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 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
63 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
66 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
67 and <mount> tags. The rationale is finish to decouple what should be handle
68 by SimGrid kernel (i.e., resources and activities) from plugins built in
69 user space (e.g., the file system plugin). This new tag currently accepts
70 only three attributes (id, read_bw, and write_bw). All the other information
71 that was declared with the storage related tags now has to be expressed as
72 properties. An example of platform using this new tag is available at
73 examples/platforms/hostsè_with_disks.xml
75 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
76 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
77 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
78 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
79 - FG#31: per-actor data
80 - FG#34: SG_BARRIER_SERIAL_THREAD?
81 - FG#35: model-checker does not like buster-produced binaries
82 - FG!13: MC: complete workaround in the error msg seen on modern systems
83 - FG!15: execute_flops now logs compute
84 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
85 - FG!17: ns-3: unblock the right number of communications + others issues
86 - FG!18: Improving the performance of the ns-3 bindings
87 - GH#219: Error in the throughput of TCP transfer
88 - GH!330: Minor polishing of English: Configuring_Simgrid
89 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
90 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
92 ----------------------------------------------------------------------------
94 SimGrid (3.23.2) July 8. 2019
97 - Nicer introduction page.
98 - Migrate the "Deploy your application" page to the new doc.
99 - Move Java as a subtree of MSG.
102 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
105 - Drop xbt_dynar_sort_strings().
108 - Really fix FG#26: Turning off a link should raise NetworkFailureException
109 - FG#27: Wrong exception thrown to wait_any when link is turned off
110 - GH#328: Java: Canceling multiple tasks in a single vm/host
112 ----------------------------------------------------------------------------
114 SimGrid (3.23) June 25. 2019
116 The Exotic Solstice Release.
119 - SunOS and Haiku OS support. Because exotic platforms are fun.
120 - Stop setting random seed with srand48() at initialization.
121 - Use addr2line as a fallback for stacktraces when backtrace is not available.
122 - Build option -Denable_documentation is now OFF by default.
123 - Network model 'NS3' was renamed into 'ns-3'.
126 - Simgrid can now hopefully be installed with pip.
129 - wait_any can now be used for asynchronous executions too.
132 - New log appenders: stdout and stderr. Use stdout for xbt_help.
133 - Drop xbt_dict_dump.
136 - 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
137 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
138 - MPI interface is now const correct for input parameters
139 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
140 - Fortran bindings for async collectives.
141 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
144 - Remove option 'model-check/record': Paths are recorded in any cases now.
145 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
146 always sparse. That's an immense gain in memory, and can even be
147 faster because of cache effects. So there is no need to clutter the
148 code to allow the user to go for the unefficient mode.
151 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
152 models proved to be more accurate than these old experiments.
154 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
155 - FG#1: Broken link in error messages
156 - FG#2: missing installation documentation
157 - FG#3: missing documentation in smpirun
158 - FG#6: Python bindings not available on PyPI
159 - FG#7: simple cmake call requires doxygen
160 - FG#8: make python bindings an optional dependency
161 - FG#10: Can not use MSG_process_set_data from SMPI any more
162 - FG#11: Auto-restart actors forget their on_exit behavior
163 - FG#12: -Denable_lto=OFF doesn't disable LTO
164 - FG#13: Installs unstripped file 'bin/graphicator'
165 - FG#14: Installs the empty directory 'doc/simgrid/html'
166 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
167 - FG#17: Dead link in doc (pls_ns3)
168 - FG#20: 'tesh --help' should return 0
169 - FG#21: Documentation link on http://simgrid.org/ broken
170 - FG#22: Debian installation instruction are broken
171 - FG#26: Turning off a link should raise NetworkFailureException exceptions
172 - GH#133: Java: a process can run on a VM even if its host is off
173 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
174 - GH#326: Valgrind-detected error for join() when energy plugin is activated
176 ----------------------------------------------------------------------------
178 SimGrid (3.22) April 1. 2019
180 The Easter Christmas Release (this one is somewhat late).
183 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
184 - Not complete yet: asynchronous activities (amongst others) are still missing
185 - Still ongoing: the interface may change in the future. We need more
186 testers! Please report any glitches.
187 - No new project using Java should start now. Please switch to Python.
190 - Some of the internal cleanups may lead to speed improvements:
191 - The hard limitation on the amount of simulated actors with
192 Java+Mac was removed. Now, the available memory is the only limit.
193 - Our refcounting was tidyied, leading to 10% speedups in some cases.
194 - We are still working on making our code robust to the actor kills
195 and hosts' churn. Things are improving, but it's not perfect yet.
196 - Replaced our own code to display a backtrace (that was forking addr2line)
197 with the Boost.Stacktrace library.
198 You won't see your backtraces without this optional dependency.
199 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
200 - Stop setting random seed with srand() at initialization.
203 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
204 XML file version remains 4.2 since old files are still compatible.
207 - Process termination which was broken at version 3.21 has been repaired.
208 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
209 - Hide the examples into examples/deprecated. New users should use Python.
212 - Drop MSG_process_create_from_stdfunc() from the API.
213 This C++ function was a pimple in the C API, made necessary at some
214 point by the Java bindings. This is fixed now.
215 - Hide the examples into examples/deprecated. New users should use S4U.
216 - MSG_process_create and MSG_process_attach now crash if the host used
217 to run this process is off.
218 - Fix the protype of MSG_process_on_exit()
219 Now use (int,void*) callbacks instead of (void*,void*) ones.
220 The implementation was ways too messy to actually work, I guess.
223 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
224 too often and provide more accurate timings.
225 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
226 - MPI_Alltoallw support
227 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
228 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
229 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
230 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
232 - Basic implementation of generalized requests (SMPI doesn't
233 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
236 - Drop sg_cmdline. Please use xbt_cmdline instead.
237 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
238 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
239 OsSemaphore is implemented in a portable way with C++11 threads.
240 This should allow much more threads to be created at the same time,
241 allowing Mac and Java users to simulate many more actors.
242 - Implement the 'thread' factory with std::thread instead of xbt ones.
243 It is not possible to set the stack size with threads anymore, but
244 -fsplit-stack is the way to go nowadays when using threads.
245 - Drop the xbt_os_thread_t module (now unused)
246 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
247 - Drop xbt_str_join_array().
248 - Drop cunit, use Catch2 instead.
251 - Many cleanups in the kernel::activity namespace. This was long
252 overdue, and shall open the path to many future endeavors.
255 - #132: Java: a process can not shut down its own host
256 - #220: S4U: detached send causes exception if sender terminates soon after sending
257 - #261: Document the parameters of parallel execution's constructor
258 - #300: [s4u] BarrierPtr is missing
259 - #314: SMPI args internal cleanup
260 - #316: Fix a bug related to the CPU utilization of multi-core VM
261 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
262 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
263 - #325: Turning off a host has different behavior on sleeping actors and computing actors
265 ----------------------------------------------------------------------------
267 SimGrid (3.21) October 3. 2018
269 The Restarting Documentation (TRD) Release.
272 - Convert the user manual to Sphinx for improved usability.
273 Unstable version now hosted on https://simgrid.frama.io/simgrid
274 Many glitches remain; stay tuned.
275 - Restrict the Doxygen reference API to the installed header files.
276 (documenting all internals this way is overkill + doxygen scales badly)
277 - New tutorial on S4U (highly inspired from the old MSG one, but with
278 a git repository to fork as a starting point, and a docker image)
279 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
280 - The Developper manual is still to be converted (not compiled until then)
281 - Some parts are still missing in this conversion (such as the
282 platform and deployment sections) while others were blindly converted
283 and would need more love (such as the configuration flags).
284 Things will certainly further improve in the future releases.
287 - Fully reimplement auto_restart mechanism that was utterly broken
288 (fix #22, #131 and #281 that were all very related)
289 - Implement semaphores (fix #303 and #291)
290 - When creating an actor from a function and its parameters,
291 move-only parameters are not allowed anymore, as it would prevent
292 the actor to be restartable if its parameters are consumed on run.
293 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
294 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
295 IO operations that can be started, waited for, or canceled as a regular
297 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
301 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
302 - New host variable: 'core_count'
305 - Remove xbt_os_thread_specific features
306 - Remove portability wrapper to condition variables
307 - Remove xbt_os_thread_yield()
310 - MPICH collective selector now mimics MPICH 3.3b
311 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
312 - MPI_Init(NULL,NULL) is now allowed
313 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
314 - Command-line arguments should now be handled properly in Fortran simulations
315 - MPI Attributes and related callbacks should work in Fortran
316 - Apps using C + Fortran should now work
317 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
318 - Support MPI_Error_String
321 - Due to an internal bug, Msg.run() must now be your last line.
322 We hope to fix it in a future release, and we are sorry for the inconvenience.
325 - #22: Process autorestart seem to only work with CAS01 cpus
326 - #56: Feature request: dynamic SMPI replay
327 - #93: simgrid should not eat --help
328 - #111: How to change the loopback link characteristics?
329 - #116: Platform cluster radical
330 - #117: Platform cluster/host order
331 - #129: Replay_multiple on steroid
332 - #131: Java : autostart does not work
333 - #143: Setting a breakpoint at a given time
334 - #235: Network energy models should be integrated
335 - #264: Add ptask L07 resource tracing
336 - #271: Dynamic and manual replay of SMPI traces
337 - #279: Breakpoints option uses time deltas instead of absolute time
338 - #281: Daemonized actor no longer daemonized after an auto-restart
339 - #282: TIT convertor erase traces when using absolute path in the trace list file
340 - #285: segfault when a s4u actor kills itself with msg process tracing activated
341 - #286: Pajé traces are not exposing the number of cores
342 - #287: Command-line parsing should stop when encountering '--'
343 - #288: MPI_Init(NULL, NULL)
344 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
345 - #290: Method to check if a mailbox has a message ready to be consumed
346 - #291: [s4u] semaphores do not seem available yet
347 - #294: Alltoall collective implementation of mpich
348 - #295: Starting a Host as OFF is not supported
349 - #296: DTD too permissive
350 - #299: Add a s4u barrier example
352 ----------------------------------------------------------------------------
354 SimGrid (3.20) Released June 24. 2018
356 The proxy snake_case() release.
359 - The whole API is now using snake_case() instead of mixing it with camelCase.
360 Compatibility wrappers in place for one year.
361 - We took the opportunity to increase the consistency of the API.
362 Please report any remaining glitches.
363 - New function simgrid::s4u::Host::get_actor_count:
364 Returns the number of actors running on a specific host.
367 - SMPI is now tested with ~45 proxy apps from various sources, with none or
368 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
369 - Replay: The replay file has been re-written in C++.
370 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
371 supported. They are stored in the trace and used when replayed.
372 - Basic support of MPI_Cancel. Robustness not guaranteed.
373 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
374 (only for MPI_COMM_TYPE_SHARED).
375 - New option: smpi/privatize-libs, to add external shared libs to be privatized
376 by SMPI. They will be copied locally and loaded separately by each process.
377 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
378 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
379 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
380 with internal index optind with multiple processes. Only works if MPI_Init has
382 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
383 syntax without preprocessing tricks.
386 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
387 behavior tracing is no longer limited to MSG
388 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
389 this means that traces now need to be updated or re-obtained; both Irecv
390 and Isend lines in your traces have now in total 5 mandatory fields:
391 <rankid> <command> <to/from rankid> <tag> <size>
392 To update your traces, it suffices to add a 0 for the tag here.
393 - TIT now also supports waiting for a distinct request via MPI_Wait.
394 Wait/Test now wait for a specific request, not just the last one that was
395 issued. This unfortunately means another update, because we need to
396 identify which request you want to wait for. We do this via the
397 triplet (sender, receiver, tag), which needs to be added:
398 <rankid> <command> <sender> <receiver> <tag>
399 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
400 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
403 - The deprecation of MSG is ongoing (but this should not impact you).
404 Many MSG functions are now simple wrappers on the C API of S4U. If
405 you wish to convert your code to S4U, find the S4U counterparts of
406 your MSG calls in src/msg/msg_legacy.cpp.
407 - MSG can still be used, but won't evolve anymore.
410 - Allow to run the Link energy plugin from the command line with
411 --cfg=plugin:link_energy
412 - Rename Energy plugin into host_energy
413 - Rename Load plugin into host_load
416 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
417 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
418 - Improve performance when handling timeouts of simix synchros.
421 - Config: the C API is now deprecated (will be removed in 3.23), and
422 the C++ API has been slightly improved.
425 - Fix several build issues on OSX.
426 - Move simgrid_config.h to simgrid/config.h (old header still working)
429 - #143: Setting a breakpoint at a given time
430 - #150: Inconsistent event names in SMPI replay
431 - #258: daemonized actors hang after all non-daemonized actors have completed
432 - #267: Linker error on unit_tmgr
433 - #269: SMPI: tracing of MPI_Wait/all/any broken
434 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
436 ----------------------------------------------------------------------------
438 SimGrid (3.19.1) Released March 22 2018
440 The Fixed ABI Release.
442 As you may know, we are currently refactoring SimGrid in deep.
443 Upcoming SimGrid4 will be really different from SimGrid3: modular,
444 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
446 Our goal is to smooth this transition, with backward compatibility
447 and automatic update paths, while still progressing toward SimGrid4.
449 SimGrid remains open during works: The last pure SimGrid3 release was
450 v3.12 while all subsequent versions are usable alpha versions of
451 SimGrid4: Existing interfaces remain unchanged, but the new S4U
452 interface is budding and the internals are deeply reorganized.
454 Since 2015, we work hard to reduce the changes to public APIs. When
455 we need to rename a public library symbol in S4U, we let your
456 compiler issue an explicative warning when you use the deprecated
457 function. These messages remain for four releases, i.e. for one full
458 year, before turning into an error. Starting with v3.15, your can
459 also adapt to API changes with the SIMGRID_VERSION macro, that is
460 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
462 Starting with this v3.19.1, our commitment to reduce the changes to
463 the public interfaces is extended from the API to the ABI: a program
464 using only MSG or SimDag and compiled against a given version of
465 simgrid can probably be used with a later version of SimGrid without
466 recompilation. We will do our best... but don't expect too much of
467 it, that's a really difficult goal during such profund refactoring.
469 The difference between v3.19 and v3.19.1 is that the former was
470 accidentally breaking the ABI of MSG, while the later is restoring
473 S4U and kernel APIs will still evolve until SimGrid4, with one-year
474 deprecation warnings as currently. In fact, cleaning up these
475 interfaces and converting them to snake_case() is one release goal of
476 v3.20. But don't worry, we are working to smooth this upgrade path.
478 In summary, new projects should start with S4U to benefit of the
479 future, but old MSG projects should still be usable with no change.
481 ----------------------------------------------------------------------------
483 SimGrid (3.19) Released March 20 2018
485 The Moscovitly-cold Spring release.
487 SMPI user-visible improvements:
488 - Support of MPI_Type_create_subarray
489 - Unimplemented MPI calls now abort (instead of issuing a warning)
490 Some harmless calls around errhandlers still only issue a warning.
491 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
492 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
493 is now possible (mmap privatization was refactored to this end).
494 Adding new SMPI instances should also be technically doable now.
496 SMPI internal cleanups:
497 - Removed index notion from SMPI -- no more "getPid() - 1"!
498 In tracing files, actors are called rank-<PROCESS-ID> but these are
499 NOT the real ranks (tracing several communicators is not supported yet).
500 - Deprecate smpi_process_index() (will be removed in 3.22)
501 - Better testing on classical proxy apps, through an external project:
502 https://github.com/simgrid/SMPI-proxy-apps/
505 - Use a graphical TOC to make it easier to find the documentation you need
506 - Revamp the MSG tutorial
509 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
510 - Remove parameter of MSG_process_killall().
511 Resetting the PID was bogus anyway (several actors could have the same PID).
514 - Execution->setHost() can be called after start(), to migrate it.
515 - Comm::test_any() is now implemented.
516 - s4u::Actor now has onCreation() and onDestruction() signals.
517 - Install some kernel header files for the users' plugins and more.
518 Warning, their API is really not stable yet.
519 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
520 include/simgrid/kernel/routing/*Zone.hpp
523 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
525 Build System and other cleanups:
526 - Remove unused run-time parameter "tracing/onelink-only".
527 - SimGrid now works with intel compilers (MC still buggy)
528 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
529 - tesh kills subprocesses on timeout (but not on windows)
532 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
535 - #194: Feature request: simgrid::s4u::Comm::test_any()
536 - #245: migrating an actor does not migrate its execution
537 - #253: Feature Request: expose clusters as objects
538 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
539 - #255: Tesh broken on Windows
540 - #256: Modernize FindSimGrid.cmake
541 - #257: Fix (ab)use of CMake install
543 ----------------------------------------------------------------------------
545 SimGrid (3.18) Released December 24 2017
547 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
549 MSG IS NOW DEPRECATED. Please use S4U for new projects.
550 - Support for MSG should not be removed from SimGrid before 2020,
551 but future evolutions will be limited to the integration of
552 user-provided patches.
555 - s4u::onDeadlock() signal, run before stopping the simulation.
556 - s4u::Exec: asynchronous executions (abstraction of background
558 - s4u::Host->getLoad() returns the achieved speed in flops/s
561 - Link::name() is deprecated and will be removed in v3.21, use
562 Link::getCname() instead.
563 - Mailbox::getName() changed to return a std::string, use
564 Mailbox::getCname() instead to get a char*.
565 - Storage::getName() changed to return a std::string, use
566 Storage::getCname() instead to get a char*.
567 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
568 to have both allocation and deallocation of the map in user space.
571 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
572 The amount of flops remaining to do is a vector, not a scalar.
573 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
574 name implies on both sequential and parallel tasks.
575 - Both changes fix GitHub's #223 using PR #237 as a basis.
576 Thanks Michael Mercier.
577 - Most examples were converted to S4U and hidden elsewhere at tests.
578 Rationale: we still want MSG to work; we want newcomers to use S4U.
581 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
582 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
585 - Switch to the faster dlopen privatization mechanism by default
586 - Documentation improvments
589 - Live migration is getting moved to a plugin. Dirty page tracking is
590 the first part of this plugin. This imply that VM migration is now
591 only possible if one this function is called:
592 - C/MSG: MSG_vm_live_migration_plugin_init()
593 - C/C++: sg_vm_live_migration_plugin_init()
594 - Java: Msg.liveMigrationInit()
595 For C and C++, "simgrid/plugins/live_migration.h" also has to be
599 - Define class simgrid::xbt::Path to manage file names.
600 - Removed unused functions:
601 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
602 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
603 xbt_graph_export_graphviz()
604 - xbt/str.h: xbt_str_join()
605 - Remove unused datatypes:
606 - xbt/heap.h: use std::priority_queue or boost::heap instead
607 - xbt/swag.h: use boost::intrusive::list instead
610 - New link_energy plugin for the consumption of the links.
611 - All of the operations on files and storage contents have been
612 packaged into a plugin (src/plugins/file_system). The current
613 public interface can be found in
614 include/simgrid/plugins/file_system.h
615 To use these functions you now have to initialize the plugin by
616 calling MSG_storage_file_system_init() just after calling
617 MSG_init() or sg_storage_file_system_init() just after creating
621 - Remove the undocumented/untested tag <include>
624 - Remove viva specific tracing as the tool is no longer maintained
627 - #248: Kill of finished processes leads to segfault
628 - #240: xbt_cond_wait_timeout should gracefully return for C apps
629 - #239: please implement signal s4u::onDeadlock()
630 - #230: segfaults when exit() without run()
631 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
632 - #223: MSG_task_get_flops_amount() not working with parallel tasks
633 - #222: Actor::kill() doesn't really kill and segfaults
634 - #221: odd LMM warning when killing an actor
635 - #120: Memory leak when the processes are forcefully killed
637 ----------------------------------------------------------------------------
639 SimGrid (3.17) Released October 8 2017
641 The Drained Leaks release: (almost) no known leaks despite the tests.
643 Even very long simulations will preserve your memory: our testsuite
644 stresses SimGrid for over 45mn on fast machines for a coverage of
645 over 80%, and there is only one single known leak, of about 4kb.
649 - Comm.detach(): start and forget about asynchronous emission. A cleanup
650 handler may be given to free resources if the comm cannot be completed.
651 - this_actor::send(mailbox) is now mailbox->put()
652 - New: simgrid::s4u::Comm::wait_all()
653 - New: Host.execute() for remote executions.
656 - Unused option network/sender-gap is removed.
659 - Deprecate MSG_task_isend_with_matching(): This unused feature
660 really complicates our internals. Will be removed in v3.20.
663 - Improved context termination. It is now possible to free resources, even
664 when a process is forcibly killed.
665 - ContextBoost: add support for Boost versions above 1.61.
668 - Bring back run-time option --cfg=exception/cutpath to remove exception
670 - Removed unused functions:
671 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
672 xbt_str_rtrim(), xbt_str_trim().
673 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
676 - Removed header files obsolete since SimGrid 3.12:
677 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
679 - #3: SD_exit should be made optional
680 - #120: Memory leak when the processes are forcefully killed
681 - #159: Threading test regression in Actor refcounting
682 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
683 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
684 - #186: Actor::killAll() segfaults if some process is blocked on wait()
685 - #191: VM migration and pstate
686 - #192: Updating the energy consumptions of all hosts crashes with VMs
687 - #195: All actors have PID=0 in the logs
688 - #204: Sometimes segfault with thread contexts and mmap privatization
689 - #222: Actor::kill() doesn't really kill and segfaults
690 - #225: Actor::kill() doesn't really kill when victims are doing a join()
692 ----------------------------------------------------------------------------
694 SimGrid (3.16) Released June 22. 2017.
696 The Blooming Spring Release: developments are budding.
699 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
700 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
701 - Java JAR file should be finally fully working on Mac OSX too.
703 XML platforms: Switch to platform v4.1 format.
704 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
705 - <zone> can be used as a synonym for the now deprecated <as>
706 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
707 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
708 - <actor> can be used as a synonym for the now deprecated <process>
709 - state_file and avail_file periodicity is now easier to express
710 (check the documentation) the old behavior should still work.
711 * Storage: not backward compatible, but it seems that nobody ever used it
712 - Remove attribute 'content_type' of <storage_type>: was never used
713 - Make attribute 'model' of <storage_type> optional: for future usage
714 - Remove Bconnection model property: was never (in)validated. Replaced by
715 the maximum of the read and write bandwidth as a resource constraint.
718 - New and Backwards Compatibility break:
719 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
720 When one wants to get the list of tasks whose states have changed during a
721 simulation round, s/he has to allocate and free a dynar and use it as
722 argument to this function. The former SD_simulate (double how_long)
726 - Allow multicore VMs, with the correct sharing (unless you overcommit)
727 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
730 - New function to update the consumption of all hosts at once.
731 - Fix the model for multi-core hosts, linear in the amount of busy
732 cores with an abnormality for pIdle (see paper).
733 BUG: Still not satisfactory for ptask on mono-cores.
736 - The netzone are now available from the MSG API.
737 The old names still work, but are now deprecated.
740 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
741 - New option: smpi/keep-temps to not cleanup temp files
742 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
743 mallocs mode is used, allows to change the size of the fake file used
744 (default 1MB), to potentially limit the number of mappings for large runs.
745 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
746 - Fortran ifort and flang compilers support
747 - New RMA calls supported (experimental):
748 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
749 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
750 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
751 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
752 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
753 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
756 - New: this_actor::isMaestro() returns whether we are in kernel mode.
757 - Behind the scene, ActivityImpl are now automatically refcounted.
758 Removing this nasty bottleneck will greatly simplify our internals.
761 - Replay: New function xbt_replay_action_get():
762 Retrieve the function previously associated to an event type.
763 - DROPPED FUNCTION: xbt_str_varsubst()
764 - DROPPED MODULE: strbuff. We don't need it anymore.
765 - DROPPED MODULE: matrix. We don't need it anymore.
766 - DROPPED MODULE: lib. We don't need it anymore.
768 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
770 SimGrid (3.15) Released March 22. 2017
772 The Spring Release: continuous integration servers become green
773 We fixed even the transient bugs on all target architectures:
774 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
775 Windows, FreeBSD, NetBSD.
777 - Rename NetCards to NetPoints.
778 This was intended to help NS3 users, but that's not a netcard.
779 That's a point in the routing algorithm, let's avoid wrong simplifications.
782 - New: MSG_process_yield(). Stop and yield to other processes.
783 - New: MSG_process_daemon(). Daemon processes are automatically killed
784 when the last non-daemon process terminates
785 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
786 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
787 to make room for the upcoming network energy plugin.
788 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
791 - Ensure that an actor can kill itself with Process::exit()
792 - Kill the obscure NativeException. Nobody want to survive the issues
793 it denotes, so use JniException that is a RuntimeException (not to
794 be caught explicitly).
795 - Partial bug fix in initialization. SimGrid flags on command line were
796 consumed at C level but stayed in the original Java String[] args.
797 This could mess users' args[i] if SG flags were not put at the end of
799 The SimGrid flags are now removed from the Java arguments. However,
800 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
801 if args.length is greater than the number of YOUR OWN ARGUMENTS.
802 It might be if you have --log or --cfg flags in the command line.
803 - Fix numerous memleaks all around the place. In particular, around VMs.
807 - s4u::onPlatformCreated: right before the simulation starts
808 - s4u::onSimulationEnd: right after the main simulation loop
809 - s4u::onTimeAdvance: right after a clock change (time is discrete)
810 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
811 event from the availability_file changes the avail speed.
812 - Links are now usable from s4u
813 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
814 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
815 - New examples: The conversion of MSG examples to S4U has begun
816 - Actors: create, kill, migration, and suspend
817 - Applications: master-worker and token-ring
818 - Action replay: communications and storage
819 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
822 - Backwards Compatibility breaks
823 - The SD_route_get_list and SD_route_get_size functions have been removed.
824 They are replaced by the unique
825 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
826 The route from src to dst is built in the links dynar whose size can
827 be obtained with xbt_dynar_length.
828 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
829 sg_host_route_bandwidth and sg_host_route_latency.
830 Macros ensure the backwards compatibility, but you should fix your code
833 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
834 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
835 - Bump our claim of support from MPI 1.1 to MPI 2.2.
836 We don't support 100% of it, but it should be enough. Contact us if not.
837 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
838 naive and probably inaccurate)
839 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
840 mapped onto a unique small file using some system magic.
841 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
842 - Fortran: cleanups, fixes, support of user-added operations
843 - RMA: MPI_Accumulate are applied in correct order.
844 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
845 - Better support for MPI_IN_PLACE
846 - Support for MPI_Win attrs and keyvals.
847 - Support MPI_Comm_group_create, MPI_Type_size_x
848 - MPI ops should be properly applied to all allowed types
849 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
850 algorithm by a non failing one from mpich.
853 - Kill the fifo data container: we don't use it anymore.
856 - A <cluster> can now be created with different speed values to represent pstates.
857 see examples/platforms/energy_cluster.xml for an example.
859 SimGrid (3.14.159) Released December 28. 2016
861 The previous release was only a crude approximation of a Pi release;
862 this one should be a bit better.
864 - Revert a last minute change that broke on Mac OSX.
865 - Fix the changelog and NEWS files.
866 - Improve the documentation, in particular of the routing module.
867 - Integrate some patches flying around in Debian.
869 SimGrid (3.14) Released December 24. 2016
872 * The whole documentation was reworked and reorganized. There is
873 still much room for improvement here, but we're on it.
876 * We now need python3 for our testing infrastructure.
877 * Model checking needs libevent
880 * Now works on FreeBSD
883 * Privatization now works on FreeBSD
885 * Privatization is now activated by default in smpirun
886 You can turn it off with -no-privatize if something goes wrong.
888 * Call-location tracing for SMPI.
889 You can add the exact location (filename / linenumber) of an MPI call to
890 your trace files and slow down or speed up the simulation between two
891 consecutive calls by using an adjustment file (see the documentation).
893 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
894 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
895 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
896 values were ignored in some cases. The timings of these functions can now
897 be significantly different.
899 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
900 smpi/running-power is renamed to smpi/host-speed
902 * smpi/grow-injected-times option to enable or disable multiplication of the
903 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
904 make simulation less precise (but also much faster).
906 * smpirun script should be (much) faster for large deployments.
908 * SMPI tracing: fixed issue with poor matching of send/receives.
910 * Replay: Fix broken waitall
912 New functions and features
913 * MSG_parallel_task_execute_with_timeout, to timeout computations.
915 Dropped / renamed functions and features
916 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
917 - MSG_mailbox_is_empty() -> Mailbox::empty()
918 - MSG_mailbox_front() -> Mailbox::front()
919 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
920 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
921 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
922 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
923 * Don't pass the free_f parameter to property related functions:
924 - MSG_host_set_property_value()
925 - MSG_as_router_set_property_value()
926 - MSG_storage_set_property_value()
927 * VM properties. Since msg_vm_t are msg_host_t, just use
928 MSG_host_get_property_value() and friends
929 * VM I/O related things:
930 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
931 - Unimplemented save/restore methods
932 * MSG_as_router_get_property_value() was redundant with
933 MSG_environment_as_get_property_value().
934 - Removed MSG_as_router_*propert*() functions
935 - Added MSG_environment_as_set_property_value() for consistency
936 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
937 Well, they are still there for now, but deprecated with a warning.
938 Please switch to xbt_dict_new_homogeneous() before this is removed
940 * Task affinity. Its intended behavior (that was very badly tested
941 and probably not really working) was deceiving what most users
942 would have hoped here.
943 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
944 * Option network/coordinates is now useless and should be dropped.
947 * Added option storage/max_file_descriptors to allow more than 1024 files opened
950 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
951 * Added Dragonfly topology support
954 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
955 EiBps, EBps, Eibps, Ebps for bandwidth.
956 They may become useful to some lucky ones.
959 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
960 * ex/app_tokenring: new example, very similar to the MSG Token Ring
961 * ex/async_waitAll: new example, on asynchronous communications
964 * Memory usage should be decreased for simulations with a large number
965 of processes. This also helps for SMPI.
967 SimGrid (3.13) stable; urgency=low
969 The Half Release, a.k.a. the Zealous Easter Trim.
971 - We removed half of the lines, that were mostly experimental cruft.
972 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
973 (not counting blanks and comments -- according to openhub.net)
974 - The internals are now compiled in C++ (and will soon be clean C++)
975 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
976 59 klines of C, etc. We added only 29 klines of C++ in replacement.
978 * Backwards Compatibility breaks
979 - Removed Lua simulation bindings (switch to C or Java for that).
980 Lua can still be used to describe platforms
981 - Removed Java kernel plug-ins.
982 Will be reintroduced after the ongoing major internals reorg.
984 - the following functions were removed.
985 They were too specific and should be reimplemented in a generic
986 way, with filter function.
987 - MSG_task_listen_from_host
988 - MSG_mailbox_get_count_host_waiting_tasks
989 - MSG_mailbox_put_with_timeout was removed.
990 Please use MSG_task_send_with_timeout instead.
992 - the SD_application_reinit function was removed. It has been a noop for a while.
993 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
994 replaced by a more flexible mechanism.
995 - The following functions thus do not exist anymore
996 - SD_workstation_get_access_mode
997 - SD_workstation_set_access_mode
998 - SD_workstation_get_current_task
999 - Basic estimation functions have been removed but can easily be replaced
1000 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1001 - SD_workstation_get_computation_time => amount / sg_host_speed()
1003 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1004 Use VM.getSpeed()*load/100 for the legacy behavior.
1006 - option enable_tracing was removed. It was not doing anything for a while.
1007 - In the ModelChecker:
1008 - the model-checker now ptraces the model-checked process which means
1009 you cannot use a debugger on the latter anymore (we might make this
1010 optional in the feature);
1011 - removed soft-dirty page tracking;
1012 - remove model-checked side snapshot management,
1013 MC_snapshot() and MC_compare_snapshot();
1014 - keep the MC_cut() function as a stub (it was not really working
1015 in the previous release).
1018 * All options are consistently in kebab-case. Old names are kept as alias.
1021 * Switch to platform v4 format.
1022 - Rename from 'power' to 'speed' the attributes describing the amount of
1023 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1024 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1025 - In <host> and <link>, attributes availability and state are gone.
1026 It was redundant with state and availability traces, and with peak values.
1027 - In <cluster>, attributes availability_file and state_file are gone.
1028 It was too complex and unused.
1029 - Kill <gpu>. Was not doing anything.
1030 - The DOCTYPE points to the right URL:
1031 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1032 (the file at this address now documents the changelog since its v1)
1033 - A warning is emitted for unit-less values (they are still accepted).
1034 - speed. Default: 'f' or 'flops'. Also defined:
1035 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1036 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1037 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1038 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1039 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1040 - latency. Default: 's' second. Also defined:
1041 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1043 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1044 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1047 * s4u::Host is now the preferred public interface to the Host features.
1048 sg_host_* functions are C bindings to the exact same behavior
1049 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1052 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1053 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1054 They were deprecated since a few years, and probably did not even compile anymore.
1057 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1058 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1059 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1060 encouraged to update their codes. A list of the modified functions can be found at the end of
1061 include/simgrid/simdag.h
1064 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1065 the lambda or closure passed as a parameter will run in kernel mode.
1066 All the callback functions should be rewritten to that interface at some point.
1069 * Reorganizing and cleaning the internals all around the place.
1072 * Remove old default barrier/bcast buggy algorithms (see #18407)
1073 * Various bug fixes to handle more codes
1074 * Remove the need for the --foreground option of smpirun (it is still
1075 accepted for backward compatibility).
1078 * Kill the setset data container: MC don't use it anymore.
1079 * Kill the queue data container: it made more sense with GRAS.
1080 * Kill the xbt_peer_t data type: it's useless without GRAS.
1081 * Kill rm_cb feature of config sets: it was never useful.
1082 * Kill graphxml parsing feature. It was not used.
1083 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1085 - xbt_dynar_sort_strings(), when the content is char*
1086 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1087 They throw exceptions on invalid input;
1088 * C++ support for declaring CLI flags (simgrid::config::Flag);
1089 * class for abstracting different signal backends (simgrid::xbt::signal).
1090 with no external dependencies (we need very simple signals).
1093 * refactoring and cleanup of the code;
1094 * ongoing process to cleanly separate the model-checking algorithms
1095 from the code model-checking support.
1097 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1099 SimGrid (3.12) stable; urgency=low
1101 The Facelift Release.
1104 * Require g++ v4.7 at least to not speak prehistorical C++.
1105 * Require Boost 1.48 (for signal2 component).
1106 * Java must be version 7 at least when activated.
1107 * Builds on Windows again (including Java bindings).
1108 * Tracing is now always enabled (no way to turn it out)
1109 * Remove GTNetS. It was not working anyway.
1110 * Various cleanups in the cmake scripts.
1111 * Move headers around to sort them out on installed systems:
1112 - instr/instr.h -> simgrid/instr.h
1113 - instr/jedule/* -> simgrid/jedule
1114 - simdag/datatypes.h was removed
1115 - simdag/simdag.h -> simgrid/simdag.h
1116 - msg/datatypes.h was removed
1117 - msg/msg.h -> simgrid/msg.h
1120 * Interface improvement:
1121 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1122 - Sanitize the interface in MSG_task_ module:
1123 - Merge two functions that were close enough but misleading:
1124 set_compute_duration(t) -> set_flops_amount(t)
1125 get_remaining_computation(t) -> get_flops_amount(t)
1126 - set_data_size(t) -> set_bytes_amount(t)
1127 get_data_size(t) -> get_bytes_amount(t)
1128 - Massive cleanups in the functions related to the energy
1129 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1130 - New: MSG_host_get_pstate()
1132 - msg/energy/onoff: switching hosts on and off
1134 * Interface improvement:
1135 - Rename Host.isAvail() to Host.isOn()
1136 - Rename Process.currentProcess() to Process.getCurrentProcess()
1137 - Rename Task.setDataSize() to Task.setBytesAmount()
1138 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1140 - #18874: Actually allows the GC to reclaim tasks
1144 - SIMIX_process_throw: raises an exception in a remote process
1145 * Refactoring: Separate sub-modules
1146 - libsmx: the public interface, as libc in a real system
1147 - popping: the strange dance that converts a user request into a kernel handling
1148 - smx_context_*: the virtualization mechanisms that embed the user code
1149 - smx_*: the handling of each simcalls
1150 * Interface cleanup:
1151 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1152 * Rename smx_action_t into smx_synchro_t, making explicit that these
1153 things are used to synchronize processes with their environment.
1154 For example, a communication is a sort of synchronization involving
1155 the communicating processes (that may block until the exchange) and
1156 the platform. The same can be said from computations, etc.
1158 - #18888: segfault when a process ends before its kill_time
1162 - 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.
1163 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1164 - MPI_Type_set_name, MPI_Type_get_name
1165 - MPI_*_c2f and MPI_*_f2c functions
1166 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1167 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1168 - Activate a lot of new tests from the mpich 3 testsuite
1170 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1171 - InfiniBand network model added: Based on the works of Jerome Vienne
1172 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1173 - When smpi/display_timing is set, also display global simulation time and application times
1174 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1175 * Collective communications
1176 - 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
1177 - 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/).
1178 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1180 - Replay now uses algorithms from wanted collective selector
1181 - Replay can be used with SMP-aware algorithms
1182 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1183 - 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.
1185 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1186 - Correctly use loopback on fat-tree clusters
1187 - Asynchronous small messages shouldn't trigger deadlocks anymore
1188 * Energy/DVFS cleanup and improvement
1189 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1190 - new: smpi_get_host_pstate()
1193 - "Full" network optimization flag was broken since Surf++
1194 - Better handling of precision flags in maxmin
1195 - Fix bug causing sometimes "Impossible" errors
1196 - Properly pass cluster properties to included hosts
1197 * Improvement of the Energy plugin.
1198 - Always update the consumption before returning that value
1199 - New property: watt_off to denote the dissipation when the host is off
1200 - New functions getWattMinAt and getWattMaxAt to retrieve the
1201 dissipation of pstates that we are not currently at.
1202 * Java: class NetworkLink renamed to Link
1203 * New function: simcall_process_get_kill_time()
1204 * Massive rename s/workstation/host/
1205 - That's intrusive, but that's good for the project consistency. Sorry.
1206 - Change config option "workstation/model" into "host/model"
1209 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1210 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1211 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1212 - xbt_dirname and xbt_basename for non-POSIX systems
1214 * The model checker now runs as a separate process.
1215 * The model checker runs must now be launched with the new simgrid-mc program.
1216 * Record/Replay: the MC can display a textual representation of a path in the
1217 execution graph. It can then be replayed outside of the model checker.
1219 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1221 SimGrid (3.11) stable; urgency=low
1226 * Normalizing pointers addresses tool for better diff between logs
1229 * Add cloud examples using new VMs
1230 - examples/msg/cloud/two_tasks_vm.tesh
1231 - examples/msg/cloud/simple_vm.tesh
1232 - examples/java/cloud/cloud.tesh
1233 - examples/java/cloud/migration/migration.tesh
1234 * Add java surf examples:
1235 - examples/java/surfPlugin/surf_plugin.tesh
1236 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1237 - examples/java/surfCpuModel/surf_cpu_model.tesh
1238 * Add SMPI+MSG example:
1239 - examples/smpi/smpi_msg_masterslave/
1244 - msg task destroy cancel test
1245 - msg_host on/off test
1246 * Move all tests in testsuite to teshsuite (adding tesh files)
1247 * Restructure teshsuites
1248 - one folder for each kind of test
1249 * Restructure AddTests.cmake
1251 - structure the order of tests (with sections)
1254 * Add virtual machine
1255 - creation of a VM on a PM
1256 - migration of a VM from a PM to another PM
1258 - MSG_process_join(msg_process_t process, double timeout)
1259 - msg_bar_t MSG_barrier_init(unsigned int count)
1260 - int MSG_barrier_wait(msg_bar_t barrier)
1261 - void MSG_barrier_destroy(msg_bar_t barrier)
1262 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1263 * New option "msg/debug_multiple_use" to help debugging when a task is used
1266 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1268 - Increase code coverage in test suites
1272 * Protect context stacks against stack overflow. The number of protected memory
1273 pages allocated on the top of each stack (1 by default) can be configured
1274 with the new command line option --cfg=contexts/guard_size:<value>.
1275 * Simcalls are now generated by a python script that
1276 - generates files included by SimGrid
1277 - checks that all the functions exist, and proposes prototypes
1279 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1280 host_set_data, host_get_data
1282 - simcall_process_join(smx_process_t process, double timeout)
1283 * Fix bug where sleeping processing could not be suspended.
1286 * Translate surf models from C to C++
1287 - Generic classes for all models: Model, Resource, Action
1288 - A generic interface for each kind of model (CPU, Network, Storage
1289 Workstation, WorkstationVM)
1291 * Translate surf routings from C to C++
1292 * Add callbacks using sigc++ or boost::signals2
1293 - Add callback functions for resource creation/destruction
1294 - Add callback functions for action state change
1295 - Handle Energy as a plugin
1296 * Replace swag by boost::intrusive
1297 * Add new routing models for clusters. For documentation, see
1298 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1299 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1300 parameters for cluster tag
1301 - Fat trees, with topology="FAT_TREE" and
1302 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1303 - see examples/platforms/torus_cluster.xml and
1304 examples/platforms/fat_tree_cluster.xml
1305 * More documentation
1308 * Hostfiles support host:nb_processes construct to deploy several processes on
1310 * Collective communication algorithms should not crash if used with
1311 improper number of nodes and report the error.
1312 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1313 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1314 MPI_Dims_create, MPI_Cart_sub are supported.
1315 * New interface to use SMPI programmatically (still depends on MSG for
1316 some parts, see examples/smpi/smpi_msg_masterslave):
1317 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1321 * Global variables privatization in MPI executables is now performed at runtime
1322 with the option smpi/privatize_global_variables (default:no).
1323 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1324 dynamic libraries loaded by the application are not privatized (static
1325 linking with these libraries is advised in this case)
1328 * Options defined in XML work correctly now.
1331 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1332 copied into simgrid.jar or not (ON by default). Use this option if you want
1333 to reduce the size of the installed simgrid.jar, *and* the native libraries
1334 are kept installed elsewhere.
1335 * Surf binding with SWIG (code generated in maintainer mode only):
1336 - plugin to handle callbacks
1337 - CPU model only for the moment
1340 * Supernovae build mode is definitively removed. It was used to improve
1341 inlining and inter-module optimizations. It is nowadays superseded by
1342 link-time optimizations commonly available in compilers.
1343 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1344 * Add boost dependency for surf++
1345 * Add new macro for tests
1346 - ADD_TESH(name <tesh_args>)
1347 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1351 - xbt_bar_t XBT_barrier_init(unsigned int count)
1352 - int XBT_barrier_wait(xbt_bar_t barrier)
1353 - void XBT_barrier_destroy(xbt_bar_t barrier)
1354 * Make the xbt_os_time module public
1356 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1358 SimGrid (3.10) stable; urgency=low
1360 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1363 * Reintegrate Java to the main archive as desynchronizing these
1364 package is not acceptable anymore (Java is now considered stable)
1365 * Add explicit synchronization facilities through semaphores
1366 * Bug fix: Task.setDataSize() only changed the C world, not the value
1367 cached in the Java world
1370 * Dramatically change the way files are handled. API and internals changed, but
1371 this part of MSG was not considered as production grade either.
1372 * Add explicit synchronization facilities through semaphores
1373 * Add a new function MSG_host_get_process_list()
1374 * Preliminary DVFS support (see examples/msg/energy/ for details)
1377 * SMPI is now included directly in the libsimgrid as the windows
1378 linker doesn't force us on splitting it anymore.
1379 * Improvements of the SMPI replay tool:
1380 - Most of the collective communications are now rooted in the same process as
1381 in the original application.
1382 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1383 used until now). Multiple data types can now be used in a trace.
1384 - The replay tool now supports traces produce either by TAU or a modified
1386 - Bug Fix: the compute part of the reduce action is now taken into account.
1387 - Gatherv collective is now supported
1388 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1389 of smpirun outputs time independent traces for the current run. One file
1390 is created per process. If too many processes are simulated, this behavior
1391 can be changed to one file for all processes by using the
1392 tracing/smpi/format/ti_one_file flag
1393 * smpirun generates the host file if needed (with given host count and platform)
1394 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1395 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1396 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1397 --cfg=smpi/coll_selector:(mpich/ompi)
1398 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1400 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1401 happening inside a collective SMPI call.
1402 * Fix the behavior of complex data types handling.
1403 * Make MPI_Wtime another synchronization point to take computations into
1405 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1406 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1407 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1408 (no privatization of global variables yet)
1409 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1412 * Verification of liveness properties is now available for SMPI applications
1413 (in addition to MSG applications)
1414 * Bugged examples using SMPI in examples/smpi/mc/
1415 * Add --cfg=model-check/visited option. Allows the verification of infinite
1416 programs. Detection of loops in the execution thanks to the system state
1417 comparison and reduction of the state space to explore. Can be combined with
1418 DPOR for safety properties.
1421 * Allow to change SimGrid configuration (see --help) within the code
1422 thanks to SD_config() as it can be done in MSG.
1423 * Add a new function SD_task_set_amount() upon user request.
1426 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1427 * Remove rule based routing (no more PCRE dependency)
1428 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1429 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1430 * Add a loopback_bw and loopback_lat options to cluster tag.
1431 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1432 routing. Now peers are encapsulated in an AS and have their own private
1433 router but this is transparent.
1436 * Our own implementation of getline is renamed xbt_getline, and gets
1437 used even if the OS provide a getline(). This should reduce the
1438 configuration complexity by using the same code on all platforms.
1439 * New type: xbt_cfg_elm_boolean.
1440 * Allow to use yes/no for boolean configuration options in the command line.
1441 * Allow to disable SimGrid cleanups at exit from command line option.
1442 There are situations where one may want a simulation to end with an exit.
1443 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1444 annoying when scripting around the simulator. Adding a
1445 --cfg=clean_atexit:no allows to circumvent this issue.
1448 * Lots of memory leaks were corrected in this release.
1449 There is no known memory leaks anymore, in all of our 600+ tests.
1450 * New command line option --version, to get SimGrid version information.
1451 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1453 * Supernovae builds are deprecated, and expected to be removed in the next
1456 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1458 SimGrid (3.9) stable; urgency=low
1460 The Grasgory release: GRAS is really dead now.
1462 * Complete overhaul of the internal host structures scheme.
1465 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1466 was considered as experimental and badly maintained since then.
1467 * Keeping it was thus a trap to our potential users, that could take
1468 it instead of MSG or SMPI by mistake despite is pity state.
1469 * GRAS seems to have very few users (if any), and no one volunteered
1470 to maintain it further. It also induces a lot of XBT code (for
1471 portability sake), that must be maintained too.
1472 * For all these reasons, we killed GRAS. If someone wants to revive it
1473 in the future, don't cry, our git history still remembers of GRAS.
1476 * Major overhaul. Merge our documentation again as time proved that
1477 splitting it was really not helping our users.
1478 * Further improve the developer documentation to help newcomers
1479 hacking on SimGrid itself. The user documentation (and in
1480 particular, the beginner documentation) is still in a sorry state.
1483 * Now works on Windows too!
1484 * Much more extensive test suite, from MPICH
1487 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1488 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1489 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1490 * Bug fix: let task be scheduled when the last dependency to be solved is
1491 a control dependency.
1492 * Remove SD_load_environment_script function.
1493 Use the C sg_platf function if you want to declare a platform
1497 * New function: MSG_process_get_number()
1498 * Old function documented: MSG_config()
1499 * Remove MSG_load_platform_script function
1500 Use the C sg_platf function if you want to declare a platform
1504 * Change the default value of the TCP_gamma constant (maximal size of TCP
1505 congestion window) to a more realistic 4MiB value. If you notice changes in
1506 your simulation results, you can fall back to the previous 20k tiny window
1507 by adding --cfg=network/TCP_gamma:20000 on command line.
1508 * (Hopefully) fix a bug wrt periodic availability and state traces
1509 * Bug fix: use default values at start when first event in availability/state
1510 trace is not at time 0.
1513 * remove the "new_" part of function name sg_platf_new_trace_connect
1514 (resulting in sg_platf_trace_connect), since it does not create
1518 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1519 use case where it's really mandatory, and maintaining it was a pain in
1521 * New: xbt_fifo_search(), search an item with a user-provided
1522 comparison function instead of dumb pointer comparison.
1525 * Fix the lua deployment:
1526 Use `simgrid.init_application()` before deployment instead of
1527 `simgrid.msg_register_application()` after.
1530 * Transfer the tracing files into the corresponding modules.
1532 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1534 SimGrid (3.8.1) stable; urgency=low
1536 The "we are told that some people want to also *install* the simgrid
1539 * Add missing file "tesh.1" to the archive.
1541 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1543 SimGrid (3.8) stable; urgency=low
1545 The Psssshiiiit release: SimGrid jumps into the Cloud.
1548 * Add an experimental interface to manipulate VMs. They are mainly
1549 process groups with very few intrinsic semantic, but they should
1550 allow you to build the semantic you want easily.
1551 * New function: MSG_host_set_property_value()
1552 * New function: MSG_process_on_exit(). To clean memory in all cases.
1553 * Bug fixes that made the host (and link) failures unusable.
1554 * Add a way to auto-restart process when the host in which they are
1555 executing comes back (ON_FAILURE="RESTART" on deployment file,
1556 MSG_process_auto_restart_set).
1557 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1558 please stop using the old ones, they are DEPRECATED.
1560 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1561 Please use MSG_init() instead. (reducing the amount of entry
1562 points in the library helps us).
1563 * Make it impossible to link against the wrong version of the lib
1564 * Deprecate MSG_clean(). No need to call it anymore.
1565 * Function MSG_get_host_number() is not deprecated anymore.
1568 * Split the doc into a user guide and a reference guide.
1569 * Start a developper guide to help people hacking on SimGrid.
1572 * Enable tracing by default. This modules rocks you should use it.
1573 * Remove option custom_flags. Now use environment variables CFLAGS
1575 * Use default cmake things to detect lua instead of home grown ones.
1576 * New option "enable_mallocators" to disable mallocators, for debugging
1577 purpose ("on" by default).
1580 * Bug fixes around the resource failures: don't let the processes
1581 survive the host they are running onto.
1582 * Add an interface to auto-restart processes when the host in which they are
1583 executing comes back.
1584 * Ensures that SIMIX_clean is called automatically. It's not part of
1585 the public interface anymore (bindings should be updated).
1588 * Bug fix for when SD_Simulate is called with a positive value: be careful
1589 when comparing doubles. Sometimes they are different for non significant
1591 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1592 parallel task whose initial work is distributed among host according
1593 to the Amdahl's law. Such tasks are created with a parameter alpha
1594 that corresponds to the non-parallelizable part of the computation.
1595 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1596 between two sets of workstations assuming a 1D block distribution (each
1597 workstation owns a similar share of data) on both sides.
1599 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1600 Data redistribution will be automatically scheduled once parent and child
1601 are both scheduled. The filling of computation_amount and
1602 communication_amount structures is now done seamlessly thanks to the chosen
1604 * New function SD_workstation_dump to display various information
1605 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1606 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1607 bandwidth on the route onto which the task is scheduled and the amount of
1609 To divide the nominal bandwidth by 2, the rate then has to be:
1610 rate = bandwidth/(2*amount)
1611 * Compute tasks that have failed can now be rescheduled and executed again
1612 (from their beginning)
1613 * Increasing source code coverage (src/simdag is now covered at 95.8%
1617 * Re-implement time-independent trace replay using SMPI (at the
1618 smpi_smp_* level) instead of MSG. This should replace
1619 examples/msg/actions/actions.c
1620 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1621 hindexed and structs)
1622 * Implement the exchange of non-contiguous data.
1623 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1624 * Correct behavior of smpi/sender_gap and set its default value to 0
1625 * Add option to asynchronously send small messages to allow better
1626 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1627 specifies the size in bytes under which messages will be asynchronously sent.
1628 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1629 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1632 * New C interface to define a platform: XML is now optional.
1633 For more info, please check include/simgrid/platf.h.
1634 * New interface to define random platforms from the C:
1635 For more info, please check include/simgrid/platf_generator.h and
1636 examples/msg/masterslave/masterslave_platfgen.c
1637 * Export a sg_cmdline dynar containing all the arguments we got from
1641 * Two new tracing options for adding comments to trace file so you
1642 can track your experiments (see --help-tracing for details).
1643 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1644 * Adding the SimGrid version that generated the trace file as a comment.
1645 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1646 * Fix to avoid key clashes on Paje links
1647 * Other minor fixes related to the Paje specification
1650 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1651 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1652 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1653 * Greatly improve the robustness of mmalloc to user errors (such as
1654 using an area after freeing it, or freeing it twice)
1656 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1658 SimGrid-java (3.8.1) stable; urgency=low
1660 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1661 Warning: all methods are visible, but only some of them are
1662 implemented so far. Check the source (src/jtrace.c)
1663 for further information.
1664 * New module: org.simgrid.msg.File (SimGrid File management functions)
1665 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1666 * Change the meaning of Process.restart: now restart the process from
1667 the begining, like MSG_process_restart in C.
1668 * Add Process.setAutoRestart: handling of process restart when failed
1670 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1671 you to retrieve the properties of the processes/hosts
1672 * Deprecate Msg.clean(): you can just forget about it now.
1673 * New function Process.getCount(), that only works when compiling
1674 with the not yet released version 3.9 of the C library.
1676 * New context factory based on Coroutines. It mandates a modified JVM
1677 but then, the simulations run about five times faster, and there is
1678 no limit to the amount of processes (beside of the available memory).
1680 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1682 SimGrid (3.7.1) stable; urgency=low
1685 * Restore the prototype of MSG_process_create_with_environment() to
1686 the pre-3.7 situation by removing the kill_time argument.
1687 * Add a MSG_process_set_kill_time() function instead.
1690 * Fix weird behaviors when dealing with parallel tasks.
1693 * Simgrid is now built as a dll.
1694 * Simgrid-java now works on Windows.
1695 * Simgrid-Java is now included into Windows package.
1698 * First pre-build package for MacOSX.
1701 * Fix compilation when using MSG_USE_DEPRECATED.
1702 * Fix some compilation issues on Macs and Windows.
1703 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1704 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1706 We discovered that the Lua console is broken, but we are missing the
1707 manpower to fix it right now. The problem existed in 3.7 too, so we
1708 are not blocking the release for that. Sorry if you depended on this
1709 feature, any help would be really welcome.
1711 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1714 SimGrid-java (3.7.1) stable; urgency=low
1716 The "Java aint got to be bloated and slow" release
1719 * Various internal cleanups and performance improvement
1720 Simulations are expected to run up to twice faster or so
1721 * Make Process.kill(process) an instance method, not a static one
1722 * User processes are not java.lang.Thread subclasses.
1723 This breaks the compatibility (sorry), but previous API was
1724 brain-dead, making it impossible to have non-trivial
1725 initializations in the process constructor.
1726 * Require a full constructor per Process sub-class.
1727 Kinda breaks the compatibility (sorry), but this allows a much more
1728 efficient way to launch the processes at simulation startup.
1729 * Do not embeed our version of semaphores, java 1.5 can be considered
1730 as sufficiently prevalent for us to not dupplicate its features.
1734 * Add examples for almost every part of the API
1735 We spotted and fixed a lot of bugs in the process
1737 * New module: asynchronous communication API
1738 * New function: Process.sleep()
1739 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1740 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1741 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1743 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1745 SimGrid (3.7) stable; urgency=low
1747 The "spring cleanups (before the next Big Project kicks in)" release.
1750 * We can specify the SMPI latency/bandwidth factor with command line
1751 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1752 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1753 You can also use the "config tag" from platform file by adding this line
1754 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1755 (see "example/platforms/tag_config.xml" to use "config tag").
1756 Note that the command line supersedes the platform file configuration.
1757 * Change the correction factors used in LMM model, according to
1758 the latest experiments described in INRIA RR-7821.
1759 Accuracy should be improved this way.
1760 * Use the partial invalidation optimization by default for the
1761 network too. Should produce the exact same results, only faster.
1762 * Major cleanup in surf to merge models and split some optimization
1763 mechanisms from the core of the models. As a result you can now
1764 specify which model to use (e.g., --cfg=network/model:LV08
1765 --cfg=cpu/model:Cas01) and which optimization mode to use
1766 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1767 Incompatible combinations should err at initialization. See
1768 --help-models for the list of all models and optimization modes.
1769 * The CLM03 workstation model was dropped for simplicity because it
1770 used the deprecated CM02 network model. Use default instead.
1771 * Rename the TCP_gamma configuration option to network/TCP_gamma
1772 * Rename the coordinates configuration option to
1773 network/coordinates, and document it
1774 * Use now crosstraffic keyword instead of the terribly misleading
1775 fullduplex keyword. It is activated by default now in the current
1776 default model, use --cfg=network/crosstraffic:0 to turn it off.
1777 * Ongoing refactoring the model parsing to make XML files optional
1778 See include/simgrid/platf.h for details (still to be completed)
1781 * Major overhaul of the documentation. Almost instructive now :/
1782 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1783 functions and friends. This interface was considered as
1784 deprecated since over 2 years, it's time to inform our users that it is.
1785 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1786 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1787 use these (crappy) functions in your code.
1788 These functions will be removed soon. Stop using them now.
1789 * Deprecate MSG_get_host_{table,number}
1790 Implement MSG_hosts_as_dynar() instead.
1791 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1792 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1795 * Stabilize the parallel execution mode of user contexts
1796 * Introduce configuration variables to control parallel execution:
1797 - contexts/synchro: Synchronization mode to use when running
1798 contexts in parallel (either futex, posix or busy_wait)
1799 - contexts/parallel_threshold: Minimal number of user contexts
1800 that must be part of a scheduling round to switch to parallel
1801 execution mode (raw contexts only)
1802 * Fix bugs that prevented to use suspend/resume along with
1803 synchronization structures.
1804 * Fix bugs in process termination that lead to invalid memory access
1805 in very specific conditions.
1808 * Introduce a parallel mode for the models (controlled by surf/nthreads
1809 configuration item). In our tests, running the models in parallel
1810 never lead to any speedups because they are so fast that the gain
1811 of computing each model in parallel does not amortizes the
1812 synchronization costs, even when ultra fast futexes are used.
1813 This is released anyway because YMMV.
1816 * Performance boost by using a swag internally to compute the set of
1817 tasks that are finished and should constitute the return value of
1821 * Enable it by default now that it is considered rather stable.
1824 * Documentation of the tracing functions.
1825 * Performance gains when tracing categorized/uncategorized resource
1826 utilization by avoiding calls to get route when updating resource
1827 variables. LMM constraints are being used instead.
1828 * API changed to set task categories. Use MSG_task_set_category instead
1829 of TRACE_msg_set_task_category, and SD_task_set_category instead
1830 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1831 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1832 * Examples for link user variables added (see at examples/msg/tracing/)
1833 * Deprecated function TRACE_msg_set_process_category completely removed
1834 * Trace header updated according to the latest Paje file format
1835 * Tracing network lazy updates, no longer obligate users to use full updates
1836 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1837 * Experimental: let user code declare/set/push/pop application states for hosts
1838 * API changed to allow the manual creation of graph configuration files
1839 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1842 * Improve the API of Lua MSG bindings, using the Lua spirit.
1843 * Each simulated process now lives in its own Lua world (globals are
1844 automatically duplicated). It helps writing simulators. Will allow
1845 to run Splay programs within SimGrid in the future.
1846 * Add a Chord example in Lua, equivalent to the MSG one.
1849 * Start the implementation of a solution to express temporal
1850 properties, not only local assertions. This is still an
1851 experimental work in progress, stay clear from it to be safe.
1855 - Add new runtime parameters --help-logs and --help-log-categories
1856 to display information about supported logging parameters and
1858 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1859 don't exists anymore.
1860 * Mallocators: allow value NULL for the reset function.
1862 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1863 create homogeneous dictionaries, where all the elements share the
1864 same free function. Non homogeneous dictionaries will be
1865 deprecated in the next release.
1866 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1867 - Multi-level dictionaries are deprecated.
1869 - new function xbt_dynar_search_or_negative() that is useful when
1870 you have less than 2 million elements in your dynar and don't
1871 want of the extra complexity of catching exceptions when the
1872 element is not found.
1874 - Make xbt_os_thread module (for thread portability) public.
1875 Documentation is still to come, sorry.
1877 - Cleanups and simplifications to make it maintainable again.
1878 - Exotic features (such as memalign and valloc) were removed.
1879 - The metadata were extended and improved so that the
1880 model-checker becomes able to explore and inspect the heaps.
1881 - This may induce a performance drop when enable_model-checking is
1882 ON in cmake (even if it's not used in the simulation), but it is
1883 necessary at this point to get MC working.
1885 Turn model-checking OFF if simulation performance matters to you.
1886 Not enabling it at runtime is not enough, disable it in cmake.
1888 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1891 SimGrid (3.6.2) stable; urgency=low
1893 The "Not coding new stuff allows to polish old things" release.
1896 * New bindings to the NS3 packet level simulator (experimental)
1897 * Use the raw (efficient) execution contexts instead of the sysv
1898 (portable) ones when possible.
1899 * libpcre is now mandatory in any cases since not using it led to
1900 severe performance loss and possibly other issues
1901 * Update the XML platforms:
1902 - G5K: include the latest machine in Nancy
1903 - GridPP and LCG: new platforms
1904 * Documentation was partially updated, at least (more to come)
1906 Bug fixes, cosmetics and small improvements
1907 * Free terminated processes before the end of the simulation to avoid
1908 exhausting the memory of users having very dynamic amount of
1910 * Bug fix and cosmetics about canceling non-running tasks
1911 * Bug fix about the dot loader's issues when using libcgraph
1914 * Create an installer for windows with nsis (amd64 and win32)
1915 - Add an hello world project to illustrate simgrid project creation.
1916 - Embed libpcre into the Simgrid installer to avoid
1917 its compilation burden
1918 * The raw execution contexts should work on Apple now
1919 * Port to Windows 64 bits
1920 - Sysv contexts now have an implementation for this arch
1921 - GRAS communication features now support this arch
1922 * Drop support for borland compiler on windows
1923 - this code was not maintained, and we kinda depend on gcc nowadays
1924 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1925 * Fix portability issue on unstable ubuntu: linker became picky on
1928 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1930 SimGrid (3.6.1) stable; urgency=low
1932 The "Oops, we broke Macs too" release
1935 * Fixed contexts detection so that raw ones are used when possible
1936 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1937 leads to a strange error, with user code segfaulting sometimes when
1938 the generated code is not perfectly aligned (which is not
1939 controllable from the user side, depends on the amount of code)
1942 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1944 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1946 SimGrid-java (3.6) unstable; urgency=low
1949 * Split of every thing from simgrid v3.5 into a separate package.
1951 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1953 SimGrid (3.6) stable; urgency=medium
1955 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1958 * Bindings now constitute their own package, separated from the main one.
1959 Rationale: reduce our maintenance nightmare by reducing the module coupling
1960 They will soon be released on their own on gforge.
1962 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1963 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1965 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1966 * It's not quite deprecated for now because we have no replacement,
1967 but it may soon become the case.
1970 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1971 * Fortran: New user-level cache variable to store the rank of the running
1972 process. This improves performance by an order of magnitude.
1973 * C: New coccinelle script to automatically locate and modify global and
1974 local static variables.
1975 * Improved SMPI network model with a sender-side gap to account for multiple
1979 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1980 only say if a communication is finished, no matter whether it succeeded or
1981 failed. You can call MSG_comm_get_status() to know the status of a finished
1983 * New function MSG_task_dsend() to send a task and detach it. When a
1984 communication is detached, you are never notified of its success or failure
1985 and the memory is released automatically once it is finished. This function
1986 is useful when you don't care about the end nor the success of a
1988 * Change the prototypes of action replay. Sorry for inconvenience,
1989 but this is really more efficient this way (and to adapt your code,
1990 you just have to fix the initialization, that shouldn't be too long)
1991 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1992 ever managed to do anything useful with it.
1993 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1994 takes a dynar of communications. It returns immediately and gives the
1995 index of a finished communication (if any).
1996 * New example: a basic implementation of the Chord P2P algorithm.
1999 * New model for multi-core CPUs. You can now use the core attribute to
2000 precise the number of cores of a host. This is a basic model. Every
2001 process running on the host receives at most the power provided in
2002 the DTD (throughput<=power). Total throughput of process cannot exceed
2004 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2005 router linked by an up-link and a down-link (possibly asymmetrical).
2006 This kind of pattern allows to easily build last-mile model style platforms.
2007 Aggregating such patterns in a rule-based AS is thus the technique of
2008 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2009 * New model for Vivaldi routing. We transformed the Vivaldi network model
2010 into a Vivaldi routing model (based on the rule-based model). This allows to
2011 combine Vivaldi based latencies with last-mile platforms.
2014 * Added a check for NaN of IEEE754 infinite in the double entries of
2016 * Introduce a new context factory "raw", highly inspirated from the
2017 ucontext factory, but using manually crafted functions in assembly to
2018 do the work in an efficient manner.
2019 * Allow to change the used context factory at run time, not only at
2020 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2021 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2022 contexts at runtime (only with raw contexts or ucontexts).
2023 * Completely rewrote this module to allow parallel execution of user
2024 processes. Use --cfg=contexts/nthreads:N to execute user processes
2025 with N parallel threads (the default is 1, meaning no parallelism).
2026 * Allow to decide dynamically between sequential and parallel modes.
2027 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2028 processes in parallel only when their number is greater than or equal to P
2030 * Added a check for NaN of IEEE754 infinite in the double entries of
2034 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2035 won't output the state of processes when interrupted with Ctrl-C
2036 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2037 NULL-terminated array. This may solve backward compatibility issues
2038 due to the change to return type of SD_simulate. See also:
2039 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2040 * Add new macros with variable number of arguments.
2041 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2042 - in xbt/asserts.h: xbt_assert
2043 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2044 - in xbt/ex.h: THROWF and RETHROWF.
2045 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2047 * Change xbt_die() to accept a format string with arguments, just like printf.
2048 * New data structure: xbt_lib_t, like a dict but more general and with better
2052 * New configuration options
2053 Options triva/categorized and triva/uncategorized can be used to generate
2054 graph configuration files for Triva visualization tool.
2055 * Configuration option tracing/platform is renamed to tracing/categorized
2056 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2057 * New TRACE_user_link_variable interface:
2058 User provides the name of the link and the tracing variable to attach to it
2059 * the declaration of tracing categories must be done after the environment
2061 * simpler tracing interface, just one way to declare categories
2062 TRACE_category or TRACE_category_with_color, it is up to you
2063 * links in the trace file are again identified by their names
2064 * trace contains the full platform hierarchy exactly as declared using the ASes
2065 * Options tracing/msg/[task|process]:1 groups the process by hosts
2066 for both cases, tasks and processes must have names that are unique during the simulation
2067 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2068 * The experimental option tracing/msg/volume is deprecated
2069 its functionality may be reincorporated if needed
2071 The tracing generates a trace file with unordered timestamped events,
2072 because of the way the core simulator (surf) works. A script available
2073 at the tools directory (fix-paje-trace.sh) can be used to put the events
2074 in order. We have changed the tracing so it can generate ordered timestamped
2075 events in the final trace, but depending on the simulator (and how much time
2076 is simulated) that can lead to a huge memory utilization. It is deactivated
2077 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2079 Build Infrastructure
2080 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2081 - We are trying hard to keep the API stable, but it may happen that
2082 some things change (we're a research project after all, not a
2083 nuclear plant operating system). If such things should happen, you
2084 could rely on that macro to adapt.
2085 - current value: 30600 for 3.06.00, aka 3.6
2086 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2087 a number that can be compared with SIMGRID_VERSION.
2088 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2089 assertions and verbose/debug logging events are disabled at compile time.
2091 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2093 SimGrid (3.5) stable; urgency=medium
2096 * New feature to any SimGrid-based simulator: Model-Checking
2097 Check SIN#1 for more details.
2100 * New Model SMPI (three-interval linear regression for correction factors)
2101 See RR-7426, available at http://hal.inria.fr/inria-00527150
2102 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2103 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2104 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2105 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2106 * See also src/smpi/README
2111 - Tracing API changes: TRACE_start and TRACE_end should not be called
2112 by user-code. They are automatically called by simulators created
2113 with SimDAG, MSG and SMPI if the toolkit is compiled with
2114 tracing_enabled=ON. Categories declaration and utilization remain the
2115 same for MSG and SimDag.
2116 - A function was added to the tracing API to declare categories with
2118 - TRACE_category_with_color (char *category, char *color)
2119 where color must be in the following format
2120 "%f %f %f", red, green, blue
2121 and red, green, blue are float values in the interval [0, 1]
2122 - User can specify NULL as color parameter, or continue calling
2123 TRACE_category (cat)
2124 On that case, the tracing system will define random colors
2125 - The following command-line options are supported:
2126 --cfg=tracing/filename:msg.trace
2127 --cfg=tracing:1 (activate tracing, needed to use others)
2128 --cfg=tracing/platform:1 (categorized resource use)
2129 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2130 --cfg=tracing/msg/task:1 (task creation)
2131 --cfg=tracing/msg/process:1 (process creation, migration)
2132 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2133 --cfg=tracing/smpi:1 (SMPI interface tracing)
2134 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2135 - examples of examples/msg/tracing updated
2137 - DAXLoader and DOTLoader functions can generate tasks with categories
2138 - A new function to attribute a category to SD tasks:
2139 TRACE_sd_set_task_category (SD_task_t task, char *category)
2140 * Tracing the MPI interface implemented by SMPI
2141 - Collective operations are traced with states
2142 - Point-to-Point operations are traced with states/links
2143 - Tracing activated by a parameter "-trace filename" passed
2144 to smpirun during execution (considering that simgrid
2145 is compiled with tracing enabled)
2146 - To run the simulation with gdb, the simulator
2147 accepts --cfg=tracing/smpi:1 to trace SMPI
2148 - tesh files to check if smpi tracing is ok
2149 - See examples/smpi/NAS/DT-trace
2150 * GTNetS tracing re-worked
2151 - adaptation to the tracing system of GTNets to cope
2152 with modifications regarding the fullduplex mode
2153 - new tesh files to check if gtnets tracing is ok
2156 * Asynchronous communications through the functions:
2157 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2158 * New function: MSG_load_platform_script()
2159 to make possible using a lua script instead of XML files to set up platforms
2160 * New function: MSG_set_function
2161 to associate functions to processes, used when bypassing the parser
2162 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2164 Platforms: Add some more examples in examples/platforms
2165 * Grid'5000: see www.grid5000.fr
2166 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2169 * Change the XML format. This is a very important modification. SimGrid
2170 3.5 introduces a new hierarchical format based on the notion of
2171 Autonomous Systems. Compatibility with old format is ensured
2172 through the perl script provided in the install bin directory
2173 bin/simgrid_update_xml.
2174 It is now possible to build platforms with specific routing
2175 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2176 connect several platforms together. We will try to provide soon
2177 set of realistic platforms exploiting these properties (have a
2178 look at examples/platforms/ for the moment).
2179 * Take the opportunity of the XML format change to be a good XML citizen:
2180 rename link:ctn to link_ctn and similar changes (also dealt with by
2182 * Add a new routing scheme (rule-based) using regular expressions. It
2183 enables to have an extremely low memory footprint when the
2184 underlying routing is simple and can be compactly described. You
2185 need to have libpcre4-dev (perl regular expressions) installed if
2186 you want to use this routing scheme.
2187 * Revive the cluster TAG and allow to easily and efficiently (both in
2188 term of memory and speed) connect clusters together. Have a look
2189 at teshsuite/simdag/platforms/ to see how this can be done. With
2190 this tag, you can create clusters with thousands of tasks at no
2191 cost (have a look at examples/platforms/).
2192 Note: clusters are implemented as ASes, so there is no need for an
2193 enclosing AS tag if you have only one cluster in your platform.
2194 * Add new generic functions in the public interface that allows the user
2195 to call SURF 'create_resource' methods from your code (same
2196 functionality as the XML bypass mechanism but with a much lighter
2198 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2199 that uses a piecewise linear approximation to produce better
2200 results when exchanging small messages.
2201 * Add a new parameter to handle correctly full duplex link and account
2202 for interferences between uplink and downlink communications
2203 (activate with --cfg=fullduplex:1).
2206 * Rename the SD_READY (all dependencies are satisfied and task is
2207 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2208 dependencies are satisfied) state.
2209 This prevents a confusion between the notion of "ready to schedule"
2210 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2211 simulated" (SD_RUNNABLE) used by the simulation kernel.
2212 * Change the way a task is considered as ready. Instead of removing
2213 dependencies when a task is done, a counter is decreased. This way,
2214 it is always possible to reach ancestors thanks to the
2215 SD_taks_get_parents function (even after the end of the simulation.)
2216 * Change the return type of SD_Simulate from (SD_task_t*) into
2217 xbt_dynar_t. This function was in handling a dynar internally and
2218 converted it into a NULL terminated array for historical reasons.
2219 * New function SD_dotload(char*) to load a DAG described in dot
2220 format. This loader and the corresponding examples require the
2221 installation of the graphviz library.
2222 * Fix a bug in the management of tasks of size 0 in the surf network
2223 models. This problem was only visible with SIMDAG and you should
2224 thus disregard results produced with earlier versions if you
2225 relied on this feature (some tasks were blocked because of this).
2226 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2227 even though your applications were doing only point-to-point
2228 communications and sequential computations. Now you can really use any
2229 model you want (of course, if you create real parallel tasks, which are
2230 not implemented in most models beside ptaskL07, this will abort).
2231 * Add an example that schedules a DAX on an heterogeneous platform
2232 using a Min-Min strategy.
2233 * New function SD_workstation_get_current_task() that returns the kind
2234 of task currently running on a workstation in the sequential access
2236 * Raise some warnings when unexecuted tasks remains at the end of the
2237 simulation. This is usually caused by cycles in the DAG.
2240 * New function: SIMIX_process_set_function() called by MSG_set_function
2241 * Change the underlying waiting queue in semaphores so that a process
2242 can wait on several of them simultaneously (as in waitany).
2243 * Fix the way to handle tokens in semaphores so that all access patterns
2244 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2245 * kill the dirty pimple SIMIX_message_sizes_output()
2246 Please use (proper) visualization instead
2249 * New data container: setset (set of sets of elements)
2250 * New module: mmalloc (mapped malloc, allowing to have several
2251 independent segments of malloc)
2252 * New function: xbt_dict_cursor_set_data()
2253 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2254 * New function: xbt_dynar_is_empty()
2255 * New function: xbt_fifo_get_last_item()
2256 * Fix xbt_dynar_shrink(): use the right element size.
2257 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2258 uninitialized areas during expand.
2259 * Fix semaphores: previous implementation was severely broken.
2260 * Use library init/fini functions for our initialization.
2261 - you can use logs and other feature as soon as you want in your
2262 code (even before the xbt_init / MSG_init)
2263 - xbt_exit is now a no-op and produce a warning when used.
2266 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2267 benefit from the latest improvement to the simulation kernel.
2268 * Kill measurement sockets for now. If you rely on them, sorry. This
2269 release is not for you. This feature will be reintroduced in the
2270 future, but we cannot delay the release any further.
2271 * New function: gras_msgtype_get_name().
2272 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2273 * Fix (at last) the pmm example: it should not randomly fail anymore.
2275 Build chain: bug fixes and overall polishing
2276 * Cmake is now stable enough. Hence, we killed the autotools.
2277 * Port to windows ( TM :)
2278 * Fix the 'make install' target.
2279 No need to use 'make install-simgrid' anymore
2280 * Introduce a 'make dist' target compiling a *source* archive
2281 'make package' compiles a binary archive
2282 * Compile java files only on need
2283 * Add --cd and --setenv command line options to tesh
2284 * Out of source builds are not fully supported yet, but we are close
2285 * Enable supernovae and optimization flags by default for our users
2288 * Add layer to set up environment directly from lua, without XML.
2289 * The effect of gras_stub_generator can be achieved through
2290 lua too (check examples/gras/console/ping_generator.lua)
2292 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2294 SimGrid (3.4.1) stable; urgency=low
2296 The "Polishing easter eggs is probably a good idea" release.
2297 This is a bug fixes release only.
2301 * Fix a bug preventing the tasks from begin garbage collected.
2304 * Fix a bug occuring when a host involved in a communication fails.
2305 This was not detected properly by the other peer involved in the
2306 communication. Now, it's reported as a network error.
2309 * Warn the user about loop dependencies in data flow of DAX files
2310 * Obey the control-flow dependencies of DAX files
2313 * Add option "enable_smpi" allowing to not compile SMPI.
2314 Probably useful for the (Mac) users experiencing a build error here
2315 * Improve the detection of lua5.1 and ruby1.8
2317 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2319 SimGrid (3.4) stable; urgency=low
2321 The "Easter in Cargese" release. Also known as (major changes):
2323 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2324 ~> bindings were greatly improved
2325 ~> new tracing infrastructure for better visualization introduced
2327 * the "Welcome to configury modernity" release.
2328 ~> we switched from autotools to cmake, and improved our cdash
2331 A more detailled list of changes follow (full detail in svn log).
2333 Java Bindings: Various Cleanups
2334 * (install java-gcj-compat-dev on debian-like to use them)
2335 * Remove put/get: no need to export deprecated interface in Java
2336 Use send/receive instead.
2337 * Cleanup the examples and add a README per directory
2338 * Remove example autoDestination (that's the only way to go now)
2339 * Remove example explicitDestination (was a plain copy of basic)
2340 * Make JniException a runtime exception, so that there is no need to
2341 declare the fact that you may encounter such a beast. I guess that
2342 nobody will ever want to survive such error.
2343 * Create specific errors for each MSG case of failure:
2344 host failure, transfer failure, timeout, task cancelled
2345 * Cleanup the exceptions that may get thrown by each function
2346 * Other internal cleanups in Java bindings. Performance still bad :/
2347 Ruby and Lua Bindings: create them
2348 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2349 * That's new and great, you should try them out.
2350 Same functionalities than Java bindings, only even less polished
2352 * Kill the useless "rate" argument of SD_task_get_execution_time()
2353 Everyone used to provide -1 as a value, it was not used, and the
2354 semantic of a possible use wasn't even clear.
2355 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2356 either as comm costs or compute costs to mean that there is no
2357 such thing for that specific task.
2358 * Add a SD_task_set_name() function
2359 * Fix SD_task_unschedule() on typed tasks
2360 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2361 * In DAX loader, accept useless 'level' attributes to <job> since
2362 LIGO DAGs have them (seem to be to ease graphical representation).
2364 * Add an example masterslave_mailbox.c using send/receive and not
2365 the deprecated put/get interface.
2366 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2367 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2369 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2370 (declare MSG_USE_DEPRECATED to still have the old name)
2371 * Add a MSG_task_set_data() function
2372 * About trace replay (see examples/msg/actions):
2374 - Allow to work with splitted trace files for each process
2375 Give the specific trace file as argument of each process,
2376 and call MSG_action_trace_run(NULL)
2377 You can still have one merged file for all processes.
2378 - Fix implementation of collective operations
2379 * Allow task_execute() on 0-sized tasks (closes #10063)
2381 * This is the first release of SimGrid where SMPI is not considered
2382 beta anymore (even if some corners should still be improved)
2383 * Port over the new SIMIX_network submodule (internal refactoring)
2384 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2385 * Implement more missing elements of the standard:
2387 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2389 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2390 MPI_Scan() MPI_Get_processor_name()
2391 - Added implementation of missing case for Alltoall (warning: it's
2392 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2393 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2394 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2395 SMPI_MPI_Allgatherv()
2396 * Bug fixes include:
2397 - MPI_Waitsome() was broken
2398 - Allow relative includes in smpicc
2399 - Command line cfg argument 'reference_speed' was ignored...
2400 - Some functions did not properly lead to auto-benching of user code
2401 - smpicc passes -O2 by default (just like openmpi one)
2403 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2404 * Bug fixes about timeouts during communications
2405 * add SIMIX_message_sizes_output() as a pimple to write to file the
2406 amount of messages per size. Use gnuplot to get histogram.
2407 Pimple because that's the only user-visible function of simix,
2408 defined directly in xbt.h (irk, sorry)
2410 - Add a SIMIX_sem_get_capacity() function
2411 - Fix interactions with processe resume/suspende
2412 - release_forever() was stupidly broken
2413 - Fix SIMIX_display_process_status() for processes in a semaphore
2414 - Make SIMIX_sem_block_onto() user-visible
2415 * Refactoring context stuff:
2416 - Use pseudo-OOP for better modularity
2417 - reimplement SIMIX_process_kill() without process_schedule() so
2418 that the latter can take as invariant that it is called from
2420 - Merge context_start into context_new for sake of simplicity
2422 * Add a Vivaldi network model, coded live during SUD'10 ;)
2423 * Rename configuration variables to start a hierarchy:
2424 o cpu_model -> cpu/model
2425 o network_model -> network/model
2426 o workstation_model -> workstation/model
2427 * New configuration variables:
2428 o network/bandwidth_factor: correction to bandwith
2429 o network/latency_factor: correction to latency
2430 o netwotk/weight_S: correction to the weight of competing streams
2431 * Add a long description to the models, that users can see with such
2432 argument on the command line: --cfg=cpu/model:help
2433 * --help-models display the long description of all known models
2435 * config: add the ability to set a default value after registration
2436 Does not override any previously set value (e.g. from cmd line)
2437 * dict: allow to have integer key and data.
2438 When so, you need to use the following functions
2439 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2440 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2441 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2442 In contrary to regular dicts, the key is not malloced before copy.
2443 Mixing scalar and regular elements in the same dict is not tested
2445 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2446 Tracing for Visualization:
2447 * SimGrid is now instrumented in order to generate a trace file for
2448 visualization analysis: to use it, need to compile SimGrid with the
2449 "tracing" option enabled, and instrument the program using SimGrid with
2450 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2451 (among other functions).
2452 * The instrumentation only traces the platform utilization for now
2453 * Documentation to use the tracing functions and how to analyze the
2454 traces with the Triva tool is written.
2455 * More information about: SimGrid FAQ (in the section Tracing Simulations
2458 * We moved to cmake as default build system. Autotools support will
2459 be dropped soon. Check the FAQ for more info about how to use it.
2460 * Greatly improved our cdash/ctest interactions
2461 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2462 * Added memory checking tests with valgrind; lot of memleak fixing.
2463 This may be the first release of simgrid with so few memory issues
2464 * Added code coverage tests.
2465 Our coverage is still improvable, but at least we see it on cdash.
2467 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2469 SimGrid (3.3.4) stable; urgency=low
2471 The "Desktop Grid needs love too" release (also called Xmas release).
2473 Models improvements:
2474 * Major speedup in the maxmin system solving by using lazy evaluation
2475 Instead of solving completely the maxmin system at each iteration,
2476 only invalidate (and recompute) the modified parts.
2477 This new feature is enabled in default models but you can try to
2478 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2479 * Cas01 IMproved as default CPU model
2480 This CPU model is the same Cas01 model, but it uses the
2481 maxmin-selective-update flag and a heap structure to manage
2482 actions on SURF kernel.
2483 It reduces the complexity to find the next action to finish and,
2484 consequently, it's faster than the old Cas01.
2485 This is the new default CPU model (Cas01).
2486 * Rename the old Cas01 model to Cas01_fullupdate
2487 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2488 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2489 * CpuTI (CPU Trace Integration)
2490 A new CPU model whose objective is simulate faster when using
2491 availability trace files.
2492 Instead of using a full featured, over engineered maxmin system for
2493 CPU modeling, this model does the pre-integration of traces files
2494 to calculate the amount of CPU power available, and so, executes
2495 faster than the old CPU models.
2496 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2497 * Use LV08 as default network model since it gives better accuracy
2498 for small messages and shouldn't change things for big ones.
2499 Use --cfg=network_model:CM02 to get the previous behavior.
2502 ******************************************
2503 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2504 ******************************************
2505 * The new CPU model may changes simulations!
2506 The point is that events occurring at the exact same timestamp
2507 are not scheduled in the same order with the old and new
2508 version. This may be enough to completely change the execution
2509 of simulations in some cases.
2510 * The new network model will change simulations!
2511 This new model is more realistic than the previous one, so you
2512 should consider redoing your old experiments with this model.
2513 Sorry for the inconvenience.
2516 * Introduce the supernovae compilation mode
2517 When compiled that way, the whole SimGrid (or almost) is put in a
2518 single compilation unit and compiled in one shoot.
2519 This is to help gcc which has difficulties to inline stuff from one
2521 The speedup seem to be above 15%, althrough more tests are needed on
2522 amd64 to confirm that gain.
2525 * Port of MSG's mailbox on top of SIMIX network
2526 The put/get mechanism was greatly simplified on the way.
2529 * New SIMIX network module. Provides:
2530 - Mailbox: rendez-vous mechanism to find with who you want to speak
2531 - Synchronous send/recv: easier and hopefully faster since the
2532 logic is handled in the maestro process directly now
2533 - Asynchronous send/recv: you dreamt of it? It's here now
2534 Too bad that nobody cared enough to propagate the change to MSG.
2535 * Add semaphores as SIMIX synchronization mechanism.
2538 * new function SD_daxload(char*) to load a DAX file
2539 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2540 * Introduce typed tasks. Specify its kind and cost at creation.
2541 At scheduling, just give where it should be placed, and the cost
2542 for each involved resource is automatically computed.
2543 Existing constructors so far (more to come of course):
2544 - SD_task_create_comm_e2e() for end-to-end communication
2545 - SD_task_create_comp_seq() for sequential computation
2546 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2547 * new function SD_task_dump() for debuging display
2548 * new function SD_task_dotty(task,FILE*) writing to file the info
2549 about the task in dotty format
2550 * SD_task_dependency_exists() can now cope with having one of its
2551 arguments NULL. If so, it tests whether the other argument has any
2553 * Add getters on list of preceding/following tasks:
2554 SD_task_get_parents(task) and SD_task_get_children(task)
2555 * Add getters on amount of workstations and list:
2556 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2557 * Add getter on task kind: SD_task_get_kind(task)
2558 * Update the start_time and finish_time of tasks on completion/failure
2559 * Bugfix: Remove task from state swags when destroyed
2562 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2565 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2566 dynar with one separate thread per value of the dynar.
2567 * Change the prototype of xbt_thread_create(), sorry.
2568 Added a boolean parameter indicating whether we want to join this
2569 thread (used in SG only for now)
2570 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2573 * GTNetS wrappers should now be usable again (and betterly tested too)
2574 * Fix a major regression from 3.2 where the timeout provided to
2575 MSG_task_put_with_timeout() was used as absolute time before which
2576 the comm should be done.
2577 * Start to fix the <cluster> tag.
2578 - Internal links should be good now (beside of the loopback, which
2579 use the private link instead)
2580 - paths to the external world is still rather broken
2581 - the <route:multi> tag is just broken. Actually that's brain-dead.
2582 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2583 to make it less stupid
2584 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2585 * Fix a source-level compatibility glitch from 3.2: after defining
2586 MSG_USE_DEPRECATED, you can use the old name
2587 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2588 * Allow to compile from the SVN with automake 1.11
2589 * Fix some problems when using the "start_time" tag in deployment XMLs.
2590 * Fix #8569: XBT/synchro.h has redundant declarations
2591 * Fix #8563: MSG return values and exceptions
2592 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2593 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2594 Thanks to Robson Peixoto.
2595 * Fix a vicious bug in dictionaries inducing that some elements were
2596 not freed on xbt_dict_free()
2598 Portability report of this version:
2599 * Main portability targets:
2600 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2601 on (amd64/i386/ia64)
2602 - mac leopard on i386
2603 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2604 but nothing critical.
2605 * Other platforms: windows, AIX and others were not tested for this release
2607 Timing report of this version:
2608 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2609 scenario parameters). From 8h to a few seconds in desktop grid settings.
2610 * Supernovae brings about 25% speedup on i386.
2612 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2614 SimGrid (3.3.3) stable; urgency=low
2616 The "Need for Speed" release.
2618 The timings done to validate the 3.3.2 were faulty.
2619 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2621 The problem was a conversion from a manually handled vector to
2622 xbt_dynar_t on the critical path.
2623 xbt_dynar_foreach calls functions, inducing stack management crap.
2625 We inlined these functions and xbt_dynar_foreach is now breath taking.
2626 We also inlined xbt_swag_belong on the way.
2628 Here are some approximate speedup measurements (on master/slaves
2629 simulations lasting between 10s and 20s each):
2630 3.3.1 -> 3.3.2: about same performance
2631 3.3.2 -> 3.3.3: 40% speedup
2632 3.3.1 -> 3.3.3: 40% speedup
2633 3.3.1 with inline patch -> 3.3.3: 30% speedup
2635 Our reading is that the refactoring which occurred in 3.3.2 made us
2636 suffer much more from the xbt_dynar_foreach low performance, but
2637 once we solved this, this refactoring proved to be very performance
2638 effective. From the 40% speedup, somehow, 10% are due to the
2639 inlining and 30% to the refactoring.
2641 That's a pitty that gcc cannot inline functions placed in other files
2642 alone. We have to choose between:
2643 - break the encapsulation (by putting private data structures and
2644 accessors in headers files to help gcc)
2645 - live with low performance
2646 - switch to a decent compiler such as icc (not quite possible).
2648 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2650 SimGrid (3.3.2) stable; urgency=low
2652 The "Simplicity does not preceed complexity, but follows it" release.
2654 The main contributors of this release were (lexical order):
2655 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2658 * Extract the routing logic into its own object.
2659 (was dupplicated in network.c and workstation_LV07.c;
2660 Allows to implement other ways of storing that info)
2661 => kill now useless network_card concept
2662 - Use dynar to represent routes (instead of void** + int*)
2663 - kill link_set (use surf_network_model->resource_set instead)
2664 - Add a command-line option to choose the routing schema to use
2665 - Add three new models:
2666 * Floyd (shortest path computed at initialization)
2667 * Dijikstra (shortest path recomputed all the time)
2668 * Cached Dijikstra (shortest path computed on need)
2669 All these models where contributed by Silas De Munck, and are
2670 described in his ICCS09 paper.
2672 * Simplify model declaration
2673 (less redirections, less function to write when defining a model)
2674 - Factorize stuff between models:
2677 surf_model_resource_set(model)
2678 surf_model_resource_by_name(model, name)
2679 - Unify the types of models in s_surf_model_t (using an union)
2680 - Embeed fields of common_public directly into s_surf_model_t
2681 - Rename model methods:
2682 action_free ~> action_unref
2683 action_change_state ~> action_state_set
2684 action_get_state ~> action_state_get
2685 - Change model methods into functions:
2686 (model)->common_public->action_use ~> surf_action_ref
2688 * Implement a generic resource; use it as ancestor to specific ones
2689 (allows to kill duplicated code in models)
2690 Drawback: timer command don't need no name nor properties;
2691 workstation_CLM03 don't need no properties
2692 (but I guess we can live with those few bytes wasted)
2694 * Improve the action object model
2695 - implement a constructor avoiding dupplicated code about field
2696 initialization in generic_action part.
2698 * Kill the SDP model: it has an external dependency, is deprecated
2699 in flavor of modern lmm models, and didn't compile since a while
2702 * Relocation of the context module from XBT to SIMIX.
2703 (the context were decoupled from the simix processes, duplicating a lot of code)
2704 => a lot of code was factorized
2705 - less overhead is introduced during scheduling
2706 - simpler API for the context factory
2707 - the logic for process creation,destruction and manipulation was simplified
2708 * Simplification of the s_smx_process_t data structure.
2709 => accesing the simix level data associated to a process is faster now,
2710 and the code is a lot more readable.
2713 * Implement some more MPI primitives:
2714 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2715 -implementation: Bcast: flat or 2-ary tree (default),
2716 Barrier: 4-ary tree,
2718 Allreduce: Reduce then Bcast
2719 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2720 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2721 Alltoallv: flat tree, like ompi
2723 * Add support for optimized collectives (Bcast is now binomial by default)
2724 * Port smpirun and smpicc to OS X
2727 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2728 Shout out if you used it.
2731 * Display the list of still queued messages in SG mode when existing
2735 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2736 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2737 * Add xbt_str_from_file(FILE*)
2738 * Add xbt_dict_get_key achieving a linear reverse search
2739 * Remove the context module
2741 Portability report of this version:
2742 * Main portability targets:
2743 - Linux(debian)/x86/context
2744 - Linux(debian)/x86/pthreads
2745 - Linux(debian)/amd64/context
2746 - Linux(debian)/amd64/pthreads
2747 On these, we still have the eratic breakages of gras/pmm and
2748 amok/saturate_sg reported in previous version. We still think
2749 that the tests are the cause of the fault, not the tested code.
2751 - Mac OSX Leopard/x86/context
2752 Still false negative in tesh autotesting.
2753 Smpi still fails, but this time because readlink does not accept -f
2754 Everything seems to work properly beside of that.
2757 - AIX version 5.3 (only tested contexts this time)
2758 Smpi still fails there because mktemp is not installed.
2759 Everything seems to work properly beside of that.
2761 I managed to compile it for the first time, but several breakages.
2762 Won't delay the release for this exotic platform.
2764 * Windows: it's still lagging behind. If you want to help, please
2767 Timing report of this version:
2768 This version seem to be more than 5% faster than 3.3.1 (on linux
2769 64bits with contextes). The gain is less than expected, we are
2770 investigating this for next release.
2772 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2774 SimGrid (3.3.1) stable; urgency=low
2777 * Implement a --cfg-help to show existing configuration variables
2778 * Build chain do not require doxygen in maintainer mode
2781 * fix a bug on struct sizeof computation, which prevented the
2782 exchange of arrays of structs in some conditions
2783 - added a regression test about this in datadesc_usage
2784 * Allow the exchange of 0-long dynamic vectors.
2785 - for that, use -1 as indicator of dynamic size instead of 0
2786 - This implied to change any size from unsigned long to long,
2787 reducing a bit communication abilities, but I guess that with
2788 64bits being quite common, this is more than enough.
2789 - This also induce a protocol change, thus bumping network protocol
2790 version from 0 to 1 (if we have external users, we have to get
2791 clean on that point too ;)
2792 - added two regression tests about this in datadesc_usage
2793 * Be more verbose when propagating local exceptions
2794 This helps debugging.
2795 * Display the status of simulated processes when receiving SIGINT in
2799 * Allow to control the simulation from a trace file.
2800 New functions MSG_action_register() and MSG_action_trace_run()
2801 The first one allows to associate a function execution to each
2802 kind of action while the second one parses a trace file and
2803 triggers the corresponding actions within the system.
2804 For now, only a toy example is provided in examples/msg/actions
2805 * Add an exemple of process migration in examples/msg/migration
2806 * Fix a bug in task exchange which broke MSG_task_get_sender()
2807 Add a teshsuite regression test for that.
2808 [Bug: if MSG_task_get_sender() is called after sender exit,
2810 * Fix a bug which prevented suspend/resume to work properly
2811 * Display the status of simulated processes when receiving SIGINT
2812 This fixes a regression of v3.3. due to the introduction of SIMIX
2813 * Bug fixing in failure management:
2814 - trace could not start by a failure at time 0
2815 - failure during communications were not working
2818 * Add SIMIX_process_set_name() to change the name of the current
2819 process in the log messages.
2820 * Store smx_hosts in a dict since we only retrieve them by name
2821 * Move the configuration infrastructure to surf
2824 * Move the configuration infrastructure to surf
2827 * Massive internal cleanups:
2828 - Store internal structures on processes instead of hosts (allows
2829 to have more than one process per host, in addition of being more
2831 - Cleanup the initialization/finalization process
2832 - Kill a whole bunch of unneeded synchronization:
2833 processes run in exclusive manner within the simulator
2834 - Move queues from global tables to process data fields
2836 - now accept -platform and -hostfile arguments
2837 - Pass the right rank value to processes according to the hostfile
2838 * Compile the examples by default, and use them as regression tests
2839 * Implement MPI_Wtime()
2840 * Change the reference speed to a command line option
2843 * TCP_gamma can now be specified as command line option using
2844 --cfg=TCP_gamma:10000000.0
2845 * Change the --surf-path cmd line option into --cfg=path:
2848 * Also include strbuff from xbt.h public header
2849 * xbt_ex_display(): do not free the exception after displaying
2850 This allows to do more with the given exception afterward.
2851 Users should call xbt_ex_free() themselves.
2855 Portability report of this version:
2856 * Main portability targets:
2857 - Linux(debian)/x86/context
2858 - Linux(debian)/x86/pthreads
2859 - Linux(debian)/amd64/context
2860 - Linux(debian)/amd64/pthreads
2861 These targets fail about 1/10 of times on gras/pmm, but we believe
2862 that this is because of the test, not because of simgrid.
2863 amok/saturate_sg fails even more rarely, and the test may not be
2866 - Mac OSX Leopard/x86/context
2867 The test suite still spits tons of errors because some obscure
2868 force prevents us from removing the temporary directories
2869 arguing that they still contain some metadata I've never heard of.
2870 Smpi fails because seq is not installed.
2871 Everything seems to work properly beside of that.
2874 - AIX version 5.3 (both contexts and pthread)
2875 Smpi still fails there because mktemp is not installed.
2876 XML inclusions seems rosty on AIX.
2878 * Windows: it's still lagging behind. If you want to help, please
2881 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2883 SimGrid (3.3) stable; urgency=high
2887 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2888 [Malek Cherier & Mt]
2890 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2893 Until now, GRAS were using MSG as an interface to SURF. It was
2894 quite difficult because both interface have several differences
2895 (MSG channels vs GRAS sockets were the most notable point).
2897 This also opens the gate to SMPI (which should occur soon) and speed
2898 up simulations by to 40% (even if it were not the main goal).
2900 **************************************
2901 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2902 **************************************
2903 The point is that events occuring at the exact same timestamp are
2904 not scheduled in the same order with the old and new version. This
2905 may be enough to completely change the execution of simulations in
2906 some cases. Sorry for the inconvenience.
2908 * Cleanup and upgrade the XML format to push further scalability
2909 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2911 * Improve the testing infrastructure with tesh. Now a very large part of
2912 the code is tested not only by being run but also by checking that the
2913 output match an expected output [Mt].
2915 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2916 is really less memory-demanding, which should allow you to use
2917 larger files in SimGrid [AL].
2919 * Inform valgrind about our contextes, so that it becomes usable
2920 with the default (and more effecient) version of SimGrid
2921 [contributed by Sékou Diakite, many thanks]
2924 * Introduce a listener thread in charge of receiving incoming
2925 messages from the network. It allows to overlap communication and
2926 computation but most notably, it removes some stupid deadlocks due
2927 to the fact that so far, a process could not send and receive at
2928 the same time. This made most non trivial communication schema
2930 * Convert the PIDs from long int to int to match the MSG ones (and
2931 linux ones too) [Mt]
2932 * New function: gras_agent_spawn() to launch a new process on
2933 current host. Only working in simulation for now. [Mt]
2934 * New function: gras_os_hostport() returning a constant form (ie,
2935 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2938 * Make the backtrace of exceptions more human readable [Mt]
2939 * New module: xbt/str [Mt]
2940 a ton of string utility functions (split, join, printf to a newly
2941 allocated buffer, trim, etc)
2942 * New module: xbt/hash [Mt]
2943 SHA1 hashing algorithm (more to come if needed)
2944 * New module: xbt/synchro [Mt]
2945 synchronization tools (mutex and conditions) working the same way
2946 in simulation and in real life (mainly useful for GRAS, but not
2948 * New module: xbt/queue [Mt]
2949 classical producer/consumer synchronization scheme
2950 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2951 (using the classical functions will get serialized) [Mt]
2952 * Make dictionary internal table dynamic. No need to specify its size
2953 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2955 * Make sure the log channels are organized as a tree under windows
2956 (because of ANSI C compatibility issue, any channel were child of
2957 root directly) [Mt].
2960 * Cleaned many thing in surf and fixed a few bugs [AL].
2961 * Add a nice command line configuration mechanism to compose models [AL].
2962 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2963 the previous one (KCCFLN05). It relies on something that looks like
2964 a max-min sharing mechanism but cannot be written as such. A new solver
2965 was thus designed [AL].
2966 * Add a new solver to lmm. Based on Lagrange optimization and
2967 gradient-based descent, it enables to efficiently maximise systems s.a
2969 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2971 This solver enables to propose two new network models for TCP Reno and
2972 TCP Vegas based on Low's work. These models still need to be fully
2973 tested though [Pedro Velho].
2976 * Bug fix in SD_simulate. Now the time bound given as argument is
2978 * Use the new parallel task model (ptask_L07) as default.
2979 * Use the SURF command line configuration mechanism.
2980 * 0-size tasks (for synchronization) should now work.
2982 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2984 SimGrid (3.2) stable; urgency=high
2988 We still experience issues on this platform, but we believe that at
2989 least MSG is usable.
2991 GRAS API BREAKAGE (for simplification purpose, sorry):
2992 * the gras_msgtype_by_name is not used anymore. Instead of
2993 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2994 you can write (and must)
2995 gras_msg_send(toserver, "request", &request);
2996 - If you still want to pass a gras_msgtype_t to the function (to cache
2997 the type and avoid the lookup time), use the gras_msg_send_() variant.
2998 - Impacted functions:
2999 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3000 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3001 * The callbacks are now expected to return 0 when everything went well
3002 (just like the main() function)
3004 GRAS new features and improvements:
3005 * New module mechanism where user code can use per process globals [Mt]
3006 This is similar to gras_userdata_*() functions, but for libraries. It
3007 factorize some code developped over and over in the examples and AMOK.
3008 It has still to be documented and used (only amok/peermanagement is
3010 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3011 loss when they were small enough to fit into the buffer and sent quickly
3012 enough so that they can all get received in one shoot.
3013 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3014 if not found. Use the *_or_null() variant for the old semantic.
3015 * In gras_msg_handle, do not discard messages without callback.
3016 They are probably messages to be explicitly awaited later (ie, proofs of
3017 mis-synchronization in userland since they are sent before being awaited)
3018 No big deal usually.
3019 * gras_socket_meas_send/recv: semantic changed!
3020 The numerical arguments used to be (1) the total amount of data to send
3021 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3022 messages. This was need for the fool willing to send more than MAXINT
3023 bytes on quite fat pipes.
3026 * Do really rename the hostmanagement module to peermanagement. [Mt]
3027 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3028 but this is rather new and this was documented in the module
3029 documentation (poor excuses, I admit)
3030 * Bandwidth measurement semantic changed! This follows the changes to
3031 gras_socket_meas_send/recv explained above.
3034 * A sequential mode has been added to the workstations. When a workstation
3035 is in sequential mode, it can execute only one task, and the other tasks
3036 are waiting in a FIFO. [Christophe Thiery]
3039 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3040 model for SIMDAG. [Christophe Thiery]
3041 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3042 equal to 0, which caused some bad behaviors in
3043 saturated_constraint_set_update. I now use a threshold mechanism like in
3047 * When running manually src/testall, you select specific units [Mt]
3048 testall is the result of our cunit mechanism, and should replace all
3049 the scripty thingy around since bash don't run easily on billware.
3051 * A mallocator system has been added. [Christophe Thiery]
3052 Mallocators allow you to recycle your unused objects instead of freeing them
3053 and allocating new ones.
3055 Documentation update:
3056 * FAQ reworking + New FAQs:
3057 - "Valgrind spits tons of errors!" [Mt]
3058 - "How to repport bugs" [Mt]
3059 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3060 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3061 - Communication time measurement within MSG [AL]
3062 - I experience weird communication times when I change the latency [AL]
3063 * GRAS tutorial [Mt]
3065 - an introduction to the framework and to the used communication model
3066 - an initiatic tour introducing the most proheminent features:
3068 . Lesson 0: Installing GRAS
3069 . Lesson 1: Setting up your own project
3070 o Part 2: Message passing
3071 . Lesson 2: Exchanging simple messages
3072 . Lesson 3: Passing arguments to the processes (in SG)
3073 . Lesson 4: Attaching callbacks to messages
3074 . Lesson 5: Using globals in processes
3075 . Lesson 6: Logging information properly
3076 . Lesson 7: Using internal timers
3077 . Lesson 8: Handling errors through exceptions
3078 . Lesson 9: Exchanging simple data
3079 . Lesson 10: Remote Procedure Calling (RPC)
3080 . Lesson 11: Explicitely waiting for messages
3081 . Recapping of message passing features in GRAS
3082 - A HOWTO section containing:
3083 o HOWTO design a GRAS application
3084 More are due, of course. They will come latter. In the meanwhile, you can
3085 check the examples which are still here.
3087 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3089 SimGrid (3.1) stable; urgency=high
3093 There was a stack corruption somewhere, visible only when optimizing
3094 with these versions. [Vince]
3097 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3098 module that enabled to play with Directed Acyclic Graphs. It is built
3099 directly on top of SURF and provides an API rather close to the old
3100 SG. Some old codes using SG are currently under rewrite to check that
3101 all needful functions are provided. [Christophe Thiery]
3104 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3105 extension of the classical CLM03 model that gracefully handles
3106 failures. This is now the default model for MSG and GRAS. It doesn't
3107 handle parallel tasks yet though. [AL]
3108 * Bug fix: Weights were not correctly set in the network part.
3109 WARNING: This may have resulted in incorrect results with simulations
3110 where there are more than one flow on a given link. [AL]
3113 * After a (long ?) discussion on simgrid-devel, we have decided that the
3114 convention we had on units was stupid. That is why it has been decided
3115 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3116 WARNING: This means that all previous platform files will not work as
3117 such with this version! A warning is issued to ask users to update
3119 A conversion script can be found in the contrib module of the CVS, under
3120 the name contrib/platform_generation/surfxml_update.pl [MQ]
3123 * Bug fix: Processes were started in reverse order, wrt deployment file.
3124 WARNING: if your code relies on this bug, please fix it. [AL]
3125 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3126 being executed on two different locations. [AL]
3127 * Bug fix: Failures are now better supported thanks to Derrick's tests
3128 (there was many failure situations I hadn't thought of and that weren't
3129 correctly handled). [AL]
3130 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3134 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3135 Exception occurring on server-side are propagated back to client (!).
3137 API CHANGE: the callback changed their prototype. Change:
3138 int my_handler(gras_socket_t expeditor, void *payload_data) {
3140 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3141 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3143 * New! function: gras_msg_handleall to deal with all messages arriving
3144 within a given period.
3145 * New! function: gras_socket_server_range to get a server socket in a
3146 range of port numbers (ease to avoid port number conflicts) [MQ]
3147 * New! gras processes display their backtrace when they get a SIGUSR1
3148 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3149 Sweet to debug RL processes [MQ]
3153 - Do not force experiment sizes to be expressed in kb, or it becomes
3154 impossible to measure the latency this way (needs one byte-long tests)
3155 WARNING: this changes the amok_bw_* function semantic. [MQ]
3156 - Implements the link saturation stuff. [MQ]
3157 * Peer management module:
3158 New! module factorizing code that we wrote over and over [MQ].
3161 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3162 - Test units are placed directly into the library code, they get extracted
3163 automatically and placed into the src/testall binary.
3164 - Convert most of the XBT tests to this system.
3165 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3166 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3167 live in the same C code. This required to change a little bit the API
3168 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3169 * New module: sparse graph structure with basic algorithms (this is work
3170 in progress and the API is not considered to be frozen yet). [AL]
3171 * Display more information on backtraces: source line & function names are
3172 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3173 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
3174 * Reworked a little bit some #include statements to load only required
3175 headers. Some user code that relied on SimGrid to include stdlib or
3176 stdio may need to include it by themselves. [AL]
3177 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3179 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3180 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3181 but I guess that almost nobody used those parts. [MQ]
3183 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3185 SimGrid (3.0.1) stable; urgency=low
3188 * Unfortunately, I had missed 5 misnamed functions:
3189 xbt_fifo_item_t xbt_fifo_newitem(void);
3190 void xbt_fifo_freeitem(xbt_fifo_item_t);
3191 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3192 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3193 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3194 They're now deprecated. Please use their new versions:
3195 xbt_fifo_item_t xbt_fifo_new_item(void);
3196 void xbt_fifo_free_item(xbt_fifo_item_t);
3197 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3198 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3199 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3201 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3202 * Documentation: xbt_log module unmercifully reworked [MQ]
3203 * Bugfix: there was a problem with the ending of contexts with
3204 the pthread backend. It caused some weird deadlock or behavior
3205 depending on the pthread implementation. [AL]
3206 * Bugfix: get the exceptions raised in the simulator repport where
3207 and why they come from when they are not caught in time [AL, MQ]
3210 * Bugfix: Do repport the error when two non-connected hosts try to
3211 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3214 * Add additionnal checkings on communications. Assert that two
3215 communicating hosts are connected by a set of links... [AL]
3218 * Add additionnal checkings on channel values in communication [AL]
3219 * New: MSG_task_get_source to see on which host a task was generated [HC]
3220 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3221 the number of tasks waiting to be received on channel and sent
3223 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3224 waits for the first task coming from a given host.. [AL]
3226 GRAS new functionnalities: [MQ]
3227 * Enhance the parsing macro to allow the size of multidimentional objects
3228 to be given thru annotations.
3229 * New example (and documentation): Matrix Multiplication a la RPC
3230 (as when I was young!) and fix a bunch of bugs found on the way.
3232 GRAS performance improvements: [MQ]
3234 * Reduce the amount of cbps creation/destruction by making it static to
3235 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3238 * Change libdata to a set so that we can search for stuff by ID (and thus
3239 reduce the insane amount of dict lookups)
3242 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3243 RL communication on top of SG since it's so uneffective.
3244 It may also be used for inter-thread communication in RL, one day.
3245 * Use gras_datadesc_copy() to exchange messages on top of SG
3247 - improve message exchange performance on top of SG
3248 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3249 * Don't exchange on the network the size of the used part of buffer,
3250 instead, specify the possible buffer size to read().
3252 - reduces the amount of read/write calls (one pair per exchange)
3253 - reduces the amount of exchanged data (the size)
3254 - allows to retrieve all arrived data on receiver side, if we don't need
3255 it right now (subsequent read will peek the buffer)
3256 - allows the receiver to proceed with the begining of the stream before
3257 everything is arrived
3258 - make it possible to build an iov transport (using readv/writev)
3260 - take care of the data with non-stable storage (like stacked data),
3262 * If possible, TCP send uses vector I/O (when writev() is here)
3263 - Don't use it for receive since we send data sizes and data on the
3264 same stream, so we wouldn't be able to chain large amount of chunks
3265 before having to flush the stuff to read the size.
3266 * Rework the transport plugin mechanism to simplify it and reduce the
3267 amount of pointer dereferencement when searching for the right function
3270 * I guess that now, we do almost as few system calls as possible while
3271 doing as few data copy as possible.
3273 To improve it further, we could try to send all the sizes first and then
3274 all the data (to use iov on receiving size), but it's only a partial
3275 solution: when you have 2 dimensional data, the sizes of the second
3276 dimension is data of the first dimension, so you need 3 streams.
3278 I'm not sure the potential performance gains justify the coding burden.
3280 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3282 SimGrid (3.00) stable; urgency=high
3285 * New! Give the possibility to hijack the surf parser and thus bypass
3286 MSG_create_environment and MSG_launch_application. Have a look at
3287 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3289 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3291 SimGrid (2.96) unstable; urgency=low
3296 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3297 This deprecates the xbt_error_t mechanisms.
3298 It modifies (simplifies) all XBT and GRAS API.
3299 MSG API keeps unchanged (exceptions raised by XBT are caught from
3300 within MSG and masked with existing error handling facilities)
3303 * New! Add a FATPIPE model. [AL]
3304 * New! Add a parallel task model. [AL]
3305 * New! Add automatically a loopback interface (in the default
3306 network model) if none was precised.
3309 * Bugfix: MSG_process_resume now works with the current running process.
3311 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3312 * Modification of MSG_task_get_compute_duration. Once a task has been
3313 processed, the value returned by this function is now equal to 0. [AL]
3314 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3315 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3316 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3317 of a m_task. For now, MSG_task_cancel only works with computation
3319 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3320 of the processor (in Mflop/s) regardless of the current load on the
3322 * API Change: use proper naming convention for MSG_getClock and
3323 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3325 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3326 This function changes the priority of a computation task. This priority
3327 doesn't affect the transfer rate. A priority of 2 will make a task
3328 receive two times more cpu power than the other ones. This function
3329 has been added to suit the needs of Nguyen The Loc and hasn't been that
3330 much tested yet. So if it fails, please report it and send me your code.
3332 * API Change: removed all functions and types that were marked "deprecated"
3333 since many months. Renamed MSG_global_init_args to MSG_global_init.
3335 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3337 SimGrid (2.95) unstable; urgency=low
3340 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3341 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3342 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3345 * Bugfix: Allow absolute paths for platform description files [MQ]
3346 * Bugfix: do free the variables after use. Leads to drastic performance
3348 * Implement max_duration (ie, timeouts) on resources [AL]
3351 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3353 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3354 the needed support (timeouts) [AL]
3357 * Implement measurement sockets. You can now get the bandwidth between two
3358 hosts thanks to AMOK (see below). [MQ]
3359 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3360 dynar over the network (yeah) [MQ]
3361 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3362 * simple/static token-ring example. [Alexandre Colucci and MQ]
3363 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3364 in place before [MQ]
3367 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3368 type parsing. "short" and "long" modifiers were also ignored; other
3369 modifier (such as reference level) are still ignored. [MQ]
3370 * Embeed the buffer size within the buffer itself on SG. [MQ]
3371 That way, send() are atomic and cannot get intermixed anymore (at least
3372 the ones which are less than 100k; bigger messages still have the issue)
3373 * Array size pushed by the field, not by the field type (or each
3374 and every long int will push stuff to the cbps) [MQ]
3375 * use select() to sleep since it allows to portably sleep less than one
3378 GRAS (minor cleanups)
3379 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3382 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3383 * type category 'ignored' killed as it was never used and were difficult
3385 * whether a type can cycle or not is now a flag, leaving room for more
3386 flags (as "ignored", if we feel the need one day ;)
3387 * Rename raw sockets to measurement sockets since "raw" has another
3388 meaning in networking community.
3391 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3392 and offers features not belonging to GRAS but that most applications
3393 need. One day, it may be a set of plugins loadable at runtime.
3394 * New module: bandwidth
3395 bandwidth measurement between arbitrary nodes running this module. [MQ]
3397 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3399 SimGrid (2.94) unstable; urgency=low
3401 The first beta release of SimGrid 3 !
3405 * Update the main page and the FAQ. Adding references to gforge.
3408 * Add a gras_os_getpid function.
3411 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3412 * Extend the logs so that they also print PID, hostname, date, ... if
3414 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3415 and kill the old version which were in testsuite/
3416 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3420 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3421 for users. Instead, each programming environment (GRAS, MSG,...) use it
3422 when needed to provide such a feature to users.
3423 Don't shortcut the mechanism or you will also shortcut the virtualization
3424 you need on the simulator.
3428 * Cleanups in configury with regard to compile optimization/warning flags.
3429 Also add -fno-loop-optimize to any powerpc since it's the optimization
3430 killing gcc (< 3.4.0).
3431 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3432 needed by MSG examples complications
3433 * Borrow configury beautifications from PHP
3436 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3437 warning (thanks loris for stumbling into this one).
3438 * Bugfix: stop loading private headers (gras_config.h) from the public
3442 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3443 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3444 RL would imply a DNS resolver, which is *hard* to do in a portable way
3445 (and therefore delayed).
3446 * Implement a real timer mechanism and use it in timing macros. This allows
3447 to avoid rounding errors and get a 0.000005 sec precision in timing
3448 macros. While I was at it, various cleanups:
3449 - allow to declare more than one timed section per file (fix a stupid bug)
3450 - move some private declaration to the right place
3451 - merge conditional execution and timing macros into emulation module
3452 - document the module
3453 - make sure the module cleanups its mess on gras_exit
3454 * Documentation improvements:
3455 - (new) how to compile applications using GRAS
3456 - (new) emulation support (timing macros)
3458 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3460 SimGrid (2.93) unstable; urgency=low
3462 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3465 - Use Paje properly where used. Still to be sanitized properly.
3466 - Portability fix: Add an implementation of the contexts using pthread
3470 - Add xbt_procname(): returns the name of the current process.
3471 Use it to show the current process's name in all logging.
3473 - fix detection of older flex version and the reaction, since we do
3474 depend on modern ones (we use lex_destroy)
3475 - Better separation of SG and RL in the libs: remove all simulation code
3476 from libgras. As a result, this lib is now only 200k when stripped.
3477 Some of the xbt modules may also be duplicated (two sets and such) and
3478 should be cleaned/killed before SG3.
3479 - Insist on using xlC on AIX because of weird problems involving gcc there.
3480 - Cleanup the make remote stuff. This is now done by scripts
3481 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3482 mainly for our private use, but we're working on changing them to user
3485 - Bugfix: flush the socket on close only if there is some *output*.
3486 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3488 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3491 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3493 SimGrid (2.92) unstable; urgency=low
3495 Alpha 3 on the path to SimGrid 3
3499 - New! First try of benchmarking macros.
3500 - New! First try so that gras_stub_generator generate deployment and
3501 remote compilation helpers.
3503 - Bugfix: Initialization fix in msg_test.
3507 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3509 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3511 - API Change: gras_os_sleep to take the amount of seconds as a double.
3512 Accepting an int was error prone since it was the only location where
3513 seconds were coded as such. It leaded to damn rounding errors.
3514 - Bugfix: Hard to belive that timers ever worked before this.
3516 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3518 SimGrid (2.91) unstable; urgency=low
3520 Alpha 2 on the path to SimGrid 3
3524 - Bug fix in the lmm_solver.
3526 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3527 through the function MSG_paje_output.
3528 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3529 - It is possible to bound the rate of a communication in MSG with
3530 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3531 to write it in the changelog).
3532 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3537 - Various cleanups to the autotools stuff
3538 - Begin to move Gras examples to examples/gras/
3539 - Let make distcheck work again (yeah!)
3541 - documentation overhauled using doxygen.
3542 gtk-doc-tools is dead in SimGrid now.
3543 - Automatically extract all existing logging categories, and add the list
3544 to the documentation (long standing one, to say the less)
3546 - Cleanup the known architecture table. Reorder the entries to group what
3547 should be, and use a more consistent naming scheme.
3548 (some of the test dataset are still to be regenerated)
3549 - New! Allow library to register globals on each process just as userdata
3551 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3552 lookup time (for now).
3553 Use it in msg and trp.
3554 This cleans a lot the internals and helps enforcing privacy of the
3555 headers between the gras components.
3556 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3557 - Bugfix: gras_os_time was delirious in RL.
3558 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3560 - Reenable GRAS now that it works.
3562 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3564 SimGrid (2.90) unstable; urgency=low
3566 Alpha 1 on the path to SimGrid 3
3568 * It is a long time since the last release of SimGrid. I'm sorry about
3569 that but as I had told you, I was rewriting a lot of things. I apologize
3570 to those who had been reporting bugs to me and that I had not answered.
3571 If your bug is still in the new version, please tell me. Here is a
3572 summary of the main changes.
3574 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3575 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3576 portability, speed, and a lot more but you'll figure it out later.
3577 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3578 I wanted to release the new MSG as soon as possible and I have
3579 broken GRAS, which is the reason why, for now, only MSG is fully
3580 functional. A laconic description of these projects is available
3581 in the documentation.
3583 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3584 kernel optimized for our needs (called SURF but only the developers
3585 should use it). Hence, MSG is now roughly 30 times faster and I think
3586 that by rewriting a little bit MSG, I could event speed it up a little
3587 bit more. Beside the gain in speed, it is also much easier to encode a
3588 new platform model with SURF than it was with SG. More to come...
3590 * REVOLUTION 3: I have tried to change a little as possible the API of
3591 MSG but a few things really had to disappear. The main differences
3592 with the previous version are:
3593 1) no more m_links_t and the corresponding functions. Platforms are
3594 directly read from a XML description and cannot be hard-coded
3595 anymore. The same format is used for application deployment
3596 description. The new format is described in the documentation.
3597 Have a look in tools/platform_generation. There is a tiny script
3598 that converts from the old platform format to the new one. Concerning
3599 the application deployment format, parsing the old one is tricky.
3600 I think most of you should however be able to convert your files. If
3601 it is really an issue, I can write a C code that does the conversion.
3603 2) the toolbox tbx does not exist anymore. We now have a library
3604 with much more data-structures but without the hash-tables (we have
3605 dictionaries that are much faster).
3607 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3609 *****************************************************************************
3610 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3611 * don't feel like converting the oldies. (Mt) *
3612 *****************************************************************************
3615 Version 2.90: "the long awaited one"
3616 - Finished rewriting and debugging MSG. Rewrote the documentation.
3617 - disable GRAS for now since it needs to be ported to the newest SG
3620 - Finish the port to windows (using mingw32 for cross-compile)
3623 - Main loop and datastructures of SURF. A cpu resource object is
3624 functional. Surf can thus be used to create cpu's with variable
3625 performance on which you can execute some actions.
3627 2004-11-15 Martin Quinson
3628 - Port to ARM. Simply added the alignment and size descriptions. Should
3629 work, but the ARM machines are so slow that I didn't had the opportunity
3630 to 'make check' over there yet.
3632 2004-11-15 Arnaud Legrand
3633 - Trace manager now written. It uses a heap structure and is therefore
3634 expected to be efficient. It may however be speeded up (particularly
3635 when many events occur at the same date) by using red and black
3636 trees. One day maybe...
3637 - Max-min linear system solver written. It uses a sparse matrix
3638 structure taking advantage of its expected use. Most operations are
3639 O(1) and free/calloc are called as few as possible. The computation of
3640 the minimum could however be improved by using a red and black tree
3643 2004-11-03 Arnaud Legrand
3644 - Rename every gras_* function that was in xbt/ to its xbt_
3646 - Add a heap and a doubly-linked list to xbt
3647 - Added a dichotomy to the dictionaries. make check works as well before
3648 so I assume that the patch is correct. I do not know however if things
3649 run effectively faster than before now. :)
3651 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3652 SimGrid, and the version number is bumped to 2.x
3654 2004-10-29 Martin Quinson
3655 - Introduction of the remote errors.
3656 They are the result of a RMI/RPC on the remote machine.
3657 ErrCodes being scalar values, you can't get the host on which those
3658 errors did happen. Extending the error mechanism as in Gnome is possible.
3659 No idea yet whether it is a good idea.
3661 2004-10-28 Martin Quinson
3662 - Interface revolution: the Starred Structure Eradication.
3663 I used to do typedef struct {} toto_t; and then handle *toto_t.
3664 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3665 - ???_t is a valid type (builded with typedef)
3666 - s_toto_t is a structure (access to fields with .)
3667 - s_toto is a structure needing 'struct' keyword to be used
3668 - e_toto_t is an enum
3669 - toto_t is an 'object' (struct*)
3671 typedef struct s_toto {} s_toto_t, *toto_t;
3672 typedef enum {} e_toto_t;
3673 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3674 s_toto_t) is private.
3676 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3677 so that it can set the variable to NULL. It was so for dicts and sets,
3678 it changed for dynars.
3680 - Fix a bunch of memleaks in dict_remove
3681 - Fix a bug in sg/server_socket opening: it failed all the time.
3683 2004-10-07 Martin Quinson
3684 - Speed up dynar lookup operation a bit.
3686 gras_dynar_get is dead.
3688 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3689 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3690 (which gives you the address of the stored data).
3692 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3693 copy of the data using an affectation to do the job and not a memcpy.
3695 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3696 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3698 It does not really speedup the dynar test because they are
3699 setting elements all the time (and look them seldom). But the dict does
3700 far more lookup than setting.
3702 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3704 2004-10-05 Martin Quinson
3705 - Allow to (en/dis)able the cycle detection at run time.
3707 Whether we should check for cycle or not is now a property of each
3708 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3709 datadesc_cycle_unset allow to remove this property when previously set.
3711 Note that the cycle detection is off by default since it impacts the
3712 performance. Watch the data you feed GRAS with ;)
3714 This property is hereditary. Any element embedded in a structure having it
3715 set have it set for the time of this data exchange.
3717 You should set it both on sender and receiver side. If you don't set it on
3718 sender side, it will enter an endless loop. If you forget on receiver
3719 side, the cycles won't be recreated after communication.
3721 - Header reorganization.
3722 Kill gras_private.h, each submodule must load the headers it needs.
3724 2004-10-04 Martin Quinson
3725 - Interface revolution: do not try to survive to malloc failure.
3727 Now, gras_malloc and friends call gras_abort() on failure.
3728 As a conclusion, malloc_error is not a valid error anymore, and all
3729 functions for which it was the only gras_error_t return value are
3730 changed. They now return void, or there result directly.
3731 This simplify the API a lot.
3733 2004-09-29 Martin Quinson
3734 - Re-enable raw sockets.
3735 Created by gras_socket_{client,server}_ext;
3736 Used with gras_raw_{send,recv}
3739 It should allow to kill the last bits of gras first version soon.
3741 This is not completely satisfactory yet (duplicate code with
3742 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3745 - Simplify transport plugin (internal) interface by not passing any
3746 argument to _server and _client, but embedding them in the socket
3749 2004-09-28 Martin Quinson
3750 - Finish the port to AIX.
3751 autoconf was my problem (segfault within the malloc replacement
3752 function. No idea why)
3754 2004-09-16 Martin Quinson
3755 - Fix some size_t madness on 64bit architectures.
3757 2004-09-08 Martin Quinson
3758 - Reduce the number of system headers loaded, overload some more system
3759 calls (such as malloc to cast the result of the system one, and work
3761 - Fix and reintroduce the config support
3763 2004-09-07 Martin Quinson
3764 - Source code reorganization to allow Arnaud to surf all over there.
3765 - Allow to document the logging categories.
3766 - Remove all uppercase from logging categories and useless cleanup in names.
3768 2004-08-18 Martin Quinson
3769 Version 0.6.2 (protocol not changed; API changed)
3770 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3771 gras_error_t), and NULL when not found. Functions expecting a msgtype
3772 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3773 hopefully usefull message.
3774 - Portability to prehistoric sparcs again
3776 2004-08-17 Martin Quinson
3777 Version 0.6.1 (protocol not changed; ABI not changed)
3778 - prealloc some buffers to speed things up
3780 2004-08-11 Martin Quinson
3781 Version 0.6 (protocol not changed; ABI expended)
3782 - The parsing macro can deal with the references, provided that you add
3783 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3785 2004-08-09 Martin Quinson
3786 Version 0.5 (protocol not changed; ABI changed)
3787 - Allow to off turn the cycle detection code in data exchange at
3788 compilation time. It should be at run time, but I'm short of time (and
3789 the config stuff is still broken). That way, we keep dict out of the
3790 critical path, which is good because the performance is poor:
3791 - search not dichotomial yet
3792 - dynar give no way to access their content and memcpy everytime
3793 - In composed data description (struct, ref and so on), stop foolness of
3794 keeping the subtype's ID, but store the type itself. This keeps sets out
3795 of the critical path, which is good since they rely on dynar and
3796 dictionnaries. The only loose of that is that we cannot detect the
3797 redeclaration of a structure/union with another content (but I'm not sure
3798 the code detected well this error before anyway). We still can detect
3799 the redefinition discrepancy for the other types.
3800 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3801 it breaks the code because of type-punning used all over the place).
3802 This breaks on all non-gcc architectures (for now).
3804 All those changes (plus the buffer of last time) allow me to gain 2 order
3805 of magnitude on cruel tests consisting of 800000 array of integers on two
3806 level of a hierarchical structure (200 secondes -> 4 secondes)
3809 - the selector of reference must now return the type it points to, not
3810 the ID of this type.
3812 2004-08-06 Martin Quinson
3813 Version 0.4 (protocol changed; ABI not changed)
3814 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3815 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3816 - Implement a buffer transport, and use it by default (it relies on tcp in
3817 real life and on sg in simulation).
3818 That's a bit hackish since I had a new field to the structure to store
3819 its data without interfering with the subtype ones. Inheritance
3820 is tricky in C. And that's a kind of reverse inheritance with one class
3821 derivating two classes. Or maybe a game with java interfaces. Anyway,
3822 that's damn hard in C (at least).
3823 Moreover, I got tired while trying to ensure plugin separation and
3824 genericity in SG mode. MSG wants me to do weird things, so let's go for
3825 cruel hacks (temporarily of course ;).
3826 See comment in transport_private.h:71
3827 - do not include all the _interface headers in private but in the files
3828 which really need them (to cut the compilation time when they are
3831 2004-07-26 Martin Quinson
3832 Version 0.3 (protocol not changed; ABI changed)
3833 - Major overhault of the datadesc interface to simplify it:
3834 - shorted the function names:
3835 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3836 - add a trivial way to push/pop integers into the cbps without malloc.
3837 This allows to make really generic sub_type description, which simply
3838 pop their size of the stack.
3839 - add a function gras_datadesc_ref_pop_arr() which does what users want
3840 most of the time: Declare a dynamic array (which pops its size of the
3841 stack) and declare a reference to it. Poor name, but anyway.
3842 - kill the post-send callback, add a post-receive one
3844 2004-07-23 Martin Quinson
3845 Version 0.2 (protocol changed; ABI changed)
3846 - add some testing for cpbs in the test cases, and fix some more bugs.
3847 This invalidate again the little64 data file, since I cannot regenerate
3849 - remove an awfull optimization in the logging stuff, allowing me to:
3850 - understand it again
3851 - learn gcc how to check that the argument match the provided format
3852 - fix all errors revealed by gcc after that
3853 - internal keys of dict are not \0 terminated. Deal with it properly in
3854 loggings instead of segfaulting when the user want to see the logs :-/
3856 2004-07-22 Martin Quinson
3857 - Fix some stupid bug preventing cbps (callback postit) from working
3859 2004-07-21 Martin Quinson
3860 - Some documentation cleanups
3861 - remove the useless last argument of msgtype_declare
3862 - rename the Virtu functions to fit into the 'os' namespace
3863 - move headers src/include -> src/include/gras/ and stop fooling with
3864 gras -> . symbolic link
3865 - make distcheck is now successful
3867 2004-07-19 Martin Quinson
3869 - Build shared library also
3870 - Install html doc to the right location
3871 - stop removing maintainer files in make clean
3872 - build tests only on make check
3874 2004-07-13 Martin Quinson
3876 - No major issue in previous version => change versioning schema
3877 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3878 corresponding dataset.
3880 2004-07-11 Martin Quinson
3882 - Get it working with any kind of structure (we can compute the padding
3883 bytes remotely for all the architectures I have access to)
3884 - Implement the structure parsing macro (still not quite robust/complete)
3885 - Improvement to the remote testing toysuite
3887 2004-07-10 Martin Quinson
3888 [autoconf mechanism]
3889 - get ride of a bunch of deprecated macros
3890 - actually run the test for two-compliment, not only compile it :-/
3891 - test whether the structures get packed (and bail out if yes. Damn.
3892 Alignment is a serious matter)
3893 - test whether the structures get compacted (but respecting the alignment
3894 constraints of each types)
3895 - test whether the array fields of structures can straddle alignment boundaries
3897 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3898 (took me 2 hours to find that bug, looking at the wrong place)
3899 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3900 before sending/receiving and are used to compute the offsets of fields
3901 - Given that padding size depend even on compiler options, keep track of
3902 alignment and aligned_size only for the current architecture. Not a big
3903 deal since we send structure fields one after the other (seems
3905 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3906 program, let it work on linux/gcc/little32. portability todo.
3908 2004-07-08 Martin Quinson
3909 - import and improve remote compilation support from FAST
3910 - make sure make check works on half a dozen of machines out there
3912 2004-07-07 Martin Quinson
3913 Let's say it's version 0.0.3 ;)
3914 - Implement conversions (yuhu!)
3915 - Let it work on solaris (beside conversion, of course)
3916 - Stupid me, using rand() to generate the conversion datatests in not wise.
3918 2004-07-06 Martin Quinson
3919 - Let make dist work, since I'm gonna need it to compile on remote hosts
3920 - Let Tests/datadesc_usage write the architecture on which the file was
3921 generated as first byte.
3922 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3923 architecture descriptions.
3924 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3925 on those architectures.
3926 - Optimization: send/recv array of scalar in one shoot
3928 2004-07-05 Martin Quinson
3929 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3931 - Plug a whole bunch of memleaks
3932 - each process now have to call gras_{init,exit}. One day, their log
3933 settings will be separated
3934 - Continue the code factorisation between SG, RL and common in Transport.
3936 2004-07-04 Martin Quinson
3938 - Redistribution between SG and RL.
3939 We wanna have to accept in SG, so move accepted related parts of RL in
3940 the common part. (more precisely, the dynar of all known sockets is no
3941 more a static in transport.c, but part of the process_data)
3943 [gras_stub_generator]
3944 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3946 2004-07-03 Martin Quinson
3947 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3948 them all up in one shot)
3950 - Fix the ugly case of reference to dynamic array.
3951 - New (semi-public) function gras_datadesc_size to allow the messaging
3952 layer to malloc the needed space for the buffer.
3954 - gras_socket_close now expect the socket to close (and not its address to
3955 put NULL in it after it). This is because the socket passed to handlers
3956 is one of their argument (=> not writable).
3958 - propagate the interface cleanup from last week in datadesc, ie remove a
3959 superfluous level of indirection. User pass adress of variable
3960 containing data (both when sending and receiving), and not of a variable
3961 being a pointer to the data. Let's say that I like it better ;)
3962 The price for that is constructs like "int msg=*(int*)payload" in
3963 handlers, but it's a fine price, IMHO.
3965 - Let it work in RL (yuhu)
3967 2004-06-21 Martin Quinson
3969 - porting SG plugin and SG select to new standards (works almost).
3970 - plug memleaks and fix bugs around.
3973 - cleanup the prototype of data recv and force users to specify when they
3974 want to handle references to objects. Test case working even for cycles.
3975 - plug memleaks. Valgrind is perfectly ok with this.
3977 2004-06-12 Martin Quinson
3979 - cleanup the separation between plugin and main code in plugin creation
3981 2004-06-11 Martin Quinson
3983 - Reput hook for raw sockets, needed for BW experiments
3984 - kill a few lines of dead code
3985 [Data description] Interface cleanup
3986 - gras_datadesc_by_name returns the searched type or NULL.
3987 That way, no variable is needed to use a type desc once, which makes
3989 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3990 two parameters were strings (field name, type name), leading to
3992 [Dicos] Interface cleanup
3993 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3994 This is consistant with the dynar API.
3996 2004-04-21 Martin Quinson
3998 - Porting to new standards.
4000 - interface cleanup.
4001 There is no bag anymore, no need to take extra provision to mask the
4002 pointers behind "ID".
4003 Better splitup of functions between files create/exchange/convert.
4004 This is still a bit artificial since convert and receive are so
4005 interleaved, but anyway.
4007 - add a queued message list to procdata (the ones not matching criteria
4009 - factorize some more code between SG and RL wrt procdata
4011 - use gras_exit in example to track memleaks
4012 - get rid of gs_example now that GS is properly integrated into gras
4013 - update run_test to integrate the latest tests (datadesc)
4015 - rename WARNINGn macros to WARNn since it prooved error-prone
4017 2004-04-19 Martin Quinson
4019 - register init/exit functions within gras module mechanism
4020 - send/receive function.
4021 Convertion is not implemented, but short-cutted if not needed.
4022 struct/array elements are sent one by one (instead of block-wise), but
4023 nobody really cares (yet). Get a prototype before optimizing.
4024 - tests (using a file socket) for DD send/receive on:
4025 - base types: int, float
4026 - array: fixed size, string (ie ref to dynamic string)
4027 - structure: homogeneous, heterogeneous
4028 - chained list, graph with cycle
4029 Believe it or not, valgrind is not too unhappy with the results. The
4030 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4034 - Bugfix when using a filename explicitely (instead of '-')
4036 2004-04-09 Martin Quinson
4038 - factorize more code between RL and SG in socket creation
4039 - Complete the implementation and tests of:
4041 o file (only in RL, and mainly for debugging)
4043 I lost 3 days to design a portable address resolver, and then decided
4044 that the prototype mainly have to run on my box.
4045 Addressing portability too early may be like optimizing too early :-/
4047 - use gras_init in the Tests instead of the crappy parse_log_opt
4048 (the latter function is removed)
4049 [Conditional execution]
4050 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4051 [Code reorganisation]
4052 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4053 Build examples against the RL library, since there is no way to disable
4054 its creation for now.
4056 For information, the beginning of coding on GRAS was back in june
4057 2003. I guess that every line has been rewritten at least twice since