1 ----------------------------------------------------------------------------
3 SimGrid (3.23.3) NOT RELEASED YET (v3.24 expected September 23. 7:50 UTC)
6 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
7 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
8 thus becomes the default resource to execute I/O operations. Examples
9 and the file system plugin have been modified to use disks instead of
10 storage elements. Storage can still be used but is doomed to disappear.
11 Users are thus adviced to update their simulators accordingly.
12 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
13 for consistency with pthread_barrier_wait()
14 - Host::get_englobing_zone() returns the englobing netzone
15 - Actor::on_destruction is now called in the destructor
16 Actor::on_termination new signal called when the actor terminates
18 - Global signals are now part of the Engine:
19 - on_platform_creation: after config settings, before the XML parsing
20 - on_platform_created: right after the XML parsing
21 - on_time_advance: each time the clock advances
22 - on_simulation_end: after simulation, before cleanups
23 - on_deadlock: as the name implies.
25 - sg_{actor,host,link}_{data,data_set}() now all exist.
26 Use them to attach user data to the object and retrieve it.
29 - Introduce an experimental Wifi model. It sounds reasonable
30 according to the state of the art, but it still has to be properly
31 validated, at least against ns-3 if not against reality.
32 - Improved the usability of ns-3. Several bugs were ironed out.
33 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
34 It is now expressed as 'idle:epsilon:allCores' to properly model the
35 consumption of non-whole tasks on mono-core hosts. More info in the doc.
38 - convert a new set of functions to the S4U C interface and move the old MSG
39 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
42 - Fortran bindings for DVFS have been removed.
43 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
44 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
45 - SMPI can now be selected by cmake's find_module(MPI) with
46 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
47 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
48 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
49 - trace-call-location can now be used with TI traces, and replayed.
50 - Default for trace-call-location is now to use file names and not full paths.
51 To revert to previous behaviour (in case of collision of filenames), option
52 "smpi/trace-call-use-absolute-path" can be set to yes.
55 - Use the included xxHash as an hash implem when C++14 is usable.
56 - Option model-checker/hash was removed. This is always activated now.
57 - New option smpi/buffering controls the MPI buffering in MC mode.
58 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
59 This is useful to check for MPI compliance.
62 - New section on plugins: how to define a new one, and existing ones.
65 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
66 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
69 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
70 and <mount> tags. The rationale is finish to decouple what should be handle
71 by SimGrid kernel (i.e., resources and activities) from plugins built in
72 user space (e.g., the file system plugin). This new tag currently accepts
73 only three attributes (id, read_bw, and write_bw). All the other information
74 that was declared with the storage related tags now has to be expressed as
75 properties. An example of platform using this new tag is available at
76 examples/platforms/hostsè_with_disks.xml
78 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
79 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
80 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
81 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
82 - FG#31: per-actor data
83 - FG#34: SG_BARRIER_SERIAL_THREAD?
84 - FG#35: model-checker does not like buster-produced binaries
85 - FG!13: MC: complete workaround in the error msg seen on modern systems
86 - FG!15: execute_flops now logs compute
87 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
88 - FG!17: ns-3: unblock the right number of communications + others issues
89 - FG!18: Improving the performance of the ns-3 bindings
90 - GH#219: Error in the throughput of TCP transfer
91 - GH!330: Minor polishing of English: Configuring_Simgrid
92 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
93 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
95 ----------------------------------------------------------------------------
97 SimGrid (3.23.2) July 8. 2019
100 - Nicer introduction page.
101 - Migrate the "Deploy your application" page to the new doc.
102 - Move Java as a subtree of MSG.
105 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
108 - Drop xbt_dynar_sort_strings().
111 - Really fix FG#26: Turning off a link should raise NetworkFailureException
112 - FG#27: Wrong exception thrown to wait_any when link is turned off
113 - GH#328: Java: Canceling multiple tasks in a single vm/host
115 ----------------------------------------------------------------------------
117 SimGrid (3.23) June 25. 2019
119 The Exotic Solstice Release.
122 - SunOS and Haiku OS support. Because exotic platforms are fun.
123 - Stop setting random seed with srand48() at initialization.
124 - Use addr2line as a fallback for stacktraces when backtrace is not available.
125 - Build option -Denable_documentation is now OFF by default.
126 - Network model 'NS3' was renamed into 'ns-3'.
129 - Simgrid can now hopefully be installed with pip.
132 - wait_any can now be used for asynchronous executions too.
135 - New log appenders: stdout and stderr. Use stdout for xbt_help.
136 - Drop xbt_dict_dump.
139 - 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
140 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
141 - MPI interface is now const correct for input parameters
142 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
143 - Fortran bindings for async collectives.
144 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
147 - Remove option 'model-check/record': Paths are recorded in any cases now.
148 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
149 always sparse. That's an immense gain in memory, and can even be
150 faster because of cache effects. So there is no need to clutter the
151 code to allow the user to go for the unefficient mode.
154 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
155 models proved to be more accurate than these old experiments.
157 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
158 - FG#1: Broken link in error messages
159 - FG#2: missing installation documentation
160 - FG#3: missing documentation in smpirun
161 - FG#6: Python bindings not available on PyPI
162 - FG#7: simple cmake call requires doxygen
163 - FG#8: make python bindings an optional dependency
164 - FG#10: Can not use MSG_process_set_data from SMPI any more
165 - FG#11: Auto-restart actors forget their on_exit behavior
166 - FG#12: -Denable_lto=OFF doesn't disable LTO
167 - FG#13: Installs unstripped file 'bin/graphicator'
168 - FG#14: Installs the empty directory 'doc/simgrid/html'
169 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
170 - FG#17: Dead link in doc (pls_ns3)
171 - FG#20: 'tesh --help' should return 0
172 - FG#21: Documentation link on http://simgrid.org/ broken
173 - FG#22: Debian installation instruction are broken
174 - FG#26: Turning off a link should raise NetworkFailureException exceptions
175 - GH#133: Java: a process can run on a VM even if its host is off
176 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
177 - GH#326: Valgrind-detected error for join() when energy plugin is activated
179 ----------------------------------------------------------------------------
181 SimGrid (3.22) April 1. 2019
183 The Easter Christmas Release (this one is somewhat late).
186 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
187 - Not complete yet: asynchronous activities (amongst others) are still missing
188 - Still ongoing: the interface may change in the future. We need more
189 testers! Please report any glitches.
190 - No new project using Java should start now. Please switch to Python.
193 - Some of the internal cleanups may lead to speed improvements:
194 - The hard limitation on the amount of simulated actors with
195 Java+Mac was removed. Now, the available memory is the only limit.
196 - Our refcounting was tidyied, leading to 10% speedups in some cases.
197 - We are still working on making our code robust to the actor kills
198 and hosts' churn. Things are improving, but it's not perfect yet.
199 - Replaced our own code to display a backtrace (that was forking addr2line)
200 with the Boost.Stacktrace library.
201 You won't see your backtraces without this optional dependency.
202 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
203 - Stop setting random seed with srand() at initialization.
206 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
207 XML file version remains 4.2 since old files are still compatible.
210 - Process termination which was broken at version 3.21 has been repaired.
211 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
212 - Hide the examples into examples/deprecated. New users should use Python.
215 - Drop MSG_process_create_from_stdfunc() from the API.
216 This C++ function was a pimple in the C API, made necessary at some
217 point by the Java bindings. This is fixed now.
218 - Hide the examples into examples/deprecated. New users should use S4U.
219 - MSG_process_create and MSG_process_attach now crash if the host used
220 to run this process is off.
221 - Fix the protype of MSG_process_on_exit()
222 Now use (int,void*) callbacks instead of (void*,void*) ones.
223 The implementation was ways too messy to actually work, I guess.
226 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
227 too often and provide more accurate timings.
228 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
229 - MPI_Alltoallw support
230 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
231 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
232 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
233 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
235 - Basic implementation of generalized requests (SMPI doesn't
236 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
239 - Drop sg_cmdline. Please use xbt_cmdline instead.
240 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
241 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
242 OsSemaphore is implemented in a portable way with C++11 threads.
243 This should allow much more threads to be created at the same time,
244 allowing Mac and Java users to simulate many more actors.
245 - Implement the 'thread' factory with std::thread instead of xbt ones.
246 It is not possible to set the stack size with threads anymore, but
247 -fsplit-stack is the way to go nowadays when using threads.
248 - Drop the xbt_os_thread_t module (now unused)
249 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
250 - Drop xbt_str_join_array().
251 - Drop cunit, use Catch2 instead.
254 - Many cleanups in the kernel::activity namespace. This was long
255 overdue, and shall open the path to many future endeavors.
258 - #132: Java: a process can not shut down its own host
259 - #220: S4U: detached send causes exception if sender terminates soon after sending
260 - #261: Document the parameters of parallel execution's constructor
261 - #300: [s4u] BarrierPtr is missing
262 - #314: SMPI args internal cleanup
263 - #316: Fix a bug related to the CPU utilization of multi-core VM
264 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
265 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
266 - #325: Turning off a host has different behavior on sleeping actors and computing actors
268 ----------------------------------------------------------------------------
270 SimGrid (3.21) October 3. 2018
272 The Restarting Documentation (TRD) Release.
275 - Convert the user manual to Sphinx for improved usability.
276 Unstable version now hosted on https://simgrid.frama.io/simgrid
277 Many glitches remain; stay tuned.
278 - Restrict the Doxygen reference API to the installed header files.
279 (documenting all internals this way is overkill + doxygen scales badly)
280 - New tutorial on S4U (highly inspired from the old MSG one, but with
281 a git repository to fork as a starting point, and a docker image)
282 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
283 - The Developper manual is still to be converted (not compiled until then)
284 - Some parts are still missing in this conversion (such as the
285 platform and deployment sections) while others were blindly converted
286 and would need more love (such as the configuration flags).
287 Things will certainly further improve in the future releases.
290 - Fully reimplement auto_restart mechanism that was utterly broken
291 (fix #22, #131 and #281 that were all very related)
292 - Implement semaphores (fix #303 and #291)
293 - When creating an actor from a function and its parameters,
294 move-only parameters are not allowed anymore, as it would prevent
295 the actor to be restartable if its parameters are consumed on run.
296 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
297 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
298 IO operations that can be started, waited for, or canceled as a regular
300 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
304 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
305 - New host variable: 'core_count'
308 - Remove xbt_os_thread_specific features
309 - Remove portability wrapper to condition variables
310 - Remove xbt_os_thread_yield()
313 - MPICH collective selector now mimics MPICH 3.3b
314 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
315 - MPI_Init(NULL,NULL) is now allowed
316 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
317 - Command-line arguments should now be handled properly in Fortran simulations
318 - MPI Attributes and related callbacks should work in Fortran
319 - Apps using C + Fortran should now work
320 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
321 - Support MPI_Error_String
324 - Due to an internal bug, Msg.run() must now be your last line.
325 We hope to fix it in a future release, and we are sorry for the inconvenience.
328 - #22: Process autorestart seem to only work with CAS01 cpus
329 - #56: Feature request: dynamic SMPI replay
330 - #93: simgrid should not eat --help
331 - #111: How to change the loopback link characteristics?
332 - #116: Platform cluster radical
333 - #117: Platform cluster/host order
334 - #129: Replay_multiple on steroid
335 - #131: Java : autostart does not work
336 - #143: Setting a breakpoint at a given time
337 - #235: Network energy models should be integrated
338 - #264: Add ptask L07 resource tracing
339 - #271: Dynamic and manual replay of SMPI traces
340 - #279: Breakpoints option uses time deltas instead of absolute time
341 - #281: Daemonized actor no longer daemonized after an auto-restart
342 - #282: TIT convertor erase traces when using absolute path in the trace list file
343 - #285: segfault when a s4u actor kills itself with msg process tracing activated
344 - #286: Pajé traces are not exposing the number of cores
345 - #287: Command-line parsing should stop when encountering '--'
346 - #288: MPI_Init(NULL, NULL)
347 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
348 - #290: Method to check if a mailbox has a message ready to be consumed
349 - #291: [s4u] semaphores do not seem available yet
350 - #294: Alltoall collective implementation of mpich
351 - #295: Starting a Host as OFF is not supported
352 - #296: DTD too permissive
353 - #299: Add a s4u barrier example
355 ----------------------------------------------------------------------------
357 SimGrid (3.20) Released June 24. 2018
359 The proxy snake_case() release.
362 - The whole API is now using snake_case() instead of mixing it with camelCase.
363 Compatibility wrappers in place for one year.
364 - We took the opportunity to increase the consistency of the API.
365 Please report any remaining glitches.
366 - New function simgrid::s4u::Host::get_actor_count:
367 Returns the number of actors running on a specific host.
370 - SMPI is now tested with ~45 proxy apps from various sources, with none or
371 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
372 - Replay: The replay file has been re-written in C++.
373 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
374 supported. They are stored in the trace and used when replayed.
375 - Basic support of MPI_Cancel. Robustness not guaranteed.
376 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
377 (only for MPI_COMM_TYPE_SHARED).
378 - New option: smpi/privatize-libs, to add external shared libs to be privatized
379 by SMPI. They will be copied locally and loaded separately by each process.
380 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
381 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
382 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
383 with internal index optind with multiple processes. Only works if MPI_Init has
385 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
386 syntax without preprocessing tricks.
389 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
390 behavior tracing is no longer limited to MSG
391 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
392 this means that traces now need to be updated or re-obtained; both Irecv
393 and Isend lines in your traces have now in total 5 mandatory fields:
394 <rankid> <command> <to/from rankid> <tag> <size>
395 To update your traces, it suffices to add a 0 for the tag here.
396 - TIT now also supports waiting for a distinct request via MPI_Wait.
397 Wait/Test now wait for a specific request, not just the last one that was
398 issued. This unfortunately means another update, because we need to
399 identify which request you want to wait for. We do this via the
400 triplet (sender, receiver, tag), which needs to be added:
401 <rankid> <command> <sender> <receiver> <tag>
402 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
403 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
406 - The deprecation of MSG is ongoing (but this should not impact you).
407 Many MSG functions are now simple wrappers on the C API of S4U. If
408 you wish to convert your code to S4U, find the S4U counterparts of
409 your MSG calls in src/msg/msg_legacy.cpp.
410 - MSG can still be used, but won't evolve anymore.
413 - Allow to run the Link energy plugin from the command line with
414 --cfg=plugin:link_energy
415 - Rename Energy plugin into host_energy
416 - Rename Load plugin into host_load
419 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
420 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
421 - Improve performance when handling timeouts of simix synchros.
424 - Config: the C API is now deprecated (will be removed in 3.23), and
425 the C++ API has been slightly improved.
428 - Fix several build issues on OSX.
429 - Move simgrid_config.h to simgrid/config.h (old header still working)
432 - #143: Setting a breakpoint at a given time
433 - #150: Inconsistent event names in SMPI replay
434 - #258: daemonized actors hang after all non-daemonized actors have completed
435 - #267: Linker error on unit_tmgr
436 - #269: SMPI: tracing of MPI_Wait/all/any broken
437 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
439 ----------------------------------------------------------------------------
441 SimGrid (3.19.1) Released March 22 2018
443 The Fixed ABI Release.
445 As you may know, we are currently refactoring SimGrid in deep.
446 Upcoming SimGrid4 will be really different from SimGrid3: modular,
447 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
449 Our goal is to smooth this transition, with backward compatibility
450 and automatic update paths, while still progressing toward SimGrid4.
452 SimGrid remains open during works: The last pure SimGrid3 release was
453 v3.12 while all subsequent versions are usable alpha versions of
454 SimGrid4: Existing interfaces remain unchanged, but the new S4U
455 interface is budding and the internals are deeply reorganized.
457 Since 2015, we work hard to reduce the changes to public APIs. When
458 we need to rename a public library symbol in S4U, we let your
459 compiler issue an explicative warning when you use the deprecated
460 function. These messages remain for four releases, i.e. for one full
461 year, before turning into an error. Starting with v3.15, your can
462 also adapt to API changes with the SIMGRID_VERSION macro, that is
463 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
465 Starting with this v3.19.1, our commitment to reduce the changes to
466 the public interfaces is extended from the API to the ABI: a program
467 using only MSG or SimDag and compiled against a given version of
468 simgrid can probably be used with a later version of SimGrid without
469 recompilation. We will do our best... but don't expect too much of
470 it, that's a really difficult goal during such profund refactoring.
472 The difference between v3.19 and v3.19.1 is that the former was
473 accidentally breaking the ABI of MSG, while the later is restoring
476 S4U and kernel APIs will still evolve until SimGrid4, with one-year
477 deprecation warnings as currently. In fact, cleaning up these
478 interfaces and converting them to snake_case() is one release goal of
479 v3.20. But don't worry, we are working to smooth this upgrade path.
481 In summary, new projects should start with S4U to benefit of the
482 future, but old MSG projects should still be usable with no change.
484 ----------------------------------------------------------------------------
486 SimGrid (3.19) Released March 20 2018
488 The Moscovitly-cold Spring release.
490 SMPI user-visible improvements:
491 - Support of MPI_Type_create_subarray
492 - Unimplemented MPI calls now abort (instead of issuing a warning)
493 Some harmless calls around errhandlers still only issue a warning.
494 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
495 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
496 is now possible (mmap privatization was refactored to this end).
497 Adding new SMPI instances should also be technically doable now.
499 SMPI internal cleanups:
500 - Removed index notion from SMPI -- no more "getPid() - 1"!
501 In tracing files, actors are called rank-<PROCESS-ID> but these are
502 NOT the real ranks (tracing several communicators is not supported yet).
503 - Deprecate smpi_process_index() (will be removed in 3.22)
504 - Better testing on classical proxy apps, through an external project:
505 https://github.com/simgrid/SMPI-proxy-apps/
508 - Use a graphical TOC to make it easier to find the documentation you need
509 - Revamp the MSG tutorial
512 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
513 - Remove parameter of MSG_process_killall().
514 Resetting the PID was bogus anyway (several actors could have the same PID).
517 - Execution->setHost() can be called after start(), to migrate it.
518 - Comm::test_any() is now implemented.
519 - s4u::Actor now has onCreation() and onDestruction() signals.
520 - Install some kernel header files for the users' plugins and more.
521 Warning, their API is really not stable yet.
522 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
523 include/simgrid/kernel/routing/*Zone.hpp
526 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
528 Build System and other cleanups:
529 - Remove unused run-time parameter "tracing/onelink-only".
530 - SimGrid now works with intel compilers (MC still buggy)
531 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
532 - tesh kills subprocesses on timeout (but not on windows)
535 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
538 - #194: Feature request: simgrid::s4u::Comm::test_any()
539 - #245: migrating an actor does not migrate its execution
540 - #253: Feature Request: expose clusters as objects
541 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
542 - #255: Tesh broken on Windows
543 - #256: Modernize FindSimGrid.cmake
544 - #257: Fix (ab)use of CMake install
546 ----------------------------------------------------------------------------
548 SimGrid (3.18) Released December 24 2017
550 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
552 MSG IS NOW DEPRECATED. Please use S4U for new projects.
553 - Support for MSG should not be removed from SimGrid before 2020,
554 but future evolutions will be limited to the integration of
555 user-provided patches.
558 - s4u::onDeadlock() signal, run before stopping the simulation.
559 - s4u::Exec: asynchronous executions (abstraction of background
561 - s4u::Host->getLoad() returns the achieved speed in flops/s
564 - Link::name() is deprecated and will be removed in v3.21, use
565 Link::getCname() instead.
566 - Mailbox::getName() changed to return a std::string, use
567 Mailbox::getCname() instead to get a char*.
568 - Storage::getName() changed to return a std::string, use
569 Storage::getCname() instead to get a char*.
570 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
571 to have both allocation and deallocation of the map in user space.
574 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
575 The amount of flops remaining to do is a vector, not a scalar.
576 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
577 name implies on both sequential and parallel tasks.
578 - Both changes fix GitHub's #223 using PR #237 as a basis.
579 Thanks Michael Mercier.
580 - Most examples were converted to S4U and hidden elsewhere at tests.
581 Rationale: we still want MSG to work; we want newcomers to use S4U.
584 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
585 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
588 - Switch to the faster dlopen privatization mechanism by default
589 - Documentation improvments
592 - Live migration is getting moved to a plugin. Dirty page tracking is
593 the first part of this plugin. This imply that VM migration is now
594 only possible if one this function is called:
595 - C/MSG: MSG_vm_live_migration_plugin_init()
596 - C/C++: sg_vm_live_migration_plugin_init()
597 - Java: Msg.liveMigrationInit()
598 For C and C++, "simgrid/plugins/live_migration.h" also has to be
602 - Define class simgrid::xbt::Path to manage file names.
603 - Removed unused functions:
604 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
605 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
606 xbt_graph_export_graphviz()
607 - xbt/str.h: xbt_str_join()
608 - Remove unused datatypes:
609 - xbt/heap.h: use std::priority_queue or boost::heap instead
610 - xbt/swag.h: use boost::intrusive::list instead
613 - New link_energy plugin for the consumption of the links.
614 - All of the operations on files and storage contents have been
615 packaged into a plugin (src/plugins/file_system). The current
616 public interface can be found in
617 include/simgrid/plugins/file_system.h
618 To use these functions you now have to initialize the plugin by
619 calling MSG_storage_file_system_init() just after calling
620 MSG_init() or sg_storage_file_system_init() just after creating
624 - Remove the undocumented/untested tag <include>
627 - Remove viva specific tracing as the tool is no longer maintained
630 - #248: Kill of finished processes leads to segfault
631 - #240: xbt_cond_wait_timeout should gracefully return for C apps
632 - #239: please implement signal s4u::onDeadlock()
633 - #230: segfaults when exit() without run()
634 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
635 - #223: MSG_task_get_flops_amount() not working with parallel tasks
636 - #222: Actor::kill() doesn't really kill and segfaults
637 - #221: odd LMM warning when killing an actor
638 - #120: Memory leak when the processes are forcefully killed
640 ----------------------------------------------------------------------------
642 SimGrid (3.17) Released October 8 2017
644 The Drained Leaks release: (almost) no known leaks despite the tests.
646 Even very long simulations will preserve your memory: our testsuite
647 stresses SimGrid for over 45mn on fast machines for a coverage of
648 over 80%, and there is only one single known leak, of about 4kb.
652 - Comm.detach(): start and forget about asynchronous emission. A cleanup
653 handler may be given to free resources if the comm cannot be completed.
654 - this_actor::send(mailbox) is now mailbox->put()
655 - New: simgrid::s4u::Comm::wait_all()
656 - New: Host.execute() for remote executions.
659 - Unused option network/sender-gap is removed.
662 - Deprecate MSG_task_isend_with_matching(): This unused feature
663 really complicates our internals. Will be removed in v3.20.
666 - Improved context termination. It is now possible to free resources, even
667 when a process is forcibly killed.
668 - ContextBoost: add support for Boost versions above 1.61.
671 - Bring back run-time option --cfg=exception/cutpath to remove exception
673 - Removed unused functions:
674 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
675 xbt_str_rtrim(), xbt_str_trim().
676 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
679 - Removed header files obsolete since SimGrid 3.12:
680 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
682 - #3: SD_exit should be made optional
683 - #120: Memory leak when the processes are forcefully killed
684 - #159: Threading test regression in Actor refcounting
685 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
686 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
687 - #186: Actor::killAll() segfaults if some process is blocked on wait()
688 - #191: VM migration and pstate
689 - #192: Updating the energy consumptions of all hosts crashes with VMs
690 - #195: All actors have PID=0 in the logs
691 - #204: Sometimes segfault with thread contexts and mmap privatization
692 - #222: Actor::kill() doesn't really kill and segfaults
693 - #225: Actor::kill() doesn't really kill when victims are doing a join()
695 ----------------------------------------------------------------------------
697 SimGrid (3.16) Released June 22. 2017.
699 The Blooming Spring Release: developments are budding.
702 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
703 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
704 - Java JAR file should be finally fully working on Mac OSX too.
706 XML platforms: Switch to platform v4.1 format.
707 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
708 - <zone> can be used as a synonym for the now deprecated <as>
709 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
710 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
711 - <actor> can be used as a synonym for the now deprecated <process>
712 - state_file and avail_file periodicity is now easier to express
713 (check the documentation) the old behavior should still work.
714 * Storage: not backward compatible, but it seems that nobody ever used it
715 - Remove attribute 'content_type' of <storage_type>: was never used
716 - Make attribute 'model' of <storage_type> optional: for future usage
717 - Remove Bconnection model property: was never (in)validated. Replaced by
718 the maximum of the read and write bandwidth as a resource constraint.
721 - New and Backwards Compatibility break:
722 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
723 When one wants to get the list of tasks whose states have changed during a
724 simulation round, s/he has to allocate and free a dynar and use it as
725 argument to this function. The former SD_simulate (double how_long)
729 - Allow multicore VMs, with the correct sharing (unless you overcommit)
730 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
733 - New function to update the consumption of all hosts at once.
734 - Fix the model for multi-core hosts, linear in the amount of busy
735 cores with an abnormality for pIdle (see paper).
736 BUG: Still not satisfactory for ptask on mono-cores.
739 - The netzone are now available from the MSG API.
740 The old names still work, but are now deprecated.
743 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
744 - New option: smpi/keep-temps to not cleanup temp files
745 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
746 mallocs mode is used, allows to change the size of the fake file used
747 (default 1MB), to potentially limit the number of mappings for large runs.
748 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
749 - Fortran ifort and flang compilers support
750 - New RMA calls supported (experimental):
751 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
752 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
753 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
754 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
755 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
756 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
759 - New: this_actor::isMaestro() returns whether we are in kernel mode.
760 - Behind the scene, ActivityImpl are now automatically refcounted.
761 Removing this nasty bottleneck will greatly simplify our internals.
764 - Replay: New function xbt_replay_action_get():
765 Retrieve the function previously associated to an event type.
766 - DROPPED FUNCTION: xbt_str_varsubst()
767 - DROPPED MODULE: strbuff. We don't need it anymore.
768 - DROPPED MODULE: matrix. We don't need it anymore.
769 - DROPPED MODULE: lib. We don't need it anymore.
771 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
773 SimGrid (3.15) Released March 22. 2017
775 The Spring Release: continuous integration servers become green
776 We fixed even the transient bugs on all target architectures:
777 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
778 Windows, FreeBSD, NetBSD.
780 - Rename NetCards to NetPoints.
781 This was intended to help NS3 users, but that's not a netcard.
782 That's a point in the routing algorithm, let's avoid wrong simplifications.
785 - New: MSG_process_yield(). Stop and yield to other processes.
786 - New: MSG_process_daemon(). Daemon processes are automatically killed
787 when the last non-daemon process terminates
788 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
789 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
790 to make room for the upcoming network energy plugin.
791 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
794 - Ensure that an actor can kill itself with Process::exit()
795 - Kill the obscure NativeException. Nobody want to survive the issues
796 it denotes, so use JniException that is a RuntimeException (not to
797 be caught explicitly).
798 - Partial bug fix in initialization. SimGrid flags on command line were
799 consumed at C level but stayed in the original Java String[] args.
800 This could mess users' args[i] if SG flags were not put at the end of
802 The SimGrid flags are now removed from the Java arguments. However,
803 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
804 if args.length is greater than the number of YOUR OWN ARGUMENTS.
805 It might be if you have --log or --cfg flags in the command line.
806 - Fix numerous memleaks all around the place. In particular, around VMs.
810 - s4u::onPlatformCreated: right before the simulation starts
811 - s4u::onSimulationEnd: right after the main simulation loop
812 - s4u::onTimeAdvance: right after a clock change (time is discrete)
813 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
814 event from the availability_file changes the avail speed.
815 - Links are now usable from s4u
816 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
817 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
818 - New examples: The conversion of MSG examples to S4U has begun
819 - Actors: create, kill, migration, and suspend
820 - Applications: master-worker and token-ring
821 - Action replay: communications and storage
822 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
825 - Backwards Compatibility breaks
826 - The SD_route_get_list and SD_route_get_size functions have been removed.
827 They are replaced by the unique
828 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
829 The route from src to dst is built in the links dynar whose size can
830 be obtained with xbt_dynar_length.
831 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
832 sg_host_route_bandwidth and sg_host_route_latency.
833 Macros ensure the backwards compatibility, but you should fix your code
836 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
837 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
838 - Bump our claim of support from MPI 1.1 to MPI 2.2.
839 We don't support 100% of it, but it should be enough. Contact us if not.
840 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
841 naive and probably inaccurate)
842 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
843 mapped onto a unique small file using some system magic.
844 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
845 - Fortran: cleanups, fixes, support of user-added operations
846 - RMA: MPI_Accumulate are applied in correct order.
847 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
848 - Better support for MPI_IN_PLACE
849 - Support for MPI_Win attrs and keyvals.
850 - Support MPI_Comm_group_create, MPI_Type_size_x
851 - MPI ops should be properly applied to all allowed types
852 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
853 algorithm by a non failing one from mpich.
856 - Kill the fifo data container: we don't use it anymore.
859 - A <cluster> can now be created with different speed values to represent pstates.
860 see examples/platforms/energy_cluster.xml for an example.
862 SimGrid (3.14.159) Released December 28. 2016
864 The previous release was only a crude approximation of a Pi release;
865 this one should be a bit better.
867 - Revert a last minute change that broke on Mac OSX.
868 - Fix the changelog and NEWS files.
869 - Improve the documentation, in particular of the routing module.
870 - Integrate some patches flying around in Debian.
872 SimGrid (3.14) Released December 24. 2016
875 * The whole documentation was reworked and reorganized. There is
876 still much room for improvement here, but we're on it.
879 * We now need python3 for our testing infrastructure.
880 * Model checking needs libevent
883 * Now works on FreeBSD
886 * Privatization now works on FreeBSD
888 * Privatization is now activated by default in smpirun
889 You can turn it off with -no-privatize if something goes wrong.
891 * Call-location tracing for SMPI.
892 You can add the exact location (filename / linenumber) of an MPI call to
893 your trace files and slow down or speed up the simulation between two
894 consecutive calls by using an adjustment file (see the documentation).
896 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
897 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
898 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
899 values were ignored in some cases. The timings of these functions can now
900 be significantly different.
902 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
903 smpi/running-power is renamed to smpi/host-speed
905 * smpi/grow-injected-times option to enable or disable multiplication of the
906 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
907 make simulation less precise (but also much faster).
909 * smpirun script should be (much) faster for large deployments.
911 * SMPI tracing: fixed issue with poor matching of send/receives.
913 * Replay: Fix broken waitall
915 New functions and features
916 * MSG_parallel_task_execute_with_timeout, to timeout computations.
918 Dropped / renamed functions and features
919 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
920 - MSG_mailbox_is_empty() -> Mailbox::empty()
921 - MSG_mailbox_front() -> Mailbox::front()
922 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
923 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
924 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
925 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
926 * Don't pass the free_f parameter to property related functions:
927 - MSG_host_set_property_value()
928 - MSG_as_router_set_property_value()
929 - MSG_storage_set_property_value()
930 * VM properties. Since msg_vm_t are msg_host_t, just use
931 MSG_host_get_property_value() and friends
932 * VM I/O related things:
933 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
934 - Unimplemented save/restore methods
935 * MSG_as_router_get_property_value() was redundant with
936 MSG_environment_as_get_property_value().
937 - Removed MSG_as_router_*propert*() functions
938 - Added MSG_environment_as_set_property_value() for consistency
939 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
940 Well, they are still there for now, but deprecated with a warning.
941 Please switch to xbt_dict_new_homogeneous() before this is removed
943 * Task affinity. Its intended behavior (that was very badly tested
944 and probably not really working) was deceiving what most users
945 would have hoped here.
946 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
947 * Option network/coordinates is now useless and should be dropped.
950 * Added option storage/max_file_descriptors to allow more than 1024 files opened
953 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
954 * Added Dragonfly topology support
957 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
958 EiBps, EBps, Eibps, Ebps for bandwidth.
959 They may become useful to some lucky ones.
962 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
963 * ex/app_tokenring: new example, very similar to the MSG Token Ring
964 * ex/async_waitAll: new example, on asynchronous communications
967 * Memory usage should be decreased for simulations with a large number
968 of processes. This also helps for SMPI.
970 SimGrid (3.13) stable; urgency=low
972 The Half Release, a.k.a. the Zealous Easter Trim.
974 - We removed half of the lines, that were mostly experimental cruft.
975 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
976 (not counting blanks and comments -- according to openhub.net)
977 - The internals are now compiled in C++ (and will soon be clean C++)
978 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
979 59 klines of C, etc. We added only 29 klines of C++ in replacement.
981 * Backwards Compatibility breaks
982 - Removed Lua simulation bindings (switch to C or Java for that).
983 Lua can still be used to describe platforms
984 - Removed Java kernel plug-ins.
985 Will be reintroduced after the ongoing major internals reorg.
987 - the following functions were removed.
988 They were too specific and should be reimplemented in a generic
989 way, with filter function.
990 - MSG_task_listen_from_host
991 - MSG_mailbox_get_count_host_waiting_tasks
992 - MSG_mailbox_put_with_timeout was removed.
993 Please use MSG_task_send_with_timeout instead.
995 - the SD_application_reinit function was removed. It has been a noop for a while.
996 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
997 replaced by a more flexible mechanism.
998 - The following functions thus do not exist anymore
999 - SD_workstation_get_access_mode
1000 - SD_workstation_set_access_mode
1001 - SD_workstation_get_current_task
1002 - Basic estimation functions have been removed but can easily be replaced
1003 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1004 - SD_workstation_get_computation_time => amount / sg_host_speed()
1006 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1007 Use VM.getSpeed()*load/100 for the legacy behavior.
1009 - option enable_tracing was removed. It was not doing anything for a while.
1010 - In the ModelChecker:
1011 - the model-checker now ptraces the model-checked process which means
1012 you cannot use a debugger on the latter anymore (we might make this
1013 optional in the feature);
1014 - removed soft-dirty page tracking;
1015 - remove model-checked side snapshot management,
1016 MC_snapshot() and MC_compare_snapshot();
1017 - keep the MC_cut() function as a stub (it was not really working
1018 in the previous release).
1021 * All options are consistently in kebab-case. Old names are kept as alias.
1024 * Switch to platform v4 format.
1025 - Rename from 'power' to 'speed' the attributes describing the amount of
1026 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1027 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1028 - In <host> and <link>, attributes availability and state are gone.
1029 It was redundant with state and availability traces, and with peak values.
1030 - In <cluster>, attributes availability_file and state_file are gone.
1031 It was too complex and unused.
1032 - Kill <gpu>. Was not doing anything.
1033 - The DOCTYPE points to the right URL:
1034 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1035 (the file at this address now documents the changelog since its v1)
1036 - A warning is emitted for unit-less values (they are still accepted).
1037 - speed. Default: 'f' or 'flops'. Also defined:
1038 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1039 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1040 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1041 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1042 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1043 - latency. Default: 's' second. Also defined:
1044 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1046 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1047 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1050 * s4u::Host is now the preferred public interface to the Host features.
1051 sg_host_* functions are C bindings to the exact same behavior
1052 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1055 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1056 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1057 They were deprecated since a few years, and probably did not even compile anymore.
1060 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1061 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1062 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1063 encouraged to update their codes. A list of the modified functions can be found at the end of
1064 include/simgrid/simdag.h
1067 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1068 the lambda or closure passed as a parameter will run in kernel mode.
1069 All the callback functions should be rewritten to that interface at some point.
1072 * Reorganizing and cleaning the internals all around the place.
1075 * Remove old default barrier/bcast buggy algorithms (see #18407)
1076 * Various bug fixes to handle more codes
1077 * Remove the need for the --foreground option of smpirun (it is still
1078 accepted for backward compatibility).
1081 * Kill the setset data container: MC don't use it anymore.
1082 * Kill the queue data container: it made more sense with GRAS.
1083 * Kill the xbt_peer_t data type: it's useless without GRAS.
1084 * Kill rm_cb feature of config sets: it was never useful.
1085 * Kill graphxml parsing feature. It was not used.
1086 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1088 - xbt_dynar_sort_strings(), when the content is char*
1089 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1090 They throw exceptions on invalid input;
1091 * C++ support for declaring CLI flags (simgrid::config::Flag);
1092 * class for abstracting different signal backends (simgrid::xbt::signal).
1093 with no external dependencies (we need very simple signals).
1096 * refactoring and cleanup of the code;
1097 * ongoing process to cleanly separate the model-checking algorithms
1098 from the code model-checking support.
1100 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1102 SimGrid (3.12) stable; urgency=low
1104 The Facelift Release.
1107 * Require g++ v4.7 at least to not speak prehistorical C++.
1108 * Require Boost 1.48 (for signal2 component).
1109 * Java must be version 7 at least when activated.
1110 * Builds on Windows again (including Java bindings).
1111 * Tracing is now always enabled (no way to turn it out)
1112 * Remove GTNetS. It was not working anyway.
1113 * Various cleanups in the cmake scripts.
1114 * Move headers around to sort them out on installed systems:
1115 - instr/instr.h -> simgrid/instr.h
1116 - instr/jedule/* -> simgrid/jedule
1117 - simdag/datatypes.h was removed
1118 - simdag/simdag.h -> simgrid/simdag.h
1119 - msg/datatypes.h was removed
1120 - msg/msg.h -> simgrid/msg.h
1123 * Interface improvement:
1124 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1125 - Sanitize the interface in MSG_task_ module:
1126 - Merge two functions that were close enough but misleading:
1127 set_compute_duration(t) -> set_flops_amount(t)
1128 get_remaining_computation(t) -> get_flops_amount(t)
1129 - set_data_size(t) -> set_bytes_amount(t)
1130 get_data_size(t) -> get_bytes_amount(t)
1131 - Massive cleanups in the functions related to the energy
1132 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1133 - New: MSG_host_get_pstate()
1135 - msg/energy/onoff: switching hosts on and off
1137 * Interface improvement:
1138 - Rename Host.isAvail() to Host.isOn()
1139 - Rename Process.currentProcess() to Process.getCurrentProcess()
1140 - Rename Task.setDataSize() to Task.setBytesAmount()
1141 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1143 - #18874: Actually allows the GC to reclaim tasks
1147 - SIMIX_process_throw: raises an exception in a remote process
1148 * Refactoring: Separate sub-modules
1149 - libsmx: the public interface, as libc in a real system
1150 - popping: the strange dance that converts a user request into a kernel handling
1151 - smx_context_*: the virtualization mechanisms that embed the user code
1152 - smx_*: the handling of each simcalls
1153 * Interface cleanup:
1154 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1155 * Rename smx_action_t into smx_synchro_t, making explicit that these
1156 things are used to synchronize processes with their environment.
1157 For example, a communication is a sort of synchronization involving
1158 the communicating processes (that may block until the exchange) and
1159 the platform. The same can be said from computations, etc.
1161 - #18888: segfault when a process ends before its kill_time
1165 - 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.
1166 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1167 - MPI_Type_set_name, MPI_Type_get_name
1168 - MPI_*_c2f and MPI_*_f2c functions
1169 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1170 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1171 - Activate a lot of new tests from the mpich 3 testsuite
1173 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1174 - InfiniBand network model added: Based on the works of Jerome Vienne
1175 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1176 - When smpi/display_timing is set, also display global simulation time and application times
1177 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1178 * Collective communications
1179 - 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
1180 - 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/).
1181 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1183 - Replay now uses algorithms from wanted collective selector
1184 - Replay can be used with SMP-aware algorithms
1185 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1186 - 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.
1188 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1189 - Correctly use loopback on fat-tree clusters
1190 - Asynchronous small messages shouldn't trigger deadlocks anymore
1191 * Energy/DVFS cleanup and improvement
1192 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1193 - new: smpi_get_host_pstate()
1196 - "Full" network optimization flag was broken since Surf++
1197 - Better handling of precision flags in maxmin
1198 - Fix bug causing sometimes "Impossible" errors
1199 - Properly pass cluster properties to included hosts
1200 * Improvement of the Energy plugin.
1201 - Always update the consumption before returning that value
1202 - New property: watt_off to denote the dissipation when the host is off
1203 - New functions getWattMinAt and getWattMaxAt to retrieve the
1204 dissipation of pstates that we are not currently at.
1205 * Java: class NetworkLink renamed to Link
1206 * New function: simcall_process_get_kill_time()
1207 * Massive rename s/workstation/host/
1208 - That's intrusive, but that's good for the project consistency. Sorry.
1209 - Change config option "workstation/model" into "host/model"
1212 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1213 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1214 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1215 - xbt_dirname and xbt_basename for non-POSIX systems
1217 * The model checker now runs as a separate process.
1218 * The model checker runs must now be launched with the new simgrid-mc program.
1219 * Record/Replay: the MC can display a textual representation of a path in the
1220 execution graph. It can then be replayed outside of the model checker.
1222 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1224 SimGrid (3.11) stable; urgency=low
1229 * Normalizing pointers addresses tool for better diff between logs
1232 * Add cloud examples using new VMs
1233 - examples/msg/cloud/two_tasks_vm.tesh
1234 - examples/msg/cloud/simple_vm.tesh
1235 - examples/java/cloud/cloud.tesh
1236 - examples/java/cloud/migration/migration.tesh
1237 * Add java surf examples:
1238 - examples/java/surfPlugin/surf_plugin.tesh
1239 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1240 - examples/java/surfCpuModel/surf_cpu_model.tesh
1241 * Add SMPI+MSG example:
1242 - examples/smpi/smpi_msg_masterslave/
1247 - msg task destroy cancel test
1248 - msg_host on/off test
1249 * Move all tests in testsuite to teshsuite (adding tesh files)
1250 * Restructure teshsuites
1251 - one folder for each kind of test
1252 * Restructure AddTests.cmake
1254 - structure the order of tests (with sections)
1257 * Add virtual machine
1258 - creation of a VM on a PM
1259 - migration of a VM from a PM to another PM
1261 - MSG_process_join(msg_process_t process, double timeout)
1262 - msg_bar_t MSG_barrier_init(unsigned int count)
1263 - int MSG_barrier_wait(msg_bar_t barrier)
1264 - void MSG_barrier_destroy(msg_bar_t barrier)
1265 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1266 * New option "msg/debug_multiple_use" to help debugging when a task is used
1269 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1271 - Increase code coverage in test suites
1275 * Protect context stacks against stack overflow. The number of protected memory
1276 pages allocated on the top of each stack (1 by default) can be configured
1277 with the new command line option --cfg=contexts/guard_size:<value>.
1278 * Simcalls are now generated by a python script that
1279 - generates files included by SimGrid
1280 - checks that all the functions exist, and proposes prototypes
1282 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1283 host_set_data, host_get_data
1285 - simcall_process_join(smx_process_t process, double timeout)
1286 * Fix bug where sleeping processing could not be suspended.
1289 * Translate surf models from C to C++
1290 - Generic classes for all models: Model, Resource, Action
1291 - A generic interface for each kind of model (CPU, Network, Storage
1292 Workstation, WorkstationVM)
1294 * Translate surf routings from C to C++
1295 * Add callbacks using sigc++ or boost::signals2
1296 - Add callback functions for resource creation/destruction
1297 - Add callback functions for action state change
1298 - Handle Energy as a plugin
1299 * Replace swag by boost::intrusive
1300 * Add new routing models for clusters. For documentation, see
1301 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1302 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1303 parameters for cluster tag
1304 - Fat trees, with topology="FAT_TREE" and
1305 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1306 - see examples/platforms/torus_cluster.xml and
1307 examples/platforms/fat_tree_cluster.xml
1308 * More documentation
1311 * Hostfiles support host:nb_processes construct to deploy several processes on
1313 * Collective communication algorithms should not crash if used with
1314 improper number of nodes and report the error.
1315 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1316 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1317 MPI_Dims_create, MPI_Cart_sub are supported.
1318 * New interface to use SMPI programmatically (still depends on MSG for
1319 some parts, see examples/smpi/smpi_msg_masterslave):
1320 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1324 * Global variables privatization in MPI executables is now performed at runtime
1325 with the option smpi/privatize_global_variables (default:no).
1326 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1327 dynamic libraries loaded by the application are not privatized (static
1328 linking with these libraries is advised in this case)
1331 * Options defined in XML work correctly now.
1334 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1335 copied into simgrid.jar or not (ON by default). Use this option if you want
1336 to reduce the size of the installed simgrid.jar, *and* the native libraries
1337 are kept installed elsewhere.
1338 * Surf binding with SWIG (code generated in maintainer mode only):
1339 - plugin to handle callbacks
1340 - CPU model only for the moment
1343 * Supernovae build mode is definitively removed. It was used to improve
1344 inlining and inter-module optimizations. It is nowadays superseded by
1345 link-time optimizations commonly available in compilers.
1346 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1347 * Add boost dependency for surf++
1348 * Add new macro for tests
1349 - ADD_TESH(name <tesh_args>)
1350 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1354 - xbt_bar_t XBT_barrier_init(unsigned int count)
1355 - int XBT_barrier_wait(xbt_bar_t barrier)
1356 - void XBT_barrier_destroy(xbt_bar_t barrier)
1357 * Make the xbt_os_time module public
1359 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1361 SimGrid (3.10) stable; urgency=low
1363 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1366 * Reintegrate Java to the main archive as desynchronizing these
1367 package is not acceptable anymore (Java is now considered stable)
1368 * Add explicit synchronization facilities through semaphores
1369 * Bug fix: Task.setDataSize() only changed the C world, not the value
1370 cached in the Java world
1373 * Dramatically change the way files are handled. API and internals changed, but
1374 this part of MSG was not considered as production grade either.
1375 * Add explicit synchronization facilities through semaphores
1376 * Add a new function MSG_host_get_process_list()
1377 * Preliminary DVFS support (see examples/msg/energy/ for details)
1380 * SMPI is now included directly in the libsimgrid as the windows
1381 linker doesn't force us on splitting it anymore.
1382 * Improvements of the SMPI replay tool:
1383 - Most of the collective communications are now rooted in the same process as
1384 in the original application.
1385 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1386 used until now). Multiple data types can now be used in a trace.
1387 - The replay tool now supports traces produce either by TAU or a modified
1389 - Bug Fix: the compute part of the reduce action is now taken into account.
1390 - Gatherv collective is now supported
1391 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1392 of smpirun outputs time independent traces for the current run. One file
1393 is created per process. If too many processes are simulated, this behavior
1394 can be changed to one file for all processes by using the
1395 tracing/smpi/format/ti_one_file flag
1396 * smpirun generates the host file if needed (with given host count and platform)
1397 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1398 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1399 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1400 --cfg=smpi/coll_selector:(mpich/ompi)
1401 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1403 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1404 happening inside a collective SMPI call.
1405 * Fix the behavior of complex data types handling.
1406 * Make MPI_Wtime another synchronization point to take computations into
1408 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1409 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1410 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1411 (no privatization of global variables yet)
1412 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1415 * Verification of liveness properties is now available for SMPI applications
1416 (in addition to MSG applications)
1417 * Bugged examples using SMPI in examples/smpi/mc/
1418 * Add --cfg=model-check/visited option. Allows the verification of infinite
1419 programs. Detection of loops in the execution thanks to the system state
1420 comparison and reduction of the state space to explore. Can be combined with
1421 DPOR for safety properties.
1424 * Allow to change SimGrid configuration (see --help) within the code
1425 thanks to SD_config() as it can be done in MSG.
1426 * Add a new function SD_task_set_amount() upon user request.
1429 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1430 * Remove rule based routing (no more PCRE dependency)
1431 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1432 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1433 * Add a loopback_bw and loopback_lat options to cluster tag.
1434 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1435 routing. Now peers are encapsulated in an AS and have their own private
1436 router but this is transparent.
1439 * Our own implementation of getline is renamed xbt_getline, and gets
1440 used even if the OS provide a getline(). This should reduce the
1441 configuration complexity by using the same code on all platforms.
1442 * New type: xbt_cfg_elm_boolean.
1443 * Allow to use yes/no for boolean configuration options in the command line.
1444 * Allow to disable SimGrid cleanups at exit from command line option.
1445 There are situations where one may want a simulation to end with an exit.
1446 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1447 annoying when scripting around the simulator. Adding a
1448 --cfg=clean_atexit:no allows to circumvent this issue.
1451 * Lots of memory leaks were corrected in this release.
1452 There is no known memory leaks anymore, in all of our 600+ tests.
1453 * New command line option --version, to get SimGrid version information.
1454 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1456 * Supernovae builds are deprecated, and expected to be removed in the next
1459 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1461 SimGrid (3.9) stable; urgency=low
1463 The Grasgory release: GRAS is really dead now.
1465 * Complete overhaul of the internal host structures scheme.
1468 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1469 was considered as experimental and badly maintained since then.
1470 * Keeping it was thus a trap to our potential users, that could take
1471 it instead of MSG or SMPI by mistake despite is pity state.
1472 * GRAS seems to have very few users (if any), and no one volunteered
1473 to maintain it further. It also induces a lot of XBT code (for
1474 portability sake), that must be maintained too.
1475 * For all these reasons, we killed GRAS. If someone wants to revive it
1476 in the future, don't cry, our git history still remembers of GRAS.
1479 * Major overhaul. Merge our documentation again as time proved that
1480 splitting it was really not helping our users.
1481 * Further improve the developer documentation to help newcomers
1482 hacking on SimGrid itself. The user documentation (and in
1483 particular, the beginner documentation) is still in a sorry state.
1486 * Now works on Windows too!
1487 * Much more extensive test suite, from MPICH
1490 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1491 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1492 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1493 * Bug fix: let task be scheduled when the last dependency to be solved is
1494 a control dependency.
1495 * Remove SD_load_environment_script function.
1496 Use the C sg_platf function if you want to declare a platform
1500 * New function: MSG_process_get_number()
1501 * Old function documented: MSG_config()
1502 * Remove MSG_load_platform_script function
1503 Use the C sg_platf function if you want to declare a platform
1507 * Change the default value of the TCP_gamma constant (maximal size of TCP
1508 congestion window) to a more realistic 4MiB value. If you notice changes in
1509 your simulation results, you can fall back to the previous 20k tiny window
1510 by adding --cfg=network/TCP_gamma:20000 on command line.
1511 * (Hopefully) fix a bug wrt periodic availability and state traces
1512 * Bug fix: use default values at start when first event in availability/state
1513 trace is not at time 0.
1516 * remove the "new_" part of function name sg_platf_new_trace_connect
1517 (resulting in sg_platf_trace_connect), since it does not create
1521 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1522 use case where it's really mandatory, and maintaining it was a pain in
1524 * New: xbt_fifo_search(), search an item with a user-provided
1525 comparison function instead of dumb pointer comparison.
1528 * Fix the lua deployment:
1529 Use `simgrid.init_application()` before deployment instead of
1530 `simgrid.msg_register_application()` after.
1533 * Transfer the tracing files into the corresponding modules.
1535 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1537 SimGrid (3.8.1) stable; urgency=low
1539 The "we are told that some people want to also *install* the simgrid
1542 * Add missing file "tesh.1" to the archive.
1544 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1546 SimGrid (3.8) stable; urgency=low
1548 The Psssshiiiit release: SimGrid jumps into the Cloud.
1551 * Add an experimental interface to manipulate VMs. They are mainly
1552 process groups with very few intrinsic semantic, but they should
1553 allow you to build the semantic you want easily.
1554 * New function: MSG_host_set_property_value()
1555 * New function: MSG_process_on_exit(). To clean memory in all cases.
1556 * Bug fixes that made the host (and link) failures unusable.
1557 * Add a way to auto-restart process when the host in which they are
1558 executing comes back (ON_FAILURE="RESTART" on deployment file,
1559 MSG_process_auto_restart_set).
1560 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1561 please stop using the old ones, they are DEPRECATED.
1563 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1564 Please use MSG_init() instead. (reducing the amount of entry
1565 points in the library helps us).
1566 * Make it impossible to link against the wrong version of the lib
1567 * Deprecate MSG_clean(). No need to call it anymore.
1568 * Function MSG_get_host_number() is not deprecated anymore.
1571 * Split the doc into a user guide and a reference guide.
1572 * Start a developper guide to help people hacking on SimGrid.
1575 * Enable tracing by default. This modules rocks you should use it.
1576 * Remove option custom_flags. Now use environment variables CFLAGS
1578 * Use default cmake things to detect lua instead of home grown ones.
1579 * New option "enable_mallocators" to disable mallocators, for debugging
1580 purpose ("on" by default).
1583 * Bug fixes around the resource failures: don't let the processes
1584 survive the host they are running onto.
1585 * Add an interface to auto-restart processes when the host in which they are
1586 executing comes back.
1587 * Ensures that SIMIX_clean is called automatically. It's not part of
1588 the public interface anymore (bindings should be updated).
1591 * Bug fix for when SD_Simulate is called with a positive value: be careful
1592 when comparing doubles. Sometimes they are different for non significant
1594 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1595 parallel task whose initial work is distributed among host according
1596 to the Amdahl's law. Such tasks are created with a parameter alpha
1597 that corresponds to the non-parallelizable part of the computation.
1598 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1599 between two sets of workstations assuming a 1D block distribution (each
1600 workstation owns a similar share of data) on both sides.
1602 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1603 Data redistribution will be automatically scheduled once parent and child
1604 are both scheduled. The filling of computation_amount and
1605 communication_amount structures is now done seamlessly thanks to the chosen
1607 * New function SD_workstation_dump to display various information
1608 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1609 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1610 bandwidth on the route onto which the task is scheduled and the amount of
1612 To divide the nominal bandwidth by 2, the rate then has to be:
1613 rate = bandwidth/(2*amount)
1614 * Compute tasks that have failed can now be rescheduled and executed again
1615 (from their beginning)
1616 * Increasing source code coverage (src/simdag is now covered at 95.8%
1620 * Re-implement time-independent trace replay using SMPI (at the
1621 smpi_smp_* level) instead of MSG. This should replace
1622 examples/msg/actions/actions.c
1623 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1624 hindexed and structs)
1625 * Implement the exchange of non-contiguous data.
1626 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1627 * Correct behavior of smpi/sender_gap and set its default value to 0
1628 * Add option to asynchronously send small messages to allow better
1629 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1630 specifies the size in bytes under which messages will be asynchronously sent.
1631 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1632 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1635 * New C interface to define a platform: XML is now optional.
1636 For more info, please check include/simgrid/platf.h.
1637 * New interface to define random platforms from the C:
1638 For more info, please check include/simgrid/platf_generator.h and
1639 examples/msg/masterslave/masterslave_platfgen.c
1640 * Export a sg_cmdline dynar containing all the arguments we got from
1644 * Two new tracing options for adding comments to trace file so you
1645 can track your experiments (see --help-tracing for details).
1646 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1647 * Adding the SimGrid version that generated the trace file as a comment.
1648 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1649 * Fix to avoid key clashes on Paje links
1650 * Other minor fixes related to the Paje specification
1653 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1654 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1655 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1656 * Greatly improve the robustness of mmalloc to user errors (such as
1657 using an area after freeing it, or freeing it twice)
1659 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1661 SimGrid-java (3.8.1) stable; urgency=low
1663 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1664 Warning: all methods are visible, but only some of them are
1665 implemented so far. Check the source (src/jtrace.c)
1666 for further information.
1667 * New module: org.simgrid.msg.File (SimGrid File management functions)
1668 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1669 * Change the meaning of Process.restart: now restart the process from
1670 the begining, like MSG_process_restart in C.
1671 * Add Process.setAutoRestart: handling of process restart when failed
1673 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1674 you to retrieve the properties of the processes/hosts
1675 * Deprecate Msg.clean(): you can just forget about it now.
1676 * New function Process.getCount(), that only works when compiling
1677 with the not yet released version 3.9 of the C library.
1679 * New context factory based on Coroutines. It mandates a modified JVM
1680 but then, the simulations run about five times faster, and there is
1681 no limit to the amount of processes (beside of the available memory).
1683 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1685 SimGrid (3.7.1) stable; urgency=low
1688 * Restore the prototype of MSG_process_create_with_environment() to
1689 the pre-3.7 situation by removing the kill_time argument.
1690 * Add a MSG_process_set_kill_time() function instead.
1693 * Fix weird behaviors when dealing with parallel tasks.
1696 * Simgrid is now built as a dll.
1697 * Simgrid-java now works on Windows.
1698 * Simgrid-Java is now included into Windows package.
1701 * First pre-build package for MacOSX.
1704 * Fix compilation when using MSG_USE_DEPRECATED.
1705 * Fix some compilation issues on Macs and Windows.
1706 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1707 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1709 We discovered that the Lua console is broken, but we are missing the
1710 manpower to fix it right now. The problem existed in 3.7 too, so we
1711 are not blocking the release for that. Sorry if you depended on this
1712 feature, any help would be really welcome.
1714 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1717 SimGrid-java (3.7.1) stable; urgency=low
1719 The "Java aint got to be bloated and slow" release
1722 * Various internal cleanups and performance improvement
1723 Simulations are expected to run up to twice faster or so
1724 * Make Process.kill(process) an instance method, not a static one
1725 * User processes are not java.lang.Thread subclasses.
1726 This breaks the compatibility (sorry), but previous API was
1727 brain-dead, making it impossible to have non-trivial
1728 initializations in the process constructor.
1729 * Require a full constructor per Process sub-class.
1730 Kinda breaks the compatibility (sorry), but this allows a much more
1731 efficient way to launch the processes at simulation startup.
1732 * Do not embeed our version of semaphores, java 1.5 can be considered
1733 as sufficiently prevalent for us to not dupplicate its features.
1737 * Add examples for almost every part of the API
1738 We spotted and fixed a lot of bugs in the process
1740 * New module: asynchronous communication API
1741 * New function: Process.sleep()
1742 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1743 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1744 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1746 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1748 SimGrid (3.7) stable; urgency=low
1750 The "spring cleanups (before the next Big Project kicks in)" release.
1753 * We can specify the SMPI latency/bandwidth factor with command line
1754 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1755 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1756 You can also use the "config tag" from platform file by adding this line
1757 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1758 (see "example/platforms/tag_config.xml" to use "config tag").
1759 Note that the command line supersedes the platform file configuration.
1760 * Change the correction factors used in LMM model, according to
1761 the latest experiments described in INRIA RR-7821.
1762 Accuracy should be improved this way.
1763 * Use the partial invalidation optimization by default for the
1764 network too. Should produce the exact same results, only faster.
1765 * Major cleanup in surf to merge models and split some optimization
1766 mechanisms from the core of the models. As a result you can now
1767 specify which model to use (e.g., --cfg=network/model:LV08
1768 --cfg=cpu/model:Cas01) and which optimization mode to use
1769 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1770 Incompatible combinations should err at initialization. See
1771 --help-models for the list of all models and optimization modes.
1772 * The CLM03 workstation model was dropped for simplicity because it
1773 used the deprecated CM02 network model. Use default instead.
1774 * Rename the TCP_gamma configuration option to network/TCP_gamma
1775 * Rename the coordinates configuration option to
1776 network/coordinates, and document it
1777 * Use now crosstraffic keyword instead of the terribly misleading
1778 fullduplex keyword. It is activated by default now in the current
1779 default model, use --cfg=network/crosstraffic:0 to turn it off.
1780 * Ongoing refactoring the model parsing to make XML files optional
1781 See include/simgrid/platf.h for details (still to be completed)
1784 * Major overhaul of the documentation. Almost instructive now :/
1785 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1786 functions and friends. This interface was considered as
1787 deprecated since over 2 years, it's time to inform our users that it is.
1788 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1789 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1790 use these (crappy) functions in your code.
1791 These functions will be removed soon. Stop using them now.
1792 * Deprecate MSG_get_host_{table,number}
1793 Implement MSG_hosts_as_dynar() instead.
1794 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1795 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1798 * Stabilize the parallel execution mode of user contexts
1799 * Introduce configuration variables to control parallel execution:
1800 - contexts/synchro: Synchronization mode to use when running
1801 contexts in parallel (either futex, posix or busy_wait)
1802 - contexts/parallel_threshold: Minimal number of user contexts
1803 that must be part of a scheduling round to switch to parallel
1804 execution mode (raw contexts only)
1805 * Fix bugs that prevented to use suspend/resume along with
1806 synchronization structures.
1807 * Fix bugs in process termination that lead to invalid memory access
1808 in very specific conditions.
1811 * Introduce a parallel mode for the models (controlled by surf/nthreads
1812 configuration item). In our tests, running the models in parallel
1813 never lead to any speedups because they are so fast that the gain
1814 of computing each model in parallel does not amortizes the
1815 synchronization costs, even when ultra fast futexes are used.
1816 This is released anyway because YMMV.
1819 * Performance boost by using a swag internally to compute the set of
1820 tasks that are finished and should constitute the return value of
1824 * Enable it by default now that it is considered rather stable.
1827 * Documentation of the tracing functions.
1828 * Performance gains when tracing categorized/uncategorized resource
1829 utilization by avoiding calls to get route when updating resource
1830 variables. LMM constraints are being used instead.
1831 * API changed to set task categories. Use MSG_task_set_category instead
1832 of TRACE_msg_set_task_category, and SD_task_set_category instead
1833 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1834 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1835 * Examples for link user variables added (see at examples/msg/tracing/)
1836 * Deprecated function TRACE_msg_set_process_category completely removed
1837 * Trace header updated according to the latest Paje file format
1838 * Tracing network lazy updates, no longer obligate users to use full updates
1839 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1840 * Experimental: let user code declare/set/push/pop application states for hosts
1841 * API changed to allow the manual creation of graph configuration files
1842 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1845 * Improve the API of Lua MSG bindings, using the Lua spirit.
1846 * Each simulated process now lives in its own Lua world (globals are
1847 automatically duplicated). It helps writing simulators. Will allow
1848 to run Splay programs within SimGrid in the future.
1849 * Add a Chord example in Lua, equivalent to the MSG one.
1852 * Start the implementation of a solution to express temporal
1853 properties, not only local assertions. This is still an
1854 experimental work in progress, stay clear from it to be safe.
1858 - Add new runtime parameters --help-logs and --help-log-categories
1859 to display information about supported logging parameters and
1861 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1862 don't exists anymore.
1863 * Mallocators: allow value NULL for the reset function.
1865 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1866 create homogeneous dictionaries, where all the elements share the
1867 same free function. Non homogeneous dictionaries will be
1868 deprecated in the next release.
1869 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1870 - Multi-level dictionaries are deprecated.
1872 - new function xbt_dynar_search_or_negative() that is useful when
1873 you have less than 2 million elements in your dynar and don't
1874 want of the extra complexity of catching exceptions when the
1875 element is not found.
1877 - Make xbt_os_thread module (for thread portability) public.
1878 Documentation is still to come, sorry.
1880 - Cleanups and simplifications to make it maintainable again.
1881 - Exotic features (such as memalign and valloc) were removed.
1882 - The metadata were extended and improved so that the
1883 model-checker becomes able to explore and inspect the heaps.
1884 - This may induce a performance drop when enable_model-checking is
1885 ON in cmake (even if it's not used in the simulation), but it is
1886 necessary at this point to get MC working.
1888 Turn model-checking OFF if simulation performance matters to you.
1889 Not enabling it at runtime is not enough, disable it in cmake.
1891 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1894 SimGrid (3.6.2) stable; urgency=low
1896 The "Not coding new stuff allows to polish old things" release.
1899 * New bindings to the NS3 packet level simulator (experimental)
1900 * Use the raw (efficient) execution contexts instead of the sysv
1901 (portable) ones when possible.
1902 * libpcre is now mandatory in any cases since not using it led to
1903 severe performance loss and possibly other issues
1904 * Update the XML platforms:
1905 - G5K: include the latest machine in Nancy
1906 - GridPP and LCG: new platforms
1907 * Documentation was partially updated, at least (more to come)
1909 Bug fixes, cosmetics and small improvements
1910 * Free terminated processes before the end of the simulation to avoid
1911 exhausting the memory of users having very dynamic amount of
1913 * Bug fix and cosmetics about canceling non-running tasks
1914 * Bug fix about the dot loader's issues when using libcgraph
1917 * Create an installer for windows with nsis (amd64 and win32)
1918 - Add an hello world project to illustrate simgrid project creation.
1919 - Embed libpcre into the Simgrid installer to avoid
1920 its compilation burden
1921 * The raw execution contexts should work on Apple now
1922 * Port to Windows 64 bits
1923 - Sysv contexts now have an implementation for this arch
1924 - GRAS communication features now support this arch
1925 * Drop support for borland compiler on windows
1926 - this code was not maintained, and we kinda depend on gcc nowadays
1927 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1928 * Fix portability issue on unstable ubuntu: linker became picky on
1931 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1933 SimGrid (3.6.1) stable; urgency=low
1935 The "Oops, we broke Macs too" release
1938 * Fixed contexts detection so that raw ones are used when possible
1939 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1940 leads to a strange error, with user code segfaulting sometimes when
1941 the generated code is not perfectly aligned (which is not
1942 controllable from the user side, depends on the amount of code)
1945 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1947 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1949 SimGrid-java (3.6) unstable; urgency=low
1952 * Split of every thing from simgrid v3.5 into a separate package.
1954 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1956 SimGrid (3.6) stable; urgency=medium
1958 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1961 * Bindings now constitute their own package, separated from the main one.
1962 Rationale: reduce our maintenance nightmare by reducing the module coupling
1963 They will soon be released on their own on gforge.
1965 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1966 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1968 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1969 * It's not quite deprecated for now because we have no replacement,
1970 but it may soon become the case.
1973 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1974 * Fortran: New user-level cache variable to store the rank of the running
1975 process. This improves performance by an order of magnitude.
1976 * C: New coccinelle script to automatically locate and modify global and
1977 local static variables.
1978 * Improved SMPI network model with a sender-side gap to account for multiple
1982 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1983 only say if a communication is finished, no matter whether it succeeded or
1984 failed. You can call MSG_comm_get_status() to know the status of a finished
1986 * New function MSG_task_dsend() to send a task and detach it. When a
1987 communication is detached, you are never notified of its success or failure
1988 and the memory is released automatically once it is finished. This function
1989 is useful when you don't care about the end nor the success of a
1991 * Change the prototypes of action replay. Sorry for inconvenience,
1992 but this is really more efficient this way (and to adapt your code,
1993 you just have to fix the initialization, that shouldn't be too long)
1994 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1995 ever managed to do anything useful with it.
1996 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1997 takes a dynar of communications. It returns immediately and gives the
1998 index of a finished communication (if any).
1999 * New example: a basic implementation of the Chord P2P algorithm.
2002 * New model for multi-core CPUs. You can now use the core attribute to
2003 precise the number of cores of a host. This is a basic model. Every
2004 process running on the host receives at most the power provided in
2005 the DTD (throughput<=power). Total throughput of process cannot exceed
2007 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2008 router linked by an up-link and a down-link (possibly asymmetrical).
2009 This kind of pattern allows to easily build last-mile model style platforms.
2010 Aggregating such patterns in a rule-based AS is thus the technique of
2011 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2012 * New model for Vivaldi routing. We transformed the Vivaldi network model
2013 into a Vivaldi routing model (based on the rule-based model). This allows to
2014 combine Vivaldi based latencies with last-mile platforms.
2017 * Added a check for NaN of IEEE754 infinite in the double entries of
2019 * Introduce a new context factory "raw", highly inspirated from the
2020 ucontext factory, but using manually crafted functions in assembly to
2021 do the work in an efficient manner.
2022 * Allow to change the used context factory at run time, not only at
2023 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2024 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2025 contexts at runtime (only with raw contexts or ucontexts).
2026 * Completely rewrote this module to allow parallel execution of user
2027 processes. Use --cfg=contexts/nthreads:N to execute user processes
2028 with N parallel threads (the default is 1, meaning no parallelism).
2029 * Allow to decide dynamically between sequential and parallel modes.
2030 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2031 processes in parallel only when their number is greater than or equal to P
2033 * Added a check for NaN of IEEE754 infinite in the double entries of
2037 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2038 won't output the state of processes when interrupted with Ctrl-C
2039 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2040 NULL-terminated array. This may solve backward compatibility issues
2041 due to the change to return type of SD_simulate. See also:
2042 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2043 * Add new macros with variable number of arguments.
2044 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2045 - in xbt/asserts.h: xbt_assert
2046 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2047 - in xbt/ex.h: THROWF and RETHROWF.
2048 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2050 * Change xbt_die() to accept a format string with arguments, just like printf.
2051 * New data structure: xbt_lib_t, like a dict but more general and with better
2055 * New configuration options
2056 Options triva/categorized and triva/uncategorized can be used to generate
2057 graph configuration files for Triva visualization tool.
2058 * Configuration option tracing/platform is renamed to tracing/categorized
2059 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2060 * New TRACE_user_link_variable interface:
2061 User provides the name of the link and the tracing variable to attach to it
2062 * the declaration of tracing categories must be done after the environment
2064 * simpler tracing interface, just one way to declare categories
2065 TRACE_category or TRACE_category_with_color, it is up to you
2066 * links in the trace file are again identified by their names
2067 * trace contains the full platform hierarchy exactly as declared using the ASes
2068 * Options tracing/msg/[task|process]:1 groups the process by hosts
2069 for both cases, tasks and processes must have names that are unique during the simulation
2070 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2071 * The experimental option tracing/msg/volume is deprecated
2072 its functionality may be reincorporated if needed
2074 The tracing generates a trace file with unordered timestamped events,
2075 because of the way the core simulator (surf) works. A script available
2076 at the tools directory (fix-paje-trace.sh) can be used to put the events
2077 in order. We have changed the tracing so it can generate ordered timestamped
2078 events in the final trace, but depending on the simulator (and how much time
2079 is simulated) that can lead to a huge memory utilization. It is deactivated
2080 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2082 Build Infrastructure
2083 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2084 - We are trying hard to keep the API stable, but it may happen that
2085 some things change (we're a research project after all, not a
2086 nuclear plant operating system). If such things should happen, you
2087 could rely on that macro to adapt.
2088 - current value: 30600 for 3.06.00, aka 3.6
2089 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2090 a number that can be compared with SIMGRID_VERSION.
2091 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2092 assertions and verbose/debug logging events are disabled at compile time.
2094 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2096 SimGrid (3.5) stable; urgency=medium
2099 * New feature to any SimGrid-based simulator: Model-Checking
2100 Check SIN#1 for more details.
2103 * New Model SMPI (three-interval linear regression for correction factors)
2104 See RR-7426, available at http://hal.inria.fr/inria-00527150
2105 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2106 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2107 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2108 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2109 * See also src/smpi/README
2114 - Tracing API changes: TRACE_start and TRACE_end should not be called
2115 by user-code. They are automatically called by simulators created
2116 with SimDAG, MSG and SMPI if the toolkit is compiled with
2117 tracing_enabled=ON. Categories declaration and utilization remain the
2118 same for MSG and SimDag.
2119 - A function was added to the tracing API to declare categories with
2121 - TRACE_category_with_color (char *category, char *color)
2122 where color must be in the following format
2123 "%f %f %f", red, green, blue
2124 and red, green, blue are float values in the interval [0, 1]
2125 - User can specify NULL as color parameter, or continue calling
2126 TRACE_category (cat)
2127 On that case, the tracing system will define random colors
2128 - The following command-line options are supported:
2129 --cfg=tracing/filename:msg.trace
2130 --cfg=tracing:1 (activate tracing, needed to use others)
2131 --cfg=tracing/platform:1 (categorized resource use)
2132 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2133 --cfg=tracing/msg/task:1 (task creation)
2134 --cfg=tracing/msg/process:1 (process creation, migration)
2135 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2136 --cfg=tracing/smpi:1 (SMPI interface tracing)
2137 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2138 - examples of examples/msg/tracing updated
2140 - DAXLoader and DOTLoader functions can generate tasks with categories
2141 - A new function to attribute a category to SD tasks:
2142 TRACE_sd_set_task_category (SD_task_t task, char *category)
2143 * Tracing the MPI interface implemented by SMPI
2144 - Collective operations are traced with states
2145 - Point-to-Point operations are traced with states/links
2146 - Tracing activated by a parameter "-trace filename" passed
2147 to smpirun during execution (considering that simgrid
2148 is compiled with tracing enabled)
2149 - To run the simulation with gdb, the simulator
2150 accepts --cfg=tracing/smpi:1 to trace SMPI
2151 - tesh files to check if smpi tracing is ok
2152 - See examples/smpi/NAS/DT-trace
2153 * GTNetS tracing re-worked
2154 - adaptation to the tracing system of GTNets to cope
2155 with modifications regarding the fullduplex mode
2156 - new tesh files to check if gtnets tracing is ok
2159 * Asynchronous communications through the functions:
2160 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2161 * New function: MSG_load_platform_script()
2162 to make possible using a lua script instead of XML files to set up platforms
2163 * New function: MSG_set_function
2164 to associate functions to processes, used when bypassing the parser
2165 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2167 Platforms: Add some more examples in examples/platforms
2168 * Grid'5000: see www.grid5000.fr
2169 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2172 * Change the XML format. This is a very important modification. SimGrid
2173 3.5 introduces a new hierarchical format based on the notion of
2174 Autonomous Systems. Compatibility with old format is ensured
2175 through the perl script provided in the install bin directory
2176 bin/simgrid_update_xml.
2177 It is now possible to build platforms with specific routing
2178 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2179 connect several platforms together. We will try to provide soon
2180 set of realistic platforms exploiting these properties (have a
2181 look at examples/platforms/ for the moment).
2182 * Take the opportunity of the XML format change to be a good XML citizen:
2183 rename link:ctn to link_ctn and similar changes (also dealt with by
2185 * Add a new routing scheme (rule-based) using regular expressions. It
2186 enables to have an extremely low memory footprint when the
2187 underlying routing is simple and can be compactly described. You
2188 need to have libpcre4-dev (perl regular expressions) installed if
2189 you want to use this routing scheme.
2190 * Revive the cluster TAG and allow to easily and efficiently (both in
2191 term of memory and speed) connect clusters together. Have a look
2192 at teshsuite/simdag/platforms/ to see how this can be done. With
2193 this tag, you can create clusters with thousands of tasks at no
2194 cost (have a look at examples/platforms/).
2195 Note: clusters are implemented as ASes, so there is no need for an
2196 enclosing AS tag if you have only one cluster in your platform.
2197 * Add new generic functions in the public interface that allows the user
2198 to call SURF 'create_resource' methods from your code (same
2199 functionality as the XML bypass mechanism but with a much lighter
2201 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2202 that uses a piecewise linear approximation to produce better
2203 results when exchanging small messages.
2204 * Add a new parameter to handle correctly full duplex link and account
2205 for interferences between uplink and downlink communications
2206 (activate with --cfg=fullduplex:1).
2209 * Rename the SD_READY (all dependencies are satisfied and task is
2210 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2211 dependencies are satisfied) state.
2212 This prevents a confusion between the notion of "ready to schedule"
2213 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2214 simulated" (SD_RUNNABLE) used by the simulation kernel.
2215 * Change the way a task is considered as ready. Instead of removing
2216 dependencies when a task is done, a counter is decreased. This way,
2217 it is always possible to reach ancestors thanks to the
2218 SD_taks_get_parents function (even after the end of the simulation.)
2219 * Change the return type of SD_Simulate from (SD_task_t*) into
2220 xbt_dynar_t. This function was in handling a dynar internally and
2221 converted it into a NULL terminated array for historical reasons.
2222 * New function SD_dotload(char*) to load a DAG described in dot
2223 format. This loader and the corresponding examples require the
2224 installation of the graphviz library.
2225 * Fix a bug in the management of tasks of size 0 in the surf network
2226 models. This problem was only visible with SIMDAG and you should
2227 thus disregard results produced with earlier versions if you
2228 relied on this feature (some tasks were blocked because of this).
2229 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2230 even though your applications were doing only point-to-point
2231 communications and sequential computations. Now you can really use any
2232 model you want (of course, if you create real parallel tasks, which are
2233 not implemented in most models beside ptaskL07, this will abort).
2234 * Add an example that schedules a DAX on an heterogeneous platform
2235 using a Min-Min strategy.
2236 * New function SD_workstation_get_current_task() that returns the kind
2237 of task currently running on a workstation in the sequential access
2239 * Raise some warnings when unexecuted tasks remains at the end of the
2240 simulation. This is usually caused by cycles in the DAG.
2243 * New function: SIMIX_process_set_function() called by MSG_set_function
2244 * Change the underlying waiting queue in semaphores so that a process
2245 can wait on several of them simultaneously (as in waitany).
2246 * Fix the way to handle tokens in semaphores so that all access patterns
2247 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2248 * kill the dirty pimple SIMIX_message_sizes_output()
2249 Please use (proper) visualization instead
2252 * New data container: setset (set of sets of elements)
2253 * New module: mmalloc (mapped malloc, allowing to have several
2254 independent segments of malloc)
2255 * New function: xbt_dict_cursor_set_data()
2256 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2257 * New function: xbt_dynar_is_empty()
2258 * New function: xbt_fifo_get_last_item()
2259 * Fix xbt_dynar_shrink(): use the right element size.
2260 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2261 uninitialized areas during expand.
2262 * Fix semaphores: previous implementation was severely broken.
2263 * Use library init/fini functions for our initialization.
2264 - you can use logs and other feature as soon as you want in your
2265 code (even before the xbt_init / MSG_init)
2266 - xbt_exit is now a no-op and produce a warning when used.
2269 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2270 benefit from the latest improvement to the simulation kernel.
2271 * Kill measurement sockets for now. If you rely on them, sorry. This
2272 release is not for you. This feature will be reintroduced in the
2273 future, but we cannot delay the release any further.
2274 * New function: gras_msgtype_get_name().
2275 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2276 * Fix (at last) the pmm example: it should not randomly fail anymore.
2278 Build chain: bug fixes and overall polishing
2279 * Cmake is now stable enough. Hence, we killed the autotools.
2280 * Port to windows ( TM :)
2281 * Fix the 'make install' target.
2282 No need to use 'make install-simgrid' anymore
2283 * Introduce a 'make dist' target compiling a *source* archive
2284 'make package' compiles a binary archive
2285 * Compile java files only on need
2286 * Add --cd and --setenv command line options to tesh
2287 * Out of source builds are not fully supported yet, but we are close
2288 * Enable supernovae and optimization flags by default for our users
2291 * Add layer to set up environment directly from lua, without XML.
2292 * The effect of gras_stub_generator can be achieved through
2293 lua too (check examples/gras/console/ping_generator.lua)
2295 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2297 SimGrid (3.4.1) stable; urgency=low
2299 The "Polishing easter eggs is probably a good idea" release.
2300 This is a bug fixes release only.
2304 * Fix a bug preventing the tasks from begin garbage collected.
2307 * Fix a bug occuring when a host involved in a communication fails.
2308 This was not detected properly by the other peer involved in the
2309 communication. Now, it's reported as a network error.
2312 * Warn the user about loop dependencies in data flow of DAX files
2313 * Obey the control-flow dependencies of DAX files
2316 * Add option "enable_smpi" allowing to not compile SMPI.
2317 Probably useful for the (Mac) users experiencing a build error here
2318 * Improve the detection of lua5.1 and ruby1.8
2320 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2322 SimGrid (3.4) stable; urgency=low
2324 The "Easter in Cargese" release. Also known as (major changes):
2326 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2327 ~> bindings were greatly improved
2328 ~> new tracing infrastructure for better visualization introduced
2330 * the "Welcome to configury modernity" release.
2331 ~> we switched from autotools to cmake, and improved our cdash
2334 A more detailled list of changes follow (full detail in svn log).
2336 Java Bindings: Various Cleanups
2337 * (install java-gcj-compat-dev on debian-like to use them)
2338 * Remove put/get: no need to export deprecated interface in Java
2339 Use send/receive instead.
2340 * Cleanup the examples and add a README per directory
2341 * Remove example autoDestination (that's the only way to go now)
2342 * Remove example explicitDestination (was a plain copy of basic)
2343 * Make JniException a runtime exception, so that there is no need to
2344 declare the fact that you may encounter such a beast. I guess that
2345 nobody will ever want to survive such error.
2346 * Create specific errors for each MSG case of failure:
2347 host failure, transfer failure, timeout, task cancelled
2348 * Cleanup the exceptions that may get thrown by each function
2349 * Other internal cleanups in Java bindings. Performance still bad :/
2350 Ruby and Lua Bindings: create them
2351 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2352 * That's new and great, you should try them out.
2353 Same functionalities than Java bindings, only even less polished
2355 * Kill the useless "rate" argument of SD_task_get_execution_time()
2356 Everyone used to provide -1 as a value, it was not used, and the
2357 semantic of a possible use wasn't even clear.
2358 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2359 either as comm costs or compute costs to mean that there is no
2360 such thing for that specific task.
2361 * Add a SD_task_set_name() function
2362 * Fix SD_task_unschedule() on typed tasks
2363 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2364 * In DAX loader, accept useless 'level' attributes to <job> since
2365 LIGO DAGs have them (seem to be to ease graphical representation).
2367 * Add an example masterslave_mailbox.c using send/receive and not
2368 the deprecated put/get interface.
2369 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2370 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2372 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2373 (declare MSG_USE_DEPRECATED to still have the old name)
2374 * Add a MSG_task_set_data() function
2375 * About trace replay (see examples/msg/actions):
2377 - Allow to work with splitted trace files for each process
2378 Give the specific trace file as argument of each process,
2379 and call MSG_action_trace_run(NULL)
2380 You can still have one merged file for all processes.
2381 - Fix implementation of collective operations
2382 * Allow task_execute() on 0-sized tasks (closes #10063)
2384 * This is the first release of SimGrid where SMPI is not considered
2385 beta anymore (even if some corners should still be improved)
2386 * Port over the new SIMIX_network submodule (internal refactoring)
2387 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2388 * Implement more missing elements of the standard:
2390 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2392 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2393 MPI_Scan() MPI_Get_processor_name()
2394 - Added implementation of missing case for Alltoall (warning: it's
2395 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2396 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2397 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2398 SMPI_MPI_Allgatherv()
2399 * Bug fixes include:
2400 - MPI_Waitsome() was broken
2401 - Allow relative includes in smpicc
2402 - Command line cfg argument 'reference_speed' was ignored...
2403 - Some functions did not properly lead to auto-benching of user code
2404 - smpicc passes -O2 by default (just like openmpi one)
2406 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2407 * Bug fixes about timeouts during communications
2408 * add SIMIX_message_sizes_output() as a pimple to write to file the
2409 amount of messages per size. Use gnuplot to get histogram.
2410 Pimple because that's the only user-visible function of simix,
2411 defined directly in xbt.h (irk, sorry)
2413 - Add a SIMIX_sem_get_capacity() function
2414 - Fix interactions with processe resume/suspende
2415 - release_forever() was stupidly broken
2416 - Fix SIMIX_display_process_status() for processes in a semaphore
2417 - Make SIMIX_sem_block_onto() user-visible
2418 * Refactoring context stuff:
2419 - Use pseudo-OOP for better modularity
2420 - reimplement SIMIX_process_kill() without process_schedule() so
2421 that the latter can take as invariant that it is called from
2423 - Merge context_start into context_new for sake of simplicity
2425 * Add a Vivaldi network model, coded live during SUD'10 ;)
2426 * Rename configuration variables to start a hierarchy:
2427 o cpu_model -> cpu/model
2428 o network_model -> network/model
2429 o workstation_model -> workstation/model
2430 * New configuration variables:
2431 o network/bandwidth_factor: correction to bandwith
2432 o network/latency_factor: correction to latency
2433 o netwotk/weight_S: correction to the weight of competing streams
2434 * Add a long description to the models, that users can see with such
2435 argument on the command line: --cfg=cpu/model:help
2436 * --help-models display the long description of all known models
2438 * config: add the ability to set a default value after registration
2439 Does not override any previously set value (e.g. from cmd line)
2440 * dict: allow to have integer key and data.
2441 When so, you need to use the following functions
2442 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2443 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2444 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2445 In contrary to regular dicts, the key is not malloced before copy.
2446 Mixing scalar and regular elements in the same dict is not tested
2448 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2449 Tracing for Visualization:
2450 * SimGrid is now instrumented in order to generate a trace file for
2451 visualization analysis: to use it, need to compile SimGrid with the
2452 "tracing" option enabled, and instrument the program using SimGrid with
2453 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2454 (among other functions).
2455 * The instrumentation only traces the platform utilization for now
2456 * Documentation to use the tracing functions and how to analyze the
2457 traces with the Triva tool is written.
2458 * More information about: SimGrid FAQ (in the section Tracing Simulations
2461 * We moved to cmake as default build system. Autotools support will
2462 be dropped soon. Check the FAQ for more info about how to use it.
2463 * Greatly improved our cdash/ctest interactions
2464 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2465 * Added memory checking tests with valgrind; lot of memleak fixing.
2466 This may be the first release of simgrid with so few memory issues
2467 * Added code coverage tests.
2468 Our coverage is still improvable, but at least we see it on cdash.
2470 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2472 SimGrid (3.3.4) stable; urgency=low
2474 The "Desktop Grid needs love too" release (also called Xmas release).
2476 Models improvements:
2477 * Major speedup in the maxmin system solving by using lazy evaluation
2478 Instead of solving completely the maxmin system at each iteration,
2479 only invalidate (and recompute) the modified parts.
2480 This new feature is enabled in default models but you can try to
2481 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2482 * Cas01 IMproved as default CPU model
2483 This CPU model is the same Cas01 model, but it uses the
2484 maxmin-selective-update flag and a heap structure to manage
2485 actions on SURF kernel.
2486 It reduces the complexity to find the next action to finish and,
2487 consequently, it's faster than the old Cas01.
2488 This is the new default CPU model (Cas01).
2489 * Rename the old Cas01 model to Cas01_fullupdate
2490 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2491 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2492 * CpuTI (CPU Trace Integration)
2493 A new CPU model whose objective is simulate faster when using
2494 availability trace files.
2495 Instead of using a full featured, over engineered maxmin system for
2496 CPU modeling, this model does the pre-integration of traces files
2497 to calculate the amount of CPU power available, and so, executes
2498 faster than the old CPU models.
2499 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2500 * Use LV08 as default network model since it gives better accuracy
2501 for small messages and shouldn't change things for big ones.
2502 Use --cfg=network_model:CM02 to get the previous behavior.
2505 ******************************************
2506 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2507 ******************************************
2508 * The new CPU model may changes simulations!
2509 The point is that events occurring at the exact same timestamp
2510 are not scheduled in the same order with the old and new
2511 version. This may be enough to completely change the execution
2512 of simulations in some cases.
2513 * The new network model will change simulations!
2514 This new model is more realistic than the previous one, so you
2515 should consider redoing your old experiments with this model.
2516 Sorry for the inconvenience.
2519 * Introduce the supernovae compilation mode
2520 When compiled that way, the whole SimGrid (or almost) is put in a
2521 single compilation unit and compiled in one shoot.
2522 This is to help gcc which has difficulties to inline stuff from one
2524 The speedup seem to be above 15%, althrough more tests are needed on
2525 amd64 to confirm that gain.
2528 * Port of MSG's mailbox on top of SIMIX network
2529 The put/get mechanism was greatly simplified on the way.
2532 * New SIMIX network module. Provides:
2533 - Mailbox: rendez-vous mechanism to find with who you want to speak
2534 - Synchronous send/recv: easier and hopefully faster since the
2535 logic is handled in the maestro process directly now
2536 - Asynchronous send/recv: you dreamt of it? It's here now
2537 Too bad that nobody cared enough to propagate the change to MSG.
2538 * Add semaphores as SIMIX synchronization mechanism.
2541 * new function SD_daxload(char*) to load a DAX file
2542 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2543 * Introduce typed tasks. Specify its kind and cost at creation.
2544 At scheduling, just give where it should be placed, and the cost
2545 for each involved resource is automatically computed.
2546 Existing constructors so far (more to come of course):
2547 - SD_task_create_comm_e2e() for end-to-end communication
2548 - SD_task_create_comp_seq() for sequential computation
2549 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2550 * new function SD_task_dump() for debuging display
2551 * new function SD_task_dotty(task,FILE*) writing to file the info
2552 about the task in dotty format
2553 * SD_task_dependency_exists() can now cope with having one of its
2554 arguments NULL. If so, it tests whether the other argument has any
2556 * Add getters on list of preceding/following tasks:
2557 SD_task_get_parents(task) and SD_task_get_children(task)
2558 * Add getters on amount of workstations and list:
2559 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2560 * Add getter on task kind: SD_task_get_kind(task)
2561 * Update the start_time and finish_time of tasks on completion/failure
2562 * Bugfix: Remove task from state swags when destroyed
2565 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2568 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2569 dynar with one separate thread per value of the dynar.
2570 * Change the prototype of xbt_thread_create(), sorry.
2571 Added a boolean parameter indicating whether we want to join this
2572 thread (used in SG only for now)
2573 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2576 * GTNetS wrappers should now be usable again (and betterly tested too)
2577 * Fix a major regression from 3.2 where the timeout provided to
2578 MSG_task_put_with_timeout() was used as absolute time before which
2579 the comm should be done.
2580 * Start to fix the <cluster> tag.
2581 - Internal links should be good now (beside of the loopback, which
2582 use the private link instead)
2583 - paths to the external world is still rather broken
2584 - the <route:multi> tag is just broken. Actually that's brain-dead.
2585 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2586 to make it less stupid
2587 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2588 * Fix a source-level compatibility glitch from 3.2: after defining
2589 MSG_USE_DEPRECATED, you can use the old name
2590 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2591 * Allow to compile from the SVN with automake 1.11
2592 * Fix some problems when using the "start_time" tag in deployment XMLs.
2593 * Fix #8569: XBT/synchro.h has redundant declarations
2594 * Fix #8563: MSG return values and exceptions
2595 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2596 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2597 Thanks to Robson Peixoto.
2598 * Fix a vicious bug in dictionaries inducing that some elements were
2599 not freed on xbt_dict_free()
2601 Portability report of this version:
2602 * Main portability targets:
2603 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2604 on (amd64/i386/ia64)
2605 - mac leopard on i386
2606 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2607 but nothing critical.
2608 * Other platforms: windows, AIX and others were not tested for this release
2610 Timing report of this version:
2611 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2612 scenario parameters). From 8h to a few seconds in desktop grid settings.
2613 * Supernovae brings about 25% speedup on i386.
2615 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2617 SimGrid (3.3.3) stable; urgency=low
2619 The "Need for Speed" release.
2621 The timings done to validate the 3.3.2 were faulty.
2622 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2624 The problem was a conversion from a manually handled vector to
2625 xbt_dynar_t on the critical path.
2626 xbt_dynar_foreach calls functions, inducing stack management crap.
2628 We inlined these functions and xbt_dynar_foreach is now breath taking.
2629 We also inlined xbt_swag_belong on the way.
2631 Here are some approximate speedup measurements (on master/slaves
2632 simulations lasting between 10s and 20s each):
2633 3.3.1 -> 3.3.2: about same performance
2634 3.3.2 -> 3.3.3: 40% speedup
2635 3.3.1 -> 3.3.3: 40% speedup
2636 3.3.1 with inline patch -> 3.3.3: 30% speedup
2638 Our reading is that the refactoring which occurred in 3.3.2 made us
2639 suffer much more from the xbt_dynar_foreach low performance, but
2640 once we solved this, this refactoring proved to be very performance
2641 effective. From the 40% speedup, somehow, 10% are due to the
2642 inlining and 30% to the refactoring.
2644 That's a pitty that gcc cannot inline functions placed in other files
2645 alone. We have to choose between:
2646 - break the encapsulation (by putting private data structures and
2647 accessors in headers files to help gcc)
2648 - live with low performance
2649 - switch to a decent compiler such as icc (not quite possible).
2651 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2653 SimGrid (3.3.2) stable; urgency=low
2655 The "Simplicity does not preceed complexity, but follows it" release.
2657 The main contributors of this release were (lexical order):
2658 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2661 * Extract the routing logic into its own object.
2662 (was dupplicated in network.c and workstation_LV07.c;
2663 Allows to implement other ways of storing that info)
2664 => kill now useless network_card concept
2665 - Use dynar to represent routes (instead of void** + int*)
2666 - kill link_set (use surf_network_model->resource_set instead)
2667 - Add a command-line option to choose the routing schema to use
2668 - Add three new models:
2669 * Floyd (shortest path computed at initialization)
2670 * Dijikstra (shortest path recomputed all the time)
2671 * Cached Dijikstra (shortest path computed on need)
2672 All these models where contributed by Silas De Munck, and are
2673 described in his ICCS09 paper.
2675 * Simplify model declaration
2676 (less redirections, less function to write when defining a model)
2677 - Factorize stuff between models:
2680 surf_model_resource_set(model)
2681 surf_model_resource_by_name(model, name)
2682 - Unify the types of models in s_surf_model_t (using an union)
2683 - Embeed fields of common_public directly into s_surf_model_t
2684 - Rename model methods:
2685 action_free ~> action_unref
2686 action_change_state ~> action_state_set
2687 action_get_state ~> action_state_get
2688 - Change model methods into functions:
2689 (model)->common_public->action_use ~> surf_action_ref
2691 * Implement a generic resource; use it as ancestor to specific ones
2692 (allows to kill duplicated code in models)
2693 Drawback: timer command don't need no name nor properties;
2694 workstation_CLM03 don't need no properties
2695 (but I guess we can live with those few bytes wasted)
2697 * Improve the action object model
2698 - implement a constructor avoiding dupplicated code about field
2699 initialization in generic_action part.
2701 * Kill the SDP model: it has an external dependency, is deprecated
2702 in flavor of modern lmm models, and didn't compile since a while
2705 * Relocation of the context module from XBT to SIMIX.
2706 (the context were decoupled from the simix processes, duplicating a lot of code)
2707 => a lot of code was factorized
2708 - less overhead is introduced during scheduling
2709 - simpler API for the context factory
2710 - the logic for process creation,destruction and manipulation was simplified
2711 * Simplification of the s_smx_process_t data structure.
2712 => accesing the simix level data associated to a process is faster now,
2713 and the code is a lot more readable.
2716 * Implement some more MPI primitives:
2717 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2718 -implementation: Bcast: flat or 2-ary tree (default),
2719 Barrier: 4-ary tree,
2721 Allreduce: Reduce then Bcast
2722 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2723 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2724 Alltoallv: flat tree, like ompi
2726 * Add support for optimized collectives (Bcast is now binomial by default)
2727 * Port smpirun and smpicc to OS X
2730 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2731 Shout out if you used it.
2734 * Display the list of still queued messages in SG mode when existing
2738 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2739 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2740 * Add xbt_str_from_file(FILE*)
2741 * Add xbt_dict_get_key achieving a linear reverse search
2742 * Remove the context module
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 On these, we still have the eratic breakages of gras/pmm and
2751 amok/saturate_sg reported in previous version. We still think
2752 that the tests are the cause of the fault, not the tested code.
2754 - Mac OSX Leopard/x86/context
2755 Still false negative in tesh autotesting.
2756 Smpi still fails, but this time because readlink does not accept -f
2757 Everything seems to work properly beside of that.
2760 - AIX version 5.3 (only tested contexts this time)
2761 Smpi still fails there because mktemp is not installed.
2762 Everything seems to work properly beside of that.
2764 I managed to compile it for the first time, but several breakages.
2765 Won't delay the release for this exotic platform.
2767 * Windows: it's still lagging behind. If you want to help, please
2770 Timing report of this version:
2771 This version seem to be more than 5% faster than 3.3.1 (on linux
2772 64bits with contextes). The gain is less than expected, we are
2773 investigating this for next release.
2775 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2777 SimGrid (3.3.1) stable; urgency=low
2780 * Implement a --cfg-help to show existing configuration variables
2781 * Build chain do not require doxygen in maintainer mode
2784 * fix a bug on struct sizeof computation, which prevented the
2785 exchange of arrays of structs in some conditions
2786 - added a regression test about this in datadesc_usage
2787 * Allow the exchange of 0-long dynamic vectors.
2788 - for that, use -1 as indicator of dynamic size instead of 0
2789 - This implied to change any size from unsigned long to long,
2790 reducing a bit communication abilities, but I guess that with
2791 64bits being quite common, this is more than enough.
2792 - This also induce a protocol change, thus bumping network protocol
2793 version from 0 to 1 (if we have external users, we have to get
2794 clean on that point too ;)
2795 - added two regression tests about this in datadesc_usage
2796 * Be more verbose when propagating local exceptions
2797 This helps debugging.
2798 * Display the status of simulated processes when receiving SIGINT in
2802 * Allow to control the simulation from a trace file.
2803 New functions MSG_action_register() and MSG_action_trace_run()
2804 The first one allows to associate a function execution to each
2805 kind of action while the second one parses a trace file and
2806 triggers the corresponding actions within the system.
2807 For now, only a toy example is provided in examples/msg/actions
2808 * Add an exemple of process migration in examples/msg/migration
2809 * Fix a bug in task exchange which broke MSG_task_get_sender()
2810 Add a teshsuite regression test for that.
2811 [Bug: if MSG_task_get_sender() is called after sender exit,
2813 * Fix a bug which prevented suspend/resume to work properly
2814 * Display the status of simulated processes when receiving SIGINT
2815 This fixes a regression of v3.3. due to the introduction of SIMIX
2816 * Bug fixing in failure management:
2817 - trace could not start by a failure at time 0
2818 - failure during communications were not working
2821 * Add SIMIX_process_set_name() to change the name of the current
2822 process in the log messages.
2823 * Store smx_hosts in a dict since we only retrieve them by name
2824 * Move the configuration infrastructure to surf
2827 * Move the configuration infrastructure to surf
2830 * Massive internal cleanups:
2831 - Store internal structures on processes instead of hosts (allows
2832 to have more than one process per host, in addition of being more
2834 - Cleanup the initialization/finalization process
2835 - Kill a whole bunch of unneeded synchronization:
2836 processes run in exclusive manner within the simulator
2837 - Move queues from global tables to process data fields
2839 - now accept -platform and -hostfile arguments
2840 - Pass the right rank value to processes according to the hostfile
2841 * Compile the examples by default, and use them as regression tests
2842 * Implement MPI_Wtime()
2843 * Change the reference speed to a command line option
2846 * TCP_gamma can now be specified as command line option using
2847 --cfg=TCP_gamma:10000000.0
2848 * Change the --surf-path cmd line option into --cfg=path:
2851 * Also include strbuff from xbt.h public header
2852 * xbt_ex_display(): do not free the exception after displaying
2853 This allows to do more with the given exception afterward.
2854 Users should call xbt_ex_free() themselves.
2858 Portability report of this version:
2859 * Main portability targets:
2860 - Linux(debian)/x86/context
2861 - Linux(debian)/x86/pthreads
2862 - Linux(debian)/amd64/context
2863 - Linux(debian)/amd64/pthreads
2864 These targets fail about 1/10 of times on gras/pmm, but we believe
2865 that this is because of the test, not because of simgrid.
2866 amok/saturate_sg fails even more rarely, and the test may not be
2869 - Mac OSX Leopard/x86/context
2870 The test suite still spits tons of errors because some obscure
2871 force prevents us from removing the temporary directories
2872 arguing that they still contain some metadata I've never heard of.
2873 Smpi fails because seq is not installed.
2874 Everything seems to work properly beside of that.
2877 - AIX version 5.3 (both contexts and pthread)
2878 Smpi still fails there because mktemp is not installed.
2879 XML inclusions seems rosty on AIX.
2881 * Windows: it's still lagging behind. If you want to help, please
2884 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2886 SimGrid (3.3) stable; urgency=high
2890 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2891 [Malek Cherier & Mt]
2893 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2896 Until now, GRAS were using MSG as an interface to SURF. It was
2897 quite difficult because both interface have several differences
2898 (MSG channels vs GRAS sockets were the most notable point).
2900 This also opens the gate to SMPI (which should occur soon) and speed
2901 up simulations by to 40% (even if it were not the main goal).
2903 **************************************
2904 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2905 **************************************
2906 The point is that events occuring at the exact same timestamp are
2907 not scheduled in the same order with the old and new version. This
2908 may be enough to completely change the execution of simulations in
2909 some cases. Sorry for the inconvenience.
2911 * Cleanup and upgrade the XML format to push further scalability
2912 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2914 * Improve the testing infrastructure with tesh. Now a very large part of
2915 the code is tested not only by being run but also by checking that the
2916 output match an expected output [Mt].
2918 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2919 is really less memory-demanding, which should allow you to use
2920 larger files in SimGrid [AL].
2922 * Inform valgrind about our contextes, so that it becomes usable
2923 with the default (and more effecient) version of SimGrid
2924 [contributed by Sékou Diakite, many thanks]
2927 * Introduce a listener thread in charge of receiving incoming
2928 messages from the network. It allows to overlap communication and
2929 computation but most notably, it removes some stupid deadlocks due
2930 to the fact that so far, a process could not send and receive at
2931 the same time. This made most non trivial communication schema
2933 * Convert the PIDs from long int to int to match the MSG ones (and
2934 linux ones too) [Mt]
2935 * New function: gras_agent_spawn() to launch a new process on
2936 current host. Only working in simulation for now. [Mt]
2937 * New function: gras_os_hostport() returning a constant form (ie,
2938 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2941 * Make the backtrace of exceptions more human readable [Mt]
2942 * New module: xbt/str [Mt]
2943 a ton of string utility functions (split, join, printf to a newly
2944 allocated buffer, trim, etc)
2945 * New module: xbt/hash [Mt]
2946 SHA1 hashing algorithm (more to come if needed)
2947 * New module: xbt/synchro [Mt]
2948 synchronization tools (mutex and conditions) working the same way
2949 in simulation and in real life (mainly useful for GRAS, but not
2951 * New module: xbt/queue [Mt]
2952 classical producer/consumer synchronization scheme
2953 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2954 (using the classical functions will get serialized) [Mt]
2955 * Make dictionary internal table dynamic. No need to specify its size
2956 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2958 * Make sure the log channels are organized as a tree under windows
2959 (because of ANSI C compatibility issue, any channel were child of
2960 root directly) [Mt].
2963 * Cleaned many thing in surf and fixed a few bugs [AL].
2964 * Add a nice command line configuration mechanism to compose models [AL].
2965 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2966 the previous one (KCCFLN05). It relies on something that looks like
2967 a max-min sharing mechanism but cannot be written as such. A new solver
2968 was thus designed [AL].
2969 * Add a new solver to lmm. Based on Lagrange optimization and
2970 gradient-based descent, it enables to efficiently maximise systems s.a
2972 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2974 This solver enables to propose two new network models for TCP Reno and
2975 TCP Vegas based on Low's work. These models still need to be fully
2976 tested though [Pedro Velho].
2979 * Bug fix in SD_simulate. Now the time bound given as argument is
2981 * Use the new parallel task model (ptask_L07) as default.
2982 * Use the SURF command line configuration mechanism.
2983 * 0-size tasks (for synchronization) should now work.
2985 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2987 SimGrid (3.2) stable; urgency=high
2991 We still experience issues on this platform, but we believe that at
2992 least MSG is usable.
2994 GRAS API BREAKAGE (for simplification purpose, sorry):
2995 * the gras_msgtype_by_name is not used anymore. Instead of
2996 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2997 you can write (and must)
2998 gras_msg_send(toserver, "request", &request);
2999 - If you still want to pass a gras_msgtype_t to the function (to cache
3000 the type and avoid the lookup time), use the gras_msg_send_() variant.
3001 - Impacted functions:
3002 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3003 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3004 * The callbacks are now expected to return 0 when everything went well
3005 (just like the main() function)
3007 GRAS new features and improvements:
3008 * New module mechanism where user code can use per process globals [Mt]
3009 This is similar to gras_userdata_*() functions, but for libraries. It
3010 factorize some code developped over and over in the examples and AMOK.
3011 It has still to be documented and used (only amok/peermanagement is
3013 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3014 loss when they were small enough to fit into the buffer and sent quickly
3015 enough so that they can all get received in one shoot.
3016 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3017 if not found. Use the *_or_null() variant for the old semantic.
3018 * In gras_msg_handle, do not discard messages without callback.
3019 They are probably messages to be explicitly awaited later (ie, proofs of
3020 mis-synchronization in userland since they are sent before being awaited)
3021 No big deal usually.
3022 * gras_socket_meas_send/recv: semantic changed!
3023 The numerical arguments used to be (1) the total amount of data to send
3024 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3025 messages. This was need for the fool willing to send more than MAXINT
3026 bytes on quite fat pipes.
3029 * Do really rename the hostmanagement module to peermanagement. [Mt]
3030 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3031 but this is rather new and this was documented in the module
3032 documentation (poor excuses, I admit)
3033 * Bandwidth measurement semantic changed! This follows the changes to
3034 gras_socket_meas_send/recv explained above.
3037 * A sequential mode has been added to the workstations. When a workstation
3038 is in sequential mode, it can execute only one task, and the other tasks
3039 are waiting in a FIFO. [Christophe Thiery]
3042 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3043 model for SIMDAG. [Christophe Thiery]
3044 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3045 equal to 0, which caused some bad behaviors in
3046 saturated_constraint_set_update. I now use a threshold mechanism like in
3050 * When running manually src/testall, you select specific units [Mt]
3051 testall is the result of our cunit mechanism, and should replace all
3052 the scripty thingy around since bash don't run easily on billware.
3054 * A mallocator system has been added. [Christophe Thiery]
3055 Mallocators allow you to recycle your unused objects instead of freeing them
3056 and allocating new ones.
3058 Documentation update:
3059 * FAQ reworking + New FAQs:
3060 - "Valgrind spits tons of errors!" [Mt]
3061 - "How to repport bugs" [Mt]
3062 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3063 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3064 - Communication time measurement within MSG [AL]
3065 - I experience weird communication times when I change the latency [AL]
3066 * GRAS tutorial [Mt]
3068 - an introduction to the framework and to the used communication model
3069 - an initiatic tour introducing the most proheminent features:
3071 . Lesson 0: Installing GRAS
3072 . Lesson 1: Setting up your own project
3073 o Part 2: Message passing
3074 . Lesson 2: Exchanging simple messages
3075 . Lesson 3: Passing arguments to the processes (in SG)
3076 . Lesson 4: Attaching callbacks to messages
3077 . Lesson 5: Using globals in processes
3078 . Lesson 6: Logging information properly
3079 . Lesson 7: Using internal timers
3080 . Lesson 8: Handling errors through exceptions
3081 . Lesson 9: Exchanging simple data
3082 . Lesson 10: Remote Procedure Calling (RPC)
3083 . Lesson 11: Explicitely waiting for messages
3084 . Recapping of message passing features in GRAS
3085 - A HOWTO section containing:
3086 o HOWTO design a GRAS application
3087 More are due, of course. They will come latter. In the meanwhile, you can
3088 check the examples which are still here.
3090 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3092 SimGrid (3.1) stable; urgency=high
3096 There was a stack corruption somewhere, visible only when optimizing
3097 with these versions. [Vince]
3100 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3101 module that enabled to play with Directed Acyclic Graphs. It is built
3102 directly on top of SURF and provides an API rather close to the old
3103 SG. Some old codes using SG are currently under rewrite to check that
3104 all needful functions are provided. [Christophe Thiery]
3107 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3108 extension of the classical CLM03 model that gracefully handles
3109 failures. This is now the default model for MSG and GRAS. It doesn't
3110 handle parallel tasks yet though. [AL]
3111 * Bug fix: Weights were not correctly set in the network part.
3112 WARNING: This may have resulted in incorrect results with simulations
3113 where there are more than one flow on a given link. [AL]
3116 * After a (long ?) discussion on simgrid-devel, we have decided that the
3117 convention we had on units was stupid. That is why it has been decided
3118 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3119 WARNING: This means that all previous platform files will not work as
3120 such with this version! A warning is issued to ask users to update
3122 A conversion script can be found in the contrib module of the CVS, under
3123 the name contrib/platform_generation/surfxml_update.pl [MQ]
3126 * Bug fix: Processes were started in reverse order, wrt deployment file.
3127 WARNING: if your code relies on this bug, please fix it. [AL]
3128 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3129 being executed on two different locations. [AL]
3130 * Bug fix: Failures are now better supported thanks to Derrick's tests
3131 (there was many failure situations I hadn't thought of and that weren't
3132 correctly handled). [AL]
3133 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3137 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3138 Exception occurring on server-side are propagated back to client (!).
3140 API CHANGE: the callback changed their prototype. Change:
3141 int my_handler(gras_socket_t expeditor, void *payload_data) {
3143 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3144 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3146 * New! function: gras_msg_handleall to deal with all messages arriving
3147 within a given period.
3148 * New! function: gras_socket_server_range to get a server socket in a
3149 range of port numbers (ease to avoid port number conflicts) [MQ]
3150 * New! gras processes display their backtrace when they get a SIGUSR1
3151 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3152 Sweet to debug RL processes [MQ]
3156 - Do not force experiment sizes to be expressed in kb, or it becomes
3157 impossible to measure the latency this way (needs one byte-long tests)
3158 WARNING: this changes the amok_bw_* function semantic. [MQ]
3159 - Implements the link saturation stuff. [MQ]
3160 * Peer management module:
3161 New! module factorizing code that we wrote over and over [MQ].
3164 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3165 - Test units are placed directly into the library code, they get extracted
3166 automatically and placed into the src/testall binary.
3167 - Convert most of the XBT tests to this system.
3168 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3169 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3170 live in the same C code. This required to change a little bit the API
3171 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3172 * New module: sparse graph structure with basic algorithms (this is work
3173 in progress and the API is not considered to be frozen yet). [AL]
3174 * Display more information on backtraces: source line & function names are
3175 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3176 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
3177 * Reworked a little bit some #include statements to load only required
3178 headers. Some user code that relied on SimGrid to include stdlib or
3179 stdio may need to include it by themselves. [AL]
3180 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3182 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3183 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3184 but I guess that almost nobody used those parts. [MQ]
3186 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3188 SimGrid (3.0.1) stable; urgency=low
3191 * Unfortunately, I had missed 5 misnamed functions:
3192 xbt_fifo_item_t xbt_fifo_newitem(void);
3193 void xbt_fifo_freeitem(xbt_fifo_item_t);
3194 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3195 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3196 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3197 They're now deprecated. Please use their new versions:
3198 xbt_fifo_item_t xbt_fifo_new_item(void);
3199 void xbt_fifo_free_item(xbt_fifo_item_t);
3200 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3201 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3202 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3204 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3205 * Documentation: xbt_log module unmercifully reworked [MQ]
3206 * Bugfix: there was a problem with the ending of contexts with
3207 the pthread backend. It caused some weird deadlock or behavior
3208 depending on the pthread implementation. [AL]
3209 * Bugfix: get the exceptions raised in the simulator repport where
3210 and why they come from when they are not caught in time [AL, MQ]
3213 * Bugfix: Do repport the error when two non-connected hosts try to
3214 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3217 * Add additionnal checkings on communications. Assert that two
3218 communicating hosts are connected by a set of links... [AL]
3221 * Add additionnal checkings on channel values in communication [AL]
3222 * New: MSG_task_get_source to see on which host a task was generated [HC]
3223 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3224 the number of tasks waiting to be received on channel and sent
3226 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3227 waits for the first task coming from a given host.. [AL]
3229 GRAS new functionnalities: [MQ]
3230 * Enhance the parsing macro to allow the size of multidimentional objects
3231 to be given thru annotations.
3232 * New example (and documentation): Matrix Multiplication a la RPC
3233 (as when I was young!) and fix a bunch of bugs found on the way.
3235 GRAS performance improvements: [MQ]
3237 * Reduce the amount of cbps creation/destruction by making it static to
3238 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3241 * Change libdata to a set so that we can search for stuff by ID (and thus
3242 reduce the insane amount of dict lookups)
3245 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3246 RL communication on top of SG since it's so uneffective.
3247 It may also be used for inter-thread communication in RL, one day.
3248 * Use gras_datadesc_copy() to exchange messages on top of SG
3250 - improve message exchange performance on top of SG
3251 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3252 * Don't exchange on the network the size of the used part of buffer,
3253 instead, specify the possible buffer size to read().
3255 - reduces the amount of read/write calls (one pair per exchange)
3256 - reduces the amount of exchanged data (the size)
3257 - allows to retrieve all arrived data on receiver side, if we don't need
3258 it right now (subsequent read will peek the buffer)
3259 - allows the receiver to proceed with the begining of the stream before
3260 everything is arrived
3261 - make it possible to build an iov transport (using readv/writev)
3263 - take care of the data with non-stable storage (like stacked data),
3265 * If possible, TCP send uses vector I/O (when writev() is here)
3266 - Don't use it for receive since we send data sizes and data on the
3267 same stream, so we wouldn't be able to chain large amount of chunks
3268 before having to flush the stuff to read the size.
3269 * Rework the transport plugin mechanism to simplify it and reduce the
3270 amount of pointer dereferencement when searching for the right function
3273 * I guess that now, we do almost as few system calls as possible while
3274 doing as few data copy as possible.
3276 To improve it further, we could try to send all the sizes first and then
3277 all the data (to use iov on receiving size), but it's only a partial
3278 solution: when you have 2 dimensional data, the sizes of the second
3279 dimension is data of the first dimension, so you need 3 streams.
3281 I'm not sure the potential performance gains justify the coding burden.
3283 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3285 SimGrid (3.00) stable; urgency=high
3288 * New! Give the possibility to hijack the surf parser and thus bypass
3289 MSG_create_environment and MSG_launch_application. Have a look at
3290 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3292 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3294 SimGrid (2.96) unstable; urgency=low
3299 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3300 This deprecates the xbt_error_t mechanisms.
3301 It modifies (simplifies) all XBT and GRAS API.
3302 MSG API keeps unchanged (exceptions raised by XBT are caught from
3303 within MSG and masked with existing error handling facilities)
3306 * New! Add a FATPIPE model. [AL]
3307 * New! Add a parallel task model. [AL]
3308 * New! Add automatically a loopback interface (in the default
3309 network model) if none was precised.
3312 * Bugfix: MSG_process_resume now works with the current running process.
3314 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3315 * Modification of MSG_task_get_compute_duration. Once a task has been
3316 processed, the value returned by this function is now equal to 0. [AL]
3317 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3318 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3319 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3320 of a m_task. For now, MSG_task_cancel only works with computation
3322 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3323 of the processor (in Mflop/s) regardless of the current load on the
3325 * API Change: use proper naming convention for MSG_getClock and
3326 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3328 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3329 This function changes the priority of a computation task. This priority
3330 doesn't affect the transfer rate. A priority of 2 will make a task
3331 receive two times more cpu power than the other ones. This function
3332 has been added to suit the needs of Nguyen The Loc and hasn't been that
3333 much tested yet. So if it fails, please report it and send me your code.
3335 * API Change: removed all functions and types that were marked "deprecated"
3336 since many months. Renamed MSG_global_init_args to MSG_global_init.
3338 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3340 SimGrid (2.95) unstable; urgency=low
3343 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3344 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3345 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3348 * Bugfix: Allow absolute paths for platform description files [MQ]
3349 * Bugfix: do free the variables after use. Leads to drastic performance
3351 * Implement max_duration (ie, timeouts) on resources [AL]
3354 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3356 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3357 the needed support (timeouts) [AL]
3360 * Implement measurement sockets. You can now get the bandwidth between two
3361 hosts thanks to AMOK (see below). [MQ]
3362 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3363 dynar over the network (yeah) [MQ]
3364 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3365 * simple/static token-ring example. [Alexandre Colucci and MQ]
3366 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3367 in place before [MQ]
3370 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3371 type parsing. "short" and "long" modifiers were also ignored; other
3372 modifier (such as reference level) are still ignored. [MQ]
3373 * Embeed the buffer size within the buffer itself on SG. [MQ]
3374 That way, send() are atomic and cannot get intermixed anymore (at least
3375 the ones which are less than 100k; bigger messages still have the issue)
3376 * Array size pushed by the field, not by the field type (or each
3377 and every long int will push stuff to the cbps) [MQ]
3378 * use select() to sleep since it allows to portably sleep less than one
3381 GRAS (minor cleanups)
3382 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3385 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3386 * type category 'ignored' killed as it was never used and were difficult
3388 * whether a type can cycle or not is now a flag, leaving room for more
3389 flags (as "ignored", if we feel the need one day ;)
3390 * Rename raw sockets to measurement sockets since "raw" has another
3391 meaning in networking community.
3394 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3395 and offers features not belonging to GRAS but that most applications
3396 need. One day, it may be a set of plugins loadable at runtime.
3397 * New module: bandwidth
3398 bandwidth measurement between arbitrary nodes running this module. [MQ]
3400 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3402 SimGrid (2.94) unstable; urgency=low
3404 The first beta release of SimGrid 3 !
3408 * Update the main page and the FAQ. Adding references to gforge.
3411 * Add a gras_os_getpid function.
3414 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3415 * Extend the logs so that they also print PID, hostname, date, ... if
3417 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3418 and kill the old version which were in testsuite/
3419 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3423 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3424 for users. Instead, each programming environment (GRAS, MSG,...) use it
3425 when needed to provide such a feature to users.
3426 Don't shortcut the mechanism or you will also shortcut the virtualization
3427 you need on the simulator.
3431 * Cleanups in configury with regard to compile optimization/warning flags.
3432 Also add -fno-loop-optimize to any powerpc since it's the optimization
3433 killing gcc (< 3.4.0).
3434 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3435 needed by MSG examples complications
3436 * Borrow configury beautifications from PHP
3439 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3440 warning (thanks loris for stumbling into this one).
3441 * Bugfix: stop loading private headers (gras_config.h) from the public
3445 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3446 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3447 RL would imply a DNS resolver, which is *hard* to do in a portable way
3448 (and therefore delayed).
3449 * Implement a real timer mechanism and use it in timing macros. This allows
3450 to avoid rounding errors and get a 0.000005 sec precision in timing
3451 macros. While I was at it, various cleanups:
3452 - allow to declare more than one timed section per file (fix a stupid bug)
3453 - move some private declaration to the right place
3454 - merge conditional execution and timing macros into emulation module
3455 - document the module
3456 - make sure the module cleanups its mess on gras_exit
3457 * Documentation improvements:
3458 - (new) how to compile applications using GRAS
3459 - (new) emulation support (timing macros)
3461 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3463 SimGrid (2.93) unstable; urgency=low
3465 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3468 - Use Paje properly where used. Still to be sanitized properly.
3469 - Portability fix: Add an implementation of the contexts using pthread
3473 - Add xbt_procname(): returns the name of the current process.
3474 Use it to show the current process's name in all logging.
3476 - fix detection of older flex version and the reaction, since we do
3477 depend on modern ones (we use lex_destroy)
3478 - Better separation of SG and RL in the libs: remove all simulation code
3479 from libgras. As a result, this lib is now only 200k when stripped.
3480 Some of the xbt modules may also be duplicated (two sets and such) and
3481 should be cleaned/killed before SG3.
3482 - Insist on using xlC on AIX because of weird problems involving gcc there.
3483 - Cleanup the make remote stuff. This is now done by scripts
3484 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3485 mainly for our private use, but we're working on changing them to user
3488 - Bugfix: flush the socket on close only if there is some *output*.
3489 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3491 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3494 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3496 SimGrid (2.92) unstable; urgency=low
3498 Alpha 3 on the path to SimGrid 3
3502 - New! First try of benchmarking macros.
3503 - New! First try so that gras_stub_generator generate deployment and
3504 remote compilation helpers.
3506 - Bugfix: Initialization fix in msg_test.
3510 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3512 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3514 - API Change: gras_os_sleep to take the amount of seconds as a double.
3515 Accepting an int was error prone since it was the only location where
3516 seconds were coded as such. It leaded to damn rounding errors.
3517 - Bugfix: Hard to belive that timers ever worked before this.
3519 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3521 SimGrid (2.91) unstable; urgency=low
3523 Alpha 2 on the path to SimGrid 3
3527 - Bug fix in the lmm_solver.
3529 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3530 through the function MSG_paje_output.
3531 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3532 - It is possible to bound the rate of a communication in MSG with
3533 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3534 to write it in the changelog).
3535 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3540 - Various cleanups to the autotools stuff
3541 - Begin to move Gras examples to examples/gras/
3542 - Let make distcheck work again (yeah!)
3544 - documentation overhauled using doxygen.
3545 gtk-doc-tools is dead in SimGrid now.
3546 - Automatically extract all existing logging categories, and add the list
3547 to the documentation (long standing one, to say the less)
3549 - Cleanup the known architecture table. Reorder the entries to group what
3550 should be, and use a more consistent naming scheme.
3551 (some of the test dataset are still to be regenerated)
3552 - New! Allow library to register globals on each process just as userdata
3554 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3555 lookup time (for now).
3556 Use it in msg and trp.
3557 This cleans a lot the internals and helps enforcing privacy of the
3558 headers between the gras components.
3559 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3560 - Bugfix: gras_os_time was delirious in RL.
3561 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3563 - Reenable GRAS now that it works.
3565 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3567 SimGrid (2.90) unstable; urgency=low
3569 Alpha 1 on the path to SimGrid 3
3571 * It is a long time since the last release of SimGrid. I'm sorry about
3572 that but as I had told you, I was rewriting a lot of things. I apologize
3573 to those who had been reporting bugs to me and that I had not answered.
3574 If your bug is still in the new version, please tell me. Here is a
3575 summary of the main changes.
3577 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3578 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3579 portability, speed, and a lot more but you'll figure it out later.
3580 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3581 I wanted to release the new MSG as soon as possible and I have
3582 broken GRAS, which is the reason why, for now, only MSG is fully
3583 functional. A laconic description of these projects is available
3584 in the documentation.
3586 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3587 kernel optimized for our needs (called SURF but only the developers
3588 should use it). Hence, MSG is now roughly 30 times faster and I think
3589 that by rewriting a little bit MSG, I could event speed it up a little
3590 bit more. Beside the gain in speed, it is also much easier to encode a
3591 new platform model with SURF than it was with SG. More to come...
3593 * REVOLUTION 3: I have tried to change a little as possible the API of
3594 MSG but a few things really had to disappear. The main differences
3595 with the previous version are:
3596 1) no more m_links_t and the corresponding functions. Platforms are
3597 directly read from a XML description and cannot be hard-coded
3598 anymore. The same format is used for application deployment
3599 description. The new format is described in the documentation.
3600 Have a look in tools/platform_generation. There is a tiny script
3601 that converts from the old platform format to the new one. Concerning
3602 the application deployment format, parsing the old one is tricky.
3603 I think most of you should however be able to convert your files. If
3604 it is really an issue, I can write a C code that does the conversion.
3606 2) the toolbox tbx does not exist anymore. We now have a library
3607 with much more data-structures but without the hash-tables (we have
3608 dictionaries that are much faster).
3610 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3612 *****************************************************************************
3613 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3614 * don't feel like converting the oldies. (Mt) *
3615 *****************************************************************************
3618 Version 2.90: "the long awaited one"
3619 - Finished rewriting and debugging MSG. Rewrote the documentation.
3620 - disable GRAS for now since it needs to be ported to the newest SG
3623 - Finish the port to windows (using mingw32 for cross-compile)
3626 - Main loop and datastructures of SURF. A cpu resource object is
3627 functional. Surf can thus be used to create cpu's with variable
3628 performance on which you can execute some actions.
3630 2004-11-15 Martin Quinson
3631 - Port to ARM. Simply added the alignment and size descriptions. Should
3632 work, but the ARM machines are so slow that I didn't had the opportunity
3633 to 'make check' over there yet.
3635 2004-11-15 Arnaud Legrand
3636 - Trace manager now written. It uses a heap structure and is therefore
3637 expected to be efficient. It may however be speeded up (particularly
3638 when many events occur at the same date) by using red and black
3639 trees. One day maybe...
3640 - Max-min linear system solver written. It uses a sparse matrix
3641 structure taking advantage of its expected use. Most operations are
3642 O(1) and free/calloc are called as few as possible. The computation of
3643 the minimum could however be improved by using a red and black tree
3646 2004-11-03 Arnaud Legrand
3647 - Rename every gras_* function that was in xbt/ to its xbt_
3649 - Add a heap and a doubly-linked list to xbt
3650 - Added a dichotomy to the dictionaries. make check works as well before
3651 so I assume that the patch is correct. I do not know however if things
3652 run effectively faster than before now. :)
3654 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3655 SimGrid, and the version number is bumped to 2.x
3657 2004-10-29 Martin Quinson
3658 - Introduction of the remote errors.
3659 They are the result of a RMI/RPC on the remote machine.
3660 ErrCodes being scalar values, you can't get the host on which those
3661 errors did happen. Extending the error mechanism as in Gnome is possible.
3662 No idea yet whether it is a good idea.
3664 2004-10-28 Martin Quinson
3665 - Interface revolution: the Starred Structure Eradication.
3666 I used to do typedef struct {} toto_t; and then handle *toto_t.
3667 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3668 - ???_t is a valid type (builded with typedef)
3669 - s_toto_t is a structure (access to fields with .)
3670 - s_toto is a structure needing 'struct' keyword to be used
3671 - e_toto_t is an enum
3672 - toto_t is an 'object' (struct*)
3674 typedef struct s_toto {} s_toto_t, *toto_t;
3675 typedef enum {} e_toto_t;
3676 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3677 s_toto_t) is private.
3679 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3680 so that it can set the variable to NULL. It was so for dicts and sets,
3681 it changed for dynars.
3683 - Fix a bunch of memleaks in dict_remove
3684 - Fix a bug in sg/server_socket opening: it failed all the time.
3686 2004-10-07 Martin Quinson
3687 - Speed up dynar lookup operation a bit.
3689 gras_dynar_get is dead.
3691 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3692 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3693 (which gives you the address of the stored data).
3695 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3696 copy of the data using an affectation to do the job and not a memcpy.
3698 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3699 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3701 It does not really speedup the dynar test because they are
3702 setting elements all the time (and look them seldom). But the dict does
3703 far more lookup than setting.
3705 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3707 2004-10-05 Martin Quinson
3708 - Allow to (en/dis)able the cycle detection at run time.
3710 Whether we should check for cycle or not is now a property of each
3711 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3712 datadesc_cycle_unset allow to remove this property when previously set.
3714 Note that the cycle detection is off by default since it impacts the
3715 performance. Watch the data you feed GRAS with ;)
3717 This property is hereditary. Any element embedded in a structure having it
3718 set have it set for the time of this data exchange.
3720 You should set it both on sender and receiver side. If you don't set it on
3721 sender side, it will enter an endless loop. If you forget on receiver
3722 side, the cycles won't be recreated after communication.
3724 - Header reorganization.
3725 Kill gras_private.h, each submodule must load the headers it needs.
3727 2004-10-04 Martin Quinson
3728 - Interface revolution: do not try to survive to malloc failure.
3730 Now, gras_malloc and friends call gras_abort() on failure.
3731 As a conclusion, malloc_error is not a valid error anymore, and all
3732 functions for which it was the only gras_error_t return value are
3733 changed. They now return void, or there result directly.
3734 This simplify the API a lot.
3736 2004-09-29 Martin Quinson
3737 - Re-enable raw sockets.
3738 Created by gras_socket_{client,server}_ext;
3739 Used with gras_raw_{send,recv}
3742 It should allow to kill the last bits of gras first version soon.
3744 This is not completely satisfactory yet (duplicate code with
3745 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3748 - Simplify transport plugin (internal) interface by not passing any
3749 argument to _server and _client, but embedding them in the socket
3752 2004-09-28 Martin Quinson
3753 - Finish the port to AIX.
3754 autoconf was my problem (segfault within the malloc replacement
3755 function. No idea why)
3757 2004-09-16 Martin Quinson
3758 - Fix some size_t madness on 64bit architectures.
3760 2004-09-08 Martin Quinson
3761 - Reduce the number of system headers loaded, overload some more system
3762 calls (such as malloc to cast the result of the system one, and work
3764 - Fix and reintroduce the config support
3766 2004-09-07 Martin Quinson
3767 - Source code reorganization to allow Arnaud to surf all over there.
3768 - Allow to document the logging categories.
3769 - Remove all uppercase from logging categories and useless cleanup in names.
3771 2004-08-18 Martin Quinson
3772 Version 0.6.2 (protocol not changed; API changed)
3773 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3774 gras_error_t), and NULL when not found. Functions expecting a msgtype
3775 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3776 hopefully usefull message.
3777 - Portability to prehistoric sparcs again
3779 2004-08-17 Martin Quinson
3780 Version 0.6.1 (protocol not changed; ABI not changed)
3781 - prealloc some buffers to speed things up
3783 2004-08-11 Martin Quinson
3784 Version 0.6 (protocol not changed; ABI expended)
3785 - The parsing macro can deal with the references, provided that you add
3786 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3788 2004-08-09 Martin Quinson
3789 Version 0.5 (protocol not changed; ABI changed)
3790 - Allow to off turn the cycle detection code in data exchange at
3791 compilation time. It should be at run time, but I'm short of time (and
3792 the config stuff is still broken). That way, we keep dict out of the
3793 critical path, which is good because the performance is poor:
3794 - search not dichotomial yet
3795 - dynar give no way to access their content and memcpy everytime
3796 - In composed data description (struct, ref and so on), stop foolness of
3797 keeping the subtype's ID, but store the type itself. This keeps sets out
3798 of the critical path, which is good since they rely on dynar and
3799 dictionnaries. The only loose of that is that we cannot detect the
3800 redeclaration of a structure/union with another content (but I'm not sure
3801 the code detected well this error before anyway). We still can detect
3802 the redefinition discrepancy for the other types.
3803 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3804 it breaks the code because of type-punning used all over the place).
3805 This breaks on all non-gcc architectures (for now).
3807 All those changes (plus the buffer of last time) allow me to gain 2 order
3808 of magnitude on cruel tests consisting of 800000 array of integers on two
3809 level of a hierarchical structure (200 secondes -> 4 secondes)
3812 - the selector of reference must now return the type it points to, not
3813 the ID of this type.
3815 2004-08-06 Martin Quinson
3816 Version 0.4 (protocol changed; ABI not changed)
3817 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3818 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3819 - Implement a buffer transport, and use it by default (it relies on tcp in
3820 real life and on sg in simulation).
3821 That's a bit hackish since I had a new field to the structure to store
3822 its data without interfering with the subtype ones. Inheritance
3823 is tricky in C. And that's a kind of reverse inheritance with one class
3824 derivating two classes. Or maybe a game with java interfaces. Anyway,
3825 that's damn hard in C (at least).
3826 Moreover, I got tired while trying to ensure plugin separation and
3827 genericity in SG mode. MSG wants me to do weird things, so let's go for
3828 cruel hacks (temporarily of course ;).
3829 See comment in transport_private.h:71
3830 - do not include all the _interface headers in private but in the files
3831 which really need them (to cut the compilation time when they are
3834 2004-07-26 Martin Quinson
3835 Version 0.3 (protocol not changed; ABI changed)
3836 - Major overhault of the datadesc interface to simplify it:
3837 - shorted the function names:
3838 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3839 - add a trivial way to push/pop integers into the cbps without malloc.
3840 This allows to make really generic sub_type description, which simply
3841 pop their size of the stack.
3842 - add a function gras_datadesc_ref_pop_arr() which does what users want
3843 most of the time: Declare a dynamic array (which pops its size of the
3844 stack) and declare a reference to it. Poor name, but anyway.
3845 - kill the post-send callback, add a post-receive one
3847 2004-07-23 Martin Quinson
3848 Version 0.2 (protocol changed; ABI changed)
3849 - add some testing for cpbs in the test cases, and fix some more bugs.
3850 This invalidate again the little64 data file, since I cannot regenerate
3852 - remove an awfull optimization in the logging stuff, allowing me to:
3853 - understand it again
3854 - learn gcc how to check that the argument match the provided format
3855 - fix all errors revealed by gcc after that
3856 - internal keys of dict are not \0 terminated. Deal with it properly in
3857 loggings instead of segfaulting when the user want to see the logs :-/
3859 2004-07-22 Martin Quinson
3860 - Fix some stupid bug preventing cbps (callback postit) from working
3862 2004-07-21 Martin Quinson
3863 - Some documentation cleanups
3864 - remove the useless last argument of msgtype_declare
3865 - rename the Virtu functions to fit into the 'os' namespace
3866 - move headers src/include -> src/include/gras/ and stop fooling with
3867 gras -> . symbolic link
3868 - make distcheck is now successful
3870 2004-07-19 Martin Quinson
3872 - Build shared library also
3873 - Install html doc to the right location
3874 - stop removing maintainer files in make clean
3875 - build tests only on make check
3877 2004-07-13 Martin Quinson
3879 - No major issue in previous version => change versioning schema
3880 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3881 corresponding dataset.
3883 2004-07-11 Martin Quinson
3885 - Get it working with any kind of structure (we can compute the padding
3886 bytes remotely for all the architectures I have access to)
3887 - Implement the structure parsing macro (still not quite robust/complete)
3888 - Improvement to the remote testing toysuite
3890 2004-07-10 Martin Quinson
3891 [autoconf mechanism]
3892 - get ride of a bunch of deprecated macros
3893 - actually run the test for two-compliment, not only compile it :-/
3894 - test whether the structures get packed (and bail out if yes. Damn.
3895 Alignment is a serious matter)
3896 - test whether the structures get compacted (but respecting the alignment
3897 constraints of each types)
3898 - test whether the array fields of structures can straddle alignment boundaries
3900 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3901 (took me 2 hours to find that bug, looking at the wrong place)
3902 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3903 before sending/receiving and are used to compute the offsets of fields
3904 - Given that padding size depend even on compiler options, keep track of
3905 alignment and aligned_size only for the current architecture. Not a big
3906 deal since we send structure fields one after the other (seems
3908 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3909 program, let it work on linux/gcc/little32. portability todo.
3911 2004-07-08 Martin Quinson
3912 - import and improve remote compilation support from FAST
3913 - make sure make check works on half a dozen of machines out there
3915 2004-07-07 Martin Quinson
3916 Let's say it's version 0.0.3 ;)
3917 - Implement conversions (yuhu!)
3918 - Let it work on solaris (beside conversion, of course)
3919 - Stupid me, using rand() to generate the conversion datatests in not wise.
3921 2004-07-06 Martin Quinson
3922 - Let make dist work, since I'm gonna need it to compile on remote hosts
3923 - Let Tests/datadesc_usage write the architecture on which the file was
3924 generated as first byte.
3925 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3926 architecture descriptions.
3927 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3928 on those architectures.
3929 - Optimization: send/recv array of scalar in one shoot
3931 2004-07-05 Martin Quinson
3932 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3934 - Plug a whole bunch of memleaks
3935 - each process now have to call gras_{init,exit}. One day, their log
3936 settings will be separated
3937 - Continue the code factorisation between SG, RL and common in Transport.
3939 2004-07-04 Martin Quinson
3941 - Redistribution between SG and RL.
3942 We wanna have to accept in SG, so move accepted related parts of RL in
3943 the common part. (more precisely, the dynar of all known sockets is no
3944 more a static in transport.c, but part of the process_data)
3946 [gras_stub_generator]
3947 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3949 2004-07-03 Martin Quinson
3950 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3951 them all up in one shot)
3953 - Fix the ugly case of reference to dynamic array.
3954 - New (semi-public) function gras_datadesc_size to allow the messaging
3955 layer to malloc the needed space for the buffer.
3957 - gras_socket_close now expect the socket to close (and not its address to
3958 put NULL in it after it). This is because the socket passed to handlers
3959 is one of their argument (=> not writable).
3961 - propagate the interface cleanup from last week in datadesc, ie remove a
3962 superfluous level of indirection. User pass adress of variable
3963 containing data (both when sending and receiving), and not of a variable
3964 being a pointer to the data. Let's say that I like it better ;)
3965 The price for that is constructs like "int msg=*(int*)payload" in
3966 handlers, but it's a fine price, IMHO.
3968 - Let it work in RL (yuhu)
3970 2004-06-21 Martin Quinson
3972 - porting SG plugin and SG select to new standards (works almost).
3973 - plug memleaks and fix bugs around.
3976 - cleanup the prototype of data recv and force users to specify when they
3977 want to handle references to objects. Test case working even for cycles.
3978 - plug memleaks. Valgrind is perfectly ok with this.
3980 2004-06-12 Martin Quinson
3982 - cleanup the separation between plugin and main code in plugin creation
3984 2004-06-11 Martin Quinson
3986 - Reput hook for raw sockets, needed for BW experiments
3987 - kill a few lines of dead code
3988 [Data description] Interface cleanup
3989 - gras_datadesc_by_name returns the searched type or NULL.
3990 That way, no variable is needed to use a type desc once, which makes
3992 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3993 two parameters were strings (field name, type name), leading to
3995 [Dicos] Interface cleanup
3996 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3997 This is consistant with the dynar API.
3999 2004-04-21 Martin Quinson
4001 - Porting to new standards.
4003 - interface cleanup.
4004 There is no bag anymore, no need to take extra provision to mask the
4005 pointers behind "ID".
4006 Better splitup of functions between files create/exchange/convert.
4007 This is still a bit artificial since convert and receive are so
4008 interleaved, but anyway.
4010 - add a queued message list to procdata (the ones not matching criteria
4012 - factorize some more code between SG and RL wrt procdata
4014 - use gras_exit in example to track memleaks
4015 - get rid of gs_example now that GS is properly integrated into gras
4016 - update run_test to integrate the latest tests (datadesc)
4018 - rename WARNINGn macros to WARNn since it prooved error-prone
4020 2004-04-19 Martin Quinson
4022 - register init/exit functions within gras module mechanism
4023 - send/receive function.
4024 Convertion is not implemented, but short-cutted if not needed.
4025 struct/array elements are sent one by one (instead of block-wise), but
4026 nobody really cares (yet). Get a prototype before optimizing.
4027 - tests (using a file socket) for DD send/receive on:
4028 - base types: int, float
4029 - array: fixed size, string (ie ref to dynamic string)
4030 - structure: homogeneous, heterogeneous
4031 - chained list, graph with cycle
4032 Believe it or not, valgrind is not too unhappy with the results. The
4033 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4037 - Bugfix when using a filename explicitely (instead of '-')
4039 2004-04-09 Martin Quinson
4041 - factorize more code between RL and SG in socket creation
4042 - Complete the implementation and tests of:
4044 o file (only in RL, and mainly for debugging)
4046 I lost 3 days to design a portable address resolver, and then decided
4047 that the prototype mainly have to run on my box.
4048 Addressing portability too early may be like optimizing too early :-/
4050 - use gras_init in the Tests instead of the crappy parse_log_opt
4051 (the latter function is removed)
4052 [Conditional execution]
4053 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4054 [Code reorganisation]
4055 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4056 Build examples against the RL library, since there is no way to disable
4057 its creation for now.
4059 For information, the beginning of coding on GRAS was back in june
4060 2003. I guess that every line has been rewritten at least twice since