1 ----------------------------------------------------------------------------
3 SimGrid (3.23.3) NOT RELEASED YET (v3.24 expected September 23. 7:50 UTC)
6 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
7 for consistency with pthread_barrier_wait()
8 - Host::get_englobing_zone() returns the englobing netzone
9 - Actor::on_destruction is now called in the destructor
10 Actor::on_termination new signal called when the actor terminates
14 - convert a new set of functions to the S4U C interface and move the old MSG
15 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
18 - Fortran bindings for DVFS have been removed.
19 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init
20 - SMPI can now be selected by cmake's find_module(MPI) with
21 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
22 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
23 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
26 - Option model-checker/hash was removed. This is always activated now.
27 - New option smpi/buffering controls the MPI buffering in MC mode.
28 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
29 This is useful to check for MPI compliance.
32 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
33 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
36 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
37 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
38 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
39 - FG#34: SG_BARRIER_SERIAL_THREAD?
40 - FG#35: model-checker does not like buster-produced binaries
42 ----------------------------------------------------------------------------
44 SimGrid (3.23.2) July 8. 2019
47 - Nicer introduction page.
48 - Migrate the "Deploy your application" page to the new doc.
49 - Move Java as a subtree of MSG.
52 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
55 - Drop xbt_dynar_sort_strings().
58 - Really fix FG#26: Turning off a link should raise NetworkFailureException
59 - FG#27: Wrong exception thrown to wait_any when link is turned off
60 - GH#328: Java: Canceling multiple tasks in a single vm/host
62 ----------------------------------------------------------------------------
64 SimGrid (3.23) June 25. 2019
66 The Exotic Solstice Release.
69 - SunOS and Haiku OS support. Because exotic platforms are fun.
70 - Stop setting random seed with srand48() at initialization.
71 - Use addr2line as a fallback for stacktraces when backtrace is not available.
72 - Build option -Denable_documentation is now OFF by default.
73 - Network model 'NS3' was renamed into 'ns-3'.
76 - Simgrid can now hopefully be installed with pip.
79 - wait_any can now be used for asynchronous executions too.
82 - New log appenders: stdout and stderr. Use stdout for xbt_help.
86 - 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
87 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
88 - MPI interface is now const correct for input parameters
89 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
90 - Fortran bindings for async collectives.
91 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
94 - Remove option 'model-check/record': Paths are recorded in any cases now.
95 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
96 always sparse. That's an immense gain in memory, and can even be
97 faster because of cache effects. So there is no need to clutter the
98 code to allow the user to go for the unefficient mode.
101 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
102 models proved to be more accurate than these old experiments.
104 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
105 - FG#1: Broken link in error messages
106 - FG#2: missing installation documentation
107 - FG#3: missing documentation in smpirun
108 - FG#6: Python bindings not available on PyPI
109 - FG#7: simple cmake call requires doxygen
110 - FG#8: make python bindings an optional dependency
111 - FG#10: Can not use MSG_process_set_data from SMPI any more
112 - FG#11: Auto-restart actors forget their on_exit behavior
113 - FG#12: -Denable_lto=OFF doesn't disable LTO
114 - FG#13: Installs unstripped file 'bin/graphicator'
115 - FG#14: Installs the empty directory 'doc/simgrid/html'
116 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
117 - FG#17: Dead link in doc (pls_ns3)
118 - FG#20: 'tesh --help' should return 0
119 - FG#21: Documentation link on http://simgrid.org/ broken
120 - FG#22: Debian installation instruction are broken
121 - FG#26: Turning off a link should raise NetworkFailureException exceptions
122 - GH#133: Java: a process can run on a VM even if its host is off
123 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
124 - GH#326: Valgrind-detected error for join() when energy plugin is activated
126 ----------------------------------------------------------------------------
128 SimGrid (3.22) April 1. 2019
130 The Easter Christmas Release (this one is somewhat late).
133 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
134 - Not complete yet: asynchronous activities (amongst others) are still missing
135 - Still ongoing: the interface may change in the future. We need more
136 testers! Please report any glitches.
137 - No new project using Java should start now. Please switch to Python.
140 - Some of the internal cleanups may lead to speed improvements:
141 - The hard limitation on the amount of simulated actors with
142 Java+Mac was removed. Now, the available memory is the only limit.
143 - Our refcounting was tidyied, leading to 10% speedups in some cases.
144 - We are still working on making our code robust to the actor kills
145 and hosts' churn. Things are improving, but it's not perfect yet.
146 - Replaced our own code to display a backtrace (that was forking addr2line)
147 with the Boost.Stacktrace library.
148 You won't see your backtraces without this optional dependency.
149 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
150 - Stop setting random seed with srand() at initialization.
153 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
154 XML file version remains 4.2 since old files are still compatible.
157 - Process termination which was broken at version 3.21 has been repaired.
158 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
159 - Hide the examples into examples/deprecated. New users should use Python.
162 - Drop MSG_process_create_from_stdfunc() from the API.
163 This C++ function was a pimple in the C API, made necessary at some
164 point by the Java bindings. This is fixed now.
165 - Hide the examples into examples/deprecated. New users should use S4U.
166 - MSG_process_create and MSG_process_attach now crash if the host used
167 to run this process is off.
168 - Fix the protype of MSG_process_on_exit()
169 Now use (int,void*) callbacks instead of (void*,void*) ones.
170 The implementation was ways too messy to actually work, I guess.
173 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
174 too often and provide more accurate timings.
175 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
176 - MPI_Alltoallw support
177 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
178 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
179 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
180 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
182 - Basic implementation of generalized requests (SMPI doesn't
183 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
186 - Drop sg_cmdline. Please use xbt_cmdline instead.
187 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
188 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
189 OsSemaphore is implemented in a portable way with C++11 threads.
190 This should allow much more threads to be created at the same time,
191 allowing Mac and Java users to simulate many more actors.
192 - Implement the 'thread' factory with std::thread instead of xbt ones.
193 It is not possible to set the stack size with threads anymore, but
194 -fsplit-stack is the way to go nowadays when using threads.
195 - Drop the xbt_os_thread_t module (now unused)
196 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
197 - Drop xbt_str_join_array().
198 - Drop cunit, use Catch2 instead.
201 - Many cleanups in the kernel::activity namespace. This was long
202 overdue, and shall open the path to many future endeavors.
205 - #132: Java: a process can not shut down its own host
206 - #220: S4U: detached send causes exception if sender terminates soon after sending
207 - #261: Document the parameters of parallel execution's constructor
208 - #300: [s4u] BarrierPtr is missing
209 - #314: SMPI args internal cleanup
210 - #316: Fix a bug related to the CPU utilization of multi-core VM
211 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
212 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
213 - #325: Turning off a host has different behavior on sleeping actors and computing actors
215 ----------------------------------------------------------------------------
217 SimGrid (3.21) October 3. 2018
219 The Restarting Documentation (TRD) Release.
222 - Convert the user manual to Sphinx for improved usability.
223 Unstable version now hosted on https://simgrid.frama.io/simgrid
224 Many glitches remain; stay tuned.
225 - Restrict the Doxygen reference API to the installed header files.
226 (documenting all internals this way is overkill + doxygen scales badly)
227 - New tutorial on S4U (highly inspired from the old MSG one, but with
228 a git repository to fork as a starting point, and a docker image)
229 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
230 - The Developper manual is still to be converted (not compiled until then)
231 - Some parts are still missing in this conversion (such as the
232 platform and deployment sections) while others were blindly converted
233 and would need more love (such as the configuration flags).
234 Things will certainly further improve in the future releases.
237 - Fully reimplement auto_restart mechanism that was utterly broken
238 (fix #22, #131 and #281 that were all very related)
239 - Implement semaphores (fix #303 and #291)
240 - When creating an actor from a function and its parameters,
241 move-only parameters are not allowed anymore, as it would prevent
242 the actor to be restartable if its parameters are consumed on run.
243 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
244 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
245 IO operations that can be started, waited for, or canceled as a regular
247 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
251 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
252 - New host variable: 'core_count'
255 - Remove xbt_os_thread_specific features
256 - Remove portability wrapper to condition variables
257 - Remove xbt_os_thread_yield()
260 - MPICH collective selector now mimics MPICH 3.3b
261 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
262 - MPI_Init(NULL,NULL) is now allowed
263 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
264 - Command-line arguments should now be handled properly in Fortran simulations
265 - MPI Attributes and related callbacks should work in Fortran
266 - Apps using C + Fortran should now work
267 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
268 - Support MPI_Error_String
271 - Due to an internal bug, Msg.run() must now be your last line.
272 We hope to fix it in a future release, and we are sorry for the inconvenience.
275 - #22: Process autorestart seem to only work with CAS01 cpus
276 - #56: Feature request: dynamic SMPI replay
277 - #93: simgrid should not eat --help
278 - #111: How to change the loopback link characteristics?
279 - #116: Platform cluster radical
280 - #117: Platform cluster/host order
281 - #129: Replay_multiple on steroid
282 - #131: Java : autostart does not work
283 - #143: Setting a breakpoint at a given time
284 - #235: Network energy models should be integrated
285 - #264: Add ptask L07 resource tracing
286 - #271: Dynamic and manual replay of SMPI traces
287 - #279: Breakpoints option uses time deltas instead of absolute time
288 - #281: Daemonized actor no longer daemonized after an auto-restart
289 - #282: TIT convertor erase traces when using absolute path in the trace list file
290 - #285: segfault when a s4u actor kills itself with msg process tracing activated
291 - #286: Pajé traces are not exposing the number of cores
292 - #287: Command-line parsing should stop when encountering '--'
293 - #288: MPI_Init(NULL, NULL)
294 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
295 - #290: Method to check if a mailbox has a message ready to be consumed
296 - #291: [s4u] semaphores do not seem available yet
297 - #294: Alltoall collective implementation of mpich
298 - #295: Starting a Host as OFF is not supported
299 - #296: DTD too permissive
300 - #299: Add a s4u barrier example
302 ----------------------------------------------------------------------------
304 SimGrid (3.20) Released June 24. 2018
306 The proxy snake_case() release.
309 - The whole API is now using snake_case() instead of mixing it with camelCase.
310 Compatibility wrappers in place for one year.
311 - We took the opportunity to increase the consistency of the API.
312 Please report any remaining glitches.
313 - New function simgrid::s4u::Host::get_actor_count:
314 Returns the number of actors running on a specific host.
317 - SMPI is now tested with ~45 proxy apps from various sources, with none or
318 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
319 - Replay: The replay file has been re-written in C++.
320 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
321 supported. They are stored in the trace and used when replayed.
322 - Basic support of MPI_Cancel. Robustness not guaranteed.
323 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
324 (only for MPI_COMM_TYPE_SHARED).
325 - New option: smpi/privatize-libs, to add external shared libs to be privatized
326 by SMPI. They will be copied locally and loaded separately by each process.
327 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
328 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
329 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
330 with internal index optind with multiple processes. Only works if MPI_Init has
332 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
333 syntax without preprocessing tricks.
336 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
337 behavior tracing is no longer limited to MSG
338 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
339 this means that traces now need to be updated or re-obtained; both Irecv
340 and Isend lines in your traces have now in total 5 mandatory fields:
341 <rankid> <command> <to/from rankid> <tag> <size>
342 To update your traces, it suffices to add a 0 for the tag here.
343 - TIT now also supports waiting for a distinct request via MPI_Wait.
344 Wait/Test now wait for a specific request, not just the last one that was
345 issued. This unfortunately means another update, because we need to
346 identify which request you want to wait for. We do this via the
347 triplet (sender, receiver, tag), which needs to be added:
348 <rankid> <command> <sender> <receiver> <tag>
349 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
350 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
353 - The deprecation of MSG is ongoing (but this should not impact you).
354 Many MSG functions are now simple wrappers on the C API of S4U. If
355 you wish to convert your code to S4U, find the S4U counterparts of
356 your MSG calls in src/msg/msg_legacy.cpp.
357 - MSG can still be used, but won't evolve anymore.
360 - Allow to run the Link energy plugin from the command line with
361 --cfg=plugin:link_energy
362 - Rename Energy plugin into host_energy
363 - Rename Load plugin into host_load
366 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
367 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
368 - Improve performance when handling timeouts of simix synchros.
371 - Config: the C API is now deprecated (will be removed in 3.23), and
372 the C++ API has been slightly improved.
375 - Fix several build issues on OSX.
376 - Move simgrid_config.h to simgrid/config.h (old header still working)
379 - #143: Setting a breakpoint at a given time
380 - #150: Inconsistent event names in SMPI replay
381 - #258: daemonized actors hang after all non-daemonized actors have completed
382 - #267: Linker error on unit_tmgr
383 - #269: SMPI: tracing of MPI_Wait/all/any broken
384 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
386 ----------------------------------------------------------------------------
388 SimGrid (3.19.1) Released March 22 2018
390 The Fixed ABI Release.
392 As you may know, we are currently refactoring SimGrid in deep.
393 Upcoming SimGrid4 will be really different from SimGrid3: modular,
394 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
396 Our goal is to smooth this transition, with backward compatibility
397 and automatic update paths, while still progressing toward SimGrid4.
399 SimGrid remains open during works: The last pure SimGrid3 release was
400 v3.12 while all subsequent versions are usable alpha versions of
401 SimGrid4: Existing interfaces remain unchanged, but the new S4U
402 interface is budding and the internals are deeply reorganized.
404 Since 2015, we work hard to reduce the changes to public APIs. When
405 we need to rename a public library symbol in S4U, we let your
406 compiler issue an explicative warning when you use the deprecated
407 function. These messages remain for four releases, i.e. for one full
408 year, before turning into an error. Starting with v3.15, your can
409 also adapt to API changes with the SIMGRID_VERSION macro, that is
410 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
412 Starting with this v3.19.1, our commitment to reduce the changes to
413 the public interfaces is extended from the API to the ABI: a program
414 using only MSG or SimDag and compiled against a given version of
415 simgrid can probably be used with a later version of SimGrid without
416 recompilation. We will do our best... but don't expect too much of
417 it, that's a really difficult goal during such profund refactoring.
419 The difference between v3.19 and v3.19.1 is that the former was
420 accidentally breaking the ABI of MSG, while the later is restoring
423 S4U and kernel APIs will still evolve until SimGrid4, with one-year
424 deprecation warnings as currently. In fact, cleaning up these
425 interfaces and converting them to snake_case() is one release goal of
426 v3.20. But don't worry, we are working to smooth this upgrade path.
428 In summary, new projects should start with S4U to benefit of the
429 future, but old MSG projects should still be usable with no change.
431 ----------------------------------------------------------------------------
433 SimGrid (3.19) Released March 20 2018
435 The Moscovitly-cold Spring release.
437 SMPI user-visible improvements:
438 - Support of MPI_Type_create_subarray
439 - Unimplemented MPI calls now abort (instead of issuing a warning)
440 Some harmless calls around errhandlers still only issue a warning.
441 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
442 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
443 is now possible (mmap privatization was refactored to this end).
444 Adding new SMPI instances should also be technically doable now.
446 SMPI internal cleanups:
447 - Removed index notion from SMPI -- no more "getPid() - 1"!
448 In tracing files, actors are called rank-<PROCESS-ID> but these are
449 NOT the real ranks (tracing several communicators is not supported yet).
450 - Deprecate smpi_process_index() (will be removed in 3.22)
451 - Better testing on classical proxy apps, through an external project:
452 https://github.com/simgrid/SMPI-proxy-apps/
455 - Use a graphical TOC to make it easier to find the documentation you need
456 - Revamp the MSG tutorial
459 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
460 - Remove parameter of MSG_process_killall().
461 Resetting the PID was bogus anyway (several actors could have the same PID).
464 - Execution->setHost() can be called after start(), to migrate it.
465 - Comm::test_any() is now implemented.
466 - s4u::Actor now has onCreation() and onDestruction() signals.
467 - Install some kernel header files for the users' plugins and more.
468 Warning, their API is really not stable yet.
469 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
470 include/simgrid/kernel/routing/*Zone.hpp
473 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
475 Build System and other cleanups:
476 - Remove unused run-time parameter "tracing/onelink-only".
477 - SimGrid now works with intel compilers (MC still buggy)
478 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
479 - tesh kills subprocesses on timeout (but not on windows)
482 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
485 - #194: Feature request: simgrid::s4u::Comm::test_any()
486 - #245: migrating an actor does not migrate its execution
487 - #253: Feature Request: expose clusters as objects
488 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
489 - #255: Tesh broken on Windows
490 - #256: Modernize FindSimGrid.cmake
491 - #257: Fix (ab)use of CMake install
493 ----------------------------------------------------------------------------
495 SimGrid (3.18) Released December 24 2017
497 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
499 MSG IS NOW DEPRECATED. Please use S4U for new projects.
500 - Support for MSG should not be removed from SimGrid before 2020,
501 but future evolutions will be limited to the integration of
502 user-provided patches.
505 - s4u::onDeadlock() signal, run before stopping the simulation.
506 - s4u::Exec: asynchronous executions (abstraction of background
508 - s4u::Host->getLoad() returns the achieved speed in flops/s
511 - Link::name() is deprecated and will be removed in v3.21, use
512 Link::getCname() instead.
513 - Mailbox::getName() changed to return a std::string, use
514 Mailbox::getCname() instead to get a char*.
515 - Storage::getName() changed to return a std::string, use
516 Storage::getCname() instead to get a char*.
517 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
518 to have both allocation and deallocation of the map in user space.
521 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
522 The amount of flops remaining to do is a vector, not a scalar.
523 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
524 name implies on both sequential and parallel tasks.
525 - Both changes fix GitHub's #223 using PR #237 as a basis.
526 Thanks Michael Mercier.
527 - Most examples were converted to S4U and hidden elsewhere at tests.
528 Rationale: we still want MSG to work; we want newcomers to use S4U.
531 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
532 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
535 - Switch to the faster dlopen privatization mechanism by default
536 - Documentation improvments
539 - Live migration is getting moved to a plugin. Dirty page tracking is
540 the first part of this plugin. This imply that VM migration is now
541 only possible if one this function is called:
542 - C/MSG: MSG_vm_live_migration_plugin_init()
543 - C/C++: sg_vm_live_migration_plugin_init()
544 - Java: Msg.liveMigrationInit()
545 For C and C++, "simgrid/plugins/live_migration.h" also has to be
549 - Define class simgrid::xbt::Path to manage file names.
550 - Removed unused functions:
551 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
552 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
553 xbt_graph_export_graphviz()
554 - xbt/str.h: xbt_str_join()
555 - Remove unused datatypes:
556 - xbt/heap.h: use std::priority_queue or boost::heap instead
557 - xbt/swag.h: use boost::intrusive::list instead
560 - New link_energy plugin for the consumption of the links.
561 - All of the operations on files and storage contents have been
562 packaged into a plugin (src/plugins/file_system). The current
563 public interface can be found in
564 include/simgrid/plugins/file_system.h
565 To use these functions you now have to initialize the plugin by
566 calling MSG_storage_file_system_init() just after calling
567 MSG_init() or sg_storage_file_system_init() just after creating
571 - Remove the undocumented/untested tag <include>
574 - Remove viva specific tracing as the tool is no longer maintained
577 - #248: Kill of finished processes leads to segfault
578 - #240: xbt_cond_wait_timeout should gracefully return for C apps
579 - #239: please implement signal s4u::onDeadlock()
580 - #230: segfaults when exit() without run()
581 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
582 - #223: MSG_task_get_flops_amount() not working with parallel tasks
583 - #222: Actor::kill() doesn't really kill and segfaults
584 - #221: odd LMM warning when killing an actor
585 - #120: Memory leak when the processes are forcefully killed
587 ----------------------------------------------------------------------------
589 SimGrid (3.17) Released October 8 2017
591 The Drained Leaks release: (almost) no known leaks despite the tests.
593 Even very long simulations will preserve your memory: our testsuite
594 stresses SimGrid for over 45mn on fast machines for a coverage of
595 over 80%, and there is only one single known leak, of about 4kb.
599 - Comm.detach(): start and forget about asynchronous emission. A cleanup
600 handler may be given to free resources if the comm cannot be completed.
601 - this_actor::send(mailbox) is now mailbox->put()
602 - New: simgrid::s4u::Comm::wait_all()
603 - New: Host.execute() for remote executions.
606 - Unused option network/sender-gap is removed.
609 - Deprecate MSG_task_isend_with_matching(): This unused feature
610 really complicates our internals. Will be removed in v3.20.
613 - Improved context termination. It is now possible to free resources, even
614 when a process is forcibly killed.
615 - ContextBoost: add support for Boost versions above 1.61.
618 - Bring back run-time option --cfg=exception/cutpath to remove exception
620 - Removed unused functions:
621 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
622 xbt_str_rtrim(), xbt_str_trim().
623 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
626 - Removed header files obsolete since SimGrid 3.12:
627 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
629 - #3: SD_exit should be made optional
630 - #120: Memory leak when the processes are forcefully killed
631 - #159: Threading test regression in Actor refcounting
632 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
633 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
634 - #186: Actor::killAll() segfaults if some process is blocked on wait()
635 - #191: VM migration and pstate
636 - #192: Updating the energy consumptions of all hosts crashes with VMs
637 - #195: All actors have PID=0 in the logs
638 - #204: Sometimes segfault with thread contexts and mmap privatization
639 - #222: Actor::kill() doesn't really kill and segfaults
640 - #225: Actor::kill() doesn't really kill when victims are doing a join()
642 ----------------------------------------------------------------------------
644 SimGrid (3.16) Released June 22. 2017.
646 The Blooming Spring Release: developments are budding.
649 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
650 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
651 - Java JAR file should be finally fully working on Mac OSX too.
653 XML platforms: Switch to platform v4.1 format.
654 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
655 - <zone> can be used as a synonym for the now deprecated <as>
656 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
657 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
658 - <actor> can be used as a synonym for the now deprecated <process>
659 - state_file and avail_file periodicity is now easier to express
660 (check the documentation) the old behavior should still work.
661 * Storage: not backward compatible, but it seems that nobody ever used it
662 - Remove attribute 'content_type' of <storage_type>: was never used
663 - Make attribute 'model' of <storage_type> optional: for future usage
664 - Remove Bconnection model property: was never (in)validated. Replaced by
665 the maximum of the read and write bandwidth as a resource constraint.
668 - New and Backwards Compatibility break:
669 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
670 When one wants to get the list of tasks whose states have changed during a
671 simulation round, s/he has to allocate and free a dynar and use it as
672 argument to this function. The former SD_simulate (double how_long)
676 - Allow multicore VMs, with the correct sharing (unless you overcommit)
677 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
680 - New function to update the consumption of all hosts at once.
681 - Fix the model for multi-core hosts, linear in the amount of busy
682 cores with an abnormality for pIdle (see paper).
683 BUG: Still not satisfactory for ptask on mono-cores.
686 - The netzone are now available from the MSG API.
687 The old names still work, but are now deprecated.
690 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
691 - New option: smpi/keep-temps to not cleanup temp files
692 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
693 mallocs mode is used, allows to change the size of the fake file used
694 (default 1MB), to potentially limit the number of mappings for large runs.
695 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
696 - Fortran ifort and flang compilers support
697 - New RMA calls supported (experimental):
698 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
699 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
700 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
701 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
702 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
703 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
706 - New: this_actor::isMaestro() returns whether we are in kernel mode.
707 - Behind the scene, ActivityImpl are now automatically refcounted.
708 Removing this nasty bottleneck will greatly simplify our internals.
711 - Replay: New function xbt_replay_action_get():
712 Retrieve the function previously associated to an event type.
713 - DROPPED FUNCTION: xbt_str_varsubst()
714 - DROPPED MODULE: strbuff. We don't need it anymore.
715 - DROPPED MODULE: matrix. We don't need it anymore.
716 - DROPPED MODULE: lib. We don't need it anymore.
718 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
720 SimGrid (3.15) Released March 22. 2017
722 The Spring Release: continuous integration servers become green
723 We fixed even the transient bugs on all target architectures:
724 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
725 Windows, FreeBSD, NetBSD.
727 - Rename NetCards to NetPoints.
728 This was intended to help NS3 users, but that's not a netcard.
729 That's a point in the routing algorithm, let's avoid wrong simplifications.
732 - New: MSG_process_yield(). Stop and yield to other processes.
733 - New: MSG_process_daemon(). Daemon processes are automatically killed
734 when the last non-daemon process terminates
735 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
736 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
737 to make room for the upcoming network energy plugin.
738 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
741 - Ensure that an actor can kill itself with Process::exit()
742 - Kill the obscure NativeException. Nobody want to survive the issues
743 it denotes, so use JniException that is a RuntimeException (not to
744 be caught explicitly).
745 - Partial bug fix in initialization. SimGrid flags on command line were
746 consumed at C level but stayed in the original Java String[] args.
747 This could mess users' args[i] if SG flags were not put at the end of
749 The SimGrid flags are now removed from the Java arguments. However,
750 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
751 if args.length is greater than the number of YOUR OWN ARGUMENTS.
752 It might be if you have --log or --cfg flags in the command line.
753 - Fix numerous memleaks all around the place. In particular, around VMs.
757 - s4u::onPlatformCreated: right before the simulation starts
758 - s4u::onSimulationEnd: right after the main simulation loop
759 - s4u::onTimeAdvance: right after a clock change (time is discrete)
760 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
761 event from the availability_file changes the avail speed.
762 - Links are now usable from s4u
763 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
764 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
765 - New examples: The conversion of MSG examples to S4U has begun
766 - Actors: create, kill, migration, and suspend
767 - Applications: master-worker and token-ring
768 - Action replay: communications and storage
769 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
772 - Backwards Compatibility breaks
773 - The SD_route_get_list and SD_route_get_size functions have been removed.
774 They are replaced by the unique
775 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
776 The route from src to dst is built in the links dynar whose size can
777 be obtained with xbt_dynar_length.
778 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
779 sg_host_route_bandwidth and sg_host_route_latency.
780 Macros ensure the backwards compatibility, but you should fix your code
783 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
784 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
785 - Bump our claim of support from MPI 1.1 to MPI 2.2.
786 We don't support 100% of it, but it should be enough. Contact us if not.
787 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
788 naive and probably inaccurate)
789 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
790 mapped onto a unique small file using some system magic.
791 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
792 - Fortran: cleanups, fixes, support of user-added operations
793 - RMA: MPI_Accumulate are applied in correct order.
794 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
795 - Better support for MPI_IN_PLACE
796 - Support for MPI_Win attrs and keyvals.
797 - Support MPI_Comm_group_create, MPI_Type_size_x
798 - MPI ops should be properly applied to all allowed types
799 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
800 algorithm by a non failing one from mpich.
803 - Kill the fifo data container: we don't use it anymore.
806 - A <cluster> can now be created with different speed values to represent pstates.
807 see examples/platforms/energy_cluster.xml for an example.
809 SimGrid (3.14.159) Released December 28. 2016
811 The previous release was only a crude approximation of a Pi release;
812 this one should be a bit better.
814 - Revert a last minute change that broke on Mac OSX.
815 - Fix the changelog and NEWS files.
816 - Improve the documentation, in particular of the routing module.
817 - Integrate some patches flying around in Debian.
819 SimGrid (3.14) Released December 24. 2016
822 * The whole documentation was reworked and reorganized. There is
823 still much room for improvement here, but we're on it.
826 * We now need python3 for our testing infrastructure.
827 * Model checking needs libevent
830 * Now works on FreeBSD
833 * Privatization now works on FreeBSD
835 * Privatization is now activated by default in smpirun
836 You can turn it off with -no-privatize if something goes wrong.
838 * Call-location tracing for SMPI.
839 You can add the exact location (filename / linenumber) of an MPI call to
840 your trace files and slow down or speed up the simulation between two
841 consecutive calls by using an adjustment file (see the documentation).
843 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
844 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
845 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
846 values were ignored in some cases. The timings of these functions can now
847 be significantly different.
849 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
850 smpi/running-power is renamed to smpi/host-speed
852 * smpi/grow-injected-times option to enable or disable multiplication of the
853 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
854 make simulation less precise (but also much faster).
856 * smpirun script should be (much) faster for large deployments.
858 * SMPI tracing: fixed issue with poor matching of send/receives.
860 * Replay: Fix broken waitall
862 New functions and features
863 * MSG_parallel_task_execute_with_timeout, to timeout computations.
865 Dropped / renamed functions and features
866 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
867 - MSG_mailbox_is_empty() -> Mailbox::empty()
868 - MSG_mailbox_front() -> Mailbox::front()
869 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
870 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
871 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
872 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
873 * Don't pass the free_f parameter to property related functions:
874 - MSG_host_set_property_value()
875 - MSG_as_router_set_property_value()
876 - MSG_storage_set_property_value()
877 * VM properties. Since msg_vm_t are msg_host_t, just use
878 MSG_host_get_property_value() and friends
879 * VM I/O related things:
880 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
881 - Unimplemented save/restore methods
882 * MSG_as_router_get_property_value() was redundant with
883 MSG_environment_as_get_property_value().
884 - Removed MSG_as_router_*propert*() functions
885 - Added MSG_environment_as_set_property_value() for consistency
886 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
887 Well, they are still there for now, but deprecated with a warning.
888 Please switch to xbt_dict_new_homogeneous() before this is removed
890 * Task affinity. Its intended behavior (that was very badly tested
891 and probably not really working) was deceiving what most users
892 would have hoped here.
893 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
894 * Option network/coordinates is now useless and should be dropped.
897 * Added option storage/max_file_descriptors to allow more than 1024 files opened
900 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
901 * Added Dragonfly topology support
904 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
905 EiBps, EBps, Eibps, Ebps for bandwidth.
906 They may become useful to some lucky ones.
909 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
910 * ex/app_tokenring: new example, very similar to the MSG Token Ring
911 * ex/async_waitAll: new example, on asynchronous communications
914 * Memory usage should be decreased for simulations with a large number
915 of processes. This also helps for SMPI.
917 SimGrid (3.13) stable; urgency=low
919 The Half Release, a.k.a. the Zealous Easter Trim.
921 - We removed half of the lines, that were mostly experimental cruft.
922 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
923 (not counting blanks and comments -- according to openhub.net)
924 - The internals are now compiled in C++ (and will soon be clean C++)
925 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
926 59 klines of C, etc. We added only 29 klines of C++ in replacement.
928 * Backwards Compatibility breaks
929 - Removed Lua simulation bindings (switch to C or Java for that).
930 Lua can still be used to describe platforms
931 - Removed Java kernel plug-ins.
932 Will be reintroduced after the ongoing major internals reorg.
934 - the following functions were removed.
935 They were too specific and should be reimplemented in a generic
936 way, with filter function.
937 - MSG_task_listen_from_host
938 - MSG_mailbox_get_count_host_waiting_tasks
939 - MSG_mailbox_put_with_timeout was removed.
940 Please use MSG_task_send_with_timeout instead.
942 - the SD_application_reinit function was removed. It has been a noop for a while.
943 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
944 replaced by a more flexible mechanism.
945 - The following functions thus do not exist anymore
946 - SD_workstation_get_access_mode
947 - SD_workstation_set_access_mode
948 - SD_workstation_get_current_task
949 - Basic estimation functions have been removed but can easily be replaced
950 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
951 - SD_workstation_get_computation_time => amount / sg_host_speed()
953 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
954 Use VM.getSpeed()*load/100 for the legacy behavior.
956 - option enable_tracing was removed. It was not doing anything for a while.
957 - In the ModelChecker:
958 - the model-checker now ptraces the model-checked process which means
959 you cannot use a debugger on the latter anymore (we might make this
960 optional in the feature);
961 - removed soft-dirty page tracking;
962 - remove model-checked side snapshot management,
963 MC_snapshot() and MC_compare_snapshot();
964 - keep the MC_cut() function as a stub (it was not really working
965 in the previous release).
968 * All options are consistently in kebab-case. Old names are kept as alias.
971 * Switch to platform v4 format.
972 - Rename from 'power' to 'speed' the attributes describing the amount of
973 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
974 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
975 - In <host> and <link>, attributes availability and state are gone.
976 It was redundant with state and availability traces, and with peak values.
977 - In <cluster>, attributes availability_file and state_file are gone.
978 It was too complex and unused.
979 - Kill <gpu>. Was not doing anything.
980 - The DOCTYPE points to the right URL:
981 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
982 (the file at this address now documents the changelog since its v1)
983 - A warning is emitted for unit-less values (they are still accepted).
984 - speed. Default: 'f' or 'flops'. Also defined:
985 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
986 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
987 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
988 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
989 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
990 - latency. Default: 's' second. Also defined:
991 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
993 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
994 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
997 * s4u::Host is now the preferred public interface to the Host features.
998 sg_host_* functions are C bindings to the exact same behavior
999 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1002 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1003 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1004 They were deprecated since a few years, and probably did not even compile anymore.
1007 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1008 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1009 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1010 encouraged to update their codes. A list of the modified functions can be found at the end of
1011 include/simgrid/simdag.h
1014 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1015 the lambda or closure passed as a parameter will run in kernel mode.
1016 All the callback functions should be rewritten to that interface at some point.
1019 * Reorganizing and cleaning the internals all around the place.
1022 * Remove old default barrier/bcast buggy algorithms (see #18407)
1023 * Various bug fixes to handle more codes
1024 * Remove the need for the --foreground option of smpirun (it is still
1025 accepted for backward compatibility).
1028 * Kill the setset data container: MC don't use it anymore.
1029 * Kill the queue data container: it made more sense with GRAS.
1030 * Kill the xbt_peer_t data type: it's useless without GRAS.
1031 * Kill rm_cb feature of config sets: it was never useful.
1032 * Kill graphxml parsing feature. It was not used.
1033 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1035 - xbt_dynar_sort_strings(), when the content is char*
1036 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1037 They throw exceptions on invalid input;
1038 * C++ support for declaring CLI flags (simgrid::config::Flag);
1039 * class for abstracting different signal backends (simgrid::xbt::signal).
1040 with no external dependencies (we need very simple signals).
1043 * refactoring and cleanup of the code;
1044 * ongoing process to cleanly separate the model-checking algorithms
1045 from the code model-checking support.
1047 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1049 SimGrid (3.12) stable; urgency=low
1051 The Facelift Release.
1054 * Require g++ v4.7 at least to not speak prehistorical C++.
1055 * Require Boost 1.48 (for signal2 component).
1056 * Java must be version 7 at least when activated.
1057 * Builds on Windows again (including Java bindings).
1058 * Tracing is now always enabled (no way to turn it out)
1059 * Remove GTNetS. It was not working anyway.
1060 * Various cleanups in the cmake scripts.
1061 * Move headers around to sort them out on installed systems:
1062 - instr/instr.h -> simgrid/instr.h
1063 - instr/jedule/* -> simgrid/jedule
1064 - simdag/datatypes.h was removed
1065 - simdag/simdag.h -> simgrid/simdag.h
1066 - msg/datatypes.h was removed
1067 - msg/msg.h -> simgrid/msg.h
1070 * Interface improvement:
1071 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1072 - Sanitize the interface in MSG_task_ module:
1073 - Merge two functions that were close enough but misleading:
1074 set_compute_duration(t) -> set_flops_amount(t)
1075 get_remaining_computation(t) -> get_flops_amount(t)
1076 - set_data_size(t) -> set_bytes_amount(t)
1077 get_data_size(t) -> get_bytes_amount(t)
1078 - Massive cleanups in the functions related to the energy
1079 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1080 - New: MSG_host_get_pstate()
1082 - msg/energy/onoff: switching hosts on and off
1084 * Interface improvement:
1085 - Rename Host.isAvail() to Host.isOn()
1086 - Rename Process.currentProcess() to Process.getCurrentProcess()
1087 - Rename Task.setDataSize() to Task.setBytesAmount()
1088 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1090 - #18874: Actually allows the GC to reclaim tasks
1094 - SIMIX_process_throw: raises an exception in a remote process
1095 * Refactoring: Separate sub-modules
1096 - libsmx: the public interface, as libc in a real system
1097 - popping: the strange dance that converts a user request into a kernel handling
1098 - smx_context_*: the virtualization mechanisms that embed the user code
1099 - smx_*: the handling of each simcalls
1100 * Interface cleanup:
1101 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1102 * Rename smx_action_t into smx_synchro_t, making explicit that these
1103 things are used to synchronize processes with their environment.
1104 For example, a communication is a sort of synchronization involving
1105 the communicating processes (that may block until the exchange) and
1106 the platform. The same can be said from computations, etc.
1108 - #18888: segfault when a process ends before its kill_time
1112 - 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.
1113 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1114 - MPI_Type_set_name, MPI_Type_get_name
1115 - MPI_*_c2f and MPI_*_f2c functions
1116 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1117 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1118 - Activate a lot of new tests from the mpich 3 testsuite
1120 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1121 - InfiniBand network model added: Based on the works of Jerome Vienne
1122 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1123 - When smpi/display_timing is set, also display global simulation time and application times
1124 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1125 * Collective communications
1126 - 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
1127 - 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/).
1128 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1130 - Replay now uses algorithms from wanted collective selector
1131 - Replay can be used with SMP-aware algorithms
1132 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1133 - 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.
1135 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1136 - Correctly use loopback on fat-tree clusters
1137 - Asynchronous small messages shouldn't trigger deadlocks anymore
1138 * Energy/DVFS cleanup and improvement
1139 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1140 - new: smpi_get_host_pstate()
1143 - "Full" network optimization flag was broken since Surf++
1144 - Better handling of precision flags in maxmin
1145 - Fix bug causing sometimes "Impossible" errors
1146 - Properly pass cluster properties to included hosts
1147 * Improvement of the Energy plugin.
1148 - Always update the consumption before returning that value
1149 - New property: watt_off to denote the dissipation when the host is off
1150 - New functions getWattMinAt and getWattMaxAt to retrieve the
1151 dissipation of pstates that we are not currently at.
1152 * Java: class NetworkLink renamed to Link
1153 * New function: simcall_process_get_kill_time()
1154 * Massive rename s/workstation/host/
1155 - That's intrusive, but that's good for the project consistency. Sorry.
1156 - Change config option "workstation/model" into "host/model"
1159 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1160 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1161 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1162 - xbt_dirname and xbt_basename for non-POSIX systems
1164 * The model checker now runs as a separate process.
1165 * The model checker runs must now be launched with the new simgrid-mc program.
1166 * Record/Replay: the MC can display a textual representation of a path in the
1167 execution graph. It can then be replayed outside of the model checker.
1169 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1171 SimGrid (3.11) stable; urgency=low
1176 * Normalizing pointers addresses tool for better diff between logs
1179 * Add cloud examples using new VMs
1180 - examples/msg/cloud/two_tasks_vm.tesh
1181 - examples/msg/cloud/simple_vm.tesh
1182 - examples/java/cloud/cloud.tesh
1183 - examples/java/cloud/migration/migration.tesh
1184 * Add java surf examples:
1185 - examples/java/surfPlugin/surf_plugin.tesh
1186 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1187 - examples/java/surfCpuModel/surf_cpu_model.tesh
1188 * Add SMPI+MSG example:
1189 - examples/smpi/smpi_msg_masterslave/
1194 - msg task destroy cancel test
1195 - msg_host on/off test
1196 * Move all tests in testsuite to teshsuite (adding tesh files)
1197 * Restructure teshsuites
1198 - one folder for each kind of test
1199 * Restructure AddTests.cmake
1201 - structure the order of tests (with sections)
1204 * Add virtual machine
1205 - creation of a VM on a PM
1206 - migration of a VM from a PM to another PM
1208 - MSG_process_join(msg_process_t process, double timeout)
1209 - msg_bar_t MSG_barrier_init(unsigned int count)
1210 - int MSG_barrier_wait(msg_bar_t barrier)
1211 - void MSG_barrier_destroy(msg_bar_t barrier)
1212 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1213 * New option "msg/debug_multiple_use" to help debugging when a task is used
1216 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1218 - Increase code coverage in test suites
1222 * Protect context stacks against stack overflow. The number of protected memory
1223 pages allocated on the top of each stack (1 by default) can be configured
1224 with the new command line option --cfg=contexts/guard_size:<value>.
1225 * Simcalls are now generated by a python script that
1226 - generates files included by SimGrid
1227 - checks that all the functions exist, and proposes prototypes
1229 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1230 host_set_data, host_get_data
1232 - simcall_process_join(smx_process_t process, double timeout)
1233 * Fix bug where sleeping processing could not be suspended.
1236 * Translate surf models from C to C++
1237 - Generic classes for all models: Model, Resource, Action
1238 - A generic interface for each kind of model (CPU, Network, Storage
1239 Workstation, WorkstationVM)
1241 * Translate surf routings from C to C++
1242 * Add callbacks using sigc++ or boost::signals2
1243 - Add callback functions for resource creation/destruction
1244 - Add callback functions for action state change
1245 - Handle Energy as a plugin
1246 * Replace swag by boost::intrusive
1247 * Add new routing models for clusters. For documentation, see
1248 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1249 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1250 parameters for cluster tag
1251 - Fat trees, with topology="FAT_TREE" and
1252 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1253 - see examples/platforms/torus_cluster.xml and
1254 examples/platforms/fat_tree_cluster.xml
1255 * More documentation
1258 * Hostfiles support host:nb_processes construct to deploy several processes on
1260 * Collective communication algorithms should not crash if used with
1261 improper number of nodes and report the error.
1262 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1263 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1264 MPI_Dims_create, MPI_Cart_sub are supported.
1265 * New interface to use SMPI programmatically (still depends on MSG for
1266 some parts, see examples/smpi/smpi_msg_masterslave):
1267 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1271 * Global variables privatization in MPI executables is now performed at runtime
1272 with the option smpi/privatize_global_variables (default:no).
1273 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1274 dynamic libraries loaded by the application are not privatized (static
1275 linking with these libraries is advised in this case)
1278 * Options defined in XML work correctly now.
1281 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1282 copied into simgrid.jar or not (ON by default). Use this option if you want
1283 to reduce the size of the installed simgrid.jar, *and* the native libraries
1284 are kept installed elsewhere.
1285 * Surf binding with SWIG (code generated in maintainer mode only):
1286 - plugin to handle callbacks
1287 - CPU model only for the moment
1290 * Supernovae build mode is definitively removed. It was used to improve
1291 inlining and inter-module optimizations. It is nowadays superseded by
1292 link-time optimizations commonly available in compilers.
1293 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1294 * Add boost dependency for surf++
1295 * Add new macro for tests
1296 - ADD_TESH(name <tesh_args>)
1297 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1301 - xbt_bar_t XBT_barrier_init(unsigned int count)
1302 - int XBT_barrier_wait(xbt_bar_t barrier)
1303 - void XBT_barrier_destroy(xbt_bar_t barrier)
1304 * Make the xbt_os_time module public
1306 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1308 SimGrid (3.10) stable; urgency=low
1310 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1313 * Reintegrate Java to the main archive as desynchronizing these
1314 package is not acceptable anymore (Java is now considered stable)
1315 * Add explicit synchronization facilities through semaphores
1316 * Bug fix: Task.setDataSize() only changed the C world, not the value
1317 cached in the Java world
1320 * Dramatically change the way files are handled. API and internals changed, but
1321 this part of MSG was not considered as production grade either.
1322 * Add explicit synchronization facilities through semaphores
1323 * Add a new function MSG_host_get_process_list()
1324 * Preliminary DVFS support (see examples/msg/energy/ for details)
1327 * SMPI is now included directly in the libsimgrid as the windows
1328 linker doesn't force us on splitting it anymore.
1329 * Improvements of the SMPI replay tool:
1330 - Most of the collective communications are now rooted in the same process as
1331 in the original application.
1332 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1333 used until now). Multiple data types can now be used in a trace.
1334 - The replay tool now supports traces produce either by TAU or a modified
1336 - Bug Fix: the compute part of the reduce action is now taken into account.
1337 - Gatherv collective is now supported
1338 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1339 of smpirun outputs time independent traces for the current run. One file
1340 is created per process. If too many processes are simulated, this behavior
1341 can be changed to one file for all processes by using the
1342 tracing/smpi/format/ti_one_file flag
1343 * smpirun generates the host file if needed (with given host count and platform)
1344 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1345 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1346 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1347 --cfg=smpi/coll_selector:(mpich/ompi)
1348 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1350 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1351 happening inside a collective SMPI call.
1352 * Fix the behavior of complex data types handling.
1353 * Make MPI_Wtime another synchronization point to take computations into
1355 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1356 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1357 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1358 (no privatization of global variables yet)
1359 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1362 * Verification of liveness properties is now available for SMPI applications
1363 (in addition to MSG applications)
1364 * Bugged examples using SMPI in examples/smpi/mc/
1365 * Add --cfg=model-check/visited option. Allows the verification of infinite
1366 programs. Detection of loops in the execution thanks to the system state
1367 comparison and reduction of the state space to explore. Can be combined with
1368 DPOR for safety properties.
1371 * Allow to change SimGrid configuration (see --help) within the code
1372 thanks to SD_config() as it can be done in MSG.
1373 * Add a new function SD_task_set_amount() upon user request.
1376 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1377 * Remove rule based routing (no more PCRE dependency)
1378 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1379 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1380 * Add a loopback_bw and loopback_lat options to cluster tag.
1381 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1382 routing. Now peers are encapsulated in an AS and have their own private
1383 router but this is transparent.
1386 * Our own implementation of getline is renamed xbt_getline, and gets
1387 used even if the OS provide a getline(). This should reduce the
1388 configuration complexity by using the same code on all platforms.
1389 * New type: xbt_cfg_elm_boolean.
1390 * Allow to use yes/no for boolean configuration options in the command line.
1391 * Allow to disable SimGrid cleanups at exit from command line option.
1392 There are situations where one may want a simulation to end with an exit.
1393 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1394 annoying when scripting around the simulator. Adding a
1395 --cfg=clean_atexit:no allows to circumvent this issue.
1398 * Lots of memory leaks were corrected in this release.
1399 There is no known memory leaks anymore, in all of our 600+ tests.
1400 * New command line option --version, to get SimGrid version information.
1401 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1403 * Supernovae builds are deprecated, and expected to be removed in the next
1406 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1408 SimGrid (3.9) stable; urgency=low
1410 The Grasgory release: GRAS is really dead now.
1412 * Complete overhaul of the internal host structures scheme.
1415 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1416 was considered as experimental and badly maintained since then.
1417 * Keeping it was thus a trap to our potential users, that could take
1418 it instead of MSG or SMPI by mistake despite is pity state.
1419 * GRAS seems to have very few users (if any), and no one volunteered
1420 to maintain it further. It also induces a lot of XBT code (for
1421 portability sake), that must be maintained too.
1422 * For all these reasons, we killed GRAS. If someone wants to revive it
1423 in the future, don't cry, our git history still remembers of GRAS.
1426 * Major overhaul. Merge our documentation again as time proved that
1427 splitting it was really not helping our users.
1428 * Further improve the developer documentation to help newcomers
1429 hacking on SimGrid itself. The user documentation (and in
1430 particular, the beginner documentation) is still in a sorry state.
1433 * Now works on Windows too!
1434 * Much more extensive test suite, from MPICH
1437 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1438 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1439 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1440 * Bug fix: let task be scheduled when the last dependency to be solved is
1441 a control dependency.
1442 * Remove SD_load_environment_script function.
1443 Use the C sg_platf function if you want to declare a platform
1447 * New function: MSG_process_get_number()
1448 * Old function documented: MSG_config()
1449 * Remove MSG_load_platform_script function
1450 Use the C sg_platf function if you want to declare a platform
1454 * Change the default value of the TCP_gamma constant (maximal size of TCP
1455 congestion window) to a more realistic 4MiB value. If you notice changes in
1456 your simulation results, you can fall back to the previous 20k tiny window
1457 by adding --cfg=network/TCP_gamma:20000 on command line.
1458 * (Hopefully) fix a bug wrt periodic availability and state traces
1459 * Bug fix: use default values at start when first event in availability/state
1460 trace is not at time 0.
1463 * remove the "new_" part of function name sg_platf_new_trace_connect
1464 (resulting in sg_platf_trace_connect), since it does not create
1468 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1469 use case where it's really mandatory, and maintaining it was a pain in
1471 * New: xbt_fifo_search(), search an item with a user-provided
1472 comparison function instead of dumb pointer comparison.
1475 * Fix the lua deployment:
1476 Use `simgrid.init_application()` before deployment instead of
1477 `simgrid.msg_register_application()` after.
1480 * Transfer the tracing files into the corresponding modules.
1482 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1484 SimGrid (3.8.1) stable; urgency=low
1486 The "we are told that some people want to also *install* the simgrid
1489 * Add missing file "tesh.1" to the archive.
1491 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1493 SimGrid (3.8) stable; urgency=low
1495 The Psssshiiiit release: SimGrid jumps into the Cloud.
1498 * Add an experimental interface to manipulate VMs. They are mainly
1499 process groups with very few intrinsic semantic, but they should
1500 allow you to build the semantic you want easily.
1501 * New function: MSG_host_set_property_value()
1502 * New function: MSG_process_on_exit(). To clean memory in all cases.
1503 * Bug fixes that made the host (and link) failures unusable.
1504 * Add a way to auto-restart process when the host in which they are
1505 executing comes back (ON_FAILURE="RESTART" on deployment file,
1506 MSG_process_auto_restart_set).
1507 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1508 please stop using the old ones, they are DEPRECATED.
1510 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1511 Please use MSG_init() instead. (reducing the amount of entry
1512 points in the library helps us).
1513 * Make it impossible to link against the wrong version of the lib
1514 * Deprecate MSG_clean(). No need to call it anymore.
1515 * Function MSG_get_host_number() is not deprecated anymore.
1518 * Split the doc into a user guide and a reference guide.
1519 * Start a developper guide to help people hacking on SimGrid.
1522 * Enable tracing by default. This modules rocks you should use it.
1523 * Remove option custom_flags. Now use environment variables CFLAGS
1525 * Use default cmake things to detect lua instead of home grown ones.
1526 * New option "enable_mallocators" to disable mallocators, for debugging
1527 purpose ("on" by default).
1530 * Bug fixes around the resource failures: don't let the processes
1531 survive the host they are running onto.
1532 * Add an interface to auto-restart processes when the host in which they are
1533 executing comes back.
1534 * Ensures that SIMIX_clean is called automatically. It's not part of
1535 the public interface anymore (bindings should be updated).
1538 * Bug fix for when SD_Simulate is called with a positive value: be careful
1539 when comparing doubles. Sometimes they are different for non significant
1541 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1542 parallel task whose initial work is distributed among host according
1543 to the Amdahl's law. Such tasks are created with a parameter alpha
1544 that corresponds to the non-parallelizable part of the computation.
1545 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1546 between two sets of workstations assuming a 1D block distribution (each
1547 workstation owns a similar share of data) on both sides.
1549 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1550 Data redistribution will be automatically scheduled once parent and child
1551 are both scheduled. The filling of computation_amount and
1552 communication_amount structures is now done seamlessly thanks to the chosen
1554 * New function SD_workstation_dump to display various information
1555 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1556 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1557 bandwidth on the route onto which the task is scheduled and the amount of
1559 To divide the nominal bandwidth by 2, the rate then has to be:
1560 rate = bandwidth/(2*amount)
1561 * Compute tasks that have failed can now be rescheduled and executed again
1562 (from their beginning)
1563 * Increasing source code coverage (src/simdag is now covered at 95.8%
1567 * Re-implement time-independent trace replay using SMPI (at the
1568 smpi_smp_* level) instead of MSG. This should replace
1569 examples/msg/actions/actions.c
1570 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1571 hindexed and structs)
1572 * Implement the exchange of non-contiguous data.
1573 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1574 * Correct behavior of smpi/sender_gap and set its default value to 0
1575 * Add option to asynchronously send small messages to allow better
1576 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1577 specifies the size in bytes under which messages will be asynchronously sent.
1578 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1579 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1582 * New C interface to define a platform: XML is now optional.
1583 For more info, please check include/simgrid/platf.h.
1584 * New interface to define random platforms from the C:
1585 For more info, please check include/simgrid/platf_generator.h and
1586 examples/msg/masterslave/masterslave_platfgen.c
1587 * Export a sg_cmdline dynar containing all the arguments we got from
1591 * Two new tracing options for adding comments to trace file so you
1592 can track your experiments (see --help-tracing for details).
1593 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1594 * Adding the SimGrid version that generated the trace file as a comment.
1595 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1596 * Fix to avoid key clashes on Paje links
1597 * Other minor fixes related to the Paje specification
1600 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1601 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1602 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1603 * Greatly improve the robustness of mmalloc to user errors (such as
1604 using an area after freeing it, or freeing it twice)
1606 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1608 SimGrid-java (3.8.1) stable; urgency=low
1610 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1611 Warning: all methods are visible, but only some of them are
1612 implemented so far. Check the source (src/jtrace.c)
1613 for further information.
1614 * New module: org.simgrid.msg.File (SimGrid File management functions)
1615 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1616 * Change the meaning of Process.restart: now restart the process from
1617 the begining, like MSG_process_restart in C.
1618 * Add Process.setAutoRestart: handling of process restart when failed
1620 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1621 you to retrieve the properties of the processes/hosts
1622 * Deprecate Msg.clean(): you can just forget about it now.
1623 * New function Process.getCount(), that only works when compiling
1624 with the not yet released version 3.9 of the C library.
1626 * New context factory based on Coroutines. It mandates a modified JVM
1627 but then, the simulations run about five times faster, and there is
1628 no limit to the amount of processes (beside of the available memory).
1630 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1632 SimGrid (3.7.1) stable; urgency=low
1635 * Restore the prototype of MSG_process_create_with_environment() to
1636 the pre-3.7 situation by removing the kill_time argument.
1637 * Add a MSG_process_set_kill_time() function instead.
1640 * Fix weird behaviors when dealing with parallel tasks.
1643 * Simgrid is now built as a dll.
1644 * Simgrid-java now works on Windows.
1645 * Simgrid-Java is now included into Windows package.
1648 * First pre-build package for MacOSX.
1651 * Fix compilation when using MSG_USE_DEPRECATED.
1652 * Fix some compilation issues on Macs and Windows.
1653 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1654 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1656 We discovered that the Lua console is broken, but we are missing the
1657 manpower to fix it right now. The problem existed in 3.7 too, so we
1658 are not blocking the release for that. Sorry if you depended on this
1659 feature, any help would be really welcome.
1661 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1664 SimGrid-java (3.7.1) stable; urgency=low
1666 The "Java aint got to be bloated and slow" release
1669 * Various internal cleanups and performance improvement
1670 Simulations are expected to run up to twice faster or so
1671 * Make Process.kill(process) an instance method, not a static one
1672 * User processes are not java.lang.Thread subclasses.
1673 This breaks the compatibility (sorry), but previous API was
1674 brain-dead, making it impossible to have non-trivial
1675 initializations in the process constructor.
1676 * Require a full constructor per Process sub-class.
1677 Kinda breaks the compatibility (sorry), but this allows a much more
1678 efficient way to launch the processes at simulation startup.
1679 * Do not embeed our version of semaphores, java 1.5 can be considered
1680 as sufficiently prevalent for us to not dupplicate its features.
1684 * Add examples for almost every part of the API
1685 We spotted and fixed a lot of bugs in the process
1687 * New module: asynchronous communication API
1688 * New function: Process.sleep()
1689 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1690 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1691 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1693 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1695 SimGrid (3.7) stable; urgency=low
1697 The "spring cleanups (before the next Big Project kicks in)" release.
1700 * We can specify the SMPI latency/bandwidth factor with command line
1701 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1702 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1703 You can also use the "config tag" from platform file by adding this line
1704 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1705 (see "example/platforms/tag_config.xml" to use "config tag").
1706 Note that the command line supersedes the platform file configuration.
1707 * Change the correction factors used in LMM model, according to
1708 the latest experiments described in INRIA RR-7821.
1709 Accuracy should be improved this way.
1710 * Use the partial invalidation optimization by default for the
1711 network too. Should produce the exact same results, only faster.
1712 * Major cleanup in surf to merge models and split some optimization
1713 mechanisms from the core of the models. As a result you can now
1714 specify which model to use (e.g., --cfg=network/model:LV08
1715 --cfg=cpu/model:Cas01) and which optimization mode to use
1716 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1717 Incompatible combinations should err at initialization. See
1718 --help-models for the list of all models and optimization modes.
1719 * The CLM03 workstation model was dropped for simplicity because it
1720 used the deprecated CM02 network model. Use default instead.
1721 * Rename the TCP_gamma configuration option to network/TCP_gamma
1722 * Rename the coordinates configuration option to
1723 network/coordinates, and document it
1724 * Use now crosstraffic keyword instead of the terribly misleading
1725 fullduplex keyword. It is activated by default now in the current
1726 default model, use --cfg=network/crosstraffic:0 to turn it off.
1727 * Ongoing refactoring the model parsing to make XML files optional
1728 See include/simgrid/platf.h for details (still to be completed)
1731 * Major overhaul of the documentation. Almost instructive now :/
1732 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1733 functions and friends. This interface was considered as
1734 deprecated since over 2 years, it's time to inform our users that it is.
1735 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1736 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1737 use these (crappy) functions in your code.
1738 These functions will be removed soon. Stop using them now.
1739 * Deprecate MSG_get_host_{table,number}
1740 Implement MSG_hosts_as_dynar() instead.
1741 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1742 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1745 * Stabilize the parallel execution mode of user contexts
1746 * Introduce configuration variables to control parallel execution:
1747 - contexts/synchro: Synchronization mode to use when running
1748 contexts in parallel (either futex, posix or busy_wait)
1749 - contexts/parallel_threshold: Minimal number of user contexts
1750 that must be part of a scheduling round to switch to parallel
1751 execution mode (raw contexts only)
1752 * Fix bugs that prevented to use suspend/resume along with
1753 synchronization structures.
1754 * Fix bugs in process termination that lead to invalid memory access
1755 in very specific conditions.
1758 * Introduce a parallel mode for the models (controlled by surf/nthreads
1759 configuration item). In our tests, running the models in parallel
1760 never lead to any speedups because they are so fast that the gain
1761 of computing each model in parallel does not amortizes the
1762 synchronization costs, even when ultra fast futexes are used.
1763 This is released anyway because YMMV.
1766 * Performance boost by using a swag internally to compute the set of
1767 tasks that are finished and should constitute the return value of
1771 * Enable it by default now that it is considered rather stable.
1774 * Documentation of the tracing functions.
1775 * Performance gains when tracing categorized/uncategorized resource
1776 utilization by avoiding calls to get route when updating resource
1777 variables. LMM constraints are being used instead.
1778 * API changed to set task categories. Use MSG_task_set_category instead
1779 of TRACE_msg_set_task_category, and SD_task_set_category instead
1780 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1781 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1782 * Examples for link user variables added (see at examples/msg/tracing/)
1783 * Deprecated function TRACE_msg_set_process_category completely removed
1784 * Trace header updated according to the latest Paje file format
1785 * Tracing network lazy updates, no longer obligate users to use full updates
1786 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1787 * Experimental: let user code declare/set/push/pop application states for hosts
1788 * API changed to allow the manual creation of graph configuration files
1789 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1792 * Improve the API of Lua MSG bindings, using the Lua spirit.
1793 * Each simulated process now lives in its own Lua world (globals are
1794 automatically duplicated). It helps writing simulators. Will allow
1795 to run Splay programs within SimGrid in the future.
1796 * Add a Chord example in Lua, equivalent to the MSG one.
1799 * Start the implementation of a solution to express temporal
1800 properties, not only local assertions. This is still an
1801 experimental work in progress, stay clear from it to be safe.
1805 - Add new runtime parameters --help-logs and --help-log-categories
1806 to display information about supported logging parameters and
1808 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1809 don't exists anymore.
1810 * Mallocators: allow value NULL for the reset function.
1812 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1813 create homogeneous dictionaries, where all the elements share the
1814 same free function. Non homogeneous dictionaries will be
1815 deprecated in the next release.
1816 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1817 - Multi-level dictionaries are deprecated.
1819 - new function xbt_dynar_search_or_negative() that is useful when
1820 you have less than 2 million elements in your dynar and don't
1821 want of the extra complexity of catching exceptions when the
1822 element is not found.
1824 - Make xbt_os_thread module (for thread portability) public.
1825 Documentation is still to come, sorry.
1827 - Cleanups and simplifications to make it maintainable again.
1828 - Exotic features (such as memalign and valloc) were removed.
1829 - The metadata were extended and improved so that the
1830 model-checker becomes able to explore and inspect the heaps.
1831 - This may induce a performance drop when enable_model-checking is
1832 ON in cmake (even if it's not used in the simulation), but it is
1833 necessary at this point to get MC working.
1835 Turn model-checking OFF if simulation performance matters to you.
1836 Not enabling it at runtime is not enough, disable it in cmake.
1838 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1841 SimGrid (3.6.2) stable; urgency=low
1843 The "Not coding new stuff allows to polish old things" release.
1846 * New bindings to the NS3 packet level simulator (experimental)
1847 * Use the raw (efficient) execution contexts instead of the sysv
1848 (portable) ones when possible.
1849 * libpcre is now mandatory in any cases since not using it led to
1850 severe performance loss and possibly other issues
1851 * Update the XML platforms:
1852 - G5K: include the latest machine in Nancy
1853 - GridPP and LCG: new platforms
1854 * Documentation was partially updated, at least (more to come)
1856 Bug fixes, cosmetics and small improvements
1857 * Free terminated processes before the end of the simulation to avoid
1858 exhausting the memory of users having very dynamic amount of
1860 * Bug fix and cosmetics about canceling non-running tasks
1861 * Bug fix about the dot loader's issues when using libcgraph
1864 * Create an installer for windows with nsis (amd64 and win32)
1865 - Add an hello world project to illustrate simgrid project creation.
1866 - Embed libpcre into the Simgrid installer to avoid
1867 its compilation burden
1868 * The raw execution contexts should work on Apple now
1869 * Port to Windows 64 bits
1870 - Sysv contexts now have an implementation for this arch
1871 - GRAS communication features now support this arch
1872 * Drop support for borland compiler on windows
1873 - this code was not maintained, and we kinda depend on gcc nowadays
1874 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1875 * Fix portability issue on unstable ubuntu: linker became picky on
1878 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1880 SimGrid (3.6.1) stable; urgency=low
1882 The "Oops, we broke Macs too" release
1885 * Fixed contexts detection so that raw ones are used when possible
1886 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1887 leads to a strange error, with user code segfaulting sometimes when
1888 the generated code is not perfectly aligned (which is not
1889 controllable from the user side, depends on the amount of code)
1892 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1894 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1896 SimGrid-java (3.6) unstable; urgency=low
1899 * Split of every thing from simgrid v3.5 into a separate package.
1901 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1903 SimGrid (3.6) stable; urgency=medium
1905 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1908 * Bindings now constitute their own package, separated from the main one.
1909 Rationale: reduce our maintenance nightmare by reducing the module coupling
1910 They will soon be released on their own on gforge.
1912 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1913 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1915 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1916 * It's not quite deprecated for now because we have no replacement,
1917 but it may soon become the case.
1920 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1921 * Fortran: New user-level cache variable to store the rank of the running
1922 process. This improves performance by an order of magnitude.
1923 * C: New coccinelle script to automatically locate and modify global and
1924 local static variables.
1925 * Improved SMPI network model with a sender-side gap to account for multiple
1929 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1930 only say if a communication is finished, no matter whether it succeeded or
1931 failed. You can call MSG_comm_get_status() to know the status of a finished
1933 * New function MSG_task_dsend() to send a task and detach it. When a
1934 communication is detached, you are never notified of its success or failure
1935 and the memory is released automatically once it is finished. This function
1936 is useful when you don't care about the end nor the success of a
1938 * Change the prototypes of action replay. Sorry for inconvenience,
1939 but this is really more efficient this way (and to adapt your code,
1940 you just have to fix the initialization, that shouldn't be too long)
1941 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1942 ever managed to do anything useful with it.
1943 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1944 takes a dynar of communications. It returns immediately and gives the
1945 index of a finished communication (if any).
1946 * New example: a basic implementation of the Chord P2P algorithm.
1949 * New model for multi-core CPUs. You can now use the core attribute to
1950 precise the number of cores of a host. This is a basic model. Every
1951 process running on the host receives at most the power provided in
1952 the DTD (throughput<=power). Total throughput of process cannot exceed
1954 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1955 router linked by an up-link and a down-link (possibly asymmetrical).
1956 This kind of pattern allows to easily build last-mile model style platforms.
1957 Aggregating such patterns in a rule-based AS is thus the technique of
1958 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1959 * New model for Vivaldi routing. We transformed the Vivaldi network model
1960 into a Vivaldi routing model (based on the rule-based model). This allows to
1961 combine Vivaldi based latencies with last-mile platforms.
1964 * Added a check for NaN of IEEE754 infinite in the double entries of
1966 * Introduce a new context factory "raw", highly inspirated from the
1967 ucontext factory, but using manually crafted functions in assembly to
1968 do the work in an efficient manner.
1969 * Allow to change the used context factory at run time, not only at
1970 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1971 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1972 contexts at runtime (only with raw contexts or ucontexts).
1973 * Completely rewrote this module to allow parallel execution of user
1974 processes. Use --cfg=contexts/nthreads:N to execute user processes
1975 with N parallel threads (the default is 1, meaning no parallelism).
1976 * Allow to decide dynamically between sequential and parallel modes.
1977 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1978 processes in parallel only when their number is greater than or equal to P
1980 * Added a check for NaN of IEEE754 infinite in the double entries of
1984 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1985 won't output the state of processes when interrupted with Ctrl-C
1986 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1987 NULL-terminated array. This may solve backward compatibility issues
1988 due to the change to return type of SD_simulate. See also:
1989 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1990 * Add new macros with variable number of arguments.
1991 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1992 - in xbt/asserts.h: xbt_assert
1993 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1994 - in xbt/ex.h: THROWF and RETHROWF.
1995 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1997 * Change xbt_die() to accept a format string with arguments, just like printf.
1998 * New data structure: xbt_lib_t, like a dict but more general and with better
2002 * New configuration options
2003 Options triva/categorized and triva/uncategorized can be used to generate
2004 graph configuration files for Triva visualization tool.
2005 * Configuration option tracing/platform is renamed to tracing/categorized
2006 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2007 * New TRACE_user_link_variable interface:
2008 User provides the name of the link and the tracing variable to attach to it
2009 * the declaration of tracing categories must be done after the environment
2011 * simpler tracing interface, just one way to declare categories
2012 TRACE_category or TRACE_category_with_color, it is up to you
2013 * links in the trace file are again identified by their names
2014 * trace contains the full platform hierarchy exactly as declared using the ASes
2015 * Options tracing/msg/[task|process]:1 groups the process by hosts
2016 for both cases, tasks and processes must have names that are unique during the simulation
2017 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2018 * The experimental option tracing/msg/volume is deprecated
2019 its functionality may be reincorporated if needed
2021 The tracing generates a trace file with unordered timestamped events,
2022 because of the way the core simulator (surf) works. A script available
2023 at the tools directory (fix-paje-trace.sh) can be used to put the events
2024 in order. We have changed the tracing so it can generate ordered timestamped
2025 events in the final trace, but depending on the simulator (and how much time
2026 is simulated) that can lead to a huge memory utilization. It is deactivated
2027 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2029 Build Infrastructure
2030 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2031 - We are trying hard to keep the API stable, but it may happen that
2032 some things change (we're a research project after all, not a
2033 nuclear plant operating system). If such things should happen, you
2034 could rely on that macro to adapt.
2035 - current value: 30600 for 3.06.00, aka 3.6
2036 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2037 a number that can be compared with SIMGRID_VERSION.
2038 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2039 assertions and verbose/debug logging events are disabled at compile time.
2041 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2043 SimGrid (3.5) stable; urgency=medium
2046 * New feature to any SimGrid-based simulator: Model-Checking
2047 Check SIN#1 for more details.
2050 * New Model SMPI (three-interval linear regression for correction factors)
2051 See RR-7426, available at http://hal.inria.fr/inria-00527150
2052 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2053 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2054 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2055 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2056 * See also src/smpi/README
2061 - Tracing API changes: TRACE_start and TRACE_end should not be called
2062 by user-code. They are automatically called by simulators created
2063 with SimDAG, MSG and SMPI if the toolkit is compiled with
2064 tracing_enabled=ON. Categories declaration and utilization remain the
2065 same for MSG and SimDag.
2066 - A function was added to the tracing API to declare categories with
2068 - TRACE_category_with_color (char *category, char *color)
2069 where color must be in the following format
2070 "%f %f %f", red, green, blue
2071 and red, green, blue are float values in the interval [0, 1]
2072 - User can specify NULL as color parameter, or continue calling
2073 TRACE_category (cat)
2074 On that case, the tracing system will define random colors
2075 - The following command-line options are supported:
2076 --cfg=tracing/filename:msg.trace
2077 --cfg=tracing:1 (activate tracing, needed to use others)
2078 --cfg=tracing/platform:1 (categorized resource use)
2079 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2080 --cfg=tracing/msg/task:1 (task creation)
2081 --cfg=tracing/msg/process:1 (process creation, migration)
2082 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2083 --cfg=tracing/smpi:1 (SMPI interface tracing)
2084 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2085 - examples of examples/msg/tracing updated
2087 - DAXLoader and DOTLoader functions can generate tasks with categories
2088 - A new function to attribute a category to SD tasks:
2089 TRACE_sd_set_task_category (SD_task_t task, char *category)
2090 * Tracing the MPI interface implemented by SMPI
2091 - Collective operations are traced with states
2092 - Point-to-Point operations are traced with states/links
2093 - Tracing activated by a parameter "-trace filename" passed
2094 to smpirun during execution (considering that simgrid
2095 is compiled with tracing enabled)
2096 - To run the simulation with gdb, the simulator
2097 accepts --cfg=tracing/smpi:1 to trace SMPI
2098 - tesh files to check if smpi tracing is ok
2099 - See examples/smpi/NAS/DT-trace
2100 * GTNetS tracing re-worked
2101 - adaptation to the tracing system of GTNets to cope
2102 with modifications regarding the fullduplex mode
2103 - new tesh files to check if gtnets tracing is ok
2106 * Asynchronous communications through the functions:
2107 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2108 * New function: MSG_load_platform_script()
2109 to make possible using a lua script instead of XML files to set up platforms
2110 * New function: MSG_set_function
2111 to associate functions to processes, used when bypassing the parser
2112 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2114 Platforms: Add some more examples in examples/platforms
2115 * Grid'5000: see www.grid5000.fr
2116 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2119 * Change the XML format. This is a very important modification. SimGrid
2120 3.5 introduces a new hierarchical format based on the notion of
2121 Autonomous Systems. Compatibility with old format is ensured
2122 through the perl script provided in the install bin directory
2123 bin/simgrid_update_xml.
2124 It is now possible to build platforms with specific routing
2125 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2126 connect several platforms together. We will try to provide soon
2127 set of realistic platforms exploiting these properties (have a
2128 look at examples/platforms/ for the moment).
2129 * Take the opportunity of the XML format change to be a good XML citizen:
2130 rename link:ctn to link_ctn and similar changes (also dealt with by
2132 * Add a new routing scheme (rule-based) using regular expressions. It
2133 enables to have an extremely low memory footprint when the
2134 underlying routing is simple and can be compactly described. You
2135 need to have libpcre4-dev (perl regular expressions) installed if
2136 you want to use this routing scheme.
2137 * Revive the cluster TAG and allow to easily and efficiently (both in
2138 term of memory and speed) connect clusters together. Have a look
2139 at teshsuite/simdag/platforms/ to see how this can be done. With
2140 this tag, you can create clusters with thousands of tasks at no
2141 cost (have a look at examples/platforms/).
2142 Note: clusters are implemented as ASes, so there is no need for an
2143 enclosing AS tag if you have only one cluster in your platform.
2144 * Add new generic functions in the public interface that allows the user
2145 to call SURF 'create_resource' methods from your code (same
2146 functionality as the XML bypass mechanism but with a much lighter
2148 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2149 that uses a piecewise linear approximation to produce better
2150 results when exchanging small messages.
2151 * Add a new parameter to handle correctly full duplex link and account
2152 for interferences between uplink and downlink communications
2153 (activate with --cfg=fullduplex:1).
2156 * Rename the SD_READY (all dependencies are satisfied and task is
2157 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2158 dependencies are satisfied) state.
2159 This prevents a confusion between the notion of "ready to schedule"
2160 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2161 simulated" (SD_RUNNABLE) used by the simulation kernel.
2162 * Change the way a task is considered as ready. Instead of removing
2163 dependencies when a task is done, a counter is decreased. This way,
2164 it is always possible to reach ancestors thanks to the
2165 SD_taks_get_parents function (even after the end of the simulation.)
2166 * Change the return type of SD_Simulate from (SD_task_t*) into
2167 xbt_dynar_t. This function was in handling a dynar internally and
2168 converted it into a NULL terminated array for historical reasons.
2169 * New function SD_dotload(char*) to load a DAG described in dot
2170 format. This loader and the corresponding examples require the
2171 installation of the graphviz library.
2172 * Fix a bug in the management of tasks of size 0 in the surf network
2173 models. This problem was only visible with SIMDAG and you should
2174 thus disregard results produced with earlier versions if you
2175 relied on this feature (some tasks were blocked because of this).
2176 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2177 even though your applications were doing only point-to-point
2178 communications and sequential computations. Now you can really use any
2179 model you want (of course, if you create real parallel tasks, which are
2180 not implemented in most models beside ptaskL07, this will abort).
2181 * Add an example that schedules a DAX on an heterogeneous platform
2182 using a Min-Min strategy.
2183 * New function SD_workstation_get_current_task() that returns the kind
2184 of task currently running on a workstation in the sequential access
2186 * Raise some warnings when unexecuted tasks remains at the end of the
2187 simulation. This is usually caused by cycles in the DAG.
2190 * New function: SIMIX_process_set_function() called by MSG_set_function
2191 * Change the underlying waiting queue in semaphores so that a process
2192 can wait on several of them simultaneously (as in waitany).
2193 * Fix the way to handle tokens in semaphores so that all access patterns
2194 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2195 * kill the dirty pimple SIMIX_message_sizes_output()
2196 Please use (proper) visualization instead
2199 * New data container: setset (set of sets of elements)
2200 * New module: mmalloc (mapped malloc, allowing to have several
2201 independent segments of malloc)
2202 * New function: xbt_dict_cursor_set_data()
2203 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2204 * New function: xbt_dynar_is_empty()
2205 * New function: xbt_fifo_get_last_item()
2206 * Fix xbt_dynar_shrink(): use the right element size.
2207 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2208 uninitialized areas during expand.
2209 * Fix semaphores: previous implementation was severely broken.
2210 * Use library init/fini functions for our initialization.
2211 - you can use logs and other feature as soon as you want in your
2212 code (even before the xbt_init / MSG_init)
2213 - xbt_exit is now a no-op and produce a warning when used.
2216 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2217 benefit from the latest improvement to the simulation kernel.
2218 * Kill measurement sockets for now. If you rely on them, sorry. This
2219 release is not for you. This feature will be reintroduced in the
2220 future, but we cannot delay the release any further.
2221 * New function: gras_msgtype_get_name().
2222 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2223 * Fix (at last) the pmm example: it should not randomly fail anymore.
2225 Build chain: bug fixes and overall polishing
2226 * Cmake is now stable enough. Hence, we killed the autotools.
2227 * Port to windows ( TM :)
2228 * Fix the 'make install' target.
2229 No need to use 'make install-simgrid' anymore
2230 * Introduce a 'make dist' target compiling a *source* archive
2231 'make package' compiles a binary archive
2232 * Compile java files only on need
2233 * Add --cd and --setenv command line options to tesh
2234 * Out of source builds are not fully supported yet, but we are close
2235 * Enable supernovae and optimization flags by default for our users
2238 * Add layer to set up environment directly from lua, without XML.
2239 * The effect of gras_stub_generator can be achieved through
2240 lua too (check examples/gras/console/ping_generator.lua)
2242 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2244 SimGrid (3.4.1) stable; urgency=low
2246 The "Polishing easter eggs is probably a good idea" release.
2247 This is a bug fixes release only.
2251 * Fix a bug preventing the tasks from begin garbage collected.
2254 * Fix a bug occuring when a host involved in a communication fails.
2255 This was not detected properly by the other peer involved in the
2256 communication. Now, it's reported as a network error.
2259 * Warn the user about loop dependencies in data flow of DAX files
2260 * Obey the control-flow dependencies of DAX files
2263 * Add option "enable_smpi" allowing to not compile SMPI.
2264 Probably useful for the (Mac) users experiencing a build error here
2265 * Improve the detection of lua5.1 and ruby1.8
2267 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2269 SimGrid (3.4) stable; urgency=low
2271 The "Easter in Cargese" release. Also known as (major changes):
2273 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2274 ~> bindings were greatly improved
2275 ~> new tracing infrastructure for better visualization introduced
2277 * the "Welcome to configury modernity" release.
2278 ~> we switched from autotools to cmake, and improved our cdash
2281 A more detailled list of changes follow (full detail in svn log).
2283 Java Bindings: Various Cleanups
2284 * (install java-gcj-compat-dev on debian-like to use them)
2285 * Remove put/get: no need to export deprecated interface in Java
2286 Use send/receive instead.
2287 * Cleanup the examples and add a README per directory
2288 * Remove example autoDestination (that's the only way to go now)
2289 * Remove example explicitDestination (was a plain copy of basic)
2290 * Make JniException a runtime exception, so that there is no need to
2291 declare the fact that you may encounter such a beast. I guess that
2292 nobody will ever want to survive such error.
2293 * Create specific errors for each MSG case of failure:
2294 host failure, transfer failure, timeout, task cancelled
2295 * Cleanup the exceptions that may get thrown by each function
2296 * Other internal cleanups in Java bindings. Performance still bad :/
2297 Ruby and Lua Bindings: create them
2298 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2299 * That's new and great, you should try them out.
2300 Same functionalities than Java bindings, only even less polished
2302 * Kill the useless "rate" argument of SD_task_get_execution_time()
2303 Everyone used to provide -1 as a value, it was not used, and the
2304 semantic of a possible use wasn't even clear.
2305 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2306 either as comm costs or compute costs to mean that there is no
2307 such thing for that specific task.
2308 * Add a SD_task_set_name() function
2309 * Fix SD_task_unschedule() on typed tasks
2310 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2311 * In DAX loader, accept useless 'level' attributes to <job> since
2312 LIGO DAGs have them (seem to be to ease graphical representation).
2314 * Add an example masterslave_mailbox.c using send/receive and not
2315 the deprecated put/get interface.
2316 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2317 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2319 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2320 (declare MSG_USE_DEPRECATED to still have the old name)
2321 * Add a MSG_task_set_data() function
2322 * About trace replay (see examples/msg/actions):
2324 - Allow to work with splitted trace files for each process
2325 Give the specific trace file as argument of each process,
2326 and call MSG_action_trace_run(NULL)
2327 You can still have one merged file for all processes.
2328 - Fix implementation of collective operations
2329 * Allow task_execute() on 0-sized tasks (closes #10063)
2331 * This is the first release of SimGrid where SMPI is not considered
2332 beta anymore (even if some corners should still be improved)
2333 * Port over the new SIMIX_network submodule (internal refactoring)
2334 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2335 * Implement more missing elements of the standard:
2337 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2339 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2340 MPI_Scan() MPI_Get_processor_name()
2341 - Added implementation of missing case for Alltoall (warning: it's
2342 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2343 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2344 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2345 SMPI_MPI_Allgatherv()
2346 * Bug fixes include:
2347 - MPI_Waitsome() was broken
2348 - Allow relative includes in smpicc
2349 - Command line cfg argument 'reference_speed' was ignored...
2350 - Some functions did not properly lead to auto-benching of user code
2351 - smpicc passes -O2 by default (just like openmpi one)
2353 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2354 * Bug fixes about timeouts during communications
2355 * add SIMIX_message_sizes_output() as a pimple to write to file the
2356 amount of messages per size. Use gnuplot to get histogram.
2357 Pimple because that's the only user-visible function of simix,
2358 defined directly in xbt.h (irk, sorry)
2360 - Add a SIMIX_sem_get_capacity() function
2361 - Fix interactions with processe resume/suspende
2362 - release_forever() was stupidly broken
2363 - Fix SIMIX_display_process_status() for processes in a semaphore
2364 - Make SIMIX_sem_block_onto() user-visible
2365 * Refactoring context stuff:
2366 - Use pseudo-OOP for better modularity
2367 - reimplement SIMIX_process_kill() without process_schedule() so
2368 that the latter can take as invariant that it is called from
2370 - Merge context_start into context_new for sake of simplicity
2372 * Add a Vivaldi network model, coded live during SUD'10 ;)
2373 * Rename configuration variables to start a hierarchy:
2374 o cpu_model -> cpu/model
2375 o network_model -> network/model
2376 o workstation_model -> workstation/model
2377 * New configuration variables:
2378 o network/bandwidth_factor: correction to bandwith
2379 o network/latency_factor: correction to latency
2380 o netwotk/weight_S: correction to the weight of competing streams
2381 * Add a long description to the models, that users can see with such
2382 argument on the command line: --cfg=cpu/model:help
2383 * --help-models display the long description of all known models
2385 * config: add the ability to set a default value after registration
2386 Does not override any previously set value (e.g. from cmd line)
2387 * dict: allow to have integer key and data.
2388 When so, you need to use the following functions
2389 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2390 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2391 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2392 In contrary to regular dicts, the key is not malloced before copy.
2393 Mixing scalar and regular elements in the same dict is not tested
2395 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2396 Tracing for Visualization:
2397 * SimGrid is now instrumented in order to generate a trace file for
2398 visualization analysis: to use it, need to compile SimGrid with the
2399 "tracing" option enabled, and instrument the program using SimGrid with
2400 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2401 (among other functions).
2402 * The instrumentation only traces the platform utilization for now
2403 * Documentation to use the tracing functions and how to analyze the
2404 traces with the Triva tool is written.
2405 * More information about: SimGrid FAQ (in the section Tracing Simulations
2408 * We moved to cmake as default build system. Autotools support will
2409 be dropped soon. Check the FAQ for more info about how to use it.
2410 * Greatly improved our cdash/ctest interactions
2411 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2412 * Added memory checking tests with valgrind; lot of memleak fixing.
2413 This may be the first release of simgrid with so few memory issues
2414 * Added code coverage tests.
2415 Our coverage is still improvable, but at least we see it on cdash.
2417 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2419 SimGrid (3.3.4) stable; urgency=low
2421 The "Desktop Grid needs love too" release (also called Xmas release).
2423 Models improvements:
2424 * Major speedup in the maxmin system solving by using lazy evaluation
2425 Instead of solving completely the maxmin system at each iteration,
2426 only invalidate (and recompute) the modified parts.
2427 This new feature is enabled in default models but you can try to
2428 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2429 * Cas01 IMproved as default CPU model
2430 This CPU model is the same Cas01 model, but it uses the
2431 maxmin-selective-update flag and a heap structure to manage
2432 actions on SURF kernel.
2433 It reduces the complexity to find the next action to finish and,
2434 consequently, it's faster than the old Cas01.
2435 This is the new default CPU model (Cas01).
2436 * Rename the old Cas01 model to Cas01_fullupdate
2437 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2438 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2439 * CpuTI (CPU Trace Integration)
2440 A new CPU model whose objective is simulate faster when using
2441 availability trace files.
2442 Instead of using a full featured, over engineered maxmin system for
2443 CPU modeling, this model does the pre-integration of traces files
2444 to calculate the amount of CPU power available, and so, executes
2445 faster than the old CPU models.
2446 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2447 * Use LV08 as default network model since it gives better accuracy
2448 for small messages and shouldn't change things for big ones.
2449 Use --cfg=network_model:CM02 to get the previous behavior.
2452 ******************************************
2453 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2454 ******************************************
2455 * The new CPU model may changes simulations!
2456 The point is that events occurring at the exact same timestamp
2457 are not scheduled in the same order with the old and new
2458 version. This may be enough to completely change the execution
2459 of simulations in some cases.
2460 * The new network model will change simulations!
2461 This new model is more realistic than the previous one, so you
2462 should consider redoing your old experiments with this model.
2463 Sorry for the inconvenience.
2466 * Introduce the supernovae compilation mode
2467 When compiled that way, the whole SimGrid (or almost) is put in a
2468 single compilation unit and compiled in one shoot.
2469 This is to help gcc which has difficulties to inline stuff from one
2471 The speedup seem to be above 15%, althrough more tests are needed on
2472 amd64 to confirm that gain.
2475 * Port of MSG's mailbox on top of SIMIX network
2476 The put/get mechanism was greatly simplified on the way.
2479 * New SIMIX network module. Provides:
2480 - Mailbox: rendez-vous mechanism to find with who you want to speak
2481 - Synchronous send/recv: easier and hopefully faster since the
2482 logic is handled in the maestro process directly now
2483 - Asynchronous send/recv: you dreamt of it? It's here now
2484 Too bad that nobody cared enough to propagate the change to MSG.
2485 * Add semaphores as SIMIX synchronization mechanism.
2488 * new function SD_daxload(char*) to load a DAX file
2489 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2490 * Introduce typed tasks. Specify its kind and cost at creation.
2491 At scheduling, just give where it should be placed, and the cost
2492 for each involved resource is automatically computed.
2493 Existing constructors so far (more to come of course):
2494 - SD_task_create_comm_e2e() for end-to-end communication
2495 - SD_task_create_comp_seq() for sequential computation
2496 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2497 * new function SD_task_dump() for debuging display
2498 * new function SD_task_dotty(task,FILE*) writing to file the info
2499 about the task in dotty format
2500 * SD_task_dependency_exists() can now cope with having one of its
2501 arguments NULL. If so, it tests whether the other argument has any
2503 * Add getters on list of preceding/following tasks:
2504 SD_task_get_parents(task) and SD_task_get_children(task)
2505 * Add getters on amount of workstations and list:
2506 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2507 * Add getter on task kind: SD_task_get_kind(task)
2508 * Update the start_time and finish_time of tasks on completion/failure
2509 * Bugfix: Remove task from state swags when destroyed
2512 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2515 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2516 dynar with one separate thread per value of the dynar.
2517 * Change the prototype of xbt_thread_create(), sorry.
2518 Added a boolean parameter indicating whether we want to join this
2519 thread (used in SG only for now)
2520 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2523 * GTNetS wrappers should now be usable again (and betterly tested too)
2524 * Fix a major regression from 3.2 where the timeout provided to
2525 MSG_task_put_with_timeout() was used as absolute time before which
2526 the comm should be done.
2527 * Start to fix the <cluster> tag.
2528 - Internal links should be good now (beside of the loopback, which
2529 use the private link instead)
2530 - paths to the external world is still rather broken
2531 - the <route:multi> tag is just broken. Actually that's brain-dead.
2532 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2533 to make it less stupid
2534 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2535 * Fix a source-level compatibility glitch from 3.2: after defining
2536 MSG_USE_DEPRECATED, you can use the old name
2537 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2538 * Allow to compile from the SVN with automake 1.11
2539 * Fix some problems when using the "start_time" tag in deployment XMLs.
2540 * Fix #8569: XBT/synchro.h has redundant declarations
2541 * Fix #8563: MSG return values and exceptions
2542 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2543 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2544 Thanks to Robson Peixoto.
2545 * Fix a vicious bug in dictionaries inducing that some elements were
2546 not freed on xbt_dict_free()
2548 Portability report of this version:
2549 * Main portability targets:
2550 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2551 on (amd64/i386/ia64)
2552 - mac leopard on i386
2553 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2554 but nothing critical.
2555 * Other platforms: windows, AIX and others were not tested for this release
2557 Timing report of this version:
2558 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2559 scenario parameters). From 8h to a few seconds in desktop grid settings.
2560 * Supernovae brings about 25% speedup on i386.
2562 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2564 SimGrid (3.3.3) stable; urgency=low
2566 The "Need for Speed" release.
2568 The timings done to validate the 3.3.2 were faulty.
2569 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2571 The problem was a conversion from a manually handled vector to
2572 xbt_dynar_t on the critical path.
2573 xbt_dynar_foreach calls functions, inducing stack management crap.
2575 We inlined these functions and xbt_dynar_foreach is now breath taking.
2576 We also inlined xbt_swag_belong on the way.
2578 Here are some approximate speedup measurements (on master/slaves
2579 simulations lasting between 10s and 20s each):
2580 3.3.1 -> 3.3.2: about same performance
2581 3.3.2 -> 3.3.3: 40% speedup
2582 3.3.1 -> 3.3.3: 40% speedup
2583 3.3.1 with inline patch -> 3.3.3: 30% speedup
2585 Our reading is that the refactoring which occurred in 3.3.2 made us
2586 suffer much more from the xbt_dynar_foreach low performance, but
2587 once we solved this, this refactoring proved to be very performance
2588 effective. From the 40% speedup, somehow, 10% are due to the
2589 inlining and 30% to the refactoring.
2591 That's a pitty that gcc cannot inline functions placed in other files
2592 alone. We have to choose between:
2593 - break the encapsulation (by putting private data structures and
2594 accessors in headers files to help gcc)
2595 - live with low performance
2596 - switch to a decent compiler such as icc (not quite possible).
2598 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2600 SimGrid (3.3.2) stable; urgency=low
2602 The "Simplicity does not preceed complexity, but follows it" release.
2604 The main contributors of this release were (lexical order):
2605 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2608 * Extract the routing logic into its own object.
2609 (was dupplicated in network.c and workstation_LV07.c;
2610 Allows to implement other ways of storing that info)
2611 => kill now useless network_card concept
2612 - Use dynar to represent routes (instead of void** + int*)
2613 - kill link_set (use surf_network_model->resource_set instead)
2614 - Add a command-line option to choose the routing schema to use
2615 - Add three new models:
2616 * Floyd (shortest path computed at initialization)
2617 * Dijikstra (shortest path recomputed all the time)
2618 * Cached Dijikstra (shortest path computed on need)
2619 All these models where contributed by Silas De Munck, and are
2620 described in his ICCS09 paper.
2622 * Simplify model declaration
2623 (less redirections, less function to write when defining a model)
2624 - Factorize stuff between models:
2627 surf_model_resource_set(model)
2628 surf_model_resource_by_name(model, name)
2629 - Unify the types of models in s_surf_model_t (using an union)
2630 - Embeed fields of common_public directly into s_surf_model_t
2631 - Rename model methods:
2632 action_free ~> action_unref
2633 action_change_state ~> action_state_set
2634 action_get_state ~> action_state_get
2635 - Change model methods into functions:
2636 (model)->common_public->action_use ~> surf_action_ref
2638 * Implement a generic resource; use it as ancestor to specific ones
2639 (allows to kill duplicated code in models)
2640 Drawback: timer command don't need no name nor properties;
2641 workstation_CLM03 don't need no properties
2642 (but I guess we can live with those few bytes wasted)
2644 * Improve the action object model
2645 - implement a constructor avoiding dupplicated code about field
2646 initialization in generic_action part.
2648 * Kill the SDP model: it has an external dependency, is deprecated
2649 in flavor of modern lmm models, and didn't compile since a while
2652 * Relocation of the context module from XBT to SIMIX.
2653 (the context were decoupled from the simix processes, duplicating a lot of code)
2654 => a lot of code was factorized
2655 - less overhead is introduced during scheduling
2656 - simpler API for the context factory
2657 - the logic for process creation,destruction and manipulation was simplified
2658 * Simplification of the s_smx_process_t data structure.
2659 => accesing the simix level data associated to a process is faster now,
2660 and the code is a lot more readable.
2663 * Implement some more MPI primitives:
2664 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2665 -implementation: Bcast: flat or 2-ary tree (default),
2666 Barrier: 4-ary tree,
2668 Allreduce: Reduce then Bcast
2669 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2670 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2671 Alltoallv: flat tree, like ompi
2673 * Add support for optimized collectives (Bcast is now binomial by default)
2674 * Port smpirun and smpicc to OS X
2677 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2678 Shout out if you used it.
2681 * Display the list of still queued messages in SG mode when existing
2685 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2686 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2687 * Add xbt_str_from_file(FILE*)
2688 * Add xbt_dict_get_key achieving a linear reverse search
2689 * Remove the context module
2691 Portability report of this version:
2692 * Main portability targets:
2693 - Linux(debian)/x86/context
2694 - Linux(debian)/x86/pthreads
2695 - Linux(debian)/amd64/context
2696 - Linux(debian)/amd64/pthreads
2697 On these, we still have the eratic breakages of gras/pmm and
2698 amok/saturate_sg reported in previous version. We still think
2699 that the tests are the cause of the fault, not the tested code.
2701 - Mac OSX Leopard/x86/context
2702 Still false negative in tesh autotesting.
2703 Smpi still fails, but this time because readlink does not accept -f
2704 Everything seems to work properly beside of that.
2707 - AIX version 5.3 (only tested contexts this time)
2708 Smpi still fails there because mktemp is not installed.
2709 Everything seems to work properly beside of that.
2711 I managed to compile it for the first time, but several breakages.
2712 Won't delay the release for this exotic platform.
2714 * Windows: it's still lagging behind. If you want to help, please
2717 Timing report of this version:
2718 This version seem to be more than 5% faster than 3.3.1 (on linux
2719 64bits with contextes). The gain is less than expected, we are
2720 investigating this for next release.
2722 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2724 SimGrid (3.3.1) stable; urgency=low
2727 * Implement a --cfg-help to show existing configuration variables
2728 * Build chain do not require doxygen in maintainer mode
2731 * fix a bug on struct sizeof computation, which prevented the
2732 exchange of arrays of structs in some conditions
2733 - added a regression test about this in datadesc_usage
2734 * Allow the exchange of 0-long dynamic vectors.
2735 - for that, use -1 as indicator of dynamic size instead of 0
2736 - This implied to change any size from unsigned long to long,
2737 reducing a bit communication abilities, but I guess that with
2738 64bits being quite common, this is more than enough.
2739 - This also induce a protocol change, thus bumping network protocol
2740 version from 0 to 1 (if we have external users, we have to get
2741 clean on that point too ;)
2742 - added two regression tests about this in datadesc_usage
2743 * Be more verbose when propagating local exceptions
2744 This helps debugging.
2745 * Display the status of simulated processes when receiving SIGINT in
2749 * Allow to control the simulation from a trace file.
2750 New functions MSG_action_register() and MSG_action_trace_run()
2751 The first one allows to associate a function execution to each
2752 kind of action while the second one parses a trace file and
2753 triggers the corresponding actions within the system.
2754 For now, only a toy example is provided in examples/msg/actions
2755 * Add an exemple of process migration in examples/msg/migration
2756 * Fix a bug in task exchange which broke MSG_task_get_sender()
2757 Add a teshsuite regression test for that.
2758 [Bug: if MSG_task_get_sender() is called after sender exit,
2760 * Fix a bug which prevented suspend/resume to work properly
2761 * Display the status of simulated processes when receiving SIGINT
2762 This fixes a regression of v3.3. due to the introduction of SIMIX
2763 * Bug fixing in failure management:
2764 - trace could not start by a failure at time 0
2765 - failure during communications were not working
2768 * Add SIMIX_process_set_name() to change the name of the current
2769 process in the log messages.
2770 * Store smx_hosts in a dict since we only retrieve them by name
2771 * Move the configuration infrastructure to surf
2774 * Move the configuration infrastructure to surf
2777 * Massive internal cleanups:
2778 - Store internal structures on processes instead of hosts (allows
2779 to have more than one process per host, in addition of being more
2781 - Cleanup the initialization/finalization process
2782 - Kill a whole bunch of unneeded synchronization:
2783 processes run in exclusive manner within the simulator
2784 - Move queues from global tables to process data fields
2786 - now accept -platform and -hostfile arguments
2787 - Pass the right rank value to processes according to the hostfile
2788 * Compile the examples by default, and use them as regression tests
2789 * Implement MPI_Wtime()
2790 * Change the reference speed to a command line option
2793 * TCP_gamma can now be specified as command line option using
2794 --cfg=TCP_gamma:10000000.0
2795 * Change the --surf-path cmd line option into --cfg=path:
2798 * Also include strbuff from xbt.h public header
2799 * xbt_ex_display(): do not free the exception after displaying
2800 This allows to do more with the given exception afterward.
2801 Users should call xbt_ex_free() themselves.
2805 Portability report of this version:
2806 * Main portability targets:
2807 - Linux(debian)/x86/context
2808 - Linux(debian)/x86/pthreads
2809 - Linux(debian)/amd64/context
2810 - Linux(debian)/amd64/pthreads
2811 These targets fail about 1/10 of times on gras/pmm, but we believe
2812 that this is because of the test, not because of simgrid.
2813 amok/saturate_sg fails even more rarely, and the test may not be
2816 - Mac OSX Leopard/x86/context
2817 The test suite still spits tons of errors because some obscure
2818 force prevents us from removing the temporary directories
2819 arguing that they still contain some metadata I've never heard of.
2820 Smpi fails because seq is not installed.
2821 Everything seems to work properly beside of that.
2824 - AIX version 5.3 (both contexts and pthread)
2825 Smpi still fails there because mktemp is not installed.
2826 XML inclusions seems rosty on AIX.
2828 * Windows: it's still lagging behind. If you want to help, please
2831 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2833 SimGrid (3.3) stable; urgency=high
2837 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2838 [Malek Cherier & Mt]
2840 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2843 Until now, GRAS were using MSG as an interface to SURF. It was
2844 quite difficult because both interface have several differences
2845 (MSG channels vs GRAS sockets were the most notable point).
2847 This also opens the gate to SMPI (which should occur soon) and speed
2848 up simulations by to 40% (even if it were not the main goal).
2850 **************************************
2851 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2852 **************************************
2853 The point is that events occuring at the exact same timestamp are
2854 not scheduled in the same order with the old and new version. This
2855 may be enough to completely change the execution of simulations in
2856 some cases. Sorry for the inconvenience.
2858 * Cleanup and upgrade the XML format to push further scalability
2859 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2861 * Improve the testing infrastructure with tesh. Now a very large part of
2862 the code is tested not only by being run but also by checking that the
2863 output match an expected output [Mt].
2865 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2866 is really less memory-demanding, which should allow you to use
2867 larger files in SimGrid [AL].
2869 * Inform valgrind about our contextes, so that it becomes usable
2870 with the default (and more effecient) version of SimGrid
2871 [contributed by Sékou Diakite, many thanks]
2874 * Introduce a listener thread in charge of receiving incoming
2875 messages from the network. It allows to overlap communication and
2876 computation but most notably, it removes some stupid deadlocks due
2877 to the fact that so far, a process could not send and receive at
2878 the same time. This made most non trivial communication schema
2880 * Convert the PIDs from long int to int to match the MSG ones (and
2881 linux ones too) [Mt]
2882 * New function: gras_agent_spawn() to launch a new process on
2883 current host. Only working in simulation for now. [Mt]
2884 * New function: gras_os_hostport() returning a constant form (ie,
2885 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2888 * Make the backtrace of exceptions more human readable [Mt]
2889 * New module: xbt/str [Mt]
2890 a ton of string utility functions (split, join, printf to a newly
2891 allocated buffer, trim, etc)
2892 * New module: xbt/hash [Mt]
2893 SHA1 hashing algorithm (more to come if needed)
2894 * New module: xbt/synchro [Mt]
2895 synchronization tools (mutex and conditions) working the same way
2896 in simulation and in real life (mainly useful for GRAS, but not
2898 * New module: xbt/queue [Mt]
2899 classical producer/consumer synchronization scheme
2900 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2901 (using the classical functions will get serialized) [Mt]
2902 * Make dictionary internal table dynamic. No need to specify its size
2903 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2905 * Make sure the log channels are organized as a tree under windows
2906 (because of ANSI C compatibility issue, any channel were child of
2907 root directly) [Mt].
2910 * Cleaned many thing in surf and fixed a few bugs [AL].
2911 * Add a nice command line configuration mechanism to compose models [AL].
2912 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2913 the previous one (KCCFLN05). It relies on something that looks like
2914 a max-min sharing mechanism but cannot be written as such. A new solver
2915 was thus designed [AL].
2916 * Add a new solver to lmm. Based on Lagrange optimization and
2917 gradient-based descent, it enables to efficiently maximise systems s.a
2919 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2921 This solver enables to propose two new network models for TCP Reno and
2922 TCP Vegas based on Low's work. These models still need to be fully
2923 tested though [Pedro Velho].
2926 * Bug fix in SD_simulate. Now the time bound given as argument is
2928 * Use the new parallel task model (ptask_L07) as default.
2929 * Use the SURF command line configuration mechanism.
2930 * 0-size tasks (for synchronization) should now work.
2932 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2934 SimGrid (3.2) stable; urgency=high
2938 We still experience issues on this platform, but we believe that at
2939 least MSG is usable.
2941 GRAS API BREAKAGE (for simplification purpose, sorry):
2942 * the gras_msgtype_by_name is not used anymore. Instead of
2943 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2944 you can write (and must)
2945 gras_msg_send(toserver, "request", &request);
2946 - If you still want to pass a gras_msgtype_t to the function (to cache
2947 the type and avoid the lookup time), use the gras_msg_send_() variant.
2948 - Impacted functions:
2949 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2950 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2951 * The callbacks are now expected to return 0 when everything went well
2952 (just like the main() function)
2954 GRAS new features and improvements:
2955 * New module mechanism where user code can use per process globals [Mt]
2956 This is similar to gras_userdata_*() functions, but for libraries. It
2957 factorize some code developped over and over in the examples and AMOK.
2958 It has still to be documented and used (only amok/peermanagement is
2960 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2961 loss when they were small enough to fit into the buffer and sent quickly
2962 enough so that they can all get received in one shoot.
2963 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2964 if not found. Use the *_or_null() variant for the old semantic.
2965 * In gras_msg_handle, do not discard messages without callback.
2966 They are probably messages to be explicitly awaited later (ie, proofs of
2967 mis-synchronization in userland since they are sent before being awaited)
2968 No big deal usually.
2969 * gras_socket_meas_send/recv: semantic changed!
2970 The numerical arguments used to be (1) the total amount of data to send
2971 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2972 messages. This was need for the fool willing to send more than MAXINT
2973 bytes on quite fat pipes.
2976 * Do really rename the hostmanagement module to peermanagement. [Mt]
2977 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2978 but this is rather new and this was documented in the module
2979 documentation (poor excuses, I admit)
2980 * Bandwidth measurement semantic changed! This follows the changes to
2981 gras_socket_meas_send/recv explained above.
2984 * A sequential mode has been added to the workstations. When a workstation
2985 is in sequential mode, it can execute only one task, and the other tasks
2986 are waiting in a FIFO. [Christophe Thiery]
2989 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2990 model for SIMDAG. [Christophe Thiery]
2991 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2992 equal to 0, which caused some bad behaviors in
2993 saturated_constraint_set_update. I now use a threshold mechanism like in
2997 * When running manually src/testall, you select specific units [Mt]
2998 testall is the result of our cunit mechanism, and should replace all
2999 the scripty thingy around since bash don't run easily on billware.
3001 * A mallocator system has been added. [Christophe Thiery]
3002 Mallocators allow you to recycle your unused objects instead of freeing them
3003 and allocating new ones.
3005 Documentation update:
3006 * FAQ reworking + New FAQs:
3007 - "Valgrind spits tons of errors!" [Mt]
3008 - "How to repport bugs" [Mt]
3009 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3010 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3011 - Communication time measurement within MSG [AL]
3012 - I experience weird communication times when I change the latency [AL]
3013 * GRAS tutorial [Mt]
3015 - an introduction to the framework and to the used communication model
3016 - an initiatic tour introducing the most proheminent features:
3018 . Lesson 0: Installing GRAS
3019 . Lesson 1: Setting up your own project
3020 o Part 2: Message passing
3021 . Lesson 2: Exchanging simple messages
3022 . Lesson 3: Passing arguments to the processes (in SG)
3023 . Lesson 4: Attaching callbacks to messages
3024 . Lesson 5: Using globals in processes
3025 . Lesson 6: Logging information properly
3026 . Lesson 7: Using internal timers
3027 . Lesson 8: Handling errors through exceptions
3028 . Lesson 9: Exchanging simple data
3029 . Lesson 10: Remote Procedure Calling (RPC)
3030 . Lesson 11: Explicitely waiting for messages
3031 . Recapping of message passing features in GRAS
3032 - A HOWTO section containing:
3033 o HOWTO design a GRAS application
3034 More are due, of course. They will come latter. In the meanwhile, you can
3035 check the examples which are still here.
3037 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3039 SimGrid (3.1) stable; urgency=high
3043 There was a stack corruption somewhere, visible only when optimizing
3044 with these versions. [Vince]
3047 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3048 module that enabled to play with Directed Acyclic Graphs. It is built
3049 directly on top of SURF and provides an API rather close to the old
3050 SG. Some old codes using SG are currently under rewrite to check that
3051 all needful functions are provided. [Christophe Thiery]
3054 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3055 extension of the classical CLM03 model that gracefully handles
3056 failures. This is now the default model for MSG and GRAS. It doesn't
3057 handle parallel tasks yet though. [AL]
3058 * Bug fix: Weights were not correctly set in the network part.
3059 WARNING: This may have resulted in incorrect results with simulations
3060 where there are more than one flow on a given link. [AL]
3063 * After a (long ?) discussion on simgrid-devel, we have decided that the
3064 convention we had on units was stupid. That is why it has been decided
3065 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3066 WARNING: This means that all previous platform files will not work as
3067 such with this version! A warning is issued to ask users to update
3069 A conversion script can be found in the contrib module of the CVS, under
3070 the name contrib/platform_generation/surfxml_update.pl [MQ]
3073 * Bug fix: Processes were started in reverse order, wrt deployment file.
3074 WARNING: if your code relies on this bug, please fix it. [AL]
3075 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3076 being executed on two different locations. [AL]
3077 * Bug fix: Failures are now better supported thanks to Derrick's tests
3078 (there was many failure situations I hadn't thought of and that weren't
3079 correctly handled). [AL]
3080 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3084 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3085 Exception occurring on server-side are propagated back to client (!).
3087 API CHANGE: the callback changed their prototype. Change:
3088 int my_handler(gras_socket_t expeditor, void *payload_data) {
3090 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3091 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3093 * New! function: gras_msg_handleall to deal with all messages arriving
3094 within a given period.
3095 * New! function: gras_socket_server_range to get a server socket in a
3096 range of port numbers (ease to avoid port number conflicts) [MQ]
3097 * New! gras processes display their backtrace when they get a SIGUSR1
3098 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3099 Sweet to debug RL processes [MQ]
3103 - Do not force experiment sizes to be expressed in kb, or it becomes
3104 impossible to measure the latency this way (needs one byte-long tests)
3105 WARNING: this changes the amok_bw_* function semantic. [MQ]
3106 - Implements the link saturation stuff. [MQ]
3107 * Peer management module:
3108 New! module factorizing code that we wrote over and over [MQ].
3111 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3112 - Test units are placed directly into the library code, they get extracted
3113 automatically and placed into the src/testall binary.
3114 - Convert most of the XBT tests to this system.
3115 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3116 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3117 live in the same C code. This required to change a little bit the API
3118 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3119 * New module: sparse graph structure with basic algorithms (this is work
3120 in progress and the API is not considered to be frozen yet). [AL]
3121 * Display more information on backtraces: source line & function names are
3122 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3123 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
3124 * Reworked a little bit some #include statements to load only required
3125 headers. Some user code that relied on SimGrid to include stdlib or
3126 stdio may need to include it by themselves. [AL]
3127 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3129 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3130 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3131 but I guess that almost nobody used those parts. [MQ]
3133 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3135 SimGrid (3.0.1) stable; urgency=low
3138 * Unfortunately, I had missed 5 misnamed functions:
3139 xbt_fifo_item_t xbt_fifo_newitem(void);
3140 void xbt_fifo_freeitem(xbt_fifo_item_t);
3141 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3142 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3143 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3144 They're now deprecated. Please use their new versions:
3145 xbt_fifo_item_t xbt_fifo_new_item(void);
3146 void xbt_fifo_free_item(xbt_fifo_item_t);
3147 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3148 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3149 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3151 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3152 * Documentation: xbt_log module unmercifully reworked [MQ]
3153 * Bugfix: there was a problem with the ending of contexts with
3154 the pthread backend. It caused some weird deadlock or behavior
3155 depending on the pthread implementation. [AL]
3156 * Bugfix: get the exceptions raised in the simulator repport where
3157 and why they come from when they are not caught in time [AL, MQ]
3160 * Bugfix: Do repport the error when two non-connected hosts try to
3161 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3164 * Add additionnal checkings on communications. Assert that two
3165 communicating hosts are connected by a set of links... [AL]
3168 * Add additionnal checkings on channel values in communication [AL]
3169 * New: MSG_task_get_source to see on which host a task was generated [HC]
3170 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3171 the number of tasks waiting to be received on channel and sent
3173 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3174 waits for the first task coming from a given host.. [AL]
3176 GRAS new functionnalities: [MQ]
3177 * Enhance the parsing macro to allow the size of multidimentional objects
3178 to be given thru annotations.
3179 * New example (and documentation): Matrix Multiplication a la RPC
3180 (as when I was young!) and fix a bunch of bugs found on the way.
3182 GRAS performance improvements: [MQ]
3184 * Reduce the amount of cbps creation/destruction by making it static to
3185 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3188 * Change libdata to a set so that we can search for stuff by ID (and thus
3189 reduce the insane amount of dict lookups)
3192 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3193 RL communication on top of SG since it's so uneffective.
3194 It may also be used for inter-thread communication in RL, one day.
3195 * Use gras_datadesc_copy() to exchange messages on top of SG
3197 - improve message exchange performance on top of SG
3198 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3199 * Don't exchange on the network the size of the used part of buffer,
3200 instead, specify the possible buffer size to read().
3202 - reduces the amount of read/write calls (one pair per exchange)
3203 - reduces the amount of exchanged data (the size)
3204 - allows to retrieve all arrived data on receiver side, if we don't need
3205 it right now (subsequent read will peek the buffer)
3206 - allows the receiver to proceed with the begining of the stream before
3207 everything is arrived
3208 - make it possible to build an iov transport (using readv/writev)
3210 - take care of the data with non-stable storage (like stacked data),
3212 * If possible, TCP send uses vector I/O (when writev() is here)
3213 - Don't use it for receive since we send data sizes and data on the
3214 same stream, so we wouldn't be able to chain large amount of chunks
3215 before having to flush the stuff to read the size.
3216 * Rework the transport plugin mechanism to simplify it and reduce the
3217 amount of pointer dereferencement when searching for the right function
3220 * I guess that now, we do almost as few system calls as possible while
3221 doing as few data copy as possible.
3223 To improve it further, we could try to send all the sizes first and then
3224 all the data (to use iov on receiving size), but it's only a partial
3225 solution: when you have 2 dimensional data, the sizes of the second
3226 dimension is data of the first dimension, so you need 3 streams.
3228 I'm not sure the potential performance gains justify the coding burden.
3230 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3232 SimGrid (3.00) stable; urgency=high
3235 * New! Give the possibility to hijack the surf parser and thus bypass
3236 MSG_create_environment and MSG_launch_application. Have a look at
3237 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3239 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3241 SimGrid (2.96) unstable; urgency=low
3246 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3247 This deprecates the xbt_error_t mechanisms.
3248 It modifies (simplifies) all XBT and GRAS API.
3249 MSG API keeps unchanged (exceptions raised by XBT are caught from
3250 within MSG and masked with existing error handling facilities)
3253 * New! Add a FATPIPE model. [AL]
3254 * New! Add a parallel task model. [AL]
3255 * New! Add automatically a loopback interface (in the default
3256 network model) if none was precised.
3259 * Bugfix: MSG_process_resume now works with the current running process.
3261 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3262 * Modification of MSG_task_get_compute_duration. Once a task has been
3263 processed, the value returned by this function is now equal to 0. [AL]
3264 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3265 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3266 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3267 of a m_task. For now, MSG_task_cancel only works with computation
3269 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3270 of the processor (in Mflop/s) regardless of the current load on the
3272 * API Change: use proper naming convention for MSG_getClock and
3273 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3275 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3276 This function changes the priority of a computation task. This priority
3277 doesn't affect the transfer rate. A priority of 2 will make a task
3278 receive two times more cpu power than the other ones. This function
3279 has been added to suit the needs of Nguyen The Loc and hasn't been that
3280 much tested yet. So if it fails, please report it and send me your code.
3282 * API Change: removed all functions and types that were marked "deprecated"
3283 since many months. Renamed MSG_global_init_args to MSG_global_init.
3285 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3287 SimGrid (2.95) unstable; urgency=low
3290 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3291 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3292 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3295 * Bugfix: Allow absolute paths for platform description files [MQ]
3296 * Bugfix: do free the variables after use. Leads to drastic performance
3298 * Implement max_duration (ie, timeouts) on resources [AL]
3301 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3303 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3304 the needed support (timeouts) [AL]
3307 * Implement measurement sockets. You can now get the bandwidth between two
3308 hosts thanks to AMOK (see below). [MQ]
3309 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3310 dynar over the network (yeah) [MQ]
3311 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3312 * simple/static token-ring example. [Alexandre Colucci and MQ]
3313 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3314 in place before [MQ]
3317 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3318 type parsing. "short" and "long" modifiers were also ignored; other
3319 modifier (such as reference level) are still ignored. [MQ]
3320 * Embeed the buffer size within the buffer itself on SG. [MQ]
3321 That way, send() are atomic and cannot get intermixed anymore (at least
3322 the ones which are less than 100k; bigger messages still have the issue)
3323 * Array size pushed by the field, not by the field type (or each
3324 and every long int will push stuff to the cbps) [MQ]
3325 * use select() to sleep since it allows to portably sleep less than one
3328 GRAS (minor cleanups)
3329 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3332 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3333 * type category 'ignored' killed as it was never used and were difficult
3335 * whether a type can cycle or not is now a flag, leaving room for more
3336 flags (as "ignored", if we feel the need one day ;)
3337 * Rename raw sockets to measurement sockets since "raw" has another
3338 meaning in networking community.
3341 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3342 and offers features not belonging to GRAS but that most applications
3343 need. One day, it may be a set of plugins loadable at runtime.
3344 * New module: bandwidth
3345 bandwidth measurement between arbitrary nodes running this module. [MQ]
3347 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3349 SimGrid (2.94) unstable; urgency=low
3351 The first beta release of SimGrid 3 !
3355 * Update the main page and the FAQ. Adding references to gforge.
3358 * Add a gras_os_getpid function.
3361 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3362 * Extend the logs so that they also print PID, hostname, date, ... if
3364 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3365 and kill the old version which were in testsuite/
3366 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3370 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3371 for users. Instead, each programming environment (GRAS, MSG,...) use it
3372 when needed to provide such a feature to users.
3373 Don't shortcut the mechanism or you will also shortcut the virtualization
3374 you need on the simulator.
3378 * Cleanups in configury with regard to compile optimization/warning flags.
3379 Also add -fno-loop-optimize to any powerpc since it's the optimization
3380 killing gcc (< 3.4.0).
3381 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3382 needed by MSG examples complications
3383 * Borrow configury beautifications from PHP
3386 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3387 warning (thanks loris for stumbling into this one).
3388 * Bugfix: stop loading private headers (gras_config.h) from the public
3392 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3393 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3394 RL would imply a DNS resolver, which is *hard* to do in a portable way
3395 (and therefore delayed).
3396 * Implement a real timer mechanism and use it in timing macros. This allows
3397 to avoid rounding errors and get a 0.000005 sec precision in timing
3398 macros. While I was at it, various cleanups:
3399 - allow to declare more than one timed section per file (fix a stupid bug)
3400 - move some private declaration to the right place
3401 - merge conditional execution and timing macros into emulation module
3402 - document the module
3403 - make sure the module cleanups its mess on gras_exit
3404 * Documentation improvements:
3405 - (new) how to compile applications using GRAS
3406 - (new) emulation support (timing macros)
3408 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3410 SimGrid (2.93) unstable; urgency=low
3412 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3415 - Use Paje properly where used. Still to be sanitized properly.
3416 - Portability fix: Add an implementation of the contexts using pthread
3420 - Add xbt_procname(): returns the name of the current process.
3421 Use it to show the current process's name in all logging.
3423 - fix detection of older flex version and the reaction, since we do
3424 depend on modern ones (we use lex_destroy)
3425 - Better separation of SG and RL in the libs: remove all simulation code
3426 from libgras. As a result, this lib is now only 200k when stripped.
3427 Some of the xbt modules may also be duplicated (two sets and such) and
3428 should be cleaned/killed before SG3.
3429 - Insist on using xlC on AIX because of weird problems involving gcc there.
3430 - Cleanup the make remote stuff. This is now done by scripts
3431 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3432 mainly for our private use, but we're working on changing them to user
3435 - Bugfix: flush the socket on close only if there is some *output*.
3436 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3438 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3441 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3443 SimGrid (2.92) unstable; urgency=low
3445 Alpha 3 on the path to SimGrid 3
3449 - New! First try of benchmarking macros.
3450 - New! First try so that gras_stub_generator generate deployment and
3451 remote compilation helpers.
3453 - Bugfix: Initialization fix in msg_test.
3457 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3459 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3461 - API Change: gras_os_sleep to take the amount of seconds as a double.
3462 Accepting an int was error prone since it was the only location where
3463 seconds were coded as such. It leaded to damn rounding errors.
3464 - Bugfix: Hard to belive that timers ever worked before this.
3466 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3468 SimGrid (2.91) unstable; urgency=low
3470 Alpha 2 on the path to SimGrid 3
3474 - Bug fix in the lmm_solver.
3476 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3477 through the function MSG_paje_output.
3478 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3479 - It is possible to bound the rate of a communication in MSG with
3480 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3481 to write it in the changelog).
3482 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3487 - Various cleanups to the autotools stuff
3488 - Begin to move Gras examples to examples/gras/
3489 - Let make distcheck work again (yeah!)
3491 - documentation overhauled using doxygen.
3492 gtk-doc-tools is dead in SimGrid now.
3493 - Automatically extract all existing logging categories, and add the list
3494 to the documentation (long standing one, to say the less)
3496 - Cleanup the known architecture table. Reorder the entries to group what
3497 should be, and use a more consistent naming scheme.
3498 (some of the test dataset are still to be regenerated)
3499 - New! Allow library to register globals on each process just as userdata
3501 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3502 lookup time (for now).
3503 Use it in msg and trp.
3504 This cleans a lot the internals and helps enforcing privacy of the
3505 headers between the gras components.
3506 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3507 - Bugfix: gras_os_time was delirious in RL.
3508 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3510 - Reenable GRAS now that it works.
3512 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3514 SimGrid (2.90) unstable; urgency=low
3516 Alpha 1 on the path to SimGrid 3
3518 * It is a long time since the last release of SimGrid. I'm sorry about
3519 that but as I had told you, I was rewriting a lot of things. I apologize
3520 to those who had been reporting bugs to me and that I had not answered.
3521 If your bug is still in the new version, please tell me. Here is a
3522 summary of the main changes.
3524 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3525 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3526 portability, speed, and a lot more but you'll figure it out later.
3527 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3528 I wanted to release the new MSG as soon as possible and I have
3529 broken GRAS, which is the reason why, for now, only MSG is fully
3530 functional. A laconic description of these projects is available
3531 in the documentation.
3533 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3534 kernel optimized for our needs (called SURF but only the developers
3535 should use it). Hence, MSG is now roughly 30 times faster and I think
3536 that by rewriting a little bit MSG, I could event speed it up a little
3537 bit more. Beside the gain in speed, it is also much easier to encode a
3538 new platform model with SURF than it was with SG. More to come...
3540 * REVOLUTION 3: I have tried to change a little as possible the API of
3541 MSG but a few things really had to disappear. The main differences
3542 with the previous version are:
3543 1) no more m_links_t and the corresponding functions. Platforms are
3544 directly read from a XML description and cannot be hard-coded
3545 anymore. The same format is used for application deployment
3546 description. The new format is described in the documentation.
3547 Have a look in tools/platform_generation. There is a tiny script
3548 that converts from the old platform format to the new one. Concerning
3549 the application deployment format, parsing the old one is tricky.
3550 I think most of you should however be able to convert your files. If
3551 it is really an issue, I can write a C code that does the conversion.
3553 2) the toolbox tbx does not exist anymore. We now have a library
3554 with much more data-structures but without the hash-tables (we have
3555 dictionaries that are much faster).
3557 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3559 *****************************************************************************
3560 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3561 * don't feel like converting the oldies. (Mt) *
3562 *****************************************************************************
3565 Version 2.90: "the long awaited one"
3566 - Finished rewriting and debugging MSG. Rewrote the documentation.
3567 - disable GRAS for now since it needs to be ported to the newest SG
3570 - Finish the port to windows (using mingw32 for cross-compile)
3573 - Main loop and datastructures of SURF. A cpu resource object is
3574 functional. Surf can thus be used to create cpu's with variable
3575 performance on which you can execute some actions.
3577 2004-11-15 Martin Quinson
3578 - Port to ARM. Simply added the alignment and size descriptions. Should
3579 work, but the ARM machines are so slow that I didn't had the opportunity
3580 to 'make check' over there yet.
3582 2004-11-15 Arnaud Legrand
3583 - Trace manager now written. It uses a heap structure and is therefore
3584 expected to be efficient. It may however be speeded up (particularly
3585 when many events occur at the same date) by using red and black
3586 trees. One day maybe...
3587 - Max-min linear system solver written. It uses a sparse matrix
3588 structure taking advantage of its expected use. Most operations are
3589 O(1) and free/calloc are called as few as possible. The computation of
3590 the minimum could however be improved by using a red and black tree
3593 2004-11-03 Arnaud Legrand
3594 - Rename every gras_* function that was in xbt/ to its xbt_
3596 - Add a heap and a doubly-linked list to xbt
3597 - Added a dichotomy to the dictionaries. make check works as well before
3598 so I assume that the patch is correct. I do not know however if things
3599 run effectively faster than before now. :)
3601 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3602 SimGrid, and the version number is bumped to 2.x
3604 2004-10-29 Martin Quinson
3605 - Introduction of the remote errors.
3606 They are the result of a RMI/RPC on the remote machine.
3607 ErrCodes being scalar values, you can't get the host on which those
3608 errors did happen. Extending the error mechanism as in Gnome is possible.
3609 No idea yet whether it is a good idea.
3611 2004-10-28 Martin Quinson
3612 - Interface revolution: the Starred Structure Eradication.
3613 I used to do typedef struct {} toto_t; and then handle *toto_t.
3614 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3615 - ???_t is a valid type (builded with typedef)
3616 - s_toto_t is a structure (access to fields with .)
3617 - s_toto is a structure needing 'struct' keyword to be used
3618 - e_toto_t is an enum
3619 - toto_t is an 'object' (struct*)
3621 typedef struct s_toto {} s_toto_t, *toto_t;
3622 typedef enum {} e_toto_t;
3623 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3624 s_toto_t) is private.
3626 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3627 so that it can set the variable to NULL. It was so for dicts and sets,
3628 it changed for dynars.
3630 - Fix a bunch of memleaks in dict_remove
3631 - Fix a bug in sg/server_socket opening: it failed all the time.
3633 2004-10-07 Martin Quinson
3634 - Speed up dynar lookup operation a bit.
3636 gras_dynar_get is dead.
3638 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3639 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3640 (which gives you the address of the stored data).
3642 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3643 copy of the data using an affectation to do the job and not a memcpy.
3645 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3646 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3648 It does not really speedup the dynar test because they are
3649 setting elements all the time (and look them seldom). But the dict does
3650 far more lookup than setting.
3652 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3654 2004-10-05 Martin Quinson
3655 - Allow to (en/dis)able the cycle detection at run time.
3657 Whether we should check for cycle or not is now a property of each
3658 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3659 datadesc_cycle_unset allow to remove this property when previously set.
3661 Note that the cycle detection is off by default since it impacts the
3662 performance. Watch the data you feed GRAS with ;)
3664 This property is hereditary. Any element embedded in a structure having it
3665 set have it set for the time of this data exchange.
3667 You should set it both on sender and receiver side. If you don't set it on
3668 sender side, it will enter an endless loop. If you forget on receiver
3669 side, the cycles won't be recreated after communication.
3671 - Header reorganization.
3672 Kill gras_private.h, each submodule must load the headers it needs.
3674 2004-10-04 Martin Quinson
3675 - Interface revolution: do not try to survive to malloc failure.
3677 Now, gras_malloc and friends call gras_abort() on failure.
3678 As a conclusion, malloc_error is not a valid error anymore, and all
3679 functions for which it was the only gras_error_t return value are
3680 changed. They now return void, or there result directly.
3681 This simplify the API a lot.
3683 2004-09-29 Martin Quinson
3684 - Re-enable raw sockets.
3685 Created by gras_socket_{client,server}_ext;
3686 Used with gras_raw_{send,recv}
3689 It should allow to kill the last bits of gras first version soon.
3691 This is not completely satisfactory yet (duplicate code with
3692 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3695 - Simplify transport plugin (internal) interface by not passing any
3696 argument to _server and _client, but embedding them in the socket
3699 2004-09-28 Martin Quinson
3700 - Finish the port to AIX.
3701 autoconf was my problem (segfault within the malloc replacement
3702 function. No idea why)
3704 2004-09-16 Martin Quinson
3705 - Fix some size_t madness on 64bit architectures.
3707 2004-09-08 Martin Quinson
3708 - Reduce the number of system headers loaded, overload some more system
3709 calls (such as malloc to cast the result of the system one, and work
3711 - Fix and reintroduce the config support
3713 2004-09-07 Martin Quinson
3714 - Source code reorganization to allow Arnaud to surf all over there.
3715 - Allow to document the logging categories.
3716 - Remove all uppercase from logging categories and useless cleanup in names.
3718 2004-08-18 Martin Quinson
3719 Version 0.6.2 (protocol not changed; API changed)
3720 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3721 gras_error_t), and NULL when not found. Functions expecting a msgtype
3722 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3723 hopefully usefull message.
3724 - Portability to prehistoric sparcs again
3726 2004-08-17 Martin Quinson
3727 Version 0.6.1 (protocol not changed; ABI not changed)
3728 - prealloc some buffers to speed things up
3730 2004-08-11 Martin Quinson
3731 Version 0.6 (protocol not changed; ABI expended)
3732 - The parsing macro can deal with the references, provided that you add
3733 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3735 2004-08-09 Martin Quinson
3736 Version 0.5 (protocol not changed; ABI changed)
3737 - Allow to off turn the cycle detection code in data exchange at
3738 compilation time. It should be at run time, but I'm short of time (and
3739 the config stuff is still broken). That way, we keep dict out of the
3740 critical path, which is good because the performance is poor:
3741 - search not dichotomial yet
3742 - dynar give no way to access their content and memcpy everytime
3743 - In composed data description (struct, ref and so on), stop foolness of
3744 keeping the subtype's ID, but store the type itself. This keeps sets out
3745 of the critical path, which is good since they rely on dynar and
3746 dictionnaries. The only loose of that is that we cannot detect the
3747 redeclaration of a structure/union with another content (but I'm not sure
3748 the code detected well this error before anyway). We still can detect
3749 the redefinition discrepancy for the other types.
3750 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3751 it breaks the code because of type-punning used all over the place).
3752 This breaks on all non-gcc architectures (for now).
3754 All those changes (plus the buffer of last time) allow me to gain 2 order
3755 of magnitude on cruel tests consisting of 800000 array of integers on two
3756 level of a hierarchical structure (200 secondes -> 4 secondes)
3759 - the selector of reference must now return the type it points to, not
3760 the ID of this type.
3762 2004-08-06 Martin Quinson
3763 Version 0.4 (protocol changed; ABI not changed)
3764 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3765 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3766 - Implement a buffer transport, and use it by default (it relies on tcp in
3767 real life and on sg in simulation).
3768 That's a bit hackish since I had a new field to the structure to store
3769 its data without interfering with the subtype ones. Inheritance
3770 is tricky in C. And that's a kind of reverse inheritance with one class
3771 derivating two classes. Or maybe a game with java interfaces. Anyway,
3772 that's damn hard in C (at least).
3773 Moreover, I got tired while trying to ensure plugin separation and
3774 genericity in SG mode. MSG wants me to do weird things, so let's go for
3775 cruel hacks (temporarily of course ;).
3776 See comment in transport_private.h:71
3777 - do not include all the _interface headers in private but in the files
3778 which really need them (to cut the compilation time when they are
3781 2004-07-26 Martin Quinson
3782 Version 0.3 (protocol not changed; ABI changed)
3783 - Major overhault of the datadesc interface to simplify it:
3784 - shorted the function names:
3785 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3786 - add a trivial way to push/pop integers into the cbps without malloc.
3787 This allows to make really generic sub_type description, which simply
3788 pop their size of the stack.
3789 - add a function gras_datadesc_ref_pop_arr() which does what users want
3790 most of the time: Declare a dynamic array (which pops its size of the
3791 stack) and declare a reference to it. Poor name, but anyway.
3792 - kill the post-send callback, add a post-receive one
3794 2004-07-23 Martin Quinson
3795 Version 0.2 (protocol changed; ABI changed)
3796 - add some testing for cpbs in the test cases, and fix some more bugs.
3797 This invalidate again the little64 data file, since I cannot regenerate
3799 - remove an awfull optimization in the logging stuff, allowing me to:
3800 - understand it again
3801 - learn gcc how to check that the argument match the provided format
3802 - fix all errors revealed by gcc after that
3803 - internal keys of dict are not \0 terminated. Deal with it properly in
3804 loggings instead of segfaulting when the user want to see the logs :-/
3806 2004-07-22 Martin Quinson
3807 - Fix some stupid bug preventing cbps (callback postit) from working
3809 2004-07-21 Martin Quinson
3810 - Some documentation cleanups
3811 - remove the useless last argument of msgtype_declare
3812 - rename the Virtu functions to fit into the 'os' namespace
3813 - move headers src/include -> src/include/gras/ and stop fooling with
3814 gras -> . symbolic link
3815 - make distcheck is now successful
3817 2004-07-19 Martin Quinson
3819 - Build shared library also
3820 - Install html doc to the right location
3821 - stop removing maintainer files in make clean
3822 - build tests only on make check
3824 2004-07-13 Martin Quinson
3826 - No major issue in previous version => change versioning schema
3827 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3828 corresponding dataset.
3830 2004-07-11 Martin Quinson
3832 - Get it working with any kind of structure (we can compute the padding
3833 bytes remotely for all the architectures I have access to)
3834 - Implement the structure parsing macro (still not quite robust/complete)
3835 - Improvement to the remote testing toysuite
3837 2004-07-10 Martin Quinson
3838 [autoconf mechanism]
3839 - get ride of a bunch of deprecated macros
3840 - actually run the test for two-compliment, not only compile it :-/
3841 - test whether the structures get packed (and bail out if yes. Damn.
3842 Alignment is a serious matter)
3843 - test whether the structures get compacted (but respecting the alignment
3844 constraints of each types)
3845 - test whether the array fields of structures can straddle alignment boundaries
3847 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3848 (took me 2 hours to find that bug, looking at the wrong place)
3849 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3850 before sending/receiving and are used to compute the offsets of fields
3851 - Given that padding size depend even on compiler options, keep track of
3852 alignment and aligned_size only for the current architecture. Not a big
3853 deal since we send structure fields one after the other (seems
3855 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3856 program, let it work on linux/gcc/little32. portability todo.
3858 2004-07-08 Martin Quinson
3859 - import and improve remote compilation support from FAST
3860 - make sure make check works on half a dozen of machines out there
3862 2004-07-07 Martin Quinson
3863 Let's say it's version 0.0.3 ;)
3864 - Implement conversions (yuhu!)
3865 - Let it work on solaris (beside conversion, of course)
3866 - Stupid me, using rand() to generate the conversion datatests in not wise.
3868 2004-07-06 Martin Quinson
3869 - Let make dist work, since I'm gonna need it to compile on remote hosts
3870 - Let Tests/datadesc_usage write the architecture on which the file was
3871 generated as first byte.
3872 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3873 architecture descriptions.
3874 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3875 on those architectures.
3876 - Optimization: send/recv array of scalar in one shoot
3878 2004-07-05 Martin Quinson
3879 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3881 - Plug a whole bunch of memleaks
3882 - each process now have to call gras_{init,exit}. One day, their log
3883 settings will be separated
3884 - Continue the code factorisation between SG, RL and common in Transport.
3886 2004-07-04 Martin Quinson
3888 - Redistribution between SG and RL.
3889 We wanna have to accept in SG, so move accepted related parts of RL in
3890 the common part. (more precisely, the dynar of all known sockets is no
3891 more a static in transport.c, but part of the process_data)
3893 [gras_stub_generator]
3894 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3896 2004-07-03 Martin Quinson
3897 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3898 them all up in one shot)
3900 - Fix the ugly case of reference to dynamic array.
3901 - New (semi-public) function gras_datadesc_size to allow the messaging
3902 layer to malloc the needed space for the buffer.
3904 - gras_socket_close now expect the socket to close (and not its address to
3905 put NULL in it after it). This is because the socket passed to handlers
3906 is one of their argument (=> not writable).
3908 - propagate the interface cleanup from last week in datadesc, ie remove a
3909 superfluous level of indirection. User pass adress of variable
3910 containing data (both when sending and receiving), and not of a variable
3911 being a pointer to the data. Let's say that I like it better ;)
3912 The price for that is constructs like "int msg=*(int*)payload" in
3913 handlers, but it's a fine price, IMHO.
3915 - Let it work in RL (yuhu)
3917 2004-06-21 Martin Quinson
3919 - porting SG plugin and SG select to new standards (works almost).
3920 - plug memleaks and fix bugs around.
3923 - cleanup the prototype of data recv and force users to specify when they
3924 want to handle references to objects. Test case working even for cycles.
3925 - plug memleaks. Valgrind is perfectly ok with this.
3927 2004-06-12 Martin Quinson
3929 - cleanup the separation between plugin and main code in plugin creation
3931 2004-06-11 Martin Quinson
3933 - Reput hook for raw sockets, needed for BW experiments
3934 - kill a few lines of dead code
3935 [Data description] Interface cleanup
3936 - gras_datadesc_by_name returns the searched type or NULL.
3937 That way, no variable is needed to use a type desc once, which makes
3939 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3940 two parameters were strings (field name, type name), leading to
3942 [Dicos] Interface cleanup
3943 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3944 This is consistant with the dynar API.
3946 2004-04-21 Martin Quinson
3948 - Porting to new standards.
3950 - interface cleanup.
3951 There is no bag anymore, no need to take extra provision to mask the
3952 pointers behind "ID".
3953 Better splitup of functions between files create/exchange/convert.
3954 This is still a bit artificial since convert and receive are so
3955 interleaved, but anyway.
3957 - add a queued message list to procdata (the ones not matching criteria
3959 - factorize some more code between SG and RL wrt procdata
3961 - use gras_exit in example to track memleaks
3962 - get rid of gs_example now that GS is properly integrated into gras
3963 - update run_test to integrate the latest tests (datadesc)
3965 - rename WARNINGn macros to WARNn since it prooved error-prone
3967 2004-04-19 Martin Quinson
3969 - register init/exit functions within gras module mechanism
3970 - send/receive function.
3971 Convertion is not implemented, but short-cutted if not needed.
3972 struct/array elements are sent one by one (instead of block-wise), but
3973 nobody really cares (yet). Get a prototype before optimizing.
3974 - tests (using a file socket) for DD send/receive on:
3975 - base types: int, float
3976 - array: fixed size, string (ie ref to dynamic string)
3977 - structure: homogeneous, heterogeneous
3978 - chained list, graph with cycle
3979 Believe it or not, valgrind is not too unhappy with the results. The
3980 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3984 - Bugfix when using a filename explicitely (instead of '-')
3986 2004-04-09 Martin Quinson
3988 - factorize more code between RL and SG in socket creation
3989 - Complete the implementation and tests of:
3991 o file (only in RL, and mainly for debugging)
3993 I lost 3 days to design a portable address resolver, and then decided
3994 that the prototype mainly have to run on my box.
3995 Addressing portability too early may be like optimizing too early :-/
3997 - use gras_init in the Tests instead of the crappy parse_log_opt
3998 (the latter function is removed)
3999 [Conditional execution]
4000 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4001 [Code reorganisation]
4002 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4003 Build examples against the RL library, since there is no way to disable
4004 its creation for now.
4006 For information, the beginning of coding on GRAS was back in june
4007 2003. I guess that every line has been rewritten at least twice since