1 ----------------------------------------------------------------------------
3 SimGrid (3.24.1) NOT RELEASED YET (v3.25 expected December 22. 2029, 04:19 UTC)
5 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
6 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
7 - GH#305: Unscheduled tasks are still excuted
8 - FG!19: Removing RngStream
10 ----------------------------------------------------------------------------
12 SimGrid (3.24) October 9. 2019
14 The Clean Disk Release.
17 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
18 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
19 thus becomes the default resource to execute I/O operations. Examples
20 and the file system plugin have been modified to use disks instead of
21 storage elements. Storage can still be used but is doomed to disappear.
22 Users are thus adviced to update their simulators accordingly.
23 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
24 for consistency with pthread_barrier_wait()
25 - Host::get_englobing_zone() returns the englobing netzone
26 - Actor::on_destruction is now called in the destructor
27 Actor::on_termination new signal called when the actor terminates
29 - Global signals are now part of the Engine:
30 - on_platform_creation: after config settings, before the XML parsing
31 - on_platform_created: right after the XML parsing
32 - on_time_advance: each time the clock advances
33 - on_simulation_end: after simulation, before cleanups
34 - on_deadlock: as the name implies.
36 - sg_{actor,host,link}_{data,data_set}() now all exist.
37 Use them to attach user data to the object and retrieve it.
40 - Introduce an experimental Wifi model. It sounds reasonable
41 according to the state of the art, but it still has to be properly
42 validated, at least against ns-3 if not against reality.
43 - Improved the usability of ns-3. Several bugs were ironed out.
44 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
45 It is now expressed as 'idle:epsilon:allCores' to properly model the
46 consumption of non-whole tasks on mono-core hosts. More info in the doc.
49 - convert a new set of functions to the S4U C interface and move the old MSG
50 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
53 - Fortran bindings for DVFS have been removed.
54 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
55 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
56 - SMPI can now be selected by cmake's find_module(MPI) with
57 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
58 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
59 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
60 - trace-call-location can be used with TI traces, and replayed, and is compatible
61 with smpi/comp-adjustment-file.
62 - sleep events are now correctly traced/replayed.
63 - Default for trace-call-location is now to use file names and not full paths.
64 To revert to previous behavior (in case of collision of filenames), option
65 "smpi/trace-call-use-absolute-path" can be set to yes.
68 - Use the included xxHash as an hash implem when C++14 is usable.
69 - Option model-checker/hash was removed. This is always activated now.
70 - New option smpi/buffering controls the MPI buffering in MC mode.
71 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
72 This is useful to check for MPI compliance.
75 - New section on plugins: how to define a new one, and existing ones.
78 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
79 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
82 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
83 and <mount> tags. The rationale is finish to decouple what should be handle
84 by SimGrid kernel (i.e., resources and activities) from plugins built in
85 user space (e.g., the file system plugin). This new tag currently accepts
86 only three attributes (id, read_bw, and write_bw). All the other information
87 that was declared with the storage related tags now has to be expressed as
88 properties. An example of platform using this new tag is available at
89 examples/platforms/hosts_with_disks.xml
92 - 'expect signal' can now accept more than one potential signal.
94 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
95 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
96 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
97 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
98 - FG#31: per-actor data
99 - FG#34: SG_BARRIER_SERIAL_THREAD?
100 - FG#35: model-checker does not like buster-produced binaries
101 - FG!13: MC: complete workaround in the error msg seen on modern systems
102 - FG!15: execute_flops now logs compute
103 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
104 - FG!17: ns-3: unblock the right number of communications + others issues
105 - FG!18: Improving the performance of the ns-3 bindings
106 - GH#189: Energy consumption of parallel tasks
107 - GH#219: Error in the throughput of TCP transfer
108 - GH!330: Minor polishing of English: Configuring_Simgrid
109 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
110 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
111 - GH!334: Rework energy plugin, again
113 ----------------------------------------------------------------------------
115 SimGrid (3.23.2) July 8. 2019
118 - Nicer introduction page.
119 - Migrate the "Deploy your application" page to the new doc.
120 - Move Java as a subtree of MSG.
123 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
126 - Drop xbt_dynar_sort_strings().
129 - Really fix FG#26: Turning off a link should raise NetworkFailureException
130 - FG#27: Wrong exception thrown to wait_any when link is turned off
131 - GH#328: Java: Canceling multiple tasks in a single vm/host
133 ----------------------------------------------------------------------------
135 SimGrid (3.23) June 25. 2019
137 The Exotic Solstice Release.
140 - SunOS and Haiku OS support. Because exotic platforms are fun.
141 - Stop setting random seed with srand48() at initialization.
142 - Use addr2line as a fallback for stacktraces when backtrace is not available.
143 - Build option -Denable_documentation is now OFF by default.
144 - Network model 'NS3' was renamed into 'ns-3'.
147 - Simgrid can now hopefully be installed with pip.
150 - wait_any can now be used for asynchronous executions too.
153 - New log appenders: stdout and stderr. Use stdout for xbt_help.
154 - Drop xbt_dict_dump.
157 - 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
158 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
159 - MPI interface is now const correct for input parameters
160 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
161 - Fortran bindings for async collectives.
162 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
165 - Remove option 'model-check/record': Paths are recorded in any cases now.
166 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
167 always sparse. That's an immense gain in memory, and can even be
168 faster because of cache effects. So there is no need to clutter the
169 code to allow the user to go for the unefficient mode.
172 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
173 models proved to be more accurate than these old experiments.
175 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
176 - FG#1: Broken link in error messages
177 - FG#2: missing installation documentation
178 - FG#3: missing documentation in smpirun
179 - FG#6: Python bindings not available on PyPI
180 - FG#7: simple cmake call requires doxygen
181 - FG#8: make python bindings an optional dependency
182 - FG#10: Can not use MSG_process_set_data from SMPI any more
183 - FG#11: Auto-restart actors forget their on_exit behavior
184 - FG#12: -Denable_lto=OFF doesn't disable LTO
185 - FG#13: Installs unstripped file 'bin/graphicator'
186 - FG#14: Installs the empty directory 'doc/simgrid/html'
187 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
188 - FG#17: Dead link in doc (pls_ns3)
189 - FG#20: 'tesh --help' should return 0
190 - FG#21: Documentation link on http://simgrid.org/ broken
191 - FG#22: Debian installation instruction are broken
192 - FG#26: Turning off a link should raise NetworkFailureException exceptions
193 - GH#133: Java: a process can run on a VM even if its host is off
194 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
195 - GH#326: Valgrind-detected error for join() when energy plugin is activated
197 ----------------------------------------------------------------------------
199 SimGrid (3.22) April 1. 2019
201 The Easter Christmas Release (this one is somewhat late).
204 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
205 - Not complete yet: asynchronous activities (amongst others) are still missing
206 - Still ongoing: the interface may change in the future. We need more
207 testers! Please report any glitches.
208 - No new project using Java should start now. Please switch to Python.
211 - Some of the internal cleanups may lead to speed improvements:
212 - The hard limitation on the amount of simulated actors with
213 Java+Mac was removed. Now, the available memory is the only limit.
214 - Our refcounting was tidyied, leading to 10% speedups in some cases.
215 - We are still working on making our code robust to the actor kills
216 and hosts' churn. Things are improving, but it's not perfect yet.
217 - Replaced our own code to display a backtrace (that was forking addr2line)
218 with the Boost.Stacktrace library.
219 You won't see your backtraces without this optional dependency.
220 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
221 - Stop setting random seed with srand() at initialization.
224 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
225 XML file version remains 4.2 since old files are still compatible.
228 - Process termination which was broken at version 3.21 has been repaired.
229 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
230 - Hide the examples into examples/deprecated. New users should use Python.
233 - Drop MSG_process_create_from_stdfunc() from the API.
234 This C++ function was a pimple in the C API, made necessary at some
235 point by the Java bindings. This is fixed now.
236 - Hide the examples into examples/deprecated. New users should use S4U.
237 - MSG_process_create and MSG_process_attach now crash if the host used
238 to run this process is off.
239 - Fix the protype of MSG_process_on_exit()
240 Now use (int,void*) callbacks instead of (void*,void*) ones.
241 The implementation was ways too messy to actually work, I guess.
244 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
245 too often and provide more accurate timings.
246 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
247 - MPI_Alltoallw support
248 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
249 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
250 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
251 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
253 - Basic implementation of generalized requests (SMPI doesn't
254 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
257 - Drop sg_cmdline. Please use xbt_cmdline instead.
258 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
259 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
260 OsSemaphore is implemented in a portable way with C++11 threads.
261 This should allow much more threads to be created at the same time,
262 allowing Mac and Java users to simulate many more actors.
263 - Implement the 'thread' factory with std::thread instead of xbt ones.
264 It is not possible to set the stack size with threads anymore, but
265 -fsplit-stack is the way to go nowadays when using threads.
266 - Drop the xbt_os_thread_t module (now unused)
267 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
268 - Drop xbt_str_join_array().
269 - Drop cunit, use Catch2 instead.
272 - Many cleanups in the kernel::activity namespace. This was long
273 overdue, and shall open the path to many future endeavors.
276 - #132: Java: a process can not shut down its own host
277 - #220: S4U: detached send causes exception if sender terminates soon after sending
278 - #261: Document the parameters of parallel execution's constructor
279 - #300: [s4u] BarrierPtr is missing
280 - #314: SMPI args internal cleanup
281 - #316: Fix a bug related to the CPU utilization of multi-core VM
282 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
283 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
284 - #325: Turning off a host has different behavior on sleeping actors and computing actors
286 ----------------------------------------------------------------------------
288 SimGrid (3.21) October 3. 2018
290 The Restarting Documentation (TRD) Release.
293 - Convert the user manual to Sphinx for improved usability.
294 Unstable version now hosted on https://simgrid.frama.io/simgrid
295 Many glitches remain; stay tuned.
296 - Restrict the Doxygen reference API to the installed header files.
297 (documenting all internals this way is overkill + doxygen scales badly)
298 - New tutorial on S4U (highly inspired from the old MSG one, but with
299 a git repository to fork as a starting point, and a docker image)
300 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
301 - The Developper manual is still to be converted (not compiled until then)
302 - Some parts are still missing in this conversion (such as the
303 platform and deployment sections) while others were blindly converted
304 and would need more love (such as the configuration flags).
305 Things will certainly further improve in the future releases.
308 - Fully reimplement auto_restart mechanism that was utterly broken
309 (fix #22, #131 and #281 that were all very related)
310 - Implement semaphores (fix #303 and #291)
311 - When creating an actor from a function and its parameters,
312 move-only parameters are not allowed anymore, as it would prevent
313 the actor to be restartable if its parameters are consumed on run.
314 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
315 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
316 IO operations that can be started, waited for, or canceled as a regular
318 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
322 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
323 - New host variable: 'core_count'
326 - Remove xbt_os_thread_specific features
327 - Remove portability wrapper to condition variables
328 - Remove xbt_os_thread_yield()
331 - MPICH collective selector now mimics MPICH 3.3b
332 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
333 - MPI_Init(NULL,NULL) is now allowed
334 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
335 - Command-line arguments should now be handled properly in Fortran simulations
336 - MPI Attributes and related callbacks should work in Fortran
337 - Apps using C + Fortran should now work
338 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
339 - Support MPI_Error_String
342 - Due to an internal bug, Msg.run() must now be your last line.
343 We hope to fix it in a future release, and we are sorry for the inconvenience.
346 - #22: Process autorestart seem to only work with CAS01 cpus
347 - #56: Feature request: dynamic SMPI replay
348 - #93: simgrid should not eat --help
349 - #111: How to change the loopback link characteristics?
350 - #116: Platform cluster radical
351 - #117: Platform cluster/host order
352 - #129: Replay_multiple on steroid
353 - #131: Java : autostart does not work
354 - #143: Setting a breakpoint at a given time
355 - #235: Network energy models should be integrated
356 - #264: Add ptask L07 resource tracing
357 - #271: Dynamic and manual replay of SMPI traces
358 - #279: Breakpoints option uses time deltas instead of absolute time
359 - #281: Daemonized actor no longer daemonized after an auto-restart
360 - #282: TIT convertor erase traces when using absolute path in the trace list file
361 - #285: segfault when a s4u actor kills itself with msg process tracing activated
362 - #286: Pajé traces are not exposing the number of cores
363 - #287: Command-line parsing should stop when encountering '--'
364 - #288: MPI_Init(NULL, NULL)
365 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
366 - #290: Method to check if a mailbox has a message ready to be consumed
367 - #291: [s4u] semaphores do not seem available yet
368 - #294: Alltoall collective implementation of mpich
369 - #295: Starting a Host as OFF is not supported
370 - #296: DTD too permissive
371 - #299: Add a s4u barrier example
373 ----------------------------------------------------------------------------
375 SimGrid (3.20) Released June 24. 2018
377 The proxy snake_case() release.
380 - The whole API is now using snake_case() instead of mixing it with camelCase.
381 Compatibility wrappers in place for one year.
382 - We took the opportunity to increase the consistency of the API.
383 Please report any remaining glitches.
384 - New function simgrid::s4u::Host::get_actor_count:
385 Returns the number of actors running on a specific host.
388 - SMPI is now tested with ~45 proxy apps from various sources, with none or
389 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
390 - Replay: The replay file has been re-written in C++.
391 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
392 supported. They are stored in the trace and used when replayed.
393 - Basic support of MPI_Cancel. Robustness not guaranteed.
394 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
395 (only for MPI_COMM_TYPE_SHARED).
396 - New option: smpi/privatize-libs, to add external shared libs to be privatized
397 by SMPI. They will be copied locally and loaded separately by each process.
398 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
399 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
400 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
401 with internal index optind with multiple processes. Only works if MPI_Init has
403 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
404 syntax without preprocessing tricks.
407 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
408 behavior tracing is no longer limited to MSG
409 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
410 this means that traces now need to be updated or re-obtained; both Irecv
411 and Isend lines in your traces have now in total 5 mandatory fields:
412 <rankid> <command> <to/from rankid> <tag> <size>
413 To update your traces, it suffices to add a 0 for the tag here.
414 - TIT now also supports waiting for a distinct request via MPI_Wait.
415 Wait/Test now wait for a specific request, not just the last one that was
416 issued. This unfortunately means another update, because we need to
417 identify which request you want to wait for. We do this via the
418 triplet (sender, receiver, tag), which needs to be added:
419 <rankid> <command> <sender> <receiver> <tag>
420 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
421 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
424 - The deprecation of MSG is ongoing (but this should not impact you).
425 Many MSG functions are now simple wrappers on the C API of S4U. If
426 you wish to convert your code to S4U, find the S4U counterparts of
427 your MSG calls in src/msg/msg_legacy.cpp.
428 - MSG can still be used, but won't evolve anymore.
431 - Allow to run the Link energy plugin from the command line with
432 --cfg=plugin:link_energy
433 - Rename Energy plugin into host_energy
434 - Rename Load plugin into host_load
437 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
438 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
439 - Improve performance when handling timeouts of simix synchros.
442 - Config: the C API is now deprecated (will be removed in 3.23), and
443 the C++ API has been slightly improved.
446 - Fix several build issues on OSX.
447 - Move simgrid_config.h to simgrid/config.h (old header still working)
450 - #143: Setting a breakpoint at a given time
451 - #150: Inconsistent event names in SMPI replay
452 - #258: daemonized actors hang after all non-daemonized actors have completed
453 - #267: Linker error on unit_tmgr
454 - #269: SMPI: tracing of MPI_Wait/all/any broken
455 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
457 ----------------------------------------------------------------------------
459 SimGrid (3.19.1) Released March 22 2018
461 The Fixed ABI Release.
463 As you may know, we are currently refactoring SimGrid in deep.
464 Upcoming SimGrid4 will be really different from SimGrid3: modular,
465 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
467 Our goal is to smooth this transition, with backward compatibility
468 and automatic update paths, while still progressing toward SimGrid4.
470 SimGrid remains open during works: The last pure SimGrid3 release was
471 v3.12 while all subsequent versions are usable alpha versions of
472 SimGrid4: Existing interfaces remain unchanged, but the new S4U
473 interface is budding and the internals are deeply reorganized.
475 Since 2015, we work hard to reduce the changes to public APIs. When
476 we need to rename a public library symbol in S4U, we let your
477 compiler issue an explicative warning when you use the deprecated
478 function. These messages remain for four releases, i.e. for one full
479 year, before turning into an error. Starting with v3.15, your can
480 also adapt to API changes with the SIMGRID_VERSION macro, that is
481 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
483 Starting with this v3.19.1, our commitment to reduce the changes to
484 the public interfaces is extended from the API to the ABI: a program
485 using only MSG or SimDag and compiled against a given version of
486 simgrid can probably be used with a later version of SimGrid without
487 recompilation. We will do our best... but don't expect too much of
488 it, that's a really difficult goal during such profund refactoring.
490 The difference between v3.19 and v3.19.1 is that the former was
491 accidentally breaking the ABI of MSG, while the later is restoring
494 S4U and kernel APIs will still evolve until SimGrid4, with one-year
495 deprecation warnings as currently. In fact, cleaning up these
496 interfaces and converting them to snake_case() is one release goal of
497 v3.20. But don't worry, we are working to smooth this upgrade path.
499 In summary, new projects should start with S4U to benefit of the
500 future, but old MSG projects should still be usable with no change.
502 ----------------------------------------------------------------------------
504 SimGrid (3.19) Released March 20 2018
506 The Moscovitly-cold Spring release.
508 SMPI user-visible improvements:
509 - Support of MPI_Type_create_subarray
510 - Unimplemented MPI calls now abort (instead of issuing a warning)
511 Some harmless calls around errhandlers still only issue a warning.
512 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
513 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
514 is now possible (mmap privatization was refactored to this end).
515 Adding new SMPI instances should also be technically doable now.
517 SMPI internal cleanups:
518 - Removed index notion from SMPI -- no more "getPid() - 1"!
519 In tracing files, actors are called rank-<PROCESS-ID> but these are
520 NOT the real ranks (tracing several communicators is not supported yet).
521 - Deprecate smpi_process_index() (will be removed in 3.22)
522 - Better testing on classical proxy apps, through an external project:
523 https://github.com/simgrid/SMPI-proxy-apps/
526 - Use a graphical TOC to make it easier to find the documentation you need
527 - Revamp the MSG tutorial
530 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
531 - Remove parameter of MSG_process_killall().
532 Resetting the PID was bogus anyway (several actors could have the same PID).
535 - Execution->setHost() can be called after start(), to migrate it.
536 - Comm::test_any() is now implemented.
537 - s4u::Actor now has onCreation() and onDestruction() signals.
538 - Install some kernel header files for the users' plugins and more.
539 Warning, their API is really not stable yet.
540 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
541 include/simgrid/kernel/routing/*Zone.hpp
544 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
546 Build System and other cleanups:
547 - Remove unused run-time parameter "tracing/onelink-only".
548 - SimGrid now works with intel compilers (MC still buggy)
549 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
550 - tesh kills subprocesses on timeout (but not on windows)
553 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
556 - #194: Feature request: simgrid::s4u::Comm::test_any()
557 - #245: migrating an actor does not migrate its execution
558 - #253: Feature Request: expose clusters as objects
559 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
560 - #255: Tesh broken on Windows
561 - #256: Modernize FindSimGrid.cmake
562 - #257: Fix (ab)use of CMake install
564 ----------------------------------------------------------------------------
566 SimGrid (3.18) Released December 24 2017
568 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
570 MSG IS NOW DEPRECATED. Please use S4U for new projects.
571 - Support for MSG should not be removed from SimGrid before 2020,
572 but future evolutions will be limited to the integration of
573 user-provided patches.
576 - s4u::onDeadlock() signal, run before stopping the simulation.
577 - s4u::Exec: asynchronous executions (abstraction of background
579 - s4u::Host->getLoad() returns the achieved speed in flops/s
582 - Link::name() is deprecated and will be removed in v3.21, use
583 Link::getCname() instead.
584 - Mailbox::getName() changed to return a std::string, use
585 Mailbox::getCname() instead to get a char*.
586 - Storage::getName() changed to return a std::string, use
587 Storage::getCname() instead to get a char*.
588 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
589 to have both allocation and deallocation of the map in user space.
592 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
593 The amount of flops remaining to do is a vector, not a scalar.
594 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
595 name implies on both sequential and parallel tasks.
596 - Both changes fix GitHub's #223 using PR #237 as a basis.
597 Thanks Michael Mercier.
598 - Most examples were converted to S4U and hidden elsewhere at tests.
599 Rationale: we still want MSG to work; we want newcomers to use S4U.
602 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
603 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
606 - Switch to the faster dlopen privatization mechanism by default
607 - Documentation improvments
610 - Live migration is getting moved to a plugin. Dirty page tracking is
611 the first part of this plugin. This imply that VM migration is now
612 only possible if one this function is called:
613 - C/MSG: MSG_vm_live_migration_plugin_init()
614 - C/C++: sg_vm_live_migration_plugin_init()
615 - Java: Msg.liveMigrationInit()
616 For C and C++, "simgrid/plugins/live_migration.h" also has to be
620 - Define class simgrid::xbt::Path to manage file names.
621 - Removed unused functions:
622 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
623 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
624 xbt_graph_export_graphviz()
625 - xbt/str.h: xbt_str_join()
626 - Remove unused datatypes:
627 - xbt/heap.h: use std::priority_queue or boost::heap instead
628 - xbt/swag.h: use boost::intrusive::list instead
631 - New link_energy plugin for the consumption of the links.
632 - All of the operations on files and storage contents have been
633 packaged into a plugin (src/plugins/file_system). The current
634 public interface can be found in
635 include/simgrid/plugins/file_system.h
636 To use these functions you now have to initialize the plugin by
637 calling MSG_storage_file_system_init() just after calling
638 MSG_init() or sg_storage_file_system_init() just after creating
642 - Remove the undocumented/untested tag <include>
645 - Remove viva specific tracing as the tool is no longer maintained
648 - #248: Kill of finished processes leads to segfault
649 - #240: xbt_cond_wait_timeout should gracefully return for C apps
650 - #239: please implement signal s4u::onDeadlock()
651 - #230: segfaults when exit() without run()
652 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
653 - #223: MSG_task_get_flops_amount() not working with parallel tasks
654 - #222: Actor::kill() doesn't really kill and segfaults
655 - #221: odd LMM warning when killing an actor
656 - #120: Memory leak when the processes are forcefully killed
658 ----------------------------------------------------------------------------
660 SimGrid (3.17) Released October 8 2017
662 The Drained Leaks release: (almost) no known leaks despite the tests.
664 Even very long simulations will preserve your memory: our testsuite
665 stresses SimGrid for over 45mn on fast machines for a coverage of
666 over 80%, and there is only one single known leak, of about 4kb.
670 - Comm.detach(): start and forget about asynchronous emission. A cleanup
671 handler may be given to free resources if the comm cannot be completed.
672 - this_actor::send(mailbox) is now mailbox->put()
673 - New: simgrid::s4u::Comm::wait_all()
674 - New: Host.execute() for remote executions.
677 - Unused option network/sender-gap is removed.
680 - Deprecate MSG_task_isend_with_matching(): This unused feature
681 really complicates our internals. Will be removed in v3.20.
684 - Improved context termination. It is now possible to free resources, even
685 when a process is forcibly killed.
686 - ContextBoost: add support for Boost versions above 1.61.
689 - Bring back run-time option --cfg=exception/cutpath to remove exception
691 - Removed unused functions:
692 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
693 xbt_str_rtrim(), xbt_str_trim().
694 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
697 - Removed header files obsolete since SimGrid 3.12:
698 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
700 - #3: SD_exit should be made optional
701 - #120: Memory leak when the processes are forcefully killed
702 - #159: Threading test regression in Actor refcounting
703 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
704 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
705 - #186: Actor::killAll() segfaults if some process is blocked on wait()
706 - #191: VM migration and pstate
707 - #192: Updating the energy consumptions of all hosts crashes with VMs
708 - #195: All actors have PID=0 in the logs
709 - #204: Sometimes segfault with thread contexts and mmap privatization
710 - #222: Actor::kill() doesn't really kill and segfaults
711 - #225: Actor::kill() doesn't really kill when victims are doing a join()
713 ----------------------------------------------------------------------------
715 SimGrid (3.16) Released June 22. 2017.
717 The Blooming Spring Release: developments are budding.
720 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
721 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
722 - Java JAR file should be finally fully working on Mac OSX too.
724 XML platforms: Switch to platform v4.1 format.
725 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
726 - <zone> can be used as a synonym for the now deprecated <as>
727 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
728 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
729 - <actor> can be used as a synonym for the now deprecated <process>
730 - state_file and avail_file periodicity is now easier to express
731 (check the documentation) the old behavior should still work.
732 * Storage: not backward compatible, but it seems that nobody ever used it
733 - Remove attribute 'content_type' of <storage_type>: was never used
734 - Make attribute 'model' of <storage_type> optional: for future usage
735 - Remove Bconnection model property: was never (in)validated. Replaced by
736 the maximum of the read and write bandwidth as a resource constraint.
739 - New and Backwards Compatibility break:
740 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
741 When one wants to get the list of tasks whose states have changed during a
742 simulation round, s/he has to allocate and free a dynar and use it as
743 argument to this function. The former SD_simulate (double how_long)
747 - Allow multicore VMs, with the correct sharing (unless you overcommit)
748 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
751 - New function to update the consumption of all hosts at once.
752 - Fix the model for multi-core hosts, linear in the amount of busy
753 cores with an abnormality for pIdle (see paper).
754 BUG: Still not satisfactory for ptask on mono-cores.
757 - The netzone are now available from the MSG API.
758 The old names still work, but are now deprecated.
761 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
762 - New option: smpi/keep-temps to not cleanup temp files
763 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
764 mallocs mode is used, allows to change the size of the fake file used
765 (default 1MB), to potentially limit the number of mappings for large runs.
766 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
767 - Fortran ifort and flang compilers support
768 - New RMA calls supported (experimental):
769 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
770 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
771 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
772 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
773 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
774 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
777 - New: this_actor::isMaestro() returns whether we are in kernel mode.
778 - Behind the scene, ActivityImpl are now automatically refcounted.
779 Removing this nasty bottleneck will greatly simplify our internals.
782 - Replay: New function xbt_replay_action_get():
783 Retrieve the function previously associated to an event type.
784 - DROPPED FUNCTION: xbt_str_varsubst()
785 - DROPPED MODULE: strbuff. We don't need it anymore.
786 - DROPPED MODULE: matrix. We don't need it anymore.
787 - DROPPED MODULE: lib. We don't need it anymore.
789 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
791 SimGrid (3.15) Released March 22. 2017
793 The Spring Release: continuous integration servers become green
794 We fixed even the transient bugs on all target architectures:
795 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
796 Windows, FreeBSD, NetBSD.
798 - Rename NetCards to NetPoints.
799 This was intended to help NS3 users, but that's not a netcard.
800 That's a point in the routing algorithm, let's avoid wrong simplifications.
803 - New: MSG_process_yield(). Stop and yield to other processes.
804 - New: MSG_process_daemon(). Daemon processes are automatically killed
805 when the last non-daemon process terminates
806 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
807 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
808 to make room for the upcoming network energy plugin.
809 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
812 - Ensure that an actor can kill itself with Process::exit()
813 - Kill the obscure NativeException. Nobody want to survive the issues
814 it denotes, so use JniException that is a RuntimeException (not to
815 be caught explicitly).
816 - Partial bug fix in initialization. SimGrid flags on command line were
817 consumed at C level but stayed in the original Java String[] args.
818 This could mess users' args[i] if SG flags were not put at the end of
820 The SimGrid flags are now removed from the Java arguments. However,
821 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
822 if args.length is greater than the number of YOUR OWN ARGUMENTS.
823 It might be if you have --log or --cfg flags in the command line.
824 - Fix numerous memleaks all around the place. In particular, around VMs.
828 - s4u::onPlatformCreated: right before the simulation starts
829 - s4u::onSimulationEnd: right after the main simulation loop
830 - s4u::onTimeAdvance: right after a clock change (time is discrete)
831 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
832 event from the availability_file changes the avail speed.
833 - Links are now usable from s4u
834 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
835 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
836 - New examples: The conversion of MSG examples to S4U has begun
837 - Actors: create, kill, migration, and suspend
838 - Applications: master-worker and token-ring
839 - Action replay: communications and storage
840 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
843 - Backwards Compatibility breaks
844 - The SD_route_get_list and SD_route_get_size functions have been removed.
845 They are replaced by the unique
846 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
847 The route from src to dst is built in the links dynar whose size can
848 be obtained with xbt_dynar_length.
849 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
850 sg_host_route_bandwidth and sg_host_route_latency.
851 Macros ensure the backwards compatibility, but you should fix your code
854 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
855 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
856 - Bump our claim of support from MPI 1.1 to MPI 2.2.
857 We don't support 100% of it, but it should be enough. Contact us if not.
858 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
859 naive and probably inaccurate)
860 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
861 mapped onto a unique small file using some system magic.
862 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
863 - Fortran: cleanups, fixes, support of user-added operations
864 - RMA: MPI_Accumulate are applied in correct order.
865 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
866 - Better support for MPI_IN_PLACE
867 - Support for MPI_Win attrs and keyvals.
868 - Support MPI_Comm_group_create, MPI_Type_size_x
869 - MPI ops should be properly applied to all allowed types
870 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
871 algorithm by a non failing one from mpich.
874 - Kill the fifo data container: we don't use it anymore.
877 - A <cluster> can now be created with different speed values to represent pstates.
878 see examples/platforms/energy_cluster.xml for an example.
880 SimGrid (3.14.159) Released December 28. 2016
882 The previous release was only a crude approximation of a Pi release;
883 this one should be a bit better.
885 - Revert a last minute change that broke on Mac OSX.
886 - Fix the changelog and NEWS files.
887 - Improve the documentation, in particular of the routing module.
888 - Integrate some patches flying around in Debian.
890 SimGrid (3.14) Released December 24. 2016
893 * The whole documentation was reworked and reorganized. There is
894 still much room for improvement here, but we're on it.
897 * We now need python3 for our testing infrastructure.
898 * Model checking needs libevent
901 * Now works on FreeBSD
904 * Privatization now works on FreeBSD
906 * Privatization is now activated by default in smpirun
907 You can turn it off with -no-privatize if something goes wrong.
909 * Call-location tracing for SMPI.
910 You can add the exact location (filename / linenumber) of an MPI call to
911 your trace files and slow down or speed up the simulation between two
912 consecutive calls by using an adjustment file (see the documentation).
914 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
915 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
916 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
917 values were ignored in some cases. The timings of these functions can now
918 be significantly different.
920 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
921 smpi/running-power is renamed to smpi/host-speed
923 * smpi/grow-injected-times option to enable or disable multiplication of the
924 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
925 make simulation less precise (but also much faster).
927 * smpirun script should be (much) faster for large deployments.
929 * SMPI tracing: fixed issue with poor matching of send/receives.
931 * Replay: Fix broken waitall
933 New functions and features
934 * MSG_parallel_task_execute_with_timeout, to timeout computations.
936 Dropped / renamed functions and features
937 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
938 - MSG_mailbox_is_empty() -> Mailbox::empty()
939 - MSG_mailbox_front() -> Mailbox::front()
940 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
941 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
942 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
943 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
944 * Don't pass the free_f parameter to property related functions:
945 - MSG_host_set_property_value()
946 - MSG_as_router_set_property_value()
947 - MSG_storage_set_property_value()
948 * VM properties. Since msg_vm_t are msg_host_t, just use
949 MSG_host_get_property_value() and friends
950 * VM I/O related things:
951 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
952 - Unimplemented save/restore methods
953 * MSG_as_router_get_property_value() was redundant with
954 MSG_environment_as_get_property_value().
955 - Removed MSG_as_router_*propert*() functions
956 - Added MSG_environment_as_set_property_value() for consistency
957 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
958 Well, they are still there for now, but deprecated with a warning.
959 Please switch to xbt_dict_new_homogeneous() before this is removed
961 * Task affinity. Its intended behavior (that was very badly tested
962 and probably not really working) was deceiving what most users
963 would have hoped here.
964 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
965 * Option network/coordinates is now useless and should be dropped.
968 * Added option storage/max_file_descriptors to allow more than 1024 files opened
971 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
972 * Added Dragonfly topology support
975 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
976 EiBps, EBps, Eibps, Ebps for bandwidth.
977 They may become useful to some lucky ones.
980 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
981 * ex/app_tokenring: new example, very similar to the MSG Token Ring
982 * ex/async_waitAll: new example, on asynchronous communications
985 * Memory usage should be decreased for simulations with a large number
986 of processes. This also helps for SMPI.
988 SimGrid (3.13) stable; urgency=low
990 The Half Release, a.k.a. the Zealous Easter Trim.
992 - We removed half of the lines, that were mostly experimental cruft.
993 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
994 (not counting blanks and comments -- according to openhub.net)
995 - The internals are now compiled in C++ (and will soon be clean C++)
996 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
997 59 klines of C, etc. We added only 29 klines of C++ in replacement.
999 * Backwards Compatibility breaks
1000 - Removed Lua simulation bindings (switch to C or Java for that).
1001 Lua can still be used to describe platforms
1002 - Removed Java kernel plug-ins.
1003 Will be reintroduced after the ongoing major internals reorg.
1005 - the following functions were removed.
1006 They were too specific and should be reimplemented in a generic
1007 way, with filter function.
1008 - MSG_task_listen_from_host
1009 - MSG_mailbox_get_count_host_waiting_tasks
1010 - MSG_mailbox_put_with_timeout was removed.
1011 Please use MSG_task_send_with_timeout instead.
1013 - the SD_application_reinit function was removed. It has been a noop for a while.
1014 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1015 replaced by a more flexible mechanism.
1016 - The following functions thus do not exist anymore
1017 - SD_workstation_get_access_mode
1018 - SD_workstation_set_access_mode
1019 - SD_workstation_get_current_task
1020 - Basic estimation functions have been removed but can easily be replaced
1021 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1022 - SD_workstation_get_computation_time => amount / sg_host_speed()
1024 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1025 Use VM.getSpeed()*load/100 for the legacy behavior.
1027 - option enable_tracing was removed. It was not doing anything for a while.
1028 - In the ModelChecker:
1029 - the model-checker now ptraces the model-checked process which means
1030 you cannot use a debugger on the latter anymore (we might make this
1031 optional in the feature);
1032 - removed soft-dirty page tracking;
1033 - remove model-checked side snapshot management,
1034 MC_snapshot() and MC_compare_snapshot();
1035 - keep the MC_cut() function as a stub (it was not really working
1036 in the previous release).
1039 * All options are consistently in kebab-case. Old names are kept as alias.
1042 * Switch to platform v4 format.
1043 - Rename from 'power' to 'speed' the attributes describing the amount of
1044 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1045 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1046 - In <host> and <link>, attributes availability and state are gone.
1047 It was redundant with state and availability traces, and with peak values.
1048 - In <cluster>, attributes availability_file and state_file are gone.
1049 It was too complex and unused.
1050 - Kill <gpu>. Was not doing anything.
1051 - The DOCTYPE points to the right URL:
1052 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1053 (the file at this address now documents the changelog since its v1)
1054 - A warning is emitted for unit-less values (they are still accepted).
1055 - speed. Default: 'f' or 'flops'. Also defined:
1056 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1057 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1058 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1059 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1060 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1061 - latency. Default: 's' second. Also defined:
1062 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1064 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1065 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1068 * s4u::Host is now the preferred public interface to the Host features.
1069 sg_host_* functions are C bindings to the exact same behavior
1070 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1073 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1074 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1075 They were deprecated since a few years, and probably did not even compile anymore.
1078 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1079 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1080 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1081 encouraged to update their codes. A list of the modified functions can be found at the end of
1082 include/simgrid/simdag.h
1085 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1086 the lambda or closure passed as a parameter will run in kernel mode.
1087 All the callback functions should be rewritten to that interface at some point.
1090 * Reorganizing and cleaning the internals all around the place.
1093 * Remove old default barrier/bcast buggy algorithms (see #18407)
1094 * Various bug fixes to handle more codes
1095 * Remove the need for the --foreground option of smpirun (it is still
1096 accepted for backward compatibility).
1099 * Kill the setset data container: MC don't use it anymore.
1100 * Kill the queue data container: it made more sense with GRAS.
1101 * Kill the xbt_peer_t data type: it's useless without GRAS.
1102 * Kill rm_cb feature of config sets: it was never useful.
1103 * Kill graphxml parsing feature. It was not used.
1104 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1106 - xbt_dynar_sort_strings(), when the content is char*
1107 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1108 They throw exceptions on invalid input;
1109 * C++ support for declaring CLI flags (simgrid::config::Flag);
1110 * class for abstracting different signal backends (simgrid::xbt::signal).
1111 with no external dependencies (we need very simple signals).
1114 * refactoring and cleanup of the code;
1115 * ongoing process to cleanly separate the model-checking algorithms
1116 from the code model-checking support.
1118 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1120 SimGrid (3.12) stable; urgency=low
1122 The Facelift Release.
1125 * Require g++ v4.7 at least to not speak prehistorical C++.
1126 * Require Boost 1.48 (for signal2 component).
1127 * Java must be version 7 at least when activated.
1128 * Builds on Windows again (including Java bindings).
1129 * Tracing is now always enabled (no way to turn it out)
1130 * Remove GTNetS. It was not working anyway.
1131 * Various cleanups in the cmake scripts.
1132 * Move headers around to sort them out on installed systems:
1133 - instr/instr.h -> simgrid/instr.h
1134 - instr/jedule/* -> simgrid/jedule
1135 - simdag/datatypes.h was removed
1136 - simdag/simdag.h -> simgrid/simdag.h
1137 - msg/datatypes.h was removed
1138 - msg/msg.h -> simgrid/msg.h
1141 * Interface improvement:
1142 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1143 - Sanitize the interface in MSG_task_ module:
1144 - Merge two functions that were close enough but misleading:
1145 set_compute_duration(t) -> set_flops_amount(t)
1146 get_remaining_computation(t) -> get_flops_amount(t)
1147 - set_data_size(t) -> set_bytes_amount(t)
1148 get_data_size(t) -> get_bytes_amount(t)
1149 - Massive cleanups in the functions related to the energy
1150 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1151 - New: MSG_host_get_pstate()
1153 - msg/energy/onoff: switching hosts on and off
1155 * Interface improvement:
1156 - Rename Host.isAvail() to Host.isOn()
1157 - Rename Process.currentProcess() to Process.getCurrentProcess()
1158 - Rename Task.setDataSize() to Task.setBytesAmount()
1159 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1161 - #18874: Actually allows the GC to reclaim tasks
1165 - SIMIX_process_throw: raises an exception in a remote process
1166 * Refactoring: Separate sub-modules
1167 - libsmx: the public interface, as libc in a real system
1168 - popping: the strange dance that converts a user request into a kernel handling
1169 - smx_context_*: the virtualization mechanisms that embed the user code
1170 - smx_*: the handling of each simcalls
1171 * Interface cleanup:
1172 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1173 * Rename smx_action_t into smx_synchro_t, making explicit that these
1174 things are used to synchronize processes with their environment.
1175 For example, a communication is a sort of synchronization involving
1176 the communicating processes (that may block until the exchange) and
1177 the platform. The same can be said from computations, etc.
1179 - #18888: segfault when a process ends before its kill_time
1183 - 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.
1184 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1185 - MPI_Type_set_name, MPI_Type_get_name
1186 - MPI_*_c2f and MPI_*_f2c functions
1187 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1188 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1189 - Activate a lot of new tests from the mpich 3 testsuite
1191 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1192 - InfiniBand network model added: Based on the works of Jerome Vienne
1193 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1194 - When smpi/display_timing is set, also display global simulation time and application times
1195 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1196 * Collective communications
1197 - 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
1198 - 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/).
1199 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1201 - Replay now uses algorithms from wanted collective selector
1202 - Replay can be used with SMP-aware algorithms
1203 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1204 - 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.
1206 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1207 - Correctly use loopback on fat-tree clusters
1208 - Asynchronous small messages shouldn't trigger deadlocks anymore
1209 * Energy/DVFS cleanup and improvement
1210 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1211 - new: smpi_get_host_pstate()
1214 - "Full" network optimization flag was broken since Surf++
1215 - Better handling of precision flags in maxmin
1216 - Fix bug causing sometimes "Impossible" errors
1217 - Properly pass cluster properties to included hosts
1218 * Improvement of the Energy plugin.
1219 - Always update the consumption before returning that value
1220 - New property: watt_off to denote the dissipation when the host is off
1221 - New functions getWattMinAt and getWattMaxAt to retrieve the
1222 dissipation of pstates that we are not currently at.
1223 * Java: class NetworkLink renamed to Link
1224 * New function: simcall_process_get_kill_time()
1225 * Massive rename s/workstation/host/
1226 - That's intrusive, but that's good for the project consistency. Sorry.
1227 - Change config option "workstation/model" into "host/model"
1230 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1231 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1232 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1233 - xbt_dirname and xbt_basename for non-POSIX systems
1235 * The model checker now runs as a separate process.
1236 * The model checker runs must now be launched with the new simgrid-mc program.
1237 * Record/Replay: the MC can display a textual representation of a path in the
1238 execution graph. It can then be replayed outside of the model checker.
1240 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1242 SimGrid (3.11) stable; urgency=low
1247 * Normalizing pointers addresses tool for better diff between logs
1250 * Add cloud examples using new VMs
1251 - examples/msg/cloud/two_tasks_vm.tesh
1252 - examples/msg/cloud/simple_vm.tesh
1253 - examples/java/cloud/cloud.tesh
1254 - examples/java/cloud/migration/migration.tesh
1255 * Add java surf examples:
1256 - examples/java/surfPlugin/surf_plugin.tesh
1257 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1258 - examples/java/surfCpuModel/surf_cpu_model.tesh
1259 * Add SMPI+MSG example:
1260 - examples/smpi/smpi_msg_masterslave/
1265 - msg task destroy cancel test
1266 - msg_host on/off test
1267 * Move all tests in testsuite to teshsuite (adding tesh files)
1268 * Restructure teshsuites
1269 - one folder for each kind of test
1270 * Restructure AddTests.cmake
1272 - structure the order of tests (with sections)
1275 * Add virtual machine
1276 - creation of a VM on a PM
1277 - migration of a VM from a PM to another PM
1279 - MSG_process_join(msg_process_t process, double timeout)
1280 - msg_bar_t MSG_barrier_init(unsigned int count)
1281 - int MSG_barrier_wait(msg_bar_t barrier)
1282 - void MSG_barrier_destroy(msg_bar_t barrier)
1283 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1284 * New option "msg/debug_multiple_use" to help debugging when a task is used
1287 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1289 - Increase code coverage in test suites
1293 * Protect context stacks against stack overflow. The number of protected memory
1294 pages allocated on the top of each stack (1 by default) can be configured
1295 with the new command line option --cfg=contexts/guard_size:<value>.
1296 * Simcalls are now generated by a python script that
1297 - generates files included by SimGrid
1298 - checks that all the functions exist, and proposes prototypes
1300 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1301 host_set_data, host_get_data
1303 - simcall_process_join(smx_process_t process, double timeout)
1304 * Fix bug where sleeping processing could not be suspended.
1307 * Translate surf models from C to C++
1308 - Generic classes for all models: Model, Resource, Action
1309 - A generic interface for each kind of model (CPU, Network, Storage
1310 Workstation, WorkstationVM)
1312 * Translate surf routings from C to C++
1313 * Add callbacks using sigc++ or boost::signals2
1314 - Add callback functions for resource creation/destruction
1315 - Add callback functions for action state change
1316 - Handle Energy as a plugin
1317 * Replace swag by boost::intrusive
1318 * Add new routing models for clusters. For documentation, see
1319 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1320 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1321 parameters for cluster tag
1322 - Fat trees, with topology="FAT_TREE" and
1323 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1324 - see examples/platforms/torus_cluster.xml and
1325 examples/platforms/fat_tree_cluster.xml
1326 * More documentation
1329 * Hostfiles support host:nb_processes construct to deploy several processes on
1331 * Collective communication algorithms should not crash if used with
1332 improper number of nodes and report the error.
1333 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1334 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1335 MPI_Dims_create, MPI_Cart_sub are supported.
1336 * New interface to use SMPI programmatically (still depends on MSG for
1337 some parts, see examples/smpi/smpi_msg_masterslave):
1338 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1342 * Global variables privatization in MPI executables is now performed at runtime
1343 with the option smpi/privatize_global_variables (default:no).
1344 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1345 dynamic libraries loaded by the application are not privatized (static
1346 linking with these libraries is advised in this case)
1349 * Options defined in XML work correctly now.
1352 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1353 copied into simgrid.jar or not (ON by default). Use this option if you want
1354 to reduce the size of the installed simgrid.jar, *and* the native libraries
1355 are kept installed elsewhere.
1356 * Surf binding with SWIG (code generated in maintainer mode only):
1357 - plugin to handle callbacks
1358 - CPU model only for the moment
1361 * Supernovae build mode is definitively removed. It was used to improve
1362 inlining and inter-module optimizations. It is nowadays superseded by
1363 link-time optimizations commonly available in compilers.
1364 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1365 * Add boost dependency for surf++
1366 * Add new macro for tests
1367 - ADD_TESH(name <tesh_args>)
1368 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1372 - xbt_bar_t XBT_barrier_init(unsigned int count)
1373 - int XBT_barrier_wait(xbt_bar_t barrier)
1374 - void XBT_barrier_destroy(xbt_bar_t barrier)
1375 * Make the xbt_os_time module public
1377 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1379 SimGrid (3.10) stable; urgency=low
1381 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1384 * Reintegrate Java to the main archive as desynchronizing these
1385 package is not acceptable anymore (Java is now considered stable)
1386 * Add explicit synchronization facilities through semaphores
1387 * Bug fix: Task.setDataSize() only changed the C world, not the value
1388 cached in the Java world
1391 * Dramatically change the way files are handled. API and internals changed, but
1392 this part of MSG was not considered as production grade either.
1393 * Add explicit synchronization facilities through semaphores
1394 * Add a new function MSG_host_get_process_list()
1395 * Preliminary DVFS support (see examples/msg/energy/ for details)
1398 * SMPI is now included directly in the libsimgrid as the windows
1399 linker doesn't force us on splitting it anymore.
1400 * Improvements of the SMPI replay tool:
1401 - Most of the collective communications are now rooted in the same process as
1402 in the original application.
1403 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1404 used until now). Multiple data types can now be used in a trace.
1405 - The replay tool now supports traces produce either by TAU or a modified
1407 - Bug Fix: the compute part of the reduce action is now taken into account.
1408 - Gatherv collective is now supported
1409 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1410 of smpirun outputs time independent traces for the current run. One file
1411 is created per process. If too many processes are simulated, this behavior
1412 can be changed to one file for all processes by using the
1413 tracing/smpi/format/ti_one_file flag
1414 * smpirun generates the host file if needed (with given host count and platform)
1415 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1416 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1417 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1418 --cfg=smpi/coll_selector:(mpich/ompi)
1419 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1421 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1422 happening inside a collective SMPI call.
1423 * Fix the behavior of complex data types handling.
1424 * Make MPI_Wtime another synchronization point to take computations into
1426 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1427 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1428 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1429 (no privatization of global variables yet)
1430 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1433 * Verification of liveness properties is now available for SMPI applications
1434 (in addition to MSG applications)
1435 * Bugged examples using SMPI in examples/smpi/mc/
1436 * Add --cfg=model-check/visited option. Allows the verification of infinite
1437 programs. Detection of loops in the execution thanks to the system state
1438 comparison and reduction of the state space to explore. Can be combined with
1439 DPOR for safety properties.
1442 * Allow to change SimGrid configuration (see --help) within the code
1443 thanks to SD_config() as it can be done in MSG.
1444 * Add a new function SD_task_set_amount() upon user request.
1447 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1448 * Remove rule based routing (no more PCRE dependency)
1449 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1450 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1451 * Add a loopback_bw and loopback_lat options to cluster tag.
1452 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1453 routing. Now peers are encapsulated in an AS and have their own private
1454 router but this is transparent.
1457 * Our own implementation of getline is renamed xbt_getline, and gets
1458 used even if the OS provide a getline(). This should reduce the
1459 configuration complexity by using the same code on all platforms.
1460 * New type: xbt_cfg_elm_boolean.
1461 * Allow to use yes/no for boolean configuration options in the command line.
1462 * Allow to disable SimGrid cleanups at exit from command line option.
1463 There are situations where one may want a simulation to end with an exit.
1464 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1465 annoying when scripting around the simulator. Adding a
1466 --cfg=clean_atexit:no allows to circumvent this issue.
1469 * Lots of memory leaks were corrected in this release.
1470 There is no known memory leaks anymore, in all of our 600+ tests.
1471 * New command line option --version, to get SimGrid version information.
1472 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1474 * Supernovae builds are deprecated, and expected to be removed in the next
1477 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1479 SimGrid (3.9) stable; urgency=low
1481 The Grasgory release: GRAS is really dead now.
1483 * Complete overhaul of the internal host structures scheme.
1486 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1487 was considered as experimental and badly maintained since then.
1488 * Keeping it was thus a trap to our potential users, that could take
1489 it instead of MSG or SMPI by mistake despite is pity state.
1490 * GRAS seems to have very few users (if any), and no one volunteered
1491 to maintain it further. It also induces a lot of XBT code (for
1492 portability sake), that must be maintained too.
1493 * For all these reasons, we killed GRAS. If someone wants to revive it
1494 in the future, don't cry, our git history still remembers of GRAS.
1497 * Major overhaul. Merge our documentation again as time proved that
1498 splitting it was really not helping our users.
1499 * Further improve the developer documentation to help newcomers
1500 hacking on SimGrid itself. The user documentation (and in
1501 particular, the beginner documentation) is still in a sorry state.
1504 * Now works on Windows too!
1505 * Much more extensive test suite, from MPICH
1508 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1509 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1510 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1511 * Bug fix: let task be scheduled when the last dependency to be solved is
1512 a control dependency.
1513 * Remove SD_load_environment_script function.
1514 Use the C sg_platf function if you want to declare a platform
1518 * New function: MSG_process_get_number()
1519 * Old function documented: MSG_config()
1520 * Remove MSG_load_platform_script function
1521 Use the C sg_platf function if you want to declare a platform
1525 * Change the default value of the TCP_gamma constant (maximal size of TCP
1526 congestion window) to a more realistic 4MiB value. If you notice changes in
1527 your simulation results, you can fall back to the previous 20k tiny window
1528 by adding --cfg=network/TCP_gamma:20000 on command line.
1529 * (Hopefully) fix a bug wrt periodic availability and state traces
1530 * Bug fix: use default values at start when first event in availability/state
1531 trace is not at time 0.
1534 * remove the "new_" part of function name sg_platf_new_trace_connect
1535 (resulting in sg_platf_trace_connect), since it does not create
1539 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1540 use case where it's really mandatory, and maintaining it was a pain in
1542 * New: xbt_fifo_search(), search an item with a user-provided
1543 comparison function instead of dumb pointer comparison.
1546 * Fix the lua deployment:
1547 Use `simgrid.init_application()` before deployment instead of
1548 `simgrid.msg_register_application()` after.
1551 * Transfer the tracing files into the corresponding modules.
1553 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1555 SimGrid (3.8.1) stable; urgency=low
1557 The "we are told that some people want to also *install* the simgrid
1560 * Add missing file "tesh.1" to the archive.
1562 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1564 SimGrid (3.8) stable; urgency=low
1566 The Psssshiiiit release: SimGrid jumps into the Cloud.
1569 * Add an experimental interface to manipulate VMs. They are mainly
1570 process groups with very few intrinsic semantic, but they should
1571 allow you to build the semantic you want easily.
1572 * New function: MSG_host_set_property_value()
1573 * New function: MSG_process_on_exit(). To clean memory in all cases.
1574 * Bug fixes that made the host (and link) failures unusable.
1575 * Add a way to auto-restart process when the host in which they are
1576 executing comes back (ON_FAILURE="RESTART" on deployment file,
1577 MSG_process_auto_restart_set).
1578 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1579 please stop using the old ones, they are DEPRECATED.
1581 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1582 Please use MSG_init() instead. (reducing the amount of entry
1583 points in the library helps us).
1584 * Make it impossible to link against the wrong version of the lib
1585 * Deprecate MSG_clean(). No need to call it anymore.
1586 * Function MSG_get_host_number() is not deprecated anymore.
1589 * Split the doc into a user guide and a reference guide.
1590 * Start a developper guide to help people hacking on SimGrid.
1593 * Enable tracing by default. This modules rocks you should use it.
1594 * Remove option custom_flags. Now use environment variables CFLAGS
1596 * Use default cmake things to detect lua instead of home grown ones.
1597 * New option "enable_mallocators" to disable mallocators, for debugging
1598 purpose ("on" by default).
1601 * Bug fixes around the resource failures: don't let the processes
1602 survive the host they are running onto.
1603 * Add an interface to auto-restart processes when the host in which they are
1604 executing comes back.
1605 * Ensures that SIMIX_clean is called automatically. It's not part of
1606 the public interface anymore (bindings should be updated).
1609 * Bug fix for when SD_Simulate is called with a positive value: be careful
1610 when comparing doubles. Sometimes they are different for non significant
1612 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1613 parallel task whose initial work is distributed among host according
1614 to the Amdahl's law. Such tasks are created with a parameter alpha
1615 that corresponds to the non-parallelizable part of the computation.
1616 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1617 between two sets of workstations assuming a 1D block distribution (each
1618 workstation owns a similar share of data) on both sides.
1620 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1621 Data redistribution will be automatically scheduled once parent and child
1622 are both scheduled. The filling of computation_amount and
1623 communication_amount structures is now done seamlessly thanks to the chosen
1625 * New function SD_workstation_dump to display various information
1626 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1627 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1628 bandwidth on the route onto which the task is scheduled and the amount of
1630 To divide the nominal bandwidth by 2, the rate then has to be:
1631 rate = bandwidth/(2*amount)
1632 * Compute tasks that have failed can now be rescheduled and executed again
1633 (from their beginning)
1634 * Increasing source code coverage (src/simdag is now covered at 95.8%
1638 * Re-implement time-independent trace replay using SMPI (at the
1639 smpi_smp_* level) instead of MSG. This should replace
1640 examples/msg/actions/actions.c
1641 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1642 hindexed and structs)
1643 * Implement the exchange of non-contiguous data.
1644 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1645 * Correct behavior of smpi/sender_gap and set its default value to 0
1646 * Add option to asynchronously send small messages to allow better
1647 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1648 specifies the size in bytes under which messages will be asynchronously sent.
1649 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1650 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1653 * New C interface to define a platform: XML is now optional.
1654 For more info, please check include/simgrid/platf.h.
1655 * New interface to define random platforms from the C:
1656 For more info, please check include/simgrid/platf_generator.h and
1657 examples/msg/masterslave/masterslave_platfgen.c
1658 * Export a sg_cmdline dynar containing all the arguments we got from
1662 * Two new tracing options for adding comments to trace file so you
1663 can track your experiments (see --help-tracing for details).
1664 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1665 * Adding the SimGrid version that generated the trace file as a comment.
1666 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1667 * Fix to avoid key clashes on Paje links
1668 * Other minor fixes related to the Paje specification
1671 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1672 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1673 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1674 * Greatly improve the robustness of mmalloc to user errors (such as
1675 using an area after freeing it, or freeing it twice)
1677 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1679 SimGrid-java (3.8.1) stable; urgency=low
1681 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1682 Warning: all methods are visible, but only some of them are
1683 implemented so far. Check the source (src/jtrace.c)
1684 for further information.
1685 * New module: org.simgrid.msg.File (SimGrid File management functions)
1686 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1687 * Change the meaning of Process.restart: now restart the process from
1688 the beginning, like MSG_process_restart in C.
1689 * Add Process.setAutoRestart: handling of process restart when failed
1691 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1692 you to retrieve the properties of the processes/hosts
1693 * Deprecate Msg.clean(): you can just forget about it now.
1694 * New function Process.getCount(), that only works when compiling
1695 with the not yet released version 3.9 of the C library.
1697 * New context factory based on Coroutines. It mandates a modified JVM
1698 but then, the simulations run about five times faster, and there is
1699 no limit to the amount of processes (beside of the available memory).
1701 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1703 SimGrid (3.7.1) stable; urgency=low
1706 * Restore the prototype of MSG_process_create_with_environment() to
1707 the pre-3.7 situation by removing the kill_time argument.
1708 * Add a MSG_process_set_kill_time() function instead.
1711 * Fix weird behaviors when dealing with parallel tasks.
1714 * Simgrid is now built as a dll.
1715 * Simgrid-java now works on Windows.
1716 * Simgrid-Java is now included into Windows package.
1719 * First pre-build package for MacOSX.
1722 * Fix compilation when using MSG_USE_DEPRECATED.
1723 * Fix some compilation issues on Macs and Windows.
1724 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1725 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1727 We discovered that the Lua console is broken, but we are missing the
1728 manpower to fix it right now. The problem existed in 3.7 too, so we
1729 are not blocking the release for that. Sorry if you depended on this
1730 feature, any help would be really welcome.
1732 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1735 SimGrid-java (3.7.1) stable; urgency=low
1737 The "Java aint got to be bloated and slow" release
1740 * Various internal cleanups and performance improvement
1741 Simulations are expected to run up to twice faster or so
1742 * Make Process.kill(process) an instance method, not a static one
1743 * User processes are not java.lang.Thread subclasses.
1744 This breaks the compatibility (sorry), but previous API was
1745 brain-dead, making it impossible to have non-trivial
1746 initializations in the process constructor.
1747 * Require a full constructor per Process sub-class.
1748 Kinda breaks the compatibility (sorry), but this allows a much more
1749 efficient way to launch the processes at simulation startup.
1750 * Do not embeed our version of semaphores, java 1.5 can be considered
1751 as sufficiently prevalent for us to not dupplicate its features.
1755 * Add examples for almost every part of the API
1756 We spotted and fixed a lot of bugs in the process
1758 * New module: asynchronous communication API
1759 * New function: Process.sleep()
1760 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1761 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1762 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1764 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1766 SimGrid (3.7) stable; urgency=low
1768 The "spring cleanups (before the next Big Project kicks in)" release.
1771 * We can specify the SMPI latency/bandwidth factor with command line
1772 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1773 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1774 You can also use the "config tag" from platform file by adding this line
1775 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1776 (see "example/platforms/tag_config.xml" to use "config tag").
1777 Note that the command line supersedes the platform file configuration.
1778 * Change the correction factors used in LMM model, according to
1779 the latest experiments described in INRIA RR-7821.
1780 Accuracy should be improved this way.
1781 * Use the partial invalidation optimization by default for the
1782 network too. Should produce the exact same results, only faster.
1783 * Major cleanup in surf to merge models and split some optimization
1784 mechanisms from the core of the models. As a result you can now
1785 specify which model to use (e.g., --cfg=network/model:LV08
1786 --cfg=cpu/model:Cas01) and which optimization mode to use
1787 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1788 Incompatible combinations should err at initialization. See
1789 --help-models for the list of all models and optimization modes.
1790 * The CLM03 workstation model was dropped for simplicity because it
1791 used the deprecated CM02 network model. Use default instead.
1792 * Rename the TCP_gamma configuration option to network/TCP_gamma
1793 * Rename the coordinates configuration option to
1794 network/coordinates, and document it
1795 * Use now crosstraffic keyword instead of the terribly misleading
1796 fullduplex keyword. It is activated by default now in the current
1797 default model, use --cfg=network/crosstraffic:0 to turn it off.
1798 * Ongoing refactoring the model parsing to make XML files optional
1799 See include/simgrid/platf.h for details (still to be completed)
1802 * Major overhaul of the documentation. Almost instructive now :/
1803 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1804 functions and friends. This interface was considered as
1805 deprecated since over 2 years, it's time to inform our users that it is.
1806 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1807 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1808 use these (crappy) functions in your code.
1809 These functions will be removed soon. Stop using them now.
1810 * Deprecate MSG_get_host_{table,number}
1811 Implement MSG_hosts_as_dynar() instead.
1812 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1813 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1816 * Stabilize the parallel execution mode of user contexts
1817 * Introduce configuration variables to control parallel execution:
1818 - contexts/synchro: Synchronization mode to use when running
1819 contexts in parallel (either futex, posix or busy_wait)
1820 - contexts/parallel_threshold: Minimal number of user contexts
1821 that must be part of a scheduling round to switch to parallel
1822 execution mode (raw contexts only)
1823 * Fix bugs that prevented to use suspend/resume along with
1824 synchronization structures.
1825 * Fix bugs in process termination that lead to invalid memory access
1826 in very specific conditions.
1829 * Introduce a parallel mode for the models (controlled by surf/nthreads
1830 configuration item). In our tests, running the models in parallel
1831 never lead to any speedups because they are so fast that the gain
1832 of computing each model in parallel does not amortizes the
1833 synchronization costs, even when ultra fast futexes are used.
1834 This is released anyway because YMMV.
1837 * Performance boost by using a swag internally to compute the set of
1838 tasks that are finished and should constitute the return value of
1842 * Enable it by default now that it is considered rather stable.
1845 * Documentation of the tracing functions.
1846 * Performance gains when tracing categorized/uncategorized resource
1847 utilization by avoiding calls to get route when updating resource
1848 variables. LMM constraints are being used instead.
1849 * API changed to set task categories. Use MSG_task_set_category instead
1850 of TRACE_msg_set_task_category, and SD_task_set_category instead
1851 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1852 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1853 * Examples for link user variables added (see at examples/msg/tracing/)
1854 * Deprecated function TRACE_msg_set_process_category completely removed
1855 * Trace header updated according to the latest Paje file format
1856 * Tracing network lazy updates, no longer obligate users to use full updates
1857 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1858 * Experimental: let user code declare/set/push/pop application states for hosts
1859 * API changed to allow the manual creation of graph configuration files
1860 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1863 * Improve the API of Lua MSG bindings, using the Lua spirit.
1864 * Each simulated process now lives in its own Lua world (globals are
1865 automatically duplicated). It helps writing simulators. Will allow
1866 to run Splay programs within SimGrid in the future.
1867 * Add a Chord example in Lua, equivalent to the MSG one.
1870 * Start the implementation of a solution to express temporal
1871 properties, not only local assertions. This is still an
1872 experimental work in progress, stay clear from it to be safe.
1876 - Add new runtime parameters --help-logs and --help-log-categories
1877 to display information about supported logging parameters and
1879 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1880 don't exists anymore.
1881 * Mallocators: allow value NULL for the reset function.
1883 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1884 create homogeneous dictionaries, where all the elements share the
1885 same free function. Non homogeneous dictionaries will be
1886 deprecated in the next release.
1887 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1888 - Multi-level dictionaries are deprecated.
1890 - new function xbt_dynar_search_or_negative() that is useful when
1891 you have less than 2 million elements in your dynar and don't
1892 want of the extra complexity of catching exceptions when the
1893 element is not found.
1895 - Make xbt_os_thread module (for thread portability) public.
1896 Documentation is still to come, sorry.
1898 - Cleanups and simplifications to make it maintainable again.
1899 - Exotic features (such as memalign and valloc) were removed.
1900 - The metadata were extended and improved so that the
1901 model-checker becomes able to explore and inspect the heaps.
1902 - This may induce a performance drop when enable_model-checking is
1903 ON in cmake (even if it's not used in the simulation), but it is
1904 necessary at this point to get MC working.
1906 Turn model-checking OFF if simulation performance matters to you.
1907 Not enabling it at runtime is not enough, disable it in cmake.
1909 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1912 SimGrid (3.6.2) stable; urgency=low
1914 The "Not coding new stuff allows to polish old things" release.
1917 * New bindings to the NS3 packet level simulator (experimental)
1918 * Use the raw (efficient) execution contexts instead of the sysv
1919 (portable) ones when possible.
1920 * libpcre is now mandatory in any cases since not using it led to
1921 severe performance loss and possibly other issues
1922 * Update the XML platforms:
1923 - G5K: include the latest machine in Nancy
1924 - GridPP and LCG: new platforms
1925 * Documentation was partially updated, at least (more to come)
1927 Bug fixes, cosmetics and small improvements
1928 * Free terminated processes before the end of the simulation to avoid
1929 exhausting the memory of users having very dynamic amount of
1931 * Bug fix and cosmetics about canceling non-running tasks
1932 * Bug fix about the dot loader's issues when using libcgraph
1935 * Create an installer for windows with nsis (amd64 and win32)
1936 - Add an hello world project to illustrate simgrid project creation.
1937 - Embed libpcre into the Simgrid installer to avoid
1938 its compilation burden
1939 * The raw execution contexts should work on Apple now
1940 * Port to Windows 64 bits
1941 - Sysv contexts now have an implementation for this arch
1942 - GRAS communication features now support this arch
1943 * Drop support for borland compiler on windows
1944 - this code was not maintained, and we kinda depend on gcc nowadays
1945 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1946 * Fix portability issue on unstable ubuntu: linker became picky on
1949 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1951 SimGrid (3.6.1) stable; urgency=low
1953 The "Oops, we broke Macs too" release
1956 * Fixed contexts detection so that raw ones are used when possible
1957 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1958 leads to a strange error, with user code segfaulting sometimes when
1959 the generated code is not perfectly aligned (which is not
1960 controllable from the user side, depends on the amount of code)
1963 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1965 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1967 SimGrid-java (3.6) unstable; urgency=low
1970 * Split of every thing from simgrid v3.5 into a separate package.
1972 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1974 SimGrid (3.6) stable; urgency=medium
1976 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1979 * Bindings now constitute their own package, separated from the main one.
1980 Rationale: reduce our maintenance nightmare by reducing the module coupling
1981 They will soon be released on their own on gforge.
1983 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1984 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1986 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1987 * It's not quite deprecated for now because we have no replacement,
1988 but it may soon become the case.
1991 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1992 * Fortran: New user-level cache variable to store the rank of the running
1993 process. This improves performance by an order of magnitude.
1994 * C: New coccinelle script to automatically locate and modify global and
1995 local static variables.
1996 * Improved SMPI network model with a sender-side gap to account for multiple
2000 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2001 only say if a communication is finished, no matter whether it succeeded or
2002 failed. You can call MSG_comm_get_status() to know the status of a finished
2004 * New function MSG_task_dsend() to send a task and detach it. When a
2005 communication is detached, you are never notified of its success or failure
2006 and the memory is released automatically once it is finished. This function
2007 is useful when you don't care about the end nor the success of a
2009 * Change the prototypes of action replay. Sorry for inconvenience,
2010 but this is really more efficient this way (and to adapt your code,
2011 you just have to fix the initialization, that shouldn't be too long)
2012 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2013 ever managed to do anything useful with it.
2014 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2015 takes a dynar of communications. It returns immediately and gives the
2016 index of a finished communication (if any).
2017 * New example: a basic implementation of the Chord P2P algorithm.
2020 * New model for multi-core CPUs. You can now use the core attribute to
2021 precise the number of cores of a host. This is a basic model. Every
2022 process running on the host receives at most the power provided in
2023 the DTD (throughput<=power). Total throughput of process cannot exceed
2025 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2026 router linked by an up-link and a down-link (possibly asymmetrical).
2027 This kind of pattern allows to easily build last-mile model style platforms.
2028 Aggregating such patterns in a rule-based AS is thus the technique of
2029 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2030 * New model for Vivaldi routing. We transformed the Vivaldi network model
2031 into a Vivaldi routing model (based on the rule-based model). This allows to
2032 combine Vivaldi based latencies with last-mile platforms.
2035 * Added a check for NaN of IEEE754 infinite in the double entries of
2037 * Introduce a new context factory "raw", highly inspirated from the
2038 ucontext factory, but using manually crafted functions in assembly to
2039 do the work in an efficient manner.
2040 * Allow to change the used context factory at run time, not only at
2041 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2042 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2043 contexts at runtime (only with raw contexts or ucontexts).
2044 * Completely rewrote this module to allow parallel execution of user
2045 processes. Use --cfg=contexts/nthreads:N to execute user processes
2046 with N parallel threads (the default is 1, meaning no parallelism).
2047 * Allow to decide dynamically between sequential and parallel modes.
2048 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2049 processes in parallel only when their number is greater than or equal to P
2051 * Added a check for NaN of IEEE754 infinite in the double entries of
2055 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2056 won't output the state of processes when interrupted with Ctrl-C
2057 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2058 NULL-terminated array. This may solve backward compatibility issues
2059 due to the change to return type of SD_simulate. See also:
2060 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2061 * Add new macros with variable number of arguments.
2062 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2063 - in xbt/asserts.h: xbt_assert
2064 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2065 - in xbt/ex.h: THROWF and RETHROWF.
2066 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2068 * Change xbt_die() to accept a format string with arguments, just like printf.
2069 * New data structure: xbt_lib_t, like a dict but more general and with better
2073 * New configuration options
2074 Options triva/categorized and triva/uncategorized can be used to generate
2075 graph configuration files for Triva visualization tool.
2076 * Configuration option tracing/platform is renamed to tracing/categorized
2077 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2078 * New TRACE_user_link_variable interface:
2079 User provides the name of the link and the tracing variable to attach to it
2080 * the declaration of tracing categories must be done after the environment
2082 * simpler tracing interface, just one way to declare categories
2083 TRACE_category or TRACE_category_with_color, it is up to you
2084 * links in the trace file are again identified by their names
2085 * trace contains the full platform hierarchy exactly as declared using the ASes
2086 * Options tracing/msg/[task|process]:1 groups the process by hosts
2087 for both cases, tasks and processes must have names that are unique during the simulation
2088 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2089 * The experimental option tracing/msg/volume is deprecated
2090 its functionality may be reincorporated if needed
2092 The tracing generates a trace file with unordered timestamped events,
2093 because of the way the core simulator (surf) works. A script available
2094 at the tools directory (fix-paje-trace.sh) can be used to put the events
2095 in order. We have changed the tracing so it can generate ordered timestamped
2096 events in the final trace, but depending on the simulator (and how much time
2097 is simulated) that can lead to a huge memory utilization. It is deactivated
2098 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2100 Build Infrastructure
2101 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2102 - We are trying hard to keep the API stable, but it may happen that
2103 some things change (we're a research project after all, not a
2104 nuclear plant operating system). If such things should happen, you
2105 could rely on that macro to adapt.
2106 - current value: 30600 for 3.06.00, aka 3.6
2107 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2108 a number that can be compared with SIMGRID_VERSION.
2109 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2110 assertions and verbose/debug logging events are disabled at compile time.
2112 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2114 SimGrid (3.5) stable; urgency=medium
2117 * New feature to any SimGrid-based simulator: Model-Checking
2118 Check SIN#1 for more details.
2121 * New Model SMPI (three-interval linear regression for correction factors)
2122 See RR-7426, available at http://hal.inria.fr/inria-00527150
2123 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2124 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2125 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2126 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2127 * See also src/smpi/README
2132 - Tracing API changes: TRACE_start and TRACE_end should not be called
2133 by user-code. They are automatically called by simulators created
2134 with SimDAG, MSG and SMPI if the toolkit is compiled with
2135 tracing_enabled=ON. Categories declaration and utilization remain the
2136 same for MSG and SimDag.
2137 - A function was added to the tracing API to declare categories with
2139 - TRACE_category_with_color (char *category, char *color)
2140 where color must be in the following format
2141 "%f %f %f", red, green, blue
2142 and red, green, blue are float values in the interval [0, 1]
2143 - User can specify NULL as color parameter, or continue calling
2144 TRACE_category (cat)
2145 On that case, the tracing system will define random colors
2146 - The following command-line options are supported:
2147 --cfg=tracing/filename:msg.trace
2148 --cfg=tracing:1 (activate tracing, needed to use others)
2149 --cfg=tracing/platform:1 (categorized resource use)
2150 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2151 --cfg=tracing/msg/task:1 (task creation)
2152 --cfg=tracing/msg/process:1 (process creation, migration)
2153 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2154 --cfg=tracing/smpi:1 (SMPI interface tracing)
2155 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2156 - examples of examples/msg/tracing updated
2158 - DAXLoader and DOTLoader functions can generate tasks with categories
2159 - A new function to attribute a category to SD tasks:
2160 TRACE_sd_set_task_category (SD_task_t task, char *category)
2161 * Tracing the MPI interface implemented by SMPI
2162 - Collective operations are traced with states
2163 - Point-to-Point operations are traced with states/links
2164 - Tracing activated by a parameter "-trace filename" passed
2165 to smpirun during execution (considering that simgrid
2166 is compiled with tracing enabled)
2167 - To run the simulation with gdb, the simulator
2168 accepts --cfg=tracing/smpi:1 to trace SMPI
2169 - tesh files to check if smpi tracing is ok
2170 - See examples/smpi/NAS/DT-trace
2171 * GTNetS tracing re-worked
2172 - adaptation to the tracing system of GTNets to cope
2173 with modifications regarding the fullduplex mode
2174 - new tesh files to check if gtnets tracing is ok
2177 * Asynchronous communications through the functions:
2178 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2179 * New function: MSG_load_platform_script()
2180 to make possible using a lua script instead of XML files to set up platforms
2181 * New function: MSG_set_function
2182 to associate functions to processes, used when bypassing the parser
2183 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2185 Platforms: Add some more examples in examples/platforms
2186 * Grid'5000: see www.grid5000.fr
2187 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2190 * Change the XML format. This is a very important modification. SimGrid
2191 3.5 introduces a new hierarchical format based on the notion of
2192 Autonomous Systems. Compatibility with old format is ensured
2193 through the perl script provided in the install bin directory
2194 bin/simgrid_update_xml.
2195 It is now possible to build platforms with specific routing
2196 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2197 connect several platforms together. We will try to provide soon
2198 set of realistic platforms exploiting these properties (have a
2199 look at examples/platforms/ for the moment).
2200 * Take the opportunity of the XML format change to be a good XML citizen:
2201 rename link:ctn to link_ctn and similar changes (also dealt with by
2203 * Add a new routing scheme (rule-based) using regular expressions. It
2204 enables to have an extremely low memory footprint when the
2205 underlying routing is simple and can be compactly described. You
2206 need to have libpcre4-dev (perl regular expressions) installed if
2207 you want to use this routing scheme.
2208 * Revive the cluster TAG and allow to easily and efficiently (both in
2209 term of memory and speed) connect clusters together. Have a look
2210 at teshsuite/simdag/platforms/ to see how this can be done. With
2211 this tag, you can create clusters with thousands of tasks at no
2212 cost (have a look at examples/platforms/).
2213 Note: clusters are implemented as ASes, so there is no need for an
2214 enclosing AS tag if you have only one cluster in your platform.
2215 * Add new generic functions in the public interface that allows the user
2216 to call SURF 'create_resource' methods from your code (same
2217 functionality as the XML bypass mechanism but with a much lighter
2219 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2220 that uses a piecewise linear approximation to produce better
2221 results when exchanging small messages.
2222 * Add a new parameter to handle correctly full duplex link and account
2223 for interferences between uplink and downlink communications
2224 (activate with --cfg=fullduplex:1).
2227 * Rename the SD_READY (all dependencies are satisfied and task is
2228 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2229 dependencies are satisfied) state.
2230 This prevents a confusion between the notion of "ready to schedule"
2231 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2232 simulated" (SD_RUNNABLE) used by the simulation kernel.
2233 * Change the way a task is considered as ready. Instead of removing
2234 dependencies when a task is done, a counter is decreased. This way,
2235 it is always possible to reach ancestors thanks to the
2236 SD_taks_get_parents function (even after the end of the simulation.)
2237 * Change the return type of SD_Simulate from (SD_task_t*) into
2238 xbt_dynar_t. This function was in handling a dynar internally and
2239 converted it into a NULL terminated array for historical reasons.
2240 * New function SD_dotload(char*) to load a DAG described in dot
2241 format. This loader and the corresponding examples require the
2242 installation of the graphviz library.
2243 * Fix a bug in the management of tasks of size 0 in the surf network
2244 models. This problem was only visible with SIMDAG and you should
2245 thus disregard results produced with earlier versions if you
2246 relied on this feature (some tasks were blocked because of this).
2247 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2248 even though your applications were doing only point-to-point
2249 communications and sequential computations. Now you can really use any
2250 model you want (of course, if you create real parallel tasks, which are
2251 not implemented in most models beside ptaskL07, this will abort).
2252 * Add an example that schedules a DAX on an heterogeneous platform
2253 using a Min-Min strategy.
2254 * New function SD_workstation_get_current_task() that returns the kind
2255 of task currently running on a workstation in the sequential access
2257 * Raise some warnings when unexecuted tasks remains at the end of the
2258 simulation. This is usually caused by cycles in the DAG.
2261 * New function: SIMIX_process_set_function() called by MSG_set_function
2262 * Change the underlying waiting queue in semaphores so that a process
2263 can wait on several of them simultaneously (as in waitany).
2264 * Fix the way to handle tokens in semaphores so that all access patterns
2265 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2266 * kill the dirty pimple SIMIX_message_sizes_output()
2267 Please use (proper) visualization instead
2270 * New data container: setset (set of sets of elements)
2271 * New module: mmalloc (mapped malloc, allowing to have several
2272 independent segments of malloc)
2273 * New function: xbt_dict_cursor_set_data()
2274 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2275 * New function: xbt_dynar_is_empty()
2276 * New function: xbt_fifo_get_last_item()
2277 * Fix xbt_dynar_shrink(): use the right element size.
2278 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2279 uninitialized areas during expand.
2280 * Fix semaphores: previous implementation was severely broken.
2281 * Use library init/fini functions for our initialization.
2282 - you can use logs and other feature as soon as you want in your
2283 code (even before the xbt_init / MSG_init)
2284 - xbt_exit is now a no-op and produce a warning when used.
2287 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2288 benefit from the latest improvement to the simulation kernel.
2289 * Kill measurement sockets for now. If you rely on them, sorry. This
2290 release is not for you. This feature will be reintroduced in the
2291 future, but we cannot delay the release any further.
2292 * New function: gras_msgtype_get_name().
2293 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2294 * Fix (at last) the pmm example: it should not randomly fail anymore.
2296 Build chain: bug fixes and overall polishing
2297 * Cmake is now stable enough. Hence, we killed the autotools.
2298 * Port to windows ( TM :)
2299 * Fix the 'make install' target.
2300 No need to use 'make install-simgrid' anymore
2301 * Introduce a 'make dist' target compiling a *source* archive
2302 'make package' compiles a binary archive
2303 * Compile java files only on need
2304 * Add --cd and --setenv command line options to tesh
2305 * Out of source builds are not fully supported yet, but we are close
2306 * Enable supernovae and optimization flags by default for our users
2309 * Add layer to set up environment directly from lua, without XML.
2310 * The effect of gras_stub_generator can be achieved through
2311 lua too (check examples/gras/console/ping_generator.lua)
2313 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2315 SimGrid (3.4.1) stable; urgency=low
2317 The "Polishing easter eggs is probably a good idea" release.
2318 This is a bug fixes release only.
2322 * Fix a bug preventing the tasks from begin garbage collected.
2325 * Fix a bug occurring when a host involved in a communication fails.
2326 This was not detected properly by the other peer involved in the
2327 communication. Now, it's reported as a network error.
2330 * Warn the user about loop dependencies in data flow of DAX files
2331 * Obey the control-flow dependencies of DAX files
2334 * Add option "enable_smpi" allowing to not compile SMPI.
2335 Probably useful for the (Mac) users experiencing a build error here
2336 * Improve the detection of lua5.1 and ruby1.8
2338 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2340 SimGrid (3.4) stable; urgency=low
2342 The "Easter in Cargese" release. Also known as (major changes):
2344 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2345 ~> bindings were greatly improved
2346 ~> new tracing infrastructure for better visualization introduced
2348 * the "Welcome to configury modernity" release.
2349 ~> we switched from autotools to cmake, and improved our cdash
2352 A more detailled list of changes follow (full detail in svn log).
2354 Java Bindings: Various Cleanups
2355 * (install java-gcj-compat-dev on debian-like to use them)
2356 * Remove put/get: no need to export deprecated interface in Java
2357 Use send/receive instead.
2358 * Cleanup the examples and add a README per directory
2359 * Remove example autoDestination (that's the only way to go now)
2360 * Remove example explicitDestination (was a plain copy of basic)
2361 * Make JniException a runtime exception, so that there is no need to
2362 declare the fact that you may encounter such a beast. I guess that
2363 nobody will ever want to survive such error.
2364 * Create specific errors for each MSG case of failure:
2365 host failure, transfer failure, timeout, task cancelled
2366 * Cleanup the exceptions that may get thrown by each function
2367 * Other internal cleanups in Java bindings. Performance still bad :/
2368 Ruby and Lua Bindings: create them
2369 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2370 * That's new and great, you should try them out.
2371 Same functionalities than Java bindings, only even less polished
2373 * Kill the useless "rate" argument of SD_task_get_execution_time()
2374 Everyone used to provide -1 as a value, it was not used, and the
2375 semantic of a possible use wasn't even clear.
2376 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2377 either as comm costs or compute costs to mean that there is no
2378 such thing for that specific task.
2379 * Add a SD_task_set_name() function
2380 * Fix SD_task_unschedule() on typed tasks
2381 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2382 * In DAX loader, accept useless 'level' attributes to <job> since
2383 LIGO DAGs have them (seem to be to ease graphical representation).
2385 * Add an example masterslave_mailbox.c using send/receive and not
2386 the deprecated put/get interface.
2387 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2388 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2390 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2391 (declare MSG_USE_DEPRECATED to still have the old name)
2392 * Add a MSG_task_set_data() function
2393 * About trace replay (see examples/msg/actions):
2395 - Allow to work with splitted trace files for each process
2396 Give the specific trace file as argument of each process,
2397 and call MSG_action_trace_run(NULL)
2398 You can still have one merged file for all processes.
2399 - Fix implementation of collective operations
2400 * Allow task_execute() on 0-sized tasks (closes #10063)
2402 * This is the first release of SimGrid where SMPI is not considered
2403 beta anymore (even if some corners should still be improved)
2404 * Port over the new SIMIX_network submodule (internal refactoring)
2405 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2406 * Implement more missing elements of the standard:
2408 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2410 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2411 MPI_Scan() MPI_Get_processor_name()
2412 - Added implementation of missing case for Alltoall (warning: it's
2413 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2414 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2415 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2416 SMPI_MPI_Allgatherv()
2417 * Bug fixes include:
2418 - MPI_Waitsome() was broken
2419 - Allow relative includes in smpicc
2420 - Command line cfg argument 'reference_speed' was ignored...
2421 - Some functions did not properly lead to auto-benching of user code
2422 - smpicc passes -O2 by default (just like openmpi one)
2424 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2425 * Bug fixes about timeouts during communications
2426 * add SIMIX_message_sizes_output() as a pimple to write to file the
2427 amount of messages per size. Use gnuplot to get histogram.
2428 Pimple because that's the only user-visible function of simix,
2429 defined directly in xbt.h (irk, sorry)
2431 - Add a SIMIX_sem_get_capacity() function
2432 - Fix interactions with processe resume/suspende
2433 - release_forever() was stupidly broken
2434 - Fix SIMIX_display_process_status() for processes in a semaphore
2435 - Make SIMIX_sem_block_onto() user-visible
2436 * Refactoring context stuff:
2437 - Use pseudo-OOP for better modularity
2438 - reimplement SIMIX_process_kill() without process_schedule() so
2439 that the latter can take as invariant that it is called from
2441 - Merge context_start into context_new for sake of simplicity
2443 * Add a Vivaldi network model, coded live during SUD'10 ;)
2444 * Rename configuration variables to start a hierarchy:
2445 o cpu_model -> cpu/model
2446 o network_model -> network/model
2447 o workstation_model -> workstation/model
2448 * New configuration variables:
2449 o network/bandwidth_factor: correction to bandwidth
2450 o network/latency_factor: correction to latency
2451 o netwotk/weight_S: correction to the weight of competing streams
2452 * Add a long description to the models, that users can see with such
2453 argument on the command line: --cfg=cpu/model:help
2454 * --help-models display the long description of all known models
2456 * config: add the ability to set a default value after registration
2457 Does not override any previously set value (e.g. from cmd line)
2458 * dict: allow to have integer key and data.
2459 When so, you need to use the following functions
2460 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2461 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2462 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2463 In contrary to regular dicts, the key is not malloced before copy.
2464 Mixing scalar and regular elements in the same dict is not tested
2466 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2467 Tracing for Visualization:
2468 * SimGrid is now instrumented in order to generate a trace file for
2469 visualization analysis: to use it, need to compile SimGrid with the
2470 "tracing" option enabled, and instrument the program using SimGrid with
2471 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2472 (among other functions).
2473 * The instrumentation only traces the platform utilization for now
2474 * Documentation to use the tracing functions and how to analyze the
2475 traces with the Triva tool is written.
2476 * More information about: SimGrid FAQ (in the section Tracing Simulations
2479 * We moved to cmake as default build system. Autotools support will
2480 be dropped soon. Check the FAQ for more info about how to use it.
2481 * Greatly improved our cdash/ctest interactions
2482 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2483 * Added memory checking tests with valgrind; lot of memleak fixing.
2484 This may be the first release of simgrid with so few memory issues
2485 * Added code coverage tests.
2486 Our coverage is still improvable, but at least we see it on cdash.
2488 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2490 SimGrid (3.3.4) stable; urgency=low
2492 The "Desktop Grid needs love too" release (also called Xmas release).
2494 Models improvements:
2495 * Major speedup in the maxmin system solving by using lazy evaluation
2496 Instead of solving completely the maxmin system at each iteration,
2497 only invalidate (and recompute) the modified parts.
2498 This new feature is enabled in default models but you can try to
2499 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2500 * Cas01 IMproved as default CPU model
2501 This CPU model is the same Cas01 model, but it uses the
2502 maxmin-selective-update flag and a heap structure to manage
2503 actions on SURF kernel.
2504 It reduces the complexity to find the next action to finish and,
2505 consequently, it's faster than the old Cas01.
2506 This is the new default CPU model (Cas01).
2507 * Rename the old Cas01 model to Cas01_fullupdate
2508 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2509 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2510 * CpuTI (CPU Trace Integration)
2511 A new CPU model whose objective is simulate faster when using
2512 availability trace files.
2513 Instead of using a full featured, over engineered maxmin system for
2514 CPU modeling, this model does the pre-integration of traces files
2515 to calculate the amount of CPU power available, and so, executes
2516 faster than the old CPU models.
2517 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2518 * Use LV08 as default network model since it gives better accuracy
2519 for small messages and shouldn't change things for big ones.
2520 Use --cfg=network_model:CM02 to get the previous behavior.
2523 ******************************************
2524 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2525 ******************************************
2526 * The new CPU model may changes simulations!
2527 The point is that events occurring at the exact same timestamp
2528 are not scheduled in the same order with the old and new
2529 version. This may be enough to completely change the execution
2530 of simulations in some cases.
2531 * The new network model will change simulations!
2532 This new model is more realistic than the previous one, so you
2533 should consider redoing your old experiments with this model.
2534 Sorry for the inconvenience.
2537 * Introduce the supernovae compilation mode
2538 When compiled that way, the whole SimGrid (or almost) is put in a
2539 single compilation unit and compiled in one shoot.
2540 This is to help gcc which has difficulties to inline stuff from one
2542 The speedup seem to be above 15%, althrough more tests are needed on
2543 amd64 to confirm that gain.
2546 * Port of MSG's mailbox on top of SIMIX network
2547 The put/get mechanism was greatly simplified on the way.
2550 * New SIMIX network module. Provides:
2551 - Mailbox: rendez-vous mechanism to find with who you want to speak
2552 - Synchronous send/recv: easier and hopefully faster since the
2553 logic is handled in the maestro process directly now
2554 - Asynchronous send/recv: you dreamt of it? It's here now
2555 Too bad that nobody cared enough to propagate the change to MSG.
2556 * Add semaphores as SIMIX synchronization mechanism.
2559 * new function SD_daxload(char*) to load a DAX file
2560 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2561 * Introduce typed tasks. Specify its kind and cost at creation.
2562 At scheduling, just give where it should be placed, and the cost
2563 for each involved resource is automatically computed.
2564 Existing constructors so far (more to come of course):
2565 - SD_task_create_comm_e2e() for end-to-end communication
2566 - SD_task_create_comp_seq() for sequential computation
2567 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2568 * new function SD_task_dump() for debugging display
2569 * new function SD_task_dotty(task,FILE*) writing to file the info
2570 about the task in dotty format
2571 * SD_task_dependency_exists() can now cope with having one of its
2572 arguments NULL. If so, it tests whether the other argument has any
2574 * Add getters on list of preceding/following tasks:
2575 SD_task_get_parents(task) and SD_task_get_children(task)
2576 * Add getters on amount of workstations and list:
2577 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2578 * Add getter on task kind: SD_task_get_kind(task)
2579 * Update the start_time and finish_time of tasks on completion/failure
2580 * Bugfix: Remove task from state swags when destroyed
2583 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2586 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2587 dynar with one separate thread per value of the dynar.
2588 * Change the prototype of xbt_thread_create(), sorry.
2589 Added a boolean parameter indicating whether we want to join this
2590 thread (used in SG only for now)
2591 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2594 * GTNetS wrappers should now be usable again (and betterly tested too)
2595 * Fix a major regression from 3.2 where the timeout provided to
2596 MSG_task_put_with_timeout() was used as absolute time before which
2597 the comm should be done.
2598 * Start to fix the <cluster> tag.
2599 - Internal links should be good now (beside of the loopback, which
2600 use the private link instead)
2601 - paths to the external world is still rather broken
2602 - the <route:multi> tag is just broken. Actually that's brain-dead.
2603 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2604 to make it less stupid
2605 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2606 * Fix a source-level compatibility glitch from 3.2: after defining
2607 MSG_USE_DEPRECATED, you can use the old name
2608 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2609 * Allow to compile from the SVN with automake 1.11
2610 * Fix some problems when using the "start_time" tag in deployment XMLs.
2611 * Fix #8569: XBT/synchro.h has redundant declarations
2612 * Fix #8563: MSG return values and exceptions
2613 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2614 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2615 Thanks to Robson Peixoto.
2616 * Fix a vicious bug in dictionaries inducing that some elements were
2617 not freed on xbt_dict_free()
2619 Portability report of this version:
2620 * Main portability targets:
2621 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2622 on (amd64/i386/ia64)
2623 - mac leopard on i386
2624 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2625 but nothing critical.
2626 * Other platforms: windows, AIX and others were not tested for this release
2628 Timing report of this version:
2629 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2630 scenario parameters). From 8h to a few seconds in desktop grid settings.
2631 * Supernovae brings about 25% speedup on i386.
2633 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2635 SimGrid (3.3.3) stable; urgency=low
2637 The "Need for Speed" release.
2639 The timings done to validate the 3.3.2 were faulty.
2640 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2642 The problem was a conversion from a manually handled vector to
2643 xbt_dynar_t on the critical path.
2644 xbt_dynar_foreach calls functions, inducing stack management crap.
2646 We inlined these functions and xbt_dynar_foreach is now breath taking.
2647 We also inlined xbt_swag_belong on the way.
2649 Here are some approximate speedup measurements (on master/slaves
2650 simulations lasting between 10s and 20s each):
2651 3.3.1 -> 3.3.2: about same performance
2652 3.3.2 -> 3.3.3: 40% speedup
2653 3.3.1 -> 3.3.3: 40% speedup
2654 3.3.1 with inline patch -> 3.3.3: 30% speedup
2656 Our reading is that the refactoring which occurred in 3.3.2 made us
2657 suffer much more from the xbt_dynar_foreach low performance, but
2658 once we solved this, this refactoring proved to be very performance
2659 effective. From the 40% speedup, somehow, 10% are due to the
2660 inlining and 30% to the refactoring.
2662 That's a pitty that gcc cannot inline functions placed in other files
2663 alone. We have to choose between:
2664 - break the encapsulation (by putting private data structures and
2665 accessors in headers files to help gcc)
2666 - live with low performance
2667 - switch to a decent compiler such as icc (not quite possible).
2669 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2671 SimGrid (3.3.2) stable; urgency=low
2673 The "Simplicity does not preceed complexity, but follows it" release.
2675 The main contributors of this release were (lexical order):
2676 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2679 * Extract the routing logic into its own object.
2680 (was dupplicated in network.c and workstation_LV07.c;
2681 Allows to implement other ways of storing that info)
2682 => kill now useless network_card concept
2683 - Use dynar to represent routes (instead of void** + int*)
2684 - kill link_set (use surf_network_model->resource_set instead)
2685 - Add a command-line option to choose the routing schema to use
2686 - Add three new models:
2687 * Floyd (shortest path computed at initialization)
2688 * Dijikstra (shortest path recomputed all the time)
2689 * Cached Dijikstra (shortest path computed on need)
2690 All these models where contributed by Silas De Munck, and are
2691 described in his ICCS09 paper.
2693 * Simplify model declaration
2694 (less redirections, less function to write when defining a model)
2695 - Factorize stuff between models:
2698 surf_model_resource_set(model)
2699 surf_model_resource_by_name(model, name)
2700 - Unify the types of models in s_surf_model_t (using an union)
2701 - Embeed fields of common_public directly into s_surf_model_t
2702 - Rename model methods:
2703 action_free ~> action_unref
2704 action_change_state ~> action_state_set
2705 action_get_state ~> action_state_get
2706 - Change model methods into functions:
2707 (model)->common_public->action_use ~> surf_action_ref
2709 * Implement a generic resource; use it as ancestor to specific ones
2710 (allows to kill duplicated code in models)
2711 Drawback: timer command don't need no name nor properties;
2712 workstation_CLM03 don't need no properties
2713 (but I guess we can live with those few bytes wasted)
2715 * Improve the action object model
2716 - implement a constructor avoiding dupplicated code about field
2717 initialization in generic_action part.
2719 * Kill the SDP model: it has an external dependency, is deprecated
2720 in flavor of modern lmm models, and didn't compile since a while
2723 * Relocation of the context module from XBT to SIMIX.
2724 (the context were decoupled from the simix processes, duplicating a lot of code)
2725 => a lot of code was factorized
2726 - less overhead is introduced during scheduling
2727 - simpler API for the context factory
2728 - the logic for process creation,destruction and manipulation was simplified
2729 * Simplification of the s_smx_process_t data structure.
2730 => accesing the simix level data associated to a process is faster now,
2731 and the code is a lot more readable.
2734 * Implement some more MPI primitives:
2735 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2736 -implementation: Bcast: flat or 2-ary tree (default),
2737 Barrier: 4-ary tree,
2739 Allreduce: Reduce then Bcast
2740 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2741 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2742 Alltoallv: flat tree, like ompi
2744 * Add support for optimized collectives (Bcast is now binomial by default)
2745 * Port smpirun and smpicc to OS X
2748 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2749 Shout out if you used it.
2752 * Display the list of still queued messages in SG mode when existing
2756 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2757 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2758 * Add xbt_str_from_file(FILE*)
2759 * Add xbt_dict_get_key achieving a linear reverse search
2760 * Remove the context module
2762 Portability report of this version:
2763 * Main portability targets:
2764 - Linux(debian)/x86/context
2765 - Linux(debian)/x86/pthreads
2766 - Linux(debian)/amd64/context
2767 - Linux(debian)/amd64/pthreads
2768 On these, we still have the eratic breakages of gras/pmm and
2769 amok/saturate_sg reported in previous version. We still think
2770 that the tests are the cause of the fault, not the tested code.
2772 - Mac OSX Leopard/x86/context
2773 Still false negative in tesh autotesting.
2774 Smpi still fails, but this time because readlink does not accept -f
2775 Everything seems to work properly beside of that.
2778 - AIX version 5.3 (only tested contexts this time)
2779 Smpi still fails there because mktemp is not installed.
2780 Everything seems to work properly beside of that.
2782 I managed to compile it for the first time, but several breakages.
2783 Won't delay the release for this exotic platform.
2785 * Windows: it's still lagging behind. If you want to help, please
2788 Timing report of this version:
2789 This version seem to be more than 5% faster than 3.3.1 (on linux
2790 64bits with contextes). The gain is less than expected, we are
2791 investigating this for next release.
2793 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2795 SimGrid (3.3.1) stable; urgency=low
2798 * Implement a --cfg-help to show existing configuration variables
2799 * Build chain do not require doxygen in maintainer mode
2802 * fix a bug on struct sizeof computation, which prevented the
2803 exchange of arrays of structs in some conditions
2804 - added a regression test about this in datadesc_usage
2805 * Allow the exchange of 0-long dynamic vectors.
2806 - for that, use -1 as indicator of dynamic size instead of 0
2807 - This implied to change any size from unsigned long to long,
2808 reducing a bit communication abilities, but I guess that with
2809 64bits being quite common, this is more than enough.
2810 - This also induce a protocol change, thus bumping network protocol
2811 version from 0 to 1 (if we have external users, we have to get
2812 clean on that point too ;)
2813 - added two regression tests about this in datadesc_usage
2814 * Be more verbose when propagating local exceptions
2815 This helps debugging.
2816 * Display the status of simulated processes when receiving SIGINT in
2820 * Allow to control the simulation from a trace file.
2821 New functions MSG_action_register() and MSG_action_trace_run()
2822 The first one allows to associate a function execution to each
2823 kind of action while the second one parses a trace file and
2824 triggers the corresponding actions within the system.
2825 For now, only a toy example is provided in examples/msg/actions
2826 * Add an example of process migration in examples/msg/migration
2827 * Fix a bug in task exchange which broke MSG_task_get_sender()
2828 Add a teshsuite regression test for that.
2829 [Bug: if MSG_task_get_sender() is called after sender exit,
2831 * Fix a bug which prevented suspend/resume to work properly
2832 * Display the status of simulated processes when receiving SIGINT
2833 This fixes a regression of v3.3. due to the introduction of SIMIX
2834 * Bug fixing in failure management:
2835 - trace could not start by a failure at time 0
2836 - failure during communications were not working
2839 * Add SIMIX_process_set_name() to change the name of the current
2840 process in the log messages.
2841 * Store smx_hosts in a dict since we only retrieve them by name
2842 * Move the configuration infrastructure to surf
2845 * Move the configuration infrastructure to surf
2848 * Massive internal cleanups:
2849 - Store internal structures on processes instead of hosts (allows
2850 to have more than one process per host, in addition of being more
2852 - Cleanup the initialization/finalization process
2853 - Kill a whole bunch of unneeded synchronization:
2854 processes run in exclusive manner within the simulator
2855 - Move queues from global tables to process data fields
2857 - now accept -platform and -hostfile arguments
2858 - Pass the right rank value to processes according to the hostfile
2859 * Compile the examples by default, and use them as regression tests
2860 * Implement MPI_Wtime()
2861 * Change the reference speed to a command line option
2864 * TCP_gamma can now be specified as command line option using
2865 --cfg=TCP_gamma:10000000.0
2866 * Change the --surf-path cmd line option into --cfg=path:
2869 * Also include strbuff from xbt.h public header
2870 * xbt_ex_display(): do not free the exception after displaying
2871 This allows to do more with the given exception afterward.
2872 Users should call xbt_ex_free() themselves.
2876 Portability report of this version:
2877 * Main portability targets:
2878 - Linux(debian)/x86/context
2879 - Linux(debian)/x86/pthreads
2880 - Linux(debian)/amd64/context
2881 - Linux(debian)/amd64/pthreads
2882 These targets fail about 1/10 of times on gras/pmm, but we believe
2883 that this is because of the test, not because of simgrid.
2884 amok/saturate_sg fails even more rarely, and the test may not be
2887 - Mac OSX Leopard/x86/context
2888 The test suite still spits tons of errors because some obscure
2889 force prevents us from removing the temporary directories
2890 arguing that they still contain some metadata I've never heard of.
2891 Smpi fails because seq is not installed.
2892 Everything seems to work properly beside of that.
2895 - AIX version 5.3 (both contexts and pthread)
2896 Smpi still fails there because mktemp is not installed.
2897 XML inclusions seems rosty on AIX.
2899 * Windows: it's still lagging behind. If you want to help, please
2902 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2904 SimGrid (3.3) stable; urgency=high
2908 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2909 [Malek Cherier & Mt]
2911 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2914 Until now, GRAS were using MSG as an interface to SURF. It was
2915 quite difficult because both interface have several differences
2916 (MSG channels vs GRAS sockets were the most notable point).
2918 This also opens the gate to SMPI (which should occur soon) and speed
2919 up simulations by to 40% (even if it were not the main goal).
2921 **************************************
2922 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2923 **************************************
2924 The point is that events occurring at the exact same timestamp are
2925 not scheduled in the same order with the old and new version. This
2926 may be enough to completely change the execution of simulations in
2927 some cases. Sorry for the inconvenience.
2929 * Cleanup and upgrade the XML format to push further scalability
2930 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2932 * Improve the testing infrastructure with tesh. Now a very large part of
2933 the code is tested not only by being run but also by checking that the
2934 output match an expected output [Mt].
2936 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2937 is really less memory-demanding, which should allow you to use
2938 larger files in SimGrid [AL].
2940 * Inform valgrind about our contextes, so that it becomes usable
2941 with the default (and more effecient) version of SimGrid
2942 [contributed by Sékou Diakite, many thanks]
2945 * Introduce a listener thread in charge of receiving incoming
2946 messages from the network. It allows to overlap communication and
2947 computation but most notably, it removes some stupid deadlocks due
2948 to the fact that so far, a process could not send and receive at
2949 the same time. This made most non trivial communication schema
2951 * Convert the PIDs from long int to int to match the MSG ones (and
2952 linux ones too) [Mt]
2953 * New function: gras_agent_spawn() to launch a new process on
2954 current host. Only working in simulation for now. [Mt]
2955 * New function: gras_os_hostport() returning a constant form (ie,
2956 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2959 * Make the backtrace of exceptions more human readable [Mt]
2960 * New module: xbt/str [Mt]
2961 a ton of string utility functions (split, join, printf to a newly
2962 allocated buffer, trim, etc)
2963 * New module: xbt/hash [Mt]
2964 SHA1 hashing algorithm (more to come if needed)
2965 * New module: xbt/synchro [Mt]
2966 synchronization tools (mutex and conditions) working the same way
2967 in simulation and in real life (mainly useful for GRAS, but not
2969 * New module: xbt/queue [Mt]
2970 classical producer/consumer synchronization scheme
2971 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2972 (using the classical functions will get serialized) [Mt]
2973 * Make dictionary internal table dynamic. No need to specify its size
2974 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2976 * Make sure the log channels are organized as a tree under windows
2977 (because of ANSI C compatibility issue, any channel were child of
2978 root directly) [Mt].
2981 * Cleaned many thing in surf and fixed a few bugs [AL].
2982 * Add a nice command line configuration mechanism to compose models [AL].
2983 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2984 the previous one (KCCFLN05). It relies on something that looks like
2985 a max-min sharing mechanism but cannot be written as such. A new solver
2986 was thus designed [AL].
2987 * Add a new solver to lmm. Based on Lagrange optimization and
2988 gradient-based descent, it enables to efficiently maximise systems s.a
2990 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2992 This solver enables to propose two new network models for TCP Reno and
2993 TCP Vegas based on Low's work. These models still need to be fully
2994 tested though [Pedro Velho].
2997 * Bug fix in SD_simulate. Now the time bound given as argument is
2999 * Use the new parallel task model (ptask_L07) as default.
3000 * Use the SURF command line configuration mechanism.
3001 * 0-size tasks (for synchronization) should now work.
3003 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3005 SimGrid (3.2) stable; urgency=high
3009 We still experience issues on this platform, but we believe that at
3010 least MSG is usable.
3012 GRAS API BREAKAGE (for simplification purpose, sorry):
3013 * the gras_msgtype_by_name is not used anymore. Instead of
3014 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3015 you can write (and must)
3016 gras_msg_send(toserver, "request", &request);
3017 - If you still want to pass a gras_msgtype_t to the function (to cache
3018 the type and avoid the lookup time), use the gras_msg_send_() variant.
3019 - Impacted functions:
3020 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3021 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3022 * The callbacks are now expected to return 0 when everything went well
3023 (just like the main() function)
3025 GRAS new features and improvements:
3026 * New module mechanism where user code can use per process globals [Mt]
3027 This is similar to gras_userdata_*() functions, but for libraries. It
3028 factorize some code developped over and over in the examples and AMOK.
3029 It has still to be documented and used (only amok/peermanagement is
3031 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3032 loss when they were small enough to fit into the buffer and sent quickly
3033 enough so that they can all get received in one shoot.
3034 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3035 if not found. Use the *_or_null() variant for the old semantic.
3036 * In gras_msg_handle, do not discard messages without callback.
3037 They are probably messages to be explicitly awaited later (ie, proofs of
3038 mis-synchronization in userland since they are sent before being awaited)
3039 No big deal usually.
3040 * gras_socket_meas_send/recv: semantic changed!
3041 The numerical arguments used to be (1) the total amount of data to send
3042 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3043 messages. This was need for the fool willing to send more than MAXINT
3044 bytes on quite fat pipes.
3047 * Do really rename the hostmanagement module to peermanagement. [Mt]
3048 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3049 but this is rather new and this was documented in the module
3050 documentation (poor excuses, I admit)
3051 * Bandwidth measurement semantic changed! This follows the changes to
3052 gras_socket_meas_send/recv explained above.
3055 * A sequential mode has been added to the workstations. When a workstation
3056 is in sequential mode, it can execute only one task, and the other tasks
3057 are waiting in a FIFO. [Christophe Thiery]
3060 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3061 model for SIMDAG. [Christophe Thiery]
3062 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3063 equal to 0, which caused some bad behaviors in
3064 saturated_constraint_set_update. I now use a threshold mechanism like in
3068 * When running manually src/testall, you select specific units [Mt]
3069 testall is the result of our cunit mechanism, and should replace all
3070 the scripty thingy around since bash don't run easily on billware.
3072 * A mallocator system has been added. [Christophe Thiery]
3073 Mallocators allow you to recycle your unused objects instead of freeing them
3074 and allocating new ones.
3076 Documentation update:
3077 * FAQ reworking + New FAQs:
3078 - "Valgrind spits tons of errors!" [Mt]
3079 - "How to repport bugs" [Mt]
3080 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3081 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3082 - Communication time measurement within MSG [AL]
3083 - I experience weird communication times when I change the latency [AL]
3084 * GRAS tutorial [Mt]
3086 - an introduction to the framework and to the used communication model
3087 - an initiatic tour introducing the most proheminent features:
3089 . Lesson 0: Installing GRAS
3090 . Lesson 1: Setting up your own project
3091 o Part 2: Message passing
3092 . Lesson 2: Exchanging simple messages
3093 . Lesson 3: Passing arguments to the processes (in SG)
3094 . Lesson 4: Attaching callbacks to messages
3095 . Lesson 5: Using globals in processes
3096 . Lesson 6: Logging information properly
3097 . Lesson 7: Using internal timers
3098 . Lesson 8: Handling errors through exceptions
3099 . Lesson 9: Exchanging simple data
3100 . Lesson 10: Remote Procedure Calling (RPC)
3101 . Lesson 11: Explicitly waiting for messages
3102 . Recapping of message passing features in GRAS
3103 - A HOWTO section containing:
3104 o HOWTO design a GRAS application
3105 More are due, of course. They will come latter. In the meanwhile, you can
3106 check the examples which are still here.
3108 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3110 SimGrid (3.1) stable; urgency=high
3114 There was a stack corruption somewhere, visible only when optimizing
3115 with these versions. [Vince]
3118 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3119 module that enabled to play with Directed Acyclic Graphs. It is built
3120 directly on top of SURF and provides an API rather close to the old
3121 SG. Some old codes using SG are currently under rewrite to check that
3122 all needful functions are provided. [Christophe Thiery]
3125 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3126 extension of the classical CLM03 model that gracefully handles
3127 failures. This is now the default model for MSG and GRAS. It doesn't
3128 handle parallel tasks yet though. [AL]
3129 * Bug fix: Weights were not correctly set in the network part.
3130 WARNING: This may have resulted in incorrect results with simulations
3131 where there are more than one flow on a given link. [AL]
3134 * After a (long ?) discussion on simgrid-devel, we have decided that the
3135 convention we had on units was stupid. That is why it has been decided
3136 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3137 WARNING: This means that all previous platform files will not work as
3138 such with this version! A warning is issued to ask users to update
3140 A conversion script can be found in the contrib module of the CVS, under
3141 the name contrib/platform_generation/surfxml_update.pl [MQ]
3144 * Bug fix: Processes were started in reverse order, wrt deployment file.
3145 WARNING: if your code relies on this bug, please fix it. [AL]
3146 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3147 being executed on two different locations. [AL]
3148 * Bug fix: Failures are now better supported thanks to Derrick's tests
3149 (there was many failure situations I hadn't thought of and that weren't
3150 correctly handled). [AL]
3151 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3155 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3156 Exception occurring on server-side are propagated back to client (!).
3158 API CHANGE: the callback changed their prototype. Change:
3159 int my_handler(gras_socket_t expeditor, void *payload_data) {
3161 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3162 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3164 * New! function: gras_msg_handleall to deal with all messages arriving
3165 within a given period.
3166 * New! function: gras_socket_server_range to get a server socket in a
3167 range of port numbers (ease to avoid port number conflicts) [MQ]
3168 * New! gras processes display their backtrace when they get a SIGUSR1
3169 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3170 Sweet to debug RL processes [MQ]
3174 - Do not force experiment sizes to be expressed in kb, or it becomes
3175 impossible to measure the latency this way (needs one byte-long tests)
3176 WARNING: this changes the amok_bw_* function semantic. [MQ]
3177 - Implements the link saturation stuff. [MQ]
3178 * Peer management module:
3179 New! module factorizing code that we wrote over and over [MQ].
3182 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3183 - Test units are placed directly into the library code, they get extracted
3184 automatically and placed into the src/testall binary.
3185 - Convert most of the XBT tests to this system.
3186 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3187 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3188 live in the same C code. This required to change a little bit the API
3189 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3190 * New module: sparse graph structure with basic algorithms (this is work
3191 in progress and the API is not considered to be frozen yet). [AL]
3192 * Display more information on backtraces: source line & function names are
3193 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3194 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3195 * Reworked a little bit some #include statements to load only required
3196 headers. Some user code that relied on SimGrid to include stdlib or
3197 stdio may need to include it by themselves. [AL]
3198 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3200 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3201 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3202 but I guess that almost nobody used those parts. [MQ]
3204 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3206 SimGrid (3.0.1) stable; urgency=low
3209 * Unfortunately, I had missed 5 misnamed functions:
3210 xbt_fifo_item_t xbt_fifo_newitem(void);
3211 void xbt_fifo_freeitem(xbt_fifo_item_t);
3212 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3213 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3214 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3215 They're now deprecated. Please use their new versions:
3216 xbt_fifo_item_t xbt_fifo_new_item(void);
3217 void xbt_fifo_free_item(xbt_fifo_item_t);
3218 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3219 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3220 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3222 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3223 * Documentation: xbt_log module unmercifully reworked [MQ]
3224 * Bugfix: there was a problem with the ending of contexts with
3225 the pthread backend. It caused some weird deadlock or behavior
3226 depending on the pthread implementation. [AL]
3227 * Bugfix: get the exceptions raised in the simulator repport where
3228 and why they come from when they are not caught in time [AL, MQ]
3231 * Bugfix: Do repport the error when two non-connected hosts try to
3232 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3235 * Add additionnal checkings on communications. Assert that two
3236 communicating hosts are connected by a set of links... [AL]
3239 * Add additionnal checkings on channel values in communication [AL]
3240 * New: MSG_task_get_source to see on which host a task was generated [HC]
3241 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3242 the number of tasks waiting to be received on channel and sent
3244 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3245 waits for the first task coming from a given host.. [AL]
3247 GRAS new functionnalities: [MQ]
3248 * Enhance the parsing macro to allow the size of multidimentional objects
3249 to be given thru annotations.
3250 * New example (and documentation): Matrix Multiplication a la RPC
3251 (as when I was young!) and fix a bunch of bugs found on the way.
3253 GRAS performance improvements: [MQ]
3255 * Reduce the amount of cbps creation/destruction by making it static to
3256 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3259 * Change libdata to a set so that we can search for stuff by ID (and thus
3260 reduce the insane amount of dict lookups)
3263 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3264 RL communication on top of SG since it's so uneffective.
3265 It may also be used for inter-thread communication in RL, one day.
3266 * Use gras_datadesc_copy() to exchange messages on top of SG
3268 - improve message exchange performance on top of SG
3269 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3270 * Don't exchange on the network the size of the used part of buffer,
3271 instead, specify the possible buffer size to read().
3273 - reduces the amount of read/write calls (one pair per exchange)
3274 - reduces the amount of exchanged data (the size)
3275 - allows to retrieve all arrived data on receiver side, if we don't need
3276 it right now (subsequent read will peek the buffer)
3277 - allows the receiver to proceed with the beginning of the stream before
3278 everything is arrived
3279 - make it possible to build an iov transport (using readv/writev)
3281 - take care of the data with non-stable storage (like stacked data),
3283 * If possible, TCP send uses vector I/O (when writev() is here)
3284 - Don't use it for receive since we send data sizes and data on the
3285 same stream, so we wouldn't be able to chain large amount of chunks
3286 before having to flush the stuff to read the size.
3287 * Rework the transport plugin mechanism to simplify it and reduce the
3288 amount of pointer dereferencement when searching for the right function
3291 * I guess that now, we do almost as few system calls as possible while
3292 doing as few data copy as possible.
3294 To improve it further, we could try to send all the sizes first and then
3295 all the data (to use iov on receiving size), but it's only a partial
3296 solution: when you have 2 dimensional data, the sizes of the second
3297 dimension is data of the first dimension, so you need 3 streams.
3299 I'm not sure the potential performance gains justify the coding burden.
3301 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3303 SimGrid (3.00) stable; urgency=high
3306 * New! Give the possibility to hijack the surf parser and thus bypass
3307 MSG_create_environment and MSG_launch_application. Have a look at
3308 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3310 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3312 SimGrid (2.96) unstable; urgency=low
3317 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3318 This deprecates the xbt_error_t mechanisms.
3319 It modifies (simplifies) all XBT and GRAS API.
3320 MSG API keeps unchanged (exceptions raised by XBT are caught from
3321 within MSG and masked with existing error handling facilities)
3324 * New! Add a FATPIPE model. [AL]
3325 * New! Add a parallel task model. [AL]
3326 * New! Add automatically a loopback interface (in the default
3327 network model) if none was precised.
3330 * Bugfix: MSG_process_resume now works with the current running process.
3332 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3333 * Modification of MSG_task_get_compute_duration. Once a task has been
3334 processed, the value returned by this function is now equal to 0. [AL]
3335 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3336 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3337 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3338 of a m_task. For now, MSG_task_cancel only works with computation
3340 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3341 of the processor (in Mflop/s) regardless of the current load on the
3343 * API Change: use proper naming convention for MSG_getClock and
3344 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3346 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3347 This function changes the priority of a computation task. This priority
3348 doesn't affect the transfer rate. A priority of 2 will make a task
3349 receive two times more cpu power than the other ones. This function
3350 has been added to suit the needs of Nguyen The Loc and hasn't been that
3351 much tested yet. So if it fails, please report it and send me your code.
3353 * API Change: removed all functions and types that were marked "deprecated"
3354 since many months. Renamed MSG_global_init_args to MSG_global_init.
3356 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3358 SimGrid (2.95) unstable; urgency=low
3361 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3362 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3363 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3366 * Bugfix: Allow absolute paths for platform description files [MQ]
3367 * Bugfix: do free the variables after use. Leads to drastic performance
3369 * Implement max_duration (ie, timeouts) on resources [AL]
3372 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3374 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3375 the needed support (timeouts) [AL]
3378 * Implement measurement sockets. You can now get the bandwidth between two
3379 hosts thanks to AMOK (see below). [MQ]
3380 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3381 dynar over the network (yeah) [MQ]
3382 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3383 * simple/static token-ring example. [Alexandre Colucci and MQ]
3384 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3385 in place before [MQ]
3388 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3389 type parsing. "short" and "long" modifiers were also ignored; other
3390 modifier (such as reference level) are still ignored. [MQ]
3391 * Embeed the buffer size within the buffer itself on SG. [MQ]
3392 That way, send() are atomic and cannot get intermixed anymore (at least
3393 the ones which are less than 100k; bigger messages still have the issue)
3394 * Array size pushed by the field, not by the field type (or each
3395 and every long int will push stuff to the cbps) [MQ]
3396 * use select() to sleep since it allows to portably sleep less than one
3399 GRAS (minor cleanups)
3400 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3403 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3404 * type category 'ignored' killed as it was never used and were difficult
3406 * whether a type can cycle or not is now a flag, leaving room for more
3407 flags (as "ignored", if we feel the need one day ;)
3408 * Rename raw sockets to measurement sockets since "raw" has another
3409 meaning in networking community.
3412 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3413 and offers features not belonging to GRAS but that most applications
3414 need. One day, it may be a set of plugins loadable at runtime.
3415 * New module: bandwidth
3416 bandwidth measurement between arbitrary nodes running this module. [MQ]
3418 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3420 SimGrid (2.94) unstable; urgency=low
3422 The first beta release of SimGrid 3 !
3426 * Update the main page and the FAQ. Adding references to gforge.
3429 * Add a gras_os_getpid function.
3432 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3433 * Extend the logs so that they also print PID, hostname, date, ... if
3435 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3436 and kill the old version which were in testsuite/
3437 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3441 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3442 for users. Instead, each programming environment (GRAS, MSG,...) use it
3443 when needed to provide such a feature to users.
3444 Don't shortcut the mechanism or you will also shortcut the virtualization
3445 you need on the simulator.
3449 * Cleanups in configury with regard to compile optimization/warning flags.
3450 Also add -fno-loop-optimize to any powerpc since it's the optimization
3451 killing gcc (< 3.4.0).
3452 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3453 needed by MSG examples complications
3454 * Borrow configury beautifications from PHP
3457 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3458 warning (thanks loris for stumbling into this one).
3459 * Bugfix: stop loading private headers (gras_config.h) from the public
3463 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3464 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3465 RL would imply a DNS resolver, which is *hard* to do in a portable way
3466 (and therefore delayed).
3467 * Implement a real timer mechanism and use it in timing macros. This allows
3468 to avoid rounding errors and get a 0.000005 sec precision in timing
3469 macros. While I was at it, various cleanups:
3470 - allow to declare more than one timed section per file (fix a stupid bug)
3471 - move some private declaration to the right place
3472 - merge conditional execution and timing macros into emulation module
3473 - document the module
3474 - make sure the module cleanups its mess on gras_exit
3475 * Documentation improvements:
3476 - (new) how to compile applications using GRAS
3477 - (new) emulation support (timing macros)
3479 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3481 SimGrid (2.93) unstable; urgency=low
3483 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3486 - Use Paje properly where used. Still to be sanitized properly.
3487 - Portability fix: Add an implementation of the contexts using pthread
3491 - Add xbt_procname(): returns the name of the current process.
3492 Use it to show the current process's name in all logging.
3494 - fix detection of older flex version and the reaction, since we do
3495 depend on modern ones (we use lex_destroy)
3496 - Better separation of SG and RL in the libs: remove all simulation code
3497 from libgras. As a result, this lib is now only 200k when stripped.
3498 Some of the xbt modules may also be duplicated (two sets and such) and
3499 should be cleaned/killed before SG3.
3500 - Insist on using xlC on AIX because of weird problems involving gcc there.
3501 - Cleanup the make remote stuff. This is now done by scripts
3502 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3503 mainly for our private use, but we're working on changing them to user
3506 - Bugfix: flush the socket on close only if there is some *output*.
3507 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3509 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3512 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3514 SimGrid (2.92) unstable; urgency=low
3516 Alpha 3 on the path to SimGrid 3
3520 - New! First try of benchmarking macros.
3521 - New! First try so that gras_stub_generator generate deployment and
3522 remote compilation helpers.
3524 - Bugfix: Initialization fix in msg_test.
3528 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3530 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3532 - API Change: gras_os_sleep to take the amount of seconds as a double.
3533 Accepting an int was error prone since it was the only location where
3534 seconds were coded as such. It leaded to damn rounding errors.
3535 - Bugfix: Hard to belive that timers ever worked before this.
3537 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3539 SimGrid (2.91) unstable; urgency=low
3541 Alpha 2 on the path to SimGrid 3
3545 - Bug fix in the lmm_solver.
3547 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3548 through the function MSG_paje_output.
3549 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3550 - It is possible to bound the rate of a communication in MSG with
3551 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3552 to write it in the changelog).
3553 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3558 - Various cleanups to the autotools stuff
3559 - Begin to move Gras examples to examples/gras/
3560 - Let make distcheck work again (yeah!)
3562 - documentation overhauled using doxygen.
3563 gtk-doc-tools is dead in SimGrid now.
3564 - Automatically extract all existing logging categories, and add the list
3565 to the documentation (long standing one, to say the less)
3567 - Cleanup the known architecture table. Reorder the entries to group what
3568 should be, and use a more consistent naming scheme.
3569 (some of the test dataset are still to be regenerated)
3570 - New! Allow library to register globals on each process just as userdata
3572 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3573 lookup time (for now).
3574 Use it in msg and trp.
3575 This cleans a lot the internals and helps enforcing privacy of the
3576 headers between the gras components.
3577 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3578 - Bugfix: gras_os_time was delirious in RL.
3579 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3581 - Reenable GRAS now that it works.
3583 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3585 SimGrid (2.90) unstable; urgency=low
3587 Alpha 1 on the path to SimGrid 3
3589 * It is a long time since the last release of SimGrid. I'm sorry about
3590 that but as I had told you, I was rewriting a lot of things. I apologize
3591 to those who had been reporting bugs to me and that I had not answered.
3592 If your bug is still in the new version, please tell me. Here is a
3593 summary of the main changes.
3595 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3596 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3597 portability, speed, and a lot more but you'll figure it out later.
3598 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3599 I wanted to release the new MSG as soon as possible and I have
3600 broken GRAS, which is the reason why, for now, only MSG is fully
3601 functional. A laconic description of these projects is available
3602 in the documentation.
3604 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3605 kernel optimized for our needs (called SURF but only the developers
3606 should use it). Hence, MSG is now roughly 30 times faster and I think
3607 that by rewriting a little bit MSG, I could event speed it up a little
3608 bit more. Beside the gain in speed, it is also much easier to encode a
3609 new platform model with SURF than it was with SG. More to come...
3611 * REVOLUTION 3: I have tried to change a little as possible the API of
3612 MSG but a few things really had to disappear. The main differences
3613 with the previous version are:
3614 1) no more m_links_t and the corresponding functions. Platforms are
3615 directly read from a XML description and cannot be hard-coded
3616 anymore. The same format is used for application deployment
3617 description. The new format is described in the documentation.
3618 Have a look in tools/platform_generation. There is a tiny script
3619 that converts from the old platform format to the new one. Concerning
3620 the application deployment format, parsing the old one is tricky.
3621 I think most of you should however be able to convert your files. If
3622 it is really an issue, I can write a C code that does the conversion.
3624 2) the toolbox tbx does not exist anymore. We now have a library
3625 with much more data-structures but without the hash-tables (we have
3626 dictionaries that are much faster).
3628 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3630 *****************************************************************************
3631 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3632 * don't feel like converting the oldies. (Mt) *
3633 *****************************************************************************
3636 Version 2.90: "the long awaited one"
3637 - Finished rewriting and debugging MSG. Rewrote the documentation.
3638 - disable GRAS for now since it needs to be ported to the newest SG
3641 - Finish the port to windows (using mingw32 for cross-compile)
3644 - Main loop and datastructures of SURF. A cpu resource object is
3645 functional. Surf can thus be used to create cpu's with variable
3646 performance on which you can execute some actions.
3648 2004-11-15 Martin Quinson
3649 - Port to ARM. Simply added the alignment and size descriptions. Should
3650 work, but the ARM machines are so slow that I didn't had the opportunity
3651 to 'make check' over there yet.
3653 2004-11-15 Arnaud Legrand
3654 - Trace manager now written. It uses a heap structure and is therefore
3655 expected to be efficient. It may however be speeded up (particularly
3656 when many events occur at the same date) by using red and black
3657 trees. One day maybe...
3658 - Max-min linear system solver written. It uses a sparse matrix
3659 structure taking advantage of its expected use. Most operations are
3660 O(1) and free/calloc are called as few as possible. The computation of
3661 the minimum could however be improved by using a red and black tree
3664 2004-11-03 Arnaud Legrand
3665 - Rename every gras_* function that was in xbt/ to its xbt_
3667 - Add a heap and a doubly-linked list to xbt
3668 - Added a dichotomy to the dictionaries. make check works as well before
3669 so I assume that the patch is correct. I do not know however if things
3670 run effectively faster than before now. :)
3672 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3673 SimGrid, and the version number is bumped to 2.x
3675 2004-10-29 Martin Quinson
3676 - Introduction of the remote errors.
3677 They are the result of a RMI/RPC on the remote machine.
3678 ErrCodes being scalar values, you can't get the host on which those
3679 errors did happen. Extending the error mechanism as in Gnome is possible.
3680 No idea yet whether it is a good idea.
3682 2004-10-28 Martin Quinson
3683 - Interface revolution: the Starred Structure Eradication.
3684 I used to do typedef struct {} toto_t; and then handle *toto_t.
3685 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3686 - ???_t is a valid type (builded with typedef)
3687 - s_toto_t is a structure (access to fields with .)
3688 - s_toto is a structure needing 'struct' keyword to be used
3689 - e_toto_t is an enum
3690 - toto_t is an 'object' (struct*)
3692 typedef struct s_toto {} s_toto_t, *toto_t;
3693 typedef enum {} e_toto_t;
3694 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3695 s_toto_t) is private.
3697 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3698 so that it can set the variable to NULL. It was so for dicts and sets,
3699 it changed for dynars.
3701 - Fix a bunch of memleaks in dict_remove
3702 - Fix a bug in sg/server_socket opening: it failed all the time.
3704 2004-10-07 Martin Quinson
3705 - Speed up dynar lookup operation a bit.
3707 gras_dynar_get is dead.
3709 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3710 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3711 (which gives you the address of the stored data).
3713 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3714 copy of the data using an affectation to do the job and not a memcpy.
3716 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3717 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3719 It does not really speedup the dynar test because they are
3720 setting elements all the time (and look them seldom). But the dict does
3721 far more lookup than setting.
3723 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3725 2004-10-05 Martin Quinson
3726 - Allow to (en/dis)able the cycle detection at run time.
3728 Whether we should check for cycle or not is now a property of each
3729 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3730 datadesc_cycle_unset allow to remove this property when previously set.
3732 Note that the cycle detection is off by default since it impacts the
3733 performance. Watch the data you feed GRAS with ;)
3735 This property is hereditary. Any element embedded in a structure having it
3736 set have it set for the time of this data exchange.
3738 You should set it both on sender and receiver side. If you don't set it on
3739 sender side, it will enter an endless loop. If you forget on receiver
3740 side, the cycles won't be recreated after communication.
3742 - Header reorganization.
3743 Kill gras_private.h, each submodule must load the headers it needs.
3745 2004-10-04 Martin Quinson
3746 - Interface revolution: do not try to survive to malloc failure.
3748 Now, gras_malloc and friends call gras_abort() on failure.
3749 As a conclusion, malloc_error is not a valid error anymore, and all
3750 functions for which it was the only gras_error_t return value are
3751 changed. They now return void, or there result directly.
3752 This simplify the API a lot.
3754 2004-09-29 Martin Quinson
3755 - Re-enable raw sockets.
3756 Created by gras_socket_{client,server}_ext;
3757 Used with gras_raw_{send,recv}
3760 It should allow to kill the last bits of gras first version soon.
3762 This is not completely satisfactory yet (duplicate code with
3763 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3766 - Simplify transport plugin (internal) interface by not passing any
3767 argument to _server and _client, but embedding them in the socket
3770 2004-09-28 Martin Quinson
3771 - Finish the port to AIX.
3772 autoconf was my problem (segfault within the malloc replacement
3773 function. No idea why)
3775 2004-09-16 Martin Quinson
3776 - Fix some size_t madness on 64bit architectures.
3778 2004-09-08 Martin Quinson
3779 - Reduce the number of system headers loaded, overload some more system
3780 calls (such as malloc to cast the result of the system one, and work
3782 - Fix and reintroduce the config support
3784 2004-09-07 Martin Quinson
3785 - Source code reorganization to allow Arnaud to surf all over there.
3786 - Allow to document the logging categories.
3787 - Remove all uppercase from logging categories and useless cleanup in names.
3789 2004-08-18 Martin Quinson
3790 Version 0.6.2 (protocol not changed; API changed)
3791 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3792 gras_error_t), and NULL when not found. Functions expecting a msgtype
3793 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3794 hopefully usefull message.
3795 - Portability to prehistoric sparcs again
3797 2004-08-17 Martin Quinson
3798 Version 0.6.1 (protocol not changed; ABI not changed)
3799 - prealloc some buffers to speed things up
3801 2004-08-11 Martin Quinson
3802 Version 0.6 (protocol not changed; ABI expended)
3803 - The parsing macro can deal with the references, provided that you add
3804 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3806 2004-08-09 Martin Quinson
3807 Version 0.5 (protocol not changed; ABI changed)
3808 - Allow to off turn the cycle detection code in data exchange at
3809 compilation time. It should be at run time, but I'm short of time (and
3810 the config stuff is still broken). That way, we keep dict out of the
3811 critical path, which is good because the performance is poor:
3812 - search not dichotomial yet
3813 - dynar give no way to access their content and memcpy everytime
3814 - In composed data description (struct, ref and so on), stop foolness of
3815 keeping the subtype's ID, but store the type itself. This keeps sets out
3816 of the critical path, which is good since they rely on dynar and
3817 dictionaries. The only loose of that is that we cannot detect the
3818 redeclaration of a structure/union with another content (but I'm not sure
3819 the code detected well this error before anyway). We still can detect
3820 the redefinition discrepancy for the other types.
3821 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3822 it breaks the code because of type-punning used all over the place).
3823 This breaks on all non-gcc architectures (for now).
3825 All those changes (plus the buffer of last time) allow me to gain 2 order
3826 of magnitude on cruel tests consisting of 800000 array of integers on two
3827 level of a hierarchical structure (200 secondes -> 4 secondes)
3830 - the selector of reference must now return the type it points to, not
3831 the ID of this type.
3833 2004-08-06 Martin Quinson
3834 Version 0.4 (protocol changed; ABI not changed)
3835 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3836 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3837 - Implement a buffer transport, and use it by default (it relies on tcp in
3838 real life and on sg in simulation).
3839 That's a bit hackish since I had a new field to the structure to store
3840 its data without interfering with the subtype ones. Inheritance
3841 is tricky in C. And that's a kind of reverse inheritance with one class
3842 derivating two classes. Or maybe a game with java interfaces. Anyway,
3843 that's damn hard in C (at least).
3844 Moreover, I got tired while trying to ensure plugin separation and
3845 genericity in SG mode. MSG wants me to do weird things, so let's go for
3846 cruel hacks (temporarily of course ;).
3847 See comment in transport_private.h:71
3848 - do not include all the _interface headers in private but in the files
3849 which really need them (to cut the compilation time when they are
3852 2004-07-26 Martin Quinson
3853 Version 0.3 (protocol not changed; ABI changed)
3854 - Major overhault of the datadesc interface to simplify it:
3855 - shorted the function names:
3856 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3857 - add a trivial way to push/pop integers into the cbps without malloc.
3858 This allows to make really generic sub_type description, which simply
3859 pop their size of the stack.
3860 - add a function gras_datadesc_ref_pop_arr() which does what users want
3861 most of the time: Declare a dynamic array (which pops its size of the
3862 stack) and declare a reference to it. Poor name, but anyway.
3863 - kill the post-send callback, add a post-receive one
3865 2004-07-23 Martin Quinson
3866 Version 0.2 (protocol changed; ABI changed)
3867 - add some testing for cpbs in the test cases, and fix some more bugs.
3868 This invalidate again the little64 data file, since I cannot regenerate
3870 - remove an awfull optimization in the logging stuff, allowing me to:
3871 - understand it again
3872 - learn gcc how to check that the argument match the provided format
3873 - fix all errors revealed by gcc after that
3874 - internal keys of dict are not \0 terminated. Deal with it properly in
3875 loggings instead of segfaulting when the user want to see the logs :-/
3877 2004-07-22 Martin Quinson
3878 - Fix some stupid bug preventing cbps (callback postit) from working
3880 2004-07-21 Martin Quinson
3881 - Some documentation cleanups
3882 - remove the useless last argument of msgtype_declare
3883 - rename the Virtu functions to fit into the 'os' namespace
3884 - move headers src/include -> src/include/gras/ and stop fooling with
3885 gras -> . symbolic link
3886 - make distcheck is now successful
3888 2004-07-19 Martin Quinson
3890 - Build shared library also
3891 - Install html doc to the right location
3892 - stop removing maintainer files in make clean
3893 - build tests only on make check
3895 2004-07-13 Martin Quinson
3897 - No major issue in previous version => change versioning schema
3898 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3899 corresponding dataset.
3901 2004-07-11 Martin Quinson
3903 - Get it working with any kind of structure (we can compute the padding
3904 bytes remotely for all the architectures I have access to)
3905 - Implement the structure parsing macro (still not quite robust/complete)
3906 - Improvement to the remote testing toysuite
3908 2004-07-10 Martin Quinson
3909 [autoconf mechanism]
3910 - get ride of a bunch of deprecated macros
3911 - actually run the test for two-compliment, not only compile it :-/
3912 - test whether the structures get packed (and bail out if yes. Damn.
3913 Alignment is a serious matter)
3914 - test whether the structures get compacted (but respecting the alignment
3915 constraints of each types)
3916 - test whether the array fields of structures can straddle alignment boundaries
3918 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3919 (took me 2 hours to find that bug, looking at the wrong place)
3920 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3921 before sending/receiving and are used to compute the offsets of fields
3922 - Given that padding size depend even on compiler options, keep track of
3923 alignment and aligned_size only for the current architecture. Not a big
3924 deal since we send structure fields one after the other (seems
3926 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3927 program, let it work on linux/gcc/little32. portability todo.
3929 2004-07-08 Martin Quinson
3930 - import and improve remote compilation support from FAST
3931 - make sure make check works on half a dozen of machines out there
3933 2004-07-07 Martin Quinson
3934 Let's say it's version 0.0.3 ;)
3935 - Implement conversions (yuhu!)
3936 - Let it work on solaris (beside conversion, of course)
3937 - Stupid me, using rand() to generate the conversion datatests in not wise.
3939 2004-07-06 Martin Quinson
3940 - Let make dist work, since I'm gonna need it to compile on remote hosts
3941 - Let Tests/datadesc_usage write the architecture on which the file was
3942 generated as first byte.
3943 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3944 architecture descriptions.
3945 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3946 on those architectures.
3947 - Optimization: send/recv array of scalar in one shoot
3949 2004-07-05 Martin Quinson
3950 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3952 - Plug a whole bunch of memleaks
3953 - each process now have to call gras_{init,exit}. One day, their log
3954 settings will be separated
3955 - Continue the code factorisation between SG, RL and common in Transport.
3957 2004-07-04 Martin Quinson
3959 - Redistribution between SG and RL.
3960 We wanna have to accept in SG, so move accepted related parts of RL in
3961 the common part. (more precisely, the dynar of all known sockets is no
3962 more a static in transport.c, but part of the process_data)
3964 [gras_stub_generator]
3965 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3967 2004-07-03 Martin Quinson
3968 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3969 them all up in one shot)
3971 - Fix the ugly case of reference to dynamic array.
3972 - New (semi-public) function gras_datadesc_size to allow the messaging
3973 layer to malloc the needed space for the buffer.
3975 - gras_socket_close now expect the socket to close (and not its address to
3976 put NULL in it after it). This is because the socket passed to handlers
3977 is one of their argument (=> not writable).
3979 - propagate the interface cleanup from last week in datadesc, ie remove a
3980 superfluous level of indirection. User pass adress of variable
3981 containing data (both when sending and receiving), and not of a variable
3982 being a pointer to the data. Let's say that I like it better ;)
3983 The price for that is constructs like "int msg=*(int*)payload" in
3984 handlers, but it's a fine price, IMHO.
3986 - Let it work in RL (yuhu)
3988 2004-06-21 Martin Quinson
3990 - porting SG plugin and SG select to new standards (works almost).
3991 - plug memleaks and fix bugs around.
3994 - cleanup the prototype of data recv and force users to specify when they
3995 want to handle references to objects. Test case working even for cycles.
3996 - plug memleaks. Valgrind is perfectly ok with this.
3998 2004-06-12 Martin Quinson
4000 - cleanup the separation between plugin and main code in plugin creation
4002 2004-06-11 Martin Quinson
4004 - Reput hook for raw sockets, needed for BW experiments
4005 - kill a few lines of dead code
4006 [Data description] Interface cleanup
4007 - gras_datadesc_by_name returns the searched type or NULL.
4008 That way, no variable is needed to use a type desc once, which makes
4010 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4011 two parameters were strings (field name, type name), leading to
4013 [Dicos] Interface cleanup
4014 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4015 This is consistant with the dynar API.
4017 2004-04-21 Martin Quinson
4019 - Porting to new standards.
4021 - interface cleanup.
4022 There is no bag anymore, no need to take extra provision to mask the
4023 pointers behind "ID".
4024 Better splitup of functions between files create/exchange/convert.
4025 This is still a bit artificial since convert and receive are so
4026 interleaved, but anyway.
4028 - add a queued message list to procdata (the ones not matching criteria
4030 - factorize some more code between SG and RL wrt procdata
4032 - use gras_exit in example to track memleaks
4033 - get rid of gs_example now that GS is properly integrated into gras
4034 - update run_test to integrate the latest tests (datadesc)
4036 - rename WARNINGn macros to WARNn since it prooved error-prone
4038 2004-04-19 Martin Quinson
4040 - register init/exit functions within gras module mechanism
4041 - send/receive function.
4042 Convertion is not implemented, but short-cutted if not needed.
4043 struct/array elements are sent one by one (instead of block-wise), but
4044 nobody really cares (yet). Get a prototype before optimizing.
4045 - tests (using a file socket) for DD send/receive on:
4046 - base types: int, float
4047 - array: fixed size, string (ie ref to dynamic string)
4048 - structure: homogeneous, heterogeneous
4049 - chained list, graph with cycle
4050 Believe it or not, valgrind is not too unhappy with the results. The
4051 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4055 - Bugfix when using a filename explicitly (instead of '-')
4057 2004-04-09 Martin Quinson
4059 - factorize more code between RL and SG in socket creation
4060 - Complete the implementation and tests of:
4062 o file (only in RL, and mainly for debugging)
4064 I lost 3 days to design a portable address resolver, and then decided
4065 that the prototype mainly have to run on my box.
4066 Addressing portability too early may be like optimizing too early :-/
4068 - use gras_init in the Tests instead of the crappy parse_log_opt
4069 (the latter function is removed)
4070 [Conditional execution]
4071 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4072 [Code reorganisation]
4073 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4074 Build examples against the RL library, since there is no way to disable
4075 its creation for now.
4077 For information, the beginning of coding on GRAS was back in june
4078 2003. I guess that every line has been rewritten at least twice since