1 ----------------------------------------------------------------------------
3 SimGrid (3.23) June 25. 2019
5 The Exotic Solstice Release.
8 - SunOS and Haiku OS support. Because exotic platforms are fun.
9 - Stop setting random seed with srand48() at initialization.
10 - Use addr2line as a fallback for stacktraces when backtrace is not available.
11 - Build option -Denable_documentation is now OFF by default.
12 - Network model 'NS3' was renamed into 'ns-3'.
15 - Simgrid can now hopefully be installed with pip.
18 - wait_any can now be used for asynchronous executions too.
21 - New log appenders: stdout and stderr. Use stdout for xbt_help.
25 - 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
26 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
27 - MPI interface is now const correct for input parameters
28 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
29 - Fortran bindings for async collectives.
30 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
33 - Remove option 'model-check/record': Paths are recorded in any cases now.
34 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
35 always sparse. That's an immense gain in memory, and can even be
36 faster because of cache effects. So there is no need to clutter the
37 code to allow the user to go for the unefficient mode.
40 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
41 models proved to be more accurate than these old experiments.
43 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
44 - FG#1: Broken link in error messages
45 - FG#2: missing installation documentation
46 - FG#3: missing documentation in smpirun
47 - FG#6: Python bindings not available on PyPI
48 - FG#7: simple cmake call requires doxygen
49 - FG#8: make python bindings an optional dependency
50 - FG#10: Can not use MSG_process_set_data from SMPI any more
51 - FG#11: Auto-restart actors forget their on_exit behavior
52 - FG#12: -Denable_lto=OFF doesn't disable LTO
53 - FG#13: Installs unstripped file 'bin/graphicator'
54 - FG#14: Installs the empty directory 'doc/simgrid/html'
55 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
56 - FG#17: Dead link in doc (pls_ns3)
57 - FG#20: 'tesh --help' should return 0
58 - FG#21: Documentation link on http://simgrid.org/ broken
59 - FG#22: Debian installation instruction are broken
60 - FG#26: Turning off a link should raise NetworkFailureException exceptions
61 - GH#133: Java: a process can run on a VM even if its host is off
62 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
63 - GH#326: Valgrind-detected error for join() when energy plugin is activated
65 ----------------------------------------------------------------------------
67 SimGrid (3.22) April 1. 2019
69 The Easter Christmas Release (this one is somewhat late).
72 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
73 - Not complete yet: asynchronous activities (amongst others) are still missing
74 - Still ongoing: the interface may change in the future. We need more
75 testers! Please report any glitches.
76 - No new project using Java should start now. Please switch to Python.
79 - Some of the internal cleanups may lead to speed improvements:
80 - The hard limitation on the amount of simulated actors with
81 Java+Mac was removed. Now, the available memory is the only limit.
82 - Our refcounting was tidyied, leading to 10% speedups in some cases.
83 - We are still working on making our code robust to the actor kills
84 and hosts' churn. Things are improving, but it's not perfect yet.
85 - Replaced our own code to display a backtrace (that was forking addr2line)
86 with the Boost.Stacktrace library.
87 You won't see your backtraces without this optional dependency.
88 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
89 - Stop setting random seed with srand() at initialization.
92 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
93 XML file version remains 4.2 since old files are still compatible.
96 - Process termination which was broken at version 3.21 has been repaired.
97 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
98 - Hide the examples into examples/deprecated. New users should use Python.
101 - Drop MSG_process_create_from_stdfunc() from the API.
102 This C++ function was a pimple in the C API, made necessary at some
103 point by the Java bindings. This is fixed now.
104 - Hide the examples into examples/deprecated. New users should use S4U.
105 - MSG_process_create and MSG_process_attach now crash if the host used
106 to run this process is off.
107 - Fix the protype of MSG_process_on_exit()
108 Now use (int,void*) callbacks instead of (void*,void*) ones.
109 The implementation was ways too messy to actually work, I guess.
112 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
113 too often and provide more accurate timings.
114 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
115 - MPI_Alltoallw support
116 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
117 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
118 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
119 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
121 - Basic implementation of generalized requests (SMPI doesn't
122 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
125 - Drop sg_cmdline. Please use xbt_cmdline instead.
126 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
127 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
128 OsSemaphore is implemented in a portable way with C++11 threads.
129 This should allow much more threads to be created at the same time,
130 allowing Mac and Java users to simulate many more actors.
131 - Implement the 'thread' factory with std::thread instead of xbt ones.
132 It is not possible to set the stack size with threads anymore, but
133 -fsplit-stack is the way to go nowadays when using threads.
134 - Drop the xbt_os_thread_t module (now unused)
135 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
136 - Drop xbt_str_join_array().
137 - Drop cunit, use Catch2 instead.
140 - Many cleanups in the kernel::activity namespace. This was long
141 overdue, and shall open the path to many future endeavors.
144 - #132: Java: a process can not shut down its own host
145 - #220: S4U: detached send causes exception if sender terminates soon after sending
146 - #261: Document the parameters of parallel execution's constructor
147 - #300: [s4u] BarrierPtr is missing
148 - #314: SMPI args internal cleanup
149 - #316: Fix a bug related to the CPU utilization of multi-core VM
150 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
151 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
152 - #325: Turning off a host has different behavior on sleeping actors and computing actors
154 ----------------------------------------------------------------------------
156 SimGrid (3.21) October 3. 2018
158 The Restarting Documentation (TRD) Release.
161 - Convert the user manual to Sphinx for improved usability.
162 Unstable version now hosted on https://simgrid.frama.io/simgrid
163 Many glitches remain; stay tuned.
164 - Restrict the Doxygen reference API to the installed header files.
165 (documenting all internals this way is overkill + doxygen scales badly)
166 - New tutorial on S4U (highly inspired from the old MSG one, but with
167 a git repository to fork as a starting point, and a docker image)
168 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
169 - The Developper manual is still to be converted (not compiled until then)
170 - Some parts are still missing in this conversion (such as the
171 platform and deployment sections) while others were blindly converted
172 and would need more love (such as the configuration flags).
173 Things will certainly further improve in the future releases.
176 - Fully reimplement auto_restart mechanism that was utterly broken
177 (fix #22, #131 and #281 that were all very related)
178 - Implement semaphores (fix #303 and #291)
179 - When creating an actor from a function and its parameters,
180 move-only parameters are not allowed anymore, as it would prevent
181 the actor to be restartable if its parameters are consumed on run.
182 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
183 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
184 IO operations that can be started, waited for, or canceled as a regular
186 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
190 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
191 - New host variable: 'core_count'
194 - Remove xbt_os_thread_specific features
195 - Remove portability wrapper to condition variables
196 - Remove xbt_os_thread_yield()
199 - MPICH collective selector now mimics MPICH 3.3b
200 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
201 - MPI_Init(NULL,NULL) is now allowed
202 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
203 - Command-line arguments should now be handled properly in Fortran simulations
204 - MPI Attributes and related callbacks should work in Fortran
205 - Apps using C + Fortran should now work
206 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
207 - Support MPI_Error_String
210 - Due to an internal bug, Msg.run() must now be your last line.
211 We hope to fix it in a future release, and we are sorry for the inconvenience.
214 - #22: Process autorestart seem to only work with CAS01 cpus
215 - #56: Feature request: dynamic SMPI replay
216 - #93: simgrid should not eat --help
217 - #111: How to change the loopback link characteristics?
218 - #116: Platform cluster radical
219 - #117: Platform cluster/host order
220 - #129: Replay_multiple on steroid
221 - #131: Java : autostart does not work
222 - #143: Setting a breakpoint at a given time
223 - #235: Network energy models should be integrated
224 - #264: Add ptask L07 resource tracing
225 - #271: Dynamic and manual replay of SMPI traces
226 - #279: Breakpoints option uses time deltas instead of absolute time
227 - #281: Daemonized actor no longer daemonized after an auto-restart
228 - #282: TIT convertor erase traces when using absolute path in the trace list file
229 - #285: segfault when a s4u actor kills itself with msg process tracing activated
230 - #286: Pajé traces are not exposing the number of cores
231 - #287: Command-line parsing should stop when encountering '--'
232 - #288: MPI_Init(NULL, NULL)
233 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
234 - #290: Method to check if a mailbox has a message ready to be consumed
235 - #291: [s4u] semaphores do not seem available yet
236 - #294: Alltoall collective implementation of mpich
237 - #295: Starting a Host as OFF is not supported
238 - #296: DTD too permissive
239 - #299: Add a s4u barrier example
241 ----------------------------------------------------------------------------
243 SimGrid (3.20) Released June 24. 2018
245 The proxy snake_case() release.
248 - The whole API is now using snake_case() instead of mixing it with camelCase.
249 Compatibility wrappers in place for one year.
250 - We took the opportunity to increase the consistency of the API.
251 Please report any remaining glitches.
252 - New function simgrid::s4u::Host::get_actor_count:
253 Returns the number of actors running on a specific host.
256 - SMPI is now tested with ~45 proxy apps from various sources, with none or
257 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
258 - Replay: The replay file has been re-written in C++.
259 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
260 supported. They are stored in the trace and used when replayed.
261 - Basic support of MPI_Cancel. Robustness not guaranteed.
262 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
263 (only for MPI_COMM_TYPE_SHARED).
264 - New option: smpi/privatize-libs, to add external shared libs to be privatized
265 by SMPI. They will be copied locally and loaded separately by each process.
266 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
267 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
268 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
269 with internal index optind with multiple processes. Only works if MPI_Init has
271 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
272 syntax without preprocessing tricks.
275 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
276 behavior tracing is no longer limited to MSG
277 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
278 this means that traces now need to be updated or re-obtained; both Irecv
279 and Isend lines in your traces have now in total 5 mandatory fields:
280 <rankid> <command> <to/from rankid> <tag> <size>
281 To update your traces, it suffices to add a 0 for the tag here.
282 - TIT now also supports waiting for a distinct request via MPI_Wait.
283 Wait/Test now wait for a specific request, not just the last one that was
284 issued. This unfortunately means another update, because we need to
285 identify which request you want to wait for. We do this via the
286 triplet (sender, receiver, tag), which needs to be added:
287 <rankid> <command> <sender> <receiver> <tag>
288 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
289 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
292 - The deprecation of MSG is ongoing (but this should not impact you).
293 Many MSG functions are now simple wrappers on the C API of S4U. If
294 you wish to convert your code to S4U, find the S4U counterparts of
295 your MSG calls in src/msg/msg_legacy.cpp.
296 - MSG can still be used, but won't evolve anymore.
299 - Allow to run the Link energy plugin from the command line with
300 --cfg=plugin:link_energy
301 - Rename Energy plugin into host_energy
302 - Rename Load plugin into host_load
305 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
306 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
307 - Improve performance when handling timeouts of simix synchros.
310 - Config: the C API is now deprecated (will be removed in 3.23), and
311 the C++ API has been slightly improved.
314 - Fix several build issues on OSX.
315 - Move simgrid_config.h to simgrid/config.h (old header still working)
318 - #143: Setting a breakpoint at a given time
319 - #150: Inconsistent event names in SMPI replay
320 - #258: daemonized actors hang after all non-daemonized actors have completed
321 - #267: Linker error on unit_tmgr
322 - #269: SMPI: tracing of MPI_Wait/all/any broken
323 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
325 ----------------------------------------------------------------------------
327 SimGrid (3.19.1) Released March 22 2018
329 The Fixed ABI Release.
331 As you may know, we are currently refactoring SimGrid in deep.
332 Upcoming SimGrid4 will be really different from SimGrid3: modular,
333 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
335 Our goal is to smooth this transition, with backward compatibility
336 and automatic update paths, while still progressing toward SimGrid4.
338 SimGrid remains open during works: The last pure SimGrid3 release was
339 v3.12 while all subsequent versions are usable alpha versions of
340 SimGrid4: Existing interfaces remain unchanged, but the new S4U
341 interface is budding and the internals are deeply reorganized.
343 Since 2015, we work hard to reduce the changes to public APIs. When
344 we need to rename a public library symbol in S4U, we let your
345 compiler issue an explicative warning when you use the deprecated
346 function. These messages remain for four releases, i.e. for one full
347 year, before turning into an error. Starting with v3.15, your can
348 also adapt to API changes with the SIMGRID_VERSION macro, that is
349 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
351 Starting with this v3.19.1, our commitment to reduce the changes to
352 the public interfaces is extended from the API to the ABI: a program
353 using only MSG or SimDag and compiled against a given version of
354 simgrid can probably be used with a later version of SimGrid without
355 recompilation. We will do our best... but don't expect too much of
356 it, that's a really difficult goal during such profund refactoring.
358 The difference between v3.19 and v3.19.1 is that the former was
359 accidentally breaking the ABI of MSG, while the later is restoring
362 S4U and kernel APIs will still evolve until SimGrid4, with one-year
363 deprecation warnings as currently. In fact, cleaning up these
364 interfaces and converting them to snake_case() is one release goal of
365 v3.20. But don't worry, we are working to smooth this upgrade path.
367 In summary, new projects should start with S4U to benefit of the
368 future, but old MSG projects should still be usable with no change.
370 ----------------------------------------------------------------------------
372 SimGrid (3.19) Released March 20 2018
374 The Moscovitly-cold Spring release.
376 SMPI user-visible improvements:
377 - Support of MPI_Type_create_subarray
378 - Unimplemented MPI calls now abort (instead of issuing a warning)
379 Some harmless calls around errhandlers still only issue a warning.
380 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
381 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
382 is now possible (mmap privatization was refactored to this end).
383 Adding new SMPI instances should also be technically doable now.
385 SMPI internal cleanups:
386 - Removed index notion from SMPI -- no more "getPid() - 1"!
387 In tracing files, actors are called rank-<PROCESS-ID> but these are
388 NOT the real ranks (tracing several communicators is not supported yet).
389 - Deprecate smpi_process_index() (will be removed in 3.22)
390 - Better testing on classical proxy apps, through an external project:
391 https://github.com/simgrid/SMPI-proxy-apps/
394 - Use a graphical TOC to make it easier to find the documentation you need
395 - Revamp the MSG tutorial
398 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
399 - Remove parameter of MSG_process_killall().
400 Resetting the PID was bogus anyway (several actors could have the same PID).
403 - Execution->setHost() can be called after start(), to migrate it.
404 - Comm::test_any() is now implemented.
405 - s4u::Actor now has onCreation() and onDestruction() signals.
406 - Install some kernel header files for the users' plugins and more.
407 Warning, their API is really not stable yet.
408 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
409 include/simgrid/kernel/routing/*Zone.hpp
412 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
414 Build System and other cleanups:
415 - Remove unused run-time parameter "tracing/onelink-only".
416 - SimGrid now works with intel compilers (MC still buggy)
417 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
418 - tesh kills subprocesses on timeout (but not on windows)
421 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
424 - #194: Feature request: simgrid::s4u::Comm::test_any()
425 - #245: migrating an actor does not migrate its execution
426 - #253: Feature Request: expose clusters as objects
427 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
428 - #255: Tesh broken on Windows
429 - #256: Modernize FindSimGrid.cmake
430 - #257: Fix (ab)use of CMake install
432 ----------------------------------------------------------------------------
434 SimGrid (3.18) Released December 24 2017
436 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
438 MSG IS NOW DEPRECATED. Please use S4U for new projects.
439 - Support for MSG should not be removed from SimGrid before 2020,
440 but future evolutions will be limited to the integration of
441 user-provided patches.
444 - s4u::onDeadlock() signal, run before stopping the simulation.
445 - s4u::Exec: asynchronous executions (abstraction of background
447 - s4u::Host->getLoad() returns the achieved speed in flops/s
450 - Link::name() is deprecated and will be removed in v3.21, use
451 Link::getCname() instead.
452 - Mailbox::getName() changed to return a std::string, use
453 Mailbox::getCname() instead to get a char*.
454 - Storage::getName() changed to return a std::string, use
455 Storage::getCname() instead to get a char*.
456 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
457 to have both allocation and deallocation of the map in user space.
460 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
461 The amount of flops remaining to do is a vector, not a scalar.
462 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
463 name implies on both sequential and parallel tasks.
464 - Both changes fix GitHub's #223 using PR #237 as a basis.
465 Thanks Michael Mercier.
466 - Most examples were converted to S4U and hidden elsewhere at tests.
467 Rationale: we still want MSG to work; we want newcomers to use S4U.
470 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
471 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
474 - Switch to the faster dlopen privatization mechanism by default
475 - Documentation improvments
478 - Live migration is getting moved to a plugin. Dirty page tracking is
479 the first part of this plugin. This imply that VM migration is now
480 only possible if one this function is called:
481 - C/MSG: MSG_vm_live_migration_plugin_init()
482 - C/C++: sg_vm_live_migration_plugin_init()
483 - Java: Msg.liveMigrationInit()
484 For C and C++, "simgrid/plugins/live_migration.h" also has to be
488 - Define class simgrid::xbt::Path to manage file names.
489 - Removed unused functions:
490 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
491 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
492 xbt_graph_export_graphviz()
493 - xbt/str.h: xbt_str_join()
494 - Remove unused datatypes:
495 - xbt/heap.h: use std::priority_queue or boost::heap instead
496 - xbt/swag.h: use boost::intrusive::list instead
499 - New link_energy plugin for the consumption of the links.
500 - All of the operations on files and storage contents have been
501 packaged into a plugin (src/plugins/file_system). The current
502 public interface can be found in
503 include/simgrid/plugins/file_system.h
504 To use these functions you now have to initialize the plugin by
505 calling MSG_storage_file_system_init() just after calling
506 MSG_init() or sg_storage_file_system_init() just after creating
510 - Remove the undocumented/untested tag <include>
513 - Remove viva specific tracing as the tool is no longer maintained
516 - #248: Kill of finished processes leads to segfault
517 - #240: xbt_cond_wait_timeout should gracefully return for C apps
518 - #239: please implement signal s4u::onDeadlock()
519 - #230: segfaults when exit() without run()
520 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
521 - #223: MSG_task_get_flops_amount() not working with parallel tasks
522 - #222: Actor::kill() doesn't really kill and segfaults
523 - #221: odd LMM warning when killing an actor
524 - #120: Memory leak when the processes are forcefully killed
526 ----------------------------------------------------------------------------
528 SimGrid (3.17) Released October 8 2017
530 The Drained Leaks release: (almost) no known leaks despite the tests.
532 Even very long simulations will preserve your memory: our testsuite
533 stresses SimGrid for over 45mn on fast machines for a coverage of
534 over 80%, and there is only one single known leak, of about 4kb.
538 - Comm.detach(): start and forget about asynchronous emission. A cleanup
539 handler may be given to free resources if the comm cannot be completed.
540 - this_actor::send(mailbox) is now mailbox->put()
541 - New: simgrid::s4u::Comm::wait_all()
542 - New: Host.execute() for remote executions.
545 - Unused option network/sender-gap is removed.
548 - Deprecate MSG_task_isend_with_matching(): This unused feature
549 really complicates our internals. Will be removed in v3.20.
552 - Improved context termination. It is now possible to free resources, even
553 when a process is forcibly killed.
554 - ContextBoost: add support for Boost versions above 1.61.
557 - Bring back run-time option --cfg=exception/cutpath to remove exception
559 - Removed unused functions:
560 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
561 xbt_str_rtrim(), xbt_str_trim().
562 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
565 - Removed header files obsolete since SimGrid 3.12:
566 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
568 - #3: SD_exit should be made optional
569 - #120: Memory leak when the processes are forcefully killed
570 - #159: Threading test regression in Actor refcounting
571 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
572 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
573 - #186: Actor::killAll() segfaults if some process is blocked on wait()
574 - #191: VM migration and pstate
575 - #192: Updating the energy consumptions of all hosts crashes with VMs
576 - #195: All actors have PID=0 in the logs
577 - #204: Sometimes segfault with thread contexts and mmap privatization
578 - #222: Actor::kill() doesn't really kill and segfaults
579 - #225: Actor::kill() doesn't really kill when victims are doing a join()
581 ----------------------------------------------------------------------------
583 SimGrid (3.16) Released June 22. 2017.
585 The Blooming Spring Release: developments are budding.
588 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
589 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
590 - Java JAR file should be finally fully working on Mac OSX too.
592 XML platforms: Switch to platform v4.1 format.
593 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
594 - <zone> can be used as a synonym for the now deprecated <as>
595 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
596 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
597 - <actor> can be used as a synonym for the now deprecated <process>
598 - state_file and avail_file periodicity is now easier to express
599 (check the documentation) the old behavior should still work.
600 * Storage: not backward compatible, but it seems that nobody ever used it
601 - Remove attribute 'content_type' of <storage_type>: was never used
602 - Make attribute 'model' of <storage_type> optional: for future usage
603 - Remove Bconnection model property: was never (in)validated. Replaced by
604 the maximum of the read and write bandwidth as a resource constraint.
607 - New and Backwards Compatibility break:
608 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
609 When one wants to get the list of tasks whose states have changed during a
610 simulation round, s/he has to allocate and free a dynar and use it as
611 argument to this function. The former SD_simulate (double how_long)
615 - Allow multicore VMs, with the correct sharing (unless you overcommit)
616 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
619 - New function to update the consumption of all hosts at once.
620 - Fix the model for multi-core hosts, linear in the amount of busy
621 cores with an abnormality for pIdle (see paper).
622 BUG: Still not satisfactory for ptask on mono-cores.
625 - The netzone are now available from the MSG API.
626 The old names still work, but are now deprecated.
629 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
630 - New option: smpi/keep-temps to not cleanup temp files
631 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
632 mallocs mode is used, allows to change the size of the fake file used
633 (default 1MB), to potentially limit the number of mappings for large runs.
634 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
635 - Fortran ifort and flang compilers support
636 - New RMA calls supported (experimental):
637 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
638 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
639 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
640 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
641 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
642 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
645 - New: this_actor::isMaestro() returns whether we are in kernel mode.
646 - Behind the scene, ActivityImpl are now automatically refcounted.
647 Removing this nasty bottleneck will greatly simplify our internals.
650 - Replay: New function xbt_replay_action_get():
651 Retrieve the function previously associated to an event type.
652 - DROPPED FUNCTION: xbt_str_varsubst()
653 - DROPPED MODULE: strbuff. We don't need it anymore.
654 - DROPPED MODULE: matrix. We don't need it anymore.
655 - DROPPED MODULE: lib. We don't need it anymore.
657 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
659 SimGrid (3.15) Released March 22. 2017
661 The Spring Release: continuous integration servers become green
662 We fixed even the transient bugs on all target architectures:
663 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
664 Windows, FreeBSD, NetBSD.
666 - Rename NetCards to NetPoints.
667 This was intended to help NS3 users, but that's not a netcard.
668 That's a point in the routing algorithm, let's avoid wrong simplifications.
671 - New: MSG_process_yield(). Stop and yield to other processes.
672 - New: MSG_process_daemon(). Daemon processes are automatically killed
673 when the last non-daemon process terminates
674 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
675 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
676 to make room for the upcoming network energy plugin.
677 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
680 - Ensure that an actor can kill itself with Process::exit()
681 - Kill the obscure NativeException. Nobody want to survive the issues
682 it denotes, so use JniException that is a RuntimeException (not to
683 be caught explicitly).
684 - Partial bug fix in initialization. SimGrid flags on command line were
685 consumed at C level but stayed in the original Java String[] args.
686 This could mess users' args[i] if SG flags were not put at the end of
688 The SimGrid flags are now removed from the Java arguments. However,
689 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
690 if args.length is greater than the number of YOUR OWN ARGUMENTS.
691 It might be if you have --log or --cfg flags in the command line.
692 - Fix numerous memleaks all around the place. In particular, around VMs.
696 - s4u::onPlatformCreated: right before the simulation starts
697 - s4u::onSimulationEnd: right after the main simulation loop
698 - s4u::onTimeAdvance: right after a clock change (time is discrete)
699 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
700 event from the availability_file changes the avail speed.
701 - Links are now usable from s4u
702 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
703 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
704 - New examples: The conversion of MSG examples to S4U has begun
705 - Actors: create, kill, migration, and suspend
706 - Applications: master-worker and token-ring
707 - Action replay: communications and storage
708 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
711 - Backwards Compatibility breaks
712 - The SD_route_get_list and SD_route_get_size functions have been removed.
713 They are replaced by the unique
714 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
715 The route from src to dst is built in the links dynar whose size can
716 be obtained with xbt_dynar_length.
717 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
718 sg_host_route_bandwidth and sg_host_route_latency.
719 Macros ensure the backwards compatibility, but you should fix your code
722 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
723 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
724 - Bump our claim of support from MPI 1.1 to MPI 2.2.
725 We don't support 100% of it, but it should be enough. Contact us if not.
726 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
727 naive and probably inaccurate)
728 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
729 mapped onto a unique small file using some system magic.
730 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
731 - Fortran: cleanups, fixes, support of user-added operations
732 - RMA: MPI_Accumulate are applied in correct order.
733 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
734 - Better support for MPI_IN_PLACE
735 - Support for MPI_Win attrs and keyvals.
736 - Support MPI_Comm_group_create, MPI_Type_size_x
737 - MPI ops should be properly applied to all allowed types
738 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
739 algorithm by a non failing one from mpich.
742 - Kill the fifo data container: we don't use it anymore.
745 - A <cluster> can now be created with different speed values to represent pstates.
746 see examples/platforms/energy_cluster.xml for an example.
748 SimGrid (3.14.159) Released December 28. 2016
750 The previous release was only a crude approximation of a Pi release;
751 this one should be a bit better.
753 - Revert a last minute change that broke on Mac OSX.
754 - Fix the changelog and NEWS files.
755 - Improve the documentation, in particular of the routing module.
756 - Integrate some patches flying around in Debian.
758 SimGrid (3.14) Released December 24. 2016
761 * The whole documentation was reworked and reorganized. There is
762 still much room for improvement here, but we're on it.
765 * We now need python3 for our testing infrastructure.
766 * Model checking needs libevent
769 * Now works on FreeBSD
772 * Privatization now works on FreeBSD
774 * Privatization is now activated by default in smpirun
775 You can turn it off with -no-privatize if something goes wrong.
777 * Call-location tracing for SMPI.
778 You can add the exact location (filename / linenumber) of an MPI call to
779 your trace files and slow down or speed up the simulation between two
780 consecutive calls by using an adjustment file (see the documentation).
782 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
783 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
784 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
785 values were ignored in some cases. The timings of these functions can now
786 be significantly different.
788 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
789 smpi/running-power is renamed to smpi/host-speed
791 * smpi/grow-injected-times option to enable or disable multiplication of the
792 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
793 make simulation less precise (but also much faster).
795 * smpirun script should be (much) faster for large deployments.
797 * SMPI tracing: fixed issue with poor matching of send/receives.
799 * Replay: Fix broken waitall
801 New functions and features
802 * MSG_parallel_task_execute_with_timeout, to timeout computations.
804 Dropped / renamed functions and features
805 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
806 - MSG_mailbox_is_empty() -> Mailbox::empty()
807 - MSG_mailbox_front() -> Mailbox::front()
808 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
809 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
810 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
811 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
812 * Don't pass the free_f parameter to property related functions:
813 - MSG_host_set_property_value()
814 - MSG_as_router_set_property_value()
815 - MSG_storage_set_property_value()
816 * VM properties. Since msg_vm_t are msg_host_t, just use
817 MSG_host_get_property_value() and friends
818 * VM I/O related things:
819 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
820 - Unimplemented save/restore methods
821 * MSG_as_router_get_property_value() was redundant with
822 MSG_environment_as_get_property_value().
823 - Removed MSG_as_router_*propert*() functions
824 - Added MSG_environment_as_set_property_value() for consistency
825 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
826 Well, they are still there for now, but deprecated with a warning.
827 Please switch to xbt_dict_new_homogeneous() before this is removed
829 * Task affinity. Its intended behavior (that was very badly tested
830 and probably not really working) was deceiving what most users
831 would have hoped here.
832 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
833 * Option network/coordinates is now useless and should be dropped.
836 * Added option storage/max_file_descriptors to allow more than 1024 files opened
839 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
840 * Added Dragonfly topology support
843 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
844 EiBps, EBps, Eibps, Ebps for bandwidth.
845 They may become useful to some lucky ones.
848 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
849 * ex/app_tokenring: new example, very similar to the MSG Token Ring
850 * ex/async_waitAll: new example, on asynchronous communications
853 * Memory usage should be decreased for simulations with a large number
854 of processes. This also helps for SMPI.
856 SimGrid (3.13) stable; urgency=low
858 The Half Release, a.k.a. the Zealous Easter Trim.
860 - We removed half of the lines, that were mostly experimental cruft.
861 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
862 (not counting blanks and comments -- according to openhub.net)
863 - The internals are now compiled in C++ (and will soon be clean C++)
864 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
865 59 klines of C, etc. We added only 29 klines of C++ in replacement.
867 * Backwards Compatibility breaks
868 - Removed Lua simulation bindings (switch to C or Java for that).
869 Lua can still be used to describe platforms
870 - Removed Java kernel plug-ins.
871 Will be reintroduced after the ongoing major internals reorg.
873 - the following functions were removed.
874 They were too specific and should be reimplemented in a generic
875 way, with filter function.
876 - MSG_task_listen_from_host
877 - MSG_mailbox_get_count_host_waiting_tasks
878 - MSG_mailbox_put_with_timeout was removed.
879 Please use MSG_task_send_with_timeout instead.
881 - the SD_application_reinit function was removed. It has been a noop for a while.
882 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
883 replaced by a more flexible mechanism.
884 - The following functions thus do not exist anymore
885 - SD_workstation_get_access_mode
886 - SD_workstation_set_access_mode
887 - SD_workstation_get_current_task
888 - Basic estimation functions have been removed but can easily be replaced
889 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
890 - SD_workstation_get_computation_time => amount / sg_host_speed()
892 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
893 Use VM.getSpeed()*load/100 for the legacy behavior.
895 - option enable_tracing was removed. It was not doing anything for a while.
896 - In the ModelChecker:
897 - the model-checker now ptraces the model-checked process which means
898 you cannot use a debugger on the latter anymore (we might make this
899 optional in the feature);
900 - removed soft-dirty page tracking;
901 - remove model-checked side snapshot management,
902 MC_snapshot() and MC_compare_snapshot();
903 - keep the MC_cut() function as a stub (it was not really working
904 in the previous release).
907 * All options are consistently in kebab-case. Old names are kept as alias.
910 * Switch to platform v4 format.
911 - Rename from 'power' to 'speed' the attributes describing the amount of
912 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
913 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
914 - In <host> and <link>, attributes availability and state are gone.
915 It was redundant with state and availability traces, and with peak values.
916 - In <cluster>, attributes availability_file and state_file are gone.
917 It was too complex and unused.
918 - Kill <gpu>. Was not doing anything.
919 - The DOCTYPE points to the right URL:
920 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
921 (the file at this address now documents the changelog since its v1)
922 - A warning is emitted for unit-less values (they are still accepted).
923 - speed. Default: 'f' or 'flops'. Also defined:
924 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
925 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
926 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
927 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
928 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
929 - latency. Default: 's' second. Also defined:
930 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
932 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
933 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
936 * s4u::Host is now the preferred public interface to the Host features.
937 sg_host_* functions are C bindings to the exact same behavior
938 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
941 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
942 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
943 They were deprecated since a few years, and probably did not even compile anymore.
946 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
947 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
948 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
949 encouraged to update their codes. A list of the modified functions can be found at the end of
950 include/simgrid/simdag.h
953 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
954 the lambda or closure passed as a parameter will run in kernel mode.
955 All the callback functions should be rewritten to that interface at some point.
958 * Reorganizing and cleaning the internals all around the place.
961 * Remove old default barrier/bcast buggy algorithms (see #18407)
962 * Various bug fixes to handle more codes
963 * Remove the need for the --foreground option of smpirun (it is still
964 accepted for backward compatibility).
967 * Kill the setset data container: MC don't use it anymore.
968 * Kill the queue data container: it made more sense with GRAS.
969 * Kill the xbt_peer_t data type: it's useless without GRAS.
970 * Kill rm_cb feature of config sets: it was never useful.
971 * Kill graphxml parsing feature. It was not used.
972 * Kill the deprecated code protected by XBT_USE_DEPRECATED
974 - xbt_dynar_sort_strings(), when the content is char*
975 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
976 They throw exceptions on invalid input;
977 * C++ support for declaring CLI flags (simgrid::config::Flag);
978 * class for abstracting different signal backends (simgrid::xbt::signal).
979 with no external dependencies (we need very simple signals).
982 * refactoring and cleanup of the code;
983 * ongoing process to cleanly separate the model-checking algorithms
984 from the code model-checking support.
986 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
988 SimGrid (3.12) stable; urgency=low
990 The Facelift Release.
993 * Require g++ v4.7 at least to not speak prehistorical C++.
994 * Require Boost 1.48 (for signal2 component).
995 * Java must be version 7 at least when activated.
996 * Builds on Windows again (including Java bindings).
997 * Tracing is now always enabled (no way to turn it out)
998 * Remove GTNetS. It was not working anyway.
999 * Various cleanups in the cmake scripts.
1000 * Move headers around to sort them out on installed systems:
1001 - instr/instr.h -> simgrid/instr.h
1002 - instr/jedule/* -> simgrid/jedule
1003 - simdag/datatypes.h was removed
1004 - simdag/simdag.h -> simgrid/simdag.h
1005 - msg/datatypes.h was removed
1006 - msg/msg.h -> simgrid/msg.h
1009 * Interface improvement:
1010 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1011 - Sanitize the interface in MSG_task_ module:
1012 - Merge two functions that were close enough but misleading:
1013 set_compute_duration(t) -> set_flops_amount(t)
1014 get_remaining_computation(t) -> get_flops_amount(t)
1015 - set_data_size(t) -> set_bytes_amount(t)
1016 get_data_size(t) -> get_bytes_amount(t)
1017 - Massive cleanups in the functions related to the energy
1018 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1019 - New: MSG_host_get_pstate()
1021 - msg/energy/onoff: switching hosts on and off
1023 * Interface improvement:
1024 - Rename Host.isAvail() to Host.isOn()
1025 - Rename Process.currentProcess() to Process.getCurrentProcess()
1026 - Rename Task.setDataSize() to Task.setBytesAmount()
1027 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1029 - #18874: Actually allows the GC to reclaim tasks
1033 - SIMIX_process_throw: raises an exception in a remote process
1034 * Refactoring: Separate sub-modules
1035 - libsmx: the public interface, as libc in a real system
1036 - popping: the strange dance that converts a user request into a kernel handling
1037 - smx_context_*: the virtualization mechanisms that embed the user code
1038 - smx_*: the handling of each simcalls
1039 * Interface cleanup:
1040 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1041 * Rename smx_action_t into smx_synchro_t, making explicit that these
1042 things are used to synchronize processes with their environment.
1043 For example, a communication is a sort of synchronization involving
1044 the communicating processes (that may block until the exchange) and
1045 the platform. The same can be said from computations, etc.
1047 - #18888: segfault when a process ends before its kill_time
1051 - 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.
1052 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1053 - MPI_Type_set_name, MPI_Type_get_name
1054 - MPI_*_c2f and MPI_*_f2c functions
1055 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1056 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1057 - Activate a lot of new tests from the mpich 3 testsuite
1059 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1060 - InfiniBand network model added: Based on the works of Jerome Vienne
1061 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1062 - When smpi/display_timing is set, also display global simulation time and application times
1063 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1064 * Collective communications
1065 - 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
1066 - 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/).
1067 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1069 - Replay now uses algorithms from wanted collective selector
1070 - Replay can be used with SMP-aware algorithms
1071 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1072 - 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.
1074 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1075 - Correctly use loopback on fat-tree clusters
1076 - Asynchronous small messages shouldn't trigger deadlocks anymore
1077 * Energy/DVFS cleanup and improvement
1078 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1079 - new: smpi_get_host_pstate()
1082 - "Full" network optimization flag was broken since Surf++
1083 - Better handling of precision flags in maxmin
1084 - Fix bug causing sometimes "Impossible" errors
1085 - Properly pass cluster properties to included hosts
1086 * Improvement of the Energy plugin.
1087 - Always update the consumption before returning that value
1088 - New property: watt_off to denote the dissipation when the host is off
1089 - New functions getWattMinAt and getWattMaxAt to retrieve the
1090 dissipation of pstates that we are not currently at.
1091 * Java: class NetworkLink renamed to Link
1092 * New function: simcall_process_get_kill_time()
1093 * Massive rename s/workstation/host/
1094 - That's intrusive, but that's good for the project consistency. Sorry.
1095 - Change config option "workstation/model" into "host/model"
1098 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1099 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1100 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1101 - xbt_dirname and xbt_basename for non-POSIX systems
1103 * The model checker now runs as a separate process.
1104 * The model checker runs must now be launched with the new simgrid-mc program.
1105 * Record/Replay: the MC can display a textual representation of a path in the
1106 execution graph. It can then be replayed outside of the model checker.
1108 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1110 SimGrid (3.11) stable; urgency=low
1115 * Normalizing pointers addresses tool for better diff between logs
1118 * Add cloud examples using new VMs
1119 - examples/msg/cloud/two_tasks_vm.tesh
1120 - examples/msg/cloud/simple_vm.tesh
1121 - examples/java/cloud/cloud.tesh
1122 - examples/java/cloud/migration/migration.tesh
1123 * Add java surf examples:
1124 - examples/java/surfPlugin/surf_plugin.tesh
1125 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1126 - examples/java/surfCpuModel/surf_cpu_model.tesh
1127 * Add SMPI+MSG example:
1128 - examples/smpi/smpi_msg_masterslave/
1133 - msg task destroy cancel test
1134 - msg_host on/off test
1135 * Move all tests in testsuite to teshsuite (adding tesh files)
1136 * Restructure teshsuites
1137 - one folder for each kind of test
1138 * Restructure AddTests.cmake
1140 - structure the order of tests (with sections)
1143 * Add virtual machine
1144 - creation of a VM on a PM
1145 - migration of a VM from a PM to another PM
1147 - MSG_process_join(msg_process_t process, double timeout)
1148 - msg_bar_t MSG_barrier_init(unsigned int count)
1149 - int MSG_barrier_wait(msg_bar_t barrier)
1150 - void MSG_barrier_destroy(msg_bar_t barrier)
1151 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1152 * New option "msg/debug_multiple_use" to help debugging when a task is used
1155 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1157 - Increase code coverage in test suites
1161 * Protect context stacks against stack overflow. The number of protected memory
1162 pages allocated on the top of each stack (1 by default) can be configured
1163 with the new command line option --cfg=contexts/guard_size:<value>.
1164 * Simcalls are now generated by a python script that
1165 - generates files included by SimGrid
1166 - checks that all the functions exist, and proposes prototypes
1168 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1169 host_set_data, host_get_data
1171 - simcall_process_join(smx_process_t process, double timeout)
1172 * Fix bug where sleeping processing could not be suspended.
1175 * Translate surf models from C to C++
1176 - Generic classes for all models: Model, Resource, Action
1177 - A generic interface for each kind of model (CPU, Network, Storage
1178 Workstation, WorkstationVM)
1180 * Translate surf routings from C to C++
1181 * Add callbacks using sigc++ or boost::signals2
1182 - Add callback functions for resource creation/destruction
1183 - Add callback functions for action state change
1184 - Handle Energy as a plugin
1185 * Replace swag by boost::intrusive
1186 * Add new routing models for clusters. For documentation, see
1187 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1188 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1189 parameters for cluster tag
1190 - Fat trees, with topology="FAT_TREE" and
1191 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1192 - see examples/platforms/torus_cluster.xml and
1193 examples/platforms/fat_tree_cluster.xml
1194 * More documentation
1197 * Hostfiles support host:nb_processes construct to deploy several processes on
1199 * Collective communication algorithms should not crash if used with
1200 improper number of nodes and report the error.
1201 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1202 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1203 MPI_Dims_create, MPI_Cart_sub are supported.
1204 * New interface to use SMPI programmatically (still depends on MSG for
1205 some parts, see examples/smpi/smpi_msg_masterslave):
1206 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1210 * Global variables privatization in MPI executables is now performed at runtime
1211 with the option smpi/privatize_global_variables (default:no).
1212 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1213 dynamic libraries loaded by the application are not privatized (static
1214 linking with these libraries is advised in this case)
1217 * Options defined in XML work correctly now.
1220 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1221 copied into simgrid.jar or not (ON by default). Use this option if you want
1222 to reduce the size of the installed simgrid.jar, *and* the native libraries
1223 are kept installed elsewhere.
1224 * Surf binding with SWIG (code generated in maintainer mode only):
1225 - plugin to handle callbacks
1226 - CPU model only for the moment
1229 * Supernovae build mode is definitively removed. It was used to improve
1230 inlining and inter-module optimizations. It is nowadays superseded by
1231 link-time optimizations commonly available in compilers.
1232 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1233 * Add boost dependency for surf++
1234 * Add new macro for tests
1235 - ADD_TESH(name <tesh_args>)
1236 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1240 - xbt_bar_t XBT_barrier_init(unsigned int count)
1241 - int XBT_barrier_wait(xbt_bar_t barrier)
1242 - void XBT_barrier_destroy(xbt_bar_t barrier)
1243 * Make the xbt_os_time module public
1245 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1247 SimGrid (3.10) stable; urgency=low
1249 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1252 * Reintegrate Java to the main archive as desynchronizing these
1253 package is not acceptable anymore (Java is now considered stable)
1254 * Add explicit synchronization facilities through semaphores
1255 * Bug fix: Task.setDataSize() only changed the C world, not the value
1256 cached in the Java world
1259 * Dramatically change the way files are handled. API and internals changed, but
1260 this part of MSG was not considered as production grade either.
1261 * Add explicit synchronization facilities through semaphores
1262 * Add a new function MSG_host_get_process_list()
1263 * Preliminary DVFS support (see examples/msg/energy/ for details)
1266 * SMPI is now included directly in the libsimgrid as the windows
1267 linker doesn't force us on splitting it anymore.
1268 * Improvements of the SMPI replay tool:
1269 - Most of the collective communications are now rooted in the same process as
1270 in the original application.
1271 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1272 used until now). Multiple data types can now be used in a trace.
1273 - The replay tool now supports traces produce either by TAU or a modified
1275 - Bug Fix: the compute part of the reduce action is now taken into account.
1276 - Gatherv collective is now supported
1277 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1278 of smpirun outputs time independent traces for the current run. One file
1279 is created per process. If too many processes are simulated, this behavior
1280 can be changed to one file for all processes by using the
1281 tracing/smpi/format/ti_one_file flag
1282 * smpirun generates the host file if needed (with given host count and platform)
1283 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1284 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1285 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1286 --cfg=smpi/coll_selector:(mpich/ompi)
1287 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1289 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1290 happening inside a collective SMPI call.
1291 * Fix the behavior of complex data types handling.
1292 * Make MPI_Wtime another synchronization point to take computations into
1294 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1295 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1296 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1297 (no privatization of global variables yet)
1298 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1301 * Verification of liveness properties is now available for SMPI applications
1302 (in addition to MSG applications)
1303 * Bugged examples using SMPI in examples/smpi/mc/
1304 * Add --cfg=model-check/visited option. Allows the verification of infinite
1305 programs. Detection of loops in the execution thanks to the system state
1306 comparison and reduction of the state space to explore. Can be combined with
1307 DPOR for safety properties.
1310 * Allow to change SimGrid configuration (see --help) within the code
1311 thanks to SD_config() as it can be done in MSG.
1312 * Add a new function SD_task_set_amount() upon user request.
1315 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1316 * Remove rule based routing (no more PCRE dependency)
1317 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1318 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1319 * Add a loopback_bw and loopback_lat options to cluster tag.
1320 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1321 routing. Now peers are encapsulated in an AS and have their own private
1322 router but this is transparent.
1325 * Our own implementation of getline is renamed xbt_getline, and gets
1326 used even if the OS provide a getline(). This should reduce the
1327 configuration complexity by using the same code on all platforms.
1328 * New type: xbt_cfg_elm_boolean.
1329 * Allow to use yes/no for boolean configuration options in the command line.
1330 * Allow to disable SimGrid cleanups at exit from command line option.
1331 There are situations where one may want a simulation to end with an exit.
1332 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1333 annoying when scripting around the simulator. Adding a
1334 --cfg=clean_atexit:no allows to circumvent this issue.
1337 * Lots of memory leaks were corrected in this release.
1338 There is no known memory leaks anymore, in all of our 600+ tests.
1339 * New command line option --version, to get SimGrid version information.
1340 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1342 * Supernovae builds are deprecated, and expected to be removed in the next
1345 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1347 SimGrid (3.9) stable; urgency=low
1349 The Grasgory release: GRAS is really dead now.
1351 * Complete overhaul of the internal host structures scheme.
1354 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1355 was considered as experimental and badly maintained since then.
1356 * Keeping it was thus a trap to our potential users, that could take
1357 it instead of MSG or SMPI by mistake despite is pity state.
1358 * GRAS seems to have very few users (if any), and no one volunteered
1359 to maintain it further. It also induces a lot of XBT code (for
1360 portability sake), that must be maintained too.
1361 * For all these reasons, we killed GRAS. If someone wants to revive it
1362 in the future, don't cry, our git history still remembers of GRAS.
1365 * Major overhaul. Merge our documentation again as time proved that
1366 splitting it was really not helping our users.
1367 * Further improve the developer documentation to help newcomers
1368 hacking on SimGrid itself. The user documentation (and in
1369 particular, the beginner documentation) is still in a sorry state.
1372 * Now works on Windows too!
1373 * Much more extensive test suite, from MPICH
1376 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1377 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1378 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1379 * Bug fix: let task be scheduled when the last dependency to be solved is
1380 a control dependency.
1381 * Remove SD_load_environment_script function.
1382 Use the C sg_platf function if you want to declare a platform
1386 * New function: MSG_process_get_number()
1387 * Old function documented: MSG_config()
1388 * Remove MSG_load_platform_script function
1389 Use the C sg_platf function if you want to declare a platform
1393 * Change the default value of the TCP_gamma constant (maximal size of TCP
1394 congestion window) to a more realistic 4MiB value. If you notice changes in
1395 your simulation results, you can fall back to the previous 20k tiny window
1396 by adding --cfg=network/TCP_gamma:20000 on command line.
1397 * (Hopefully) fix a bug wrt periodic availability and state traces
1398 * Bug fix: use default values at start when first event in availability/state
1399 trace is not at time 0.
1402 * remove the "new_" part of function name sg_platf_new_trace_connect
1403 (resulting in sg_platf_trace_connect), since it does not create
1407 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1408 use case where it's really mandatory, and maintaining it was a pain in
1410 * New: xbt_fifo_search(), search an item with a user-provided
1411 comparison function instead of dumb pointer comparison.
1414 * Fix the lua deployment:
1415 Use `simgrid.init_application()` before deployment instead of
1416 `simgrid.msg_register_application()` after.
1419 * Transfer the tracing files into the corresponding modules.
1421 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1423 SimGrid (3.8.1) stable; urgency=low
1425 The "we are told that some people want to also *install* the simgrid
1428 * Add missing file "tesh.1" to the archive.
1430 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1432 SimGrid (3.8) stable; urgency=low
1434 The Psssshiiiit release: SimGrid jumps into the Cloud.
1437 * Add an experimental interface to manipulate VMs. They are mainly
1438 process groups with very few intrinsic semantic, but they should
1439 allow you to build the semantic you want easily.
1440 * New function: MSG_host_set_property_value()
1441 * New function: MSG_process_on_exit(). To clean memory in all cases.
1442 * Bug fixes that made the host (and link) failures unusable.
1443 * Add a way to auto-restart process when the host in which they are
1444 executing comes back (ON_FAILURE="RESTART" on deployment file,
1445 MSG_process_auto_restart_set).
1446 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1447 please stop using the old ones, they are DEPRECATED.
1449 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1450 Please use MSG_init() instead. (reducing the amount of entry
1451 points in the library helps us).
1452 * Make it impossible to link against the wrong version of the lib
1453 * Deprecate MSG_clean(). No need to call it anymore.
1454 * Function MSG_get_host_number() is not deprecated anymore.
1457 * Split the doc into a user guide and a reference guide.
1458 * Start a developper guide to help people hacking on SimGrid.
1461 * Enable tracing by default. This modules rocks you should use it.
1462 * Remove option custom_flags. Now use environment variables CFLAGS
1464 * Use default cmake things to detect lua instead of home grown ones.
1465 * New option "enable_mallocators" to disable mallocators, for debugging
1466 purpose ("on" by default).
1469 * Bug fixes around the resource failures: don't let the processes
1470 survive the host they are running onto.
1471 * Add an interface to auto-restart processes when the host in which they are
1472 executing comes back.
1473 * Ensures that SIMIX_clean is called automatically. It's not part of
1474 the public interface anymore (bindings should be updated).
1477 * Bug fix for when SD_Simulate is called with a positive value: be careful
1478 when comparing doubles. Sometimes they are different for non significant
1480 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1481 parallel task whose initial work is distributed among host according
1482 to the Amdahl's law. Such tasks are created with a parameter alpha
1483 that corresponds to the non-parallelizable part of the computation.
1484 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1485 between two sets of workstations assuming a 1D block distribution (each
1486 workstation owns a similar share of data) on both sides.
1488 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1489 Data redistribution will be automatically scheduled once parent and child
1490 are both scheduled. The filling of computation_amount and
1491 communication_amount structures is now done seamlessly thanks to the chosen
1493 * New function SD_workstation_dump to display various information
1494 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1495 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1496 bandwidth on the route onto which the task is scheduled and the amount of
1498 To divide the nominal bandwidth by 2, the rate then has to be:
1499 rate = bandwidth/(2*amount)
1500 * Compute tasks that have failed can now be rescheduled and executed again
1501 (from their beginning)
1502 * Increasing source code coverage (src/simdag is now covered at 95.8%
1506 * Re-implement time-independent trace replay using SMPI (at the
1507 smpi_smp_* level) instead of MSG. This should replace
1508 examples/msg/actions/actions.c
1509 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1510 hindexed and structs)
1511 * Implement the exchange of non-contiguous data.
1512 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1513 * Correct behavior of smpi/sender_gap and set its default value to 0
1514 * Add option to asynchronously send small messages to allow better
1515 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1516 specifies the size in bytes under which messages will be asynchronously sent.
1517 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1518 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1521 * New C interface to define a platform: XML is now optional.
1522 For more info, please check include/simgrid/platf.h.
1523 * New interface to define random platforms from the C:
1524 For more info, please check include/simgrid/platf_generator.h and
1525 examples/msg/masterslave/masterslave_platfgen.c
1526 * Export a sg_cmdline dynar containing all the arguments we got from
1530 * Two new tracing options for adding comments to trace file so you
1531 can track your experiments (see --help-tracing for details).
1532 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1533 * Adding the SimGrid version that generated the trace file as a comment.
1534 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1535 * Fix to avoid key clashes on Paje links
1536 * Other minor fixes related to the Paje specification
1539 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1540 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1541 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1542 * Greatly improve the robustness of mmalloc to user errors (such as
1543 using an area after freeing it, or freeing it twice)
1545 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1547 SimGrid-java (3.8.1) stable; urgency=low
1549 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1550 Warning: all methods are visible, but only some of them are
1551 implemented so far. Check the source (src/jtrace.c)
1552 for further information.
1553 * New module: org.simgrid.msg.File (SimGrid File management functions)
1554 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1555 * Change the meaning of Process.restart: now restart the process from
1556 the begining, like MSG_process_restart in C.
1557 * Add Process.setAutoRestart: handling of process restart when failed
1559 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1560 you to retrieve the properties of the processes/hosts
1561 * Deprecate Msg.clean(): you can just forget about it now.
1562 * New function Process.getCount(), that only works when compiling
1563 with the not yet released version 3.9 of the C library.
1565 * New context factory based on Coroutines. It mandates a modified JVM
1566 but then, the simulations run about five times faster, and there is
1567 no limit to the amount of processes (beside of the available memory).
1569 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1571 SimGrid (3.7.1) stable; urgency=low
1574 * Restore the prototype of MSG_process_create_with_environment() to
1575 the pre-3.7 situation by removing the kill_time argument.
1576 * Add a MSG_process_set_kill_time() function instead.
1579 * Fix weird behaviors when dealing with parallel tasks.
1582 * Simgrid is now built as a dll.
1583 * Simgrid-java now works on Windows.
1584 * Simgrid-Java is now included into Windows package.
1587 * First pre-build package for MacOSX.
1590 * Fix compilation when using MSG_USE_DEPRECATED.
1591 * Fix some compilation issues on Macs and Windows.
1592 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1593 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1595 We discovered that the Lua console is broken, but we are missing the
1596 manpower to fix it right now. The problem existed in 3.7 too, so we
1597 are not blocking the release for that. Sorry if you depended on this
1598 feature, any help would be really welcome.
1600 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1603 SimGrid-java (3.7.1) stable; urgency=low
1605 The "Java aint got to be bloated and slow" release
1608 * Various internal cleanups and performance improvement
1609 Simulations are expected to run up to twice faster or so
1610 * Make Process.kill(process) an instance method, not a static one
1611 * User processes are not java.lang.Thread subclasses.
1612 This breaks the compatibility (sorry), but previous API was
1613 brain-dead, making it impossible to have non-trivial
1614 initializations in the process constructor.
1615 * Require a full constructor per Process sub-class.
1616 Kinda breaks the compatibility (sorry), but this allows a much more
1617 efficient way to launch the processes at simulation startup.
1618 * Do not embeed our version of semaphores, java 1.5 can be considered
1619 as sufficiently prevalent for us to not dupplicate its features.
1623 * Add examples for almost every part of the API
1624 We spotted and fixed a lot of bugs in the process
1626 * New module: asynchronous communication API
1627 * New function: Process.sleep()
1628 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1629 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1630 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1632 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1634 SimGrid (3.7) stable; urgency=low
1636 The "spring cleanups (before the next Big Project kicks in)" release.
1639 * We can specify the SMPI latency/bandwidth factor with command line
1640 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1641 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1642 You can also use the "config tag" from platform file by adding this line
1643 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1644 (see "example/platforms/tag_config.xml" to use "config tag").
1645 Note that the command line supersedes the platform file configuration.
1646 * Change the correction factors used in LMM model, according to
1647 the latest experiments described in INRIA RR-7821.
1648 Accuracy should be improved this way.
1649 * Use the partial invalidation optimization by default for the
1650 network too. Should produce the exact same results, only faster.
1651 * Major cleanup in surf to merge models and split some optimization
1652 mechanisms from the core of the models. As a result you can now
1653 specify which model to use (e.g., --cfg=network/model:LV08
1654 --cfg=cpu/model:Cas01) and which optimization mode to use
1655 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1656 Incompatible combinations should err at initialization. See
1657 --help-models for the list of all models and optimization modes.
1658 * The CLM03 workstation model was dropped for simplicity because it
1659 used the deprecated CM02 network model. Use default instead.
1660 * Rename the TCP_gamma configuration option to network/TCP_gamma
1661 * Rename the coordinates configuration option to
1662 network/coordinates, and document it
1663 * Use now crosstraffic keyword instead of the terribly misleading
1664 fullduplex keyword. It is activated by default now in the current
1665 default model, use --cfg=network/crosstraffic:0 to turn it off.
1666 * Ongoing refactoring the model parsing to make XML files optional
1667 See include/simgrid/platf.h for details (still to be completed)
1670 * Major overhaul of the documentation. Almost instructive now :/
1671 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1672 functions and friends. This interface was considered as
1673 deprecated since over 2 years, it's time to inform our users that it is.
1674 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1675 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1676 use these (crappy) functions in your code.
1677 These functions will be removed soon. Stop using them now.
1678 * Deprecate MSG_get_host_{table,number}
1679 Implement MSG_hosts_as_dynar() instead.
1680 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1681 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1684 * Stabilize the parallel execution mode of user contexts
1685 * Introduce configuration variables to control parallel execution:
1686 - contexts/synchro: Synchronization mode to use when running
1687 contexts in parallel (either futex, posix or busy_wait)
1688 - contexts/parallel_threshold: Minimal number of user contexts
1689 that must be part of a scheduling round to switch to parallel
1690 execution mode (raw contexts only)
1691 * Fix bugs that prevented to use suspend/resume along with
1692 synchronization structures.
1693 * Fix bugs in process termination that lead to invalid memory access
1694 in very specific conditions.
1697 * Introduce a parallel mode for the models (controlled by surf/nthreads
1698 configuration item). In our tests, running the models in parallel
1699 never lead to any speedups because they are so fast that the gain
1700 of computing each model in parallel does not amortizes the
1701 synchronization costs, even when ultra fast futexes are used.
1702 This is released anyway because YMMV.
1705 * Performance boost by using a swag internally to compute the set of
1706 tasks that are finished and should constitute the return value of
1710 * Enable it by default now that it is considered rather stable.
1713 * Documentation of the tracing functions.
1714 * Performance gains when tracing categorized/uncategorized resource
1715 utilization by avoiding calls to get route when updating resource
1716 variables. LMM constraints are being used instead.
1717 * API changed to set task categories. Use MSG_task_set_category instead
1718 of TRACE_msg_set_task_category, and SD_task_set_category instead
1719 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1720 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1721 * Examples for link user variables added (see at examples/msg/tracing/)
1722 * Deprecated function TRACE_msg_set_process_category completely removed
1723 * Trace header updated according to the latest Paje file format
1724 * Tracing network lazy updates, no longer obligate users to use full updates
1725 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1726 * Experimental: let user code declare/set/push/pop application states for hosts
1727 * API changed to allow the manual creation of graph configuration files
1728 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1731 * Improve the API of Lua MSG bindings, using the Lua spirit.
1732 * Each simulated process now lives in its own Lua world (globals are
1733 automatically duplicated). It helps writing simulators. Will allow
1734 to run Splay programs within SimGrid in the future.
1735 * Add a Chord example in Lua, equivalent to the MSG one.
1738 * Start the implementation of a solution to express temporal
1739 properties, not only local assertions. This is still an
1740 experimental work in progress, stay clear from it to be safe.
1744 - Add new runtime parameters --help-logs and --help-log-categories
1745 to display information about supported logging parameters and
1747 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1748 don't exists anymore.
1749 * Mallocators: allow value NULL for the reset function.
1751 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1752 create homogeneous dictionaries, where all the elements share the
1753 same free function. Non homogeneous dictionaries will be
1754 deprecated in the next release.
1755 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1756 - Multi-level dictionaries are deprecated.
1758 - new function xbt_dynar_search_or_negative() that is useful when
1759 you have less than 2 million elements in your dynar and don't
1760 want of the extra complexity of catching exceptions when the
1761 element is not found.
1763 - Make xbt_os_thread module (for thread portability) public.
1764 Documentation is still to come, sorry.
1766 - Cleanups and simplifications to make it maintainable again.
1767 - Exotic features (such as memalign and valloc) were removed.
1768 - The metadata were extended and improved so that the
1769 model-checker becomes able to explore and inspect the heaps.
1770 - This may induce a performance drop when enable_model-checking is
1771 ON in cmake (even if it's not used in the simulation), but it is
1772 necessary at this point to get MC working.
1774 Turn model-checking OFF if simulation performance matters to you.
1775 Not enabling it at runtime is not enough, disable it in cmake.
1777 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1780 SimGrid (3.6.2) stable; urgency=low
1782 The "Not coding new stuff allows to polish old things" release.
1785 * New bindings to the NS3 packet level simulator (experimental)
1786 * Use the raw (efficient) execution contexts instead of the sysv
1787 (portable) ones when possible.
1788 * libpcre is now mandatory in any cases since not using it led to
1789 severe performance loss and possibly other issues
1790 * Update the XML platforms:
1791 - G5K: include the latest machine in Nancy
1792 - GridPP and LCG: new platforms
1793 * Documentation was partially updated, at least (more to come)
1795 Bug fixes, cosmetics and small improvements
1796 * Free terminated processes before the end of the simulation to avoid
1797 exhausting the memory of users having very dynamic amount of
1799 * Bug fix and cosmetics about canceling non-running tasks
1800 * Bug fix about the dot loader's issues when using libcgraph
1803 * Create an installer for windows with nsis (amd64 and win32)
1804 - Add an hello world project to illustrate simgrid project creation.
1805 - Embed libpcre into the Simgrid installer to avoid
1806 its compilation burden
1807 * The raw execution contexts should work on Apple now
1808 * Port to Windows 64 bits
1809 - Sysv contexts now have an implementation for this arch
1810 - GRAS communication features now support this arch
1811 * Drop support for borland compiler on windows
1812 - this code was not maintained, and we kinda depend on gcc nowadays
1813 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1814 * Fix portability issue on unstable ubuntu: linker became picky on
1817 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1819 SimGrid (3.6.1) stable; urgency=low
1821 The "Oops, we broke Macs too" release
1824 * Fixed contexts detection so that raw ones are used when possible
1825 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1826 leads to a strange error, with user code segfaulting sometimes when
1827 the generated code is not perfectly aligned (which is not
1828 controllable from the user side, depends on the amount of code)
1831 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1833 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1835 SimGrid-java (3.6) unstable; urgency=low
1838 * Split of every thing from simgrid v3.5 into a separate package.
1840 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1842 SimGrid (3.6) stable; urgency=medium
1844 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1847 * Bindings now constitute their own package, separated from the main one.
1848 Rationale: reduce our maintenance nightmare by reducing the module coupling
1849 They will soon be released on their own on gforge.
1851 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1852 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1854 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1855 * It's not quite deprecated for now because we have no replacement,
1856 but it may soon become the case.
1859 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1860 * Fortran: New user-level cache variable to store the rank of the running
1861 process. This improves performance by an order of magnitude.
1862 * C: New coccinelle script to automatically locate and modify global and
1863 local static variables.
1864 * Improved SMPI network model with a sender-side gap to account for multiple
1868 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1869 only say if a communication is finished, no matter whether it succeeded or
1870 failed. You can call MSG_comm_get_status() to know the status of a finished
1872 * New function MSG_task_dsend() to send a task and detach it. When a
1873 communication is detached, you are never notified of its success or failure
1874 and the memory is released automatically once it is finished. This function
1875 is useful when you don't care about the end nor the success of a
1877 * Change the prototypes of action replay. Sorry for inconvenience,
1878 but this is really more efficient this way (and to adapt your code,
1879 you just have to fix the initialization, that shouldn't be too long)
1880 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1881 ever managed to do anything useful with it.
1882 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1883 takes a dynar of communications. It returns immediately and gives the
1884 index of a finished communication (if any).
1885 * New example: a basic implementation of the Chord P2P algorithm.
1888 * New model for multi-core CPUs. You can now use the core attribute to
1889 precise the number of cores of a host. This is a basic model. Every
1890 process running on the host receives at most the power provided in
1891 the DTD (throughput<=power). Total throughput of process cannot exceed
1893 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1894 router linked by an up-link and a down-link (possibly asymmetrical).
1895 This kind of pattern allows to easily build last-mile model style platforms.
1896 Aggregating such patterns in a rule-based AS is thus the technique of
1897 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1898 * New model for Vivaldi routing. We transformed the Vivaldi network model
1899 into a Vivaldi routing model (based on the rule-based model). This allows to
1900 combine Vivaldi based latencies with last-mile platforms.
1903 * Added a check for NaN of IEEE754 infinite in the double entries of
1905 * Introduce a new context factory "raw", highly inspirated from the
1906 ucontext factory, but using manually crafted functions in assembly to
1907 do the work in an efficient manner.
1908 * Allow to change the used context factory at run time, not only at
1909 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1910 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1911 contexts at runtime (only with raw contexts or ucontexts).
1912 * Completely rewrote this module to allow parallel execution of user
1913 processes. Use --cfg=contexts/nthreads:N to execute user processes
1914 with N parallel threads (the default is 1, meaning no parallelism).
1915 * Allow to decide dynamically between sequential and parallel modes.
1916 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1917 processes in parallel only when their number is greater than or equal to P
1919 * Added a check for NaN of IEEE754 infinite in the double entries of
1923 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1924 won't output the state of processes when interrupted with Ctrl-C
1925 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1926 NULL-terminated array. This may solve backward compatibility issues
1927 due to the change to return type of SD_simulate. See also:
1928 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1929 * Add new macros with variable number of arguments.
1930 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1931 - in xbt/asserts.h: xbt_assert
1932 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1933 - in xbt/ex.h: THROWF and RETHROWF.
1934 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1936 * Change xbt_die() to accept a format string with arguments, just like printf.
1937 * New data structure: xbt_lib_t, like a dict but more general and with better
1941 * New configuration options
1942 Options triva/categorized and triva/uncategorized can be used to generate
1943 graph configuration files for Triva visualization tool.
1944 * Configuration option tracing/platform is renamed to tracing/categorized
1945 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1946 * New TRACE_user_link_variable interface:
1947 User provides the name of the link and the tracing variable to attach to it
1948 * the declaration of tracing categories must be done after the environment
1950 * simpler tracing interface, just one way to declare categories
1951 TRACE_category or TRACE_category_with_color, it is up to you
1952 * links in the trace file are again identified by their names
1953 * trace contains the full platform hierarchy exactly as declared using the ASes
1954 * Options tracing/msg/[task|process]:1 groups the process by hosts
1955 for both cases, tasks and processes must have names that are unique during the simulation
1956 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1957 * The experimental option tracing/msg/volume is deprecated
1958 its functionality may be reincorporated if needed
1960 The tracing generates a trace file with unordered timestamped events,
1961 because of the way the core simulator (surf) works. A script available
1962 at the tools directory (fix-paje-trace.sh) can be used to put the events
1963 in order. We have changed the tracing so it can generate ordered timestamped
1964 events in the final trace, but depending on the simulator (and how much time
1965 is simulated) that can lead to a huge memory utilization. It is deactivated
1966 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1968 Build Infrastructure
1969 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1970 - We are trying hard to keep the API stable, but it may happen that
1971 some things change (we're a research project after all, not a
1972 nuclear plant operating system). If such things should happen, you
1973 could rely on that macro to adapt.
1974 - current value: 30600 for 3.06.00, aka 3.6
1975 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1976 a number that can be compared with SIMGRID_VERSION.
1977 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1978 assertions and verbose/debug logging events are disabled at compile time.
1980 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1982 SimGrid (3.5) stable; urgency=medium
1985 * New feature to any SimGrid-based simulator: Model-Checking
1986 Check SIN#1 for more details.
1989 * New Model SMPI (three-interval linear regression for correction factors)
1990 See RR-7426, available at http://hal.inria.fr/inria-00527150
1991 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1992 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1993 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1994 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1995 * See also src/smpi/README
2000 - Tracing API changes: TRACE_start and TRACE_end should not be called
2001 by user-code. They are automatically called by simulators created
2002 with SimDAG, MSG and SMPI if the toolkit is compiled with
2003 tracing_enabled=ON. Categories declaration and utilization remain the
2004 same for MSG and SimDag.
2005 - A function was added to the tracing API to declare categories with
2007 - TRACE_category_with_color (char *category, char *color)
2008 where color must be in the following format
2009 "%f %f %f", red, green, blue
2010 and red, green, blue are float values in the interval [0, 1]
2011 - User can specify NULL as color parameter, or continue calling
2012 TRACE_category (cat)
2013 On that case, the tracing system will define random colors
2014 - The following command-line options are supported:
2015 --cfg=tracing/filename:msg.trace
2016 --cfg=tracing:1 (activate tracing, needed to use others)
2017 --cfg=tracing/platform:1 (categorized resource use)
2018 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2019 --cfg=tracing/msg/task:1 (task creation)
2020 --cfg=tracing/msg/process:1 (process creation, migration)
2021 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2022 --cfg=tracing/smpi:1 (SMPI interface tracing)
2023 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2024 - examples of examples/msg/tracing updated
2026 - DAXLoader and DOTLoader functions can generate tasks with categories
2027 - A new function to attribute a category to SD tasks:
2028 TRACE_sd_set_task_category (SD_task_t task, char *category)
2029 * Tracing the MPI interface implemented by SMPI
2030 - Collective operations are traced with states
2031 - Point-to-Point operations are traced with states/links
2032 - Tracing activated by a parameter "-trace filename" passed
2033 to smpirun during execution (considering that simgrid
2034 is compiled with tracing enabled)
2035 - To run the simulation with gdb, the simulator
2036 accepts --cfg=tracing/smpi:1 to trace SMPI
2037 - tesh files to check if smpi tracing is ok
2038 - See examples/smpi/NAS/DT-trace
2039 * GTNetS tracing re-worked
2040 - adaptation to the tracing system of GTNets to cope
2041 with modifications regarding the fullduplex mode
2042 - new tesh files to check if gtnets tracing is ok
2045 * Asynchronous communications through the functions:
2046 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2047 * New function: MSG_load_platform_script()
2048 to make possible using a lua script instead of XML files to set up platforms
2049 * New function: MSG_set_function
2050 to associate functions to processes, used when bypassing the parser
2051 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2053 Platforms: Add some more examples in examples/platforms
2054 * Grid'5000: see www.grid5000.fr
2055 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2058 * Change the XML format. This is a very important modification. SimGrid
2059 3.5 introduces a new hierarchical format based on the notion of
2060 Autonomous Systems. Compatibility with old format is ensured
2061 through the perl script provided in the install bin directory
2062 bin/simgrid_update_xml.
2063 It is now possible to build platforms with specific routing
2064 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2065 connect several platforms together. We will try to provide soon
2066 set of realistic platforms exploiting these properties (have a
2067 look at examples/platforms/ for the moment).
2068 * Take the opportunity of the XML format change to be a good XML citizen:
2069 rename link:ctn to link_ctn and similar changes (also dealt with by
2071 * Add a new routing scheme (rule-based) using regular expressions. It
2072 enables to have an extremely low memory footprint when the
2073 underlying routing is simple and can be compactly described. You
2074 need to have libpcre4-dev (perl regular expressions) installed if
2075 you want to use this routing scheme.
2076 * Revive the cluster TAG and allow to easily and efficiently (both in
2077 term of memory and speed) connect clusters together. Have a look
2078 at teshsuite/simdag/platforms/ to see how this can be done. With
2079 this tag, you can create clusters with thousands of tasks at no
2080 cost (have a look at examples/platforms/).
2081 Note: clusters are implemented as ASes, so there is no need for an
2082 enclosing AS tag if you have only one cluster in your platform.
2083 * Add new generic functions in the public interface that allows the user
2084 to call SURF 'create_resource' methods from your code (same
2085 functionality as the XML bypass mechanism but with a much lighter
2087 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2088 that uses a piecewise linear approximation to produce better
2089 results when exchanging small messages.
2090 * Add a new parameter to handle correctly full duplex link and account
2091 for interferences between uplink and downlink communications
2092 (activate with --cfg=fullduplex:1).
2095 * Rename the SD_READY (all dependencies are satisfied and task is
2096 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2097 dependencies are satisfied) state.
2098 This prevents a confusion between the notion of "ready to schedule"
2099 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2100 simulated" (SD_RUNNABLE) used by the simulation kernel.
2101 * Change the way a task is considered as ready. Instead of removing
2102 dependencies when a task is done, a counter is decreased. This way,
2103 it is always possible to reach ancestors thanks to the
2104 SD_taks_get_parents function (even after the end of the simulation.)
2105 * Change the return type of SD_Simulate from (SD_task_t*) into
2106 xbt_dynar_t. This function was in handling a dynar internally and
2107 converted it into a NULL terminated array for historical reasons.
2108 * New function SD_dotload(char*) to load a DAG described in dot
2109 format. This loader and the corresponding examples require the
2110 installation of the graphviz library.
2111 * Fix a bug in the management of tasks of size 0 in the surf network
2112 models. This problem was only visible with SIMDAG and you should
2113 thus disregard results produced with earlier versions if you
2114 relied on this feature (some tasks were blocked because of this).
2115 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2116 even though your applications were doing only point-to-point
2117 communications and sequential computations. Now you can really use any
2118 model you want (of course, if you create real parallel tasks, which are
2119 not implemented in most models beside ptaskL07, this will abort).
2120 * Add an example that schedules a DAX on an heterogeneous platform
2121 using a Min-Min strategy.
2122 * New function SD_workstation_get_current_task() that returns the kind
2123 of task currently running on a workstation in the sequential access
2125 * Raise some warnings when unexecuted tasks remains at the end of the
2126 simulation. This is usually caused by cycles in the DAG.
2129 * New function: SIMIX_process_set_function() called by MSG_set_function
2130 * Change the underlying waiting queue in semaphores so that a process
2131 can wait on several of them simultaneously (as in waitany).
2132 * Fix the way to handle tokens in semaphores so that all access patterns
2133 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2134 * kill the dirty pimple SIMIX_message_sizes_output()
2135 Please use (proper) visualization instead
2138 * New data container: setset (set of sets of elements)
2139 * New module: mmalloc (mapped malloc, allowing to have several
2140 independent segments of malloc)
2141 * New function: xbt_dict_cursor_set_data()
2142 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2143 * New function: xbt_dynar_is_empty()
2144 * New function: xbt_fifo_get_last_item()
2145 * Fix xbt_dynar_shrink(): use the right element size.
2146 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2147 uninitialized areas during expand.
2148 * Fix semaphores: previous implementation was severely broken.
2149 * Use library init/fini functions for our initialization.
2150 - you can use logs and other feature as soon as you want in your
2151 code (even before the xbt_init / MSG_init)
2152 - xbt_exit is now a no-op and produce a warning when used.
2155 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2156 benefit from the latest improvement to the simulation kernel.
2157 * Kill measurement sockets for now. If you rely on them, sorry. This
2158 release is not for you. This feature will be reintroduced in the
2159 future, but we cannot delay the release any further.
2160 * New function: gras_msgtype_get_name().
2161 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2162 * Fix (at last) the pmm example: it should not randomly fail anymore.
2164 Build chain: bug fixes and overall polishing
2165 * Cmake is now stable enough. Hence, we killed the autotools.
2166 * Port to windows ( TM :)
2167 * Fix the 'make install' target.
2168 No need to use 'make install-simgrid' anymore
2169 * Introduce a 'make dist' target compiling a *source* archive
2170 'make package' compiles a binary archive
2171 * Compile java files only on need
2172 * Add --cd and --setenv command line options to tesh
2173 * Out of source builds are not fully supported yet, but we are close
2174 * Enable supernovae and optimization flags by default for our users
2177 * Add layer to set up environment directly from lua, without XML.
2178 * The effect of gras_stub_generator can be achieved through
2179 lua too (check examples/gras/console/ping_generator.lua)
2181 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2183 SimGrid (3.4.1) stable; urgency=low
2185 The "Polishing easter eggs is probably a good idea" release.
2186 This is a bug fixes release only.
2190 * Fix a bug preventing the tasks from begin garbage collected.
2193 * Fix a bug occuring when a host involved in a communication fails.
2194 This was not detected properly by the other peer involved in the
2195 communication. Now, it's reported as a network error.
2198 * Warn the user about loop dependencies in data flow of DAX files
2199 * Obey the control-flow dependencies of DAX files
2202 * Add option "enable_smpi" allowing to not compile SMPI.
2203 Probably useful for the (Mac) users experiencing a build error here
2204 * Improve the detection of lua5.1 and ruby1.8
2206 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2208 SimGrid (3.4) stable; urgency=low
2210 The "Easter in Cargese" release. Also known as (major changes):
2212 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2213 ~> bindings were greatly improved
2214 ~> new tracing infrastructure for better visualization introduced
2216 * the "Welcome to configury modernity" release.
2217 ~> we switched from autotools to cmake, and improved our cdash
2220 A more detailled list of changes follow (full detail in svn log).
2222 Java Bindings: Various Cleanups
2223 * (install java-gcj-compat-dev on debian-like to use them)
2224 * Remove put/get: no need to export deprecated interface in Java
2225 Use send/receive instead.
2226 * Cleanup the examples and add a README per directory
2227 * Remove example autoDestination (that's the only way to go now)
2228 * Remove example explicitDestination (was a plain copy of basic)
2229 * Make JniException a runtime exception, so that there is no need to
2230 declare the fact that you may encounter such a beast. I guess that
2231 nobody will ever want to survive such error.
2232 * Create specific errors for each MSG case of failure:
2233 host failure, transfer failure, timeout, task cancelled
2234 * Cleanup the exceptions that may get thrown by each function
2235 * Other internal cleanups in Java bindings. Performance still bad :/
2236 Ruby and Lua Bindings: create them
2237 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2238 * That's new and great, you should try them out.
2239 Same functionalities than Java bindings, only even less polished
2241 * Kill the useless "rate" argument of SD_task_get_execution_time()
2242 Everyone used to provide -1 as a value, it was not used, and the
2243 semantic of a possible use wasn't even clear.
2244 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2245 either as comm costs or compute costs to mean that there is no
2246 such thing for that specific task.
2247 * Add a SD_task_set_name() function
2248 * Fix SD_task_unschedule() on typed tasks
2249 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2250 * In DAX loader, accept useless 'level' attributes to <job> since
2251 LIGO DAGs have them (seem to be to ease graphical representation).
2253 * Add an example masterslave_mailbox.c using send/receive and not
2254 the deprecated put/get interface.
2255 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2256 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2258 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2259 (declare MSG_USE_DEPRECATED to still have the old name)
2260 * Add a MSG_task_set_data() function
2261 * About trace replay (see examples/msg/actions):
2263 - Allow to work with splitted trace files for each process
2264 Give the specific trace file as argument of each process,
2265 and call MSG_action_trace_run(NULL)
2266 You can still have one merged file for all processes.
2267 - Fix implementation of collective operations
2268 * Allow task_execute() on 0-sized tasks (closes #10063)
2270 * This is the first release of SimGrid where SMPI is not considered
2271 beta anymore (even if some corners should still be improved)
2272 * Port over the new SIMIX_network submodule (internal refactoring)
2273 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2274 * Implement more missing elements of the standard:
2276 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2278 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2279 MPI_Scan() MPI_Get_processor_name()
2280 - Added implementation of missing case for Alltoall (warning: it's
2281 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2282 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2283 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2284 SMPI_MPI_Allgatherv()
2285 * Bug fixes include:
2286 - MPI_Waitsome() was broken
2287 - Allow relative includes in smpicc
2288 - Command line cfg argument 'reference_speed' was ignored...
2289 - Some functions did not properly lead to auto-benching of user code
2290 - smpicc passes -O2 by default (just like openmpi one)
2292 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2293 * Bug fixes about timeouts during communications
2294 * add SIMIX_message_sizes_output() as a pimple to write to file the
2295 amount of messages per size. Use gnuplot to get histogram.
2296 Pimple because that's the only user-visible function of simix,
2297 defined directly in xbt.h (irk, sorry)
2299 - Add a SIMIX_sem_get_capacity() function
2300 - Fix interactions with processe resume/suspende
2301 - release_forever() was stupidly broken
2302 - Fix SIMIX_display_process_status() for processes in a semaphore
2303 - Make SIMIX_sem_block_onto() user-visible
2304 * Refactoring context stuff:
2305 - Use pseudo-OOP for better modularity
2306 - reimplement SIMIX_process_kill() without process_schedule() so
2307 that the latter can take as invariant that it is called from
2309 - Merge context_start into context_new for sake of simplicity
2311 * Add a Vivaldi network model, coded live during SUD'10 ;)
2312 * Rename configuration variables to start a hierarchy:
2313 o cpu_model -> cpu/model
2314 o network_model -> network/model
2315 o workstation_model -> workstation/model
2316 * New configuration variables:
2317 o network/bandwidth_factor: correction to bandwith
2318 o network/latency_factor: correction to latency
2319 o netwotk/weight_S: correction to the weight of competing streams
2320 * Add a long description to the models, that users can see with such
2321 argument on the command line: --cfg=cpu/model:help
2322 * --help-models display the long description of all known models
2324 * config: add the ability to set a default value after registration
2325 Does not override any previously set value (e.g. from cmd line)
2326 * dict: allow to have integer key and data.
2327 When so, you need to use the following functions
2328 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2329 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2330 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2331 In contrary to regular dicts, the key is not malloced before copy.
2332 Mixing scalar and regular elements in the same dict is not tested
2334 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2335 Tracing for Visualization:
2336 * SimGrid is now instrumented in order to generate a trace file for
2337 visualization analysis: to use it, need to compile SimGrid with the
2338 "tracing" option enabled, and instrument the program using SimGrid with
2339 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2340 (among other functions).
2341 * The instrumentation only traces the platform utilization for now
2342 * Documentation to use the tracing functions and how to analyze the
2343 traces with the Triva tool is written.
2344 * More information about: SimGrid FAQ (in the section Tracing Simulations
2347 * We moved to cmake as default build system. Autotools support will
2348 be dropped soon. Check the FAQ for more info about how to use it.
2349 * Greatly improved our cdash/ctest interactions
2350 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2351 * Added memory checking tests with valgrind; lot of memleak fixing.
2352 This may be the first release of simgrid with so few memory issues
2353 * Added code coverage tests.
2354 Our coverage is still improvable, but at least we see it on cdash.
2356 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2358 SimGrid (3.3.4) stable; urgency=low
2360 The "Desktop Grid needs love too" release (also called Xmas release).
2362 Models improvements:
2363 * Major speedup in the maxmin system solving by using lazy evaluation
2364 Instead of solving completely the maxmin system at each iteration,
2365 only invalidate (and recompute) the modified parts.
2366 This new feature is enabled in default models but you can try to
2367 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2368 * Cas01 IMproved as default CPU model
2369 This CPU model is the same Cas01 model, but it uses the
2370 maxmin-selective-update flag and a heap structure to manage
2371 actions on SURF kernel.
2372 It reduces the complexity to find the next action to finish and,
2373 consequently, it's faster than the old Cas01.
2374 This is the new default CPU model (Cas01).
2375 * Rename the old Cas01 model to Cas01_fullupdate
2376 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2377 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2378 * CpuTI (CPU Trace Integration)
2379 A new CPU model whose objective is simulate faster when using
2380 availability trace files.
2381 Instead of using a full featured, over engineered maxmin system for
2382 CPU modeling, this model does the pre-integration of traces files
2383 to calculate the amount of CPU power available, and so, executes
2384 faster than the old CPU models.
2385 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2386 * Use LV08 as default network model since it gives better accuracy
2387 for small messages and shouldn't change things for big ones.
2388 Use --cfg=network_model:CM02 to get the previous behavior.
2391 ******************************************
2392 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2393 ******************************************
2394 * The new CPU model may changes simulations!
2395 The point is that events occurring at the exact same timestamp
2396 are not scheduled in the same order with the old and new
2397 version. This may be enough to completely change the execution
2398 of simulations in some cases.
2399 * The new network model will change simulations!
2400 This new model is more realistic than the previous one, so you
2401 should consider redoing your old experiments with this model.
2402 Sorry for the inconvenience.
2405 * Introduce the supernovae compilation mode
2406 When compiled that way, the whole SimGrid (or almost) is put in a
2407 single compilation unit and compiled in one shoot.
2408 This is to help gcc which has difficulties to inline stuff from one
2410 The speedup seem to be above 15%, althrough more tests are needed on
2411 amd64 to confirm that gain.
2414 * Port of MSG's mailbox on top of SIMIX network
2415 The put/get mechanism was greatly simplified on the way.
2418 * New SIMIX network module. Provides:
2419 - Mailbox: rendez-vous mechanism to find with who you want to speak
2420 - Synchronous send/recv: easier and hopefully faster since the
2421 logic is handled in the maestro process directly now
2422 - Asynchronous send/recv: you dreamt of it? It's here now
2423 Too bad that nobody cared enough to propagate the change to MSG.
2424 * Add semaphores as SIMIX synchronization mechanism.
2427 * new function SD_daxload(char*) to load a DAX file
2428 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2429 * Introduce typed tasks. Specify its kind and cost at creation.
2430 At scheduling, just give where it should be placed, and the cost
2431 for each involved resource is automatically computed.
2432 Existing constructors so far (more to come of course):
2433 - SD_task_create_comm_e2e() for end-to-end communication
2434 - SD_task_create_comp_seq() for sequential computation
2435 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2436 * new function SD_task_dump() for debuging display
2437 * new function SD_task_dotty(task,FILE*) writing to file the info
2438 about the task in dotty format
2439 * SD_task_dependency_exists() can now cope with having one of its
2440 arguments NULL. If so, it tests whether the other argument has any
2442 * Add getters on list of preceding/following tasks:
2443 SD_task_get_parents(task) and SD_task_get_children(task)
2444 * Add getters on amount of workstations and list:
2445 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2446 * Add getter on task kind: SD_task_get_kind(task)
2447 * Update the start_time and finish_time of tasks on completion/failure
2448 * Bugfix: Remove task from state swags when destroyed
2451 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2454 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2455 dynar with one separate thread per value of the dynar.
2456 * Change the prototype of xbt_thread_create(), sorry.
2457 Added a boolean parameter indicating whether we want to join this
2458 thread (used in SG only for now)
2459 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2462 * GTNetS wrappers should now be usable again (and betterly tested too)
2463 * Fix a major regression from 3.2 where the timeout provided to
2464 MSG_task_put_with_timeout() was used as absolute time before which
2465 the comm should be done.
2466 * Start to fix the <cluster> tag.
2467 - Internal links should be good now (beside of the loopback, which
2468 use the private link instead)
2469 - paths to the external world is still rather broken
2470 - the <route:multi> tag is just broken. Actually that's brain-dead.
2471 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2472 to make it less stupid
2473 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2474 * Fix a source-level compatibility glitch from 3.2: after defining
2475 MSG_USE_DEPRECATED, you can use the old name
2476 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2477 * Allow to compile from the SVN with automake 1.11
2478 * Fix some problems when using the "start_time" tag in deployment XMLs.
2479 * Fix #8569: XBT/synchro.h has redundant declarations
2480 * Fix #8563: MSG return values and exceptions
2481 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2482 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2483 Thanks to Robson Peixoto.
2484 * Fix a vicious bug in dictionaries inducing that some elements were
2485 not freed on xbt_dict_free()
2487 Portability report of this version:
2488 * Main portability targets:
2489 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2490 on (amd64/i386/ia64)
2491 - mac leopard on i386
2492 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2493 but nothing critical.
2494 * Other platforms: windows, AIX and others were not tested for this release
2496 Timing report of this version:
2497 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2498 scenario parameters). From 8h to a few seconds in desktop grid settings.
2499 * Supernovae brings about 25% speedup on i386.
2501 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2503 SimGrid (3.3.3) stable; urgency=low
2505 The "Need for Speed" release.
2507 The timings done to validate the 3.3.2 were faulty.
2508 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2510 The problem was a conversion from a manually handled vector to
2511 xbt_dynar_t on the critical path.
2512 xbt_dynar_foreach calls functions, inducing stack management crap.
2514 We inlined these functions and xbt_dynar_foreach is now breath taking.
2515 We also inlined xbt_swag_belong on the way.
2517 Here are some approximate speedup measurements (on master/slaves
2518 simulations lasting between 10s and 20s each):
2519 3.3.1 -> 3.3.2: about same performance
2520 3.3.2 -> 3.3.3: 40% speedup
2521 3.3.1 -> 3.3.3: 40% speedup
2522 3.3.1 with inline patch -> 3.3.3: 30% speedup
2524 Our reading is that the refactoring which occurred in 3.3.2 made us
2525 suffer much more from the xbt_dynar_foreach low performance, but
2526 once we solved this, this refactoring proved to be very performance
2527 effective. From the 40% speedup, somehow, 10% are due to the
2528 inlining and 30% to the refactoring.
2530 That's a pitty that gcc cannot inline functions placed in other files
2531 alone. We have to choose between:
2532 - break the encapsulation (by putting private data structures and
2533 accessors in headers files to help gcc)
2534 - live with low performance
2535 - switch to a decent compiler such as icc (not quite possible).
2537 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2539 SimGrid (3.3.2) stable; urgency=low
2541 The "Simplicity does not preceed complexity, but follows it" release.
2543 The main contributors of this release were (lexical order):
2544 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2547 * Extract the routing logic into its own object.
2548 (was dupplicated in network.c and workstation_LV07.c;
2549 Allows to implement other ways of storing that info)
2550 => kill now useless network_card concept
2551 - Use dynar to represent routes (instead of void** + int*)
2552 - kill link_set (use surf_network_model->resource_set instead)
2553 - Add a command-line option to choose the routing schema to use
2554 - Add three new models:
2555 * Floyd (shortest path computed at initialization)
2556 * Dijikstra (shortest path recomputed all the time)
2557 * Cached Dijikstra (shortest path computed on need)
2558 All these models where contributed by Silas De Munck, and are
2559 described in his ICCS09 paper.
2561 * Simplify model declaration
2562 (less redirections, less function to write when defining a model)
2563 - Factorize stuff between models:
2566 surf_model_resource_set(model)
2567 surf_model_resource_by_name(model, name)
2568 - Unify the types of models in s_surf_model_t (using an union)
2569 - Embeed fields of common_public directly into s_surf_model_t
2570 - Rename model methods:
2571 action_free ~> action_unref
2572 action_change_state ~> action_state_set
2573 action_get_state ~> action_state_get
2574 - Change model methods into functions:
2575 (model)->common_public->action_use ~> surf_action_ref
2577 * Implement a generic resource; use it as ancestor to specific ones
2578 (allows to kill duplicated code in models)
2579 Drawback: timer command don't need no name nor properties;
2580 workstation_CLM03 don't need no properties
2581 (but I guess we can live with those few bytes wasted)
2583 * Improve the action object model
2584 - implement a constructor avoiding dupplicated code about field
2585 initialization in generic_action part.
2587 * Kill the SDP model: it has an external dependency, is deprecated
2588 in flavor of modern lmm models, and didn't compile since a while
2591 * Relocation of the context module from XBT to SIMIX.
2592 (the context were decoupled from the simix processes, duplicating a lot of code)
2593 => a lot of code was factorized
2594 - less overhead is introduced during scheduling
2595 - simpler API for the context factory
2596 - the logic for process creation,destruction and manipulation was simplified
2597 * Simplification of the s_smx_process_t data structure.
2598 => accesing the simix level data associated to a process is faster now,
2599 and the code is a lot more readable.
2602 * Implement some more MPI primitives:
2603 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2604 -implementation: Bcast: flat or 2-ary tree (default),
2605 Barrier: 4-ary tree,
2607 Allreduce: Reduce then Bcast
2608 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2609 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2610 Alltoallv: flat tree, like ompi
2612 * Add support for optimized collectives (Bcast is now binomial by default)
2613 * Port smpirun and smpicc to OS X
2616 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2617 Shout out if you used it.
2620 * Display the list of still queued messages in SG mode when existing
2624 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2625 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2626 * Add xbt_str_from_file(FILE*)
2627 * Add xbt_dict_get_key achieving a linear reverse search
2628 * Remove the context module
2630 Portability report of this version:
2631 * Main portability targets:
2632 - Linux(debian)/x86/context
2633 - Linux(debian)/x86/pthreads
2634 - Linux(debian)/amd64/context
2635 - Linux(debian)/amd64/pthreads
2636 On these, we still have the eratic breakages of gras/pmm and
2637 amok/saturate_sg reported in previous version. We still think
2638 that the tests are the cause of the fault, not the tested code.
2640 - Mac OSX Leopard/x86/context
2641 Still false negative in tesh autotesting.
2642 Smpi still fails, but this time because readlink does not accept -f
2643 Everything seems to work properly beside of that.
2646 - AIX version 5.3 (only tested contexts this time)
2647 Smpi still fails there because mktemp is not installed.
2648 Everything seems to work properly beside of that.
2650 I managed to compile it for the first time, but several breakages.
2651 Won't delay the release for this exotic platform.
2653 * Windows: it's still lagging behind. If you want to help, please
2656 Timing report of this version:
2657 This version seem to be more than 5% faster than 3.3.1 (on linux
2658 64bits with contextes). The gain is less than expected, we are
2659 investigating this for next release.
2661 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2663 SimGrid (3.3.1) stable; urgency=low
2666 * Implement a --cfg-help to show existing configuration variables
2667 * Build chain do not require doxygen in maintainer mode
2670 * fix a bug on struct sizeof computation, which prevented the
2671 exchange of arrays of structs in some conditions
2672 - added a regression test about this in datadesc_usage
2673 * Allow the exchange of 0-long dynamic vectors.
2674 - for that, use -1 as indicator of dynamic size instead of 0
2675 - This implied to change any size from unsigned long to long,
2676 reducing a bit communication abilities, but I guess that with
2677 64bits being quite common, this is more than enough.
2678 - This also induce a protocol change, thus bumping network protocol
2679 version from 0 to 1 (if we have external users, we have to get
2680 clean on that point too ;)
2681 - added two regression tests about this in datadesc_usage
2682 * Be more verbose when propagating local exceptions
2683 This helps debugging.
2684 * Display the status of simulated processes when receiving SIGINT in
2688 * Allow to control the simulation from a trace file.
2689 New functions MSG_action_register() and MSG_action_trace_run()
2690 The first one allows to associate a function execution to each
2691 kind of action while the second one parses a trace file and
2692 triggers the corresponding actions within the system.
2693 For now, only a toy example is provided in examples/msg/actions
2694 * Add an exemple of process migration in examples/msg/migration
2695 * Fix a bug in task exchange which broke MSG_task_get_sender()
2696 Add a teshsuite regression test for that.
2697 [Bug: if MSG_task_get_sender() is called after sender exit,
2699 * Fix a bug which prevented suspend/resume to work properly
2700 * Display the status of simulated processes when receiving SIGINT
2701 This fixes a regression of v3.3. due to the introduction of SIMIX
2702 * Bug fixing in failure management:
2703 - trace could not start by a failure at time 0
2704 - failure during communications were not working
2707 * Add SIMIX_process_set_name() to change the name of the current
2708 process in the log messages.
2709 * Store smx_hosts in a dict since we only retrieve them by name
2710 * Move the configuration infrastructure to surf
2713 * Move the configuration infrastructure to surf
2716 * Massive internal cleanups:
2717 - Store internal structures on processes instead of hosts (allows
2718 to have more than one process per host, in addition of being more
2720 - Cleanup the initialization/finalization process
2721 - Kill a whole bunch of unneeded synchronization:
2722 processes run in exclusive manner within the simulator
2723 - Move queues from global tables to process data fields
2725 - now accept -platform and -hostfile arguments
2726 - Pass the right rank value to processes according to the hostfile
2727 * Compile the examples by default, and use them as regression tests
2728 * Implement MPI_Wtime()
2729 * Change the reference speed to a command line option
2732 * TCP_gamma can now be specified as command line option using
2733 --cfg=TCP_gamma:10000000.0
2734 * Change the --surf-path cmd line option into --cfg=path:
2737 * Also include strbuff from xbt.h public header
2738 * xbt_ex_display(): do not free the exception after displaying
2739 This allows to do more with the given exception afterward.
2740 Users should call xbt_ex_free() themselves.
2744 Portability report of this version:
2745 * Main portability targets:
2746 - Linux(debian)/x86/context
2747 - Linux(debian)/x86/pthreads
2748 - Linux(debian)/amd64/context
2749 - Linux(debian)/amd64/pthreads
2750 These targets fail about 1/10 of times on gras/pmm, but we believe
2751 that this is because of the test, not because of simgrid.
2752 amok/saturate_sg fails even more rarely, and the test may not be
2755 - Mac OSX Leopard/x86/context
2756 The test suite still spits tons of errors because some obscure
2757 force prevents us from removing the temporary directories
2758 arguing that they still contain some metadata I've never heard of.
2759 Smpi fails because seq is not installed.
2760 Everything seems to work properly beside of that.
2763 - AIX version 5.3 (both contexts and pthread)
2764 Smpi still fails there because mktemp is not installed.
2765 XML inclusions seems rosty on AIX.
2767 * Windows: it's still lagging behind. If you want to help, please
2770 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2772 SimGrid (3.3) stable; urgency=high
2776 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2777 [Malek Cherier & Mt]
2779 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2782 Until now, GRAS were using MSG as an interface to SURF. It was
2783 quite difficult because both interface have several differences
2784 (MSG channels vs GRAS sockets were the most notable point).
2786 This also opens the gate to SMPI (which should occur soon) and speed
2787 up simulations by to 40% (even if it were not the main goal).
2789 **************************************
2790 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2791 **************************************
2792 The point is that events occuring at the exact same timestamp are
2793 not scheduled in the same order with the old and new version. This
2794 may be enough to completely change the execution of simulations in
2795 some cases. Sorry for the inconvenience.
2797 * Cleanup and upgrade the XML format to push further scalability
2798 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2800 * Improve the testing infrastructure with tesh. Now a very large part of
2801 the code is tested not only by being run but also by checking that the
2802 output match an expected output [Mt].
2804 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2805 is really less memory-demanding, which should allow you to use
2806 larger files in SimGrid [AL].
2808 * Inform valgrind about our contextes, so that it becomes usable
2809 with the default (and more effecient) version of SimGrid
2810 [contributed by Sékou Diakite, many thanks]
2813 * Introduce a listener thread in charge of receiving incoming
2814 messages from the network. It allows to overlap communication and
2815 computation but most notably, it removes some stupid deadlocks due
2816 to the fact that so far, a process could not send and receive at
2817 the same time. This made most non trivial communication schema
2819 * Convert the PIDs from long int to int to match the MSG ones (and
2820 linux ones too) [Mt]
2821 * New function: gras_agent_spawn() to launch a new process on
2822 current host. Only working in simulation for now. [Mt]
2823 * New function: gras_os_hostport() returning a constant form (ie,
2824 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2827 * Make the backtrace of exceptions more human readable [Mt]
2828 * New module: xbt/str [Mt]
2829 a ton of string utility functions (split, join, printf to a newly
2830 allocated buffer, trim, etc)
2831 * New module: xbt/hash [Mt]
2832 SHA1 hashing algorithm (more to come if needed)
2833 * New module: xbt/synchro [Mt]
2834 synchronization tools (mutex and conditions) working the same way
2835 in simulation and in real life (mainly useful for GRAS, but not
2837 * New module: xbt/queue [Mt]
2838 classical producer/consumer synchronization scheme
2839 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2840 (using the classical functions will get serialized) [Mt]
2841 * Make dictionary internal table dynamic. No need to specify its size
2842 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2844 * Make sure the log channels are organized as a tree under windows
2845 (because of ANSI C compatibility issue, any channel were child of
2846 root directly) [Mt].
2849 * Cleaned many thing in surf and fixed a few bugs [AL].
2850 * Add a nice command line configuration mechanism to compose models [AL].
2851 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2852 the previous one (KCCFLN05). It relies on something that looks like
2853 a max-min sharing mechanism but cannot be written as such. A new solver
2854 was thus designed [AL].
2855 * Add a new solver to lmm. Based on Lagrange optimization and
2856 gradient-based descent, it enables to efficiently maximise systems s.a
2858 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2860 This solver enables to propose two new network models for TCP Reno and
2861 TCP Vegas based on Low's work. These models still need to be fully
2862 tested though [Pedro Velho].
2865 * Bug fix in SD_simulate. Now the time bound given as argument is
2867 * Use the new parallel task model (ptask_L07) as default.
2868 * Use the SURF command line configuration mechanism.
2869 * 0-size tasks (for synchronization) should now work.
2871 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2873 SimGrid (3.2) stable; urgency=high
2877 We still experience issues on this platform, but we believe that at
2878 least MSG is usable.
2880 GRAS API BREAKAGE (for simplification purpose, sorry):
2881 * the gras_msgtype_by_name is not used anymore. Instead of
2882 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2883 you can write (and must)
2884 gras_msg_send(toserver, "request", &request);
2885 - If you still want to pass a gras_msgtype_t to the function (to cache
2886 the type and avoid the lookup time), use the gras_msg_send_() variant.
2887 - Impacted functions:
2888 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2889 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2890 * The callbacks are now expected to return 0 when everything went well
2891 (just like the main() function)
2893 GRAS new features and improvements:
2894 * New module mechanism where user code can use per process globals [Mt]
2895 This is similar to gras_userdata_*() functions, but for libraries. It
2896 factorize some code developped over and over in the examples and AMOK.
2897 It has still to be documented and used (only amok/peermanagement is
2899 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2900 loss when they were small enough to fit into the buffer and sent quickly
2901 enough so that they can all get received in one shoot.
2902 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2903 if not found. Use the *_or_null() variant for the old semantic.
2904 * In gras_msg_handle, do not discard messages without callback.
2905 They are probably messages to be explicitly awaited later (ie, proofs of
2906 mis-synchronization in userland since they are sent before being awaited)
2907 No big deal usually.
2908 * gras_socket_meas_send/recv: semantic changed!
2909 The numerical arguments used to be (1) the total amount of data to send
2910 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2911 messages. This was need for the fool willing to send more than MAXINT
2912 bytes on quite fat pipes.
2915 * Do really rename the hostmanagement module to peermanagement. [Mt]
2916 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2917 but this is rather new and this was documented in the module
2918 documentation (poor excuses, I admit)
2919 * Bandwidth measurement semantic changed! This follows the changes to
2920 gras_socket_meas_send/recv explained above.
2923 * A sequential mode has been added to the workstations. When a workstation
2924 is in sequential mode, it can execute only one task, and the other tasks
2925 are waiting in a FIFO. [Christophe Thiery]
2928 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2929 model for SIMDAG. [Christophe Thiery]
2930 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2931 equal to 0, which caused some bad behaviors in
2932 saturated_constraint_set_update. I now use a threshold mechanism like in
2936 * When running manually src/testall, you select specific units [Mt]
2937 testall is the result of our cunit mechanism, and should replace all
2938 the scripty thingy around since bash don't run easily on billware.
2940 * A mallocator system has been added. [Christophe Thiery]
2941 Mallocators allow you to recycle your unused objects instead of freeing them
2942 and allocating new ones.
2944 Documentation update:
2945 * FAQ reworking + New FAQs:
2946 - "Valgrind spits tons of errors!" [Mt]
2947 - "How to repport bugs" [Mt]
2948 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2949 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2950 - Communication time measurement within MSG [AL]
2951 - I experience weird communication times when I change the latency [AL]
2952 * GRAS tutorial [Mt]
2954 - an introduction to the framework and to the used communication model
2955 - an initiatic tour introducing the most proheminent features:
2957 . Lesson 0: Installing GRAS
2958 . Lesson 1: Setting up your own project
2959 o Part 2: Message passing
2960 . Lesson 2: Exchanging simple messages
2961 . Lesson 3: Passing arguments to the processes (in SG)
2962 . Lesson 4: Attaching callbacks to messages
2963 . Lesson 5: Using globals in processes
2964 . Lesson 6: Logging information properly
2965 . Lesson 7: Using internal timers
2966 . Lesson 8: Handling errors through exceptions
2967 . Lesson 9: Exchanging simple data
2968 . Lesson 10: Remote Procedure Calling (RPC)
2969 . Lesson 11: Explicitely waiting for messages
2970 . Recapping of message passing features in GRAS
2971 - A HOWTO section containing:
2972 o HOWTO design a GRAS application
2973 More are due, of course. They will come latter. In the meanwhile, you can
2974 check the examples which are still here.
2976 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2978 SimGrid (3.1) stable; urgency=high
2982 There was a stack corruption somewhere, visible only when optimizing
2983 with these versions. [Vince]
2986 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2987 module that enabled to play with Directed Acyclic Graphs. It is built
2988 directly on top of SURF and provides an API rather close to the old
2989 SG. Some old codes using SG are currently under rewrite to check that
2990 all needful functions are provided. [Christophe Thiery]
2993 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2994 extension of the classical CLM03 model that gracefully handles
2995 failures. This is now the default model for MSG and GRAS. It doesn't
2996 handle parallel tasks yet though. [AL]
2997 * Bug fix: Weights were not correctly set in the network part.
2998 WARNING: This may have resulted in incorrect results with simulations
2999 where there are more than one flow on a given link. [AL]
3002 * After a (long ?) discussion on simgrid-devel, we have decided that the
3003 convention we had on units was stupid. That is why it has been decided
3004 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3005 WARNING: This means that all previous platform files will not work as
3006 such with this version! A warning is issued to ask users to update
3008 A conversion script can be found in the contrib module of the CVS, under
3009 the name contrib/platform_generation/surfxml_update.pl [MQ]
3012 * Bug fix: Processes were started in reverse order, wrt deployment file.
3013 WARNING: if your code relies on this bug, please fix it. [AL]
3014 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3015 being executed on two different locations. [AL]
3016 * Bug fix: Failures are now better supported thanks to Derrick's tests
3017 (there was many failure situations I hadn't thought of and that weren't
3018 correctly handled). [AL]
3019 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3023 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3024 Exception occurring on server-side are propagated back to client (!).
3026 API CHANGE: the callback changed their prototype. Change:
3027 int my_handler(gras_socket_t expeditor, void *payload_data) {
3029 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3030 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3032 * New! function: gras_msg_handleall to deal with all messages arriving
3033 within a given period.
3034 * New! function: gras_socket_server_range to get a server socket in a
3035 range of port numbers (ease to avoid port number conflicts) [MQ]
3036 * New! gras processes display their backtrace when they get a SIGUSR1
3037 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3038 Sweet to debug RL processes [MQ]
3042 - Do not force experiment sizes to be expressed in kb, or it becomes
3043 impossible to measure the latency this way (needs one byte-long tests)
3044 WARNING: this changes the amok_bw_* function semantic. [MQ]
3045 - Implements the link saturation stuff. [MQ]
3046 * Peer management module:
3047 New! module factorizing code that we wrote over and over [MQ].
3050 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3051 - Test units are placed directly into the library code, they get extracted
3052 automatically and placed into the src/testall binary.
3053 - Convert most of the XBT tests to this system.
3054 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3055 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3056 live in the same C code. This required to change a little bit the API
3057 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3058 * New module: sparse graph structure with basic algorithms (this is work
3059 in progress and the API is not considered to be frozen yet). [AL]
3060 * Display more information on backtraces: source line & function names are
3061 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3062 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
3063 * Reworked a little bit some #include statements to load only required
3064 headers. Some user code that relied on SimGrid to include stdlib or
3065 stdio may need to include it by themselves. [AL]
3066 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3068 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3069 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3070 but I guess that almost nobody used those parts. [MQ]
3072 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3074 SimGrid (3.0.1) stable; urgency=low
3077 * Unfortunately, I had missed 5 misnamed functions:
3078 xbt_fifo_item_t xbt_fifo_newitem(void);
3079 void xbt_fifo_freeitem(xbt_fifo_item_t);
3080 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3081 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3082 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3083 They're now deprecated. Please use their new versions:
3084 xbt_fifo_item_t xbt_fifo_new_item(void);
3085 void xbt_fifo_free_item(xbt_fifo_item_t);
3086 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3087 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3088 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3090 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3091 * Documentation: xbt_log module unmercifully reworked [MQ]
3092 * Bugfix: there was a problem with the ending of contexts with
3093 the pthread backend. It caused some weird deadlock or behavior
3094 depending on the pthread implementation. [AL]
3095 * Bugfix: get the exceptions raised in the simulator repport where
3096 and why they come from when they are not caught in time [AL, MQ]
3099 * Bugfix: Do repport the error when two non-connected hosts try to
3100 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3103 * Add additionnal checkings on communications. Assert that two
3104 communicating hosts are connected by a set of links... [AL]
3107 * Add additionnal checkings on channel values in communication [AL]
3108 * New: MSG_task_get_source to see on which host a task was generated [HC]
3109 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3110 the number of tasks waiting to be received on channel and sent
3112 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3113 waits for the first task coming from a given host.. [AL]
3115 GRAS new functionnalities: [MQ]
3116 * Enhance the parsing macro to allow the size of multidimentional objects
3117 to be given thru annotations.
3118 * New example (and documentation): Matrix Multiplication a la RPC
3119 (as when I was young!) and fix a bunch of bugs found on the way.
3121 GRAS performance improvements: [MQ]
3123 * Reduce the amount of cbps creation/destruction by making it static to
3124 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3127 * Change libdata to a set so that we can search for stuff by ID (and thus
3128 reduce the insane amount of dict lookups)
3131 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3132 RL communication on top of SG since it's so uneffective.
3133 It may also be used for inter-thread communication in RL, one day.
3134 * Use gras_datadesc_copy() to exchange messages on top of SG
3136 - improve message exchange performance on top of SG
3137 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3138 * Don't exchange on the network the size of the used part of buffer,
3139 instead, specify the possible buffer size to read().
3141 - reduces the amount of read/write calls (one pair per exchange)
3142 - reduces the amount of exchanged data (the size)
3143 - allows to retrieve all arrived data on receiver side, if we don't need
3144 it right now (subsequent read will peek the buffer)
3145 - allows the receiver to proceed with the begining of the stream before
3146 everything is arrived
3147 - make it possible to build an iov transport (using readv/writev)
3149 - take care of the data with non-stable storage (like stacked data),
3151 * If possible, TCP send uses vector I/O (when writev() is here)
3152 - Don't use it for receive since we send data sizes and data on the
3153 same stream, so we wouldn't be able to chain large amount of chunks
3154 before having to flush the stuff to read the size.
3155 * Rework the transport plugin mechanism to simplify it and reduce the
3156 amount of pointer dereferencement when searching for the right function
3159 * I guess that now, we do almost as few system calls as possible while
3160 doing as few data copy as possible.
3162 To improve it further, we could try to send all the sizes first and then
3163 all the data (to use iov on receiving size), but it's only a partial
3164 solution: when you have 2 dimensional data, the sizes of the second
3165 dimension is data of the first dimension, so you need 3 streams.
3167 I'm not sure the potential performance gains justify the coding burden.
3169 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3171 SimGrid (3.00) stable; urgency=high
3174 * New! Give the possibility to hijack the surf parser and thus bypass
3175 MSG_create_environment and MSG_launch_application. Have a look at
3176 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3178 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3180 SimGrid (2.96) unstable; urgency=low
3185 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3186 This deprecates the xbt_error_t mechanisms.
3187 It modifies (simplifies) all XBT and GRAS API.
3188 MSG API keeps unchanged (exceptions raised by XBT are caught from
3189 within MSG and masked with existing error handling facilities)
3192 * New! Add a FATPIPE model. [AL]
3193 * New! Add a parallel task model. [AL]
3194 * New! Add automatically a loopback interface (in the default
3195 network model) if none was precised.
3198 * Bugfix: MSG_process_resume now works with the current running process.
3200 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3201 * Modification of MSG_task_get_compute_duration. Once a task has been
3202 processed, the value returned by this function is now equal to 0. [AL]
3203 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3204 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3205 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3206 of a m_task. For now, MSG_task_cancel only works with computation
3208 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3209 of the processor (in Mflop/s) regardless of the current load on the
3211 * API Change: use proper naming convention for MSG_getClock and
3212 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3214 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3215 This function changes the priority of a computation task. This priority
3216 doesn't affect the transfer rate. A priority of 2 will make a task
3217 receive two times more cpu power than the other ones. This function
3218 has been added to suit the needs of Nguyen The Loc and hasn't been that
3219 much tested yet. So if it fails, please report it and send me your code.
3221 * API Change: removed all functions and types that were marked "deprecated"
3222 since many months. Renamed MSG_global_init_args to MSG_global_init.
3224 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3226 SimGrid (2.95) unstable; urgency=low
3229 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3230 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3231 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3234 * Bugfix: Allow absolute paths for platform description files [MQ]
3235 * Bugfix: do free the variables after use. Leads to drastic performance
3237 * Implement max_duration (ie, timeouts) on resources [AL]
3240 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3242 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3243 the needed support (timeouts) [AL]
3246 * Implement measurement sockets. You can now get the bandwidth between two
3247 hosts thanks to AMOK (see below). [MQ]
3248 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3249 dynar over the network (yeah) [MQ]
3250 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3251 * simple/static token-ring example. [Alexandre Colucci and MQ]
3252 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3253 in place before [MQ]
3256 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3257 type parsing. "short" and "long" modifiers were also ignored; other
3258 modifier (such as reference level) are still ignored. [MQ]
3259 * Embeed the buffer size within the buffer itself on SG. [MQ]
3260 That way, send() are atomic and cannot get intermixed anymore (at least
3261 the ones which are less than 100k; bigger messages still have the issue)
3262 * Array size pushed by the field, not by the field type (or each
3263 and every long int will push stuff to the cbps) [MQ]
3264 * use select() to sleep since it allows to portably sleep less than one
3267 GRAS (minor cleanups)
3268 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3271 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3272 * type category 'ignored' killed as it was never used and were difficult
3274 * whether a type can cycle or not is now a flag, leaving room for more
3275 flags (as "ignored", if we feel the need one day ;)
3276 * Rename raw sockets to measurement sockets since "raw" has another
3277 meaning in networking community.
3280 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3281 and offers features not belonging to GRAS but that most applications
3282 need. One day, it may be a set of plugins loadable at runtime.
3283 * New module: bandwidth
3284 bandwidth measurement between arbitrary nodes running this module. [MQ]
3286 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3288 SimGrid (2.94) unstable; urgency=low
3290 The first beta release of SimGrid 3 !
3294 * Update the main page and the FAQ. Adding references to gforge.
3297 * Add a gras_os_getpid function.
3300 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3301 * Extend the logs so that they also print PID, hostname, date, ... if
3303 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3304 and kill the old version which were in testsuite/
3305 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3309 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3310 for users. Instead, each programming environment (GRAS, MSG,...) use it
3311 when needed to provide such a feature to users.
3312 Don't shortcut the mechanism or you will also shortcut the virtualization
3313 you need on the simulator.
3317 * Cleanups in configury with regard to compile optimization/warning flags.
3318 Also add -fno-loop-optimize to any powerpc since it's the optimization
3319 killing gcc (< 3.4.0).
3320 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3321 needed by MSG examples complications
3322 * Borrow configury beautifications from PHP
3325 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3326 warning (thanks loris for stumbling into this one).
3327 * Bugfix: stop loading private headers (gras_config.h) from the public
3331 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3332 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3333 RL would imply a DNS resolver, which is *hard* to do in a portable way
3334 (and therefore delayed).
3335 * Implement a real timer mechanism and use it in timing macros. This allows
3336 to avoid rounding errors and get a 0.000005 sec precision in timing
3337 macros. While I was at it, various cleanups:
3338 - allow to declare more than one timed section per file (fix a stupid bug)
3339 - move some private declaration to the right place
3340 - merge conditional execution and timing macros into emulation module
3341 - document the module
3342 - make sure the module cleanups its mess on gras_exit
3343 * Documentation improvements:
3344 - (new) how to compile applications using GRAS
3345 - (new) emulation support (timing macros)
3347 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3349 SimGrid (2.93) unstable; urgency=low
3351 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3354 - Use Paje properly where used. Still to be sanitized properly.
3355 - Portability fix: Add an implementation of the contexts using pthread
3359 - Add xbt_procname(): returns the name of the current process.
3360 Use it to show the current process's name in all logging.
3362 - fix detection of older flex version and the reaction, since we do
3363 depend on modern ones (we use lex_destroy)
3364 - Better separation of SG and RL in the libs: remove all simulation code
3365 from libgras. As a result, this lib is now only 200k when stripped.
3366 Some of the xbt modules may also be duplicated (two sets and such) and
3367 should be cleaned/killed before SG3.
3368 - Insist on using xlC on AIX because of weird problems involving gcc there.
3369 - Cleanup the make remote stuff. This is now done by scripts
3370 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3371 mainly for our private use, but we're working on changing them to user
3374 - Bugfix: flush the socket on close only if there is some *output*.
3375 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3377 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3380 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3382 SimGrid (2.92) unstable; urgency=low
3384 Alpha 3 on the path to SimGrid 3
3388 - New! First try of benchmarking macros.
3389 - New! First try so that gras_stub_generator generate deployment and
3390 remote compilation helpers.
3392 - Bugfix: Initialization fix in msg_test.
3396 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3398 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3400 - API Change: gras_os_sleep to take the amount of seconds as a double.
3401 Accepting an int was error prone since it was the only location where
3402 seconds were coded as such. It leaded to damn rounding errors.
3403 - Bugfix: Hard to belive that timers ever worked before this.
3405 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3407 SimGrid (2.91) unstable; urgency=low
3409 Alpha 2 on the path to SimGrid 3
3413 - Bug fix in the lmm_solver.
3415 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3416 through the function MSG_paje_output.
3417 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3418 - It is possible to bound the rate of a communication in MSG with
3419 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3420 to write it in the changelog).
3421 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3426 - Various cleanups to the autotools stuff
3427 - Begin to move Gras examples to examples/gras/
3428 - Let make distcheck work again (yeah!)
3430 - documentation overhauled using doxygen.
3431 gtk-doc-tools is dead in SimGrid now.
3432 - Automatically extract all existing logging categories, and add the list
3433 to the documentation (long standing one, to say the less)
3435 - Cleanup the known architecture table. Reorder the entries to group what
3436 should be, and use a more consistent naming scheme.
3437 (some of the test dataset are still to be regenerated)
3438 - New! Allow library to register globals on each process just as userdata
3440 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3441 lookup time (for now).
3442 Use it in msg and trp.
3443 This cleans a lot the internals and helps enforcing privacy of the
3444 headers between the gras components.
3445 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3446 - Bugfix: gras_os_time was delirious in RL.
3447 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3449 - Reenable GRAS now that it works.
3451 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3453 SimGrid (2.90) unstable; urgency=low
3455 Alpha 1 on the path to SimGrid 3
3457 * It is a long time since the last release of SimGrid. I'm sorry about
3458 that but as I had told you, I was rewriting a lot of things. I apologize
3459 to those who had been reporting bugs to me and that I had not answered.
3460 If your bug is still in the new version, please tell me. Here is a
3461 summary of the main changes.
3463 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3464 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3465 portability, speed, and a lot more but you'll figure it out later.
3466 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3467 I wanted to release the new MSG as soon as possible and I have
3468 broken GRAS, which is the reason why, for now, only MSG is fully
3469 functional. A laconic description of these projects is available
3470 in the documentation.
3472 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3473 kernel optimized for our needs (called SURF but only the developers
3474 should use it). Hence, MSG is now roughly 30 times faster and I think
3475 that by rewriting a little bit MSG, I could event speed it up a little
3476 bit more. Beside the gain in speed, it is also much easier to encode a
3477 new platform model with SURF than it was with SG. More to come...
3479 * REVOLUTION 3: I have tried to change a little as possible the API of
3480 MSG but a few things really had to disappear. The main differences
3481 with the previous version are:
3482 1) no more m_links_t and the corresponding functions. Platforms are
3483 directly read from a XML description and cannot be hard-coded
3484 anymore. The same format is used for application deployment
3485 description. The new format is described in the documentation.
3486 Have a look in tools/platform_generation. There is a tiny script
3487 that converts from the old platform format to the new one. Concerning
3488 the application deployment format, parsing the old one is tricky.
3489 I think most of you should however be able to convert your files. If
3490 it is really an issue, I can write a C code that does the conversion.
3492 2) the toolbox tbx does not exist anymore. We now have a library
3493 with much more data-structures but without the hash-tables (we have
3494 dictionaries that are much faster).
3496 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3498 *****************************************************************************
3499 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3500 * don't feel like converting the oldies. (Mt) *
3501 *****************************************************************************
3504 Version 2.90: "the long awaited one"
3505 - Finished rewriting and debugging MSG. Rewrote the documentation.
3506 - disable GRAS for now since it needs to be ported to the newest SG
3509 - Finish the port to windows (using mingw32 for cross-compile)
3512 - Main loop and datastructures of SURF. A cpu resource object is
3513 functional. Surf can thus be used to create cpu's with variable
3514 performance on which you can execute some actions.
3516 2004-11-15 Martin Quinson
3517 - Port to ARM. Simply added the alignment and size descriptions. Should
3518 work, but the ARM machines are so slow that I didn't had the opportunity
3519 to 'make check' over there yet.
3521 2004-11-15 Arnaud Legrand
3522 - Trace manager now written. It uses a heap structure and is therefore
3523 expected to be efficient. It may however be speeded up (particularly
3524 when many events occur at the same date) by using red and black
3525 trees. One day maybe...
3526 - Max-min linear system solver written. It uses a sparse matrix
3527 structure taking advantage of its expected use. Most operations are
3528 O(1) and free/calloc are called as few as possible. The computation of
3529 the minimum could however be improved by using a red and black tree
3532 2004-11-03 Arnaud Legrand
3533 - Rename every gras_* function that was in xbt/ to its xbt_
3535 - Add a heap and a doubly-linked list to xbt
3536 - Added a dichotomy to the dictionaries. make check works as well before
3537 so I assume that the patch is correct. I do not know however if things
3538 run effectively faster than before now. :)
3540 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3541 SimGrid, and the version number is bumped to 2.x
3543 2004-10-29 Martin Quinson
3544 - Introduction of the remote errors.
3545 They are the result of a RMI/RPC on the remote machine.
3546 ErrCodes being scalar values, you can't get the host on which those
3547 errors did happen. Extending the error mechanism as in Gnome is possible.
3548 No idea yet whether it is a good idea.
3550 2004-10-28 Martin Quinson
3551 - Interface revolution: the Starred Structure Eradication.
3552 I used to do typedef struct {} toto_t; and then handle *toto_t.
3553 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3554 - ???_t is a valid type (builded with typedef)
3555 - s_toto_t is a structure (access to fields with .)
3556 - s_toto is a structure needing 'struct' keyword to be used
3557 - e_toto_t is an enum
3558 - toto_t is an 'object' (struct*)
3560 typedef struct s_toto {} s_toto_t, *toto_t;
3561 typedef enum {} e_toto_t;
3562 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3563 s_toto_t) is private.
3565 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3566 so that it can set the variable to NULL. It was so for dicts and sets,
3567 it changed for dynars.
3569 - Fix a bunch of memleaks in dict_remove
3570 - Fix a bug in sg/server_socket opening: it failed all the time.
3572 2004-10-07 Martin Quinson
3573 - Speed up dynar lookup operation a bit.
3575 gras_dynar_get is dead.
3577 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3578 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3579 (which gives you the address of the stored data).
3581 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3582 copy of the data using an affectation to do the job and not a memcpy.
3584 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3585 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3587 It does not really speedup the dynar test because they are
3588 setting elements all the time (and look them seldom). But the dict does
3589 far more lookup than setting.
3591 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3593 2004-10-05 Martin Quinson
3594 - Allow to (en/dis)able the cycle detection at run time.
3596 Whether we should check for cycle or not is now a property of each
3597 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3598 datadesc_cycle_unset allow to remove this property when previously set.
3600 Note that the cycle detection is off by default since it impacts the
3601 performance. Watch the data you feed GRAS with ;)
3603 This property is hereditary. Any element embedded in a structure having it
3604 set have it set for the time of this data exchange.
3606 You should set it both on sender and receiver side. If you don't set it on
3607 sender side, it will enter an endless loop. If you forget on receiver
3608 side, the cycles won't be recreated after communication.
3610 - Header reorganization.
3611 Kill gras_private.h, each submodule must load the headers it needs.
3613 2004-10-04 Martin Quinson
3614 - Interface revolution: do not try to survive to malloc failure.
3616 Now, gras_malloc and friends call gras_abort() on failure.
3617 As a conclusion, malloc_error is not a valid error anymore, and all
3618 functions for which it was the only gras_error_t return value are
3619 changed. They now return void, or there result directly.
3620 This simplify the API a lot.
3622 2004-09-29 Martin Quinson
3623 - Re-enable raw sockets.
3624 Created by gras_socket_{client,server}_ext;
3625 Used with gras_raw_{send,recv}
3628 It should allow to kill the last bits of gras first version soon.
3630 This is not completely satisfactory yet (duplicate code with
3631 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3634 - Simplify transport plugin (internal) interface by not passing any
3635 argument to _server and _client, but embedding them in the socket
3638 2004-09-28 Martin Quinson
3639 - Finish the port to AIX.
3640 autoconf was my problem (segfault within the malloc replacement
3641 function. No idea why)
3643 2004-09-16 Martin Quinson
3644 - Fix some size_t madness on 64bit architectures.
3646 2004-09-08 Martin Quinson
3647 - Reduce the number of system headers loaded, overload some more system
3648 calls (such as malloc to cast the result of the system one, and work
3650 - Fix and reintroduce the config support
3652 2004-09-07 Martin Quinson
3653 - Source code reorganization to allow Arnaud to surf all over there.
3654 - Allow to document the logging categories.
3655 - Remove all uppercase from logging categories and useless cleanup in names.
3657 2004-08-18 Martin Quinson
3658 Version 0.6.2 (protocol not changed; API changed)
3659 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3660 gras_error_t), and NULL when not found. Functions expecting a msgtype
3661 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3662 hopefully usefull message.
3663 - Portability to prehistoric sparcs again
3665 2004-08-17 Martin Quinson
3666 Version 0.6.1 (protocol not changed; ABI not changed)
3667 - prealloc some buffers to speed things up
3669 2004-08-11 Martin Quinson
3670 Version 0.6 (protocol not changed; ABI expended)
3671 - The parsing macro can deal with the references, provided that you add
3672 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3674 2004-08-09 Martin Quinson
3675 Version 0.5 (protocol not changed; ABI changed)
3676 - Allow to off turn the cycle detection code in data exchange at
3677 compilation time. It should be at run time, but I'm short of time (and
3678 the config stuff is still broken). That way, we keep dict out of the
3679 critical path, which is good because the performance is poor:
3680 - search not dichotomial yet
3681 - dynar give no way to access their content and memcpy everytime
3682 - In composed data description (struct, ref and so on), stop foolness of
3683 keeping the subtype's ID, but store the type itself. This keeps sets out
3684 of the critical path, which is good since they rely on dynar and
3685 dictionnaries. The only loose of that is that we cannot detect the
3686 redeclaration of a structure/union with another content (but I'm not sure
3687 the code detected well this error before anyway). We still can detect
3688 the redefinition discrepancy for the other types.
3689 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3690 it breaks the code because of type-punning used all over the place).
3691 This breaks on all non-gcc architectures (for now).
3693 All those changes (plus the buffer of last time) allow me to gain 2 order
3694 of magnitude on cruel tests consisting of 800000 array of integers on two
3695 level of a hierarchical structure (200 secondes -> 4 secondes)
3698 - the selector of reference must now return the type it points to, not
3699 the ID of this type.
3701 2004-08-06 Martin Quinson
3702 Version 0.4 (protocol changed; ABI not changed)
3703 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3704 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3705 - Implement a buffer transport, and use it by default (it relies on tcp in
3706 real life and on sg in simulation).
3707 That's a bit hackish since I had a new field to the structure to store
3708 its data without interfering with the subtype ones. Inheritance
3709 is tricky in C. And that's a kind of reverse inheritance with one class
3710 derivating two classes. Or maybe a game with java interfaces. Anyway,
3711 that's damn hard in C (at least).
3712 Moreover, I got tired while trying to ensure plugin separation and
3713 genericity in SG mode. MSG wants me to do weird things, so let's go for
3714 cruel hacks (temporarily of course ;).
3715 See comment in transport_private.h:71
3716 - do not include all the _interface headers in private but in the files
3717 which really need them (to cut the compilation time when they are
3720 2004-07-26 Martin Quinson
3721 Version 0.3 (protocol not changed; ABI changed)
3722 - Major overhault of the datadesc interface to simplify it:
3723 - shorted the function names:
3724 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3725 - add a trivial way to push/pop integers into the cbps without malloc.
3726 This allows to make really generic sub_type description, which simply
3727 pop their size of the stack.
3728 - add a function gras_datadesc_ref_pop_arr() which does what users want
3729 most of the time: Declare a dynamic array (which pops its size of the
3730 stack) and declare a reference to it. Poor name, but anyway.
3731 - kill the post-send callback, add a post-receive one
3733 2004-07-23 Martin Quinson
3734 Version 0.2 (protocol changed; ABI changed)
3735 - add some testing for cpbs in the test cases, and fix some more bugs.
3736 This invalidate again the little64 data file, since I cannot regenerate
3738 - remove an awfull optimization in the logging stuff, allowing me to:
3739 - understand it again
3740 - learn gcc how to check that the argument match the provided format
3741 - fix all errors revealed by gcc after that
3742 - internal keys of dict are not \0 terminated. Deal with it properly in
3743 loggings instead of segfaulting when the user want to see the logs :-/
3745 2004-07-22 Martin Quinson
3746 - Fix some stupid bug preventing cbps (callback postit) from working
3748 2004-07-21 Martin Quinson
3749 - Some documentation cleanups
3750 - remove the useless last argument of msgtype_declare
3751 - rename the Virtu functions to fit into the 'os' namespace
3752 - move headers src/include -> src/include/gras/ and stop fooling with
3753 gras -> . symbolic link
3754 - make distcheck is now successful
3756 2004-07-19 Martin Quinson
3758 - Build shared library also
3759 - Install html doc to the right location
3760 - stop removing maintainer files in make clean
3761 - build tests only on make check
3763 2004-07-13 Martin Quinson
3765 - No major issue in previous version => change versioning schema
3766 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3767 corresponding dataset.
3769 2004-07-11 Martin Quinson
3771 - Get it working with any kind of structure (we can compute the padding
3772 bytes remotely for all the architectures I have access to)
3773 - Implement the structure parsing macro (still not quite robust/complete)
3774 - Improvement to the remote testing toysuite
3776 2004-07-10 Martin Quinson
3777 [autoconf mechanism]
3778 - get ride of a bunch of deprecated macros
3779 - actually run the test for two-compliment, not only compile it :-/
3780 - test whether the structures get packed (and bail out if yes. Damn.
3781 Alignment is a serious matter)
3782 - test whether the structures get compacted (but respecting the alignment
3783 constraints of each types)
3784 - test whether the array fields of structures can straddle alignment boundaries
3786 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3787 (took me 2 hours to find that bug, looking at the wrong place)
3788 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3789 before sending/receiving and are used to compute the offsets of fields
3790 - Given that padding size depend even on compiler options, keep track of
3791 alignment and aligned_size only for the current architecture. Not a big
3792 deal since we send structure fields one after the other (seems
3794 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3795 program, let it work on linux/gcc/little32. portability todo.
3797 2004-07-08 Martin Quinson
3798 - import and improve remote compilation support from FAST
3799 - make sure make check works on half a dozen of machines out there
3801 2004-07-07 Martin Quinson
3802 Let's say it's version 0.0.3 ;)
3803 - Implement conversions (yuhu!)
3804 - Let it work on solaris (beside conversion, of course)
3805 - Stupid me, using rand() to generate the conversion datatests in not wise.
3807 2004-07-06 Martin Quinson
3808 - Let make dist work, since I'm gonna need it to compile on remote hosts
3809 - Let Tests/datadesc_usage write the architecture on which the file was
3810 generated as first byte.
3811 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3812 architecture descriptions.
3813 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3814 on those architectures.
3815 - Optimization: send/recv array of scalar in one shoot
3817 2004-07-05 Martin Quinson
3818 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3820 - Plug a whole bunch of memleaks
3821 - each process now have to call gras_{init,exit}. One day, their log
3822 settings will be separated
3823 - Continue the code factorisation between SG, RL and common in Transport.
3825 2004-07-04 Martin Quinson
3827 - Redistribution between SG and RL.
3828 We wanna have to accept in SG, so move accepted related parts of RL in
3829 the common part. (more precisely, the dynar of all known sockets is no
3830 more a static in transport.c, but part of the process_data)
3832 [gras_stub_generator]
3833 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3835 2004-07-03 Martin Quinson
3836 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3837 them all up in one shot)
3839 - Fix the ugly case of reference to dynamic array.
3840 - New (semi-public) function gras_datadesc_size to allow the messaging
3841 layer to malloc the needed space for the buffer.
3843 - gras_socket_close now expect the socket to close (and not its address to
3844 put NULL in it after it). This is because the socket passed to handlers
3845 is one of their argument (=> not writable).
3847 - propagate the interface cleanup from last week in datadesc, ie remove a
3848 superfluous level of indirection. User pass adress of variable
3849 containing data (both when sending and receiving), and not of a variable
3850 being a pointer to the data. Let's say that I like it better ;)
3851 The price for that is constructs like "int msg=*(int*)payload" in
3852 handlers, but it's a fine price, IMHO.
3854 - Let it work in RL (yuhu)
3856 2004-06-21 Martin Quinson
3858 - porting SG plugin and SG select to new standards (works almost).
3859 - plug memleaks and fix bugs around.
3862 - cleanup the prototype of data recv and force users to specify when they
3863 want to handle references to objects. Test case working even for cycles.
3864 - plug memleaks. Valgrind is perfectly ok with this.
3866 2004-06-12 Martin Quinson
3868 - cleanup the separation between plugin and main code in plugin creation
3870 2004-06-11 Martin Quinson
3872 - Reput hook for raw sockets, needed for BW experiments
3873 - kill a few lines of dead code
3874 [Data description] Interface cleanup
3875 - gras_datadesc_by_name returns the searched type or NULL.
3876 That way, no variable is needed to use a type desc once, which makes
3878 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3879 two parameters were strings (field name, type name), leading to
3881 [Dicos] Interface cleanup
3882 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3883 This is consistant with the dynar API.
3885 2004-04-21 Martin Quinson
3887 - Porting to new standards.
3889 - interface cleanup.
3890 There is no bag anymore, no need to take extra provision to mask the
3891 pointers behind "ID".
3892 Better splitup of functions between files create/exchange/convert.
3893 This is still a bit artificial since convert and receive are so
3894 interleaved, but anyway.
3896 - add a queued message list to procdata (the ones not matching criteria
3898 - factorize some more code between SG and RL wrt procdata
3900 - use gras_exit in example to track memleaks
3901 - get rid of gs_example now that GS is properly integrated into gras
3902 - update run_test to integrate the latest tests (datadesc)
3904 - rename WARNINGn macros to WARNn since it prooved error-prone
3906 2004-04-19 Martin Quinson
3908 - register init/exit functions within gras module mechanism
3909 - send/receive function.
3910 Convertion is not implemented, but short-cutted if not needed.
3911 struct/array elements are sent one by one (instead of block-wise), but
3912 nobody really cares (yet). Get a prototype before optimizing.
3913 - tests (using a file socket) for DD send/receive on:
3914 - base types: int, float
3915 - array: fixed size, string (ie ref to dynamic string)
3916 - structure: homogeneous, heterogeneous
3917 - chained list, graph with cycle
3918 Believe it or not, valgrind is not too unhappy with the results. The
3919 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3923 - Bugfix when using a filename explicitely (instead of '-')
3925 2004-04-09 Martin Quinson
3927 - factorize more code between RL and SG in socket creation
3928 - Complete the implementation and tests of:
3930 o file (only in RL, and mainly for debugging)
3932 I lost 3 days to design a portable address resolver, and then decided
3933 that the prototype mainly have to run on my box.
3934 Addressing portability too early may be like optimizing too early :-/
3936 - use gras_init in the Tests instead of the crappy parse_log_opt
3937 (the latter function is removed)
3938 [Conditional execution]
3939 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3940 [Code reorganisation]
3941 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3942 Build examples against the RL library, since there is no way to disable
3943 its creation for now.
3945 For information, the beginning of coding on GRAS was back in june
3946 2003. I guess that every line has been rewritten at least twice since