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
12 - Global signals are now part of the Engine:
13 - on_platform_creation: after config settings, before the XML parsing
14 - on_platform_created: right after the XML parsing
15 - on_time_advance: each time the clock advances
16 - on_simulation_end: after simulation, before cleanups
17 - on_deadlock: as the name implies.
20 - convert a new set of functions to the S4U C interface and move the old MSG
21 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
24 - Fortran bindings for DVFS have been removed.
25 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init
26 - SMPI can now be selected by cmake's find_module(MPI) with
27 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
28 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
29 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
32 - Option model-checker/hash was removed. This is always activated now.
33 - New option smpi/buffering controls the MPI buffering in MC mode.
34 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
35 This is useful to check for MPI compliance.
38 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
39 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
42 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
43 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
44 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
45 - FG#34: SG_BARRIER_SERIAL_THREAD?
46 - FG#35: model-checker does not like buster-produced binaries
48 ----------------------------------------------------------------------------
50 SimGrid (3.23.2) July 8. 2019
53 - Nicer introduction page.
54 - Migrate the "Deploy your application" page to the new doc.
55 - Move Java as a subtree of MSG.
58 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
61 - Drop xbt_dynar_sort_strings().
64 - Really fix FG#26: Turning off a link should raise NetworkFailureException
65 - FG#27: Wrong exception thrown to wait_any when link is turned off
66 - GH#328: Java: Canceling multiple tasks in a single vm/host
68 ----------------------------------------------------------------------------
70 SimGrid (3.23) June 25. 2019
72 The Exotic Solstice Release.
75 - SunOS and Haiku OS support. Because exotic platforms are fun.
76 - Stop setting random seed with srand48() at initialization.
77 - Use addr2line as a fallback for stacktraces when backtrace is not available.
78 - Build option -Denable_documentation is now OFF by default.
79 - Network model 'NS3' was renamed into 'ns-3'.
82 - Simgrid can now hopefully be installed with pip.
85 - wait_any can now be used for asynchronous executions too.
88 - New log appenders: stdout and stderr. Use stdout for xbt_help.
92 - 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
93 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
94 - MPI interface is now const correct for input parameters
95 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
96 - Fortran bindings for async collectives.
97 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
100 - Remove option 'model-check/record': Paths are recorded in any cases now.
101 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
102 always sparse. That's an immense gain in memory, and can even be
103 faster because of cache effects. So there is no need to clutter the
104 code to allow the user to go for the unefficient mode.
107 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
108 models proved to be more accurate than these old experiments.
110 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
111 - FG#1: Broken link in error messages
112 - FG#2: missing installation documentation
113 - FG#3: missing documentation in smpirun
114 - FG#6: Python bindings not available on PyPI
115 - FG#7: simple cmake call requires doxygen
116 - FG#8: make python bindings an optional dependency
117 - FG#10: Can not use MSG_process_set_data from SMPI any more
118 - FG#11: Auto-restart actors forget their on_exit behavior
119 - FG#12: -Denable_lto=OFF doesn't disable LTO
120 - FG#13: Installs unstripped file 'bin/graphicator'
121 - FG#14: Installs the empty directory 'doc/simgrid/html'
122 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
123 - FG#17: Dead link in doc (pls_ns3)
124 - FG#20: 'tesh --help' should return 0
125 - FG#21: Documentation link on http://simgrid.org/ broken
126 - FG#22: Debian installation instruction are broken
127 - FG#26: Turning off a link should raise NetworkFailureException exceptions
128 - GH#133: Java: a process can run on a VM even if its host is off
129 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
130 - GH#326: Valgrind-detected error for join() when energy plugin is activated
132 ----------------------------------------------------------------------------
134 SimGrid (3.22) April 1. 2019
136 The Easter Christmas Release (this one is somewhat late).
139 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
140 - Not complete yet: asynchronous activities (amongst others) are still missing
141 - Still ongoing: the interface may change in the future. We need more
142 testers! Please report any glitches.
143 - No new project using Java should start now. Please switch to Python.
146 - Some of the internal cleanups may lead to speed improvements:
147 - The hard limitation on the amount of simulated actors with
148 Java+Mac was removed. Now, the available memory is the only limit.
149 - Our refcounting was tidyied, leading to 10% speedups in some cases.
150 - We are still working on making our code robust to the actor kills
151 and hosts' churn. Things are improving, but it's not perfect yet.
152 - Replaced our own code to display a backtrace (that was forking addr2line)
153 with the Boost.Stacktrace library.
154 You won't see your backtraces without this optional dependency.
155 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
156 - Stop setting random seed with srand() at initialization.
159 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
160 XML file version remains 4.2 since old files are still compatible.
163 - Process termination which was broken at version 3.21 has been repaired.
164 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
165 - Hide the examples into examples/deprecated. New users should use Python.
168 - Drop MSG_process_create_from_stdfunc() from the API.
169 This C++ function was a pimple in the C API, made necessary at some
170 point by the Java bindings. This is fixed now.
171 - Hide the examples into examples/deprecated. New users should use S4U.
172 - MSG_process_create and MSG_process_attach now crash if the host used
173 to run this process is off.
174 - Fix the protype of MSG_process_on_exit()
175 Now use (int,void*) callbacks instead of (void*,void*) ones.
176 The implementation was ways too messy to actually work, I guess.
179 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
180 too often and provide more accurate timings.
181 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
182 - MPI_Alltoallw support
183 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
184 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
185 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
186 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
188 - Basic implementation of generalized requests (SMPI doesn't
189 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
192 - Drop sg_cmdline. Please use xbt_cmdline instead.
193 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
194 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
195 OsSemaphore is implemented in a portable way with C++11 threads.
196 This should allow much more threads to be created at the same time,
197 allowing Mac and Java users to simulate many more actors.
198 - Implement the 'thread' factory with std::thread instead of xbt ones.
199 It is not possible to set the stack size with threads anymore, but
200 -fsplit-stack is the way to go nowadays when using threads.
201 - Drop the xbt_os_thread_t module (now unused)
202 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
203 - Drop xbt_str_join_array().
204 - Drop cunit, use Catch2 instead.
207 - Many cleanups in the kernel::activity namespace. This was long
208 overdue, and shall open the path to many future endeavors.
211 - #132: Java: a process can not shut down its own host
212 - #220: S4U: detached send causes exception if sender terminates soon after sending
213 - #261: Document the parameters of parallel execution's constructor
214 - #300: [s4u] BarrierPtr is missing
215 - #314: SMPI args internal cleanup
216 - #316: Fix a bug related to the CPU utilization of multi-core VM
217 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
218 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
219 - #325: Turning off a host has different behavior on sleeping actors and computing actors
221 ----------------------------------------------------------------------------
223 SimGrid (3.21) October 3. 2018
225 The Restarting Documentation (TRD) Release.
228 - Convert the user manual to Sphinx for improved usability.
229 Unstable version now hosted on https://simgrid.frama.io/simgrid
230 Many glitches remain; stay tuned.
231 - Restrict the Doxygen reference API to the installed header files.
232 (documenting all internals this way is overkill + doxygen scales badly)
233 - New tutorial on S4U (highly inspired from the old MSG one, but with
234 a git repository to fork as a starting point, and a docker image)
235 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
236 - The Developper manual is still to be converted (not compiled until then)
237 - Some parts are still missing in this conversion (such as the
238 platform and deployment sections) while others were blindly converted
239 and would need more love (such as the configuration flags).
240 Things will certainly further improve in the future releases.
243 - Fully reimplement auto_restart mechanism that was utterly broken
244 (fix #22, #131 and #281 that were all very related)
245 - Implement semaphores (fix #303 and #291)
246 - When creating an actor from a function and its parameters,
247 move-only parameters are not allowed anymore, as it would prevent
248 the actor to be restartable if its parameters are consumed on run.
249 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
250 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
251 IO operations that can be started, waited for, or canceled as a regular
253 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
257 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
258 - New host variable: 'core_count'
261 - Remove xbt_os_thread_specific features
262 - Remove portability wrapper to condition variables
263 - Remove xbt_os_thread_yield()
266 - MPICH collective selector now mimics MPICH 3.3b
267 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
268 - MPI_Init(NULL,NULL) is now allowed
269 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
270 - Command-line arguments should now be handled properly in Fortran simulations
271 - MPI Attributes and related callbacks should work in Fortran
272 - Apps using C + Fortran should now work
273 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
274 - Support MPI_Error_String
277 - Due to an internal bug, Msg.run() must now be your last line.
278 We hope to fix it in a future release, and we are sorry for the inconvenience.
281 - #22: Process autorestart seem to only work with CAS01 cpus
282 - #56: Feature request: dynamic SMPI replay
283 - #93: simgrid should not eat --help
284 - #111: How to change the loopback link characteristics?
285 - #116: Platform cluster radical
286 - #117: Platform cluster/host order
287 - #129: Replay_multiple on steroid
288 - #131: Java : autostart does not work
289 - #143: Setting a breakpoint at a given time
290 - #235: Network energy models should be integrated
291 - #264: Add ptask L07 resource tracing
292 - #271: Dynamic and manual replay of SMPI traces
293 - #279: Breakpoints option uses time deltas instead of absolute time
294 - #281: Daemonized actor no longer daemonized after an auto-restart
295 - #282: TIT convertor erase traces when using absolute path in the trace list file
296 - #285: segfault when a s4u actor kills itself with msg process tracing activated
297 - #286: Pajé traces are not exposing the number of cores
298 - #287: Command-line parsing should stop when encountering '--'
299 - #288: MPI_Init(NULL, NULL)
300 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
301 - #290: Method to check if a mailbox has a message ready to be consumed
302 - #291: [s4u] semaphores do not seem available yet
303 - #294: Alltoall collective implementation of mpich
304 - #295: Starting a Host as OFF is not supported
305 - #296: DTD too permissive
306 - #299: Add a s4u barrier example
308 ----------------------------------------------------------------------------
310 SimGrid (3.20) Released June 24. 2018
312 The proxy snake_case() release.
315 - The whole API is now using snake_case() instead of mixing it with camelCase.
316 Compatibility wrappers in place for one year.
317 - We took the opportunity to increase the consistency of the API.
318 Please report any remaining glitches.
319 - New function simgrid::s4u::Host::get_actor_count:
320 Returns the number of actors running on a specific host.
323 - SMPI is now tested with ~45 proxy apps from various sources, with none or
324 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
325 - Replay: The replay file has been re-written in C++.
326 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
327 supported. They are stored in the trace and used when replayed.
328 - Basic support of MPI_Cancel. Robustness not guaranteed.
329 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
330 (only for MPI_COMM_TYPE_SHARED).
331 - New option: smpi/privatize-libs, to add external shared libs to be privatized
332 by SMPI. They will be copied locally and loaded separately by each process.
333 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
334 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
335 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
336 with internal index optind with multiple processes. Only works if MPI_Init has
338 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
339 syntax without preprocessing tricks.
342 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
343 behavior tracing is no longer limited to MSG
344 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
345 this means that traces now need to be updated or re-obtained; both Irecv
346 and Isend lines in your traces have now in total 5 mandatory fields:
347 <rankid> <command> <to/from rankid> <tag> <size>
348 To update your traces, it suffices to add a 0 for the tag here.
349 - TIT now also supports waiting for a distinct request via MPI_Wait.
350 Wait/Test now wait for a specific request, not just the last one that was
351 issued. This unfortunately means another update, because we need to
352 identify which request you want to wait for. We do this via the
353 triplet (sender, receiver, tag), which needs to be added:
354 <rankid> <command> <sender> <receiver> <tag>
355 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
356 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
359 - The deprecation of MSG is ongoing (but this should not impact you).
360 Many MSG functions are now simple wrappers on the C API of S4U. If
361 you wish to convert your code to S4U, find the S4U counterparts of
362 your MSG calls in src/msg/msg_legacy.cpp.
363 - MSG can still be used, but won't evolve anymore.
366 - Allow to run the Link energy plugin from the command line with
367 --cfg=plugin:link_energy
368 - Rename Energy plugin into host_energy
369 - Rename Load plugin into host_load
372 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
373 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
374 - Improve performance when handling timeouts of simix synchros.
377 - Config: the C API is now deprecated (will be removed in 3.23), and
378 the C++ API has been slightly improved.
381 - Fix several build issues on OSX.
382 - Move simgrid_config.h to simgrid/config.h (old header still working)
385 - #143: Setting a breakpoint at a given time
386 - #150: Inconsistent event names in SMPI replay
387 - #258: daemonized actors hang after all non-daemonized actors have completed
388 - #267: Linker error on unit_tmgr
389 - #269: SMPI: tracing of MPI_Wait/all/any broken
390 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
392 ----------------------------------------------------------------------------
394 SimGrid (3.19.1) Released March 22 2018
396 The Fixed ABI Release.
398 As you may know, we are currently refactoring SimGrid in deep.
399 Upcoming SimGrid4 will be really different from SimGrid3: modular,
400 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
402 Our goal is to smooth this transition, with backward compatibility
403 and automatic update paths, while still progressing toward SimGrid4.
405 SimGrid remains open during works: The last pure SimGrid3 release was
406 v3.12 while all subsequent versions are usable alpha versions of
407 SimGrid4: Existing interfaces remain unchanged, but the new S4U
408 interface is budding and the internals are deeply reorganized.
410 Since 2015, we work hard to reduce the changes to public APIs. When
411 we need to rename a public library symbol in S4U, we let your
412 compiler issue an explicative warning when you use the deprecated
413 function. These messages remain for four releases, i.e. for one full
414 year, before turning into an error. Starting with v3.15, your can
415 also adapt to API changes with the SIMGRID_VERSION macro, that is
416 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
418 Starting with this v3.19.1, our commitment to reduce the changes to
419 the public interfaces is extended from the API to the ABI: a program
420 using only MSG or SimDag and compiled against a given version of
421 simgrid can probably be used with a later version of SimGrid without
422 recompilation. We will do our best... but don't expect too much of
423 it, that's a really difficult goal during such profund refactoring.
425 The difference between v3.19 and v3.19.1 is that the former was
426 accidentally breaking the ABI of MSG, while the later is restoring
429 S4U and kernel APIs will still evolve until SimGrid4, with one-year
430 deprecation warnings as currently. In fact, cleaning up these
431 interfaces and converting them to snake_case() is one release goal of
432 v3.20. But don't worry, we are working to smooth this upgrade path.
434 In summary, new projects should start with S4U to benefit of the
435 future, but old MSG projects should still be usable with no change.
437 ----------------------------------------------------------------------------
439 SimGrid (3.19) Released March 20 2018
441 The Moscovitly-cold Spring release.
443 SMPI user-visible improvements:
444 - Support of MPI_Type_create_subarray
445 - Unimplemented MPI calls now abort (instead of issuing a warning)
446 Some harmless calls around errhandlers still only issue a warning.
447 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
448 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
449 is now possible (mmap privatization was refactored to this end).
450 Adding new SMPI instances should also be technically doable now.
452 SMPI internal cleanups:
453 - Removed index notion from SMPI -- no more "getPid() - 1"!
454 In tracing files, actors are called rank-<PROCESS-ID> but these are
455 NOT the real ranks (tracing several communicators is not supported yet).
456 - Deprecate smpi_process_index() (will be removed in 3.22)
457 - Better testing on classical proxy apps, through an external project:
458 https://github.com/simgrid/SMPI-proxy-apps/
461 - Use a graphical TOC to make it easier to find the documentation you need
462 - Revamp the MSG tutorial
465 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
466 - Remove parameter of MSG_process_killall().
467 Resetting the PID was bogus anyway (several actors could have the same PID).
470 - Execution->setHost() can be called after start(), to migrate it.
471 - Comm::test_any() is now implemented.
472 - s4u::Actor now has onCreation() and onDestruction() signals.
473 - Install some kernel header files for the users' plugins and more.
474 Warning, their API is really not stable yet.
475 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
476 include/simgrid/kernel/routing/*Zone.hpp
479 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
481 Build System and other cleanups:
482 - Remove unused run-time parameter "tracing/onelink-only".
483 - SimGrid now works with intel compilers (MC still buggy)
484 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
485 - tesh kills subprocesses on timeout (but not on windows)
488 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
491 - #194: Feature request: simgrid::s4u::Comm::test_any()
492 - #245: migrating an actor does not migrate its execution
493 - #253: Feature Request: expose clusters as objects
494 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
495 - #255: Tesh broken on Windows
496 - #256: Modernize FindSimGrid.cmake
497 - #257: Fix (ab)use of CMake install
499 ----------------------------------------------------------------------------
501 SimGrid (3.18) Released December 24 2017
503 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
505 MSG IS NOW DEPRECATED. Please use S4U for new projects.
506 - Support for MSG should not be removed from SimGrid before 2020,
507 but future evolutions will be limited to the integration of
508 user-provided patches.
511 - s4u::onDeadlock() signal, run before stopping the simulation.
512 - s4u::Exec: asynchronous executions (abstraction of background
514 - s4u::Host->getLoad() returns the achieved speed in flops/s
517 - Link::name() is deprecated and will be removed in v3.21, use
518 Link::getCname() instead.
519 - Mailbox::getName() changed to return a std::string, use
520 Mailbox::getCname() instead to get a char*.
521 - Storage::getName() changed to return a std::string, use
522 Storage::getCname() instead to get a char*.
523 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
524 to have both allocation and deallocation of the map in user space.
527 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
528 The amount of flops remaining to do is a vector, not a scalar.
529 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
530 name implies on both sequential and parallel tasks.
531 - Both changes fix GitHub's #223 using PR #237 as a basis.
532 Thanks Michael Mercier.
533 - Most examples were converted to S4U and hidden elsewhere at tests.
534 Rationale: we still want MSG to work; we want newcomers to use S4U.
537 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
538 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
541 - Switch to the faster dlopen privatization mechanism by default
542 - Documentation improvments
545 - Live migration is getting moved to a plugin. Dirty page tracking is
546 the first part of this plugin. This imply that VM migration is now
547 only possible if one this function is called:
548 - C/MSG: MSG_vm_live_migration_plugin_init()
549 - C/C++: sg_vm_live_migration_plugin_init()
550 - Java: Msg.liveMigrationInit()
551 For C and C++, "simgrid/plugins/live_migration.h" also has to be
555 - Define class simgrid::xbt::Path to manage file names.
556 - Removed unused functions:
557 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
558 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
559 xbt_graph_export_graphviz()
560 - xbt/str.h: xbt_str_join()
561 - Remove unused datatypes:
562 - xbt/heap.h: use std::priority_queue or boost::heap instead
563 - xbt/swag.h: use boost::intrusive::list instead
566 - New link_energy plugin for the consumption of the links.
567 - All of the operations on files and storage contents have been
568 packaged into a plugin (src/plugins/file_system). The current
569 public interface can be found in
570 include/simgrid/plugins/file_system.h
571 To use these functions you now have to initialize the plugin by
572 calling MSG_storage_file_system_init() just after calling
573 MSG_init() or sg_storage_file_system_init() just after creating
577 - Remove the undocumented/untested tag <include>
580 - Remove viva specific tracing as the tool is no longer maintained
583 - #248: Kill of finished processes leads to segfault
584 - #240: xbt_cond_wait_timeout should gracefully return for C apps
585 - #239: please implement signal s4u::onDeadlock()
586 - #230: segfaults when exit() without run()
587 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
588 - #223: MSG_task_get_flops_amount() not working with parallel tasks
589 - #222: Actor::kill() doesn't really kill and segfaults
590 - #221: odd LMM warning when killing an actor
591 - #120: Memory leak when the processes are forcefully killed
593 ----------------------------------------------------------------------------
595 SimGrid (3.17) Released October 8 2017
597 The Drained Leaks release: (almost) no known leaks despite the tests.
599 Even very long simulations will preserve your memory: our testsuite
600 stresses SimGrid for over 45mn on fast machines for a coverage of
601 over 80%, and there is only one single known leak, of about 4kb.
605 - Comm.detach(): start and forget about asynchronous emission. A cleanup
606 handler may be given to free resources if the comm cannot be completed.
607 - this_actor::send(mailbox) is now mailbox->put()
608 - New: simgrid::s4u::Comm::wait_all()
609 - New: Host.execute() for remote executions.
612 - Unused option network/sender-gap is removed.
615 - Deprecate MSG_task_isend_with_matching(): This unused feature
616 really complicates our internals. Will be removed in v3.20.
619 - Improved context termination. It is now possible to free resources, even
620 when a process is forcibly killed.
621 - ContextBoost: add support for Boost versions above 1.61.
624 - Bring back run-time option --cfg=exception/cutpath to remove exception
626 - Removed unused functions:
627 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
628 xbt_str_rtrim(), xbt_str_trim().
629 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
632 - Removed header files obsolete since SimGrid 3.12:
633 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
635 - #3: SD_exit should be made optional
636 - #120: Memory leak when the processes are forcefully killed
637 - #159: Threading test regression in Actor refcounting
638 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
639 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
640 - #186: Actor::killAll() segfaults if some process is blocked on wait()
641 - #191: VM migration and pstate
642 - #192: Updating the energy consumptions of all hosts crashes with VMs
643 - #195: All actors have PID=0 in the logs
644 - #204: Sometimes segfault with thread contexts and mmap privatization
645 - #222: Actor::kill() doesn't really kill and segfaults
646 - #225: Actor::kill() doesn't really kill when victims are doing a join()
648 ----------------------------------------------------------------------------
650 SimGrid (3.16) Released June 22. 2017.
652 The Blooming Spring Release: developments are budding.
655 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
656 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
657 - Java JAR file should be finally fully working on Mac OSX too.
659 XML platforms: Switch to platform v4.1 format.
660 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
661 - <zone> can be used as a synonym for the now deprecated <as>
662 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
663 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
664 - <actor> can be used as a synonym for the now deprecated <process>
665 - state_file and avail_file periodicity is now easier to express
666 (check the documentation) the old behavior should still work.
667 * Storage: not backward compatible, but it seems that nobody ever used it
668 - Remove attribute 'content_type' of <storage_type>: was never used
669 - Make attribute 'model' of <storage_type> optional: for future usage
670 - Remove Bconnection model property: was never (in)validated. Replaced by
671 the maximum of the read and write bandwidth as a resource constraint.
674 - New and Backwards Compatibility break:
675 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
676 When one wants to get the list of tasks whose states have changed during a
677 simulation round, s/he has to allocate and free a dynar and use it as
678 argument to this function. The former SD_simulate (double how_long)
682 - Allow multicore VMs, with the correct sharing (unless you overcommit)
683 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
686 - New function to update the consumption of all hosts at once.
687 - Fix the model for multi-core hosts, linear in the amount of busy
688 cores with an abnormality for pIdle (see paper).
689 BUG: Still not satisfactory for ptask on mono-cores.
692 - The netzone are now available from the MSG API.
693 The old names still work, but are now deprecated.
696 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
697 - New option: smpi/keep-temps to not cleanup temp files
698 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
699 mallocs mode is used, allows to change the size of the fake file used
700 (default 1MB), to potentially limit the number of mappings for large runs.
701 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
702 - Fortran ifort and flang compilers support
703 - New RMA calls supported (experimental):
704 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
705 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
706 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
707 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
708 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
709 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
712 - New: this_actor::isMaestro() returns whether we are in kernel mode.
713 - Behind the scene, ActivityImpl are now automatically refcounted.
714 Removing this nasty bottleneck will greatly simplify our internals.
717 - Replay: New function xbt_replay_action_get():
718 Retrieve the function previously associated to an event type.
719 - DROPPED FUNCTION: xbt_str_varsubst()
720 - DROPPED MODULE: strbuff. We don't need it anymore.
721 - DROPPED MODULE: matrix. We don't need it anymore.
722 - DROPPED MODULE: lib. We don't need it anymore.
724 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
726 SimGrid (3.15) Released March 22. 2017
728 The Spring Release: continuous integration servers become green
729 We fixed even the transient bugs on all target architectures:
730 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
731 Windows, FreeBSD, NetBSD.
733 - Rename NetCards to NetPoints.
734 This was intended to help NS3 users, but that's not a netcard.
735 That's a point in the routing algorithm, let's avoid wrong simplifications.
738 - New: MSG_process_yield(). Stop and yield to other processes.
739 - New: MSG_process_daemon(). Daemon processes are automatically killed
740 when the last non-daemon process terminates
741 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
742 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
743 to make room for the upcoming network energy plugin.
744 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
747 - Ensure that an actor can kill itself with Process::exit()
748 - Kill the obscure NativeException. Nobody want to survive the issues
749 it denotes, so use JniException that is a RuntimeException (not to
750 be caught explicitly).
751 - Partial bug fix in initialization. SimGrid flags on command line were
752 consumed at C level but stayed in the original Java String[] args.
753 This could mess users' args[i] if SG flags were not put at the end of
755 The SimGrid flags are now removed from the Java arguments. However,
756 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
757 if args.length is greater than the number of YOUR OWN ARGUMENTS.
758 It might be if you have --log or --cfg flags in the command line.
759 - Fix numerous memleaks all around the place. In particular, around VMs.
763 - s4u::onPlatformCreated: right before the simulation starts
764 - s4u::onSimulationEnd: right after the main simulation loop
765 - s4u::onTimeAdvance: right after a clock change (time is discrete)
766 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
767 event from the availability_file changes the avail speed.
768 - Links are now usable from s4u
769 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
770 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
771 - New examples: The conversion of MSG examples to S4U has begun
772 - Actors: create, kill, migration, and suspend
773 - Applications: master-worker and token-ring
774 - Action replay: communications and storage
775 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
778 - Backwards Compatibility breaks
779 - The SD_route_get_list and SD_route_get_size functions have been removed.
780 They are replaced by the unique
781 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
782 The route from src to dst is built in the links dynar whose size can
783 be obtained with xbt_dynar_length.
784 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
785 sg_host_route_bandwidth and sg_host_route_latency.
786 Macros ensure the backwards compatibility, but you should fix your code
789 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
790 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
791 - Bump our claim of support from MPI 1.1 to MPI 2.2.
792 We don't support 100% of it, but it should be enough. Contact us if not.
793 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
794 naive and probably inaccurate)
795 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
796 mapped onto a unique small file using some system magic.
797 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
798 - Fortran: cleanups, fixes, support of user-added operations
799 - RMA: MPI_Accumulate are applied in correct order.
800 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
801 - Better support for MPI_IN_PLACE
802 - Support for MPI_Win attrs and keyvals.
803 - Support MPI_Comm_group_create, MPI_Type_size_x
804 - MPI ops should be properly applied to all allowed types
805 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
806 algorithm by a non failing one from mpich.
809 - Kill the fifo data container: we don't use it anymore.
812 - A <cluster> can now be created with different speed values to represent pstates.
813 see examples/platforms/energy_cluster.xml for an example.
815 SimGrid (3.14.159) Released December 28. 2016
817 The previous release was only a crude approximation of a Pi release;
818 this one should be a bit better.
820 - Revert a last minute change that broke on Mac OSX.
821 - Fix the changelog and NEWS files.
822 - Improve the documentation, in particular of the routing module.
823 - Integrate some patches flying around in Debian.
825 SimGrid (3.14) Released December 24. 2016
828 * The whole documentation was reworked and reorganized. There is
829 still much room for improvement here, but we're on it.
832 * We now need python3 for our testing infrastructure.
833 * Model checking needs libevent
836 * Now works on FreeBSD
839 * Privatization now works on FreeBSD
841 * Privatization is now activated by default in smpirun
842 You can turn it off with -no-privatize if something goes wrong.
844 * Call-location tracing for SMPI.
845 You can add the exact location (filename / linenumber) of an MPI call to
846 your trace files and slow down or speed up the simulation between two
847 consecutive calls by using an adjustment file (see the documentation).
849 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
850 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
851 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
852 values were ignored in some cases. The timings of these functions can now
853 be significantly different.
855 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
856 smpi/running-power is renamed to smpi/host-speed
858 * smpi/grow-injected-times option to enable or disable multiplication of the
859 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
860 make simulation less precise (but also much faster).
862 * smpirun script should be (much) faster for large deployments.
864 * SMPI tracing: fixed issue with poor matching of send/receives.
866 * Replay: Fix broken waitall
868 New functions and features
869 * MSG_parallel_task_execute_with_timeout, to timeout computations.
871 Dropped / renamed functions and features
872 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
873 - MSG_mailbox_is_empty() -> Mailbox::empty()
874 - MSG_mailbox_front() -> Mailbox::front()
875 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
876 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
877 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
878 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
879 * Don't pass the free_f parameter to property related functions:
880 - MSG_host_set_property_value()
881 - MSG_as_router_set_property_value()
882 - MSG_storage_set_property_value()
883 * VM properties. Since msg_vm_t are msg_host_t, just use
884 MSG_host_get_property_value() and friends
885 * VM I/O related things:
886 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
887 - Unimplemented save/restore methods
888 * MSG_as_router_get_property_value() was redundant with
889 MSG_environment_as_get_property_value().
890 - Removed MSG_as_router_*propert*() functions
891 - Added MSG_environment_as_set_property_value() for consistency
892 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
893 Well, they are still there for now, but deprecated with a warning.
894 Please switch to xbt_dict_new_homogeneous() before this is removed
896 * Task affinity. Its intended behavior (that was very badly tested
897 and probably not really working) was deceiving what most users
898 would have hoped here.
899 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
900 * Option network/coordinates is now useless and should be dropped.
903 * Added option storage/max_file_descriptors to allow more than 1024 files opened
906 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
907 * Added Dragonfly topology support
910 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
911 EiBps, EBps, Eibps, Ebps for bandwidth.
912 They may become useful to some lucky ones.
915 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
916 * ex/app_tokenring: new example, very similar to the MSG Token Ring
917 * ex/async_waitAll: new example, on asynchronous communications
920 * Memory usage should be decreased for simulations with a large number
921 of processes. This also helps for SMPI.
923 SimGrid (3.13) stable; urgency=low
925 The Half Release, a.k.a. the Zealous Easter Trim.
927 - We removed half of the lines, that were mostly experimental cruft.
928 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
929 (not counting blanks and comments -- according to openhub.net)
930 - The internals are now compiled in C++ (and will soon be clean C++)
931 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
932 59 klines of C, etc. We added only 29 klines of C++ in replacement.
934 * Backwards Compatibility breaks
935 - Removed Lua simulation bindings (switch to C or Java for that).
936 Lua can still be used to describe platforms
937 - Removed Java kernel plug-ins.
938 Will be reintroduced after the ongoing major internals reorg.
940 - the following functions were removed.
941 They were too specific and should be reimplemented in a generic
942 way, with filter function.
943 - MSG_task_listen_from_host
944 - MSG_mailbox_get_count_host_waiting_tasks
945 - MSG_mailbox_put_with_timeout was removed.
946 Please use MSG_task_send_with_timeout instead.
948 - the SD_application_reinit function was removed. It has been a noop for a while.
949 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
950 replaced by a more flexible mechanism.
951 - The following functions thus do not exist anymore
952 - SD_workstation_get_access_mode
953 - SD_workstation_set_access_mode
954 - SD_workstation_get_current_task
955 - Basic estimation functions have been removed but can easily be replaced
956 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
957 - SD_workstation_get_computation_time => amount / sg_host_speed()
959 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
960 Use VM.getSpeed()*load/100 for the legacy behavior.
962 - option enable_tracing was removed. It was not doing anything for a while.
963 - In the ModelChecker:
964 - the model-checker now ptraces the model-checked process which means
965 you cannot use a debugger on the latter anymore (we might make this
966 optional in the feature);
967 - removed soft-dirty page tracking;
968 - remove model-checked side snapshot management,
969 MC_snapshot() and MC_compare_snapshot();
970 - keep the MC_cut() function as a stub (it was not really working
971 in the previous release).
974 * All options are consistently in kebab-case. Old names are kept as alias.
977 * Switch to platform v4 format.
978 - Rename from 'power' to 'speed' the attributes describing the amount of
979 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
980 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
981 - In <host> and <link>, attributes availability and state are gone.
982 It was redundant with state and availability traces, and with peak values.
983 - In <cluster>, attributes availability_file and state_file are gone.
984 It was too complex and unused.
985 - Kill <gpu>. Was not doing anything.
986 - The DOCTYPE points to the right URL:
987 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
988 (the file at this address now documents the changelog since its v1)
989 - A warning is emitted for unit-less values (they are still accepted).
990 - speed. Default: 'f' or 'flops'. Also defined:
991 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
992 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
993 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
994 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
995 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
996 - latency. Default: 's' second. Also defined:
997 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
999 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1000 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1003 * s4u::Host is now the preferred public interface to the Host features.
1004 sg_host_* functions are C bindings to the exact same behavior
1005 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1008 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1009 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1010 They were deprecated since a few years, and probably did not even compile anymore.
1013 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1014 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1015 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1016 encouraged to update their codes. A list of the modified functions can be found at the end of
1017 include/simgrid/simdag.h
1020 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1021 the lambda or closure passed as a parameter will run in kernel mode.
1022 All the callback functions should be rewritten to that interface at some point.
1025 * Reorganizing and cleaning the internals all around the place.
1028 * Remove old default barrier/bcast buggy algorithms (see #18407)
1029 * Various bug fixes to handle more codes
1030 * Remove the need for the --foreground option of smpirun (it is still
1031 accepted for backward compatibility).
1034 * Kill the setset data container: MC don't use it anymore.
1035 * Kill the queue data container: it made more sense with GRAS.
1036 * Kill the xbt_peer_t data type: it's useless without GRAS.
1037 * Kill rm_cb feature of config sets: it was never useful.
1038 * Kill graphxml parsing feature. It was not used.
1039 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1041 - xbt_dynar_sort_strings(), when the content is char*
1042 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1043 They throw exceptions on invalid input;
1044 * C++ support for declaring CLI flags (simgrid::config::Flag);
1045 * class for abstracting different signal backends (simgrid::xbt::signal).
1046 with no external dependencies (we need very simple signals).
1049 * refactoring and cleanup of the code;
1050 * ongoing process to cleanly separate the model-checking algorithms
1051 from the code model-checking support.
1053 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1055 SimGrid (3.12) stable; urgency=low
1057 The Facelift Release.
1060 * Require g++ v4.7 at least to not speak prehistorical C++.
1061 * Require Boost 1.48 (for signal2 component).
1062 * Java must be version 7 at least when activated.
1063 * Builds on Windows again (including Java bindings).
1064 * Tracing is now always enabled (no way to turn it out)
1065 * Remove GTNetS. It was not working anyway.
1066 * Various cleanups in the cmake scripts.
1067 * Move headers around to sort them out on installed systems:
1068 - instr/instr.h -> simgrid/instr.h
1069 - instr/jedule/* -> simgrid/jedule
1070 - simdag/datatypes.h was removed
1071 - simdag/simdag.h -> simgrid/simdag.h
1072 - msg/datatypes.h was removed
1073 - msg/msg.h -> simgrid/msg.h
1076 * Interface improvement:
1077 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1078 - Sanitize the interface in MSG_task_ module:
1079 - Merge two functions that were close enough but misleading:
1080 set_compute_duration(t) -> set_flops_amount(t)
1081 get_remaining_computation(t) -> get_flops_amount(t)
1082 - set_data_size(t) -> set_bytes_amount(t)
1083 get_data_size(t) -> get_bytes_amount(t)
1084 - Massive cleanups in the functions related to the energy
1085 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1086 - New: MSG_host_get_pstate()
1088 - msg/energy/onoff: switching hosts on and off
1090 * Interface improvement:
1091 - Rename Host.isAvail() to Host.isOn()
1092 - Rename Process.currentProcess() to Process.getCurrentProcess()
1093 - Rename Task.setDataSize() to Task.setBytesAmount()
1094 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1096 - #18874: Actually allows the GC to reclaim tasks
1100 - SIMIX_process_throw: raises an exception in a remote process
1101 * Refactoring: Separate sub-modules
1102 - libsmx: the public interface, as libc in a real system
1103 - popping: the strange dance that converts a user request into a kernel handling
1104 - smx_context_*: the virtualization mechanisms that embed the user code
1105 - smx_*: the handling of each simcalls
1106 * Interface cleanup:
1107 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1108 * Rename smx_action_t into smx_synchro_t, making explicit that these
1109 things are used to synchronize processes with their environment.
1110 For example, a communication is a sort of synchronization involving
1111 the communicating processes (that may block until the exchange) and
1112 the platform. The same can be said from computations, etc.
1114 - #18888: segfault when a process ends before its kill_time
1118 - 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.
1119 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1120 - MPI_Type_set_name, MPI_Type_get_name
1121 - MPI_*_c2f and MPI_*_f2c functions
1122 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1123 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1124 - Activate a lot of new tests from the mpich 3 testsuite
1126 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1127 - InfiniBand network model added: Based on the works of Jerome Vienne
1128 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1129 - When smpi/display_timing is set, also display global simulation time and application times
1130 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1131 * Collective communications
1132 - 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
1133 - 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/).
1134 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1136 - Replay now uses algorithms from wanted collective selector
1137 - Replay can be used with SMP-aware algorithms
1138 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1139 - 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.
1141 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1142 - Correctly use loopback on fat-tree clusters
1143 - Asynchronous small messages shouldn't trigger deadlocks anymore
1144 * Energy/DVFS cleanup and improvement
1145 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1146 - new: smpi_get_host_pstate()
1149 - "Full" network optimization flag was broken since Surf++
1150 - Better handling of precision flags in maxmin
1151 - Fix bug causing sometimes "Impossible" errors
1152 - Properly pass cluster properties to included hosts
1153 * Improvement of the Energy plugin.
1154 - Always update the consumption before returning that value
1155 - New property: watt_off to denote the dissipation when the host is off
1156 - New functions getWattMinAt and getWattMaxAt to retrieve the
1157 dissipation of pstates that we are not currently at.
1158 * Java: class NetworkLink renamed to Link
1159 * New function: simcall_process_get_kill_time()
1160 * Massive rename s/workstation/host/
1161 - That's intrusive, but that's good for the project consistency. Sorry.
1162 - Change config option "workstation/model" into "host/model"
1165 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1166 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1167 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1168 - xbt_dirname and xbt_basename for non-POSIX systems
1170 * The model checker now runs as a separate process.
1171 * The model checker runs must now be launched with the new simgrid-mc program.
1172 * Record/Replay: the MC can display a textual representation of a path in the
1173 execution graph. It can then be replayed outside of the model checker.
1175 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1177 SimGrid (3.11) stable; urgency=low
1182 * Normalizing pointers addresses tool for better diff between logs
1185 * Add cloud examples using new VMs
1186 - examples/msg/cloud/two_tasks_vm.tesh
1187 - examples/msg/cloud/simple_vm.tesh
1188 - examples/java/cloud/cloud.tesh
1189 - examples/java/cloud/migration/migration.tesh
1190 * Add java surf examples:
1191 - examples/java/surfPlugin/surf_plugin.tesh
1192 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1193 - examples/java/surfCpuModel/surf_cpu_model.tesh
1194 * Add SMPI+MSG example:
1195 - examples/smpi/smpi_msg_masterslave/
1200 - msg task destroy cancel test
1201 - msg_host on/off test
1202 * Move all tests in testsuite to teshsuite (adding tesh files)
1203 * Restructure teshsuites
1204 - one folder for each kind of test
1205 * Restructure AddTests.cmake
1207 - structure the order of tests (with sections)
1210 * Add virtual machine
1211 - creation of a VM on a PM
1212 - migration of a VM from a PM to another PM
1214 - MSG_process_join(msg_process_t process, double timeout)
1215 - msg_bar_t MSG_barrier_init(unsigned int count)
1216 - int MSG_barrier_wait(msg_bar_t barrier)
1217 - void MSG_barrier_destroy(msg_bar_t barrier)
1218 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1219 * New option "msg/debug_multiple_use" to help debugging when a task is used
1222 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1224 - Increase code coverage in test suites
1228 * Protect context stacks against stack overflow. The number of protected memory
1229 pages allocated on the top of each stack (1 by default) can be configured
1230 with the new command line option --cfg=contexts/guard_size:<value>.
1231 * Simcalls are now generated by a python script that
1232 - generates files included by SimGrid
1233 - checks that all the functions exist, and proposes prototypes
1235 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1236 host_set_data, host_get_data
1238 - simcall_process_join(smx_process_t process, double timeout)
1239 * Fix bug where sleeping processing could not be suspended.
1242 * Translate surf models from C to C++
1243 - Generic classes for all models: Model, Resource, Action
1244 - A generic interface for each kind of model (CPU, Network, Storage
1245 Workstation, WorkstationVM)
1247 * Translate surf routings from C to C++
1248 * Add callbacks using sigc++ or boost::signals2
1249 - Add callback functions for resource creation/destruction
1250 - Add callback functions for action state change
1251 - Handle Energy as a plugin
1252 * Replace swag by boost::intrusive
1253 * Add new routing models for clusters. For documentation, see
1254 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1255 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1256 parameters for cluster tag
1257 - Fat trees, with topology="FAT_TREE" and
1258 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1259 - see examples/platforms/torus_cluster.xml and
1260 examples/platforms/fat_tree_cluster.xml
1261 * More documentation
1264 * Hostfiles support host:nb_processes construct to deploy several processes on
1266 * Collective communication algorithms should not crash if used with
1267 improper number of nodes and report the error.
1268 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1269 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1270 MPI_Dims_create, MPI_Cart_sub are supported.
1271 * New interface to use SMPI programmatically (still depends on MSG for
1272 some parts, see examples/smpi/smpi_msg_masterslave):
1273 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1277 * Global variables privatization in MPI executables is now performed at runtime
1278 with the option smpi/privatize_global_variables (default:no).
1279 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1280 dynamic libraries loaded by the application are not privatized (static
1281 linking with these libraries is advised in this case)
1284 * Options defined in XML work correctly now.
1287 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1288 copied into simgrid.jar or not (ON by default). Use this option if you want
1289 to reduce the size of the installed simgrid.jar, *and* the native libraries
1290 are kept installed elsewhere.
1291 * Surf binding with SWIG (code generated in maintainer mode only):
1292 - plugin to handle callbacks
1293 - CPU model only for the moment
1296 * Supernovae build mode is definitively removed. It was used to improve
1297 inlining and inter-module optimizations. It is nowadays superseded by
1298 link-time optimizations commonly available in compilers.
1299 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1300 * Add boost dependency for surf++
1301 * Add new macro for tests
1302 - ADD_TESH(name <tesh_args>)
1303 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1307 - xbt_bar_t XBT_barrier_init(unsigned int count)
1308 - int XBT_barrier_wait(xbt_bar_t barrier)
1309 - void XBT_barrier_destroy(xbt_bar_t barrier)
1310 * Make the xbt_os_time module public
1312 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1314 SimGrid (3.10) stable; urgency=low
1316 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1319 * Reintegrate Java to the main archive as desynchronizing these
1320 package is not acceptable anymore (Java is now considered stable)
1321 * Add explicit synchronization facilities through semaphores
1322 * Bug fix: Task.setDataSize() only changed the C world, not the value
1323 cached in the Java world
1326 * Dramatically change the way files are handled. API and internals changed, but
1327 this part of MSG was not considered as production grade either.
1328 * Add explicit synchronization facilities through semaphores
1329 * Add a new function MSG_host_get_process_list()
1330 * Preliminary DVFS support (see examples/msg/energy/ for details)
1333 * SMPI is now included directly in the libsimgrid as the windows
1334 linker doesn't force us on splitting it anymore.
1335 * Improvements of the SMPI replay tool:
1336 - Most of the collective communications are now rooted in the same process as
1337 in the original application.
1338 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1339 used until now). Multiple data types can now be used in a trace.
1340 - The replay tool now supports traces produce either by TAU or a modified
1342 - Bug Fix: the compute part of the reduce action is now taken into account.
1343 - Gatherv collective is now supported
1344 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1345 of smpirun outputs time independent traces for the current run. One file
1346 is created per process. If too many processes are simulated, this behavior
1347 can be changed to one file for all processes by using the
1348 tracing/smpi/format/ti_one_file flag
1349 * smpirun generates the host file if needed (with given host count and platform)
1350 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1351 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1352 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1353 --cfg=smpi/coll_selector:(mpich/ompi)
1354 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1356 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1357 happening inside a collective SMPI call.
1358 * Fix the behavior of complex data types handling.
1359 * Make MPI_Wtime another synchronization point to take computations into
1361 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1362 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1363 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1364 (no privatization of global variables yet)
1365 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1368 * Verification of liveness properties is now available for SMPI applications
1369 (in addition to MSG applications)
1370 * Bugged examples using SMPI in examples/smpi/mc/
1371 * Add --cfg=model-check/visited option. Allows the verification of infinite
1372 programs. Detection of loops in the execution thanks to the system state
1373 comparison and reduction of the state space to explore. Can be combined with
1374 DPOR for safety properties.
1377 * Allow to change SimGrid configuration (see --help) within the code
1378 thanks to SD_config() as it can be done in MSG.
1379 * Add a new function SD_task_set_amount() upon user request.
1382 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1383 * Remove rule based routing (no more PCRE dependency)
1384 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1385 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1386 * Add a loopback_bw and loopback_lat options to cluster tag.
1387 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1388 routing. Now peers are encapsulated in an AS and have their own private
1389 router but this is transparent.
1392 * Our own implementation of getline is renamed xbt_getline, and gets
1393 used even if the OS provide a getline(). This should reduce the
1394 configuration complexity by using the same code on all platforms.
1395 * New type: xbt_cfg_elm_boolean.
1396 * Allow to use yes/no for boolean configuration options in the command line.
1397 * Allow to disable SimGrid cleanups at exit from command line option.
1398 There are situations where one may want a simulation to end with an exit.
1399 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1400 annoying when scripting around the simulator. Adding a
1401 --cfg=clean_atexit:no allows to circumvent this issue.
1404 * Lots of memory leaks were corrected in this release.
1405 There is no known memory leaks anymore, in all of our 600+ tests.
1406 * New command line option --version, to get SimGrid version information.
1407 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1409 * Supernovae builds are deprecated, and expected to be removed in the next
1412 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1414 SimGrid (3.9) stable; urgency=low
1416 The Grasgory release: GRAS is really dead now.
1418 * Complete overhaul of the internal host structures scheme.
1421 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1422 was considered as experimental and badly maintained since then.
1423 * Keeping it was thus a trap to our potential users, that could take
1424 it instead of MSG or SMPI by mistake despite is pity state.
1425 * GRAS seems to have very few users (if any), and no one volunteered
1426 to maintain it further. It also induces a lot of XBT code (for
1427 portability sake), that must be maintained too.
1428 * For all these reasons, we killed GRAS. If someone wants to revive it
1429 in the future, don't cry, our git history still remembers of GRAS.
1432 * Major overhaul. Merge our documentation again as time proved that
1433 splitting it was really not helping our users.
1434 * Further improve the developer documentation to help newcomers
1435 hacking on SimGrid itself. The user documentation (and in
1436 particular, the beginner documentation) is still in a sorry state.
1439 * Now works on Windows too!
1440 * Much more extensive test suite, from MPICH
1443 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1444 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1445 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1446 * Bug fix: let task be scheduled when the last dependency to be solved is
1447 a control dependency.
1448 * Remove SD_load_environment_script function.
1449 Use the C sg_platf function if you want to declare a platform
1453 * New function: MSG_process_get_number()
1454 * Old function documented: MSG_config()
1455 * Remove MSG_load_platform_script function
1456 Use the C sg_platf function if you want to declare a platform
1460 * Change the default value of the TCP_gamma constant (maximal size of TCP
1461 congestion window) to a more realistic 4MiB value. If you notice changes in
1462 your simulation results, you can fall back to the previous 20k tiny window
1463 by adding --cfg=network/TCP_gamma:20000 on command line.
1464 * (Hopefully) fix a bug wrt periodic availability and state traces
1465 * Bug fix: use default values at start when first event in availability/state
1466 trace is not at time 0.
1469 * remove the "new_" part of function name sg_platf_new_trace_connect
1470 (resulting in sg_platf_trace_connect), since it does not create
1474 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1475 use case where it's really mandatory, and maintaining it was a pain in
1477 * New: xbt_fifo_search(), search an item with a user-provided
1478 comparison function instead of dumb pointer comparison.
1481 * Fix the lua deployment:
1482 Use `simgrid.init_application()` before deployment instead of
1483 `simgrid.msg_register_application()` after.
1486 * Transfer the tracing files into the corresponding modules.
1488 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1490 SimGrid (3.8.1) stable; urgency=low
1492 The "we are told that some people want to also *install* the simgrid
1495 * Add missing file "tesh.1" to the archive.
1497 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1499 SimGrid (3.8) stable; urgency=low
1501 The Psssshiiiit release: SimGrid jumps into the Cloud.
1504 * Add an experimental interface to manipulate VMs. They are mainly
1505 process groups with very few intrinsic semantic, but they should
1506 allow you to build the semantic you want easily.
1507 * New function: MSG_host_set_property_value()
1508 * New function: MSG_process_on_exit(). To clean memory in all cases.
1509 * Bug fixes that made the host (and link) failures unusable.
1510 * Add a way to auto-restart process when the host in which they are
1511 executing comes back (ON_FAILURE="RESTART" on deployment file,
1512 MSG_process_auto_restart_set).
1513 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1514 please stop using the old ones, they are DEPRECATED.
1516 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1517 Please use MSG_init() instead. (reducing the amount of entry
1518 points in the library helps us).
1519 * Make it impossible to link against the wrong version of the lib
1520 * Deprecate MSG_clean(). No need to call it anymore.
1521 * Function MSG_get_host_number() is not deprecated anymore.
1524 * Split the doc into a user guide and a reference guide.
1525 * Start a developper guide to help people hacking on SimGrid.
1528 * Enable tracing by default. This modules rocks you should use it.
1529 * Remove option custom_flags. Now use environment variables CFLAGS
1531 * Use default cmake things to detect lua instead of home grown ones.
1532 * New option "enable_mallocators" to disable mallocators, for debugging
1533 purpose ("on" by default).
1536 * Bug fixes around the resource failures: don't let the processes
1537 survive the host they are running onto.
1538 * Add an interface to auto-restart processes when the host in which they are
1539 executing comes back.
1540 * Ensures that SIMIX_clean is called automatically. It's not part of
1541 the public interface anymore (bindings should be updated).
1544 * Bug fix for when SD_Simulate is called with a positive value: be careful
1545 when comparing doubles. Sometimes they are different for non significant
1547 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1548 parallel task whose initial work is distributed among host according
1549 to the Amdahl's law. Such tasks are created with a parameter alpha
1550 that corresponds to the non-parallelizable part of the computation.
1551 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1552 between two sets of workstations assuming a 1D block distribution (each
1553 workstation owns a similar share of data) on both sides.
1555 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1556 Data redistribution will be automatically scheduled once parent and child
1557 are both scheduled. The filling of computation_amount and
1558 communication_amount structures is now done seamlessly thanks to the chosen
1560 * New function SD_workstation_dump to display various information
1561 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1562 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1563 bandwidth on the route onto which the task is scheduled and the amount of
1565 To divide the nominal bandwidth by 2, the rate then has to be:
1566 rate = bandwidth/(2*amount)
1567 * Compute tasks that have failed can now be rescheduled and executed again
1568 (from their beginning)
1569 * Increasing source code coverage (src/simdag is now covered at 95.8%
1573 * Re-implement time-independent trace replay using SMPI (at the
1574 smpi_smp_* level) instead of MSG. This should replace
1575 examples/msg/actions/actions.c
1576 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1577 hindexed and structs)
1578 * Implement the exchange of non-contiguous data.
1579 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1580 * Correct behavior of smpi/sender_gap and set its default value to 0
1581 * Add option to asynchronously send small messages to allow better
1582 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1583 specifies the size in bytes under which messages will be asynchronously sent.
1584 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1585 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1588 * New C interface to define a platform: XML is now optional.
1589 For more info, please check include/simgrid/platf.h.
1590 * New interface to define random platforms from the C:
1591 For more info, please check include/simgrid/platf_generator.h and
1592 examples/msg/masterslave/masterslave_platfgen.c
1593 * Export a sg_cmdline dynar containing all the arguments we got from
1597 * Two new tracing options for adding comments to trace file so you
1598 can track your experiments (see --help-tracing for details).
1599 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1600 * Adding the SimGrid version that generated the trace file as a comment.
1601 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1602 * Fix to avoid key clashes on Paje links
1603 * Other minor fixes related to the Paje specification
1606 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1607 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1608 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1609 * Greatly improve the robustness of mmalloc to user errors (such as
1610 using an area after freeing it, or freeing it twice)
1612 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1614 SimGrid-java (3.8.1) stable; urgency=low
1616 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1617 Warning: all methods are visible, but only some of them are
1618 implemented so far. Check the source (src/jtrace.c)
1619 for further information.
1620 * New module: org.simgrid.msg.File (SimGrid File management functions)
1621 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1622 * Change the meaning of Process.restart: now restart the process from
1623 the begining, like MSG_process_restart in C.
1624 * Add Process.setAutoRestart: handling of process restart when failed
1626 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1627 you to retrieve the properties of the processes/hosts
1628 * Deprecate Msg.clean(): you can just forget about it now.
1629 * New function Process.getCount(), that only works when compiling
1630 with the not yet released version 3.9 of the C library.
1632 * New context factory based on Coroutines. It mandates a modified JVM
1633 but then, the simulations run about five times faster, and there is
1634 no limit to the amount of processes (beside of the available memory).
1636 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1638 SimGrid (3.7.1) stable; urgency=low
1641 * Restore the prototype of MSG_process_create_with_environment() to
1642 the pre-3.7 situation by removing the kill_time argument.
1643 * Add a MSG_process_set_kill_time() function instead.
1646 * Fix weird behaviors when dealing with parallel tasks.
1649 * Simgrid is now built as a dll.
1650 * Simgrid-java now works on Windows.
1651 * Simgrid-Java is now included into Windows package.
1654 * First pre-build package for MacOSX.
1657 * Fix compilation when using MSG_USE_DEPRECATED.
1658 * Fix some compilation issues on Macs and Windows.
1659 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1660 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1662 We discovered that the Lua console is broken, but we are missing the
1663 manpower to fix it right now. The problem existed in 3.7 too, so we
1664 are not blocking the release for that. Sorry if you depended on this
1665 feature, any help would be really welcome.
1667 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1670 SimGrid-java (3.7.1) stable; urgency=low
1672 The "Java aint got to be bloated and slow" release
1675 * Various internal cleanups and performance improvement
1676 Simulations are expected to run up to twice faster or so
1677 * Make Process.kill(process) an instance method, not a static one
1678 * User processes are not java.lang.Thread subclasses.
1679 This breaks the compatibility (sorry), but previous API was
1680 brain-dead, making it impossible to have non-trivial
1681 initializations in the process constructor.
1682 * Require a full constructor per Process sub-class.
1683 Kinda breaks the compatibility (sorry), but this allows a much more
1684 efficient way to launch the processes at simulation startup.
1685 * Do not embeed our version of semaphores, java 1.5 can be considered
1686 as sufficiently prevalent for us to not dupplicate its features.
1690 * Add examples for almost every part of the API
1691 We spotted and fixed a lot of bugs in the process
1693 * New module: asynchronous communication API
1694 * New function: Process.sleep()
1695 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1696 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1697 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1699 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1701 SimGrid (3.7) stable; urgency=low
1703 The "spring cleanups (before the next Big Project kicks in)" release.
1706 * We can specify the SMPI latency/bandwidth factor with command line
1707 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1708 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1709 You can also use the "config tag" from platform file by adding this line
1710 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1711 (see "example/platforms/tag_config.xml" to use "config tag").
1712 Note that the command line supersedes the platform file configuration.
1713 * Change the correction factors used in LMM model, according to
1714 the latest experiments described in INRIA RR-7821.
1715 Accuracy should be improved this way.
1716 * Use the partial invalidation optimization by default for the
1717 network too. Should produce the exact same results, only faster.
1718 * Major cleanup in surf to merge models and split some optimization
1719 mechanisms from the core of the models. As a result you can now
1720 specify which model to use (e.g., --cfg=network/model:LV08
1721 --cfg=cpu/model:Cas01) and which optimization mode to use
1722 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1723 Incompatible combinations should err at initialization. See
1724 --help-models for the list of all models and optimization modes.
1725 * The CLM03 workstation model was dropped for simplicity because it
1726 used the deprecated CM02 network model. Use default instead.
1727 * Rename the TCP_gamma configuration option to network/TCP_gamma
1728 * Rename the coordinates configuration option to
1729 network/coordinates, and document it
1730 * Use now crosstraffic keyword instead of the terribly misleading
1731 fullduplex keyword. It is activated by default now in the current
1732 default model, use --cfg=network/crosstraffic:0 to turn it off.
1733 * Ongoing refactoring the model parsing to make XML files optional
1734 See include/simgrid/platf.h for details (still to be completed)
1737 * Major overhaul of the documentation. Almost instructive now :/
1738 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1739 functions and friends. This interface was considered as
1740 deprecated since over 2 years, it's time to inform our users that it is.
1741 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1742 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1743 use these (crappy) functions in your code.
1744 These functions will be removed soon. Stop using them now.
1745 * Deprecate MSG_get_host_{table,number}
1746 Implement MSG_hosts_as_dynar() instead.
1747 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1748 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1751 * Stabilize the parallel execution mode of user contexts
1752 * Introduce configuration variables to control parallel execution:
1753 - contexts/synchro: Synchronization mode to use when running
1754 contexts in parallel (either futex, posix or busy_wait)
1755 - contexts/parallel_threshold: Minimal number of user contexts
1756 that must be part of a scheduling round to switch to parallel
1757 execution mode (raw contexts only)
1758 * Fix bugs that prevented to use suspend/resume along with
1759 synchronization structures.
1760 * Fix bugs in process termination that lead to invalid memory access
1761 in very specific conditions.
1764 * Introduce a parallel mode for the models (controlled by surf/nthreads
1765 configuration item). In our tests, running the models in parallel
1766 never lead to any speedups because they are so fast that the gain
1767 of computing each model in parallel does not amortizes the
1768 synchronization costs, even when ultra fast futexes are used.
1769 This is released anyway because YMMV.
1772 * Performance boost by using a swag internally to compute the set of
1773 tasks that are finished and should constitute the return value of
1777 * Enable it by default now that it is considered rather stable.
1780 * Documentation of the tracing functions.
1781 * Performance gains when tracing categorized/uncategorized resource
1782 utilization by avoiding calls to get route when updating resource
1783 variables. LMM constraints are being used instead.
1784 * API changed to set task categories. Use MSG_task_set_category instead
1785 of TRACE_msg_set_task_category, and SD_task_set_category instead
1786 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1787 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1788 * Examples for link user variables added (see at examples/msg/tracing/)
1789 * Deprecated function TRACE_msg_set_process_category completely removed
1790 * Trace header updated according to the latest Paje file format
1791 * Tracing network lazy updates, no longer obligate users to use full updates
1792 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1793 * Experimental: let user code declare/set/push/pop application states for hosts
1794 * API changed to allow the manual creation of graph configuration files
1795 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1798 * Improve the API of Lua MSG bindings, using the Lua spirit.
1799 * Each simulated process now lives in its own Lua world (globals are
1800 automatically duplicated). It helps writing simulators. Will allow
1801 to run Splay programs within SimGrid in the future.
1802 * Add a Chord example in Lua, equivalent to the MSG one.
1805 * Start the implementation of a solution to express temporal
1806 properties, not only local assertions. This is still an
1807 experimental work in progress, stay clear from it to be safe.
1811 - Add new runtime parameters --help-logs and --help-log-categories
1812 to display information about supported logging parameters and
1814 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1815 don't exists anymore.
1816 * Mallocators: allow value NULL for the reset function.
1818 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1819 create homogeneous dictionaries, where all the elements share the
1820 same free function. Non homogeneous dictionaries will be
1821 deprecated in the next release.
1822 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1823 - Multi-level dictionaries are deprecated.
1825 - new function xbt_dynar_search_or_negative() that is useful when
1826 you have less than 2 million elements in your dynar and don't
1827 want of the extra complexity of catching exceptions when the
1828 element is not found.
1830 - Make xbt_os_thread module (for thread portability) public.
1831 Documentation is still to come, sorry.
1833 - Cleanups and simplifications to make it maintainable again.
1834 - Exotic features (such as memalign and valloc) were removed.
1835 - The metadata were extended and improved so that the
1836 model-checker becomes able to explore and inspect the heaps.
1837 - This may induce a performance drop when enable_model-checking is
1838 ON in cmake (even if it's not used in the simulation), but it is
1839 necessary at this point to get MC working.
1841 Turn model-checking OFF if simulation performance matters to you.
1842 Not enabling it at runtime is not enough, disable it in cmake.
1844 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1847 SimGrid (3.6.2) stable; urgency=low
1849 The "Not coding new stuff allows to polish old things" release.
1852 * New bindings to the NS3 packet level simulator (experimental)
1853 * Use the raw (efficient) execution contexts instead of the sysv
1854 (portable) ones when possible.
1855 * libpcre is now mandatory in any cases since not using it led to
1856 severe performance loss and possibly other issues
1857 * Update the XML platforms:
1858 - G5K: include the latest machine in Nancy
1859 - GridPP and LCG: new platforms
1860 * Documentation was partially updated, at least (more to come)
1862 Bug fixes, cosmetics and small improvements
1863 * Free terminated processes before the end of the simulation to avoid
1864 exhausting the memory of users having very dynamic amount of
1866 * Bug fix and cosmetics about canceling non-running tasks
1867 * Bug fix about the dot loader's issues when using libcgraph
1870 * Create an installer for windows with nsis (amd64 and win32)
1871 - Add an hello world project to illustrate simgrid project creation.
1872 - Embed libpcre into the Simgrid installer to avoid
1873 its compilation burden
1874 * The raw execution contexts should work on Apple now
1875 * Port to Windows 64 bits
1876 - Sysv contexts now have an implementation for this arch
1877 - GRAS communication features now support this arch
1878 * Drop support for borland compiler on windows
1879 - this code was not maintained, and we kinda depend on gcc nowadays
1880 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1881 * Fix portability issue on unstable ubuntu: linker became picky on
1884 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1886 SimGrid (3.6.1) stable; urgency=low
1888 The "Oops, we broke Macs too" release
1891 * Fixed contexts detection so that raw ones are used when possible
1892 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1893 leads to a strange error, with user code segfaulting sometimes when
1894 the generated code is not perfectly aligned (which is not
1895 controllable from the user side, depends on the amount of code)
1898 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1900 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1902 SimGrid-java (3.6) unstable; urgency=low
1905 * Split of every thing from simgrid v3.5 into a separate package.
1907 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1909 SimGrid (3.6) stable; urgency=medium
1911 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1914 * Bindings now constitute their own package, separated from the main one.
1915 Rationale: reduce our maintenance nightmare by reducing the module coupling
1916 They will soon be released on their own on gforge.
1918 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1919 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1921 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1922 * It's not quite deprecated for now because we have no replacement,
1923 but it may soon become the case.
1926 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1927 * Fortran: New user-level cache variable to store the rank of the running
1928 process. This improves performance by an order of magnitude.
1929 * C: New coccinelle script to automatically locate and modify global and
1930 local static variables.
1931 * Improved SMPI network model with a sender-side gap to account for multiple
1935 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1936 only say if a communication is finished, no matter whether it succeeded or
1937 failed. You can call MSG_comm_get_status() to know the status of a finished
1939 * New function MSG_task_dsend() to send a task and detach it. When a
1940 communication is detached, you are never notified of its success or failure
1941 and the memory is released automatically once it is finished. This function
1942 is useful when you don't care about the end nor the success of a
1944 * Change the prototypes of action replay. Sorry for inconvenience,
1945 but this is really more efficient this way (and to adapt your code,
1946 you just have to fix the initialization, that shouldn't be too long)
1947 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1948 ever managed to do anything useful with it.
1949 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1950 takes a dynar of communications. It returns immediately and gives the
1951 index of a finished communication (if any).
1952 * New example: a basic implementation of the Chord P2P algorithm.
1955 * New model for multi-core CPUs. You can now use the core attribute to
1956 precise the number of cores of a host. This is a basic model. Every
1957 process running on the host receives at most the power provided in
1958 the DTD (throughput<=power). Total throughput of process cannot exceed
1960 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1961 router linked by an up-link and a down-link (possibly asymmetrical).
1962 This kind of pattern allows to easily build last-mile model style platforms.
1963 Aggregating such patterns in a rule-based AS is thus the technique of
1964 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1965 * New model for Vivaldi routing. We transformed the Vivaldi network model
1966 into a Vivaldi routing model (based on the rule-based model). This allows to
1967 combine Vivaldi based latencies with last-mile platforms.
1970 * Added a check for NaN of IEEE754 infinite in the double entries of
1972 * Introduce a new context factory "raw", highly inspirated from the
1973 ucontext factory, but using manually crafted functions in assembly to
1974 do the work in an efficient manner.
1975 * Allow to change the used context factory at run time, not only at
1976 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1977 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1978 contexts at runtime (only with raw contexts or ucontexts).
1979 * Completely rewrote this module to allow parallel execution of user
1980 processes. Use --cfg=contexts/nthreads:N to execute user processes
1981 with N parallel threads (the default is 1, meaning no parallelism).
1982 * Allow to decide dynamically between sequential and parallel modes.
1983 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1984 processes in parallel only when their number is greater than or equal to P
1986 * Added a check for NaN of IEEE754 infinite in the double entries of
1990 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1991 won't output the state of processes when interrupted with Ctrl-C
1992 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1993 NULL-terminated array. This may solve backward compatibility issues
1994 due to the change to return type of SD_simulate. See also:
1995 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1996 * Add new macros with variable number of arguments.
1997 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1998 - in xbt/asserts.h: xbt_assert
1999 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2000 - in xbt/ex.h: THROWF and RETHROWF.
2001 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2003 * Change xbt_die() to accept a format string with arguments, just like printf.
2004 * New data structure: xbt_lib_t, like a dict but more general and with better
2008 * New configuration options
2009 Options triva/categorized and triva/uncategorized can be used to generate
2010 graph configuration files for Triva visualization tool.
2011 * Configuration option tracing/platform is renamed to tracing/categorized
2012 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2013 * New TRACE_user_link_variable interface:
2014 User provides the name of the link and the tracing variable to attach to it
2015 * the declaration of tracing categories must be done after the environment
2017 * simpler tracing interface, just one way to declare categories
2018 TRACE_category or TRACE_category_with_color, it is up to you
2019 * links in the trace file are again identified by their names
2020 * trace contains the full platform hierarchy exactly as declared using the ASes
2021 * Options tracing/msg/[task|process]:1 groups the process by hosts
2022 for both cases, tasks and processes must have names that are unique during the simulation
2023 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2024 * The experimental option tracing/msg/volume is deprecated
2025 its functionality may be reincorporated if needed
2027 The tracing generates a trace file with unordered timestamped events,
2028 because of the way the core simulator (surf) works. A script available
2029 at the tools directory (fix-paje-trace.sh) can be used to put the events
2030 in order. We have changed the tracing so it can generate ordered timestamped
2031 events in the final trace, but depending on the simulator (and how much time
2032 is simulated) that can lead to a huge memory utilization. It is deactivated
2033 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2035 Build Infrastructure
2036 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2037 - We are trying hard to keep the API stable, but it may happen that
2038 some things change (we're a research project after all, not a
2039 nuclear plant operating system). If such things should happen, you
2040 could rely on that macro to adapt.
2041 - current value: 30600 for 3.06.00, aka 3.6
2042 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2043 a number that can be compared with SIMGRID_VERSION.
2044 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2045 assertions and verbose/debug logging events are disabled at compile time.
2047 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2049 SimGrid (3.5) stable; urgency=medium
2052 * New feature to any SimGrid-based simulator: Model-Checking
2053 Check SIN#1 for more details.
2056 * New Model SMPI (three-interval linear regression for correction factors)
2057 See RR-7426, available at http://hal.inria.fr/inria-00527150
2058 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2059 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2060 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2061 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2062 * See also src/smpi/README
2067 - Tracing API changes: TRACE_start and TRACE_end should not be called
2068 by user-code. They are automatically called by simulators created
2069 with SimDAG, MSG and SMPI if the toolkit is compiled with
2070 tracing_enabled=ON. Categories declaration and utilization remain the
2071 same for MSG and SimDag.
2072 - A function was added to the tracing API to declare categories with
2074 - TRACE_category_with_color (char *category, char *color)
2075 where color must be in the following format
2076 "%f %f %f", red, green, blue
2077 and red, green, blue are float values in the interval [0, 1]
2078 - User can specify NULL as color parameter, or continue calling
2079 TRACE_category (cat)
2080 On that case, the tracing system will define random colors
2081 - The following command-line options are supported:
2082 --cfg=tracing/filename:msg.trace
2083 --cfg=tracing:1 (activate tracing, needed to use others)
2084 --cfg=tracing/platform:1 (categorized resource use)
2085 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2086 --cfg=tracing/msg/task:1 (task creation)
2087 --cfg=tracing/msg/process:1 (process creation, migration)
2088 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2089 --cfg=tracing/smpi:1 (SMPI interface tracing)
2090 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2091 - examples of examples/msg/tracing updated
2093 - DAXLoader and DOTLoader functions can generate tasks with categories
2094 - A new function to attribute a category to SD tasks:
2095 TRACE_sd_set_task_category (SD_task_t task, char *category)
2096 * Tracing the MPI interface implemented by SMPI
2097 - Collective operations are traced with states
2098 - Point-to-Point operations are traced with states/links
2099 - Tracing activated by a parameter "-trace filename" passed
2100 to smpirun during execution (considering that simgrid
2101 is compiled with tracing enabled)
2102 - To run the simulation with gdb, the simulator
2103 accepts --cfg=tracing/smpi:1 to trace SMPI
2104 - tesh files to check if smpi tracing is ok
2105 - See examples/smpi/NAS/DT-trace
2106 * GTNetS tracing re-worked
2107 - adaptation to the tracing system of GTNets to cope
2108 with modifications regarding the fullduplex mode
2109 - new tesh files to check if gtnets tracing is ok
2112 * Asynchronous communications through the functions:
2113 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2114 * New function: MSG_load_platform_script()
2115 to make possible using a lua script instead of XML files to set up platforms
2116 * New function: MSG_set_function
2117 to associate functions to processes, used when bypassing the parser
2118 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2120 Platforms: Add some more examples in examples/platforms
2121 * Grid'5000: see www.grid5000.fr
2122 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2125 * Change the XML format. This is a very important modification. SimGrid
2126 3.5 introduces a new hierarchical format based on the notion of
2127 Autonomous Systems. Compatibility with old format is ensured
2128 through the perl script provided in the install bin directory
2129 bin/simgrid_update_xml.
2130 It is now possible to build platforms with specific routing
2131 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2132 connect several platforms together. We will try to provide soon
2133 set of realistic platforms exploiting these properties (have a
2134 look at examples/platforms/ for the moment).
2135 * Take the opportunity of the XML format change to be a good XML citizen:
2136 rename link:ctn to link_ctn and similar changes (also dealt with by
2138 * Add a new routing scheme (rule-based) using regular expressions. It
2139 enables to have an extremely low memory footprint when the
2140 underlying routing is simple and can be compactly described. You
2141 need to have libpcre4-dev (perl regular expressions) installed if
2142 you want to use this routing scheme.
2143 * Revive the cluster TAG and allow to easily and efficiently (both in
2144 term of memory and speed) connect clusters together. Have a look
2145 at teshsuite/simdag/platforms/ to see how this can be done. With
2146 this tag, you can create clusters with thousands of tasks at no
2147 cost (have a look at examples/platforms/).
2148 Note: clusters are implemented as ASes, so there is no need for an
2149 enclosing AS tag if you have only one cluster in your platform.
2150 * Add new generic functions in the public interface that allows the user
2151 to call SURF 'create_resource' methods from your code (same
2152 functionality as the XML bypass mechanism but with a much lighter
2154 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2155 that uses a piecewise linear approximation to produce better
2156 results when exchanging small messages.
2157 * Add a new parameter to handle correctly full duplex link and account
2158 for interferences between uplink and downlink communications
2159 (activate with --cfg=fullduplex:1).
2162 * Rename the SD_READY (all dependencies are satisfied and task is
2163 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2164 dependencies are satisfied) state.
2165 This prevents a confusion between the notion of "ready to schedule"
2166 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2167 simulated" (SD_RUNNABLE) used by the simulation kernel.
2168 * Change the way a task is considered as ready. Instead of removing
2169 dependencies when a task is done, a counter is decreased. This way,
2170 it is always possible to reach ancestors thanks to the
2171 SD_taks_get_parents function (even after the end of the simulation.)
2172 * Change the return type of SD_Simulate from (SD_task_t*) into
2173 xbt_dynar_t. This function was in handling a dynar internally and
2174 converted it into a NULL terminated array for historical reasons.
2175 * New function SD_dotload(char*) to load a DAG described in dot
2176 format. This loader and the corresponding examples require the
2177 installation of the graphviz library.
2178 * Fix a bug in the management of tasks of size 0 in the surf network
2179 models. This problem was only visible with SIMDAG and you should
2180 thus disregard results produced with earlier versions if you
2181 relied on this feature (some tasks were blocked because of this).
2182 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2183 even though your applications were doing only point-to-point
2184 communications and sequential computations. Now you can really use any
2185 model you want (of course, if you create real parallel tasks, which are
2186 not implemented in most models beside ptaskL07, this will abort).
2187 * Add an example that schedules a DAX on an heterogeneous platform
2188 using a Min-Min strategy.
2189 * New function SD_workstation_get_current_task() that returns the kind
2190 of task currently running on a workstation in the sequential access
2192 * Raise some warnings when unexecuted tasks remains at the end of the
2193 simulation. This is usually caused by cycles in the DAG.
2196 * New function: SIMIX_process_set_function() called by MSG_set_function
2197 * Change the underlying waiting queue in semaphores so that a process
2198 can wait on several of them simultaneously (as in waitany).
2199 * Fix the way to handle tokens in semaphores so that all access patterns
2200 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2201 * kill the dirty pimple SIMIX_message_sizes_output()
2202 Please use (proper) visualization instead
2205 * New data container: setset (set of sets of elements)
2206 * New module: mmalloc (mapped malloc, allowing to have several
2207 independent segments of malloc)
2208 * New function: xbt_dict_cursor_set_data()
2209 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2210 * New function: xbt_dynar_is_empty()
2211 * New function: xbt_fifo_get_last_item()
2212 * Fix xbt_dynar_shrink(): use the right element size.
2213 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2214 uninitialized areas during expand.
2215 * Fix semaphores: previous implementation was severely broken.
2216 * Use library init/fini functions for our initialization.
2217 - you can use logs and other feature as soon as you want in your
2218 code (even before the xbt_init / MSG_init)
2219 - xbt_exit is now a no-op and produce a warning when used.
2222 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2223 benefit from the latest improvement to the simulation kernel.
2224 * Kill measurement sockets for now. If you rely on them, sorry. This
2225 release is not for you. This feature will be reintroduced in the
2226 future, but we cannot delay the release any further.
2227 * New function: gras_msgtype_get_name().
2228 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2229 * Fix (at last) the pmm example: it should not randomly fail anymore.
2231 Build chain: bug fixes and overall polishing
2232 * Cmake is now stable enough. Hence, we killed the autotools.
2233 * Port to windows ( TM :)
2234 * Fix the 'make install' target.
2235 No need to use 'make install-simgrid' anymore
2236 * Introduce a 'make dist' target compiling a *source* archive
2237 'make package' compiles a binary archive
2238 * Compile java files only on need
2239 * Add --cd and --setenv command line options to tesh
2240 * Out of source builds are not fully supported yet, but we are close
2241 * Enable supernovae and optimization flags by default for our users
2244 * Add layer to set up environment directly from lua, without XML.
2245 * The effect of gras_stub_generator can be achieved through
2246 lua too (check examples/gras/console/ping_generator.lua)
2248 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2250 SimGrid (3.4.1) stable; urgency=low
2252 The "Polishing easter eggs is probably a good idea" release.
2253 This is a bug fixes release only.
2257 * Fix a bug preventing the tasks from begin garbage collected.
2260 * Fix a bug occuring when a host involved in a communication fails.
2261 This was not detected properly by the other peer involved in the
2262 communication. Now, it's reported as a network error.
2265 * Warn the user about loop dependencies in data flow of DAX files
2266 * Obey the control-flow dependencies of DAX files
2269 * Add option "enable_smpi" allowing to not compile SMPI.
2270 Probably useful for the (Mac) users experiencing a build error here
2271 * Improve the detection of lua5.1 and ruby1.8
2273 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2275 SimGrid (3.4) stable; urgency=low
2277 The "Easter in Cargese" release. Also known as (major changes):
2279 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2280 ~> bindings were greatly improved
2281 ~> new tracing infrastructure for better visualization introduced
2283 * the "Welcome to configury modernity" release.
2284 ~> we switched from autotools to cmake, and improved our cdash
2287 A more detailled list of changes follow (full detail in svn log).
2289 Java Bindings: Various Cleanups
2290 * (install java-gcj-compat-dev on debian-like to use them)
2291 * Remove put/get: no need to export deprecated interface in Java
2292 Use send/receive instead.
2293 * Cleanup the examples and add a README per directory
2294 * Remove example autoDestination (that's the only way to go now)
2295 * Remove example explicitDestination (was a plain copy of basic)
2296 * Make JniException a runtime exception, so that there is no need to
2297 declare the fact that you may encounter such a beast. I guess that
2298 nobody will ever want to survive such error.
2299 * Create specific errors for each MSG case of failure:
2300 host failure, transfer failure, timeout, task cancelled
2301 * Cleanup the exceptions that may get thrown by each function
2302 * Other internal cleanups in Java bindings. Performance still bad :/
2303 Ruby and Lua Bindings: create them
2304 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2305 * That's new and great, you should try them out.
2306 Same functionalities than Java bindings, only even less polished
2308 * Kill the useless "rate" argument of SD_task_get_execution_time()
2309 Everyone used to provide -1 as a value, it was not used, and the
2310 semantic of a possible use wasn't even clear.
2311 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2312 either as comm costs or compute costs to mean that there is no
2313 such thing for that specific task.
2314 * Add a SD_task_set_name() function
2315 * Fix SD_task_unschedule() on typed tasks
2316 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2317 * In DAX loader, accept useless 'level' attributes to <job> since
2318 LIGO DAGs have them (seem to be to ease graphical representation).
2320 * Add an example masterslave_mailbox.c using send/receive and not
2321 the deprecated put/get interface.
2322 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2323 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2325 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2326 (declare MSG_USE_DEPRECATED to still have the old name)
2327 * Add a MSG_task_set_data() function
2328 * About trace replay (see examples/msg/actions):
2330 - Allow to work with splitted trace files for each process
2331 Give the specific trace file as argument of each process,
2332 and call MSG_action_trace_run(NULL)
2333 You can still have one merged file for all processes.
2334 - Fix implementation of collective operations
2335 * Allow task_execute() on 0-sized tasks (closes #10063)
2337 * This is the first release of SimGrid where SMPI is not considered
2338 beta anymore (even if some corners should still be improved)
2339 * Port over the new SIMIX_network submodule (internal refactoring)
2340 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2341 * Implement more missing elements of the standard:
2343 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2345 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2346 MPI_Scan() MPI_Get_processor_name()
2347 - Added implementation of missing case for Alltoall (warning: it's
2348 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2349 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2350 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2351 SMPI_MPI_Allgatherv()
2352 * Bug fixes include:
2353 - MPI_Waitsome() was broken
2354 - Allow relative includes in smpicc
2355 - Command line cfg argument 'reference_speed' was ignored...
2356 - Some functions did not properly lead to auto-benching of user code
2357 - smpicc passes -O2 by default (just like openmpi one)
2359 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2360 * Bug fixes about timeouts during communications
2361 * add SIMIX_message_sizes_output() as a pimple to write to file the
2362 amount of messages per size. Use gnuplot to get histogram.
2363 Pimple because that's the only user-visible function of simix,
2364 defined directly in xbt.h (irk, sorry)
2366 - Add a SIMIX_sem_get_capacity() function
2367 - Fix interactions with processe resume/suspende
2368 - release_forever() was stupidly broken
2369 - Fix SIMIX_display_process_status() for processes in a semaphore
2370 - Make SIMIX_sem_block_onto() user-visible
2371 * Refactoring context stuff:
2372 - Use pseudo-OOP for better modularity
2373 - reimplement SIMIX_process_kill() without process_schedule() so
2374 that the latter can take as invariant that it is called from
2376 - Merge context_start into context_new for sake of simplicity
2378 * Add a Vivaldi network model, coded live during SUD'10 ;)
2379 * Rename configuration variables to start a hierarchy:
2380 o cpu_model -> cpu/model
2381 o network_model -> network/model
2382 o workstation_model -> workstation/model
2383 * New configuration variables:
2384 o network/bandwidth_factor: correction to bandwith
2385 o network/latency_factor: correction to latency
2386 o netwotk/weight_S: correction to the weight of competing streams
2387 * Add a long description to the models, that users can see with such
2388 argument on the command line: --cfg=cpu/model:help
2389 * --help-models display the long description of all known models
2391 * config: add the ability to set a default value after registration
2392 Does not override any previously set value (e.g. from cmd line)
2393 * dict: allow to have integer key and data.
2394 When so, you need to use the following functions
2395 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2396 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2397 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2398 In contrary to regular dicts, the key is not malloced before copy.
2399 Mixing scalar and regular elements in the same dict is not tested
2401 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2402 Tracing for Visualization:
2403 * SimGrid is now instrumented in order to generate a trace file for
2404 visualization analysis: to use it, need to compile SimGrid with the
2405 "tracing" option enabled, and instrument the program using SimGrid with
2406 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2407 (among other functions).
2408 * The instrumentation only traces the platform utilization for now
2409 * Documentation to use the tracing functions and how to analyze the
2410 traces with the Triva tool is written.
2411 * More information about: SimGrid FAQ (in the section Tracing Simulations
2414 * We moved to cmake as default build system. Autotools support will
2415 be dropped soon. Check the FAQ for more info about how to use it.
2416 * Greatly improved our cdash/ctest interactions
2417 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2418 * Added memory checking tests with valgrind; lot of memleak fixing.
2419 This may be the first release of simgrid with so few memory issues
2420 * Added code coverage tests.
2421 Our coverage is still improvable, but at least we see it on cdash.
2423 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2425 SimGrid (3.3.4) stable; urgency=low
2427 The "Desktop Grid needs love too" release (also called Xmas release).
2429 Models improvements:
2430 * Major speedup in the maxmin system solving by using lazy evaluation
2431 Instead of solving completely the maxmin system at each iteration,
2432 only invalidate (and recompute) the modified parts.
2433 This new feature is enabled in default models but you can try to
2434 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2435 * Cas01 IMproved as default CPU model
2436 This CPU model is the same Cas01 model, but it uses the
2437 maxmin-selective-update flag and a heap structure to manage
2438 actions on SURF kernel.
2439 It reduces the complexity to find the next action to finish and,
2440 consequently, it's faster than the old Cas01.
2441 This is the new default CPU model (Cas01).
2442 * Rename the old Cas01 model to Cas01_fullupdate
2443 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2444 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2445 * CpuTI (CPU Trace Integration)
2446 A new CPU model whose objective is simulate faster when using
2447 availability trace files.
2448 Instead of using a full featured, over engineered maxmin system for
2449 CPU modeling, this model does the pre-integration of traces files
2450 to calculate the amount of CPU power available, and so, executes
2451 faster than the old CPU models.
2452 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2453 * Use LV08 as default network model since it gives better accuracy
2454 for small messages and shouldn't change things for big ones.
2455 Use --cfg=network_model:CM02 to get the previous behavior.
2458 ******************************************
2459 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2460 ******************************************
2461 * The new CPU model may changes simulations!
2462 The point is that events occurring at the exact same timestamp
2463 are not scheduled in the same order with the old and new
2464 version. This may be enough to completely change the execution
2465 of simulations in some cases.
2466 * The new network model will change simulations!
2467 This new model is more realistic than the previous one, so you
2468 should consider redoing your old experiments with this model.
2469 Sorry for the inconvenience.
2472 * Introduce the supernovae compilation mode
2473 When compiled that way, the whole SimGrid (or almost) is put in a
2474 single compilation unit and compiled in one shoot.
2475 This is to help gcc which has difficulties to inline stuff from one
2477 The speedup seem to be above 15%, althrough more tests are needed on
2478 amd64 to confirm that gain.
2481 * Port of MSG's mailbox on top of SIMIX network
2482 The put/get mechanism was greatly simplified on the way.
2485 * New SIMIX network module. Provides:
2486 - Mailbox: rendez-vous mechanism to find with who you want to speak
2487 - Synchronous send/recv: easier and hopefully faster since the
2488 logic is handled in the maestro process directly now
2489 - Asynchronous send/recv: you dreamt of it? It's here now
2490 Too bad that nobody cared enough to propagate the change to MSG.
2491 * Add semaphores as SIMIX synchronization mechanism.
2494 * new function SD_daxload(char*) to load a DAX file
2495 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2496 * Introduce typed tasks. Specify its kind and cost at creation.
2497 At scheduling, just give where it should be placed, and the cost
2498 for each involved resource is automatically computed.
2499 Existing constructors so far (more to come of course):
2500 - SD_task_create_comm_e2e() for end-to-end communication
2501 - SD_task_create_comp_seq() for sequential computation
2502 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2503 * new function SD_task_dump() for debuging display
2504 * new function SD_task_dotty(task,FILE*) writing to file the info
2505 about the task in dotty format
2506 * SD_task_dependency_exists() can now cope with having one of its
2507 arguments NULL. If so, it tests whether the other argument has any
2509 * Add getters on list of preceding/following tasks:
2510 SD_task_get_parents(task) and SD_task_get_children(task)
2511 * Add getters on amount of workstations and list:
2512 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2513 * Add getter on task kind: SD_task_get_kind(task)
2514 * Update the start_time and finish_time of tasks on completion/failure
2515 * Bugfix: Remove task from state swags when destroyed
2518 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2521 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2522 dynar with one separate thread per value of the dynar.
2523 * Change the prototype of xbt_thread_create(), sorry.
2524 Added a boolean parameter indicating whether we want to join this
2525 thread (used in SG only for now)
2526 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2529 * GTNetS wrappers should now be usable again (and betterly tested too)
2530 * Fix a major regression from 3.2 where the timeout provided to
2531 MSG_task_put_with_timeout() was used as absolute time before which
2532 the comm should be done.
2533 * Start to fix the <cluster> tag.
2534 - Internal links should be good now (beside of the loopback, which
2535 use the private link instead)
2536 - paths to the external world is still rather broken
2537 - the <route:multi> tag is just broken. Actually that's brain-dead.
2538 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2539 to make it less stupid
2540 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2541 * Fix a source-level compatibility glitch from 3.2: after defining
2542 MSG_USE_DEPRECATED, you can use the old name
2543 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2544 * Allow to compile from the SVN with automake 1.11
2545 * Fix some problems when using the "start_time" tag in deployment XMLs.
2546 * Fix #8569: XBT/synchro.h has redundant declarations
2547 * Fix #8563: MSG return values and exceptions
2548 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2549 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2550 Thanks to Robson Peixoto.
2551 * Fix a vicious bug in dictionaries inducing that some elements were
2552 not freed on xbt_dict_free()
2554 Portability report of this version:
2555 * Main portability targets:
2556 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2557 on (amd64/i386/ia64)
2558 - mac leopard on i386
2559 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2560 but nothing critical.
2561 * Other platforms: windows, AIX and others were not tested for this release
2563 Timing report of this version:
2564 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2565 scenario parameters). From 8h to a few seconds in desktop grid settings.
2566 * Supernovae brings about 25% speedup on i386.
2568 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2570 SimGrid (3.3.3) stable; urgency=low
2572 The "Need for Speed" release.
2574 The timings done to validate the 3.3.2 were faulty.
2575 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2577 The problem was a conversion from a manually handled vector to
2578 xbt_dynar_t on the critical path.
2579 xbt_dynar_foreach calls functions, inducing stack management crap.
2581 We inlined these functions and xbt_dynar_foreach is now breath taking.
2582 We also inlined xbt_swag_belong on the way.
2584 Here are some approximate speedup measurements (on master/slaves
2585 simulations lasting between 10s and 20s each):
2586 3.3.1 -> 3.3.2: about same performance
2587 3.3.2 -> 3.3.3: 40% speedup
2588 3.3.1 -> 3.3.3: 40% speedup
2589 3.3.1 with inline patch -> 3.3.3: 30% speedup
2591 Our reading is that the refactoring which occurred in 3.3.2 made us
2592 suffer much more from the xbt_dynar_foreach low performance, but
2593 once we solved this, this refactoring proved to be very performance
2594 effective. From the 40% speedup, somehow, 10% are due to the
2595 inlining and 30% to the refactoring.
2597 That's a pitty that gcc cannot inline functions placed in other files
2598 alone. We have to choose between:
2599 - break the encapsulation (by putting private data structures and
2600 accessors in headers files to help gcc)
2601 - live with low performance
2602 - switch to a decent compiler such as icc (not quite possible).
2604 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2606 SimGrid (3.3.2) stable; urgency=low
2608 The "Simplicity does not preceed complexity, but follows it" release.
2610 The main contributors of this release were (lexical order):
2611 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2614 * Extract the routing logic into its own object.
2615 (was dupplicated in network.c and workstation_LV07.c;
2616 Allows to implement other ways of storing that info)
2617 => kill now useless network_card concept
2618 - Use dynar to represent routes (instead of void** + int*)
2619 - kill link_set (use surf_network_model->resource_set instead)
2620 - Add a command-line option to choose the routing schema to use
2621 - Add three new models:
2622 * Floyd (shortest path computed at initialization)
2623 * Dijikstra (shortest path recomputed all the time)
2624 * Cached Dijikstra (shortest path computed on need)
2625 All these models where contributed by Silas De Munck, and are
2626 described in his ICCS09 paper.
2628 * Simplify model declaration
2629 (less redirections, less function to write when defining a model)
2630 - Factorize stuff between models:
2633 surf_model_resource_set(model)
2634 surf_model_resource_by_name(model, name)
2635 - Unify the types of models in s_surf_model_t (using an union)
2636 - Embeed fields of common_public directly into s_surf_model_t
2637 - Rename model methods:
2638 action_free ~> action_unref
2639 action_change_state ~> action_state_set
2640 action_get_state ~> action_state_get
2641 - Change model methods into functions:
2642 (model)->common_public->action_use ~> surf_action_ref
2644 * Implement a generic resource; use it as ancestor to specific ones
2645 (allows to kill duplicated code in models)
2646 Drawback: timer command don't need no name nor properties;
2647 workstation_CLM03 don't need no properties
2648 (but I guess we can live with those few bytes wasted)
2650 * Improve the action object model
2651 - implement a constructor avoiding dupplicated code about field
2652 initialization in generic_action part.
2654 * Kill the SDP model: it has an external dependency, is deprecated
2655 in flavor of modern lmm models, and didn't compile since a while
2658 * Relocation of the context module from XBT to SIMIX.
2659 (the context were decoupled from the simix processes, duplicating a lot of code)
2660 => a lot of code was factorized
2661 - less overhead is introduced during scheduling
2662 - simpler API for the context factory
2663 - the logic for process creation,destruction and manipulation was simplified
2664 * Simplification of the s_smx_process_t data structure.
2665 => accesing the simix level data associated to a process is faster now,
2666 and the code is a lot more readable.
2669 * Implement some more MPI primitives:
2670 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2671 -implementation: Bcast: flat or 2-ary tree (default),
2672 Barrier: 4-ary tree,
2674 Allreduce: Reduce then Bcast
2675 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2676 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2677 Alltoallv: flat tree, like ompi
2679 * Add support for optimized collectives (Bcast is now binomial by default)
2680 * Port smpirun and smpicc to OS X
2683 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2684 Shout out if you used it.
2687 * Display the list of still queued messages in SG mode when existing
2691 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2692 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2693 * Add xbt_str_from_file(FILE*)
2694 * Add xbt_dict_get_key achieving a linear reverse search
2695 * Remove the context module
2697 Portability report of this version:
2698 * Main portability targets:
2699 - Linux(debian)/x86/context
2700 - Linux(debian)/x86/pthreads
2701 - Linux(debian)/amd64/context
2702 - Linux(debian)/amd64/pthreads
2703 On these, we still have the eratic breakages of gras/pmm and
2704 amok/saturate_sg reported in previous version. We still think
2705 that the tests are the cause of the fault, not the tested code.
2707 - Mac OSX Leopard/x86/context
2708 Still false negative in tesh autotesting.
2709 Smpi still fails, but this time because readlink does not accept -f
2710 Everything seems to work properly beside of that.
2713 - AIX version 5.3 (only tested contexts this time)
2714 Smpi still fails there because mktemp is not installed.
2715 Everything seems to work properly beside of that.
2717 I managed to compile it for the first time, but several breakages.
2718 Won't delay the release for this exotic platform.
2720 * Windows: it's still lagging behind. If you want to help, please
2723 Timing report of this version:
2724 This version seem to be more than 5% faster than 3.3.1 (on linux
2725 64bits with contextes). The gain is less than expected, we are
2726 investigating this for next release.
2728 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2730 SimGrid (3.3.1) stable; urgency=low
2733 * Implement a --cfg-help to show existing configuration variables
2734 * Build chain do not require doxygen in maintainer mode
2737 * fix a bug on struct sizeof computation, which prevented the
2738 exchange of arrays of structs in some conditions
2739 - added a regression test about this in datadesc_usage
2740 * Allow the exchange of 0-long dynamic vectors.
2741 - for that, use -1 as indicator of dynamic size instead of 0
2742 - This implied to change any size from unsigned long to long,
2743 reducing a bit communication abilities, but I guess that with
2744 64bits being quite common, this is more than enough.
2745 - This also induce a protocol change, thus bumping network protocol
2746 version from 0 to 1 (if we have external users, we have to get
2747 clean on that point too ;)
2748 - added two regression tests about this in datadesc_usage
2749 * Be more verbose when propagating local exceptions
2750 This helps debugging.
2751 * Display the status of simulated processes when receiving SIGINT in
2755 * Allow to control the simulation from a trace file.
2756 New functions MSG_action_register() and MSG_action_trace_run()
2757 The first one allows to associate a function execution to each
2758 kind of action while the second one parses a trace file and
2759 triggers the corresponding actions within the system.
2760 For now, only a toy example is provided in examples/msg/actions
2761 * Add an exemple of process migration in examples/msg/migration
2762 * Fix a bug in task exchange which broke MSG_task_get_sender()
2763 Add a teshsuite regression test for that.
2764 [Bug: if MSG_task_get_sender() is called after sender exit,
2766 * Fix a bug which prevented suspend/resume to work properly
2767 * Display the status of simulated processes when receiving SIGINT
2768 This fixes a regression of v3.3. due to the introduction of SIMIX
2769 * Bug fixing in failure management:
2770 - trace could not start by a failure at time 0
2771 - failure during communications were not working
2774 * Add SIMIX_process_set_name() to change the name of the current
2775 process in the log messages.
2776 * Store smx_hosts in a dict since we only retrieve them by name
2777 * Move the configuration infrastructure to surf
2780 * Move the configuration infrastructure to surf
2783 * Massive internal cleanups:
2784 - Store internal structures on processes instead of hosts (allows
2785 to have more than one process per host, in addition of being more
2787 - Cleanup the initialization/finalization process
2788 - Kill a whole bunch of unneeded synchronization:
2789 processes run in exclusive manner within the simulator
2790 - Move queues from global tables to process data fields
2792 - now accept -platform and -hostfile arguments
2793 - Pass the right rank value to processes according to the hostfile
2794 * Compile the examples by default, and use them as regression tests
2795 * Implement MPI_Wtime()
2796 * Change the reference speed to a command line option
2799 * TCP_gamma can now be specified as command line option using
2800 --cfg=TCP_gamma:10000000.0
2801 * Change the --surf-path cmd line option into --cfg=path:
2804 * Also include strbuff from xbt.h public header
2805 * xbt_ex_display(): do not free the exception after displaying
2806 This allows to do more with the given exception afterward.
2807 Users should call xbt_ex_free() themselves.
2811 Portability report of this version:
2812 * Main portability targets:
2813 - Linux(debian)/x86/context
2814 - Linux(debian)/x86/pthreads
2815 - Linux(debian)/amd64/context
2816 - Linux(debian)/amd64/pthreads
2817 These targets fail about 1/10 of times on gras/pmm, but we believe
2818 that this is because of the test, not because of simgrid.
2819 amok/saturate_sg fails even more rarely, and the test may not be
2822 - Mac OSX Leopard/x86/context
2823 The test suite still spits tons of errors because some obscure
2824 force prevents us from removing the temporary directories
2825 arguing that they still contain some metadata I've never heard of.
2826 Smpi fails because seq is not installed.
2827 Everything seems to work properly beside of that.
2830 - AIX version 5.3 (both contexts and pthread)
2831 Smpi still fails there because mktemp is not installed.
2832 XML inclusions seems rosty on AIX.
2834 * Windows: it's still lagging behind. If you want to help, please
2837 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2839 SimGrid (3.3) stable; urgency=high
2843 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2844 [Malek Cherier & Mt]
2846 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2849 Until now, GRAS were using MSG as an interface to SURF. It was
2850 quite difficult because both interface have several differences
2851 (MSG channels vs GRAS sockets were the most notable point).
2853 This also opens the gate to SMPI (which should occur soon) and speed
2854 up simulations by to 40% (even if it were not the main goal).
2856 **************************************
2857 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2858 **************************************
2859 The point is that events occuring at the exact same timestamp are
2860 not scheduled in the same order with the old and new version. This
2861 may be enough to completely change the execution of simulations in
2862 some cases. Sorry for the inconvenience.
2864 * Cleanup and upgrade the XML format to push further scalability
2865 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2867 * Improve the testing infrastructure with tesh. Now a very large part of
2868 the code is tested not only by being run but also by checking that the
2869 output match an expected output [Mt].
2871 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2872 is really less memory-demanding, which should allow you to use
2873 larger files in SimGrid [AL].
2875 * Inform valgrind about our contextes, so that it becomes usable
2876 with the default (and more effecient) version of SimGrid
2877 [contributed by Sékou Diakite, many thanks]
2880 * Introduce a listener thread in charge of receiving incoming
2881 messages from the network. It allows to overlap communication and
2882 computation but most notably, it removes some stupid deadlocks due
2883 to the fact that so far, a process could not send and receive at
2884 the same time. This made most non trivial communication schema
2886 * Convert the PIDs from long int to int to match the MSG ones (and
2887 linux ones too) [Mt]
2888 * New function: gras_agent_spawn() to launch a new process on
2889 current host. Only working in simulation for now. [Mt]
2890 * New function: gras_os_hostport() returning a constant form (ie,
2891 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2894 * Make the backtrace of exceptions more human readable [Mt]
2895 * New module: xbt/str [Mt]
2896 a ton of string utility functions (split, join, printf to a newly
2897 allocated buffer, trim, etc)
2898 * New module: xbt/hash [Mt]
2899 SHA1 hashing algorithm (more to come if needed)
2900 * New module: xbt/synchro [Mt]
2901 synchronization tools (mutex and conditions) working the same way
2902 in simulation and in real life (mainly useful for GRAS, but not
2904 * New module: xbt/queue [Mt]
2905 classical producer/consumer synchronization scheme
2906 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2907 (using the classical functions will get serialized) [Mt]
2908 * Make dictionary internal table dynamic. No need to specify its size
2909 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2911 * Make sure the log channels are organized as a tree under windows
2912 (because of ANSI C compatibility issue, any channel were child of
2913 root directly) [Mt].
2916 * Cleaned many thing in surf and fixed a few bugs [AL].
2917 * Add a nice command line configuration mechanism to compose models [AL].
2918 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2919 the previous one (KCCFLN05). It relies on something that looks like
2920 a max-min sharing mechanism but cannot be written as such. A new solver
2921 was thus designed [AL].
2922 * Add a new solver to lmm. Based on Lagrange optimization and
2923 gradient-based descent, it enables to efficiently maximise systems s.a
2925 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2927 This solver enables to propose two new network models for TCP Reno and
2928 TCP Vegas based on Low's work. These models still need to be fully
2929 tested though [Pedro Velho].
2932 * Bug fix in SD_simulate. Now the time bound given as argument is
2934 * Use the new parallel task model (ptask_L07) as default.
2935 * Use the SURF command line configuration mechanism.
2936 * 0-size tasks (for synchronization) should now work.
2938 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2940 SimGrid (3.2) stable; urgency=high
2944 We still experience issues on this platform, but we believe that at
2945 least MSG is usable.
2947 GRAS API BREAKAGE (for simplification purpose, sorry):
2948 * the gras_msgtype_by_name is not used anymore. Instead of
2949 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2950 you can write (and must)
2951 gras_msg_send(toserver, "request", &request);
2952 - If you still want to pass a gras_msgtype_t to the function (to cache
2953 the type and avoid the lookup time), use the gras_msg_send_() variant.
2954 - Impacted functions:
2955 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2956 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2957 * The callbacks are now expected to return 0 when everything went well
2958 (just like the main() function)
2960 GRAS new features and improvements:
2961 * New module mechanism where user code can use per process globals [Mt]
2962 This is similar to gras_userdata_*() functions, but for libraries. It
2963 factorize some code developped over and over in the examples and AMOK.
2964 It has still to be documented and used (only amok/peermanagement is
2966 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2967 loss when they were small enough to fit into the buffer and sent quickly
2968 enough so that they can all get received in one shoot.
2969 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2970 if not found. Use the *_or_null() variant for the old semantic.
2971 * In gras_msg_handle, do not discard messages without callback.
2972 They are probably messages to be explicitly awaited later (ie, proofs of
2973 mis-synchronization in userland since they are sent before being awaited)
2974 No big deal usually.
2975 * gras_socket_meas_send/recv: semantic changed!
2976 The numerical arguments used to be (1) the total amount of data to send
2977 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2978 messages. This was need for the fool willing to send more than MAXINT
2979 bytes on quite fat pipes.
2982 * Do really rename the hostmanagement module to peermanagement. [Mt]
2983 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2984 but this is rather new and this was documented in the module
2985 documentation (poor excuses, I admit)
2986 * Bandwidth measurement semantic changed! This follows the changes to
2987 gras_socket_meas_send/recv explained above.
2990 * A sequential mode has been added to the workstations. When a workstation
2991 is in sequential mode, it can execute only one task, and the other tasks
2992 are waiting in a FIFO. [Christophe Thiery]
2995 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2996 model for SIMDAG. [Christophe Thiery]
2997 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2998 equal to 0, which caused some bad behaviors in
2999 saturated_constraint_set_update. I now use a threshold mechanism like in
3003 * When running manually src/testall, you select specific units [Mt]
3004 testall is the result of our cunit mechanism, and should replace all
3005 the scripty thingy around since bash don't run easily on billware.
3007 * A mallocator system has been added. [Christophe Thiery]
3008 Mallocators allow you to recycle your unused objects instead of freeing them
3009 and allocating new ones.
3011 Documentation update:
3012 * FAQ reworking + New FAQs:
3013 - "Valgrind spits tons of errors!" [Mt]
3014 - "How to repport bugs" [Mt]
3015 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3016 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3017 - Communication time measurement within MSG [AL]
3018 - I experience weird communication times when I change the latency [AL]
3019 * GRAS tutorial [Mt]
3021 - an introduction to the framework and to the used communication model
3022 - an initiatic tour introducing the most proheminent features:
3024 . Lesson 0: Installing GRAS
3025 . Lesson 1: Setting up your own project
3026 o Part 2: Message passing
3027 . Lesson 2: Exchanging simple messages
3028 . Lesson 3: Passing arguments to the processes (in SG)
3029 . Lesson 4: Attaching callbacks to messages
3030 . Lesson 5: Using globals in processes
3031 . Lesson 6: Logging information properly
3032 . Lesson 7: Using internal timers
3033 . Lesson 8: Handling errors through exceptions
3034 . Lesson 9: Exchanging simple data
3035 . Lesson 10: Remote Procedure Calling (RPC)
3036 . Lesson 11: Explicitely waiting for messages
3037 . Recapping of message passing features in GRAS
3038 - A HOWTO section containing:
3039 o HOWTO design a GRAS application
3040 More are due, of course. They will come latter. In the meanwhile, you can
3041 check the examples which are still here.
3043 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3045 SimGrid (3.1) stable; urgency=high
3049 There was a stack corruption somewhere, visible only when optimizing
3050 with these versions. [Vince]
3053 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3054 module that enabled to play with Directed Acyclic Graphs. It is built
3055 directly on top of SURF and provides an API rather close to the old
3056 SG. Some old codes using SG are currently under rewrite to check that
3057 all needful functions are provided. [Christophe Thiery]
3060 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3061 extension of the classical CLM03 model that gracefully handles
3062 failures. This is now the default model for MSG and GRAS. It doesn't
3063 handle parallel tasks yet though. [AL]
3064 * Bug fix: Weights were not correctly set in the network part.
3065 WARNING: This may have resulted in incorrect results with simulations
3066 where there are more than one flow on a given link. [AL]
3069 * After a (long ?) discussion on simgrid-devel, we have decided that the
3070 convention we had on units was stupid. That is why it has been decided
3071 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3072 WARNING: This means that all previous platform files will not work as
3073 such with this version! A warning is issued to ask users to update
3075 A conversion script can be found in the contrib module of the CVS, under
3076 the name contrib/platform_generation/surfxml_update.pl [MQ]
3079 * Bug fix: Processes were started in reverse order, wrt deployment file.
3080 WARNING: if your code relies on this bug, please fix it. [AL]
3081 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3082 being executed on two different locations. [AL]
3083 * Bug fix: Failures are now better supported thanks to Derrick's tests
3084 (there was many failure situations I hadn't thought of and that weren't
3085 correctly handled). [AL]
3086 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3090 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3091 Exception occurring on server-side are propagated back to client (!).
3093 API CHANGE: the callback changed their prototype. Change:
3094 int my_handler(gras_socket_t expeditor, void *payload_data) {
3096 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3097 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3099 * New! function: gras_msg_handleall to deal with all messages arriving
3100 within a given period.
3101 * New! function: gras_socket_server_range to get a server socket in a
3102 range of port numbers (ease to avoid port number conflicts) [MQ]
3103 * New! gras processes display their backtrace when they get a SIGUSR1
3104 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3105 Sweet to debug RL processes [MQ]
3109 - Do not force experiment sizes to be expressed in kb, or it becomes
3110 impossible to measure the latency this way (needs one byte-long tests)
3111 WARNING: this changes the amok_bw_* function semantic. [MQ]
3112 - Implements the link saturation stuff. [MQ]
3113 * Peer management module:
3114 New! module factorizing code that we wrote over and over [MQ].
3117 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3118 - Test units are placed directly into the library code, they get extracted
3119 automatically and placed into the src/testall binary.
3120 - Convert most of the XBT tests to this system.
3121 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3122 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3123 live in the same C code. This required to change a little bit the API
3124 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3125 * New module: sparse graph structure with basic algorithms (this is work
3126 in progress and the API is not considered to be frozen yet). [AL]
3127 * Display more information on backtraces: source line & function names are
3128 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3129 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
3130 * Reworked a little bit some #include statements to load only required
3131 headers. Some user code that relied on SimGrid to include stdlib or
3132 stdio may need to include it by themselves. [AL]
3133 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3135 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3136 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3137 but I guess that almost nobody used those parts. [MQ]
3139 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3141 SimGrid (3.0.1) stable; urgency=low
3144 * Unfortunately, I had missed 5 misnamed functions:
3145 xbt_fifo_item_t xbt_fifo_newitem(void);
3146 void xbt_fifo_freeitem(xbt_fifo_item_t);
3147 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3148 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3149 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3150 They're now deprecated. Please use their new versions:
3151 xbt_fifo_item_t xbt_fifo_new_item(void);
3152 void xbt_fifo_free_item(xbt_fifo_item_t);
3153 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3154 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3155 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3157 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3158 * Documentation: xbt_log module unmercifully reworked [MQ]
3159 * Bugfix: there was a problem with the ending of contexts with
3160 the pthread backend. It caused some weird deadlock or behavior
3161 depending on the pthread implementation. [AL]
3162 * Bugfix: get the exceptions raised in the simulator repport where
3163 and why they come from when they are not caught in time [AL, MQ]
3166 * Bugfix: Do repport the error when two non-connected hosts try to
3167 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3170 * Add additionnal checkings on communications. Assert that two
3171 communicating hosts are connected by a set of links... [AL]
3174 * Add additionnal checkings on channel values in communication [AL]
3175 * New: MSG_task_get_source to see on which host a task was generated [HC]
3176 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3177 the number of tasks waiting to be received on channel and sent
3179 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3180 waits for the first task coming from a given host.. [AL]
3182 GRAS new functionnalities: [MQ]
3183 * Enhance the parsing macro to allow the size of multidimentional objects
3184 to be given thru annotations.
3185 * New example (and documentation): Matrix Multiplication a la RPC
3186 (as when I was young!) and fix a bunch of bugs found on the way.
3188 GRAS performance improvements: [MQ]
3190 * Reduce the amount of cbps creation/destruction by making it static to
3191 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3194 * Change libdata to a set so that we can search for stuff by ID (and thus
3195 reduce the insane amount of dict lookups)
3198 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3199 RL communication on top of SG since it's so uneffective.
3200 It may also be used for inter-thread communication in RL, one day.
3201 * Use gras_datadesc_copy() to exchange messages on top of SG
3203 - improve message exchange performance on top of SG
3204 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3205 * Don't exchange on the network the size of the used part of buffer,
3206 instead, specify the possible buffer size to read().
3208 - reduces the amount of read/write calls (one pair per exchange)
3209 - reduces the amount of exchanged data (the size)
3210 - allows to retrieve all arrived data on receiver side, if we don't need
3211 it right now (subsequent read will peek the buffer)
3212 - allows the receiver to proceed with the begining of the stream before
3213 everything is arrived
3214 - make it possible to build an iov transport (using readv/writev)
3216 - take care of the data with non-stable storage (like stacked data),
3218 * If possible, TCP send uses vector I/O (when writev() is here)
3219 - Don't use it for receive since we send data sizes and data on the
3220 same stream, so we wouldn't be able to chain large amount of chunks
3221 before having to flush the stuff to read the size.
3222 * Rework the transport plugin mechanism to simplify it and reduce the
3223 amount of pointer dereferencement when searching for the right function
3226 * I guess that now, we do almost as few system calls as possible while
3227 doing as few data copy as possible.
3229 To improve it further, we could try to send all the sizes first and then
3230 all the data (to use iov on receiving size), but it's only a partial
3231 solution: when you have 2 dimensional data, the sizes of the second
3232 dimension is data of the first dimension, so you need 3 streams.
3234 I'm not sure the potential performance gains justify the coding burden.
3236 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3238 SimGrid (3.00) stable; urgency=high
3241 * New! Give the possibility to hijack the surf parser and thus bypass
3242 MSG_create_environment and MSG_launch_application. Have a look at
3243 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3245 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3247 SimGrid (2.96) unstable; urgency=low
3252 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3253 This deprecates the xbt_error_t mechanisms.
3254 It modifies (simplifies) all XBT and GRAS API.
3255 MSG API keeps unchanged (exceptions raised by XBT are caught from
3256 within MSG and masked with existing error handling facilities)
3259 * New! Add a FATPIPE model. [AL]
3260 * New! Add a parallel task model. [AL]
3261 * New! Add automatically a loopback interface (in the default
3262 network model) if none was precised.
3265 * Bugfix: MSG_process_resume now works with the current running process.
3267 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3268 * Modification of MSG_task_get_compute_duration. Once a task has been
3269 processed, the value returned by this function is now equal to 0. [AL]
3270 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3271 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3272 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3273 of a m_task. For now, MSG_task_cancel only works with computation
3275 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3276 of the processor (in Mflop/s) regardless of the current load on the
3278 * API Change: use proper naming convention for MSG_getClock and
3279 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3281 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3282 This function changes the priority of a computation task. This priority
3283 doesn't affect the transfer rate. A priority of 2 will make a task
3284 receive two times more cpu power than the other ones. This function
3285 has been added to suit the needs of Nguyen The Loc and hasn't been that
3286 much tested yet. So if it fails, please report it and send me your code.
3288 * API Change: removed all functions and types that were marked "deprecated"
3289 since many months. Renamed MSG_global_init_args to MSG_global_init.
3291 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3293 SimGrid (2.95) unstable; urgency=low
3296 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3297 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3298 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3301 * Bugfix: Allow absolute paths for platform description files [MQ]
3302 * Bugfix: do free the variables after use. Leads to drastic performance
3304 * Implement max_duration (ie, timeouts) on resources [AL]
3307 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3309 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3310 the needed support (timeouts) [AL]
3313 * Implement measurement sockets. You can now get the bandwidth between two
3314 hosts thanks to AMOK (see below). [MQ]
3315 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3316 dynar over the network (yeah) [MQ]
3317 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3318 * simple/static token-ring example. [Alexandre Colucci and MQ]
3319 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3320 in place before [MQ]
3323 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3324 type parsing. "short" and "long" modifiers were also ignored; other
3325 modifier (such as reference level) are still ignored. [MQ]
3326 * Embeed the buffer size within the buffer itself on SG. [MQ]
3327 That way, send() are atomic and cannot get intermixed anymore (at least
3328 the ones which are less than 100k; bigger messages still have the issue)
3329 * Array size pushed by the field, not by the field type (or each
3330 and every long int will push stuff to the cbps) [MQ]
3331 * use select() to sleep since it allows to portably sleep less than one
3334 GRAS (minor cleanups)
3335 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3338 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3339 * type category 'ignored' killed as it was never used and were difficult
3341 * whether a type can cycle or not is now a flag, leaving room for more
3342 flags (as "ignored", if we feel the need one day ;)
3343 * Rename raw sockets to measurement sockets since "raw" has another
3344 meaning in networking community.
3347 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3348 and offers features not belonging to GRAS but that most applications
3349 need. One day, it may be a set of plugins loadable at runtime.
3350 * New module: bandwidth
3351 bandwidth measurement between arbitrary nodes running this module. [MQ]
3353 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3355 SimGrid (2.94) unstable; urgency=low
3357 The first beta release of SimGrid 3 !
3361 * Update the main page and the FAQ. Adding references to gforge.
3364 * Add a gras_os_getpid function.
3367 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3368 * Extend the logs so that they also print PID, hostname, date, ... if
3370 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3371 and kill the old version which were in testsuite/
3372 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3376 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3377 for users. Instead, each programming environment (GRAS, MSG,...) use it
3378 when needed to provide such a feature to users.
3379 Don't shortcut the mechanism or you will also shortcut the virtualization
3380 you need on the simulator.
3384 * Cleanups in configury with regard to compile optimization/warning flags.
3385 Also add -fno-loop-optimize to any powerpc since it's the optimization
3386 killing gcc (< 3.4.0).
3387 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3388 needed by MSG examples complications
3389 * Borrow configury beautifications from PHP
3392 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3393 warning (thanks loris for stumbling into this one).
3394 * Bugfix: stop loading private headers (gras_config.h) from the public
3398 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3399 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3400 RL would imply a DNS resolver, which is *hard* to do in a portable way
3401 (and therefore delayed).
3402 * Implement a real timer mechanism and use it in timing macros. This allows
3403 to avoid rounding errors and get a 0.000005 sec precision in timing
3404 macros. While I was at it, various cleanups:
3405 - allow to declare more than one timed section per file (fix a stupid bug)
3406 - move some private declaration to the right place
3407 - merge conditional execution and timing macros into emulation module
3408 - document the module
3409 - make sure the module cleanups its mess on gras_exit
3410 * Documentation improvements:
3411 - (new) how to compile applications using GRAS
3412 - (new) emulation support (timing macros)
3414 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3416 SimGrid (2.93) unstable; urgency=low
3418 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3421 - Use Paje properly where used. Still to be sanitized properly.
3422 - Portability fix: Add an implementation of the contexts using pthread
3426 - Add xbt_procname(): returns the name of the current process.
3427 Use it to show the current process's name in all logging.
3429 - fix detection of older flex version and the reaction, since we do
3430 depend on modern ones (we use lex_destroy)
3431 - Better separation of SG and RL in the libs: remove all simulation code
3432 from libgras. As a result, this lib is now only 200k when stripped.
3433 Some of the xbt modules may also be duplicated (two sets and such) and
3434 should be cleaned/killed before SG3.
3435 - Insist on using xlC on AIX because of weird problems involving gcc there.
3436 - Cleanup the make remote stuff. This is now done by scripts
3437 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3438 mainly for our private use, but we're working on changing them to user
3441 - Bugfix: flush the socket on close only if there is some *output*.
3442 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3444 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3447 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3449 SimGrid (2.92) unstable; urgency=low
3451 Alpha 3 on the path to SimGrid 3
3455 - New! First try of benchmarking macros.
3456 - New! First try so that gras_stub_generator generate deployment and
3457 remote compilation helpers.
3459 - Bugfix: Initialization fix in msg_test.
3463 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3465 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3467 - API Change: gras_os_sleep to take the amount of seconds as a double.
3468 Accepting an int was error prone since it was the only location where
3469 seconds were coded as such. It leaded to damn rounding errors.
3470 - Bugfix: Hard to belive that timers ever worked before this.
3472 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3474 SimGrid (2.91) unstable; urgency=low
3476 Alpha 2 on the path to SimGrid 3
3480 - Bug fix in the lmm_solver.
3482 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3483 through the function MSG_paje_output.
3484 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3485 - It is possible to bound the rate of a communication in MSG with
3486 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3487 to write it in the changelog).
3488 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3493 - Various cleanups to the autotools stuff
3494 - Begin to move Gras examples to examples/gras/
3495 - Let make distcheck work again (yeah!)
3497 - documentation overhauled using doxygen.
3498 gtk-doc-tools is dead in SimGrid now.
3499 - Automatically extract all existing logging categories, and add the list
3500 to the documentation (long standing one, to say the less)
3502 - Cleanup the known architecture table. Reorder the entries to group what
3503 should be, and use a more consistent naming scheme.
3504 (some of the test dataset are still to be regenerated)
3505 - New! Allow library to register globals on each process just as userdata
3507 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3508 lookup time (for now).
3509 Use it in msg and trp.
3510 This cleans a lot the internals and helps enforcing privacy of the
3511 headers between the gras components.
3512 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3513 - Bugfix: gras_os_time was delirious in RL.
3514 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3516 - Reenable GRAS now that it works.
3518 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3520 SimGrid (2.90) unstable; urgency=low
3522 Alpha 1 on the path to SimGrid 3
3524 * It is a long time since the last release of SimGrid. I'm sorry about
3525 that but as I had told you, I was rewriting a lot of things. I apologize
3526 to those who had been reporting bugs to me and that I had not answered.
3527 If your bug is still in the new version, please tell me. Here is a
3528 summary of the main changes.
3530 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3531 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3532 portability, speed, and a lot more but you'll figure it out later.
3533 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3534 I wanted to release the new MSG as soon as possible and I have
3535 broken GRAS, which is the reason why, for now, only MSG is fully
3536 functional. A laconic description of these projects is available
3537 in the documentation.
3539 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3540 kernel optimized for our needs (called SURF but only the developers
3541 should use it). Hence, MSG is now roughly 30 times faster and I think
3542 that by rewriting a little bit MSG, I could event speed it up a little
3543 bit more. Beside the gain in speed, it is also much easier to encode a
3544 new platform model with SURF than it was with SG. More to come...
3546 * REVOLUTION 3: I have tried to change a little as possible the API of
3547 MSG but a few things really had to disappear. The main differences
3548 with the previous version are:
3549 1) no more m_links_t and the corresponding functions. Platforms are
3550 directly read from a XML description and cannot be hard-coded
3551 anymore. The same format is used for application deployment
3552 description. The new format is described in the documentation.
3553 Have a look in tools/platform_generation. There is a tiny script
3554 that converts from the old platform format to the new one. Concerning
3555 the application deployment format, parsing the old one is tricky.
3556 I think most of you should however be able to convert your files. If
3557 it is really an issue, I can write a C code that does the conversion.
3559 2) the toolbox tbx does not exist anymore. We now have a library
3560 with much more data-structures but without the hash-tables (we have
3561 dictionaries that are much faster).
3563 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3565 *****************************************************************************
3566 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3567 * don't feel like converting the oldies. (Mt) *
3568 *****************************************************************************
3571 Version 2.90: "the long awaited one"
3572 - Finished rewriting and debugging MSG. Rewrote the documentation.
3573 - disable GRAS for now since it needs to be ported to the newest SG
3576 - Finish the port to windows (using mingw32 for cross-compile)
3579 - Main loop and datastructures of SURF. A cpu resource object is
3580 functional. Surf can thus be used to create cpu's with variable
3581 performance on which you can execute some actions.
3583 2004-11-15 Martin Quinson
3584 - Port to ARM. Simply added the alignment and size descriptions. Should
3585 work, but the ARM machines are so slow that I didn't had the opportunity
3586 to 'make check' over there yet.
3588 2004-11-15 Arnaud Legrand
3589 - Trace manager now written. It uses a heap structure and is therefore
3590 expected to be efficient. It may however be speeded up (particularly
3591 when many events occur at the same date) by using red and black
3592 trees. One day maybe...
3593 - Max-min linear system solver written. It uses a sparse matrix
3594 structure taking advantage of its expected use. Most operations are
3595 O(1) and free/calloc are called as few as possible. The computation of
3596 the minimum could however be improved by using a red and black tree
3599 2004-11-03 Arnaud Legrand
3600 - Rename every gras_* function that was in xbt/ to its xbt_
3602 - Add a heap and a doubly-linked list to xbt
3603 - Added a dichotomy to the dictionaries. make check works as well before
3604 so I assume that the patch is correct. I do not know however if things
3605 run effectively faster than before now. :)
3607 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3608 SimGrid, and the version number is bumped to 2.x
3610 2004-10-29 Martin Quinson
3611 - Introduction of the remote errors.
3612 They are the result of a RMI/RPC on the remote machine.
3613 ErrCodes being scalar values, you can't get the host on which those
3614 errors did happen. Extending the error mechanism as in Gnome is possible.
3615 No idea yet whether it is a good idea.
3617 2004-10-28 Martin Quinson
3618 - Interface revolution: the Starred Structure Eradication.
3619 I used to do typedef struct {} toto_t; and then handle *toto_t.
3620 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3621 - ???_t is a valid type (builded with typedef)
3622 - s_toto_t is a structure (access to fields with .)
3623 - s_toto is a structure needing 'struct' keyword to be used
3624 - e_toto_t is an enum
3625 - toto_t is an 'object' (struct*)
3627 typedef struct s_toto {} s_toto_t, *toto_t;
3628 typedef enum {} e_toto_t;
3629 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3630 s_toto_t) is private.
3632 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3633 so that it can set the variable to NULL. It was so for dicts and sets,
3634 it changed for dynars.
3636 - Fix a bunch of memleaks in dict_remove
3637 - Fix a bug in sg/server_socket opening: it failed all the time.
3639 2004-10-07 Martin Quinson
3640 - Speed up dynar lookup operation a bit.
3642 gras_dynar_get is dead.
3644 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3645 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3646 (which gives you the address of the stored data).
3648 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3649 copy of the data using an affectation to do the job and not a memcpy.
3651 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3652 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3654 It does not really speedup the dynar test because they are
3655 setting elements all the time (and look them seldom). But the dict does
3656 far more lookup than setting.
3658 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3660 2004-10-05 Martin Quinson
3661 - Allow to (en/dis)able the cycle detection at run time.
3663 Whether we should check for cycle or not is now a property of each
3664 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3665 datadesc_cycle_unset allow to remove this property when previously set.
3667 Note that the cycle detection is off by default since it impacts the
3668 performance. Watch the data you feed GRAS with ;)
3670 This property is hereditary. Any element embedded in a structure having it
3671 set have it set for the time of this data exchange.
3673 You should set it both on sender and receiver side. If you don't set it on
3674 sender side, it will enter an endless loop. If you forget on receiver
3675 side, the cycles won't be recreated after communication.
3677 - Header reorganization.
3678 Kill gras_private.h, each submodule must load the headers it needs.
3680 2004-10-04 Martin Quinson
3681 - Interface revolution: do not try to survive to malloc failure.
3683 Now, gras_malloc and friends call gras_abort() on failure.
3684 As a conclusion, malloc_error is not a valid error anymore, and all
3685 functions for which it was the only gras_error_t return value are
3686 changed. They now return void, or there result directly.
3687 This simplify the API a lot.
3689 2004-09-29 Martin Quinson
3690 - Re-enable raw sockets.
3691 Created by gras_socket_{client,server}_ext;
3692 Used with gras_raw_{send,recv}
3695 It should allow to kill the last bits of gras first version soon.
3697 This is not completely satisfactory yet (duplicate code with
3698 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3701 - Simplify transport plugin (internal) interface by not passing any
3702 argument to _server and _client, but embedding them in the socket
3705 2004-09-28 Martin Quinson
3706 - Finish the port to AIX.
3707 autoconf was my problem (segfault within the malloc replacement
3708 function. No idea why)
3710 2004-09-16 Martin Quinson
3711 - Fix some size_t madness on 64bit architectures.
3713 2004-09-08 Martin Quinson
3714 - Reduce the number of system headers loaded, overload some more system
3715 calls (such as malloc to cast the result of the system one, and work
3717 - Fix and reintroduce the config support
3719 2004-09-07 Martin Quinson
3720 - Source code reorganization to allow Arnaud to surf all over there.
3721 - Allow to document the logging categories.
3722 - Remove all uppercase from logging categories and useless cleanup in names.
3724 2004-08-18 Martin Quinson
3725 Version 0.6.2 (protocol not changed; API changed)
3726 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3727 gras_error_t), and NULL when not found. Functions expecting a msgtype
3728 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3729 hopefully usefull message.
3730 - Portability to prehistoric sparcs again
3732 2004-08-17 Martin Quinson
3733 Version 0.6.1 (protocol not changed; ABI not changed)
3734 - prealloc some buffers to speed things up
3736 2004-08-11 Martin Quinson
3737 Version 0.6 (protocol not changed; ABI expended)
3738 - The parsing macro can deal with the references, provided that you add
3739 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3741 2004-08-09 Martin Quinson
3742 Version 0.5 (protocol not changed; ABI changed)
3743 - Allow to off turn the cycle detection code in data exchange at
3744 compilation time. It should be at run time, but I'm short of time (and
3745 the config stuff is still broken). That way, we keep dict out of the
3746 critical path, which is good because the performance is poor:
3747 - search not dichotomial yet
3748 - dynar give no way to access their content and memcpy everytime
3749 - In composed data description (struct, ref and so on), stop foolness of
3750 keeping the subtype's ID, but store the type itself. This keeps sets out
3751 of the critical path, which is good since they rely on dynar and
3752 dictionnaries. The only loose of that is that we cannot detect the
3753 redeclaration of a structure/union with another content (but I'm not sure
3754 the code detected well this error before anyway). We still can detect
3755 the redefinition discrepancy for the other types.
3756 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3757 it breaks the code because of type-punning used all over the place).
3758 This breaks on all non-gcc architectures (for now).
3760 All those changes (plus the buffer of last time) allow me to gain 2 order
3761 of magnitude on cruel tests consisting of 800000 array of integers on two
3762 level of a hierarchical structure (200 secondes -> 4 secondes)
3765 - the selector of reference must now return the type it points to, not
3766 the ID of this type.
3768 2004-08-06 Martin Quinson
3769 Version 0.4 (protocol changed; ABI not changed)
3770 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3771 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3772 - Implement a buffer transport, and use it by default (it relies on tcp in
3773 real life and on sg in simulation).
3774 That's a bit hackish since I had a new field to the structure to store
3775 its data without interfering with the subtype ones. Inheritance
3776 is tricky in C. And that's a kind of reverse inheritance with one class
3777 derivating two classes. Or maybe a game with java interfaces. Anyway,
3778 that's damn hard in C (at least).
3779 Moreover, I got tired while trying to ensure plugin separation and
3780 genericity in SG mode. MSG wants me to do weird things, so let's go for
3781 cruel hacks (temporarily of course ;).
3782 See comment in transport_private.h:71
3783 - do not include all the _interface headers in private but in the files
3784 which really need them (to cut the compilation time when they are
3787 2004-07-26 Martin Quinson
3788 Version 0.3 (protocol not changed; ABI changed)
3789 - Major overhault of the datadesc interface to simplify it:
3790 - shorted the function names:
3791 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3792 - add a trivial way to push/pop integers into the cbps without malloc.
3793 This allows to make really generic sub_type description, which simply
3794 pop their size of the stack.
3795 - add a function gras_datadesc_ref_pop_arr() which does what users want
3796 most of the time: Declare a dynamic array (which pops its size of the
3797 stack) and declare a reference to it. Poor name, but anyway.
3798 - kill the post-send callback, add a post-receive one
3800 2004-07-23 Martin Quinson
3801 Version 0.2 (protocol changed; ABI changed)
3802 - add some testing for cpbs in the test cases, and fix some more bugs.
3803 This invalidate again the little64 data file, since I cannot regenerate
3805 - remove an awfull optimization in the logging stuff, allowing me to:
3806 - understand it again
3807 - learn gcc how to check that the argument match the provided format
3808 - fix all errors revealed by gcc after that
3809 - internal keys of dict are not \0 terminated. Deal with it properly in
3810 loggings instead of segfaulting when the user want to see the logs :-/
3812 2004-07-22 Martin Quinson
3813 - Fix some stupid bug preventing cbps (callback postit) from working
3815 2004-07-21 Martin Quinson
3816 - Some documentation cleanups
3817 - remove the useless last argument of msgtype_declare
3818 - rename the Virtu functions to fit into the 'os' namespace
3819 - move headers src/include -> src/include/gras/ and stop fooling with
3820 gras -> . symbolic link
3821 - make distcheck is now successful
3823 2004-07-19 Martin Quinson
3825 - Build shared library also
3826 - Install html doc to the right location
3827 - stop removing maintainer files in make clean
3828 - build tests only on make check
3830 2004-07-13 Martin Quinson
3832 - No major issue in previous version => change versioning schema
3833 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3834 corresponding dataset.
3836 2004-07-11 Martin Quinson
3838 - Get it working with any kind of structure (we can compute the padding
3839 bytes remotely for all the architectures I have access to)
3840 - Implement the structure parsing macro (still not quite robust/complete)
3841 - Improvement to the remote testing toysuite
3843 2004-07-10 Martin Quinson
3844 [autoconf mechanism]
3845 - get ride of a bunch of deprecated macros
3846 - actually run the test for two-compliment, not only compile it :-/
3847 - test whether the structures get packed (and bail out if yes. Damn.
3848 Alignment is a serious matter)
3849 - test whether the structures get compacted (but respecting the alignment
3850 constraints of each types)
3851 - test whether the array fields of structures can straddle alignment boundaries
3853 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3854 (took me 2 hours to find that bug, looking at the wrong place)
3855 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3856 before sending/receiving and are used to compute the offsets of fields
3857 - Given that padding size depend even on compiler options, keep track of
3858 alignment and aligned_size only for the current architecture. Not a big
3859 deal since we send structure fields one after the other (seems
3861 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3862 program, let it work on linux/gcc/little32. portability todo.
3864 2004-07-08 Martin Quinson
3865 - import and improve remote compilation support from FAST
3866 - make sure make check works on half a dozen of machines out there
3868 2004-07-07 Martin Quinson
3869 Let's say it's version 0.0.3 ;)
3870 - Implement conversions (yuhu!)
3871 - Let it work on solaris (beside conversion, of course)
3872 - Stupid me, using rand() to generate the conversion datatests in not wise.
3874 2004-07-06 Martin Quinson
3875 - Let make dist work, since I'm gonna need it to compile on remote hosts
3876 - Let Tests/datadesc_usage write the architecture on which the file was
3877 generated as first byte.
3878 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3879 architecture descriptions.
3880 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3881 on those architectures.
3882 - Optimization: send/recv array of scalar in one shoot
3884 2004-07-05 Martin Quinson
3885 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3887 - Plug a whole bunch of memleaks
3888 - each process now have to call gras_{init,exit}. One day, their log
3889 settings will be separated
3890 - Continue the code factorisation between SG, RL and common in Transport.
3892 2004-07-04 Martin Quinson
3894 - Redistribution between SG and RL.
3895 We wanna have to accept in SG, so move accepted related parts of RL in
3896 the common part. (more precisely, the dynar of all known sockets is no
3897 more a static in transport.c, but part of the process_data)
3899 [gras_stub_generator]
3900 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3902 2004-07-03 Martin Quinson
3903 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3904 them all up in one shot)
3906 - Fix the ugly case of reference to dynamic array.
3907 - New (semi-public) function gras_datadesc_size to allow the messaging
3908 layer to malloc the needed space for the buffer.
3910 - gras_socket_close now expect the socket to close (and not its address to
3911 put NULL in it after it). This is because the socket passed to handlers
3912 is one of their argument (=> not writable).
3914 - propagate the interface cleanup from last week in datadesc, ie remove a
3915 superfluous level of indirection. User pass adress of variable
3916 containing data (both when sending and receiving), and not of a variable
3917 being a pointer to the data. Let's say that I like it better ;)
3918 The price for that is constructs like "int msg=*(int*)payload" in
3919 handlers, but it's a fine price, IMHO.
3921 - Let it work in RL (yuhu)
3923 2004-06-21 Martin Quinson
3925 - porting SG plugin and SG select to new standards (works almost).
3926 - plug memleaks and fix bugs around.
3929 - cleanup the prototype of data recv and force users to specify when they
3930 want to handle references to objects. Test case working even for cycles.
3931 - plug memleaks. Valgrind is perfectly ok with this.
3933 2004-06-12 Martin Quinson
3935 - cleanup the separation between plugin and main code in plugin creation
3937 2004-06-11 Martin Quinson
3939 - Reput hook for raw sockets, needed for BW experiments
3940 - kill a few lines of dead code
3941 [Data description] Interface cleanup
3942 - gras_datadesc_by_name returns the searched type or NULL.
3943 That way, no variable is needed to use a type desc once, which makes
3945 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3946 two parameters were strings (field name, type name), leading to
3948 [Dicos] Interface cleanup
3949 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3950 This is consistant with the dynar API.
3952 2004-04-21 Martin Quinson
3954 - Porting to new standards.
3956 - interface cleanup.
3957 There is no bag anymore, no need to take extra provision to mask the
3958 pointers behind "ID".
3959 Better splitup of functions between files create/exchange/convert.
3960 This is still a bit artificial since convert and receive are so
3961 interleaved, but anyway.
3963 - add a queued message list to procdata (the ones not matching criteria
3965 - factorize some more code between SG and RL wrt procdata
3967 - use gras_exit in example to track memleaks
3968 - get rid of gs_example now that GS is properly integrated into gras
3969 - update run_test to integrate the latest tests (datadesc)
3971 - rename WARNINGn macros to WARNn since it prooved error-prone
3973 2004-04-19 Martin Quinson
3975 - register init/exit functions within gras module mechanism
3976 - send/receive function.
3977 Convertion is not implemented, but short-cutted if not needed.
3978 struct/array elements are sent one by one (instead of block-wise), but
3979 nobody really cares (yet). Get a prototype before optimizing.
3980 - tests (using a file socket) for DD send/receive on:
3981 - base types: int, float
3982 - array: fixed size, string (ie ref to dynamic string)
3983 - structure: homogeneous, heterogeneous
3984 - chained list, graph with cycle
3985 Believe it or not, valgrind is not too unhappy with the results. The
3986 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3990 - Bugfix when using a filename explicitely (instead of '-')
3992 2004-04-09 Martin Quinson
3994 - factorize more code between RL and SG in socket creation
3995 - Complete the implementation and tests of:
3997 o file (only in RL, and mainly for debugging)
3999 I lost 3 days to design a portable address resolver, and then decided
4000 that the prototype mainly have to run on my box.
4001 Addressing portability too early may be like optimizing too early :-/
4003 - use gras_init in the Tests instead of the crappy parse_log_opt
4004 (the latter function is removed)
4005 [Conditional execution]
4006 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4007 [Code reorganisation]
4008 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4009 Build examples against the RL library, since there is no way to disable
4010 its creation for now.
4012 For information, the beginning of coding on GRAS was back in june
4013 2003. I guess that every line has been rewritten at least twice since