since many months. Renamed MSG_global_init_args to MSG_global_init.
3186 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3188 SimGrid (2.95) unstable; urgency=low
3191 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3192 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3193 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3196 * Bugfix: Allow absolute paths for platform description files [MQ]
3197 * Bugfix: do free the variables after use. Leads to drastic performance
3199 * Implement max_duration (ie, timeouts) on resources [AL]
3202 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3204 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3205 the needed support (timeouts) [AL]
3208 * Implement measurement sockets. You can now get the bandwidth between two
3209 hosts thanks to AMOK (see below). [MQ]
3210 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3211 dynar over the network (yeah) [MQ]
3212 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3213 * simple/static token-ring example. [Alexandre Colucci and MQ]
3214 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3215 in place before [MQ]
3218 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3219 type parsing. "short" and "long" modifiers were also ignored; other
3220 modifier (such as reference level) are still ignored. [MQ]
3221 * Embeed the buffer size within the buffer itself on SG. [MQ]
3222 That way, send() are atomic and cannot get intermixed anymore (at least
3223 the ones which are less than 100k; bigger messages still have the issue)
3224 * Array size pushed by the field, not by the field type (or each
3225 and every long int will push stuff to the cbps) [MQ]
3226 * use select() to sleep since it allows to portably sleep less than one
3229 GRAS (minor cleanups)
3230 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3233 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3234 * type category 'ignored' killed as it was never used and were difficult
3236 * whether a type can cycle or not is now a flag, leaving room for more
3237 flags (as "ignored", if we feel the need one day ;)
3238 * Rename raw sockets to measurement sockets since "raw" has another
3239 meaning in networking community.
3242 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3243 and offers features not belonging to GRAS but that most applications
3244 need. One day, it may be a set of plugins loadable at runtime.
3245 * New module: bandwidth
3246 bandwidth measurement between arbitrary nodes running this module. [MQ]
3248 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3250 SimGrid (2.94) unstable; urgency=low
3252 The first beta release of SimGrid 3 !
3256 * Update the main page and the FAQ. Adding references to gforge.
3259 * Add a gras_os_getpid function.
3262 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3263 * Extend the logs so that they also print PID, hostname, date, ... if
3265 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3266 and kill the old version which were in testsuite/
3267 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3271 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3272 for users. Instead, each programming environment (GRAS, MSG,...) use it
3273 when needed to provide such a feature to users.
3274 Don't shortcut the mechanism or you will also shortcut the virtualization
3275 you need on the simulator.
3279 * Cleanups in configury with regard to compile optimization/warning flags.
3280 Also add -fno-loop-optimize to any powerpc since it's the optimization
3281 killing gcc (< 3.4.0).
3282 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3283 needed by MSG examples complications
3284 * Borrow configury beautifications from PHP
3287 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3288 warning (thanks loris for stumbling into this one).
3289 * Bugfix: stop loading private headers (gras_config.h) from the public
3293 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3294 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3295 RL would imply a DNS resolver, which is *hard* to do in a portable way
3296 (and therefore delayed).
3297 * Implement a real timer mechanism and use it in timing macros. This allows
3298 to avoid rounding errors and get a 0.000005 sec precision in timing
3299 macros. While I was at it, various cleanups:
3300 - allow to declare more than one timed section per file (fix a stupid bug)
3301 - move some private declaration to the right place
3302 - merge conditional execution and timing macros into emulation module
3303 - document the module
3304 - make sure the module cleanups its mess on gras_exit
3305 * Documentation improvements:
3306 - (new) how to compile applications using GRAS
3307 - (new) emulation support (timing macros)
3309 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3311 SimGrid (2.93) unstable; urgency=low
3313 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3316 - Use Paje properly where used. Still to be sanitized properly.
3317 - Portability fix: Add an implementation of the contexts using pthread
3321 - Add xbt_procname(): returns the name of the current process.
3322 Use it to show the current process's name in all logging.
3324 - fix detection of older flex version and the reaction, since we do
3325 depend on modern ones (we use lex_destroy)
3326 - Better separation of SG and RL in the libs: remove all simulation code
3327 from libgras. As a result, this lib is now only 200k when stripped.
3328 Some of the xbt modules may also be duplicated (two sets and such) and
3329 should be cleaned/killed before SG3.
3330 - Insist on using xlC on AIX because of weird problems involving gcc there.
3331 - Cleanup the make remote stuff. This is now done by scripts
3332 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3333 mainly for our private use, but we're working on changing them to user
3336 - Bugfix: flush the socket on close only if there is some *output*.
3337 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3339 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3342 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3344 SimGrid (2.92) unstable; urgency=low
3346 Alpha 3 on the path to SimGrid 3
3350 - New! First try of benchmarking macros.
3351 - New! First try so that gras_stub_generator generate deployment and
3352 remote compilation helpers.
3354 - Bugfix: Initialization fix in msg_test.
3358 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3360 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3362 - API Change: gras_os_sleep to take the amount of seconds as a double.
3363 Accepting an int was error prone since it was the only location where
3364 seconds were coded as such. It leaded to damn rounding errors.
3365 - Bugfix: Hard to belive that timers ever worked before this.
3367 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3369 SimGrid (2.91) unstable; urgency=low
3371 Alpha 2 on the path to SimGrid 3
3375 - Bug fix in the lmm_solver.
3377 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3378 through the function MSG_paje_output.
3379 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3380 - It is possible to bound the rate of a communication in MSG with
3381 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3382 to write it in the changelog).
3383 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3388 - Various cleanups to the autotools stuff
3389 - Begin to move Gras examples to examples/gras/
3390 - Let make distcheck work again (yeah!)
3392 - documentation overhauled using doxygen.
3393 gtk-doc-tools is dead in SimGrid now.
3394 - Automatically extract all existing logging categories, and add the list
3395 to the documentation (long standing one, to say the less)
3397 - Cleanup the known architecture table. Reorder the entries to group what
3398 should be, and use a more consistent naming scheme.
3399 (some of the test dataset are still to be regenerated)
3400 - New! Allow library to register globals on each process just as userdata
3402 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3403 lookup time (for now).
3404 Use it in msg and trp.
3405 This cleans a lot the internals and helps enforcing privacy of the
3406 headers between the gras components.
3407 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3408 - Bugfix: gras_os_time was delirious in RL.
3409 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3411 - Reenable GRAS now that it works.
3413 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3415 SimGrid (2.90) unstable; urgency=low
3417 Alpha 1 on the path to SimGrid 3
3419 * It is a long time since the last release of SimGrid. I'm sorry about
3420 that but as I had told you, I was rewriting a lot of things. I apologize
3421 to those who had been reporting bugs to me and that I had not answered.
3422 If your bug is still in the new version, please tell me. Here is a
3423 summary of the main changes.
3425 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3426 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3427 portability, speed, and a lot more but you'll figure it out later.
3428 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3429 I wanted to release the new MSG as soon as possible and I have
3430 broken GRAS, which is the reason why, for now, only MSG is fully
3431 functional. A laconic description of these projects is available
3432 in the documentation.
3434 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3435 kernel optimized for our needs (called SURF but only the developers
3436 should use it). Hence, MSG is now roughly 30 times faster and I think
3437 that by rewriting a little bit MSG, I could event speed it up a little
3438 bit more. Beside the gain in speed, it is also much easier to encode a
3439 new platform model with SURF than it was with SG. More to come...
3441 * REVOLUTION 3: I have tried to change a little as possible the API of
3442 MSG but a few things really had to disappear. The main differences
3443 with the previous version are:
3444 1) no more m_links_t and the corresponding functions. Platforms are
3445 directly read from a XML description and cannot be hard-coded
3446 anymore. The same format is used for application deployment
3447 description. The new format is described in the documentation.
3448 Have a look in tools/platform_generation. There is a tiny script
3449 that converts from the old platform format to the new one. Concerning
3450 the application deployment format, parsing the old one is tricky.
3451 I think most of you should however be able to convert your files. If
3452 it is really an issue, I can write a C code that does the conversion.
3454 2) the toolbox tbx does not exist anymore. We now have a library
3455 with much more data-structures but without the hash-tables (we have
3456 dictionaries that are much faster).
3458 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3460 *****************************************************************************
3461 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3462 * don't feel like converting the oldies. (Mt) *
3463 *****************************************************************************
3466 Version 2.90: "the long awaited one"
3467 - Finished rewriting and debugging MSG. Rewrote the documentation.
3468 - disable GRAS for now since it needs to be ported to the newest SG
3471 - Finish the port to windows (using mingw32 for cross-compile)
3474 - Main loop and datastructures of SURF. A cpu resource object is
3475 functional. Surf can thus be used to create cpu's with variable
3476 performance on which you can execute some actions.
3478 2004-11-15 Martin Quinson
3479 - Port to ARM. Simply added the alignment and size descriptions. Should
3480 work, but the ARM machines are so slow that I didn't had the opportunity
3481 to 'make check' over there yet.
3483 2004-11-15 Arnaud Legrand
3484 - Trace manager now written. It uses a heap structure and is therefore
3485 expected to be efficient. It may however be speeded up (particularly
3486 when many events occur at the same date) by using red and black
3487 trees. One day maybe...
3488 - Max-min linear system solver written. It uses a sparse matrix
3489 structure taking advantage of its expected use. Most operations are
3490 O(1) and free/calloc are called as few as possible. The computation of
3491 the minimum could however be improved by using a red and black tree
3494 2004-11-03 Arnaud Legrand
3495 - Rename every gras_* function that was in xbt/ to its xbt_
3497 - Add a heap and a doubly-linked list to xbt
3498 - Added a dichotomy to the dictionaries. make check works as well before
3499 so I assume that the patch is correct. I do not know however if things
3500 run effectively faster than before now. :)
3502 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3503 SimGrid, and the version number is bumped to 2.x
3505 2004-10-29 Martin Quinson
3506 - Introduction of the remote errors.
3507 They are the result of a RMI/RPC on the remote machine.
3508 ErrCodes being scalar values, you can't get the host on which those
3509 errors did happen. Extending the error mechanism as in Gnome is possible.
3510 No idea yet whether it is a good idea.
3512 2004-10-28 Martin Quinson
3513 - Interface revolution: the Starred Structure Eradication.
3514 I used to do typedef struct {} toto_t; and then handle *toto_t.
3515 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3516 - ???_t is a valid type (builded with typedef)
3517 - s_toto_t is a structure (access to fields with .)
3518 - s_toto is a structure needing 'struct' keyword to be used
3519 - e_toto_t is an enum
3520 - toto_t is an 'object' (struct*)
3522 typedef struct s_toto {} s_toto_t, *toto_t;
3523 typedef enum {} e_toto_t;
3524 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3525 s_toto_t) is private.
3527 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3528 so that it can set the variable to NULL. It was so for dicts and sets,
3529 it changed for dynars.
3531 - Fix a bunch of memleaks in dict_remove
3532 - Fix a bug in sg/server_socket opening: it failed all the time.
3534 2004-10-07 Martin Quinson
3535 - Speed up dynar lookup operation a bit.
3537 gras_dynar_get is dead.
3539 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3540 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3541 (which gives you the address of the stored data).
3543 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3544 copy of the data using an affectation to do the job and not a memcpy.
3546 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3547 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3549 It does not really speedup the dynar test because they are
3550 setting elements all the time (and look them seldom). But the dict does
3551 far more lookup than setting.
3553 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3555 2004-10-05 Martin Quinson
3556 - Allow to (en/dis)able the cycle detection at run time.
3558 Whether we should check for cycle or not is now a property of each
3559 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3560 datadesc_cycle_unset allow to remove this property when previously set.
3562 Note that the cycle detection is off by default since it impacts the
3563 performance. Watch the data you feed GRAS with ;)
3565 This property is hereditary. Any element embedded in a structure having it
3566 set have it set for the time of this data exchange.
3568 You should set it both on sender and receiver side. If you don't set it on
3569 sender side, it will enter an endless loop. If you forget on receiver
3570 side, the cycles won't be recreated after communication.
3572 - Header reorganization.
3573 Kill gras_private.h, each submodule must load the headers it needs.
3575 2004-10-04 Martin Quinson
3576 - Interface revolution: do not try to survive to malloc failure.
3578 Now, gras_malloc and friends call gras_abort() on failure.
3579 As a conclusion, malloc_error is not a valid error anymore, and all
3580 functions for which it was the only gras_error_t return value are
3581 changed. They now return void, or there result directly.
3582 This simplify the API a lot.
3584 2004-09-29 Martin Quinson
3585 - Re-enable raw sockets.
3586 Created by gras_socket_{client,server}_ext;
3587 Used with gras_raw_{send,recv}
3590 It should allow to kill the last bits of gras first version soon.
3592 This is not completely satisfactory yet (duplicate code with
3593 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3596 - Simplify transport plugin (internal) interface by not passing any
3597 argument to _server and _client, but embedding them in the socket
3600 2004-09-28 Martin Quinson
3601 - Finish the port to AIX.
3602 autoconf was my problem (segfault within the malloc replacement
3603 function. No idea why)
3605 2004-09-16 Martin Quinson
3606 - Fix some size_t madness on 64bit architectures.
3608 2004-09-08 Martin Quinson
3609 - Reduce the number of system headers loaded, overload some more system
3610 calls (such as malloc to cast the result of the system one, and work
3612 - Fix and reintroduce the config support
3614 2004-09-07 Martin Quinson
3615 - Source code reorganization to allow Arnaud to surf all over there.
3616 - Allow to document the logging categories.
3617 - Remove all uppercase from logging categories and useless cleanup in names.
3619 2004-08-18 Martin Quinson
3620 Version 0.6.2 (protocol not changed; API changed)
3621 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3622 gras_error_t), and NULL when not found. Functions expecting a msgtype
3623 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3624 hopefully usefull message.
3625 - Portability to prehistoric sparcs again
3627 2004-08-17 Martin Quinson
3628 Version 0.6.1 (protocol not changed; ABI not changed)
3629 - prealloc some buffers to speed things up
3631 2004-08-11 Martin Quinson
3632 Version 0.6 (protocol not changed; ABI expended)
3633 - The parsing macro can deal with the references, provided that you add
3634 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3636 2004-08-09 Martin Quinson
3637 Version 0.5 (protocol not changed; ABI changed)
3638 - Allow to off turn the cycle detection code in data exchange at
3639 compilation time. It should be at run time, but I'm short of time (and
3640 the config stuff is still broken). That way, we keep dict out of the
3641 critical path, which is good because the performance is poor:
3642 - search not dichotomial yet
3643 - dynar give no way to access their content and memcpy everytime
3644 - In composed data description (struct, ref and so on), stop foolness of
3645 keeping the subtype's ID, but store the type itself. This keeps sets out
3646 of the critical path, which is good since they rely on dynar and
3647 dictionnaries. The only loose of that is that we cannot detect the
3648 redeclaration of a structure/union with another content (but I'm not sure
3649 the code detected well this error before anyway). We still can detect
3650 the redefinition discrepancy for the other types.
3651 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3652 it breaks the code because of type-punning used all over the place).
3653 This breaks on all non-gcc architectures (for now).
3655 All those changes (plus the buffer of last time) allow me to gain 2 order
3656 of magnitude on cruel tests consisting of 800000 array of integers on two
3657 level of a hierarchical structure (200 secondes -> 4 secondes)
3660 - the selector of reference must now return the type it points to, not
3661 the ID of this type.
3663 2004-08-06 Martin Quinson
3664 Version 0.4 (protocol changed; ABI not changed)
3665 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3666 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3667 - Implement a buffer transport, and use it by default (it relies on tcp in
3668 real life and on sg in simulation).
3669 That's a bit hackish since I had a new field to the structure to store
3670 its data without interfering with the subtype ones. Inheritance
3671 is tricky in C. And that's a kind of reverse inheritance with one class
3672 derivating two classes. Or maybe a game with java interfaces. Anyway,
3673 that's damn hard in C (at least).
3674 Moreover, I got tired while trying to ensure plugin separation and
3675 genericity in SG mode. MSG wants me to do weird things, so let's go for
3676 cruel hacks (temporarily of course ;).
3677 See comment in transport_private.h:71
3678 - do not include all the _interface headers in private but in the files
3679 which really need them (to cut the compilation time when they are
3682 2004-07-26 Martin Quinson
3683 Version 0.3 (protocol not changed; ABI changed)
3684 - Major overhault of the datadesc interface to simplify it:
3685 - shorted the function names:
3686 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3687 - add a trivial way to push/pop integers into the cbps without malloc.
3688 This allows to make really generic sub_type description, which simply
3689 pop their size of the stack.
3690 - add a function gras_datadesc_ref_pop_arr() which does what users want
3691 most of the time: Declare a dynamic array (which pops its size of the
3692 stack) and declare a reference to it. Poor name, but anyway.
3693 - kill the post-send callback, add a post-receive one
3695 2004-07-23 Martin Quinson
3696 Version 0.2 (protocol changed; ABI changed)
3697 - add some testing for cpbs in the test cases, and fix some more bugs.
3698 This invalidate again the little64 data file, since I cannot regenerate
3700 - remove an awfull optimization in the logging stuff, allowing me to:
3701 - understand it again
3702 - learn gcc how to check that the argument match the provided format
3703 - fix all errors revealed by gcc after that
3704 - internal keys of dict are not \0 terminated. Deal with it properly in
3705 loggings instead of segfaulting when the user want to see the logs :-/
3707 2004-07-22 Martin Quinson
3708 - Fix some stupid bug preventing cbps (callback postit) from working
3710 2004-07-21 Martin Quinson
3711 - Some documentation cleanups
3712 - remove the useless last argument of msgtype_declare
3713 - rename the Virtu functions to fit into the 'os' namespace
3714 - move headers src/include -> src/include/gras/ and stop fooling with
3715 gras -> . symbolic link
3716 - make distcheck is now successful
3718 2004-07-19 Martin Quinson
3720 - Build shared library also
3721 - Install html doc to the right location
3722 - stop removing maintainer files in make clean
3723 - build tests only on make check
3725 2004-07-13 Martin Quinson
3727 - No major issue in previous version => change versioning schema
3728 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3729 corresponding dataset.
3731 2004-07-11 Martin Quinson
3733 - Get it working with any kind of structure (we can compute the padding
3734 bytes remotely for all the architectures I have access to)
3735 - Implement the structure parsing macro (still not quite robust/complete)
3736 - Improvement to the remote testing toysuite
3738 2004-07-10 Martin Quinson
3739 [autoconf mechanism]
3740 - get ride of a bunch of deprecated macros
3741 - actually run the test for two-compliment, not only compile it :-/
3742 - test whether the structures get packed (and bail out if yes. Damn.
3743 Alignment is a serious matter)
3744 - test whether the structures get compacted (but respecting the alignment
3745 constraints of each types)
3746 - test whether the array fields of structures can straddle alignment boundaries
3748 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3749 (took me 2 hours to find that bug, looking at the wrong place)
3750 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3751 before sending/receiving and are used to compute the offsets of fields
3752 - Given that padding size depend even on compiler options, keep track of
3753 alignment and aligned_size only for the current architecture. Not a big
3754 deal since we send structure fields one after the other (seems
3756 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3757 program, let it work on linux/gcc/little32. portability todo.
3759 2004-07-08 Martin Quinson
3760 - import and improve remote compilation support from FAST
3761 - make sure make check works on half a dozen of machines out there
3763 2004-07-07 Martin Quinson
3764 Let's say it's version 0.0.3 ;)
3765 - Implement conversions (yuhu!)
3766 - Let it work on solaris (beside conversion, of course)
3767 - Stupid me, using rand() to generate the conversion datatests in not wise.
3769 2004-07-06 Martin Quinson
3770 - Let make dist work, since I'm gonna need it to compile on remote hosts
3771 - Let Tests/datadesc_usage write the architecture on which the file was
3772 generated as first byte.
3773 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3774 architecture descriptions.
3775 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3776 on those architectures.
3777 - Optimization: send/recv array of scalar in one shoot
3779 2004-07-05 Martin Quinson
3780 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3782 - Plug a whole bunch of memleaks
3783 - each process now have to call gras_{init,exit}. One day, their log
3784 settings will be separated
3785 - Continue the code factorisation between SG, RL and common in Transport.
3787 2004-07-04 Martin Quinson
3789 - Redistribution between SG and RL.
3790 We wanna have to accept in SG, so move accepted related parts of RL in
3791 the common part. (more precisely, the dynar of all known sockets is no
3792 more a static in transport.c, but part of the process_data)
3794 [gras_stub_generator]
3795 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3797 2004-07-03 Martin Quinson
3798 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3799 them all up in one shot)
3801 - Fix the ugly case of reference to dynamic array.
3802 - New (semi-public) function gras_datadesc_size to allow the messaging
3803 layer to malloc the needed space for the buffer.
3805 - gras_socket_close now expect the socket to close (and not its address to
3806 put NULL in it after it). This is because the socket passed to handlers
3807 is one of their argument (=> not writable).
3809 - propagate the interface cleanup from last week in datadesc, ie remove a
3810 superfluous level of indirection. User pass adress of variable
3811 containing data (both when sending and receiving), and not of a variable
3812 being a pointer to the data. Let's say that I like it better ;)
3813 The price for that is constructs like "int msg=*(int*)payload" in
3814 handlers, but it's a fine price, IMHO.
3816 - Let it work in RL (yuhu)
3818 2004-06-21 Martin Quinson
3820 - porting SG plugin and SG select to new standards (works almost).
3821 - plug memleaks and fix bugs around.
3824 - cleanup the prototype of data recv and force users to specify when they
3825 want to handle references to objects. Test case working even for cycles.
3826 - plug memleaks. Valgrind is perfectly ok with this.
3828 2004-06-12 Martin Quinson
3830 - cleanup the separation between plugin and main code in plugin creation
3832 2004-06-11 Martin Quinson
3834 - Reput hook for raw sockets, needed for BW experiments
3835 - kill a few lines of dead code
3836 [Data description] Interface cleanup
3837 - gras_datadesc_by_name returns the searched type or NULL.
3838 That way, no variable is needed to use a type desc once, which makes
3840 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3841 two parameters were strings (field name, type name), leading to
3843 [Dicos] Interface cleanup
3844 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3845 This is consistant with the dynar API.
3847 2004-04-21 Martin Quinson
3849 - Porting to new standards.
3851 - interface cleanup.
3852 There is no bag anymore, no need to take extra provision to mask the
3853 pointers behind "ID".
3854 Better splitup of functions between files create/exchange/convert.
3855 This is still a bit artificial since convert and receive are so
3856 interleaved, but anyway.
3858 - add a queued message list to procdata (the ones not matching criteria
3860 - factorize some more code between SG and RL wrt procdata
3862 - use gras_exit in example to track memleaks
3863 - get rid of gs_example now that GS is properly integrated into gras
3864 - update run_test to integrate the latest tests (datadesc)
3866 - rename WARNINGn macros to WARNn since it prooved error-prone
3868 2004-04-19 Martin Quinson
3870 - register init/exit functions within gras module mechanism
3871 - send/receive function.
3872 Convertion is not implemented, but short-cutted if not needed.
3873 struct/array elements are sent one by one (instead of block-wise), but
3874 nobody really cares (yet). Get a prototype before optimizing.
3875 - tests (using a file socket) for DD send/receive on:
3876 - base types: int, float
3877 - array: fixed size, string (ie ref to dynamic string)
3878 - structure: homogeneous, heterogeneous
3879 - chained list, graph with cycle
3880 Believe it or not, valgrind is not too unhappy with the results. The
3881 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3885 - Bugfix when using a filename explicitely (instead of '-')
3887 2004-04-09 Martin Quinson
3889 - factorize more code between RL and SG in socket creation
3890 - Complete the implementation and tests of:
3892 o file (only in RL, and mainly for debugging)
3894 I lost 3 days to design a portable address resolver, and then decided
3895 that the prototype mainly have to run on my box.
3896 Addressing portability too early may be like optimizing too early :-/
3898 - use gras_init in the Tests instead of the crappy parse_log_opt
3899 (the latter function is removed)
3900 [Conditional execution]
3901 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3902 [Code reorganisation]
3903 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3904 Build examples against the RL library, since there is no way to disable
3905 its creation for now.
3907 For information, the beginning of coding on GRAS was back in june
3908 2003. I guess that every line has been rewritten at least twice since