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.
24 - Option model-checker/hash was removed. This is always activated now.
27 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
28 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
31 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
32 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
33 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
34 - FG#34: SG_BARRIER_SERIAL_THREAD?
35 - FG#35: model-checker does not like buster-produced binaries
37 ----------------------------------------------------------------------------
39 SimGrid (3.23.2) July 8. 2019
42 - Nicer introduction page.
43 - Migrate the "Deploy your application" page to the new doc.
44 - Move Java as a subtree of MSG.
47 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
50 - Drop xbt_dynar_sort_strings().
53 - Really fix FG#26: Turning off a link should raise NetworkFailureException
54 - FG#27: Wrong exception thrown to wait_any when link is turned off
55 - GH#328: Java: Canceling multiple tasks in a single vm/host
57 ----------------------------------------------------------------------------
59 SimGrid (3.23) June 25. 2019
61 The Exotic Solstice Release.
64 - SunOS and Haiku OS support. Because exotic platforms are fun.
65 - Stop setting random seed with srand48() at initialization.
66 - Use addr2line as a fallback for stacktraces when backtrace is not available.
67 - Build option -Denable_documentation is now OFF by default.
68 - Network model 'NS3' was renamed into 'ns-3'.
71 - Simgrid can now hopefully be installed with pip.
74 - wait_any can now be used for asynchronous executions too.
77 - New log appenders: stdout and stderr. Use stdout for xbt_help.
81 - 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
82 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
83 - MPI interface is now const correct for input parameters
84 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
85 - Fortran bindings for async collectives.
86 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
89 - Remove option 'model-check/record': Paths are recorded in any cases now.
90 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
91 always sparse. That's an immense gain in memory, and can even be
92 faster because of cache effects. So there is no need to clutter the
93 code to allow the user to go for the unefficient mode.
96 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
97 models proved to be more accurate than these old experiments.
99 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
100 - FG#1: Broken link in error messages
101 - FG#2: missing installation documentation
102 - FG#3: missing documentation in smpirun
103 - FG#6: Python bindings not available on PyPI
104 - FG#7: simple cmake call requires doxygen
105 - FG#8: make python bindings an optional dependency
106 - FG#10: Can not use MSG_process_set_data from SMPI any more
107 - FG#11: Auto-restart actors forget their on_exit behavior
108 - FG#12: -Denable_lto=OFF doesn't disable LTO
109 - FG#13: Installs unstripped file 'bin/graphicator'
110 - FG#14: Installs the empty directory 'doc/simgrid/html'
111 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
112 - FG#17: Dead link in doc (pls_ns3)
113 - FG#20: 'tesh --help' should return 0
114 - FG#21: Documentation link on http://simgrid.org/ broken
115 - FG#22: Debian installation instruction are broken
116 - FG#26: Turning off a link should raise NetworkFailureException exceptions
117 - GH#133: Java: a process can run on a VM even if its host is off
118 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
119 - GH#326: Valgrind-detected error for join() when energy plugin is activated
121 ----------------------------------------------------------------------------
123 SimGrid (3.22) April 1. 2019
125 The Easter Christmas Release (this one is somewhat late).
128 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
129 - Not complete yet: asynchronous activities (amongst others) are still missing
130 - Still ongoing: the interface may change in the future. We need more
131 testers! Please report any glitches.
132 - No new project using Java should start now. Please switch to Python.
135 - Some of the internal cleanups may lead to speed improvements:
136 - The hard limitation on the amount of simulated actors with
137 Java+Mac was removed. Now, the available memory is the only limit.
138 - Our refcounting was tidyied, leading to 10% speedups in some cases.
139 - We are still working on making our code robust to the actor kills
140 and hosts' churn. Things are improving, but it's not perfect yet.
141 - Replaced our own code to display a backtrace (that was forking addr2line)
142 with the Boost.Stacktrace library.
143 You won't see your backtraces without this optional dependency.
144 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
145 - Stop setting random seed with srand() at initialization.
148 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
149 XML file version remains 4.2 since old files are still compatible.
152 - Process termination which was broken at version 3.21 has been repaired.
153 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
154 - Hide the examples into examples/deprecated. New users should use Python.
157 - Drop MSG_process_create_from_stdfunc() from the API.
158 This C++ function was a pimple in the C API, made necessary at some
159 point by the Java bindings. This is fixed now.
160 - Hide the examples into examples/deprecated. New users should use S4U.
161 - MSG_process_create and MSG_process_attach now crash if the host used
162 to run this process is off.
163 - Fix the protype of MSG_process_on_exit()
164 Now use (int,void*) callbacks instead of (void*,void*) ones.
165 The implementation was ways too messy to actually work, I guess.
168 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
169 too often and provide more accurate timings.
170 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
171 - MPI_Alltoallw support
172 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
173 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
174 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
175 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
177 - Basic implementation of generalized requests (SMPI doesn't
178 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
181 - Drop sg_cmdline. Please use xbt_cmdline instead.
182 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
183 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
184 OsSemaphore is implemented in a portable way with C++11 threads.
185 This should allow much more threads to be created at the same time,
186 allowing Mac and Java users to simulate many more actors.
187 - Implement the 'thread' factory with std::thread instead of xbt ones.
188 It is not possible to set the stack size with threads anymore, but
189 -fsplit-stack is the way to go nowadays when using threads.
190 - Drop the xbt_os_thread_t module (now unused)
191 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
192 - Drop xbt_str_join_array().
193 - Drop cunit, use Catch2 instead.
196 - Many cleanups in the kernel::activity namespace. This was long
197 overdue, and shall open the path to many future endeavors.
200 - #132: Java: a process can not shut down its own host
201 - #220: S4U: detached send causes exception if sender terminates soon after sending
202 - #261: Document the parameters of parallel execution's constructor
203 - #300: [s4u] BarrierPtr is missing
204 - #314: SMPI args internal cleanup
205 - #316: Fix a bug related to the CPU utilization of multi-core VM
206 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
207 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
208 - #325: Turning off a host has different behavior on sleeping actors and computing actors
210 ----------------------------------------------------------------------------
212 SimGrid (3.21) October 3. 2018
214 The Restarting Documentation (TRD) Release.
217 - Convert the user manual to Sphinx for improved usability.
218 Unstable version now hosted on https://simgrid.frama.io/simgrid
219 Many glitches remain; stay tuned.
220 - Restrict the Doxygen reference API to the installed header files.
221 (documenting all internals this way is overkill + doxygen scales badly)
222 - New tutorial on S4U (highly inspired from the old MSG one, but with
223 a git repository to fork as a starting point, and a docker image)
224 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
225 - The Developper manual is still to be converted (not compiled until then)
226 - Some parts are still missing in this conversion (such as the
227 platform and deployment sections) while others were blindly converted
228 and would need more love (such as the configuration flags).
229 Things will certainly further improve in the future releases.
232 - Fully reimplement auto_restart mechanism that was utterly broken
233 (fix #22, #131 and #281 that were all very related)
234 - Implement semaphores (fix #303 and #291)
235 - When creating an actor from a function and its parameters,
236 move-only parameters are not allowed anymore, as it would prevent
237 the actor to be restartable if its parameters are consumed on run.
238 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
239 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
240 IO operations that can be started, waited for, or canceled as a regular
242 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
246 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
247 - New host variable: 'core_count'
250 - Remove xbt_os_thread_specific features
251 - Remove portability wrapper to condition variables
252 - Remove xbt_os_thread_yield()
255 - MPICH collective selector now mimics MPICH 3.3b
256 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
257 - MPI_Init(NULL,NULL) is now allowed
258 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
259 - Command-line arguments should now be handled properly in Fortran simulations
260 - MPI Attributes and related callbacks should work in Fortran
261 - Apps using C + Fortran should now work
262 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
263 - Support MPI_Error_String
266 - Due to an internal bug, Msg.run() must now be your last line.
267 We hope to fix it in a future release, and we are sorry for the inconvenience.
270 - #22: Process autorestart seem to only work with CAS01 cpus
271 - #56: Feature request: dynamic SMPI replay
272 - #93: simgrid should not eat --help
273 - #111: How to change the loopback link characteristics?
274 - #116: Platform cluster radical
275 - #117: Platform cluster/host order
276 - #129: Replay_multiple on steroid
277 - #131: Java : autostart does not work
278 - #143: Setting a breakpoint at a given time
279 - #235: Network energy models should be integrated
280 - #264: Add ptask L07 resource tracing
281 - #271: Dynamic and manual replay of SMPI traces
282 - #279: Breakpoints option uses time deltas instead of absolute time
283 - #281: Daemonized actor no longer daemonized after an auto-restart
284 - #282: TIT convertor erase traces when using absolute path in the trace list file
285 - #285: segfault when a s4u actor kills itself with msg process tracing activated
286 - #286: Pajé traces are not exposing the number of cores
287 - #287: Command-line parsing should stop when encountering '--'
288 - #288: MPI_Init(NULL, NULL)
289 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
290 - #290: Method to check if a mailbox has a message ready to be consumed
291 - #291: [s4u] semaphores do not seem available yet
292 - #294: Alltoall collective implementation of mpich
293 - #295: Starting a Host as OFF is not supported
294 - #296: DTD too permissive
295 - #299: Add a s4u barrier example
297 ----------------------------------------------------------------------------
299 SimGrid (3.20) Released June 24. 2018
301 The proxy snake_case() release.
304 - The whole API is now using snake_case() instead of mixing it with camelCase.
305 Compatibility wrappers in place for one year.
306 - We took the opportunity to increase the consistency of the API.
307 Please report any remaining glitches.
308 - New function simgrid::s4u::Host::get_actor_count:
309 Returns the number of actors running on a specific host.
312 - SMPI is now tested with ~45 proxy apps from various sources, with none or
313 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
314 - Replay: The replay file has been re-written in C++.
315 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
316 supported. They are stored in the trace and used when replayed.
317 - Basic support of MPI_Cancel. Robustness not guaranteed.
318 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
319 (only for MPI_COMM_TYPE_SHARED).
320 - New option: smpi/privatize-libs, to add external shared libs to be privatized
321 by SMPI. They will be copied locally and loaded separately by each process.
322 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
323 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
324 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
325 with internal index optind with multiple processes. Only works if MPI_Init has
327 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
328 syntax without preprocessing tricks.
331 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
332 behavior tracing is no longer limited to MSG
333 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
334 this means that traces now need to be updated or re-obtained; both Irecv
335 and Isend lines in your traces have now in total 5 mandatory fields:
336 <rankid> <command> <to/from rankid> <tag> <size>
337 To update your traces, it suffices to add a 0 for the tag here.
338 - TIT now also supports waiting for a distinct request via MPI_Wait.
339 Wait/Test now wait for a specific request, not just the last one that was
340 issued. This unfortunately means another update, because we need to
341 identify which request you want to wait for. We do this via the
342 triplet (sender, receiver, tag), which needs to be added:
343 <rankid> <command> <sender> <receiver> <tag>
344 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
345 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
348 - The deprecation of MSG is ongoing (but this should not impact you).
349 Many MSG functions are now simple wrappers on the C API of S4U. If
350 you wish to convert your code to S4U, find the S4U counterparts of
351 your MSG calls in src/msg/msg_legacy.cpp.
352 - MSG can still be used, but won't evolve anymore.
355 - Allow to run the Link energy plugin from the command line with
356 --cfg=plugin:link_energy
357 - Rename Energy plugin into host_energy
358 - Rename Load plugin into host_load
361 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
362 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
363 - Improve performance when handling timeouts of simix synchros.
366 - Config: the C API is now deprecated (will be removed in 3.23), and
367 the C++ API has been slightly improved.
370 - Fix several build issues on OSX.
371 - Move simgrid_config.h to simgrid/config.h (old header still working)
374 - #143: Setting a breakpoint at a given time
375 - #150: Inconsistent event names in SMPI replay
376 - #258: daemonized actors hang after all non-daemonized actors have completed
377 - #267: Linker error on unit_tmgr
378 - #269: SMPI: tracing of MPI_Wait/all/any broken
379 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
381 ----------------------------------------------------------------------------
383 SimGrid (3.19.1) Released March 22 2018
385 The Fixed ABI Release.
387 As you may know, we are currently refactoring SimGrid in deep.
388 Upcoming SimGrid4 will be really different from SimGrid3: modular,
389 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
391 Our goal is to smooth this transition, with backward compatibility
392 and automatic update paths, while still progressing toward SimGrid4.
394 SimGrid remains open during works: The last pure SimGrid3 release was
395 v3.12 while all subsequent versions are usable alpha versions of
396 SimGrid4: Existing interfaces remain unchanged, but the new S4U
397 interface is budding and the internals are deeply reorganized.
399 Since 2015, we work hard to reduce the changes to public APIs. When
400 we need to rename a public library symbol in S4U, we let your
401 compiler issue an explicative warning when you use the deprecated
402 function. These messages remain for four releases, i.e. for one full
403 year, before turning into an error. Starting with v3.15, your can
404 also adapt to API changes with the SIMGRID_VERSION macro, that is
405 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
407 Starting with this v3.19.1, our commitment to reduce the changes to
408 the public interfaces is extended from the API to the ABI: a program
409 using only MSG or SimDag and compiled against a given version of
410 simgrid can probably be used with a later version of SimGrid without
411 recompilation. We will do our best... but don't expect too much of
412 it, that's a really difficult goal during such profund refactoring.
414 The difference between v3.19 and v3.19.1 is that the former was
415 accidentally breaking the ABI of MSG, while the later is restoring
418 S4U and kernel APIs will still evolve until SimGrid4, with one-year
419 deprecation warnings as currently. In fact, cleaning up these
420 interfaces and converting them to snake_case() is one release goal of
421 v3.20. But don't worry, we are working to smooth this upgrade path.
423 In summary, new projects should start with S4U to benefit of the
424 future, but old MSG projects should still be usable with no change.
426 ----------------------------------------------------------------------------
428 SimGrid (3.19) Released March 20 2018
430 The Moscovitly-cold Spring release.
432 SMPI user-visible improvements:
433 - Support of MPI_Type_create_subarray
434 - Unimplemented MPI calls now abort (instead of issuing a warning)
435 Some harmless calls around errhandlers still only issue a warning.
436 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
437 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
438 is now possible (mmap privatization was refactored to this end).
439 Adding new SMPI instances should also be technically doable now.
441 SMPI internal cleanups:
442 - Removed index notion from SMPI -- no more "getPid() - 1"!
443 In tracing files, actors are called rank-<PROCESS-ID> but these are
444 NOT the real ranks (tracing several communicators is not supported yet).
445 - Deprecate smpi_process_index() (will be removed in 3.22)
446 - Better testing on classical proxy apps, through an external project:
447 https://github.com/simgrid/SMPI-proxy-apps/
450 - Use a graphical TOC to make it easier to find the documentation you need
451 - Revamp the MSG tutorial
454 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
455 - Remove parameter of MSG_process_killall().
456 Resetting the PID was bogus anyway (several actors could have the same PID).
459 - Execution->setHost() can be called after start(), to migrate it.
460 - Comm::test_any() is now implemented.
461 - s4u::Actor now has onCreation() and onDestruction() signals.
462 - Install some kernel header files for the users' plugins and more.
463 Warning, their API is really not stable yet.
464 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
465 include/simgrid/kernel/routing/*Zone.hpp
468 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
470 Build System and other cleanups:
471 - Remove unused run-time parameter "tracing/onelink-only".
472 - SimGrid now works with intel compilers (MC still buggy)
473 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
474 - tesh kills subprocesses on timeout (but not on windows)
477 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
480 - #194: Feature request: simgrid::s4u::Comm::test_any()
481 - #245: migrating an actor does not migrate its execution
482 - #253: Feature Request: expose clusters as objects
483 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
484 - #255: Tesh broken on Windows
485 - #256: Modernize FindSimGrid.cmake
486 - #257: Fix (ab)use of CMake install
488 ----------------------------------------------------------------------------
490 SimGrid (3.18) Released December 24 2017
492 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
494 MSG IS NOW DEPRECATED. Please use S4U for new projects.
495 - Support for MSG should not be removed from SimGrid before 2020,
496 but future evolutions will be limited to the integration of
497 user-provided patches.
500 - s4u::onDeadlock() signal, run before stopping the simulation.
501 - s4u::Exec: asynchronous executions (abstraction of background
503 - s4u::Host->getLoad() returns the achieved speed in flops/s
506 - Link::name() is deprecated and will be removed in v3.21, use
507 Link::getCname() instead.
508 - Mailbox::getName() changed to return a std::string, use
509 Mailbox::getCname() instead to get a char*.
510 - Storage::getName() changed to return a std::string, use
511 Storage::getCname() instead to get a char*.
512 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
513 to have both allocation and deallocation of the map in user space.
516 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
517 The amount of flops remaining to do is a vector, not a scalar.
518 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
519 name implies on both sequential and parallel tasks.
520 - Both changes fix GitHub's #223 using PR #237 as a basis.
521 Thanks Michael Mercier.
522 - Most examples were converted to S4U and hidden elsewhere at tests.
523 Rationale: we still want MSG to work; we want newcomers to use S4U.
526 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
527 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
530 - Switch to the faster dlopen privatization mechanism by default
531 - Documentation improvments
534 - Live migration is getting moved to a plugin. Dirty page tracking is
535 the first part of this plugin. This imply that VM migration is now
536 only possible if one this function is called:
537 - C/MSG: MSG_vm_live_migration_plugin_init()
538 - C/C++: sg_vm_live_migration_plugin_init()
539 - Java: Msg.liveMigrationInit()
540 For C and C++, "simgrid/plugins/live_migration.h" also has to be
544 - Define class simgrid::xbt::Path to manage file names.
545 - Removed unused functions:
546 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
547 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
548 xbt_graph_export_graphviz()
549 - xbt/str.h: xbt_str_join()
550 - Remove unused datatypes:
551 - xbt/heap.h: use std::priority_queue or boost::heap instead
552 - xbt/swag.h: use boost::intrusive::list instead
555 - New link_energy plugin for the consumption of the links.
556 - All of the operations on files and storage contents have been
557 packaged into a plugin (src/plugins/file_system). The current
558 public interface can be found in
559 include/simgrid/plugins/file_system.h
560 To use these functions you now have to initialize the plugin by
561 calling MSG_storage_file_system_init() just after calling
562 MSG_init() or sg_storage_file_system_init() just after creating
566 - Remove the undocumented/untested tag <include>
569 - Remove viva specific tracing as the tool is no longer maintained
572 - #248: Kill of finished processes leads to segfault
573 - #240: xbt_cond_wait_timeout should gracefully return for C apps
574 - #239: please implement signal s4u::onDeadlock()
575 - #230: segfaults when exit() without run()
576 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
577 - #223: MSG_task_get_flops_amount() not working with parallel tasks
578 - #222: Actor::kill() doesn't really kill and segfaults
579 - #221: odd LMM warning when killing an actor
580 - #120: Memory leak when the processes are forcefully killed
582 ----------------------------------------------------------------------------
584 SimGrid (3.17) Released October 8 2017
586 The Drained Leaks release: (almost) no known leaks despite the tests.
588 Even very long simulations will preserve your memory: our testsuite
589 stresses SimGrid for over 45mn on fast machines for a coverage of
590 over 80%, and there is only one single known leak, of about 4kb.
594 - Comm.detach(): start and forget about asynchronous emission. A cleanup
595 handler may be given to free resources if the comm cannot be completed.
596 - this_actor::send(mailbox) is now mailbox->put()
597 - New: simgrid::s4u::Comm::wait_all()
598 - New: Host.execute() for remote executions.
601 - Unused option network/sender-gap is removed.
604 - Deprecate MSG_task_isend_with_matching(): This unused feature
605 really complicates our internals. Will be removed in v3.20.
608 - Improved context termination. It is now possible to free resources, even
609 when a process is forcibly killed.
610 - ContextBoost: add support for Boost versions above 1.61.
613 - Bring back run-time option --cfg=exception/cutpath to remove exception
615 - Removed unused functions:
616 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
617 xbt_str_rtrim(), xbt_str_trim().
618 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
621 - Removed header files obsolete since SimGrid 3.12:
622 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
624 - #3: SD_exit should be made optional
625 - #120: Memory leak when the processes are forcefully killed
626 - #159: Threading test regression in Actor refcounting
627 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
628 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
629 - #186: Actor::killAll() segfaults if some process is blocked on wait()
630 - #191: VM migration and pstate
631 - #192: Updating the energy consumptions of all hosts crashes with VMs
632 - #195: All actors have PID=0 in the logs
633 - #204: Sometimes segfault with thread contexts and mmap privatization
634 - #222: Actor::kill() doesn't really kill and segfaults
635 - #225: Actor::kill() doesn't really kill when victims are doing a join()
637 ----------------------------------------------------------------------------
639 SimGrid (3.16) Released June 22. 2017.
641 The Blooming Spring Release: developments are budding.
644 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
645 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
646 - Java JAR file should be finally fully working on Mac OSX too.
648 XML platforms: Switch to platform v4.1 format.
649 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
650 - <zone> can be used as a synonym for the now deprecated <as>
651 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
652 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
653 - <actor> can be used as a synonym for the now deprecated <process>
654 - state_file and avail_file periodicity is now easier to express
655 (check the documentation) the old behavior should still work.
656 * Storage: not backward compatible, but it seems that nobody ever used it
657 - Remove attribute 'content_type' of <storage_type>: was never used
658 - Make attribute 'model' of <storage_type> optional: for future usage
659 - Remove Bconnection model property: was never (in)validated. Replaced by
660 the maximum of the read and write bandwidth as a resource constraint.
663 - New and Backwards Compatibility break:
664 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
665 When one wants to get the list of tasks whose states have changed during a
666 simulation round, s/he has to allocate and free a dynar and use it as
667 argument to this function. The former SD_simulate (double how_long)
671 - Allow multicore VMs, with the correct sharing (unless you overcommit)
672 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
675 - New function to update the consumption of all hosts at once.
676 - Fix the model for multi-core hosts, linear in the amount of busy
677 cores with an abnormality for pIdle (see paper).
678 BUG: Still not satisfactory for ptask on mono-cores.
681 - The netzone are now available from the MSG API.
682 The old names still work, but are now deprecated.
685 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
686 - New option: smpi/keep-temps to not cleanup temp files
687 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
688 mallocs mode is used, allows to change the size of the fake file used
689 (default 1MB), to potentially limit the number of mappings for large runs.
690 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
691 - Fortran ifort and flang compilers support
692 - New RMA calls supported (experimental):
693 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
694 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
695 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
696 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
697 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
698 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
701 - New: this_actor::isMaestro() returns whether we are in kernel mode.
702 - Behind the scene, ActivityImpl are now automatically refcounted.
703 Removing this nasty bottleneck will greatly simplify our internals.
706 - Replay: New function xbt_replay_action_get():
707 Retrieve the function previously associated to an event type.
708 - DROPPED FUNCTION: xbt_str_varsubst()
709 - DROPPED MODULE: strbuff. We don't need it anymore.
710 - DROPPED MODULE: matrix. We don't need it anymore.
711 - DROPPED MODULE: lib. We don't need it anymore.
713 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
715 SimGrid (3.15) Released March 22. 2017
717 The Spring Release: continuous integration servers become green
718 We fixed even the transient bugs on all target architectures:
719 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
720 Windows, FreeBSD, NetBSD.
722 - Rename NetCards to NetPoints.
723 This was intended to help NS3 users, but that's not a netcard.
724 That's a point in the routing algorithm, let's avoid wrong simplifications.
727 - New: MSG_process_yield(). Stop and yield to other processes.
728 - New: MSG_process_daemon(). Daemon processes are automatically killed
729 when the last non-daemon process terminates
730 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
731 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
732 to make room for the upcoming network energy plugin.
733 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
736 - Ensure that an actor can kill itself with Process::exit()
737 - Kill the obscure NativeException. Nobody want to survive the issues
738 it denotes, so use JniException that is a RuntimeException (not to
739 be caught explicitly).
740 - Partial bug fix in initialization. SimGrid flags on command line were
741 consumed at C level but stayed in the original Java String[] args.
742 This could mess users' args[i] if SG flags were not put at the end of
744 The SimGrid flags are now removed from the Java arguments. However,
745 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
746 if args.length is greater than the number of YOUR OWN ARGUMENTS.
747 It might be if you have --log or --cfg flags in the command line.
748 - Fix numerous memleaks all around the place. In particular, around VMs.
752 - s4u::onPlatformCreated: right before the simulation starts
753 - s4u::onSimulationEnd: right after the main simulation loop
754 - s4u::onTimeAdvance: right after a clock change (time is discrete)
755 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
756 event from the availability_file changes the avail speed.
757 - Links are now usable from s4u
758 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
759 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
760 - New examples: The conversion of MSG examples to S4U has begun
761 - Actors: create, kill, migration, and suspend
762 - Applications: master-worker and token-ring
763 - Action replay: communications and storage
764 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
767 - Backwards Compatibility breaks
768 - The SD_route_get_list and SD_route_get_size functions have been removed.
769 They are replaced by the unique
770 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
771 The route from src to dst is built in the links dynar whose size can
772 be obtained with xbt_dynar_length.
773 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
774 sg_host_route_bandwidth and sg_host_route_latency.
775 Macros ensure the backwards compatibility, but you should fix your code
778 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
779 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
780 - Bump our claim of support from MPI 1.1 to MPI 2.2.
781 We don't support 100% of it, but it should be enough. Contact us if not.
782 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
783 naive and probably inaccurate)
784 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
785 mapped onto a unique small file using some system magic.
786 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
787 - Fortran: cleanups, fixes, support of user-added operations
788 - RMA: MPI_Accumulate are applied in correct order.
789 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
790 - Better support for MPI_IN_PLACE
791 - Support for MPI_Win attrs and keyvals.
792 - Support MPI_Comm_group_create, MPI_Type_size_x
793 - MPI ops should be properly applied to all allowed types
794 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
795 algorithm by a non failing one from mpich.
798 - Kill the fifo data container: we don't use it anymore.
801 - A <cluster> can now be created with different speed values to represent pstates.
802 see examples/platforms/energy_cluster.xml for an example.
804 SimGrid (3.14.159) Released December 28. 2016
806 The previous release was only a crude approximation of a Pi release;
807 this one should be a bit better.
809 - Revert a last minute change that broke on Mac OSX.
810 - Fix the changelog and NEWS files.
811 - Improve the documentation, in particular of the routing module.
812 - Integrate some patches flying around in Debian.
814 SimGrid (3.14) Released December 24. 2016
817 * The whole documentation was reworked and reorganized. There is
818 still much room for improvement here, but we're on it.
821 * We now need python3 for our testing infrastructure.
822 * Model checking needs libevent
825 * Now works on FreeBSD
828 * Privatization now works on FreeBSD
830 * Privatization is now activated by default in smpirun
831 You can turn it off with -no-privatize if something goes wrong.
833 * Call-location tracing for SMPI.
834 You can add the exact location (filename / linenumber) of an MPI call to
835 your trace files and slow down or speed up the simulation between two
836 consecutive calls by using an adjustment file (see the documentation).
838 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
839 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
840 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
841 values were ignored in some cases. The timings of these functions can now
842 be significantly different.
844 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
845 smpi/running-power is renamed to smpi/host-speed
847 * smpi/grow-injected-times option to enable or disable multiplication of the
848 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
849 make simulation less precise (but also much faster).
851 * smpirun script should be (much) faster for large deployments.
853 * SMPI tracing: fixed issue with poor matching of send/receives.
855 * Replay: Fix broken waitall
857 New functions and features
858 * MSG_parallel_task_execute_with_timeout, to timeout computations.
860 Dropped / renamed functions and features
861 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
862 - MSG_mailbox_is_empty() -> Mailbox::empty()
863 - MSG_mailbox_front() -> Mailbox::front()
864 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
865 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
866 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
867 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
868 * Don't pass the free_f parameter to property related functions:
869 - MSG_host_set_property_value()
870 - MSG_as_router_set_property_value()
871 - MSG_storage_set_property_value()
872 * VM properties. Since msg_vm_t are msg_host_t, just use
873 MSG_host_get_property_value() and friends
874 * VM I/O related things:
875 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
876 - Unimplemented save/restore methods
877 * MSG_as_router_get_property_value() was redundant with
878 MSG_environment_as_get_property_value().
879 - Removed MSG_as_router_*propert*() functions
880 - Added MSG_environment_as_set_property_value() for consistency
881 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
882 Well, they are still there for now, but deprecated with a warning.
883 Please switch to xbt_dict_new_homogeneous() before this is removed
885 * Task affinity. Its intended behavior (that was very badly tested
886 and probably not really working) was deceiving what most users
887 would have hoped here.
888 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
889 * Option network/coordinates is now useless and should be dropped.
892 * Added option storage/max_file_descriptors to allow more than 1024 files opened
895 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
896 * Added Dragonfly topology support
899 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
900 EiBps, EBps, Eibps, Ebps for bandwidth.
901 They may become useful to some lucky ones.
904 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
905 * ex/app_tokenring: new example, very similar to the MSG Token Ring
906 * ex/async_waitAll: new example, on asynchronous communications
909 * Memory usage should be decreased for simulations with a large number
910 of processes. This also helps for SMPI.
912 SimGrid (3.13) stable; urgency=low
914 The Half Release, a.k.a. the Zealous Easter Trim.
916 - We removed half of the lines, that were mostly experimental cruft.
917 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
918 (not counting blanks and comments -- according to openhub.net)
919 - The internals are now compiled in C++ (and will soon be clean C++)
920 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
921 59 klines of C, etc. We added only 29 klines of C++ in replacement.
923 * Backwards Compatibility breaks
924 - Removed Lua simulation bindings (switch to C or Java for that).
925 Lua can still be used to describe platforms
926 - Removed Java kernel plug-ins.
927 Will be reintroduced after the ongoing major internals reorg.
929 - the following functions were removed.
930 They were too specific and should be reimplemented in a generic
931 way, with filter function.
932 - MSG_task_listen_from_host
933 - MSG_mailbox_get_count_host_waiting_tasks
934 - MSG_mailbox_put_with_timeout was removed.
935 Please use MSG_task_send_with_timeout instead.
937 - the SD_application_reinit function was removed. It has been a noop for a while.
938 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
939 replaced by a more flexible mechanism.
940 - The following functions thus do not exist anymore
941 - SD_workstation_get_access_mode
942 - SD_workstation_set_access_mode
943 - SD_workstation_get_current_task
944 - Basic estimation functions have been removed but can easily be replaced
945 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
946 - SD_workstation_get_computation_time => amount / sg_host_speed()
948 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
949 Use VM.getSpeed()*load/100 for the legacy behavior.
951 - option enable_tracing was removed. It was not doing anything for a while.
952 - In the ModelChecker:
953 - the model-checker now ptraces the model-checked process which means
954 you cannot use a debugger on the latter anymore (we might make this
955 optional in the feature);
956 - removed soft-dirty page tracking;
957 - remove model-checked side snapshot management,
958 MC_snapshot() and MC_compare_snapshot();
959 - keep the MC_cut() function as a stub (it was not really working
960 in the previous release).
963 * All options are consistently in kebab-case. Old names are kept as alias.
966 * Switch to platform v4 format.
967 - Rename from 'power' to 'speed' the attributes describing the amount of
968 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
969 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
970 - In <host> and <link>, attributes availability and state are gone.
971 It was redundant with state and availability traces, and with peak values.
972 - In <cluster>, attributes availability_file and state_file are gone.
973 It was too complex and unused.
974 - Kill <gpu>. Was not doing anything.
975 - The DOCTYPE points to the right URL:
976 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
977 (the file at this address now documents the changelog since its v1)
978 - A warning is emitted for unit-less values (they are still accepted).
979 - speed. Default: 'f' or 'flops'. Also defined:
980 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
981 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
982 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
983 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
984 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
985 - latency. Default: 's' second. Also defined:
986 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
988 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
989 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
992 * s4u::Host is now the preferred public interface to the Host features.
993 sg_host_* functions are C bindings to the exact same behavior
994 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
997 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
998 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
999 They were deprecated since a few years, and probably did not even compile anymore.
1002 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1003 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1004 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1005 encouraged to update their codes. A list of the modified functions can be found at the end of
1006 include/simgrid/simdag.h
1009 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1010 the lambda or closure passed as a parameter will run in kernel mode.
1011 All the callback functions should be rewritten to that interface at some point.
1014 * Reorganizing and cleaning the internals all around the place.
1017 * Remove old default barrier/bcast buggy algorithms (see #18407)
1018 * Various bug fixes to handle more codes
1019 * Remove the need for the --foreground option of smpirun (it is still
1020 accepted for backward compatibility).
1023 * Kill the setset data container: MC don't use it anymore.
1024 * Kill the queue data container: it made more sense with GRAS.
1025 * Kill the xbt_peer_t data type: it's useless without GRAS.
1026 * Kill rm_cb feature of config sets: it was never useful.
1027 * Kill graphxml parsing feature. It was not used.
1028 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1030 - xbt_dynar_sort_strings(), when the content is char*
1031 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1032 They throw exceptions on invalid input;
1033 * C++ support for declaring CLI flags (simgrid::config::Flag);
1034 * class for abstracting different signal backends (simgrid::xbt::signal).
1035 with no external dependencies (we need very simple signals).
1038 * refactoring and cleanup of the code;
1039 * ongoing process to cleanly separate the model-checking algorithms
1040 from the code model-checking support.
1042 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1044 SimGrid (3.12) stable; urgency=low
1046 The Facelift Release.
1049 * Require g++ v4.7 at least to not speak prehistorical C++.
1050 * Require Boost 1.48 (for signal2 component).
1051 * Java must be version 7 at least when activated.
1052 * Builds on Windows again (including Java bindings).
1053 * Tracing is now always enabled (no way to turn it out)
1054 * Remove GTNetS. It was not working anyway.
1055 * Various cleanups in the cmake scripts.
1056 * Move headers around to sort them out on installed systems:
1057 - instr/instr.h -> simgrid/instr.h
1058 - instr/jedule/* -> simgrid/jedule
1059 - simdag/datatypes.h was removed
1060 - simdag/simdag.h -> simgrid/simdag.h
1061 - msg/datatypes.h was removed
1062 - msg/msg.h -> simgrid/msg.h
1065 * Interface improvement:
1066 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1067 - Sanitize the interface in MSG_task_ module:
1068 - Merge two functions that were close enough but misleading:
1069 set_compute_duration(t) -> set_flops_amount(t)
1070 get_remaining_computation(t) -> get_flops_amount(t)
1071 - set_data_size(t) -> set_bytes_amount(t)
1072 get_data_size(t) -> get_bytes_amount(t)
1073 - Massive cleanups in the functions related to the energy
1074 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1075 - New: MSG_host_get_pstate()
1077 - msg/energy/onoff: switching hosts on and off
1079 * Interface improvement:
1080 - Rename Host.isAvail() to Host.isOn()
1081 - Rename Process.currentProcess() to Process.getCurrentProcess()
1082 - Rename Task.setDataSize() to Task.setBytesAmount()
1083 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1085 - #18874: Actually allows the GC to reclaim tasks
1089 - SIMIX_process_throw: raises an exception in a remote process
1090 * Refactoring: Separate sub-modules
1091 - libsmx: the public interface, as libc in a real system
1092 - popping: the strange dance that converts a user request into a kernel handling
1093 - smx_context_*: the virtualization mechanisms that embed the user code
1094 - smx_*: the handling of each simcalls
1095 * Interface cleanup:
1096 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1097 * Rename smx_action_t into smx_synchro_t, making explicit that these
1098 things are used to synchronize processes with their environment.
1099 For example, a communication is a sort of synchronization involving
1100 the communicating processes (that may block until the exchange) and
1101 the platform. The same can be said from computations, etc.
1103 - #18888: segfault when a process ends before its kill_time
1107 - 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.
1108 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1109 - MPI_Type_set_name, MPI_Type_get_name
1110 - MPI_*_c2f and MPI_*_f2c functions
1111 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1112 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1113 - Activate a lot of new tests from the mpich 3 testsuite
1115 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1116 - InfiniBand network model added: Based on the works of Jerome Vienne
1117 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1118 - When smpi/display_timing is set, also display global simulation time and application times
1119 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1120 * Collective communications
1121 - 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
1122 - 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/).
1123 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1125 - Replay now uses algorithms from wanted collective selector
1126 - Replay can be used with SMP-aware algorithms
1127 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1128 - 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.
1130 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1131 - Correctly use loopback on fat-tree clusters
1132 - Asynchronous small messages shouldn't trigger deadlocks anymore
1133 * Energy/DVFS cleanup and improvement
1134 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1135 - new: smpi_get_host_pstate()
1138 - "Full" network optimization flag was broken since Surf++
1139 - Better handling of precision flags in maxmin
1140 - Fix bug causing sometimes "Impossible" errors
1141 - Properly pass cluster properties to included hosts
1142 * Improvement of the Energy plugin.
1143 - Always update the consumption before returning that value
1144 - New property: watt_off to denote the dissipation when the host is off
1145 - New functions getWattMinAt and getWattMaxAt to retrieve the
1146 dissipation of pstates that we are not currently at.
1147 * Java: class NetworkLink renamed to Link
1148 * New function: simcall_process_get_kill_time()
1149 * Massive rename s/workstation/host/
1150 - That's intrusive, but that's good for the project consistency. Sorry.
1151 - Change config option "workstation/model" into "host/model"
1154 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1155 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1156 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1157 - xbt_dirname and xbt_basename for non-POSIX systems
1159 * The model checker now runs as a separate process.
1160 * The model checker runs must now be launched with the new simgrid-mc program.
1161 * Record/Replay: the MC can display a textual representation of a path in the
1162 execution graph. It can then be replayed outside of the model checker.
1164 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1166 SimGrid (3.11) stable; urgency=low
1171 * Normalizing pointers addresses tool for better diff between logs
1174 * Add cloud examples using new VMs
1175 - examples/msg/cloud/two_tasks_vm.tesh
1176 - examples/msg/cloud/simple_vm.tesh
1177 - examples/java/cloud/cloud.tesh
1178 - examples/java/cloud/migration/migration.tesh
1179 * Add java surf examples:
1180 - examples/java/surfPlugin/surf_plugin.tesh
1181 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1182 - examples/java/surfCpuModel/surf_cpu_model.tesh
1183 * Add SMPI+MSG example:
1184 - examples/smpi/smpi_msg_masterslave/
1189 - msg task destroy cancel test
1190 - msg_host on/off test
1191 * Move all tests in testsuite to teshsuite (adding tesh files)
1192 * Restructure teshsuites
1193 - one folder for each kind of test
1194 * Restructure AddTests.cmake
1196 - structure the order of tests (with sections)
1199 * Add virtual machine
1200 - creation of a VM on a PM
1201 - migration of a VM from a PM to another PM
1203 - MSG_process_join(msg_process_t process, double timeout)
1204 - msg_bar_t MSG_barrier_init(unsigned int count)
1205 - int MSG_barrier_wait(msg_bar_t barrier)
1206 - void MSG_barrier_destroy(msg_bar_t barrier)
1207 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1208 * New option "msg/debug_multiple_use" to help debugging when a task is used
1211 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1213 - Increase code coverage in test suites
1217 * Protect context stacks against stack overflow. The number of protected memory
1218 pages allocated on the top of each stack (1 by default) can be configured
1219 with the new command line option --cfg=contexts/guard_size:<value>.
1220 * Simcalls are now generated by a python script that
1221 - generates files included by SimGrid
1222 - checks that all the functions exist, and proposes prototypes
1224 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1225 host_set_data, host_get_data
1227 - simcall_process_join(smx_process_t process, double timeout)
1228 * Fix bug where sleeping processing could not be suspended.
1231 * Translate surf models from C to C++
1232 - Generic classes for all models: Model, Resource, Action
1233 - A generic interface for each kind of model (CPU, Network, Storage
1234 Workstation, WorkstationVM)
1236 * Translate surf routings from C to C++
1237 * Add callbacks using sigc++ or boost::signals2
1238 - Add callback functions for resource creation/destruction
1239 - Add callback functions for action state change
1240 - Handle Energy as a plugin
1241 * Replace swag by boost::intrusive
1242 * Add new routing models for clusters. For documentation, see
1243 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1244 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1245 parameters for cluster tag
1246 - Fat trees, with topology="FAT_TREE" and
1247 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1248 - see examples/platforms/torus_cluster.xml and
1249 examples/platforms/fat_tree_cluster.xml
1250 * More documentation
1253 * Hostfiles support host:nb_processes construct to deploy several processes on
1255 * Collective communication algorithms should not crash if used with
1256 improper number of nodes and report the error.
1257 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1258 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1259 MPI_Dims_create, MPI_Cart_sub are supported.
1260 * New interface to use SMPI programmatically (still depends on MSG for
1261 some parts, see examples/smpi/smpi_msg_masterslave):
1262 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1266 * Global variables privatization in MPI executables is now performed at runtime
1267 with the option smpi/privatize_global_variables (default:no).
1268 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1269 dynamic libraries loaded by the application are not privatized (static
1270 linking with these libraries is advised in this case)
1273 * Options defined in XML work correctly now.
1276 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1277 copied into simgrid.jar or not (ON by default). Use this option if you want
1278 to reduce the size of the installed simgrid.jar, *and* the native libraries
1279 are kept installed elsewhere.
1280 * Surf binding with SWIG (code generated in maintainer mode only):
1281 - plugin to handle callbacks
1282 - CPU model only for the moment
1285 * Supernovae build mode is definitively removed. It was used to improve
1286 inlining and inter-module optimizations. It is nowadays superseded by
1287 link-time optimizations commonly available in compilers.
1288 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1289 * Add boost dependency for surf++
1290 * Add new macro for tests
1291 - ADD_TESH(name <tesh_args>)
1292 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1296 - xbt_bar_t XBT_barrier_init(unsigned int count)
1297 - int XBT_barrier_wait(xbt_bar_t barrier)
1298 - void XBT_barrier_destroy(xbt_bar_t barrier)
1299 * Make the xbt_os_time module public
1301 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1303 SimGrid (3.10) stable; urgency=low
1305 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1308 * Reintegrate Java to the main archive as desynchronizing these
1309 package is not acceptable anymore (Java is now considered stable)
1310 * Add explicit synchronization facilities through semaphores
1311 * Bug fix: Task.setDataSize() only changed the C world, not the value
1312 cached in the Java world
1315 * Dramatically change the way files are handled. API and internals changed, but
1316 this part of MSG was not considered as production grade either.
1317 * Add explicit synchronization facilities through semaphores
1318 * Add a new function MSG_host_get_process_list()
1319 * Preliminary DVFS support (see examples/msg/energy/ for details)
1322 * SMPI is now included directly in the libsimgrid as the windows
1323 linker doesn't force us on splitting it anymore.
1324 * Improvements of the SMPI replay tool:
1325 - Most of the collective communications are now rooted in the same process as
1326 in the original application.
1327 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1328 used until now). Multiple data types can now be used in a trace.
1329 - The replay tool now supports traces produce either by TAU or a modified
1331 - Bug Fix: the compute part of the reduce action is now taken into account.
1332 - Gatherv collective is now supported
1333 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1334 of smpirun outputs time independent traces for the current run. One file
1335 is created per process. If too many processes are simulated, this behavior
1336 can be changed to one file for all processes by using the
1337 tracing/smpi/format/ti_one_file flag
1338 * smpirun generates the host file if needed (with given host count and platform)
1339 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1340 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1341 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1342 --cfg=smpi/coll_selector:(mpich/ompi)
1343 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1345 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1346 happening inside a collective SMPI call.
1347 * Fix the behavior of complex data types handling.
1348 * Make MPI_Wtime another synchronization point to take computations into
1350 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1351 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1352 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1353 (no privatization of global variables yet)
1354 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1357 * Verification of liveness properties is now available for SMPI applications
1358 (in addition to MSG applications)
1359 * Bugged examples using SMPI in examples/smpi/mc/
1360 * Add --cfg=model-check/visited option. Allows the verification of infinite
1361 programs. Detection of loops in the execution thanks to the system state
1362 comparison and reduction of the state space to explore. Can be combined with
1363 DPOR for safety properties.
1366 * Allow to change SimGrid configuration (see --help) within the code
1367 thanks to SD_config() as it can be done in MSG.
1368 * Add a new function SD_task_set_amount() upon user request.
1371 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1372 * Remove rule based routing (no more PCRE dependency)
1373 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1374 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1375 * Add a loopback_bw and loopback_lat options to cluster tag.
1376 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1377 routing. Now peers are encapsulated in an AS and have their own private
1378 router but this is transparent.
1381 * Our own implementation of getline is renamed xbt_getline, and gets
1382 used even if the OS provide a getline(). This should reduce the
1383 configuration complexity by using the same code on all platforms.
1384 * New type: xbt_cfg_elm_boolean.
1385 * Allow to use yes/no for boolean configuration options in the command line.
1386 * Allow to disable SimGrid cleanups at exit from command line option.
1387 There are situations where one may want a simulation to end with an exit.
1388 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1389 annoying when scripting around the simulator. Adding a
1390 --cfg=clean_atexit:no allows to circumvent this issue.
1393 * Lots of memory leaks were corrected in this release.
1394 There is no known memory leaks anymore, in all of our 600+ tests.
1395 * New command line option --version, to get SimGrid version information.
1396 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1398 * Supernovae builds are deprecated, and expected to be removed in the next
1401 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1403 SimGrid (3.9) stable; urgency=low
1405 The Grasgory release: GRAS is really dead now.
1407 * Complete overhaul of the internal host structures scheme.
1410 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1411 was considered as experimental and badly maintained since then.
1412 * Keeping it was thus a trap to our potential users, that could take
1413 it instead of MSG or SMPI by mistake despite is pity state.
1414 * GRAS seems to have very few users (if any), and no one volunteered
1415 to maintain it further. It also induces a lot of XBT code (for
1416 portability sake), that must be maintained too.
1417 * For all these reasons, we killed GRAS. If someone wants to revive it
1418 in the future, don't cry, our git history still remembers of GRAS.
1421 * Major overhaul. Merge our documentation again as time proved that
1422 splitting it was really not helping our users.
1423 * Further improve the developer documentation to help newcomers
1424 hacking on SimGrid itself. The user documentation (and in
1425 particular, the beginner documentation) is still in a sorry state.
1428 * Now works on Windows too!
1429 * Much more extensive test suite, from MPICH
1432 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1433 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1434 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1435 * Bug fix: let task be scheduled when the last dependency to be solved is
1436 a control dependency.
1437 * Remove SD_load_environment_script function.
1438 Use the C sg_platf function if you want to declare a platform
1442 * New function: MSG_process_get_number()
1443 * Old function documented: MSG_config()
1444 * Remove MSG_load_platform_script function
1445 Use the C sg_platf function if you want to declare a platform
1449 * Change the default value of the TCP_gamma constant (maximal size of TCP
1450 congestion window) to a more realistic 4MiB value. If you notice changes in
1451 your simulation results, you can fall back to the previous 20k tiny window
1452 by adding --cfg=network/TCP_gamma:20000 on command line.
1453 * (Hopefully) fix a bug wrt periodic availability and state traces
1454 * Bug fix: use default values at start when first event in availability/state
1455 trace is not at time 0.
1458 * remove the "new_" part of function name sg_platf_new_trace_connect
1459 (resulting in sg_platf_trace_connect), since it does not create
1463 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1464 use case where it's really mandatory, and maintaining it was a pain in
1466 * New: xbt_fifo_search(), search an item with a user-provided
1467 comparison function instead of dumb pointer comparison.
1470 * Fix the lua deployment:
1471 Use `simgrid.init_application()` before deployment instead of
1472 `simgrid.msg_register_application()` after.
1475 * Transfer the tracing files into the corresponding modules.
1477 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1479 SimGrid (3.8.1) stable; urgency=low
1481 The "we are told that some people want to also *install* the simgrid
1484 * Add missing file "tesh.1" to the archive.
1486 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1488 SimGrid (3.8) stable; urgency=low
1490 The Psssshiiiit release: SimGrid jumps into the Cloud.
1493 * Add an experimental interface to manipulate VMs. They are mainly
1494 process groups with very few intrinsic semantic, but they should
1495 allow you to build the semantic you want easily.
1496 * New function: MSG_host_set_property_value()
1497 * New function: MSG_process_on_exit(). To clean memory in all cases.
1498 * Bug fixes that made the host (and link) failures unusable.
1499 * Add a way to auto-restart process when the host in which they are
1500 executing comes back (ON_FAILURE="RESTART" on deployment file,
1501 MSG_process_auto_restart_set).
1502 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1503 please stop using the old ones, they are DEPRECATED.
1505 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1506 Please use MSG_init() instead. (reducing the amount of entry
1507 points in the library helps us).
1508 * Make it impossible to link against the wrong version of the lib
1509 * Deprecate MSG_clean(). No need to call it anymore.
1510 * Function MSG_get_host_number() is not deprecated anymore.
1513 * Split the doc into a user guide and a reference guide.
1514 * Start a developper guide to help people hacking on SimGrid.
1517 * Enable tracing by default. This modules rocks you should use it.
1518 * Remove option custom_flags. Now use environment variables CFLAGS
1520 * Use default cmake things to detect lua instead of home grown ones.
1521 * New option "enable_mallocators" to disable mallocators, for debugging
1522 purpose ("on" by default).
1525 * Bug fixes around the resource failures: don't let the processes
1526 survive the host they are running onto.
1527 * Add an interface to auto-restart processes when the host in which they are
1528 executing comes back.
1529 * Ensures that SIMIX_clean is called automatically. It's not part of
1530 the public interface anymore (bindings should be updated).
1533 * Bug fix for when SD_Simulate is called with a positive value: be careful
1534 when comparing doubles. Sometimes they are different for non significant
1536 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1537 parallel task whose initial work is distributed among host according
1538 to the Amdahl's law. Such tasks are created with a parameter alpha
1539 that corresponds to the non-parallelizable part of the computation.
1540 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1541 between two sets of workstations assuming a 1D block distribution (each
1542 workstation owns a similar share of data) on both sides.
1544 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1545 Data redistribution will be automatically scheduled once parent and child
1546 are both scheduled. The filling of computation_amount and
1547 communication_amount structures is now done seamlessly thanks to the chosen
1549 * New function SD_workstation_dump to display various information
1550 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1551 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1552 bandwidth on the route onto which the task is scheduled and the amount of
1554 To divide the nominal bandwidth by 2, the rate then has to be:
1555 rate = bandwidth/(2*amount)
1556 * Compute tasks that have failed can now be rescheduled and executed again
1557 (from their beginning)
1558 * Increasing source code coverage (src/simdag is now covered at 95.8%
1562 * Re-implement time-independent trace replay using SMPI (at the
1563 smpi_smp_* level) instead of MSG. This should replace
1564 examples/msg/actions/actions.c
1565 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1566 hindexed and structs)
1567 * Implement the exchange of non-contiguous data.
1568 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1569 * Correct behavior of smpi/sender_gap and set its default value to 0
1570 * Add option to asynchronously send small messages to allow better
1571 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1572 specifies the size in bytes under which messages will be asynchronously sent.
1573 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1574 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1577 * New C interface to define a platform: XML is now optional.
1578 For more info, please check include/simgrid/platf.h.
1579 * New interface to define random platforms from the C:
1580 For more info, please check include/simgrid/platf_generator.h and
1581 examples/msg/masterslave/masterslave_platfgen.c
1582 * Export a sg_cmdline dynar containing all the arguments we got from
1586 * Two new tracing options for adding comments to trace file so you
1587 can track your experiments (see --help-tracing for details).
1588 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1589 * Adding the SimGrid version that generated the trace file as a comment.
1590 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1591 * Fix to avoid key clashes on Paje links
1592 * Other minor fixes related to the Paje specification
1595 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1596 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1597 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1598 * Greatly improve the robustness of mmalloc to user errors (such as
1599 using an area after freeing it, or freeing it twice)
1601 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1603 SimGrid-java (3.8.1) stable; urgency=low
1605 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1606 Warning: all methods are visible, but only some of them are
1607 implemented so far. Check the source (src/jtrace.c)
1608 for further information.
1609 * New module: org.simgrid.msg.File (SimGrid File management functions)
1610 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1611 * Change the meaning of Process.restart: now restart the process from
1612 the begining, like MSG_process_restart in C.
1613 * Add Process.setAutoRestart: handling of process restart when failed
1615 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1616 you to retrieve the properties of the processes/hosts
1617 * Deprecate Msg.clean(): you can just forget about it now.
1618 * New function Process.getCount(), that only works when compiling
1619 with the not yet released version 3.9 of the C library.
1621 * New context factory based on Coroutines. It mandates a modified JVM
1622 but then, the simulations run about five times faster, and there is
1623 no limit to the amount of processes (beside of the available memory).
1625 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1627 SimGrid (3.7.1) stable; urgency=low
1630 * Restore the prototype of MSG_process_create_with_environment() to
1631 the pre-3.7 situation by removing the kill_time argument.
1632 * Add a MSG_process_set_kill_time() function instead.
1635 * Fix weird behaviors when dealing with parallel tasks.
1638 * Simgrid is now built as a dll.
1639 * Simgrid-java now works on Windows.
1640 * Simgrid-Java is now included into Windows package.
1643 * First pre-build package for MacOSX.
1646 * Fix compilation when using MSG_USE_DEPRECATED.
1647 * Fix some compilation issues on Macs and Windows.
1648 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1649 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1651 We discovered that the Lua console is broken, but we are missing the
1652 manpower to fix it right now. The problem existed in 3.7 too, so we
1653 are not blocking the release for that. Sorry if you depended on this
1654 feature, any help would be really welcome.
1656 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1659 SimGrid-java (3.7.1) stable; urgency=low
1661 The "Java aint got to be bloated and slow" release
1664 * Various internal cleanups and performance improvement
1665 Simulations are expected to run up to twice faster or so
1666 * Make Process.kill(process) an instance method, not a static one
1667 * User processes are not java.lang.Thread subclasses.
1668 This breaks the compatibility (sorry), but previous API was
1669 brain-dead, making it impossible to have non-trivial
1670 initializations in the process constructor.
1671 * Require a full constructor per Process sub-class.
1672 Kinda breaks the compatibility (sorry), but this allows a much more
1673 efficient way to launch the processes at simulation startup.
1674 * Do not embeed our version of semaphores, java 1.5 can be considered
1675 as sufficiently prevalent for us to not dupplicate its features.
1679 * Add examples for almost every part of the API
1680 We spotted and fixed a lot of bugs in the process
1682 * New module: asynchronous communication API
1683 * New function: Process.sleep()
1684 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1685 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1686 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1688 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1690 SimGrid (3.7) stable; urgency=low
1692 The "spring cleanups (before the next Big Project kicks in)" release.
1695 * We can specify the SMPI latency/bandwidth factor with command line
1696 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1697 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1698 You can also use the "config tag" from platform file by adding this line
1699 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1700 (see "example/platforms/tag_config.xml" to use "config tag").
1701 Note that the command line supersedes the platform file configuration.
1702 * Change the correction factors used in LMM model, according to
1703 the latest experiments described in INRIA RR-7821.
1704 Accuracy should be improved this way.
1705 * Use the partial invalidation optimization by default for the
1706 network too. Should produce the exact same results, only faster.
1707 * Major cleanup in surf to merge models and split some optimization
1708 mechanisms from the core of the models. As a result you can now
1709 specify which model to use (e.g., --cfg=network/model:LV08
1710 --cfg=cpu/model:Cas01) and which optimization mode to use
1711 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1712 Incompatible combinations should err at initialization. See
1713 --help-models for the list of all models and optimization modes.
1714 * The CLM03 workstation model was dropped for simplicity because it
1715 used the deprecated CM02 network model. Use default instead.
1716 * Rename the TCP_gamma configuration option to network/TCP_gamma
1717 * Rename the coordinates configuration option to
1718 network/coordinates, and document it
1719 * Use now crosstraffic keyword instead of the terribly misleading
1720 fullduplex keyword. It is activated by default now in the current
1721 default model, use --cfg=network/crosstraffic:0 to turn it off.
1722 * Ongoing refactoring the model parsing to make XML files optional
1723 See include/simgrid/platf.h for details (still to be completed)
1726 * Major overhaul of the documentation. Almost instructive now :/
1727 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1728 functions and friends. This interface was considered as
1729 deprecated since over 2 years, it's time to inform our users that it is.
1730 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1731 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1732 use these (crappy) functions in your code.
1733 These functions will be removed soon. Stop using them now.
1734 * Deprecate MSG_get_host_{table,number}
1735 Implement MSG_hosts_as_dynar() instead.
1736 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1737 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1740 * Stabilize the parallel execution mode of user contexts
1741 * Introduce configuration variables to control parallel execution:
1742 - contexts/synchro: Synchronization mode to use when running
1743 contexts in parallel (either futex, posix or busy_wait)
1744 - contexts/parallel_threshold: Minimal number of user contexts
1745 that must be part of a scheduling round to switch to parallel
1746 execution mode (raw contexts only)
1747 * Fix bugs that prevented to use suspend/resume along with
1748 synchronization structures.
1749 * Fix bugs in process termination that lead to invalid memory access
1750 in very specific conditions.
1753 * Introduce a parallel mode for the models (controlled by surf/nthreads
1754 configuration item). In our tests, running the models in parallel
1755 never lead to any speedups because they are so fast that the gain
1756 of computing each model in parallel does not amortizes the
1757 synchronization costs, even when ultra fast futexes are used.
1758 This is released anyway because YMMV.
1761 * Performance boost by using a swag internally to compute the set of
1762 tasks that are finished and should constitute the return value of
1766 * Enable it by default now that it is considered rather stable.
1769 * Documentation of the tracing functions.
1770 * Performance gains when tracing categorized/uncategorized resource
1771 utilization by avoiding calls to get route when updating resource
1772 variables. LMM constraints are being used instead.
1773 * API changed to set task categories. Use MSG_task_set_category instead
1774 of TRACE_msg_set_task_category, and SD_task_set_category instead
1775 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1776 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1777 * Examples for link user variables added (see at examples/msg/tracing/)
1778 * Deprecated function TRACE_msg_set_process_category completely removed
1779 * Trace header updated according to the latest Paje file format
1780 * Tracing network lazy updates, no longer obligate users to use full updates
1781 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1782 * Experimental: let user code declare/set/push/pop application states for hosts
1783 * API changed to allow the manual creation of graph configuration files
1784 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1787 * Improve the API of Lua MSG bindings, using the Lua spirit.
1788 * Each simulated process now lives in its own Lua world (globals are
1789 automatically duplicated). It helps writing simulators. Will allow
1790 to run Splay programs within SimGrid in the future.
1791 * Add a Chord example in Lua, equivalent to the MSG one.
1794 * Start the implementation of a solution to express temporal
1795 properties, not only local assertions. This is still an
1796 experimental work in progress, stay clear from it to be safe.
1800 - Add new runtime parameters --help-logs and --help-log-categories
1801 to display information about supported logging parameters and
1803 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1804 don't exists anymore.
1805 * Mallocators: allow value NULL for the reset function.
1807 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1808 create homogeneous dictionaries, where all the elements share the
1809 same free function. Non homogeneous dictionaries will be
1810 deprecated in the next release.
1811 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1812 - Multi-level dictionaries are deprecated.
1814 - new function xbt_dynar_search_or_negative() that is useful when
1815 you have less than 2 million elements in your dynar and don't
1816 want of the extra complexity of catching exceptions when the
1817 element is not found.
1819 - Make xbt_os_thread module (for thread portability) public.
1820 Documentation is still to come, sorry.
1822 - Cleanups and simplifications to make it maintainable again.
1823 - Exotic features (such as memalign and valloc) were removed.
1824 - The metadata were extended and improved so that the
1825 model-checker becomes able to explore and inspect the heaps.
1826 - This may induce a performance drop when enable_model-checking is
1827 ON in cmake (even if it's not used in the simulation), but it is
1828 necessary at this point to get MC working.
1830 Turn model-checking OFF if simulation performance matters to you.
1831 Not enabling it at runtime is not enough, disable it in cmake.
1833 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1836 SimGrid (3.6.2) stable; urgency=low
1838 The "Not coding new stuff allows to polish old things" release.
1841 * New bindings to the NS3 packet level simulator (experimental)
1842 * Use the raw (efficient) execution contexts instead of the sysv
1843 (portable) ones when possible.
1844 * libpcre is now mandatory in any cases since not using it led to
1845 severe performance loss and possibly other issues
1846 * Update the XML platforms:
1847 - G5K: include the latest machine in Nancy
1848 - GridPP and LCG: new platforms
1849 * Documentation was partially updated, at least (more to come)
1851 Bug fixes, cosmetics and small improvements
1852 * Free terminated processes before the end of the simulation to avoid
1853 exhausting the memory of users having very dynamic amount of
1855 * Bug fix and cosmetics about canceling non-running tasks
1856 * Bug fix about the dot loader's issues when using libcgraph
1859 * Create an installer for windows with nsis (amd64 and win32)
1860 - Add an hello world project to illustrate simgrid project creation.
1861 - Embed libpcre into the Simgrid installer to avoid
1862 its compilation burden
1863 * The raw execution contexts should work on Apple now
1864 * Port to Windows 64 bits
1865 - Sysv contexts now have an implementation for this arch
1866 - GRAS communication features now support this arch
1867 * Drop support for borland compiler on windows
1868 - this code was not maintained, and we kinda depend on gcc nowadays
1869 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1870 * Fix portability issue on unstable ubuntu: linker became picky on
1873 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1875 SimGrid (3.6.1) stable; urgency=low
1877 The "Oops, we broke Macs too" release
1880 * Fixed contexts detection so that raw ones are used when possible
1881 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1882 leads to a strange error, with user code segfaulting sometimes when
1883 the generated code is not perfectly aligned (which is not
1884 controllable from the user side, depends on the amount of code)
1887 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1889 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1891 SimGrid-java (3.6) unstable; urgency=low
1894 * Split of every thing from simgrid v3.5 into a separate package.
1896 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1898 SimGrid (3.6) stable; urgency=medium
1900 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1903 * Bindings now constitute their own package, separated from the main one.
1904 Rationale: reduce our maintenance nightmare by reducing the module coupling
1905 They will soon be released on their own on gforge.
1907 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1908 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1910 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1911 * It's not quite deprecated for now because we have no replacement,
1912 but it may soon become the case.
1915 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1916 * Fortran: New user-level cache variable to store the rank of the running
1917 process. This improves performance by an order of magnitude.
1918 * C: New coccinelle script to automatically locate and modify global and
1919 local static variables.
1920 * Improved SMPI network model with a sender-side gap to account for multiple
1924 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1925 only say if a communication is finished, no matter whether it succeeded or
1926 failed. You can call MSG_comm_get_status() to know the status of a finished
1928 * New function MSG_task_dsend() to send a task and detach it. When a
1929 communication is detached, you are never notified of its success or failure
1930 and the memory is released automatically once it is finished. This function
1931 is useful when you don't care about the end nor the success of a
1933 * Change the prototypes of action replay. Sorry for inconvenience,
1934 but this is really more efficient this way (and to adapt your code,
1935 you just have to fix the initialization, that shouldn't be too long)
1936 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1937 ever managed to do anything useful with it.
1938 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1939 takes a dynar of communications. It returns immediately and gives the
1940 index of a finished communication (if any).
1941 * New example: a basic implementation of the Chord P2P algorithm.
1944 * New model for multi-core CPUs. You can now use the core attribute to
1945 precise the number of cores of a host. This is a basic model. Every
1946 process running on the host receives at most the power provided in
1947 the DTD (throughput<=power). Total throughput of process cannot exceed
1949 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1950 router linked by an up-link and a down-link (possibly asymmetrical).
1951 This kind of pattern allows to easily build last-mile model style platforms.
1952 Aggregating such patterns in a rule-based AS is thus the technique of
1953 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1954 * New model for Vivaldi routing. We transformed the Vivaldi network model
1955 into a Vivaldi routing model (based on the rule-based model). This allows to
1956 combine Vivaldi based latencies with last-mile platforms.
1959 * Added a check for NaN of IEEE754 infinite in the double entries of
1961 * Introduce a new context factory "raw", highly inspirated from the
1962 ucontext factory, but using manually crafted functions in assembly to
1963 do the work in an efficient manner.
1964 * Allow to change the used context factory at run time, not only at
1965 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1966 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1967 contexts at runtime (only with raw contexts or ucontexts).
1968 * Completely rewrote this module to allow parallel execution of user
1969 processes. Use --cfg=contexts/nthreads:N to execute user processes
1970 with N parallel threads (the default is 1, meaning no parallelism).
1971 * Allow to decide dynamically between sequential and parallel modes.
1972 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1973 processes in parallel only when their number is greater than or equal to P
1975 * Added a check for NaN of IEEE754 infinite in the double entries of
1979 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1980 won't output the state of processes when interrupted with Ctrl-C
1981 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1982 NULL-terminated array. This may solve backward compatibility issues
1983 due to the change to return type of SD_simulate. See also:
1984 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1985 * Add new macros with variable number of arguments.
1986 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1987 - in xbt/asserts.h: xbt_assert
1988 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1989 - in xbt/ex.h: THROWF and RETHROWF.
1990 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1992 * Change xbt_die() to accept a format string with arguments, just like printf.
1993 * New data structure: xbt_lib_t, like a dict but more general and with better
1997 * New configuration options
1998 Options triva/categorized and triva/uncategorized can be used to generate
1999 graph configuration files for Triva visualization tool.
2000 * Configuration option tracing/platform is renamed to tracing/categorized
2001 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2002 * New TRACE_user_link_variable interface:
2003 User provides the name of the link and the tracing variable to attach to it
2004 * the declaration of tracing categories must be done after the environment
2006 * simpler tracing interface, just one way to declare categories
2007 TRACE_category or TRACE_category_with_color, it is up to you
2008 * links in the trace file are again identified by their names
2009 * trace contains the full platform hierarchy exactly as declared using the ASes
2010 * Options tracing/msg/[task|process]:1 groups the process by hosts
2011 for both cases, tasks and processes must have names that are unique during the simulation
2012 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2013 * The experimental option tracing/msg/volume is deprecated
2014 its functionality may be reincorporated if needed
2016 The tracing generates a trace file with unordered timestamped events,
2017 because of the way the core simulator (surf) works. A script available
2018 at the tools directory (fix-paje-trace.sh) can be used to put the events
2019 in order. We have changed the tracing so it can generate ordered timestamped
2020 events in the final trace, but depending on the simulator (and how much time
2021 is simulated) that can lead to a huge memory utilization. It is deactivated
2022 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2024 Build Infrastructure
2025 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2026 - We are trying hard to keep the API stable, but it may happen that
2027 some things change (we're a research project after all, not a
2028 nuclear plant operating system). If such things should happen, you
2029 could rely on that macro to adapt.
2030 - current value: 30600 for 3.06.00, aka 3.6
2031 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2032 a number that can be compared with SIMGRID_VERSION.
2033 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2034 assertions and verbose/debug logging events are disabled at compile time.
2036 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2038 SimGrid (3.5) stable; urgency=medium
2041 * New feature to any SimGrid-based simulator: Model-Checking
2042 Check SIN#1 for more details.
2045 * New Model SMPI (three-interval linear regression for correction factors)
2046 See RR-7426, available at http://hal.inria.fr/inria-00527150
2047 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2048 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2049 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2050 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2051 * See also src/smpi/README
2056 - Tracing API changes: TRACE_start and TRACE_end should not be called
2057 by user-code. They are automatically called by simulators created
2058 with SimDAG, MSG and SMPI if the toolkit is compiled with
2059 tracing_enabled=ON. Categories declaration and utilization remain the
2060 same for MSG and SimDag.
2061 - A function was added to the tracing API to declare categories with
2063 - TRACE_category_with_color (char *category, char *color)
2064 where color must be in the following format
2065 "%f %f %f", red, green, blue
2066 and red, green, blue are float values in the interval [0, 1]
2067 - User can specify NULL as color parameter, or continue calling
2068 TRACE_category (cat)
2069 On that case, the tracing system will define random colors
2070 - The following command-line options are supported:
2071 --cfg=tracing/filename:msg.trace
2072 --cfg=tracing:1 (activate tracing, needed to use others)
2073 --cfg=tracing/platform:1 (categorized resource use)
2074 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2075 --cfg=tracing/msg/task:1 (task creation)
2076 --cfg=tracing/msg/process:1 (process creation, migration)
2077 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2078 --cfg=tracing/smpi:1 (SMPI interface tracing)
2079 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2080 - examples of examples/msg/tracing updated
2082 - DAXLoader and DOTLoader functions can generate tasks with categories
2083 - A new function to attribute a category to SD tasks:
2084 TRACE_sd_set_task_category (SD_task_t task, char *category)
2085 * Tracing the MPI interface implemented by SMPI
2086 - Collective operations are traced with states
2087 - Point-to-Point operations are traced with states/links
2088 - Tracing activated by a parameter "-trace filename" passed
2089 to smpirun during execution (considering that simgrid
2090 is compiled with tracing enabled)
2091 - To run the simulation with gdb, the simulator
2092 accepts --cfg=tracing/smpi:1 to trace SMPI
2093 - tesh files to check if smpi tracing is ok
2094 - See examples/smpi/NAS/DT-trace
2095 * GTNetS tracing re-worked
2096 - adaptation to the tracing system of GTNets to cope
2097 with modifications regarding the fullduplex mode
2098 - new tesh files to check if gtnets tracing is ok
2101 * Asynchronous communications through the functions:
2102 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2103 * New function: MSG_load_platform_script()
2104 to make possible using a lua script instead of XML files to set up platforms
2105 * New function: MSG_set_function
2106 to associate functions to processes, used when bypassing the parser
2107 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2109 Platforms: Add some more examples in examples/platforms
2110 * Grid'5000: see www.grid5000.fr
2111 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2114 * Change the XML format. This is a very important modification. SimGrid
2115 3.5 introduces a new hierarchical format based on the notion of
2116 Autonomous Systems. Compatibility with old format is ensured
2117 through the perl script provided in the install bin directory
2118 bin/simgrid_update_xml.
2119 It is now possible to build platforms with specific routing
2120 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2121 connect several platforms together. We will try to provide soon
2122 set of realistic platforms exploiting these properties (have a
2123 look at examples/platforms/ for the moment).
2124 * Take the opportunity of the XML format change to be a good XML citizen:
2125 rename link:ctn to link_ctn and similar changes (also dealt with by
2127 * Add a new routing scheme (rule-based) using regular expressions. It
2128 enables to have an extremely low memory footprint when the
2129 underlying routing is simple and can be compactly described. You
2130 need to have libpcre4-dev (perl regular expressions) installed if
2131 you want to use this routing scheme.
2132 * Revive the cluster TAG and allow to easily and efficiently (both in
2133 term of memory and speed) connect clusters together. Have a look
2134 at teshsuite/simdag/platforms/ to see how this can be done. With
2135 this tag, you can create clusters with thousands of tasks at no
2136 cost (have a look at examples/platforms/).
2137 Note: clusters are implemented as ASes, so there is no need for an
2138 enclosing AS tag if you have only one cluster in your platform.
2139 * Add new generic functions in the public interface that allows the user
2140 to call SURF 'create_resource' methods from your code (same
2141 functionality as the XML bypass mechanism but with a much lighter
2143 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2144 that uses a piecewise linear approximation to produce better
2145 results when exchanging small messages.
2146 * Add a new parameter to handle correctly full duplex link and account
2147 for interferences between uplink and downlink communications
2148 (activate with --cfg=fullduplex:1).
2151 * Rename the SD_READY (all dependencies are satisfied and task is
2152 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2153 dependencies are satisfied) state.
2154 This prevents a confusion between the notion of "ready to schedule"
2155 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2156 simulated" (SD_RUNNABLE) used by the simulation kernel.
2157 * Change the way a task is considered as ready. Instead of removing
2158 dependencies when a task is done, a counter is decreased. This way,
2159 it is always possible to reach ancestors thanks to the
2160 SD_taks_get_parents function (even after the end of the simulation.)
2161 * Change the return type of SD_Simulate from (SD_task_t*) into
2162 xbt_dynar_t. This function was in handling a dynar internally and
2163 converted it into a NULL terminated array for historical reasons.
2164 * New function SD_dotload(char*) to load a DAG described in dot
2165 format. This loader and the corresponding examples require the
2166 installation of the graphviz library.
2167 * Fix a bug in the management of tasks of size 0 in the surf network
2168 models. This problem was only visible with SIMDAG and you should
2169 thus disregard results produced with earlier versions if you
2170 relied on this feature (some tasks were blocked because of this).
2171 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2172 even though your applications were doing only point-to-point
2173 communications and sequential computations. Now you can really use any
2174 model you want (of course, if you create real parallel tasks, which are
2175 not implemented in most models beside ptaskL07, this will abort).
2176 * Add an example that schedules a DAX on an heterogeneous platform
2177 using a Min-Min strategy.
2178 * New function SD_workstation_get_current_task() that returns the kind
2179 of task currently running on a workstation in the sequential access
2181 * Raise some warnings when unexecuted tasks remains at the end of the
2182 simulation. This is usually caused by cycles in the DAG.
2185 * New function: SIMIX_process_set_function() called by MSG_set_function
2186 * Change the underlying waiting queue in semaphores so that a process
2187 can wait on several of them simultaneously (as in waitany).
2188 * Fix the way to handle tokens in semaphores so that all access patterns
2189 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2190 * kill the dirty pimple SIMIX_message_sizes_output()
2191 Please use (proper) visualization instead
2194 * New data container: setset (set of sets of elements)
2195 * New module: mmalloc (mapped malloc, allowing to have several
2196 independent segments of malloc)
2197 * New function: xbt_dict_cursor_set_data()
2198 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2199 * New function: xbt_dynar_is_empty()
2200 * New function: xbt_fifo_get_last_item()
2201 * Fix xbt_dynar_shrink(): use the right element size.
2202 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2203 uninitialized areas during expand.
2204 * Fix semaphores: previous implementation was severely broken.
2205 * Use library init/fini functions for our initialization.
2206 - you can use logs and other feature as soon as you want in your
2207 code (even before the xbt_init / MSG_init)
2208 - xbt_exit is now a no-op and produce a warning when used.
2211 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2212 benefit from the latest improvement to the simulation kernel.
2213 * Kill measurement sockets for now. If you rely on them, sorry. This
2214 release is not for you. This feature will be reintroduced in the
2215 future, but we cannot delay the release any further.
2216 * New function: gras_msgtype_get_name().
2217 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2218 * Fix (at last) the pmm example: it should not randomly fail anymore.
2220 Build chain: bug fixes and overall polishing
2221 * Cmake is now stable enough. Hence, we killed the autotools.
2222 * Port to windows ( TM :)
2223 * Fix the 'make install' target.
2224 No need to use 'make install-simgrid' anymore
2225 * Introduce a 'make dist' target compiling a *source* archive
2226 'make package' compiles a binary archive
2227 * Compile java files only on need
2228 * Add --cd and --setenv command line options to tesh
2229 * Out of source builds are not fully supported yet, but we are close
2230 * Enable supernovae and optimization flags by default for our users
2233 * Add layer to set up environment directly from lua, without XML.
2234 * The effect of gras_stub_generator can be achieved through
2235 lua too (check examples/gras/console/ping_generator.lua)
2237 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2239 SimGrid (3.4.1) stable; urgency=low
2241 The "Polishing easter eggs is probably a good idea" release.
2242 This is a bug fixes release only.
2246 * Fix a bug preventing the tasks from begin garbage collected.
2249 * Fix a bug occuring when a host involved in a communication fails.
2250 This was not detected properly by the other peer involved in the
2251 communication. Now, it's reported as a network error.
2254 * Warn the user about loop dependencies in data flow of DAX files
2255 * Obey the control-flow dependencies of DAX files
2258 * Add option "enable_smpi" allowing to not compile SMPI.
2259 Probably useful for the (Mac) users experiencing a build error here
2260 * Improve the detection of lua5.1 and ruby1.8
2262 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2264 SimGrid (3.4) stable; urgency=low
2266 The "Easter in Cargese" release. Also known as (major changes):
2268 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2269 ~> bindings were greatly improved
2270 ~> new tracing infrastructure for better visualization introduced
2272 * the "Welcome to configury modernity" release.
2273 ~> we switched from autotools to cmake, and improved our cdash
2276 A more detailled list of changes follow (full detail in svn log).
2278 Java Bindings: Various Cleanups
2279 * (install java-gcj-compat-dev on debian-like to use them)
2280 * Remove put/get: no need to export deprecated interface in Java
2281 Use send/receive instead.
2282 * Cleanup the examples and add a README per directory
2283 * Remove example autoDestination (that's the only way to go now)
2284 * Remove example explicitDestination (was a plain copy of basic)
2285 * Make JniException a runtime exception, so that there is no need to
2286 declare the fact that you may encounter such a beast. I guess that
2287 nobody will ever want to survive such error.
2288 * Create specific errors for each MSG case of failure:
2289 host failure, transfer failure, timeout, task cancelled
2290 * Cleanup the exceptions that may get thrown by each function
2291 * Other internal cleanups in Java bindings. Performance still bad :/
2292 Ruby and Lua Bindings: create them
2293 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2294 * That's new and great, you should try them out.
2295 Same functionalities than Java bindings, only even less polished
2297 * Kill the useless "rate" argument of SD_task_get_execution_time()
2298 Everyone used to provide -1 as a value, it was not used, and the
2299 semantic of a possible use wasn't even clear.
2300 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2301 either as comm costs or compute costs to mean that there is no
2302 such thing for that specific task.
2303 * Add a SD_task_set_name() function
2304 * Fix SD_task_unschedule() on typed tasks
2305 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2306 * In DAX loader, accept useless 'level' attributes to <job> since
2307 LIGO DAGs have them (seem to be to ease graphical representation).
2309 * Add an example masterslave_mailbox.c using send/receive and not
2310 the deprecated put/get interface.
2311 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2312 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2314 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2315 (declare MSG_USE_DEPRECATED to still have the old name)
2316 * Add a MSG_task_set_data() function
2317 * About trace replay (see examples/msg/actions):
2319 - Allow to work with splitted trace files for each process
2320 Give the specific trace file as argument of each process,
2321 and call MSG_action_trace_run(NULL)
2322 You can still have one merged file for all processes.
2323 - Fix implementation of collective operations
2324 * Allow task_execute() on 0-sized tasks (closes #10063)
2326 * This is the first release of SimGrid where SMPI is not considered
2327 beta anymore (even if some corners should still be improved)
2328 * Port over the new SIMIX_network submodule (internal refactoring)
2329 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2330 * Implement more missing elements of the standard:
2332 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2334 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2335 MPI_Scan() MPI_Get_processor_name()
2336 - Added implementation of missing case for Alltoall (warning: it's
2337 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2338 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2339 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2340 SMPI_MPI_Allgatherv()
2341 * Bug fixes include:
2342 - MPI_Waitsome() was broken
2343 - Allow relative includes in smpicc
2344 - Command line cfg argument 'reference_speed' was ignored...
2345 - Some functions did not properly lead to auto-benching of user code
2346 - smpicc passes -O2 by default (just like openmpi one)
2348 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2349 * Bug fixes about timeouts during communications
2350 * add SIMIX_message_sizes_output() as a pimple to write to file the
2351 amount of messages per size. Use gnuplot to get histogram.
2352 Pimple because that's the only user-visible function of simix,
2353 defined directly in xbt.h (irk, sorry)
2355 - Add a SIMIX_sem_get_capacity() function
2356 - Fix interactions with processe resume/suspende
2357 - release_forever() was stupidly broken
2358 - Fix SIMIX_display_process_status() for processes in a semaphore
2359 - Make SIMIX_sem_block_onto() user-visible
2360 * Refactoring context stuff:
2361 - Use pseudo-OOP for better modularity
2362 - reimplement SIMIX_process_kill() without process_schedule() so
2363 that the latter can take as invariant that it is called from
2365 - Merge context_start into context_new for sake of simplicity
2367 * Add a Vivaldi network model, coded live during SUD'10 ;)
2368 * Rename configuration variables to start a hierarchy:
2369 o cpu_model -> cpu/model
2370 o network_model -> network/model
2371 o workstation_model -> workstation/model
2372 * New configuration variables:
2373 o network/bandwidth_factor: correction to bandwith
2374 o network/latency_factor: correction to latency
2375 o netwotk/weight_S: correction to the weight of competing streams
2376 * Add a long description to the models, that users can see with such
2377 argument on the command line: --cfg=cpu/model:help
2378 * --help-models display the long description of all known models
2380 * config: add the ability to set a default value after registration
2381 Does not override any previously set value (e.g. from cmd line)
2382 * dict: allow to have integer key and data.
2383 When so, you need to use the following functions
2384 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2385 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2386 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2387 In contrary to regular dicts, the key is not malloced before copy.
2388 Mixing scalar and regular elements in the same dict is not tested
2390 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2391 Tracing for Visualization:
2392 * SimGrid is now instrumented in order to generate a trace file for
2393 visualization analysis: to use it, need to compile SimGrid with the
2394 "tracing" option enabled, and instrument the program using SimGrid with
2395 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2396 (among other functions).
2397 * The instrumentation only traces the platform utilization for now
2398 * Documentation to use the tracing functions and how to analyze the
2399 traces with the Triva tool is written.
2400 * More information about: SimGrid FAQ (in the section Tracing Simulations
2403 * We moved to cmake as default build system. Autotools support will
2404 be dropped soon. Check the FAQ for more info about how to use it.
2405 * Greatly improved our cdash/ctest interactions
2406 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2407 * Added memory checking tests with valgrind; lot of memleak fixing.
2408 This may be the first release of simgrid with so few memory issues
2409 * Added code coverage tests.
2410 Our coverage is still improvable, but at least we see it on cdash.
2412 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2414 SimGrid (3.3.4) stable; urgency=low
2416 The "Desktop Grid needs love too" release (also called Xmas release).
2418 Models improvements:
2419 * Major speedup in the maxmin system solving by using lazy evaluation
2420 Instead of solving completely the maxmin system at each iteration,
2421 only invalidate (and recompute) the modified parts.
2422 This new feature is enabled in default models but you can try to
2423 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2424 * Cas01 IMproved as default CPU model
2425 This CPU model is the same Cas01 model, but it uses the
2426 maxmin-selective-update flag and a heap structure to manage
2427 actions on SURF kernel.
2428 It reduces the complexity to find the next action to finish and,
2429 consequently, it's faster than the old Cas01.
2430 This is the new default CPU model (Cas01).
2431 * Rename the old Cas01 model to Cas01_fullupdate
2432 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2433 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2434 * CpuTI (CPU Trace Integration)
2435 A new CPU model whose objective is simulate faster when using
2436 availability trace files.
2437 Instead of using a full featured, over engineered maxmin system for
2438 CPU modeling, this model does the pre-integration of traces files
2439 to calculate the amount of CPU power available, and so, executes
2440 faster than the old CPU models.
2441 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2442 * Use LV08 as default network model since it gives better accuracy
2443 for small messages and shouldn't change things for big ones.
2444 Use --cfg=network_model:CM02 to get the previous behavior.
2447 ******************************************
2448 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2449 ******************************************
2450 * The new CPU model may changes simulations!
2451 The point is that events occurring at the exact same timestamp
2452 are not scheduled in the same order with the old and new
2453 version. This may be enough to completely change the execution
2454 of simulations in some cases.
2455 * The new network model will change simulations!
2456 This new model is more realistic than the previous one, so you
2457 should consider redoing your old experiments with this model.
2458 Sorry for the inconvenience.
2461 * Introduce the supernovae compilation mode
2462 When compiled that way, the whole SimGrid (or almost) is put in a
2463 single compilation unit and compiled in one shoot.
2464 This is to help gcc which has difficulties to inline stuff from one
2466 The speedup seem to be above 15%, althrough more tests are needed on
2467 amd64 to confirm that gain.
2470 * Port of MSG's mailbox on top of SIMIX network
2471 The put/get mechanism was greatly simplified on the way.
2474 * New SIMIX network module. Provides:
2475 - Mailbox: rendez-vous mechanism to find with who you want to speak
2476 - Synchronous send/recv: easier and hopefully faster since the
2477 logic is handled in the maestro process directly now
2478 - Asynchronous send/recv: you dreamt of it? It's here now
2479 Too bad that nobody cared enough to propagate the change to MSG.
2480 * Add semaphores as SIMIX synchronization mechanism.
2483 * new function SD_daxload(char*) to load a DAX file
2484 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2485 * Introduce typed tasks. Specify its kind and cost at creation.
2486 At scheduling, just give where it should be placed, and the cost
2487 for each involved resource is automatically computed.
2488 Existing constructors so far (more to come of course):
2489 - SD_task_create_comm_e2e() for end-to-end communication
2490 - SD_task_create_comp_seq() for sequential computation
2491 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2492 * new function SD_task_dump() for debuging display
2493 * new function SD_task_dotty(task,FILE*) writing to file the info
2494 about the task in dotty format
2495 * SD_task_dependency_exists() can now cope with having one of its
2496 arguments NULL. If so, it tests whether the other argument has any
2498 * Add getters on list of preceding/following tasks:
2499 SD_task_get_parents(task) and SD_task_get_children(task)
2500 * Add getters on amount of workstations and list:
2501 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2502 * Add getter on task kind: SD_task_get_kind(task)
2503 * Update the start_time and finish_time of tasks on completion/failure
2504 * Bugfix: Remove task from state swags when destroyed
2507 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2510 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2511 dynar with one separate thread per value of the dynar.
2512 * Change the prototype of xbt_thread_create(), sorry.
2513 Added a boolean parameter indicating whether we want to join this
2514 thread (used in SG only for now)
2515 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2518 * GTNetS wrappers should now be usable again (and betterly tested too)
2519 * Fix a major regression from 3.2 where the timeout provided to
2520 MSG_task_put_with_timeout() was used as absolute time before which
2521 the comm should be done.
2522 * Start to fix the <cluster> tag.
2523 - Internal links should be good now (beside of the loopback, which
2524 use the private link instead)
2525 - paths to the external world is still rather broken
2526 - the <route:multi> tag is just broken. Actually that's brain-dead.
2527 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2528 to make it less stupid
2529 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2530 * Fix a source-level compatibility glitch from 3.2: after defining
2531 MSG_USE_DEPRECATED, you can use the old name
2532 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2533 * Allow to compile from the SVN with automake 1.11
2534 * Fix some problems when using the "start_time" tag in deployment XMLs.
2535 * Fix #8569: XBT/synchro.h has redundant declarations
2536 * Fix #8563: MSG return values and exceptions
2537 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2538 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2539 Thanks to Robson Peixoto.
2540 * Fix a vicious bug in dictionaries inducing that some elements were
2541 not freed on xbt_dict_free()
2543 Portability report of this version:
2544 * Main portability targets:
2545 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2546 on (amd64/i386/ia64)
2547 - mac leopard on i386
2548 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2549 but nothing critical.
2550 * Other platforms: windows, AIX and others were not tested for this release
2552 Timing report of this version:
2553 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2554 scenario parameters). From 8h to a few seconds in desktop grid settings.
2555 * Supernovae brings about 25% speedup on i386.
2557 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2559 SimGrid (3.3.3) stable; urgency=low
2561 The "Need for Speed" release.
2563 The timings done to validate the 3.3.2 were faulty.
2564 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2566 The problem was a conversion from a manually handled vector to
2567 xbt_dynar_t on the critical path.
2568 xbt_dynar_foreach calls functions, inducing stack management crap.
2570 We inlined these functions and xbt_dynar_foreach is now breath taking.
2571 We also inlined xbt_swag_belong on the way.
2573 Here are some approximate speedup measurements (on master/slaves
2574 simulations lasting between 10s and 20s each):
2575 3.3.1 -> 3.3.2: about same performance
2576 3.3.2 -> 3.3.3: 40% speedup
2577 3.3.1 -> 3.3.3: 40% speedup
2578 3.3.1 with inline patch -> 3.3.3: 30% speedup
2580 Our reading is that the refactoring which occurred in 3.3.2 made us
2581 suffer much more from the xbt_dynar_foreach low performance, but
2582 once we solved this, this refactoring proved to be very performance
2583 effective. From the 40% speedup, somehow, 10% are due to the
2584 inlining and 30% to the refactoring.
2586 That's a pitty that gcc cannot inline functions placed in other files
2587 alone. We have to choose between:
2588 - break the encapsulation (by putting private data structures and
2589 accessors in headers files to help gcc)
2590 - live with low performance
2591 - switch to a decent compiler such as icc (not quite possible).
2593 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2595 SimGrid (3.3.2) stable; urgency=low
2597 The "Simplicity does not preceed complexity, but follows it" release.
2599 The main contributors of this release were (lexical order):
2600 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2603 * Extract the routing logic into its own object.
2604 (was dupplicated in network.c and workstation_LV07.c;
2605 Allows to implement other ways of storing that info)
2606 => kill now useless network_card concept
2607 - Use dynar to represent routes (instead of void** + int*)
2608 - kill link_set (use surf_network_model->resource_set instead)
2609 - Add a command-line option to choose the routing schema to use
2610 - Add three new models:
2611 * Floyd (shortest path computed at initialization)
2612 * Dijikstra (shortest path recomputed all the time)
2613 * Cached Dijikstra (shortest path computed on need)
2614 All these models where contributed by Silas De Munck, and are
2615 described in his ICCS09 paper.
2617 * Simplify model declaration
2618 (less redirections, less function to write when defining a model)
2619 - Factorize stuff between models:
2622 surf_model_resource_set(model)
2623 surf_model_resource_by_name(model, name)
2624 - Unify the types of models in s_surf_model_t (using an union)
2625 - Embeed fields of common_public directly into s_surf_model_t
2626 - Rename model methods:
2627 action_free ~> action_unref
2628 action_change_state ~> action_state_set
2629 action_get_state ~> action_state_get
2630 - Change model methods into functions:
2631 (model)->common_public->action_use ~> surf_action_ref
2633 * Implement a generic resource; use it as ancestor to specific ones
2634 (allows to kill duplicated code in models)
2635 Drawback: timer command don't need no name nor properties;
2636 workstation_CLM03 don't need no properties
2637 (but I guess we can live with those few bytes wasted)
2639 * Improve the action object model
2640 - implement a constructor avoiding dupplicated code about field
2641 initialization in generic_action part.
2643 * Kill the SDP model: it has an external dependency, is deprecated
2644 in flavor of modern lmm models, and didn't compile since a while
2647 * Relocation of the context module from XBT to SIMIX.
2648 (the context were decoupled from the simix processes, duplicating a lot of code)
2649 => a lot of code was factorized
2650 - less overhead is introduced during scheduling
2651 - simpler API for the context factory
2652 - the logic for process creation,destruction and manipulation was simplified
2653 * Simplification of the s_smx_process_t data structure.
2654 => accesing the simix level data associated to a process is faster now,
2655 and the code is a lot more readable.
2658 * Implement some more MPI primitives:
2659 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2660 -implementation: Bcast: flat or 2-ary tree (default),
2661 Barrier: 4-ary tree,
2663 Allreduce: Reduce then Bcast
2664 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2665 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2666 Alltoallv: flat tree, like ompi
2668 * Add support for optimized collectives (Bcast is now binomial by default)
2669 * Port smpirun and smpicc to OS X
2672 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2673 Shout out if you used it.
2676 * Display the list of still queued messages in SG mode when existing
2680 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2681 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2682 * Add xbt_str_from_file(FILE*)
2683 * Add xbt_dict_get_key achieving a linear reverse search
2684 * Remove the context module
2686 Portability report of this version:
2687 * Main portability targets:
2688 - Linux(debian)/x86/context
2689 - Linux(debian)/x86/pthreads
2690 - Linux(debian)/amd64/context
2691 - Linux(debian)/amd64/pthreads
2692 On these, we still have the eratic breakages of gras/pmm and
2693 amok/saturate_sg reported in previous version. We still think
2694 that the tests are the cause of the fault, not the tested code.
2696 - Mac OSX Leopard/x86/context
2697 Still false negative in tesh autotesting.
2698 Smpi still fails, but this time because readlink does not accept -f
2699 Everything seems to work properly beside of that.
2702 - AIX version 5.3 (only tested contexts this time)
2703 Smpi still fails there because mktemp is not installed.
2704 Everything seems to work properly beside of that.
2706 I managed to compile it for the first time, but several breakages.
2707 Won't delay the release for this exotic platform.
2709 * Windows: it's still lagging behind. If you want to help, please
2712 Timing report of this version:
2713 This version seem to be more than 5% faster than 3.3.1 (on linux
2714 64bits with contextes). The gain is less than expected, we are
2715 investigating this for next release.
2717 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2719 SimGrid (3.3.1) stable; urgency=low
2722 * Implement a --cfg-help to show existing configuration variables
2723 * Build chain do not require doxygen in maintainer mode
2726 * fix a bug on struct sizeof computation, which prevented the
2727 exchange of arrays of structs in some conditions
2728 - added a regression test about this in datadesc_usage
2729 * Allow the exchange of 0-long dynamic vectors.
2730 - for that, use -1 as indicator of dynamic size instead of 0
2731 - This implied to change any size from unsigned long to long,
2732 reducing a bit communication abilities, but I guess that with
2733 64bits being quite common, this is more than enough.
2734 - This also induce a protocol change, thus bumping network protocol
2735 version from 0 to 1 (if we have external users, we have to get
2736 clean on that point too ;)
2737 - added two regression tests about this in datadesc_usage
2738 * Be more verbose when propagating local exceptions
2739 This helps debugging.
2740 * Display the status of simulated processes when receiving SIGINT in
2744 * Allow to control the simulation from a trace file.
2745 New functions MSG_action_register() and MSG_action_trace_run()
2746 The first one allows to associate a function execution to each
2747 kind of action while the second one parses a trace file and
2748 triggers the corresponding actions within the system.
2749 For now, only a toy example is provided in examples/msg/actions
2750 * Add an exemple of process migration in examples/msg/migration
2751 * Fix a bug in task exchange which broke MSG_task_get_sender()
2752 Add a teshsuite regression test for that.
2753 [Bug: if MSG_task_get_sender() is called after sender exit,
2755 * Fix a bug which prevented suspend/resume to work properly
2756 * Display the status of simulated processes when receiving SIGINT
2757 This fixes a regression of v3.3. due to the introduction of SIMIX
2758 * Bug fixing in failure management:
2759 - trace could not start by a failure at time 0
2760 - failure during communications were not working
2763 * Add SIMIX_process_set_name() to change the name of the current
2764 process in the log messages.
2765 * Store smx_hosts in a dict since we only retrieve them by name
2766 * Move the configuration infrastructure to surf
2769 * Move the configuration infrastructure to surf
2772 * Massive internal cleanups:
2773 - Store internal structures on processes instead of hosts (allows
2774 to have more than one process per host, in addition of being more
2776 - Cleanup the initialization/finalization process
2777 - Kill a whole bunch of unneeded synchronization:
2778 processes run in exclusive manner within the simulator
2779 - Move queues from global tables to process data fields
2781 - now accept -platform and -hostfile arguments
2782 - Pass the right rank value to processes according to the hostfile
2783 * Compile the examples by default, and use them as regression tests
2784 * Implement MPI_Wtime()
2785 * Change the reference speed to a command line option
2788 * TCP_gamma can now be specified as command line option using
2789 --cfg=TCP_gamma:10000000.0
2790 * Change the --surf-path cmd line option into --cfg=path:
2793 * Also include strbuff from xbt.h public header
2794 * xbt_ex_display(): do not free the exception after displaying
2795 This allows to do more with the given exception afterward.
2796 Users should call xbt_ex_free() themselves.
2800 Portability report of this version:
2801 * Main portability targets:
2802 - Linux(debian)/x86/context
2803 - Linux(debian)/x86/pthreads
2804 - Linux(debian)/amd64/context
2805 - Linux(debian)/amd64/pthreads
2806 These targets fail about 1/10 of times on gras/pmm, but we believe
2807 that this is because of the test, not because of simgrid.
2808 amok/saturate_sg fails even more rarely, and the test may not be
2811 - Mac OSX Leopard/x86/context
2812 The test suite still spits tons of errors because some obscure
2813 force prevents us from removing the temporary directories
2814 arguing that they still contain some metadata I've never heard of.
2815 Smpi fails because seq is not installed.
2816 Everything seems to work properly beside of that.
2819 - AIX version 5.3 (both contexts and pthread)
2820 Smpi still fails there because mktemp is not installed.
2821 XML inclusions seems rosty on AIX.
2823 * Windows: it's still lagging behind. If you want to help, please
2826 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2828 SimGrid (3.3) stable; urgency=high
2832 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2833 [Malek Cherier & Mt]
2835 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2838 Until now, GRAS were using MSG as an interface to SURF. It was
2839 quite difficult because both interface have several differences
2840 (MSG channels vs GRAS sockets were the most notable point).
2842 This also opens the gate to SMPI (which should occur soon) and speed
2843 up simulations by to 40% (even if it were not the main goal).
2845 **************************************
2846 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2847 **************************************
2848 The point is that events occuring at the exact same timestamp are
2849 not scheduled in the same order with the old and new version. This
2850 may be enough to completely change the execution of simulations in
2851 some cases. Sorry for the inconvenience.
2853 * Cleanup and upgrade the XML format to push further scalability
2854 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2856 * Improve the testing infrastructure with tesh. Now a very large part of
2857 the code is tested not only by being run but also by checking that the
2858 output match an expected output [Mt].
2860 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2861 is really less memory-demanding, which should allow you to use
2862 larger files in SimGrid [AL].
2864 * Inform valgrind about our contextes, so that it becomes usable
2865 with the default (and more effecient) version of SimGrid
2866 [contributed by Sékou Diakite, many thanks]
2869 * Introduce a listener thread in charge of receiving incoming
2870 messages from the network. It allows to overlap communication and
2871 computation but most notably, it removes some stupid deadlocks due
2872 to the fact that so far, a process could not send and receive at
2873 the same time. This made most non trivial communication schema
2875 * Convert the PIDs from long int to int to match the MSG ones (and
2876 linux ones too) [Mt]
2877 * New function: gras_agent_spawn() to launch a new process on
2878 current host. Only working in simulation for now. [Mt]
2879 * New function: gras_os_hostport() returning a constant form (ie,
2880 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2883 * Make the backtrace of exceptions more human readable [Mt]
2884 * New module: xbt/str [Mt]
2885 a ton of string utility functions (split, join, printf to a newly
2886 allocated buffer, trim, etc)
2887 * New module: xbt/hash [Mt]
2888 SHA1 hashing algorithm (more to come if needed)
2889 * New module: xbt/synchro [Mt]
2890 synchronization tools (mutex and conditions) working the same way
2891 in simulation and in real life (mainly useful for GRAS, but not
2893 * New module: xbt/queue [Mt]
2894 classical producer/consumer synchronization scheme
2895 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2896 (using the classical functions will get serialized) [Mt]
2897 * Make dictionary internal table dynamic. No need to specify its size
2898 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2900 * Make sure the log channels are organized as a tree under windows
2901 (because of ANSI C compatibility issue, any channel were child of
2902 root directly) [Mt].
2905 * Cleaned many thing in surf and fixed a few bugs [AL].
2906 * Add a nice command line configuration mechanism to compose models [AL].
2907 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2908 the previous one (KCCFLN05). It relies on something that looks like
2909 a max-min sharing mechanism but cannot be written as such. A new solver
2910 was thus designed [AL].
2911 * Add a new solver to lmm. Based on Lagrange optimization and
2912 gradient-based descent, it enables to efficiently maximise systems s.a
2914 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2916 This solver enables to propose two new network models for TCP Reno and
2917 TCP Vegas based on Low's work. These models still need to be fully
2918 tested though [Pedro Velho].
2921 * Bug fix in SD_simulate. Now the time bound given as argument is
2923 * Use the new parallel task model (ptask_L07) as default.
2924 * Use the SURF command line configuration mechanism.
2925 * 0-size tasks (for synchronization) should now work.
2927 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2929 SimGrid (3.2) stable; urgency=high
2933 We still experience issues on this platform, but we believe that at
2934 least MSG is usable.
2936 GRAS API BREAKAGE (for simplification purpose, sorry):
2937 * the gras_msgtype_by_name is not used anymore. Instead of
2938 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2939 you can write (and must)
2940 gras_msg_send(toserver, "request", &request);
2941 - If you still want to pass a gras_msgtype_t to the function (to cache
2942 the type and avoid the lookup time), use the gras_msg_send_() variant.
2943 - Impacted functions:
2944 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2945 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2946 * The callbacks are now expected to return 0 when everything went well
2947 (just like the main() function)
2949 GRAS new features and improvements:
2950 * New module mechanism where user code can use per process globals [Mt]
2951 This is similar to gras_userdata_*() functions, but for libraries. It
2952 factorize some code developped over and over in the examples and AMOK.
2953 It has still to be documented and used (only amok/peermanagement is
2955 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2956 loss when they were small enough to fit into the buffer and sent quickly
2957 enough so that they can all get received in one shoot.
2958 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2959 if not found. Use the *_or_null() variant for the old semantic.
2960 * In gras_msg_handle, do not discard messages without callback.
2961 They are probably messages to be explicitly awaited later (ie, proofs of
2962 mis-synchronization in userland since they are sent before being awaited)
2963 No big deal usually.
2964 * gras_socket_meas_send/recv: semantic changed!
2965 The numerical arguments used to be (1) the total amount of data to send
2966 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2967 messages. This was need for the fool willing to send more than MAXINT
2968 bytes on quite fat pipes.
2971 * Do really rename the hostmanagement module to peermanagement. [Mt]
2972 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2973 but this is rather new and this was documented in the module
2974 documentation (poor excuses, I admit)
2975 * Bandwidth measurement semantic changed! This follows the changes to
2976 gras_socket_meas_send/recv explained above.
2979 * A sequential mode has been added to the workstations. When a workstation
2980 is in sequential mode, it can execute only one task, and the other tasks
2981 are waiting in a FIFO. [Christophe Thiery]
2984 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2985 model for SIMDAG. [Christophe Thiery]
2986 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2987 equal to 0, which caused some bad behaviors in
2988 saturated_constraint_set_update. I now use a threshold mechanism like in
2992 * When running manually src/testall, you select specific units [Mt]
2993 testall is the result of our cunit mechanism, and should replace all
2994 the scripty thingy around since bash don't run easily on billware.
2996 * A mallocator system has been added. [Christophe Thiery]
2997 Mallocators allow you to recycle your unused objects instead of freeing them
2998 and allocating new ones.
3000 Documentation update:
3001 * FAQ reworking + New FAQs:
3002 - "Valgrind spits tons of errors!" [Mt]
3003 - "How to repport bugs" [Mt]
3004 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3005 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3006 - Communication time measurement within MSG [AL]
3007 - I experience weird communication times when I change the latency [AL]
3008 * GRAS tutorial [Mt]
3010 - an introduction to the framework and to the used communication model
3011 - an initiatic tour introducing the most proheminent features:
3013 . Lesson 0: Installing GRAS
3014 . Lesson 1: Setting up your own project
3015 o Part 2: Message passing
3016 . Lesson 2: Exchanging simple messages
3017 . Lesson 3: Passing arguments to the processes (in SG)
3018 . Lesson 4: Attaching callbacks to messages
3019 . Lesson 5: Using globals in processes
3020 . Lesson 6: Logging information properly
3021 . Lesson 7: Using internal timers
3022 . Lesson 8: Handling errors through exceptions
3023 . Lesson 9: Exchanging simple data
3024 . Lesson 10: Remote Procedure Calling (RPC)
3025 . Lesson 11: Explicitely waiting for messages
3026 . Recapping of message passing features in GRAS
3027 - A HOWTO section containing:
3028 o HOWTO design a GRAS application
3029 More are due, of course. They will come latter. In the meanwhile, you can
3030 check the examples which are still here.
3032 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3034 SimGrid (3.1) stable; urgency=high
3038 There was a stack corruption somewhere, visible only when optimizing
3039 with these versions. [Vince]
3042 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3043 module that enabled to play with Directed Acyclic Graphs. It is built
3044 directly on top of SURF and provides an API rather close to the old
3045 SG. Some old codes using SG are currently under rewrite to check that
3046 all needful functions are provided. [Christophe Thiery]
3049 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3050 extension of the classical CLM03 model that gracefully handles
3051 failures. This is now the default model for MSG and GRAS. It doesn't
3052 handle parallel tasks yet though. [AL]
3053 * Bug fix: Weights were not correctly set in the network part.
3054 WARNING: This may have resulted in incorrect results with simulations
3055 where there are more than one flow on a given link. [AL]
3058 * After a (long ?) discussion on simgrid-devel, we have decided that the
3059 convention we had on units was stupid. That is why it has been decided
3060 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3061 WARNING: This means that all previous platform files will not work as
3062 such with this version! A warning is issued to ask users to update
3064 A conversion script can be found in the contrib module of the CVS, under
3065 the name contrib/platform_generation/surfxml_update.pl [MQ]
3068 * Bug fix: Processes were started in reverse order, wrt deployment file.
3069 WARNING: if your code relies on this bug, please fix it. [AL]
3070 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3071 being executed on two different locations. [AL]
3072 * Bug fix: Failures are now better supported thanks to Derrick's tests
3073 (there was many failure situations I hadn't thought of and that weren't
3074 correctly handled). [AL]
3075 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3079 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3080 Exception occurring on server-side are propagated back to client (!).
3082 API CHANGE: the callback changed their prototype. Change:
3083 int my_handler(gras_socket_t expeditor, void *payload_data) {
3085 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3086 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3088 * New! function: gras_msg_handleall to deal with all messages arriving
3089 within a given period.
3090 * New! function: gras_socket_server_range to get a server socket in a
3091 range of port numbers (ease to avoid port number conflicts) [MQ]
3092 * New! gras processes display their backtrace when they get a SIGUSR1
3093 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3094 Sweet to debug RL processes [MQ]
3098 - Do not force experiment sizes to be expressed in kb, or it becomes
3099 impossible to measure the latency this way (needs one byte-long tests)
3100 WARNING: this changes the amok_bw_* function semantic. [MQ]
3101 - Implements the link saturation stuff. [MQ]
3102 * Peer management module:
3103 New! module factorizing code that we wrote over and over [MQ].
3106 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3107 - Test units are placed directly into the library code, they get extracted
3108 automatically and placed into the src/testall binary.
3109 - Convert most of the XBT tests to this system.
3110 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3111 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3112 live in the same C code. This required to change a little bit the API
3113 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3114 * New module: sparse graph structure with basic algorithms (this is work
3115 in progress and the API is not considered to be frozen yet). [AL]
3116 * Display more information on backtraces: source line & function names are
3117 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3118 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
3119 * Reworked a little bit some #include statements to load only required
3120 headers. Some user code that relied on SimGrid to include stdlib or
3121 stdio may need to include it by themselves. [AL]
3122 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3124 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3125 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3126 but I guess that almost nobody used those parts. [MQ]
3128 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3130 SimGrid (3.0.1) stable; urgency=low
3133 * Unfortunately, I had missed 5 misnamed functions:
3134 xbt_fifo_item_t xbt_fifo_newitem(void);
3135 void xbt_fifo_freeitem(xbt_fifo_item_t);
3136 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3137 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3138 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3139 They're now deprecated. Please use their new versions:
3140 xbt_fifo_item_t xbt_fifo_new_item(void);
3141 void xbt_fifo_free_item(xbt_fifo_item_t);
3142 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3143 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3144 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3146 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3147 * Documentation: xbt_log module unmercifully reworked [MQ]
3148 * Bugfix: there was a problem with the ending of contexts with
3149 the pthread backend. It caused some weird deadlock or behavior
3150 depending on the pthread implementation. [AL]
3151 * Bugfix: get the exceptions raised in the simulator repport where
3152 and why they come from when they are not caught in time [AL, MQ]
3155 * Bugfix: Do repport the error when two non-connected hosts try to
3156 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3159 * Add additionnal checkings on communications. Assert that two
3160 communicating hosts are connected by a set of links... [AL]
3163 * Add additionnal checkings on channel values in communication [AL]
3164 * New: MSG_task_get_source to see on which host a task was generated [HC]
3165 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3166 the number of tasks waiting to be received on channel and sent
3168 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3169 waits for the first task coming from a given host.. [AL]
3171 GRAS new functionnalities: [MQ]
3172 * Enhance the parsing macro to allow the size of multidimentional objects
3173 to be given thru annotations.
3174 * New example (and documentation): Matrix Multiplication a la RPC
3175 (as when I was young!) and fix a bunch of bugs found on the way.
3177 GRAS performance improvements: [MQ]
3179 * Reduce the amount of cbps creation/destruction by making it static to
3180 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3183 * Change libdata to a set so that we can search for stuff by ID (and thus
3184 reduce the insane amount of dict lookups)
3187 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3188 RL communication on top of SG since it's so uneffective.
3189 It may also be used for inter-thread communication in RL, one day.
3190 * Use gras_datadesc_copy() to exchange messages on top of SG
3192 - improve message exchange performance on top of SG
3193 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3194 * Don't exchange on the network the size of the used part of buffer,
3195 instead, specify the possible buffer size to read().
3197 - reduces the amount of read/write calls (one pair per exchange)
3198 - reduces the amount of exchanged data (the size)
3199 - allows to retrieve all arrived data on receiver side, if we don't need
3200 it right now (subsequent read will peek the buffer)
3201 - allows the receiver to proceed with the begining of the stream before
3202 everything is arrived
3203 - make it possible to build an iov transport (using readv/writev)
3205 - take care of the data with non-stable storage (like stacked data),
3207 * If possible, TCP send uses vector I/O (when writev() is here)
3208 - Don't use it for receive since we send data sizes and data on the
3209 same stream, so we wouldn't be able to chain large amount of chunks
3210 before having to flush the stuff to read the size.
3211 * Rework the transport plugin mechanism to simplify it and reduce the
3212 amount of pointer dereferencement when searching for the right function
3215 * I guess that now, we do almost as few system calls as possible while
3216 doing as few data copy as possible.
3218 To improve it further, we could try to send all the sizes first and then
3219 all the data (to use iov on receiving size), but it's only a partial
3220 solution: when you have 2 dimensional data, the sizes of the second
3221 dimension is data of the first dimension, so you need 3 streams.
3223 I'm not sure the potential performance gains justify the coding burden.
3225 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3227 SimGrid (3.00) stable; urgency=high
3230 * New! Give the possibility to hijack the surf parser and thus bypass
3231 MSG_create_environment and MSG_launch_application. Have a look at
3232 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3234 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3236 SimGrid (2.96) unstable; urgency=low
3241 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3242 This deprecates the xbt_error_t mechanisms.
3243 It modifies (simplifies) all XBT and GRAS API.
3244 MSG API keeps unchanged (exceptions raised by XBT are caught from
3245 within MSG and masked with existing error handling facilities)
3248 * New! Add a FATPIPE model. [AL]
3249 * New! Add a parallel task model. [AL]
3250 * New! Add automatically a loopback interface (in the default
3251 network model) if none was precised.
3254 * Bugfix: MSG_process_resume now works with the current running process.
3256 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3257 * Modification of MSG_task_get_compute_duration. Once a task has been
3258 processed, the value returned by this function is now equal to 0. [AL]
3259 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3260 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3261 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3262 of a m_task. For now, MSG_task_cancel only works with computation
3264 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3265 of the processor (in Mflop/s) regardless of the current load on the
3267 * API Change: use proper naming convention for MSG_getClock and
3268 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3270 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3271 This function changes the priority of a computation task. This priority
3272 doesn't affect the transfer rate. A priority of 2 will make a task
3273 receive two times more cpu power than the other ones. This function
3274 has been added to suit the needs of Nguyen The Loc and hasn't been that
3275 much tested yet. So if it fails, please report it and send me your code.
3277 * API Change: removed all functions and types that were marked "deprecated"
3278 since many months. Renamed MSG_global_init_args to MSG_global_init.
3280 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3282 SimGrid (2.95) unstable; urgency=low
3285 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3286 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3287 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3290 * Bugfix: Allow absolute paths for platform description files [MQ]
3291 * Bugfix: do free the variables after use. Leads to drastic performance
3293 * Implement max_duration (ie, timeouts) on resources [AL]
3296 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3298 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3299 the needed support (timeouts) [AL]
3302 * Implement measurement sockets. You can now get the bandwidth between two
3303 hosts thanks to AMOK (see below). [MQ]
3304 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3305 dynar over the network (yeah) [MQ]
3306 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3307 * simple/static token-ring example. [Alexandre Colucci and MQ]
3308 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3309 in place before [MQ]
3312 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3313 type parsing. "short" and "long" modifiers were also ignored; other
3314 modifier (such as reference level) are still ignored. [MQ]
3315 * Embeed the buffer size within the buffer itself on SG. [MQ]
3316 That way, send() are atomic and cannot get intermixed anymore (at least
3317 the ones which are less than 100k; bigger messages still have the issue)
3318 * Array size pushed by the field, not by the field type (or each
3319 and every long int will push stuff to the cbps) [MQ]
3320 * use select() to sleep since it allows to portably sleep less than one
3323 GRAS (minor cleanups)
3324 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3327 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3328 * type category 'ignored' killed as it was never used and were difficult
3330 * whether a type can cycle or not is now a flag, leaving room for more
3331 flags (as "ignored", if we feel the need one day ;)
3332 * Rename raw sockets to measurement sockets since "raw" has another
3333 meaning in networking community.
3336 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3337 and offers features not belonging to GRAS but that most applications
3338 need. One day, it may be a set of plugins loadable at runtime.
3339 * New module: bandwidth
3340 bandwidth measurement between arbitrary nodes running this module. [MQ]
3342 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3344 SimGrid (2.94) unstable; urgency=low
3346 The first beta release of SimGrid 3 !
3350 * Update the main page and the FAQ. Adding references to gforge.
3353 * Add a gras_os_getpid function.
3356 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3357 * Extend the logs so that they also print PID, hostname, date, ... if
3359 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3360 and kill the old version which were in testsuite/
3361 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3365 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3366 for users. Instead, each programming environment (GRAS, MSG,...) use it
3367 when needed to provide such a feature to users.
3368 Don't shortcut the mechanism or you will also shortcut the virtualization
3369 you need on the simulator.
3373 * Cleanups in configury with regard to compile optimization/warning flags.
3374 Also add -fno-loop-optimize to any powerpc since it's the optimization
3375 killing gcc (< 3.4.0).
3376 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3377 needed by MSG examples complications
3378 * Borrow configury beautifications from PHP
3381 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3382 warning (thanks loris for stumbling into this one).
3383 * Bugfix: stop loading private headers (gras_config.h) from the public
3387 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3388 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3389 RL would imply a DNS resolver, which is *hard* to do in a portable way
3390 (and therefore delayed).
3391 * Implement a real timer mechanism and use it in timing macros. This allows
3392 to avoid rounding errors and get a 0.000005 sec precision in timing
3393 macros. While I was at it, various cleanups:
3394 - allow to declare more than one timed section per file (fix a stupid bug)
3395 - move some private declaration to the right place
3396 - merge conditional execution and timing macros into emulation module
3397 - document the module
3398 - make sure the module cleanups its mess on gras_exit
3399 * Documentation improvements:
3400 - (new) how to compile applications using GRAS
3401 - (new) emulation support (timing macros)
3403 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3405 SimGrid (2.93) unstable; urgency=low
3407 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3410 - Use Paje properly where used. Still to be sanitized properly.
3411 - Portability fix: Add an implementation of the contexts using pthread
3415 - Add xbt_procname(): returns the name of the current process.
3416 Use it to show the current process's name in all logging.
3418 - fix detection of older flex version and the reaction, since we do
3419 depend on modern ones (we use lex_destroy)
3420 - Better separation of SG and RL in the libs: remove all simulation code
3421 from libgras. As a result, this lib is now only 200k when stripped.
3422 Some of the xbt modules may also be duplicated (two sets and such) and
3423 should be cleaned/killed before SG3.
3424 - Insist on using xlC on AIX because of weird problems involving gcc there.
3425 - Cleanup the make remote stuff. This is now done by scripts
3426 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3427 mainly for our private use, but we're working on changing them to user
3430 - Bugfix: flush the socket on close only if there is some *output*.
3431 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3433 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3436 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3438 SimGrid (2.92) unstable; urgency=low
3440 Alpha 3 on the path to SimGrid 3
3444 - New! First try of benchmarking macros.
3445 - New! First try so that gras_stub_generator generate deployment and
3446 remote compilation helpers.
3448 - Bugfix: Initialization fix in msg_test.
3452 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3454 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3456 - API Change: gras_os_sleep to take the amount of seconds as a double.
3457 Accepting an int was error prone since it was the only location where
3458 seconds were coded as such. It leaded to damn rounding errors.
3459 - Bugfix: Hard to belive that timers ever worked before this.
3461 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3463 SimGrid (2.91) unstable; urgency=low
3465 Alpha 2 on the path to SimGrid 3
3469 - Bug fix in the lmm_solver.
3471 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3472 through the function MSG_paje_output.
3473 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3474 - It is possible to bound the rate of a communication in MSG with
3475 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3476 to write it in the changelog).
3477 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3482 - Various cleanups to the autotools stuff
3483 - Begin to move Gras examples to examples/gras/
3484 - Let make distcheck work again (yeah!)
3486 - documentation overhauled using doxygen.
3487 gtk-doc-tools is dead in SimGrid now.
3488 - Automatically extract all existing logging categories, and add the list
3489 to the documentation (long standing one, to say the less)
3491 - Cleanup the known architecture table. Reorder the entries to group what
3492 should be, and use a more consistent naming scheme.
3493 (some of the test dataset are still to be regenerated)
3494 - New! Allow library to register globals on each process just as userdata
3496 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3497 lookup time (for now).
3498 Use it in msg and trp.
3499 This cleans a lot the internals and helps enforcing privacy of the
3500 headers between the gras components.
3501 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3502 - Bugfix: gras_os_time was delirious in RL.
3503 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3505 - Reenable GRAS now that it works.
3507 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3509 SimGrid (2.90) unstable; urgency=low
3511 Alpha 1 on the path to SimGrid 3
3513 * It is a long time since the last release of SimGrid. I'm sorry about
3514 that but as I had told you, I was rewriting a lot of things. I apologize
3515 to those who had been reporting bugs to me and that I had not answered.
3516 If your bug is still in the new version, please tell me. Here is a
3517 summary of the main changes.
3519 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3520 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3521 portability, speed, and a lot more but you'll figure it out later.
3522 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3523 I wanted to release the new MSG as soon as possible and I have
3524 broken GRAS, which is the reason why, for now, only MSG is fully
3525 functional. A laconic description of these projects is available
3526 in the documentation.
3528 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3529 kernel optimized for our needs (called SURF but only the developers
3530 should use it). Hence, MSG is now roughly 30 times faster and I think
3531 that by rewriting a little bit MSG, I could event speed it up a little
3532 bit more. Beside the gain in speed, it is also much easier to encode a
3533 new platform model with SURF than it was with SG. More to come...
3535 * REVOLUTION 3: I have tried to change a little as possible the API of
3536 MSG but a few things really had to disappear. The main differences
3537 with the previous version are:
3538 1) no more m_links_t and the corresponding functions. Platforms are
3539 directly read from a XML description and cannot be hard-coded
3540 anymore. The same format is used for application deployment
3541 description. The new format is described in the documentation.
3542 Have a look in tools/platform_generation. There is a tiny script
3543 that converts from the old platform format to the new one. Concerning
3544 the application deployment format, parsing the old one is tricky.
3545 I think most of you should however be able to convert your files. If
3546 it is really an issue, I can write a C code that does the conversion.
3548 2) the toolbox tbx does not exist anymore. We now have a library
3549 with much more data-structures but without the hash-tables (we have
3550 dictionaries that are much faster).
3552 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3554 *****************************************************************************
3555 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3556 * don't feel like converting the oldies. (Mt) *
3557 *****************************************************************************
3560 Version 2.90: "the long awaited one"
3561 - Finished rewriting and debugging MSG. Rewrote the documentation.
3562 - disable GRAS for now since it needs to be ported to the newest SG
3565 - Finish the port to windows (using mingw32 for cross-compile)
3568 - Main loop and datastructures of SURF. A cpu resource object is
3569 functional. Surf can thus be used to create cpu's with variable
3570 performance on which you can execute some actions.
3572 2004-11-15 Martin Quinson
3573 - Port to ARM. Simply added the alignment and size descriptions. Should
3574 work, but the ARM machines are so slow that I didn't had the opportunity
3575 to 'make check' over there yet.
3577 2004-11-15 Arnaud Legrand
3578 - Trace manager now written. It uses a heap structure and is therefore
3579 expected to be efficient. It may however be speeded up (particularly
3580 when many events occur at the same date) by using red and black
3581 trees. One day maybe...
3582 - Max-min linear system solver written. It uses a sparse matrix
3583 structure taking advantage of its expected use. Most operations are
3584 O(1) and free/calloc are called as few as possible. The computation of
3585 the minimum could however be improved by using a red and black tree
3588 2004-11-03 Arnaud Legrand
3589 - Rename every gras_* function that was in xbt/ to its xbt_
3591 - Add a heap and a doubly-linked list to xbt
3592 - Added a dichotomy to the dictionaries. make check works as well before
3593 so I assume that the patch is correct. I do not know however if things
3594 run effectively faster than before now. :)
3596 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3597 SimGrid, and the version number is bumped to 2.x
3599 2004-10-29 Martin Quinson
3600 - Introduction of the remote errors.
3601 They are the result of a RMI/RPC on the remote machine.
3602 ErrCodes being scalar values, you can't get the host on which those
3603 errors did happen. Extending the error mechanism as in Gnome is possible.
3604 No idea yet whether it is a good idea.
3606 2004-10-28 Martin Quinson
3607 - Interface revolution: the Starred Structure Eradication.
3608 I used to do typedef struct {} toto_t; and then handle *toto_t.
3609 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3610 - ???_t is a valid type (builded with typedef)
3611 - s_toto_t is a structure (access to fields with .)
3612 - s_toto is a structure needing 'struct' keyword to be used
3613 - e_toto_t is an enum
3614 - toto_t is an 'object' (struct*)
3616 typedef struct s_toto {} s_toto_t, *toto_t;
3617 typedef enum {} e_toto_t;
3618 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3619 s_toto_t) is private.
3621 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3622 so that it can set the variable to NULL. It was so for dicts and sets,
3623 it changed for dynars.
3625 - Fix a bunch of memleaks in dict_remove
3626 - Fix a bug in sg/server_socket opening: it failed all the time.
3628 2004-10-07 Martin Quinson
3629 - Speed up dynar lookup operation a bit.
3631 gras_dynar_get is dead.
3633 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3634 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3635 (which gives you the address of the stored data).
3637 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3638 copy of the data using an affectation to do the job and not a memcpy.
3640 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3641 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3643 It does not really speedup the dynar test because they are
3644 setting elements all the time (and look them seldom). But the dict does
3645 far more lookup than setting.
3647 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3649 2004-10-05 Martin Quinson
3650 - Allow to (en/dis)able the cycle detection at run time.
3652 Whether we should check for cycle or not is now a property of each
3653 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3654 datadesc_cycle_unset allow to remove this property when previously set.
3656 Note that the cycle detection is off by default since it impacts the
3657 performance. Watch the data you feed GRAS with ;)
3659 This property is hereditary. Any element embedded in a structure having it
3660 set have it set for the time of this data exchange.
3662 You should set it both on sender and receiver side. If you don't set it on
3663 sender side, it will enter an endless loop. If you forget on receiver
3664 side, the cycles won't be recreated after communication.
3666 - Header reorganization.
3667 Kill gras_private.h, each submodule must load the headers it needs.
3669 2004-10-04 Martin Quinson
3670 - Interface revolution: do not try to survive to malloc failure.
3672 Now, gras_malloc and friends call gras_abort() on failure.
3673 As a conclusion, malloc_error is not a valid error anymore, and all
3674 functions for which it was the only gras_error_t return value are
3675 changed. They now return void, or there result directly.
3676 This simplify the API a lot.
3678 2004-09-29 Martin Quinson
3679 - Re-enable raw sockets.
3680 Created by gras_socket_{client,server}_ext;
3681 Used with gras_raw_{send,recv}
3684 It should allow to kill the last bits of gras first version soon.
3686 This is not completely satisfactory yet (duplicate code with
3687 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3690 - Simplify transport plugin (internal) interface by not passing any
3691 argument to _server and _client, but embedding them in the socket
3694 2004-09-28 Martin Quinson
3695 - Finish the port to AIX.
3696 autoconf was my problem (segfault within the malloc replacement
3697 function. No idea why)
3699 2004-09-16 Martin Quinson
3700 - Fix some size_t madness on 64bit architectures.
3702 2004-09-08 Martin Quinson
3703 - Reduce the number of system headers loaded, overload some more system
3704 calls (such as malloc to cast the result of the system one, and work
3706 - Fix and reintroduce the config support
3708 2004-09-07 Martin Quinson
3709 - Source code reorganization to allow Arnaud to surf all over there.
3710 - Allow to document the logging categories.
3711 - Remove all uppercase from logging categories and useless cleanup in names.
3713 2004-08-18 Martin Quinson
3714 Version 0.6.2 (protocol not changed; API changed)
3715 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3716 gras_error_t), and NULL when not found. Functions expecting a msgtype
3717 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3718 hopefully usefull message.
3719 - Portability to prehistoric sparcs again
3721 2004-08-17 Martin Quinson
3722 Version 0.6.1 (protocol not changed; ABI not changed)
3723 - prealloc some buffers to speed things up
3725 2004-08-11 Martin Quinson
3726 Version 0.6 (protocol not changed; ABI expended)
3727 - The parsing macro can deal with the references, provided that you add
3728 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3730 2004-08-09 Martin Quinson
3731 Version 0.5 (protocol not changed; ABI changed)
3732 - Allow to off turn the cycle detection code in data exchange at
3733 compilation time. It should be at run time, but I'm short of time (and
3734 the config stuff is still broken). That way, we keep dict out of the
3735 critical path, which is good because the performance is poor:
3736 - search not dichotomial yet
3737 - dynar give no way to access their content and memcpy everytime
3738 - In composed data description (struct, ref and so on), stop foolness of
3739 keeping the subtype's ID, but store the type itself. This keeps sets out
3740 of the critical path, which is good since they rely on dynar and
3741 dictionnaries. The only loose of that is that we cannot detect the
3742 redeclaration of a structure/union with another content (but I'm not sure
3743 the code detected well this error before anyway). We still can detect
3744 the redefinition discrepancy for the other types.
3745 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3746 it breaks the code because of type-punning used all over the place).
3747 This breaks on all non-gcc architectures (for now).
3749 All those changes (plus the buffer of last time) allow me to gain 2 order
3750 of magnitude on cruel tests consisting of 800000 array of integers on two
3751 level of a hierarchical structure (200 secondes -> 4 secondes)
3754 - the selector of reference must now return the type it points to, not
3755 the ID of this type.
3757 2004-08-06 Martin Quinson
3758 Version 0.4 (protocol changed; ABI not changed)
3759 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3760 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3761 - Implement a buffer transport, and use it by default (it relies on tcp in
3762 real life and on sg in simulation).
3763 That's a bit hackish since I had a new field to the structure to store
3764 its data without interfering with the subtype ones. Inheritance
3765 is tricky in C. And that's a kind of reverse inheritance with one class
3766 derivating two classes. Or maybe a game with java interfaces. Anyway,
3767 that's damn hard in C (at least).
3768 Moreover, I got tired while trying to ensure plugin separation and
3769 genericity in SG mode. MSG wants me to do weird things, so let's go for
3770 cruel hacks (temporarily of course ;).
3771 See comment in transport_private.h:71
3772 - do not include all the _interface headers in private but in the files
3773 which really need them (to cut the compilation time when they are
3776 2004-07-26 Martin Quinson
3777 Version 0.3 (protocol not changed; ABI changed)
3778 - Major overhault of the datadesc interface to simplify it:
3779 - shorted the function names:
3780 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3781 - add a trivial way to push/pop integers into the cbps without malloc.
3782 This allows to make really generic sub_type description, which simply
3783 pop their size of the stack.
3784 - add a function gras_datadesc_ref_pop_arr() which does what users want
3785 most of the time: Declare a dynamic array (which pops its size of the
3786 stack) and declare a reference to it. Poor name, but anyway.
3787 - kill the post-send callback, add a post-receive one
3789 2004-07-23 Martin Quinson
3790 Version 0.2 (protocol changed; ABI changed)
3791 - add some testing for cpbs in the test cases, and fix some more bugs.
3792 This invalidate again the little64 data file, since I cannot regenerate
3794 - remove an awfull optimization in the logging stuff, allowing me to:
3795 - understand it again
3796 - learn gcc how to check that the argument match the provided format
3797 - fix all errors revealed by gcc after that
3798 - internal keys of dict are not \0 terminated. Deal with it properly in
3799 loggings instead of segfaulting when the user want to see the logs :-/
3801 2004-07-22 Martin Quinson
3802 - Fix some stupid bug preventing cbps (callback postit) from working
3804 2004-07-21 Martin Quinson
3805 - Some documentation cleanups
3806 - remove the useless last argument of msgtype_declare
3807 - rename the Virtu functions to fit into the 'os' namespace
3808 - move headers src/include -> src/include/gras/ and stop fooling with
3809 gras -> . symbolic link
3810 - make distcheck is now successful
3812 2004-07-19 Martin Quinson
3814 - Build shared library also
3815 - Install html doc to the right location
3816 - stop removing maintainer files in make clean
3817 - build tests only on make check
3819 2004-07-13 Martin Quinson
3821 - No major issue in previous version => change versioning schema
3822 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3823 corresponding dataset.
3825 2004-07-11 Martin Quinson
3827 - Get it working with any kind of structure (we can compute the padding
3828 bytes remotely for all the architectures I have access to)
3829 - Implement the structure parsing macro (still not quite robust/complete)
3830 - Improvement to the remote testing toysuite
3832 2004-07-10 Martin Quinson
3833 [autoconf mechanism]
3834 - get ride of a bunch of deprecated macros
3835 - actually run the test for two-compliment, not only compile it :-/
3836 - test whether the structures get packed (and bail out if yes. Damn.
3837 Alignment is a serious matter)
3838 - test whether the structures get compacted (but respecting the alignment
3839 constraints of each types)
3840 - test whether the array fields of structures can straddle alignment boundaries
3842 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3843 (took me 2 hours to find that bug, looking at the wrong place)
3844 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3845 before sending/receiving and are used to compute the offsets of fields
3846 - Given that padding size depend even on compiler options, keep track of
3847 alignment and aligned_size only for the current architecture. Not a big
3848 deal since we send structure fields one after the other (seems
3850 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3851 program, let it work on linux/gcc/little32. portability todo.
3853 2004-07-08 Martin Quinson
3854 - import and improve remote compilation support from FAST
3855 - make sure make check works on half a dozen of machines out there
3857 2004-07-07 Martin Quinson
3858 Let's say it's version 0.0.3 ;)
3859 - Implement conversions (yuhu!)
3860 - Let it work on solaris (beside conversion, of course)
3861 - Stupid me, using rand() to generate the conversion datatests in not wise.
3863 2004-07-06 Martin Quinson
3864 - Let make dist work, since I'm gonna need it to compile on remote hosts
3865 - Let Tests/datadesc_usage write the architecture on which the file was
3866 generated as first byte.
3867 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3868 architecture descriptions.
3869 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3870 on those architectures.
3871 - Optimization: send/recv array of scalar in one shoot
3873 2004-07-05 Martin Quinson
3874 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3876 - Plug a whole bunch of memleaks
3877 - each process now have to call gras_{init,exit}. One day, their log
3878 settings will be separated
3879 - Continue the code factorisation between SG, RL and common in Transport.
3881 2004-07-04 Martin Quinson
3883 - Redistribution between SG and RL.
3884 We wanna have to accept in SG, so move accepted related parts of RL in
3885 the common part. (more precisely, the dynar of all known sockets is no
3886 more a static in transport.c, but part of the process_data)
3888 [gras_stub_generator]
3889 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3891 2004-07-03 Martin Quinson
3892 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3893 them all up in one shot)
3895 - Fix the ugly case of reference to dynamic array.
3896 - New (semi-public) function gras_datadesc_size to allow the messaging
3897 layer to malloc the needed space for the buffer.
3899 - gras_socket_close now expect the socket to close (and not its address to
3900 put NULL in it after it). This is because the socket passed to handlers
3901 is one of their argument (=> not writable).
3903 - propagate the interface cleanup from last week in datadesc, ie remove a
3904 superfluous level of indirection. User pass adress of variable
3905 containing data (both when sending and receiving), and not of a variable
3906 being a pointer to the data. Let's say that I like it better ;)
3907 The price for that is constructs like "int msg=*(int*)payload" in
3908 handlers, but it's a fine price, IMHO.
3910 - Let it work in RL (yuhu)
3912 2004-06-21 Martin Quinson
3914 - porting SG plugin and SG select to new standards (works almost).
3915 - plug memleaks and fix bugs around.
3918 - cleanup the prototype of data recv and force users to specify when they
3919 want to handle references to objects. Test case working even for cycles.
3920 - plug memleaks. Valgrind is perfectly ok with this.
3922 2004-06-12 Martin Quinson
3924 - cleanup the separation between plugin and main code in plugin creation
3926 2004-06-11 Martin Quinson
3928 - Reput hook for raw sockets, needed for BW experiments
3929 - kill a few lines of dead code
3930 [Data description] Interface cleanup
3931 - gras_datadesc_by_name returns the searched type or NULL.
3932 That way, no variable is needed to use a type desc once, which makes
3934 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3935 two parameters were strings (field name, type name), leading to
3937 [Dicos] Interface cleanup
3938 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3939 This is consistant with the dynar API.
3941 2004-04-21 Martin Quinson
3943 - Porting to new standards.
3945 - interface cleanup.
3946 There is no bag anymore, no need to take extra provision to mask the
3947 pointers behind "ID".
3948 Better splitup of functions between files create/exchange/convert.
3949 This is still a bit artificial since convert and receive are so
3950 interleaved, but anyway.
3952 - add a queued message list to procdata (the ones not matching criteria
3954 - factorize some more code between SG and RL wrt procdata
3956 - use gras_exit in example to track memleaks
3957 - get rid of gs_example now that GS is properly integrated into gras
3958 - update run_test to integrate the latest tests (datadesc)
3960 - rename WARNINGn macros to WARNn since it prooved error-prone
3962 2004-04-19 Martin Quinson
3964 - register init/exit functions within gras module mechanism
3965 - send/receive function.
3966 Convertion is not implemented, but short-cutted if not needed.
3967 struct/array elements are sent one by one (instead of block-wise), but
3968 nobody really cares (yet). Get a prototype before optimizing.
3969 - tests (using a file socket) for DD send/receive on:
3970 - base types: int, float
3971 - array: fixed size, string (ie ref to dynamic string)
3972 - structure: homogeneous, heterogeneous
3973 - chained list, graph with cycle
3974 Believe it or not, valgrind is not too unhappy with the results. The
3975 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3979 - Bugfix when using a filename explicitely (instead of '-')
3981 2004-04-09 Martin Quinson
3983 - factorize more code between RL and SG in socket creation
3984 - Complete the implementation and tests of:
3986 o file (only in RL, and mainly for debugging)
3988 I lost 3 days to design a portable address resolver, and then decided
3989 that the prototype mainly have to run on my box.
3990 Addressing portability too early may be like optimizing too early :-/
3992 - use gras_init in the Tests instead of the crappy parse_log_opt
3993 (the latter function is removed)
3994 [Conditional execution]
3995 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3996 [Code reorganisation]
3997 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3998 Build examples against the RL library, since there is no way to disable
3999 its creation for now.
4001 For information, the beginning of coding on GRAS was back in june
4002 2003. I guess that every line has been rewritten at least twice since