1 ----------------------------------------------------------------------------
3 SimGrid (3.23.3) NOT RELEASED YET (v3.24 expected September 23. 7:50 UTC)
6 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
7 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
8 thus becomes the default resource to execute I/O operations. Examples
9 and the file system plugin have been modified to use disks instead of
10 storage elements. Storage can still be used but is doomed to disappear.
11 Users are thus adviced to update their simulators accordingly.
12 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
13 for consistency with pthread_barrier_wait()
14 - Host::get_englobing_zone() returns the englobing netzone
15 - Actor::on_destruction is now called in the destructor
16 Actor::on_termination new signal called when the actor terminates
18 - Global signals are now part of the Engine:
19 - on_platform_creation: after config settings, before the XML parsing
20 - on_platform_created: right after the XML parsing
21 - on_time_advance: each time the clock advances
22 - on_simulation_end: after simulation, before cleanups
23 - on_deadlock: as the name implies.
25 - sg_{actor,host,link}_{data,data_set}() now all exist.
26 Use them to attach user data to the object and retrieve it.
29 - Improved the usability of ns-3. Several bugs were ironed out.
30 - Introduce an experimental Wifi model. It sounds reasonable
31 according to the state of the art, but it still has to be properly
32 validated, at least against ns-3.
35 - convert a new set of functions to the S4U C interface and move the old MSG
36 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
39 - Fortran bindings for DVFS have been removed.
40 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
41 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
42 - SMPI can now be selected by cmake's find_module(MPI) with
43 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
44 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
45 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
46 - trace-call-location can now be used with TI traces, and replayed.
47 - Default for trace-call-location is now to use file names and not full paths.
48 To revert to previous behaviour (in case of collision of filenames), option
49 "smpi/trace-call-use-absolute-path" can be set to yes.
52 - Use the included xxHash as an hash implem when C++14 is usable.
53 - Option model-checker/hash was removed. This is always activated now.
54 - New option smpi/buffering controls the MPI buffering in MC mode.
55 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
56 This is useful to check for MPI compliance.
59 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
60 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
63 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
64 and <mount> tags. The rationale is finish to decouple what should be handle
65 by SimGrid kernel (i.e., resources and activities) from plugins built in
66 user space (e.g., the file system plugin). This new tag currently accepts
67 only three attributes (id, read_bw, and write_bw). All the other information
68 that was declared with the storage related tags now has to be expressed as
69 properties. An example of platform using this new tag is available at
70 examples/platforms/hostsè_with_disks.xml
72 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
73 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
74 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
75 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
76 - FG#31: per-actor data
77 - FG#34: SG_BARRIER_SERIAL_THREAD?
78 - FG#35: model-checker does not like buster-produced binaries
79 - FG!13: MC: complete workaround in the error msg seen on modern systems
80 - FG!15: execute_flops now logs compute
81 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
82 - FG!17: ns-3: unblock the right number of communications + others issues
83 - FG!18: Improving the performance of the ns-3 bindings
84 - GH#219: Error in the throughput of TCP transfer
85 ----------------------------------------------------------------------------
87 SimGrid (3.23.2) July 8. 2019
90 - Nicer introduction page.
91 - Migrate the "Deploy your application" page to the new doc.
92 - Move Java as a subtree of MSG.
95 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
98 - Drop xbt_dynar_sort_strings().
101 - Really fix FG#26: Turning off a link should raise NetworkFailureException
102 - FG#27: Wrong exception thrown to wait_any when link is turned off
103 - GH#328: Java: Canceling multiple tasks in a single vm/host
105 ----------------------------------------------------------------------------
107 SimGrid (3.23) June 25. 2019
109 The Exotic Solstice Release.
112 - SunOS and Haiku OS support. Because exotic platforms are fun.
113 - Stop setting random seed with srand48() at initialization.
114 - Use addr2line as a fallback for stacktraces when backtrace is not available.
115 - Build option -Denable_documentation is now OFF by default.
116 - Network model 'NS3' was renamed into 'ns-3'.
119 - Simgrid can now hopefully be installed with pip.
122 - wait_any can now be used for asynchronous executions too.
125 - New log appenders: stdout and stderr. Use stdout for xbt_help.
126 - Drop xbt_dict_dump.
129 - 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
130 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
131 - MPI interface is now const correct for input parameters
132 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
133 - Fortran bindings for async collectives.
134 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
137 - Remove option 'model-check/record': Paths are recorded in any cases now.
138 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
139 always sparse. That's an immense gain in memory, and can even be
140 faster because of cache effects. So there is no need to clutter the
141 code to allow the user to go for the unefficient mode.
144 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
145 models proved to be more accurate than these old experiments.
147 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
148 - FG#1: Broken link in error messages
149 - FG#2: missing installation documentation
150 - FG#3: missing documentation in smpirun
151 - FG#6: Python bindings not available on PyPI
152 - FG#7: simple cmake call requires doxygen
153 - FG#8: make python bindings an optional dependency
154 - FG#10: Can not use MSG_process_set_data from SMPI any more
155 - FG#11: Auto-restart actors forget their on_exit behavior
156 - FG#12: -Denable_lto=OFF doesn't disable LTO
157 - FG#13: Installs unstripped file 'bin/graphicator'
158 - FG#14: Installs the empty directory 'doc/simgrid/html'
159 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
160 - FG#17: Dead link in doc (pls_ns3)
161 - FG#20: 'tesh --help' should return 0
162 - FG#21: Documentation link on http://simgrid.org/ broken
163 - FG#22: Debian installation instruction are broken
164 - FG#26: Turning off a link should raise NetworkFailureException exceptions
165 - GH#133: Java: a process can run on a VM even if its host is off
166 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
167 - GH#326: Valgrind-detected error for join() when energy plugin is activated
169 ----------------------------------------------------------------------------
171 SimGrid (3.22) April 1. 2019
173 The Easter Christmas Release (this one is somewhat late).
176 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
177 - Not complete yet: asynchronous activities (amongst others) are still missing
178 - Still ongoing: the interface may change in the future. We need more
179 testers! Please report any glitches.
180 - No new project using Java should start now. Please switch to Python.
183 - Some of the internal cleanups may lead to speed improvements:
184 - The hard limitation on the amount of simulated actors with
185 Java+Mac was removed. Now, the available memory is the only limit.
186 - Our refcounting was tidyied, leading to 10% speedups in some cases.
187 - We are still working on making our code robust to the actor kills
188 and hosts' churn. Things are improving, but it's not perfect yet.
189 - Replaced our own code to display a backtrace (that was forking addr2line)
190 with the Boost.Stacktrace library.
191 You won't see your backtraces without this optional dependency.
192 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
193 - Stop setting random seed with srand() at initialization.
196 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
197 XML file version remains 4.2 since old files are still compatible.
200 - Process termination which was broken at version 3.21 has been repaired.
201 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
202 - Hide the examples into examples/deprecated. New users should use Python.
205 - Drop MSG_process_create_from_stdfunc() from the API.
206 This C++ function was a pimple in the C API, made necessary at some
207 point by the Java bindings. This is fixed now.
208 - Hide the examples into examples/deprecated. New users should use S4U.
209 - MSG_process_create and MSG_process_attach now crash if the host used
210 to run this process is off.
211 - Fix the protype of MSG_process_on_exit()
212 Now use (int,void*) callbacks instead of (void*,void*) ones.
213 The implementation was ways too messy to actually work, I guess.
216 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
217 too often and provide more accurate timings.
218 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
219 - MPI_Alltoallw support
220 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
221 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
222 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
223 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
225 - Basic implementation of generalized requests (SMPI doesn't
226 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
229 - Drop sg_cmdline. Please use xbt_cmdline instead.
230 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
231 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
232 OsSemaphore is implemented in a portable way with C++11 threads.
233 This should allow much more threads to be created at the same time,
234 allowing Mac and Java users to simulate many more actors.
235 - Implement the 'thread' factory with std::thread instead of xbt ones.
236 It is not possible to set the stack size with threads anymore, but
237 -fsplit-stack is the way to go nowadays when using threads.
238 - Drop the xbt_os_thread_t module (now unused)
239 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
240 - Drop xbt_str_join_array().
241 - Drop cunit, use Catch2 instead.
244 - Many cleanups in the kernel::activity namespace. This was long
245 overdue, and shall open the path to many future endeavors.
248 - #132: Java: a process can not shut down its own host
249 - #220: S4U: detached send causes exception if sender terminates soon after sending
250 - #261: Document the parameters of parallel execution's constructor
251 - #300: [s4u] BarrierPtr is missing
252 - #314: SMPI args internal cleanup
253 - #316: Fix a bug related to the CPU utilization of multi-core VM
254 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
255 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
256 - #325: Turning off a host has different behavior on sleeping actors and computing actors
258 ----------------------------------------------------------------------------
260 SimGrid (3.21) October 3. 2018
262 The Restarting Documentation (TRD) Release.
265 - Convert the user manual to Sphinx for improved usability.
266 Unstable version now hosted on https://simgrid.frama.io/simgrid
267 Many glitches remain; stay tuned.
268 - Restrict the Doxygen reference API to the installed header files.
269 (documenting all internals this way is overkill + doxygen scales badly)
270 - New tutorial on S4U (highly inspired from the old MSG one, but with
271 a git repository to fork as a starting point, and a docker image)
272 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
273 - The Developper manual is still to be converted (not compiled until then)
274 - Some parts are still missing in this conversion (such as the
275 platform and deployment sections) while others were blindly converted
276 and would need more love (such as the configuration flags).
277 Things will certainly further improve in the future releases.
280 - Fully reimplement auto_restart mechanism that was utterly broken
281 (fix #22, #131 and #281 that were all very related)
282 - Implement semaphores (fix #303 and #291)
283 - When creating an actor from a function and its parameters,
284 move-only parameters are not allowed anymore, as it would prevent
285 the actor to be restartable if its parameters are consumed on run.
286 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
287 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
288 IO operations that can be started, waited for, or canceled as a regular
290 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
294 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
295 - New host variable: 'core_count'
298 - Remove xbt_os_thread_specific features
299 - Remove portability wrapper to condition variables
300 - Remove xbt_os_thread_yield()
303 - MPICH collective selector now mimics MPICH 3.3b
304 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
305 - MPI_Init(NULL,NULL) is now allowed
306 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
307 - Command-line arguments should now be handled properly in Fortran simulations
308 - MPI Attributes and related callbacks should work in Fortran
309 - Apps using C + Fortran should now work
310 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
311 - Support MPI_Error_String
314 - Due to an internal bug, Msg.run() must now be your last line.
315 We hope to fix it in a future release, and we are sorry for the inconvenience.
318 - #22: Process autorestart seem to only work with CAS01 cpus
319 - #56: Feature request: dynamic SMPI replay
320 - #93: simgrid should not eat --help
321 - #111: How to change the loopback link characteristics?
322 - #116: Platform cluster radical
323 - #117: Platform cluster/host order
324 - #129: Replay_multiple on steroid
325 - #131: Java : autostart does not work
326 - #143: Setting a breakpoint at a given time
327 - #235: Network energy models should be integrated
328 - #264: Add ptask L07 resource tracing
329 - #271: Dynamic and manual replay of SMPI traces
330 - #279: Breakpoints option uses time deltas instead of absolute time
331 - #281: Daemonized actor no longer daemonized after an auto-restart
332 - #282: TIT convertor erase traces when using absolute path in the trace list file
333 - #285: segfault when a s4u actor kills itself with msg process tracing activated
334 - #286: Pajé traces are not exposing the number of cores
335 - #287: Command-line parsing should stop when encountering '--'
336 - #288: MPI_Init(NULL, NULL)
337 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
338 - #290: Method to check if a mailbox has a message ready to be consumed
339 - #291: [s4u] semaphores do not seem available yet
340 - #294: Alltoall collective implementation of mpich
341 - #295: Starting a Host as OFF is not supported
342 - #296: DTD too permissive
343 - #299: Add a s4u barrier example
345 ----------------------------------------------------------------------------
347 SimGrid (3.20) Released June 24. 2018
349 The proxy snake_case() release.
352 - The whole API is now using snake_case() instead of mixing it with camelCase.
353 Compatibility wrappers in place for one year.
354 - We took the opportunity to increase the consistency of the API.
355 Please report any remaining glitches.
356 - New function simgrid::s4u::Host::get_actor_count:
357 Returns the number of actors running on a specific host.
360 - SMPI is now tested with ~45 proxy apps from various sources, with none or
361 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
362 - Replay: The replay file has been re-written in C++.
363 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
364 supported. They are stored in the trace and used when replayed.
365 - Basic support of MPI_Cancel. Robustness not guaranteed.
366 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
367 (only for MPI_COMM_TYPE_SHARED).
368 - New option: smpi/privatize-libs, to add external shared libs to be privatized
369 by SMPI. They will be copied locally and loaded separately by each process.
370 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
371 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
372 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
373 with internal index optind with multiple processes. Only works if MPI_Init has
375 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
376 syntax without preprocessing tricks.
379 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
380 behavior tracing is no longer limited to MSG
381 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
382 this means that traces now need to be updated or re-obtained; both Irecv
383 and Isend lines in your traces have now in total 5 mandatory fields:
384 <rankid> <command> <to/from rankid> <tag> <size>
385 To update your traces, it suffices to add a 0 for the tag here.
386 - TIT now also supports waiting for a distinct request via MPI_Wait.
387 Wait/Test now wait for a specific request, not just the last one that was
388 issued. This unfortunately means another update, because we need to
389 identify which request you want to wait for. We do this via the
390 triplet (sender, receiver, tag), which needs to be added:
391 <rankid> <command> <sender> <receiver> <tag>
392 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
393 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
396 - The deprecation of MSG is ongoing (but this should not impact you).
397 Many MSG functions are now simple wrappers on the C API of S4U. If
398 you wish to convert your code to S4U, find the S4U counterparts of
399 your MSG calls in src/msg/msg_legacy.cpp.
400 - MSG can still be used, but won't evolve anymore.
403 - Allow to run the Link energy plugin from the command line with
404 --cfg=plugin:link_energy
405 - Rename Energy plugin into host_energy
406 - Rename Load plugin into host_load
409 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
410 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
411 - Improve performance when handling timeouts of simix synchros.
414 - Config: the C API is now deprecated (will be removed in 3.23), and
415 the C++ API has been slightly improved.
418 - Fix several build issues on OSX.
419 - Move simgrid_config.h to simgrid/config.h (old header still working)
422 - #143: Setting a breakpoint at a given time
423 - #150: Inconsistent event names in SMPI replay
424 - #258: daemonized actors hang after all non-daemonized actors have completed
425 - #267: Linker error on unit_tmgr
426 - #269: SMPI: tracing of MPI_Wait/all/any broken
427 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
429 ----------------------------------------------------------------------------
431 SimGrid (3.19.1) Released March 22 2018
433 The Fixed ABI Release.
435 As you may know, we are currently refactoring SimGrid in deep.
436 Upcoming SimGrid4 will be really different from SimGrid3: modular,
437 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
439 Our goal is to smooth this transition, with backward compatibility
440 and automatic update paths, while still progressing toward SimGrid4.
442 SimGrid remains open during works: The last pure SimGrid3 release was
443 v3.12 while all subsequent versions are usable alpha versions of
444 SimGrid4: Existing interfaces remain unchanged, but the new S4U
445 interface is budding and the internals are deeply reorganized.
447 Since 2015, we work hard to reduce the changes to public APIs. When
448 we need to rename a public library symbol in S4U, we let your
449 compiler issue an explicative warning when you use the deprecated
450 function. These messages remain for four releases, i.e. for one full
451 year, before turning into an error. Starting with v3.15, your can
452 also adapt to API changes with the SIMGRID_VERSION macro, that is
453 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
455 Starting with this v3.19.1, our commitment to reduce the changes to
456 the public interfaces is extended from the API to the ABI: a program
457 using only MSG or SimDag and compiled against a given version of
458 simgrid can probably be used with a later version of SimGrid without
459 recompilation. We will do our best... but don't expect too much of
460 it, that's a really difficult goal during such profund refactoring.
462 The difference between v3.19 and v3.19.1 is that the former was
463 accidentally breaking the ABI of MSG, while the later is restoring
466 S4U and kernel APIs will still evolve until SimGrid4, with one-year
467 deprecation warnings as currently. In fact, cleaning up these
468 interfaces and converting them to snake_case() is one release goal of
469 v3.20. But don't worry, we are working to smooth this upgrade path.
471 In summary, new projects should start with S4U to benefit of the
472 future, but old MSG projects should still be usable with no change.
474 ----------------------------------------------------------------------------
476 SimGrid (3.19) Released March 20 2018
478 The Moscovitly-cold Spring release.
480 SMPI user-visible improvements:
481 - Support of MPI_Type_create_subarray
482 - Unimplemented MPI calls now abort (instead of issuing a warning)
483 Some harmless calls around errhandlers still only issue a warning.
484 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
485 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
486 is now possible (mmap privatization was refactored to this end).
487 Adding new SMPI instances should also be technically doable now.
489 SMPI internal cleanups:
490 - Removed index notion from SMPI -- no more "getPid() - 1"!
491 In tracing files, actors are called rank-<PROCESS-ID> but these are
492 NOT the real ranks (tracing several communicators is not supported yet).
493 - Deprecate smpi_process_index() (will be removed in 3.22)
494 - Better testing on classical proxy apps, through an external project:
495 https://github.com/simgrid/SMPI-proxy-apps/
498 - Use a graphical TOC to make it easier to find the documentation you need
499 - Revamp the MSG tutorial
502 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
503 - Remove parameter of MSG_process_killall().
504 Resetting the PID was bogus anyway (several actors could have the same PID).
507 - Execution->setHost() can be called after start(), to migrate it.
508 - Comm::test_any() is now implemented.
509 - s4u::Actor now has onCreation() and onDestruction() signals.
510 - Install some kernel header files for the users' plugins and more.
511 Warning, their API is really not stable yet.
512 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
513 include/simgrid/kernel/routing/*Zone.hpp
516 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
518 Build System and other cleanups:
519 - Remove unused run-time parameter "tracing/onelink-only".
520 - SimGrid now works with intel compilers (MC still buggy)
521 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
522 - tesh kills subprocesses on timeout (but not on windows)
525 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
528 - #194: Feature request: simgrid::s4u::Comm::test_any()
529 - #245: migrating an actor does not migrate its execution
530 - #253: Feature Request: expose clusters as objects
531 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
532 - #255: Tesh broken on Windows
533 - #256: Modernize FindSimGrid.cmake
534 - #257: Fix (ab)use of CMake install
536 ----------------------------------------------------------------------------
538 SimGrid (3.18) Released December 24 2017
540 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
542 MSG IS NOW DEPRECATED. Please use S4U for new projects.
543 - Support for MSG should not be removed from SimGrid before 2020,
544 but future evolutions will be limited to the integration of
545 user-provided patches.
548 - s4u::onDeadlock() signal, run before stopping the simulation.
549 - s4u::Exec: asynchronous executions (abstraction of background
551 - s4u::Host->getLoad() returns the achieved speed in flops/s
554 - Link::name() is deprecated and will be removed in v3.21, use
555 Link::getCname() instead.
556 - Mailbox::getName() changed to return a std::string, use
557 Mailbox::getCname() instead to get a char*.
558 - Storage::getName() changed to return a std::string, use
559 Storage::getCname() instead to get a char*.
560 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
561 to have both allocation and deallocation of the map in user space.
564 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
565 The amount of flops remaining to do is a vector, not a scalar.
566 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
567 name implies on both sequential and parallel tasks.
568 - Both changes fix GitHub's #223 using PR #237 as a basis.
569 Thanks Michael Mercier.
570 - Most examples were converted to S4U and hidden elsewhere at tests.
571 Rationale: we still want MSG to work; we want newcomers to use S4U.
574 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
575 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
578 - Switch to the faster dlopen privatization mechanism by default
579 - Documentation improvments
582 - Live migration is getting moved to a plugin. Dirty page tracking is
583 the first part of this plugin. This imply that VM migration is now
584 only possible if one this function is called:
585 - C/MSG: MSG_vm_live_migration_plugin_init()
586 - C/C++: sg_vm_live_migration_plugin_init()
587 - Java: Msg.liveMigrationInit()
588 For C and C++, "simgrid/plugins/live_migration.h" also has to be
592 - Define class simgrid::xbt::Path to manage file names.
593 - Removed unused functions:
594 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
595 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
596 xbt_graph_export_graphviz()
597 - xbt/str.h: xbt_str_join()
598 - Remove unused datatypes:
599 - xbt/heap.h: use std::priority_queue or boost::heap instead
600 - xbt/swag.h: use boost::intrusive::list instead
603 - New link_energy plugin for the consumption of the links.
604 - All of the operations on files and storage contents have been
605 packaged into a plugin (src/plugins/file_system). The current
606 public interface can be found in
607 include/simgrid/plugins/file_system.h
608 To use these functions you now have to initialize the plugin by
609 calling MSG_storage_file_system_init() just after calling
610 MSG_init() or sg_storage_file_system_init() just after creating
614 - Remove the undocumented/untested tag <include>
617 - Remove viva specific tracing as the tool is no longer maintained
620 - #248: Kill of finished processes leads to segfault
621 - #240: xbt_cond_wait_timeout should gracefully return for C apps
622 - #239: please implement signal s4u::onDeadlock()
623 - #230: segfaults when exit() without run()
624 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
625 - #223: MSG_task_get_flops_amount() not working with parallel tasks
626 - #222: Actor::kill() doesn't really kill and segfaults
627 - #221: odd LMM warning when killing an actor
628 - #120: Memory leak when the processes are forcefully killed
630 ----------------------------------------------------------------------------
632 SimGrid (3.17) Released October 8 2017
634 The Drained Leaks release: (almost) no known leaks despite the tests.
636 Even very long simulations will preserve your memory: our testsuite
637 stresses SimGrid for over 45mn on fast machines for a coverage of
638 over 80%, and there is only one single known leak, of about 4kb.
642 - Comm.detach(): start and forget about asynchronous emission. A cleanup
643 handler may be given to free resources if the comm cannot be completed.
644 - this_actor::send(mailbox) is now mailbox->put()
645 - New: simgrid::s4u::Comm::wait_all()
646 - New: Host.execute() for remote executions.
649 - Unused option network/sender-gap is removed.
652 - Deprecate MSG_task_isend_with_matching(): This unused feature
653 really complicates our internals. Will be removed in v3.20.
656 - Improved context termination. It is now possible to free resources, even
657 when a process is forcibly killed.
658 - ContextBoost: add support for Boost versions above 1.61.
661 - Bring back run-time option --cfg=exception/cutpath to remove exception
663 - Removed unused functions:
664 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
665 xbt_str_rtrim(), xbt_str_trim().
666 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
669 - Removed header files obsolete since SimGrid 3.12:
670 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
672 - #3: SD_exit should be made optional
673 - #120: Memory leak when the processes are forcefully killed
674 - #159: Threading test regression in Actor refcounting
675 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
676 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
677 - #186: Actor::killAll() segfaults if some process is blocked on wait()
678 - #191: VM migration and pstate
679 - #192: Updating the energy consumptions of all hosts crashes with VMs
680 - #195: All actors have PID=0 in the logs
681 - #204: Sometimes segfault with thread contexts and mmap privatization
682 - #222: Actor::kill() doesn't really kill and segfaults
683 - #225: Actor::kill() doesn't really kill when victims are doing a join()
685 ----------------------------------------------------------------------------
687 SimGrid (3.16) Released June 22. 2017.
689 The Blooming Spring Release: developments are budding.
692 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
693 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
694 - Java JAR file should be finally fully working on Mac OSX too.
696 XML platforms: Switch to platform v4.1 format.
697 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
698 - <zone> can be used as a synonym for the now deprecated <as>
699 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
700 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
701 - <actor> can be used as a synonym for the now deprecated <process>
702 - state_file and avail_file periodicity is now easier to express
703 (check the documentation) the old behavior should still work.
704 * Storage: not backward compatible, but it seems that nobody ever used it
705 - Remove attribute 'content_type' of <storage_type>: was never used
706 - Make attribute 'model' of <storage_type> optional: for future usage
707 - Remove Bconnection model property: was never (in)validated. Replaced by
708 the maximum of the read and write bandwidth as a resource constraint.
711 - New and Backwards Compatibility break:
712 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
713 When one wants to get the list of tasks whose states have changed during a
714 simulation round, s/he has to allocate and free a dynar and use it as
715 argument to this function. The former SD_simulate (double how_long)
719 - Allow multicore VMs, with the correct sharing (unless you overcommit)
720 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
723 - New function to update the consumption of all hosts at once.
724 - Fix the model for multi-core hosts, linear in the amount of busy
725 cores with an abnormality for pIdle (see paper).
726 BUG: Still not satisfactory for ptask on mono-cores.
729 - The netzone are now available from the MSG API.
730 The old names still work, but are now deprecated.
733 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
734 - New option: smpi/keep-temps to not cleanup temp files
735 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
736 mallocs mode is used, allows to change the size of the fake file used
737 (default 1MB), to potentially limit the number of mappings for large runs.
738 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
739 - Fortran ifort and flang compilers support
740 - New RMA calls supported (experimental):
741 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
742 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
743 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
744 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
745 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
746 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
749 - New: this_actor::isMaestro() returns whether we are in kernel mode.
750 - Behind the scene, ActivityImpl are now automatically refcounted.
751 Removing this nasty bottleneck will greatly simplify our internals.
754 - Replay: New function xbt_replay_action_get():
755 Retrieve the function previously associated to an event type.
756 - DROPPED FUNCTION: xbt_str_varsubst()
757 - DROPPED MODULE: strbuff. We don't need it anymore.
758 - DROPPED MODULE: matrix. We don't need it anymore.
759 - DROPPED MODULE: lib. We don't need it anymore.
761 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
763 SimGrid (3.15) Released March 22. 2017
765 The Spring Release: continuous integration servers become green
766 We fixed even the transient bugs on all target architectures:
767 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
768 Windows, FreeBSD, NetBSD.
770 - Rename NetCards to NetPoints.
771 This was intended to help NS3 users, but that's not a netcard.
772 That's a point in the routing algorithm, let's avoid wrong simplifications.
775 - New: MSG_process_yield(). Stop and yield to other processes.
776 - New: MSG_process_daemon(). Daemon processes are automatically killed
777 when the last non-daemon process terminates
778 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
779 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
780 to make room for the upcoming network energy plugin.
781 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
784 - Ensure that an actor can kill itself with Process::exit()
785 - Kill the obscure NativeException. Nobody want to survive the issues
786 it denotes, so use JniException that is a RuntimeException (not to
787 be caught explicitly).
788 - Partial bug fix in initialization. SimGrid flags on command line were
789 consumed at C level but stayed in the original Java String[] args.
790 This could mess users' args[i] if SG flags were not put at the end of
792 The SimGrid flags are now removed from the Java arguments. However,
793 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
794 if args.length is greater than the number of YOUR OWN ARGUMENTS.
795 It might be if you have --log or --cfg flags in the command line.
796 - Fix numerous memleaks all around the place. In particular, around VMs.
800 - s4u::onPlatformCreated: right before the simulation starts
801 - s4u::onSimulationEnd: right after the main simulation loop
802 - s4u::onTimeAdvance: right after a clock change (time is discrete)
803 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
804 event from the availability_file changes the avail speed.
805 - Links are now usable from s4u
806 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
807 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
808 - New examples: The conversion of MSG examples to S4U has begun
809 - Actors: create, kill, migration, and suspend
810 - Applications: master-worker and token-ring
811 - Action replay: communications and storage
812 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
815 - Backwards Compatibility breaks
816 - The SD_route_get_list and SD_route_get_size functions have been removed.
817 They are replaced by the unique
818 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
819 The route from src to dst is built in the links dynar whose size can
820 be obtained with xbt_dynar_length.
821 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
822 sg_host_route_bandwidth and sg_host_route_latency.
823 Macros ensure the backwards compatibility, but you should fix your code
826 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
827 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
828 - Bump our claim of support from MPI 1.1 to MPI 2.2.
829 We don't support 100% of it, but it should be enough. Contact us if not.
830 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
831 naive and probably inaccurate)
832 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
833 mapped onto a unique small file using some system magic.
834 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
835 - Fortran: cleanups, fixes, support of user-added operations
836 - RMA: MPI_Accumulate are applied in correct order.
837 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
838 - Better support for MPI_IN_PLACE
839 - Support for MPI_Win attrs and keyvals.
840 - Support MPI_Comm_group_create, MPI_Type_size_x
841 - MPI ops should be properly applied to all allowed types
842 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
843 algorithm by a non failing one from mpich.
846 - Kill the fifo data container: we don't use it anymore.
849 - A <cluster> can now be created with different speed values to represent pstates.
850 see examples/platforms/energy_cluster.xml for an example.
852 SimGrid (3.14.159) Released December 28. 2016
854 The previous release was only a crude approximation of a Pi release;
855 this one should be a bit better.
857 - Revert a last minute change that broke on Mac OSX.
858 - Fix the changelog and NEWS files.
859 - Improve the documentation, in particular of the routing module.
860 - Integrate some patches flying around in Debian.
862 SimGrid (3.14) Released December 24. 2016
865 * The whole documentation was reworked and reorganized. There is
866 still much room for improvement here, but we're on it.
869 * We now need python3 for our testing infrastructure.
870 * Model checking needs libevent
873 * Now works on FreeBSD
876 * Privatization now works on FreeBSD
878 * Privatization is now activated by default in smpirun
879 You can turn it off with -no-privatize if something goes wrong.
881 * Call-location tracing for SMPI.
882 You can add the exact location (filename / linenumber) of an MPI call to
883 your trace files and slow down or speed up the simulation between two
884 consecutive calls by using an adjustment file (see the documentation).
886 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
887 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
888 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
889 values were ignored in some cases. The timings of these functions can now
890 be significantly different.
892 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
893 smpi/running-power is renamed to smpi/host-speed
895 * smpi/grow-injected-times option to enable or disable multiplication of the
896 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
897 make simulation less precise (but also much faster).
899 * smpirun script should be (much) faster for large deployments.
901 * SMPI tracing: fixed issue with poor matching of send/receives.
903 * Replay: Fix broken waitall
905 New functions and features
906 * MSG_parallel_task_execute_with_timeout, to timeout computations.
908 Dropped / renamed functions and features
909 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
910 - MSG_mailbox_is_empty() -> Mailbox::empty()
911 - MSG_mailbox_front() -> Mailbox::front()
912 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
913 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
914 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
915 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
916 * Don't pass the free_f parameter to property related functions:
917 - MSG_host_set_property_value()
918 - MSG_as_router_set_property_value()
919 - MSG_storage_set_property_value()
920 * VM properties. Since msg_vm_t are msg_host_t, just use
921 MSG_host_get_property_value() and friends
922 * VM I/O related things:
923 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
924 - Unimplemented save/restore methods
925 * MSG_as_router_get_property_value() was redundant with
926 MSG_environment_as_get_property_value().
927 - Removed MSG_as_router_*propert*() functions
928 - Added MSG_environment_as_set_property_value() for consistency
929 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
930 Well, they are still there for now, but deprecated with a warning.
931 Please switch to xbt_dict_new_homogeneous() before this is removed
933 * Task affinity. Its intended behavior (that was very badly tested
934 and probably not really working) was deceiving what most users
935 would have hoped here.
936 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
937 * Option network/coordinates is now useless and should be dropped.
940 * Added option storage/max_file_descriptors to allow more than 1024 files opened
943 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
944 * Added Dragonfly topology support
947 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
948 EiBps, EBps, Eibps, Ebps for bandwidth.
949 They may become useful to some lucky ones.
952 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
953 * ex/app_tokenring: new example, very similar to the MSG Token Ring
954 * ex/async_waitAll: new example, on asynchronous communications
957 * Memory usage should be decreased for simulations with a large number
958 of processes. This also helps for SMPI.
960 SimGrid (3.13) stable; urgency=low
962 The Half Release, a.k.a. the Zealous Easter Trim.
964 - We removed half of the lines, that were mostly experimental cruft.
965 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
966 (not counting blanks and comments -- according to openhub.net)
967 - The internals are now compiled in C++ (and will soon be clean C++)
968 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
969 59 klines of C, etc. We added only 29 klines of C++ in replacement.
971 * Backwards Compatibility breaks
972 - Removed Lua simulation bindings (switch to C or Java for that).
973 Lua can still be used to describe platforms
974 - Removed Java kernel plug-ins.
975 Will be reintroduced after the ongoing major internals reorg.
977 - the following functions were removed.
978 They were too specific and should be reimplemented in a generic
979 way, with filter function.
980 - MSG_task_listen_from_host
981 - MSG_mailbox_get_count_host_waiting_tasks
982 - MSG_mailbox_put_with_timeout was removed.
983 Please use MSG_task_send_with_timeout instead.
985 - the SD_application_reinit function was removed. It has been a noop for a while.
986 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
987 replaced by a more flexible mechanism.
988 - The following functions thus do not exist anymore
989 - SD_workstation_get_access_mode
990 - SD_workstation_set_access_mode
991 - SD_workstation_get_current_task
992 - Basic estimation functions have been removed but can easily be replaced
993 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
994 - SD_workstation_get_computation_time => amount / sg_host_speed()
996 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
997 Use VM.getSpeed()*load/100 for the legacy behavior.
999 - option enable_tracing was removed. It was not doing anything for a while.
1000 - In the ModelChecker:
1001 - the model-checker now ptraces the model-checked process which means
1002 you cannot use a debugger on the latter anymore (we might make this
1003 optional in the feature);
1004 - removed soft-dirty page tracking;
1005 - remove model-checked side snapshot management,
1006 MC_snapshot() and MC_compare_snapshot();
1007 - keep the MC_cut() function as a stub (it was not really working
1008 in the previous release).
1011 * All options are consistently in kebab-case. Old names are kept as alias.
1014 * Switch to platform v4 format.
1015 - Rename from 'power' to 'speed' the attributes describing the amount of
1016 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1017 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1018 - In <host> and <link>, attributes availability and state are gone.
1019 It was redundant with state and availability traces, and with peak values.
1020 - In <cluster>, attributes availability_file and state_file are gone.
1021 It was too complex and unused.
1022 - Kill <gpu>. Was not doing anything.
1023 - The DOCTYPE points to the right URL:
1024 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1025 (the file at this address now documents the changelog since its v1)
1026 - A warning is emitted for unit-less values (they are still accepted).
1027 - speed. Default: 'f' or 'flops'. Also defined:
1028 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1029 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1030 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1031 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1032 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1033 - latency. Default: 's' second. Also defined:
1034 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1036 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1037 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1040 * s4u::Host is now the preferred public interface to the Host features.
1041 sg_host_* functions are C bindings to the exact same behavior
1042 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1045 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1046 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1047 They were deprecated since a few years, and probably did not even compile anymore.
1050 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1051 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1052 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1053 encouraged to update their codes. A list of the modified functions can be found at the end of
1054 include/simgrid/simdag.h
1057 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1058 the lambda or closure passed as a parameter will run in kernel mode.
1059 All the callback functions should be rewritten to that interface at some point.
1062 * Reorganizing and cleaning the internals all around the place.
1065 * Remove old default barrier/bcast buggy algorithms (see #18407)
1066 * Various bug fixes to handle more codes
1067 * Remove the need for the --foreground option of smpirun (it is still
1068 accepted for backward compatibility).
1071 * Kill the setset data container: MC don't use it anymore.
1072 * Kill the queue data container: it made more sense with GRAS.
1073 * Kill the xbt_peer_t data type: it's useless without GRAS.
1074 * Kill rm_cb feature of config sets: it was never useful.
1075 * Kill graphxml parsing feature. It was not used.
1076 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1078 - xbt_dynar_sort_strings(), when the content is char*
1079 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1080 They throw exceptions on invalid input;
1081 * C++ support for declaring CLI flags (simgrid::config::Flag);
1082 * class for abstracting different signal backends (simgrid::xbt::signal).
1083 with no external dependencies (we need very simple signals).
1086 * refactoring and cleanup of the code;
1087 * ongoing process to cleanly separate the model-checking algorithms
1088 from the code model-checking support.
1090 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1092 SimGrid (3.12) stable; urgency=low
1094 The Facelift Release.
1097 * Require g++ v4.7 at least to not speak prehistorical C++.
1098 * Require Boost 1.48 (for signal2 component).
1099 * Java must be version 7 at least when activated.
1100 * Builds on Windows again (including Java bindings).
1101 * Tracing is now always enabled (no way to turn it out)
1102 * Remove GTNetS. It was not working anyway.
1103 * Various cleanups in the cmake scripts.
1104 * Move headers around to sort them out on installed systems:
1105 - instr/instr.h -> simgrid/instr.h
1106 - instr/jedule/* -> simgrid/jedule
1107 - simdag/datatypes.h was removed
1108 - simdag/simdag.h -> simgrid/simdag.h
1109 - msg/datatypes.h was removed
1110 - msg/msg.h -> simgrid/msg.h
1113 * Interface improvement:
1114 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1115 - Sanitize the interface in MSG_task_ module:
1116 - Merge two functions that were close enough but misleading:
1117 set_compute_duration(t) -> set_flops_amount(t)
1118 get_remaining_computation(t) -> get_flops_amount(t)
1119 - set_data_size(t) -> set_bytes_amount(t)
1120 get_data_size(t) -> get_bytes_amount(t)
1121 - Massive cleanups in the functions related to the energy
1122 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1123 - New: MSG_host_get_pstate()
1125 - msg/energy/onoff: switching hosts on and off
1127 * Interface improvement:
1128 - Rename Host.isAvail() to Host.isOn()
1129 - Rename Process.currentProcess() to Process.getCurrentProcess()
1130 - Rename Task.setDataSize() to Task.setBytesAmount()
1131 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1133 - #18874: Actually allows the GC to reclaim tasks
1137 - SIMIX_process_throw: raises an exception in a remote process
1138 * Refactoring: Separate sub-modules
1139 - libsmx: the public interface, as libc in a real system
1140 - popping: the strange dance that converts a user request into a kernel handling
1141 - smx_context_*: the virtualization mechanisms that embed the user code
1142 - smx_*: the handling of each simcalls
1143 * Interface cleanup:
1144 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1145 * Rename smx_action_t into smx_synchro_t, making explicit that these
1146 things are used to synchronize processes with their environment.
1147 For example, a communication is a sort of synchronization involving
1148 the communicating processes (that may block until the exchange) and
1149 the platform. The same can be said from computations, etc.
1151 - #18888: segfault when a process ends before its kill_time
1155 - 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.
1156 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1157 - MPI_Type_set_name, MPI_Type_get_name
1158 - MPI_*_c2f and MPI_*_f2c functions
1159 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1160 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1161 - Activate a lot of new tests from the mpich 3 testsuite
1163 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1164 - InfiniBand network model added: Based on the works of Jerome Vienne
1165 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1166 - When smpi/display_timing is set, also display global simulation time and application times
1167 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1168 * Collective communications
1169 - 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
1170 - 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/).
1171 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1173 - Replay now uses algorithms from wanted collective selector
1174 - Replay can be used with SMP-aware algorithms
1175 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1176 - 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.
1178 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1179 - Correctly use loopback on fat-tree clusters
1180 - Asynchronous small messages shouldn't trigger deadlocks anymore
1181 * Energy/DVFS cleanup and improvement
1182 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1183 - new: smpi_get_host_pstate()
1186 - "Full" network optimization flag was broken since Surf++
1187 - Better handling of precision flags in maxmin
1188 - Fix bug causing sometimes "Impossible" errors
1189 - Properly pass cluster properties to included hosts
1190 * Improvement of the Energy plugin.
1191 - Always update the consumption before returning that value
1192 - New property: watt_off to denote the dissipation when the host is off
1193 - New functions getWattMinAt and getWattMaxAt to retrieve the
1194 dissipation of pstates that we are not currently at.
1195 * Java: class NetworkLink renamed to Link
1196 * New function: simcall_process_get_kill_time()
1197 * Massive rename s/workstation/host/
1198 - That's intrusive, but that's good for the project consistency. Sorry.
1199 - Change config option "workstation/model" into "host/model"
1202 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1203 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1204 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1205 - xbt_dirname and xbt_basename for non-POSIX systems
1207 * The model checker now runs as a separate process.
1208 * The model checker runs must now be launched with the new simgrid-mc program.
1209 * Record/Replay: the MC can display a textual representation of a path in the
1210 execution graph. It can then be replayed outside of the model checker.
1212 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1214 SimGrid (3.11) stable; urgency=low
1219 * Normalizing pointers addresses tool for better diff between logs
1222 * Add cloud examples using new VMs
1223 - examples/msg/cloud/two_tasks_vm.tesh
1224 - examples/msg/cloud/simple_vm.tesh
1225 - examples/java/cloud/cloud.tesh
1226 - examples/java/cloud/migration/migration.tesh
1227 * Add java surf examples:
1228 - examples/java/surfPlugin/surf_plugin.tesh
1229 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1230 - examples/java/surfCpuModel/surf_cpu_model.tesh
1231 * Add SMPI+MSG example:
1232 - examples/smpi/smpi_msg_masterslave/
1237 - msg task destroy cancel test
1238 - msg_host on/off test
1239 * Move all tests in testsuite to teshsuite (adding tesh files)
1240 * Restructure teshsuites
1241 - one folder for each kind of test
1242 * Restructure AddTests.cmake
1244 - structure the order of tests (with sections)
1247 * Add virtual machine
1248 - creation of a VM on a PM
1249 - migration of a VM from a PM to another PM
1251 - MSG_process_join(msg_process_t process, double timeout)
1252 - msg_bar_t MSG_barrier_init(unsigned int count)
1253 - int MSG_barrier_wait(msg_bar_t barrier)
1254 - void MSG_barrier_destroy(msg_bar_t barrier)
1255 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1256 * New option "msg/debug_multiple_use" to help debugging when a task is used
1259 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1261 - Increase code coverage in test suites
1265 * Protect context stacks against stack overflow. The number of protected memory
1266 pages allocated on the top of each stack (1 by default) can be configured
1267 with the new command line option --cfg=contexts/guard_size:<value>.
1268 * Simcalls are now generated by a python script that
1269 - generates files included by SimGrid
1270 - checks that all the functions exist, and proposes prototypes
1272 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1273 host_set_data, host_get_data
1275 - simcall_process_join(smx_process_t process, double timeout)
1276 * Fix bug where sleeping processing could not be suspended.
1279 * Translate surf models from C to C++
1280 - Generic classes for all models: Model, Resource, Action
1281 - A generic interface for each kind of model (CPU, Network, Storage
1282 Workstation, WorkstationVM)
1284 * Translate surf routings from C to C++
1285 * Add callbacks using sigc++ or boost::signals2
1286 - Add callback functions for resource creation/destruction
1287 - Add callback functions for action state change
1288 - Handle Energy as a plugin
1289 * Replace swag by boost::intrusive
1290 * Add new routing models for clusters. For documentation, see
1291 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1292 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1293 parameters for cluster tag
1294 - Fat trees, with topology="FAT_TREE" and
1295 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1296 - see examples/platforms/torus_cluster.xml and
1297 examples/platforms/fat_tree_cluster.xml
1298 * More documentation
1301 * Hostfiles support host:nb_processes construct to deploy several processes on
1303 * Collective communication algorithms should not crash if used with
1304 improper number of nodes and report the error.
1305 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1306 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1307 MPI_Dims_create, MPI_Cart_sub are supported.
1308 * New interface to use SMPI programmatically (still depends on MSG for
1309 some parts, see examples/smpi/smpi_msg_masterslave):
1310 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1314 * Global variables privatization in MPI executables is now performed at runtime
1315 with the option smpi/privatize_global_variables (default:no).
1316 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1317 dynamic libraries loaded by the application are not privatized (static
1318 linking with these libraries is advised in this case)
1321 * Options defined in XML work correctly now.
1324 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1325 copied into simgrid.jar or not (ON by default). Use this option if you want
1326 to reduce the size of the installed simgrid.jar, *and* the native libraries
1327 are kept installed elsewhere.
1328 * Surf binding with SWIG (code generated in maintainer mode only):
1329 - plugin to handle callbacks
1330 - CPU model only for the moment
1333 * Supernovae build mode is definitively removed. It was used to improve
1334 inlining and inter-module optimizations. It is nowadays superseded by
1335 link-time optimizations commonly available in compilers.
1336 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1337 * Add boost dependency for surf++
1338 * Add new macro for tests
1339 - ADD_TESH(name <tesh_args>)
1340 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1344 - xbt_bar_t XBT_barrier_init(unsigned int count)
1345 - int XBT_barrier_wait(xbt_bar_t barrier)
1346 - void XBT_barrier_destroy(xbt_bar_t barrier)
1347 * Make the xbt_os_time module public
1349 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1351 SimGrid (3.10) stable; urgency=low
1353 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1356 * Reintegrate Java to the main archive as desynchronizing these
1357 package is not acceptable anymore (Java is now considered stable)
1358 * Add explicit synchronization facilities through semaphores
1359 * Bug fix: Task.setDataSize() only changed the C world, not the value
1360 cached in the Java world
1363 * Dramatically change the way files are handled. API and internals changed, but
1364 this part of MSG was not considered as production grade either.
1365 * Add explicit synchronization facilities through semaphores
1366 * Add a new function MSG_host_get_process_list()
1367 * Preliminary DVFS support (see examples/msg/energy/ for details)
1370 * SMPI is now included directly in the libsimgrid as the windows
1371 linker doesn't force us on splitting it anymore.
1372 * Improvements of the SMPI replay tool:
1373 - Most of the collective communications are now rooted in the same process as
1374 in the original application.
1375 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1376 used until now). Multiple data types can now be used in a trace.
1377 - The replay tool now supports traces produce either by TAU or a modified
1379 - Bug Fix: the compute part of the reduce action is now taken into account.
1380 - Gatherv collective is now supported
1381 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1382 of smpirun outputs time independent traces for the current run. One file
1383 is created per process. If too many processes are simulated, this behavior
1384 can be changed to one file for all processes by using the
1385 tracing/smpi/format/ti_one_file flag
1386 * smpirun generates the host file if needed (with given host count and platform)
1387 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1388 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1389 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1390 --cfg=smpi/coll_selector:(mpich/ompi)
1391 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1393 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1394 happening inside a collective SMPI call.
1395 * Fix the behavior of complex data types handling.
1396 * Make MPI_Wtime another synchronization point to take computations into
1398 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1399 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1400 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1401 (no privatization of global variables yet)
1402 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1405 * Verification of liveness properties is now available for SMPI applications
1406 (in addition to MSG applications)
1407 * Bugged examples using SMPI in examples/smpi/mc/
1408 * Add --cfg=model-check/visited option. Allows the verification of infinite
1409 programs. Detection of loops in the execution thanks to the system state
1410 comparison and reduction of the state space to explore. Can be combined with
1411 DPOR for safety properties.
1414 * Allow to change SimGrid configuration (see --help) within the code
1415 thanks to SD_config() as it can be done in MSG.
1416 * Add a new function SD_task_set_amount() upon user request.
1419 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1420 * Remove rule based routing (no more PCRE dependency)
1421 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1422 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1423 * Add a loopback_bw and loopback_lat options to cluster tag.
1424 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1425 routing. Now peers are encapsulated in an AS and have their own private
1426 router but this is transparent.
1429 * Our own implementation of getline is renamed xbt_getline, and gets
1430 used even if the OS provide a getline(). This should reduce the
1431 configuration complexity by using the same code on all platforms.
1432 * New type: xbt_cfg_elm_boolean.
1433 * Allow to use yes/no for boolean configuration options in the command line.
1434 * Allow to disable SimGrid cleanups at exit from command line option.
1435 There are situations where one may want a simulation to end with an exit.
1436 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1437 annoying when scripting around the simulator. Adding a
1438 --cfg=clean_atexit:no allows to circumvent this issue.
1441 * Lots of memory leaks were corrected in this release.
1442 There is no known memory leaks anymore, in all of our 600+ tests.
1443 * New command line option --version, to get SimGrid version information.
1444 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1446 * Supernovae builds are deprecated, and expected to be removed in the next
1449 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1451 SimGrid (3.9) stable; urgency=low
1453 The Grasgory release: GRAS is really dead now.
1455 * Complete overhaul of the internal host structures scheme.
1458 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1459 was considered as experimental and badly maintained since then.
1460 * Keeping it was thus a trap to our potential users, that could take
1461 it instead of MSG or SMPI by mistake despite is pity state.
1462 * GRAS seems to have very few users (if any), and no one volunteered
1463 to maintain it further. It also induces a lot of XBT code (for
1464 portability sake), that must be maintained too.
1465 * For all these reasons, we killed GRAS. If someone wants to revive it
1466 in the future, don't cry, our git history still remembers of GRAS.
1469 * Major overhaul. Merge our documentation again as time proved that
1470 splitting it was really not helping our users.
1471 * Further improve the developer documentation to help newcomers
1472 hacking on SimGrid itself. The user documentation (and in
1473 particular, the beginner documentation) is still in a sorry state.
1476 * Now works on Windows too!
1477 * Much more extensive test suite, from MPICH
1480 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1481 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1482 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1483 * Bug fix: let task be scheduled when the last dependency to be solved is
1484 a control dependency.
1485 * Remove SD_load_environment_script function.
1486 Use the C sg_platf function if you want to declare a platform
1490 * New function: MSG_process_get_number()
1491 * Old function documented: MSG_config()
1492 * Remove MSG_load_platform_script function
1493 Use the C sg_platf function if you want to declare a platform
1497 * Change the default value of the TCP_gamma constant (maximal size of TCP
1498 congestion window) to a more realistic 4MiB value. If you notice changes in
1499 your simulation results, you can fall back to the previous 20k tiny window
1500 by adding --cfg=network/TCP_gamma:20000 on command line.
1501 * (Hopefully) fix a bug wrt periodic availability and state traces
1502 * Bug fix: use default values at start when first event in availability/state
1503 trace is not at time 0.
1506 * remove the "new_" part of function name sg_platf_new_trace_connect
1507 (resulting in sg_platf_trace_connect), since it does not create
1511 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1512 use case where it's really mandatory, and maintaining it was a pain in
1514 * New: xbt_fifo_search(), search an item with a user-provided
1515 comparison function instead of dumb pointer comparison.
1518 * Fix the lua deployment:
1519 Use `simgrid.init_application()` before deployment instead of
1520 `simgrid.msg_register_application()` after.
1523 * Transfer the tracing files into the corresponding modules.
1525 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1527 SimGrid (3.8.1) stable; urgency=low
1529 The "we are told that some people want to also *install* the simgrid
1532 * Add missing file "tesh.1" to the archive.
1534 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1536 SimGrid (3.8) stable; urgency=low
1538 The Psssshiiiit release: SimGrid jumps into the Cloud.
1541 * Add an experimental interface to manipulate VMs. They are mainly
1542 process groups with very few intrinsic semantic, but they should
1543 allow you to build the semantic you want easily.
1544 * New function: MSG_host_set_property_value()
1545 * New function: MSG_process_on_exit(). To clean memory in all cases.
1546 * Bug fixes that made the host (and link) failures unusable.
1547 * Add a way to auto-restart process when the host in which they are
1548 executing comes back (ON_FAILURE="RESTART" on deployment file,
1549 MSG_process_auto_restart_set).
1550 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1551 please stop using the old ones, they are DEPRECATED.
1553 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1554 Please use MSG_init() instead. (reducing the amount of entry
1555 points in the library helps us).
1556 * Make it impossible to link against the wrong version of the lib
1557 * Deprecate MSG_clean(). No need to call it anymore.
1558 * Function MSG_get_host_number() is not deprecated anymore.
1561 * Split the doc into a user guide and a reference guide.
1562 * Start a developper guide to help people hacking on SimGrid.
1565 * Enable tracing by default. This modules rocks you should use it.
1566 * Remove option custom_flags. Now use environment variables CFLAGS
1568 * Use default cmake things to detect lua instead of home grown ones.
1569 * New option "enable_mallocators" to disable mallocators, for debugging
1570 purpose ("on" by default).
1573 * Bug fixes around the resource failures: don't let the processes
1574 survive the host they are running onto.
1575 * Add an interface to auto-restart processes when the host in which they are
1576 executing comes back.
1577 * Ensures that SIMIX_clean is called automatically. It's not part of
1578 the public interface anymore (bindings should be updated).
1581 * Bug fix for when SD_Simulate is called with a positive value: be careful
1582 when comparing doubles. Sometimes they are different for non significant
1584 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1585 parallel task whose initial work is distributed among host according
1586 to the Amdahl's law. Such tasks are created with a parameter alpha
1587 that corresponds to the non-parallelizable part of the computation.
1588 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1589 between two sets of workstations assuming a 1D block distribution (each
1590 workstation owns a similar share of data) on both sides.
1592 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1593 Data redistribution will be automatically scheduled once parent and child
1594 are both scheduled. The filling of computation_amount and
1595 communication_amount structures is now done seamlessly thanks to the chosen
1597 * New function SD_workstation_dump to display various information
1598 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1599 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1600 bandwidth on the route onto which the task is scheduled and the amount of
1602 To divide the nominal bandwidth by 2, the rate then has to be:
1603 rate = bandwidth/(2*amount)
1604 * Compute tasks that have failed can now be rescheduled and executed again
1605 (from their beginning)
1606 * Increasing source code coverage (src/simdag is now covered at 95.8%
1610 * Re-implement time-independent trace replay using SMPI (at the
1611 smpi_smp_* level) instead of MSG. This should replace
1612 examples/msg/actions/actions.c
1613 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1614 hindexed and structs)
1615 * Implement the exchange of non-contiguous data.
1616 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1617 * Correct behavior of smpi/sender_gap and set its default value to 0
1618 * Add option to asynchronously send small messages to allow better
1619 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1620 specifies the size in bytes under which messages will be asynchronously sent.
1621 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1622 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1625 * New C interface to define a platform: XML is now optional.
1626 For more info, please check include/simgrid/platf.h.
1627 * New interface to define random platforms from the C:
1628 For more info, please check include/simgrid/platf_generator.h and
1629 examples/msg/masterslave/masterslave_platfgen.c
1630 * Export a sg_cmdline dynar containing all the arguments we got from
1634 * Two new tracing options for adding comments to trace file so you
1635 can track your experiments (see --help-tracing for details).
1636 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1637 * Adding the SimGrid version that generated the trace file as a comment.
1638 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1639 * Fix to avoid key clashes on Paje links
1640 * Other minor fixes related to the Paje specification
1643 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1644 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1645 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1646 * Greatly improve the robustness of mmalloc to user errors (such as
1647 using an area after freeing it, or freeing it twice)
1649 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1651 SimGrid-java (3.8.1) stable; urgency=low
1653 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1654 Warning: all methods are visible, but only some of them are
1655 implemented so far. Check the source (src/jtrace.c)
1656 for further information.
1657 * New module: org.simgrid.msg.File (SimGrid File management functions)
1658 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1659 * Change the meaning of Process.restart: now restart the process from
1660 the begining, like MSG_process_restart in C.
1661 * Add Process.setAutoRestart: handling of process restart when failed
1663 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1664 you to retrieve the properties of the processes/hosts
1665 * Deprecate Msg.clean(): you can just forget about it now.
1666 * New function Process.getCount(), that only works when compiling
1667 with the not yet released version 3.9 of the C library.
1669 * New context factory based on Coroutines. It mandates a modified JVM
1670 but then, the simulations run about five times faster, and there is
1671 no limit to the amount of processes (beside of the available memory).
1673 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1675 SimGrid (3.7.1) stable; urgency=low
1678 * Restore the prototype of MSG_process_create_with_environment() to
1679 the pre-3.7 situation by removing the kill_time argument.
1680 * Add a MSG_process_set_kill_time() function instead.
1683 * Fix weird behaviors when dealing with parallel tasks.
1686 * Simgrid is now built as a dll.
1687 * Simgrid-java now works on Windows.
1688 * Simgrid-Java is now included into Windows package.
1691 * First pre-build package for MacOSX.
1694 * Fix compilation when using MSG_USE_DEPRECATED.
1695 * Fix some compilation issues on Macs and Windows.
1696 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1697 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1699 We discovered that the Lua console is broken, but we are missing the
1700 manpower to fix it right now. The problem existed in 3.7 too, so we
1701 are not blocking the release for that. Sorry if you depended on this
1702 feature, any help would be really welcome.
1704 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1707 SimGrid-java (3.7.1) stable; urgency=low
1709 The "Java aint got to be bloated and slow" release
1712 * Various internal cleanups and performance improvement
1713 Simulations are expected to run up to twice faster or so
1714 * Make Process.kill(process) an instance method, not a static one
1715 * User processes are not java.lang.Thread subclasses.
1716 This breaks the compatibility (sorry), but previous API was
1717 brain-dead, making it impossible to have non-trivial
1718 initializations in the process constructor.
1719 * Require a full constructor per Process sub-class.
1720 Kinda breaks the compatibility (sorry), but this allows a much more
1721 efficient way to launch the processes at simulation startup.
1722 * Do not embeed our version of semaphores, java 1.5 can be considered
1723 as sufficiently prevalent for us to not dupplicate its features.
1727 * Add examples for almost every part of the API
1728 We spotted and fixed a lot of bugs in the process
1730 * New module: asynchronous communication API
1731 * New function: Process.sleep()
1732 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1733 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1734 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1736 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1738 SimGrid (3.7) stable; urgency=low
1740 The "spring cleanups (before the next Big Project kicks in)" release.
1743 * We can specify the SMPI latency/bandwidth factor with command line
1744 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1745 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1746 You can also use the "config tag" from platform file by adding this line
1747 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1748 (see "example/platforms/tag_config.xml" to use "config tag").
1749 Note that the command line supersedes the platform file configuration.
1750 * Change the correction factors used in LMM model, according to
1751 the latest experiments described in INRIA RR-7821.
1752 Accuracy should be improved this way.
1753 * Use the partial invalidation optimization by default for the
1754 network too. Should produce the exact same results, only faster.
1755 * Major cleanup in surf to merge models and split some optimization
1756 mechanisms from the core of the models. As a result you can now
1757 specify which model to use (e.g., --cfg=network/model:LV08
1758 --cfg=cpu/model:Cas01) and which optimization mode to use
1759 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1760 Incompatible combinations should err at initialization. See
1761 --help-models for the list of all models and optimization modes.
1762 * The CLM03 workstation model was dropped for simplicity because it
1763 used the deprecated CM02 network model. Use default instead.
1764 * Rename the TCP_gamma configuration option to network/TCP_gamma
1765 * Rename the coordinates configuration option to
1766 network/coordinates, and document it
1767 * Use now crosstraffic keyword instead of the terribly misleading
1768 fullduplex keyword. It is activated by default now in the current
1769 default model, use --cfg=network/crosstraffic:0 to turn it off.
1770 * Ongoing refactoring the model parsing to make XML files optional
1771 See include/simgrid/platf.h for details (still to be completed)
1774 * Major overhaul of the documentation. Almost instructive now :/
1775 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1776 functions and friends. This interface was considered as
1777 deprecated since over 2 years, it's time to inform our users that it is.
1778 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1779 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1780 use these (crappy) functions in your code.
1781 These functions will be removed soon. Stop using them now.
1782 * Deprecate MSG_get_host_{table,number}
1783 Implement MSG_hosts_as_dynar() instead.
1784 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1785 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1788 * Stabilize the parallel execution mode of user contexts
1789 * Introduce configuration variables to control parallel execution:
1790 - contexts/synchro: Synchronization mode to use when running
1791 contexts in parallel (either futex, posix or busy_wait)
1792 - contexts/parallel_threshold: Minimal number of user contexts
1793 that must be part of a scheduling round to switch to parallel
1794 execution mode (raw contexts only)
1795 * Fix bugs that prevented to use suspend/resume along with
1796 synchronization structures.
1797 * Fix bugs in process termination that lead to invalid memory access
1798 in very specific conditions.
1801 * Introduce a parallel mode for the models (controlled by surf/nthreads
1802 configuration item). In our tests, running the models in parallel
1803 never lead to any speedups because they are so fast that the gain
1804 of computing each model in parallel does not amortizes the
1805 synchronization costs, even when ultra fast futexes are used.
1806 This is released anyway because YMMV.
1809 * Performance boost by using a swag internally to compute the set of
1810 tasks that are finished and should constitute the return value of
1814 * Enable it by default now that it is considered rather stable.
1817 * Documentation of the tracing functions.
1818 * Performance gains when tracing categorized/uncategorized resource
1819 utilization by avoiding calls to get route when updating resource
1820 variables. LMM constraints are being used instead.
1821 * API changed to set task categories. Use MSG_task_set_category instead
1822 of TRACE_msg_set_task_category, and SD_task_set_category instead
1823 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1824 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1825 * Examples for link user variables added (see at examples/msg/tracing/)
1826 * Deprecated function TRACE_msg_set_process_category completely removed
1827 * Trace header updated according to the latest Paje file format
1828 * Tracing network lazy updates, no longer obligate users to use full updates
1829 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1830 * Experimental: let user code declare/set/push/pop application states for hosts
1831 * API changed to allow the manual creation of graph configuration files
1832 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1835 * Improve the API of Lua MSG bindings, using the Lua spirit.
1836 * Each simulated process now lives in its own Lua world (globals are
1837 automatically duplicated). It helps writing simulators. Will allow
1838 to run Splay programs within SimGrid in the future.
1839 * Add a Chord example in Lua, equivalent to the MSG one.
1842 * Start the implementation of a solution to express temporal
1843 properties, not only local assertions. This is still an
1844 experimental work in progress, stay clear from it to be safe.
1848 - Add new runtime parameters --help-logs and --help-log-categories
1849 to display information about supported logging parameters and
1851 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1852 don't exists anymore.
1853 * Mallocators: allow value NULL for the reset function.
1855 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1856 create homogeneous dictionaries, where all the elements share the
1857 same free function. Non homogeneous dictionaries will be
1858 deprecated in the next release.
1859 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1860 - Multi-level dictionaries are deprecated.
1862 - new function xbt_dynar_search_or_negative() that is useful when
1863 you have less than 2 million elements in your dynar and don't
1864 want of the extra complexity of catching exceptions when the
1865 element is not found.
1867 - Make xbt_os_thread module (for thread portability) public.
1868 Documentation is still to come, sorry.
1870 - Cleanups and simplifications to make it maintainable again.
1871 - Exotic features (such as memalign and valloc) were removed.
1872 - The metadata were extended and improved so that the
1873 model-checker becomes able to explore and inspect the heaps.
1874 - This may induce a performance drop when enable_model-checking is
1875 ON in cmake (even if it's not used in the simulation), but it is
1876 necessary at this point to get MC working.
1878 Turn model-checking OFF if simulation performance matters to you.
1879 Not enabling it at runtime is not enough, disable it in cmake.
1881 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1884 SimGrid (3.6.2) stable; urgency=low
1886 The "Not coding new stuff allows to polish old things" release.
1889 * New bindings to the NS3 packet level simulator (experimental)
1890 * Use the raw (efficient) execution contexts instead of the sysv
1891 (portable) ones when possible.
1892 * libpcre is now mandatory in any cases since not using it led to
1893 severe performance loss and possibly other issues
1894 * Update the XML platforms:
1895 - G5K: include the latest machine in Nancy
1896 - GridPP and LCG: new platforms
1897 * Documentation was partially updated, at least (more to come)
1899 Bug fixes, cosmetics and small improvements
1900 * Free terminated processes before the end of the simulation to avoid
1901 exhausting the memory of users having very dynamic amount of
1903 * Bug fix and cosmetics about canceling non-running tasks
1904 * Bug fix about the dot loader's issues when using libcgraph
1907 * Create an installer for windows with nsis (amd64 and win32)
1908 - Add an hello world project to illustrate simgrid project creation.
1909 - Embed libpcre into the Simgrid installer to avoid
1910 its compilation burden
1911 * The raw execution contexts should work on Apple now
1912 * Port to Windows 64 bits
1913 - Sysv contexts now have an implementation for this arch
1914 - GRAS communication features now support this arch
1915 * Drop support for borland compiler on windows
1916 - this code was not maintained, and we kinda depend on gcc nowadays
1917 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1918 * Fix portability issue on unstable ubuntu: linker became picky on
1921 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1923 SimGrid (3.6.1) stable; urgency=low
1925 The "Oops, we broke Macs too" release
1928 * Fixed contexts detection so that raw ones are used when possible
1929 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1930 leads to a strange error, with user code segfaulting sometimes when
1931 the generated code is not perfectly aligned (which is not
1932 controllable from the user side, depends on the amount of code)
1935 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1937 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1939 SimGrid-java (3.6) unstable; urgency=low
1942 * Split of every thing from simgrid v3.5 into a separate package.
1944 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1946 SimGrid (3.6) stable; urgency=medium
1948 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1951 * Bindings now constitute their own package, separated from the main one.
1952 Rationale: reduce our maintenance nightmare by reducing the module coupling
1953 They will soon be released on their own on gforge.
1955 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1956 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1958 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1959 * It's not quite deprecated for now because we have no replacement,
1960 but it may soon become the case.
1963 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1964 * Fortran: New user-level cache variable to store the rank of the running
1965 process. This improves performance by an order of magnitude.
1966 * C: New coccinelle script to automatically locate and modify global and
1967 local static variables.
1968 * Improved SMPI network model with a sender-side gap to account for multiple
1972 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1973 only say if a communication is finished, no matter whether it succeeded or
1974 failed. You can call MSG_comm_get_status() to know the status of a finished
1976 * New function MSG_task_dsend() to send a task and detach it. When a
1977 communication is detached, you are never notified of its success or failure
1978 and the memory is released automatically once it is finished. This function
1979 is useful when you don't care about the end nor the success of a
1981 * Change the prototypes of action replay. Sorry for inconvenience,
1982 but this is really more efficient this way (and to adapt your code,
1983 you just have to fix the initialization, that shouldn't be too long)
1984 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1985 ever managed to do anything useful with it.
1986 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1987 takes a dynar of communications. It returns immediately and gives the
1988 index of a finished communication (if any).
1989 * New example: a basic implementation of the Chord P2P algorithm.
1992 * New model for multi-core CPUs. You can now use the core attribute to
1993 precise the number of cores of a host. This is a basic model. Every
1994 process running on the host receives at most the power provided in
1995 the DTD (throughput<=power). Total throughput of process cannot exceed
1997 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1998 router linked by an up-link and a down-link (possibly asymmetrical).
1999 This kind of pattern allows to easily build last-mile model style platforms.
2000 Aggregating such patterns in a rule-based AS is thus the technique of
2001 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2002 * New model for Vivaldi routing. We transformed the Vivaldi network model
2003 into a Vivaldi routing model (based on the rule-based model). This allows to
2004 combine Vivaldi based latencies with last-mile platforms.
2007 * Added a check for NaN of IEEE754 infinite in the double entries of
2009 * Introduce a new context factory "raw", highly inspirated from the
2010 ucontext factory, but using manually crafted functions in assembly to
2011 do the work in an efficient manner.
2012 * Allow to change the used context factory at run time, not only at
2013 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2014 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2015 contexts at runtime (only with raw contexts or ucontexts).
2016 * Completely rewrote this module to allow parallel execution of user
2017 processes. Use --cfg=contexts/nthreads:N to execute user processes
2018 with N parallel threads (the default is 1, meaning no parallelism).
2019 * Allow to decide dynamically between sequential and parallel modes.
2020 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2021 processes in parallel only when their number is greater than or equal to P
2023 * Added a check for NaN of IEEE754 infinite in the double entries of
2027 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2028 won't output the state of processes when interrupted with Ctrl-C
2029 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2030 NULL-terminated array. This may solve backward compatibility issues
2031 due to the change to return type of SD_simulate. See also:
2032 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2033 * Add new macros with variable number of arguments.
2034 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2035 - in xbt/asserts.h: xbt_assert
2036 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2037 - in xbt/ex.h: THROWF and RETHROWF.
2038 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2040 * Change xbt_die() to accept a format string with arguments, just like printf.
2041 * New data structure: xbt_lib_t, like a dict but more general and with better
2045 * New configuration options
2046 Options triva/categorized and triva/uncategorized can be used to generate
2047 graph configuration files for Triva visualization tool.
2048 * Configuration option tracing/platform is renamed to tracing/categorized
2049 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2050 * New TRACE_user_link_variable interface:
2051 User provides the name of the link and the tracing variable to attach to it
2052 * the declaration of tracing categories must be done after the environment
2054 * simpler tracing interface, just one way to declare categories
2055 TRACE_category or TRACE_category_with_color, it is up to you
2056 * links in the trace file are again identified by their names
2057 * trace contains the full platform hierarchy exactly as declared using the ASes
2058 * Options tracing/msg/[task|process]:1 groups the process by hosts
2059 for both cases, tasks and processes must have names that are unique during the simulation
2060 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2061 * The experimental option tracing/msg/volume is deprecated
2062 its functionality may be reincorporated if needed
2064 The tracing generates a trace file with unordered timestamped events,
2065 because of the way the core simulator (surf) works. A script available
2066 at the tools directory (fix-paje-trace.sh) can be used to put the events
2067 in order. We have changed the tracing so it can generate ordered timestamped
2068 events in the final trace, but depending on the simulator (and how much time
2069 is simulated) that can lead to a huge memory utilization. It is deactivated
2070 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2072 Build Infrastructure
2073 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2074 - We are trying hard to keep the API stable, but it may happen that
2075 some things change (we're a research project after all, not a
2076 nuclear plant operating system). If such things should happen, you
2077 could rely on that macro to adapt.
2078 - current value: 30600 for 3.06.00, aka 3.6
2079 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2080 a number that can be compared with SIMGRID_VERSION.
2081 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2082 assertions and verbose/debug logging events are disabled at compile time.
2084 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2086 SimGrid (3.5) stable; urgency=medium
2089 * New feature to any SimGrid-based simulator: Model-Checking
2090 Check SIN#1 for more details.
2093 * New Model SMPI (three-interval linear regression for correction factors)
2094 See RR-7426, available at http://hal.inria.fr/inria-00527150
2095 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2096 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2097 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2098 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2099 * See also src/smpi/README
2104 - Tracing API changes: TRACE_start and TRACE_end should not be called
2105 by user-code. They are automatically called by simulators created
2106 with SimDAG, MSG and SMPI if the toolkit is compiled with
2107 tracing_enabled=ON. Categories declaration and utilization remain the
2108 same for MSG and SimDag.
2109 - A function was added to the tracing API to declare categories with
2111 - TRACE_category_with_color (char *category, char *color)
2112 where color must be in the following format
2113 "%f %f %f", red, green, blue
2114 and red, green, blue are float values in the interval [0, 1]
2115 - User can specify NULL as color parameter, or continue calling
2116 TRACE_category (cat)
2117 On that case, the tracing system will define random colors
2118 - The following command-line options are supported:
2119 --cfg=tracing/filename:msg.trace
2120 --cfg=tracing:1 (activate tracing, needed to use others)
2121 --cfg=tracing/platform:1 (categorized resource use)
2122 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2123 --cfg=tracing/msg/task:1 (task creation)
2124 --cfg=tracing/msg/process:1 (process creation, migration)
2125 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2126 --cfg=tracing/smpi:1 (SMPI interface tracing)
2127 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2128 - examples of examples/msg/tracing updated
2130 - DAXLoader and DOTLoader functions can generate tasks with categories
2131 - A new function to attribute a category to SD tasks:
2132 TRACE_sd_set_task_category (SD_task_t task, char *category)
2133 * Tracing the MPI interface implemented by SMPI
2134 - Collective operations are traced with states
2135 - Point-to-Point operations are traced with states/links
2136 - Tracing activated by a parameter "-trace filename" passed
2137 to smpirun during execution (considering that simgrid
2138 is compiled with tracing enabled)
2139 - To run the simulation with gdb, the simulator
2140 accepts --cfg=tracing/smpi:1 to trace SMPI
2141 - tesh files to check if smpi tracing is ok
2142 - See examples/smpi/NAS/DT-trace
2143 * GTNetS tracing re-worked
2144 - adaptation to the tracing system of GTNets to cope
2145 with modifications regarding the fullduplex mode
2146 - new tesh files to check if gtnets tracing is ok
2149 * Asynchronous communications through the functions:
2150 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2151 * New function: MSG_load_platform_script()
2152 to make possible using a lua script instead of XML files to set up platforms
2153 * New function: MSG_set_function
2154 to associate functions to processes, used when bypassing the parser
2155 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2157 Platforms: Add some more examples in examples/platforms
2158 * Grid'5000: see www.grid5000.fr
2159 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2162 * Change the XML format. This is a very important modification. SimGrid
2163 3.5 introduces a new hierarchical format based on the notion of
2164 Autonomous Systems. Compatibility with old format is ensured
2165 through the perl script provided in the install bin directory
2166 bin/simgrid_update_xml.
2167 It is now possible to build platforms with specific routing
2168 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2169 connect several platforms together. We will try to provide soon
2170 set of realistic platforms exploiting these properties (have a
2171 look at examples/platforms/ for the moment).
2172 * Take the opportunity of the XML format change to be a good XML citizen:
2173 rename link:ctn to link_ctn and similar changes (also dealt with by
2175 * Add a new routing scheme (rule-based) using regular expressions. It
2176 enables to have an extremely low memory footprint when the
2177 underlying routing is simple and can be compactly described. You
2178 need to have libpcre4-dev (perl regular expressions) installed if
2179 you want to use this routing scheme.
2180 * Revive the cluster TAG and allow to easily and efficiently (both in
2181 term of memory and speed) connect clusters together. Have a look
2182 at teshsuite/simdag/platforms/ to see how this can be done. With
2183 this tag, you can create clusters with thousands of tasks at no
2184 cost (have a look at examples/platforms/).
2185 Note: clusters are implemented as ASes, so there is no need for an
2186 enclosing AS tag if you have only one cluster in your platform.
2187 * Add new generic functions in the public interface that allows the user
2188 to call SURF 'create_resource' methods from your code (same
2189 functionality as the XML bypass mechanism but with a much lighter
2191 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2192 that uses a piecewise linear approximation to produce better
2193 results when exchanging small messages.
2194 * Add a new parameter to handle correctly full duplex link and account
2195 for interferences between uplink and downlink communications
2196 (activate with --cfg=fullduplex:1).
2199 * Rename the SD_READY (all dependencies are satisfied and task is
2200 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2201 dependencies are satisfied) state.
2202 This prevents a confusion between the notion of "ready to schedule"
2203 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2204 simulated" (SD_RUNNABLE) used by the simulation kernel.
2205 * Change the way a task is considered as ready. Instead of removing
2206 dependencies when a task is done, a counter is decreased. This way,
2207 it is always possible to reach ancestors thanks to the
2208 SD_taks_get_parents function (even after the end of the simulation.)
2209 * Change the return type of SD_Simulate from (SD_task_t*) into
2210 xbt_dynar_t. This function was in handling a dynar internally and
2211 converted it into a NULL terminated array for historical reasons.
2212 * New function SD_dotload(char*) to load a DAG described in dot
2213 format. This loader and the corresponding examples require the
2214 installation of the graphviz library.
2215 * Fix a bug in the management of tasks of size 0 in the surf network
2216 models. This problem was only visible with SIMDAG and you should
2217 thus disregard results produced with earlier versions if you
2218 relied on this feature (some tasks were blocked because of this).
2219 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2220 even though your applications were doing only point-to-point
2221 communications and sequential computations. Now you can really use any
2222 model you want (of course, if you create real parallel tasks, which are
2223 not implemented in most models beside ptaskL07, this will abort).
2224 * Add an example that schedules a DAX on an heterogeneous platform
2225 using a Min-Min strategy.
2226 * New function SD_workstation_get_current_task() that returns the kind
2227 of task currently running on a workstation in the sequential access
2229 * Raise some warnings when unexecuted tasks remains at the end of the
2230 simulation. This is usually caused by cycles in the DAG.
2233 * New function: SIMIX_process_set_function() called by MSG_set_function
2234 * Change the underlying waiting queue in semaphores so that a process
2235 can wait on several of them simultaneously (as in waitany).
2236 * Fix the way to handle tokens in semaphores so that all access patterns
2237 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2238 * kill the dirty pimple SIMIX_message_sizes_output()
2239 Please use (proper) visualization instead
2242 * New data container: setset (set of sets of elements)
2243 * New module: mmalloc (mapped malloc, allowing to have several
2244 independent segments of malloc)
2245 * New function: xbt_dict_cursor_set_data()
2246 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2247 * New function: xbt_dynar_is_empty()
2248 * New function: xbt_fifo_get_last_item()
2249 * Fix xbt_dynar_shrink(): use the right element size.
2250 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2251 uninitialized areas during expand.
2252 * Fix semaphores: previous implementation was severely broken.
2253 * Use library init/fini functions for our initialization.
2254 - you can use logs and other feature as soon as you want in your
2255 code (even before the xbt_init / MSG_init)
2256 - xbt_exit is now a no-op and produce a warning when used.
2259 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2260 benefit from the latest improvement to the simulation kernel.
2261 * Kill measurement sockets for now. If you rely on them, sorry. This
2262 release is not for you. This feature will be reintroduced in the
2263 future, but we cannot delay the release any further.
2264 * New function: gras_msgtype_get_name().
2265 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2266 * Fix (at last) the pmm example: it should not randomly fail anymore.
2268 Build chain: bug fixes and overall polishing
2269 * Cmake is now stable enough. Hence, we killed the autotools.
2270 * Port to windows ( TM :)
2271 * Fix the 'make install' target.
2272 No need to use 'make install-simgrid' anymore
2273 * Introduce a 'make dist' target compiling a *source* archive
2274 'make package' compiles a binary archive
2275 * Compile java files only on need
2276 * Add --cd and --setenv command line options to tesh
2277 * Out of source builds are not fully supported yet, but we are close
2278 * Enable supernovae and optimization flags by default for our users
2281 * Add layer to set up environment directly from lua, without XML.
2282 * The effect of gras_stub_generator can be achieved through
2283 lua too (check examples/gras/console/ping_generator.lua)
2285 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2287 SimGrid (3.4.1) stable; urgency=low
2289 The "Polishing easter eggs is probably a good idea" release.
2290 This is a bug fixes release only.
2294 * Fix a bug preventing the tasks from begin garbage collected.
2297 * Fix a bug occuring when a host involved in a communication fails.
2298 This was not detected properly by the other peer involved in the
2299 communication. Now, it's reported as a network error.
2302 * Warn the user about loop dependencies in data flow of DAX files
2303 * Obey the control-flow dependencies of DAX files
2306 * Add option "enable_smpi" allowing to not compile SMPI.
2307 Probably useful for the (Mac) users experiencing a build error here
2308 * Improve the detection of lua5.1 and ruby1.8
2310 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2312 SimGrid (3.4) stable; urgency=low
2314 The "Easter in Cargese" release. Also known as (major changes):
2316 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2317 ~> bindings were greatly improved
2318 ~> new tracing infrastructure for better visualization introduced
2320 * the "Welcome to configury modernity" release.
2321 ~> we switched from autotools to cmake, and improved our cdash
2324 A more detailled list of changes follow (full detail in svn log).
2326 Java Bindings: Various Cleanups
2327 * (install java-gcj-compat-dev on debian-like to use them)
2328 * Remove put/get: no need to export deprecated interface in Java
2329 Use send/receive instead.
2330 * Cleanup the examples and add a README per directory
2331 * Remove example autoDestination (that's the only way to go now)
2332 * Remove example explicitDestination (was a plain copy of basic)
2333 * Make JniException a runtime exception, so that there is no need to
2334 declare the fact that you may encounter such a beast. I guess that
2335 nobody will ever want to survive such error.
2336 * Create specific errors for each MSG case of failure:
2337 host failure, transfer failure, timeout, task cancelled
2338 * Cleanup the exceptions that may get thrown by each function
2339 * Other internal cleanups in Java bindings. Performance still bad :/
2340 Ruby and Lua Bindings: create them
2341 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2342 * That's new and great, you should try them out.
2343 Same functionalities than Java bindings, only even less polished
2345 * Kill the useless "rate" argument of SD_task_get_execution_time()
2346 Everyone used to provide -1 as a value, it was not used, and the
2347 semantic of a possible use wasn't even clear.
2348 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2349 either as comm costs or compute costs to mean that there is no
2350 such thing for that specific task.
2351 * Add a SD_task_set_name() function
2352 * Fix SD_task_unschedule() on typed tasks
2353 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2354 * In DAX loader, accept useless 'level' attributes to <job> since
2355 LIGO DAGs have them (seem to be to ease graphical representation).
2357 * Add an example masterslave_mailbox.c using send/receive and not
2358 the deprecated put/get interface.
2359 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2360 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2362 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2363 (declare MSG_USE_DEPRECATED to still have the old name)
2364 * Add a MSG_task_set_data() function
2365 * About trace replay (see examples/msg/actions):
2367 - Allow to work with splitted trace files for each process
2368 Give the specific trace file as argument of each process,
2369 and call MSG_action_trace_run(NULL)
2370 You can still have one merged file for all processes.
2371 - Fix implementation of collective operations
2372 * Allow task_execute() on 0-sized tasks (closes #10063)
2374 * This is the first release of SimGrid where SMPI is not considered
2375 beta anymore (even if some corners should still be improved)
2376 * Port over the new SIMIX_network submodule (internal refactoring)
2377 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2378 * Implement more missing elements of the standard:
2380 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2382 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2383 MPI_Scan() MPI_Get_processor_name()
2384 - Added implementation of missing case for Alltoall (warning: it's
2385 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2386 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2387 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2388 SMPI_MPI_Allgatherv()
2389 * Bug fixes include:
2390 - MPI_Waitsome() was broken
2391 - Allow relative includes in smpicc
2392 - Command line cfg argument 'reference_speed' was ignored...
2393 - Some functions did not properly lead to auto-benching of user code
2394 - smpicc passes -O2 by default (just like openmpi one)
2396 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2397 * Bug fixes about timeouts during communications
2398 * add SIMIX_message_sizes_output() as a pimple to write to file the
2399 amount of messages per size. Use gnuplot to get histogram.
2400 Pimple because that's the only user-visible function of simix,
2401 defined directly in xbt.h (irk, sorry)
2403 - Add a SIMIX_sem_get_capacity() function
2404 - Fix interactions with processe resume/suspende
2405 - release_forever() was stupidly broken
2406 - Fix SIMIX_display_process_status() for processes in a semaphore
2407 - Make SIMIX_sem_block_onto() user-visible
2408 * Refactoring context stuff:
2409 - Use pseudo-OOP for better modularity
2410 - reimplement SIMIX_process_kill() without process_schedule() so
2411 that the latter can take as invariant that it is called from
2413 - Merge context_start into context_new for sake of simplicity
2415 * Add a Vivaldi network model, coded live during SUD'10 ;)
2416 * Rename configuration variables to start a hierarchy:
2417 o cpu_model -> cpu/model
2418 o network_model -> network/model
2419 o workstation_model -> workstation/model
2420 * New configuration variables:
2421 o network/bandwidth_factor: correction to bandwith
2422 o network/latency_factor: correction to latency
2423 o netwotk/weight_S: correction to the weight of competing streams
2424 * Add a long description to the models, that users can see with such
2425 argument on the command line: --cfg=cpu/model:help
2426 * --help-models display the long description of all known models
2428 * config: add the ability to set a default value after registration
2429 Does not override any previously set value (e.g. from cmd line)
2430 * dict: allow to have integer key and data.
2431 When so, you need to use the following functions
2432 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2433 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2434 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2435 In contrary to regular dicts, the key is not malloced before copy.
2436 Mixing scalar and regular elements in the same dict is not tested
2438 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2439 Tracing for Visualization:
2440 * SimGrid is now instrumented in order to generate a trace file for
2441 visualization analysis: to use it, need to compile SimGrid with the
2442 "tracing" option enabled, and instrument the program using SimGrid with
2443 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2444 (among other functions).
2445 * The instrumentation only traces the platform utilization for now
2446 * Documentation to use the tracing functions and how to analyze the
2447 traces with the Triva tool is written.
2448 * More information about: SimGrid FAQ (in the section Tracing Simulations
2451 * We moved to cmake as default build system. Autotools support will
2452 be dropped soon. Check the FAQ for more info about how to use it.
2453 * Greatly improved our cdash/ctest interactions
2454 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2455 * Added memory checking tests with valgrind; lot of memleak fixing.
2456 This may be the first release of simgrid with so few memory issues
2457 * Added code coverage tests.
2458 Our coverage is still improvable, but at least we see it on cdash.
2460 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2462 SimGrid (3.3.4) stable; urgency=low
2464 The "Desktop Grid needs love too" release (also called Xmas release).
2466 Models improvements:
2467 * Major speedup in the maxmin system solving by using lazy evaluation
2468 Instead of solving completely the maxmin system at each iteration,
2469 only invalidate (and recompute) the modified parts.
2470 This new feature is enabled in default models but you can try to
2471 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2472 * Cas01 IMproved as default CPU model
2473 This CPU model is the same Cas01 model, but it uses the
2474 maxmin-selective-update flag and a heap structure to manage
2475 actions on SURF kernel.
2476 It reduces the complexity to find the next action to finish and,
2477 consequently, it's faster than the old Cas01.
2478 This is the new default CPU model (Cas01).
2479 * Rename the old Cas01 model to Cas01_fullupdate
2480 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2481 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2482 * CpuTI (CPU Trace Integration)
2483 A new CPU model whose objective is simulate faster when using
2484 availability trace files.
2485 Instead of using a full featured, over engineered maxmin system for
2486 CPU modeling, this model does the pre-integration of traces files
2487 to calculate the amount of CPU power available, and so, executes
2488 faster than the old CPU models.
2489 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2490 * Use LV08 as default network model since it gives better accuracy
2491 for small messages and shouldn't change things for big ones.
2492 Use --cfg=network_model:CM02 to get the previous behavior.
2495 ******************************************
2496 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2497 ******************************************
2498 * The new CPU model may changes simulations!
2499 The point is that events occurring at the exact same timestamp
2500 are not scheduled in the same order with the old and new
2501 version. This may be enough to completely change the execution
2502 of simulations in some cases.
2503 * The new network model will change simulations!
2504 This new model is more realistic than the previous one, so you
2505 should consider redoing your old experiments with this model.
2506 Sorry for the inconvenience.
2509 * Introduce the supernovae compilation mode
2510 When compiled that way, the whole SimGrid (or almost) is put in a
2511 single compilation unit and compiled in one shoot.
2512 This is to help gcc which has difficulties to inline stuff from one
2514 The speedup seem to be above 15%, althrough more tests are needed on
2515 amd64 to confirm that gain.
2518 * Port of MSG's mailbox on top of SIMIX network
2519 The put/get mechanism was greatly simplified on the way.
2522 * New SIMIX network module. Provides:
2523 - Mailbox: rendez-vous mechanism to find with who you want to speak
2524 - Synchronous send/recv: easier and hopefully faster since the
2525 logic is handled in the maestro process directly now
2526 - Asynchronous send/recv: you dreamt of it? It's here now
2527 Too bad that nobody cared enough to propagate the change to MSG.
2528 * Add semaphores as SIMIX synchronization mechanism.
2531 * new function SD_daxload(char*) to load a DAX file
2532 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2533 * Introduce typed tasks. Specify its kind and cost at creation.
2534 At scheduling, just give where it should be placed, and the cost
2535 for each involved resource is automatically computed.
2536 Existing constructors so far (more to come of course):
2537 - SD_task_create_comm_e2e() for end-to-end communication
2538 - SD_task_create_comp_seq() for sequential computation
2539 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2540 * new function SD_task_dump() for debuging display
2541 * new function SD_task_dotty(task,FILE*) writing to file the info
2542 about the task in dotty format
2543 * SD_task_dependency_exists() can now cope with having one of its
2544 arguments NULL. If so, it tests whether the other argument has any
2546 * Add getters on list of preceding/following tasks:
2547 SD_task_get_parents(task) and SD_task_get_children(task)
2548 * Add getters on amount of workstations and list:
2549 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2550 * Add getter on task kind: SD_task_get_kind(task)
2551 * Update the start_time and finish_time of tasks on completion/failure
2552 * Bugfix: Remove task from state swags when destroyed
2555 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2558 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2559 dynar with one separate thread per value of the dynar.
2560 * Change the prototype of xbt_thread_create(), sorry.
2561 Added a boolean parameter indicating whether we want to join this
2562 thread (used in SG only for now)
2563 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2566 * GTNetS wrappers should now be usable again (and betterly tested too)
2567 * Fix a major regression from 3.2 where the timeout provided to
2568 MSG_task_put_with_timeout() was used as absolute time before which
2569 the comm should be done.
2570 * Start to fix the <cluster> tag.
2571 - Internal links should be good now (beside of the loopback, which
2572 use the private link instead)
2573 - paths to the external world is still rather broken
2574 - the <route:multi> tag is just broken. Actually that's brain-dead.
2575 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2576 to make it less stupid
2577 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2578 * Fix a source-level compatibility glitch from 3.2: after defining
2579 MSG_USE_DEPRECATED, you can use the old name
2580 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2581 * Allow to compile from the SVN with automake 1.11
2582 * Fix some problems when using the "start_time" tag in deployment XMLs.
2583 * Fix #8569: XBT/synchro.h has redundant declarations
2584 * Fix #8563: MSG return values and exceptions
2585 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2586 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2587 Thanks to Robson Peixoto.
2588 * Fix a vicious bug in dictionaries inducing that some elements were
2589 not freed on xbt_dict_free()
2591 Portability report of this version:
2592 * Main portability targets:
2593 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2594 on (amd64/i386/ia64)
2595 - mac leopard on i386
2596 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2597 but nothing critical.
2598 * Other platforms: windows, AIX and others were not tested for this release
2600 Timing report of this version:
2601 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2602 scenario parameters). From 8h to a few seconds in desktop grid settings.
2603 * Supernovae brings about 25% speedup on i386.
2605 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2607 SimGrid (3.3.3) stable; urgency=low
2609 The "Need for Speed" release.
2611 The timings done to validate the 3.3.2 were faulty.
2612 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2614 The problem was a conversion from a manually handled vector to
2615 xbt_dynar_t on the critical path.
2616 xbt_dynar_foreach calls functions, inducing stack management crap.
2618 We inlined these functions and xbt_dynar_foreach is now breath taking.
2619 We also inlined xbt_swag_belong on the way.
2621 Here are some approximate speedup measurements (on master/slaves
2622 simulations lasting between 10s and 20s each):
2623 3.3.1 -> 3.3.2: about same performance
2624 3.3.2 -> 3.3.3: 40% speedup
2625 3.3.1 -> 3.3.3: 40% speedup
2626 3.3.1 with inline patch -> 3.3.3: 30% speedup
2628 Our reading is that the refactoring which occurred in 3.3.2 made us
2629 suffer much more from the xbt_dynar_foreach low performance, but
2630 once we solved this, this refactoring proved to be very performance
2631 effective. From the 40% speedup, somehow, 10% are due to the
2632 inlining and 30% to the refactoring.
2634 That's a pitty that gcc cannot inline functions placed in other files
2635 alone. We have to choose between:
2636 - break the encapsulation (by putting private data structures and
2637 accessors in headers files to help gcc)
2638 - live with low performance
2639 - switch to a decent compiler such as icc (not quite possible).
2641 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2643 SimGrid (3.3.2) stable; urgency=low
2645 The "Simplicity does not preceed complexity, but follows it" release.
2647 The main contributors of this release were (lexical order):
2648 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2651 * Extract the routing logic into its own object.
2652 (was dupplicated in network.c and workstation_LV07.c;
2653 Allows to implement other ways of storing that info)
2654 => kill now useless network_card concept
2655 - Use dynar to represent routes (instead of void** + int*)
2656 - kill link_set (use surf_network_model->resource_set instead)
2657 - Add a command-line option to choose the routing schema to use
2658 - Add three new models:
2659 * Floyd (shortest path computed at initialization)
2660 * Dijikstra (shortest path recomputed all the time)
2661 * Cached Dijikstra (shortest path computed on need)
2662 All these models where contributed by Silas De Munck, and are
2663 described in his ICCS09 paper.
2665 * Simplify model declaration
2666 (less redirections, less function to write when defining a model)
2667 - Factorize stuff between models:
2670 surf_model_resource_set(model)
2671 surf_model_resource_by_name(model, name)
2672 - Unify the types of models in s_surf_model_t (using an union)
2673 - Embeed fields of common_public directly into s_surf_model_t
2674 - Rename model methods:
2675 action_free ~> action_unref
2676 action_change_state ~> action_state_set
2677 action_get_state ~> action_state_get
2678 - Change model methods into functions:
2679 (model)->common_public->action_use ~> surf_action_ref
2681 * Implement a generic resource; use it as ancestor to specific ones
2682 (allows to kill duplicated code in models)
2683 Drawback: timer command don't need no name nor properties;
2684 workstation_CLM03 don't need no properties
2685 (but I guess we can live with those few bytes wasted)
2687 * Improve the action object model
2688 - implement a constructor avoiding dupplicated code about field
2689 initialization in generic_action part.
2691 * Kill the SDP model: it has an external dependency, is deprecated
2692 in flavor of modern lmm models, and didn't compile since a while
2695 * Relocation of the context module from XBT to SIMIX.
2696 (the context were decoupled from the simix processes, duplicating a lot of code)
2697 => a lot of code was factorized
2698 - less overhead is introduced during scheduling
2699 - simpler API for the context factory
2700 - the logic for process creation,destruction and manipulation was simplified
2701 * Simplification of the s_smx_process_t data structure.
2702 => accesing the simix level data associated to a process is faster now,
2703 and the code is a lot more readable.
2706 * Implement some more MPI primitives:
2707 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2708 -implementation: Bcast: flat or 2-ary tree (default),
2709 Barrier: 4-ary tree,
2711 Allreduce: Reduce then Bcast
2712 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2713 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2714 Alltoallv: flat tree, like ompi
2716 * Add support for optimized collectives (Bcast is now binomial by default)
2717 * Port smpirun and smpicc to OS X
2720 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2721 Shout out if you used it.
2724 * Display the list of still queued messages in SG mode when existing
2728 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2729 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2730 * Add xbt_str_from_file(FILE*)
2731 * Add xbt_dict_get_key achieving a linear reverse search
2732 * Remove the context module
2734 Portability report of this version:
2735 * Main portability targets:
2736 - Linux(debian)/x86/context
2737 - Linux(debian)/x86/pthreads
2738 - Linux(debian)/amd64/context
2739 - Linux(debian)/amd64/pthreads
2740 On these, we still have the eratic breakages of gras/pmm and
2741 amok/saturate_sg reported in previous version. We still think
2742 that the tests are the cause of the fault, not the tested code.
2744 - Mac OSX Leopard/x86/context
2745 Still false negative in tesh autotesting.
2746 Smpi still fails, but this time because readlink does not accept -f
2747 Everything seems to work properly beside of that.
2750 - AIX version 5.3 (only tested contexts this time)
2751 Smpi still fails there because mktemp is not installed.
2752 Everything seems to work properly beside of that.
2754 I managed to compile it for the first time, but several breakages.
2755 Won't delay the release for this exotic platform.
2757 * Windows: it's still lagging behind. If you want to help, please
2760 Timing report of this version:
2761 This version seem to be more than 5% faster than 3.3.1 (on linux
2762 64bits with contextes). The gain is less than expected, we are
2763 investigating this for next release.
2765 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2767 SimGrid (3.3.1) stable; urgency=low
2770 * Implement a --cfg-help to show existing configuration variables
2771 * Build chain do not require doxygen in maintainer mode
2774 * fix a bug on struct sizeof computation, which prevented the
2775 exchange of arrays of structs in some conditions
2776 - added a regression test about this in datadesc_usage
2777 * Allow the exchange of 0-long dynamic vectors.
2778 - for that, use -1 as indicator of dynamic size instead of 0
2779 - This implied to change any size from unsigned long to long,
2780 reducing a bit communication abilities, but I guess that with
2781 64bits being quite common, this is more than enough.
2782 - This also induce a protocol change, thus bumping network protocol
2783 version from 0 to 1 (if we have external users, we have to get
2784 clean on that point too ;)
2785 - added two regression tests about this in datadesc_usage
2786 * Be more verbose when propagating local exceptions
2787 This helps debugging.
2788 * Display the status of simulated processes when receiving SIGINT in
2792 * Allow to control the simulation from a trace file.
2793 New functions MSG_action_register() and MSG_action_trace_run()
2794 The first one allows to associate a function execution to each
2795 kind of action while the second one parses a trace file and
2796 triggers the corresponding actions within the system.
2797 For now, only a toy example is provided in examples/msg/actions
2798 * Add an exemple of process migration in examples/msg/migration
2799 * Fix a bug in task exchange which broke MSG_task_get_sender()
2800 Add a teshsuite regression test for that.
2801 [Bug: if MSG_task_get_sender() is called after sender exit,
2803 * Fix a bug which prevented suspend/resume to work properly
2804 * Display the status of simulated processes when receiving SIGINT
2805 This fixes a regression of v3.3. due to the introduction of SIMIX
2806 * Bug fixing in failure management:
2807 - trace could not start by a failure at time 0
2808 - failure during communications were not working
2811 * Add SIMIX_process_set_name() to change the name of the current
2812 process in the log messages.
2813 * Store smx_hosts in a dict since we only retrieve them by name
2814 * Move the configuration infrastructure to surf
2817 * Move the configuration infrastructure to surf
2820 * Massive internal cleanups:
2821 - Store internal structures on processes instead of hosts (allows
2822 to have more than one process per host, in addition of being more
2824 - Cleanup the initialization/finalization process
2825 - Kill a whole bunch of unneeded synchronization:
2826 processes run in exclusive manner within the simulator
2827 - Move queues from global tables to process data fields
2829 - now accept -platform and -hostfile arguments
2830 - Pass the right rank value to processes according to the hostfile
2831 * Compile the examples by default, and use them as regression tests
2832 * Implement MPI_Wtime()
2833 * Change the reference speed to a command line option
2836 * TCP_gamma can now be specified as command line option using
2837 --cfg=TCP_gamma:10000000.0
2838 * Change the --surf-path cmd line option into --cfg=path:
2841 * Also include strbuff from xbt.h public header
2842 * xbt_ex_display(): do not free the exception after displaying
2843 This allows to do more with the given exception afterward.
2844 Users should call xbt_ex_free() themselves.
2848 Portability report of this version:
2849 * Main portability targets:
2850 - Linux(debian)/x86/context
2851 - Linux(debian)/x86/pthreads
2852 - Linux(debian)/amd64/context
2853 - Linux(debian)/amd64/pthreads
2854 These targets fail about 1/10 of times on gras/pmm, but we believe
2855 that this is because of the test, not because of simgrid.
2856 amok/saturate_sg fails even more rarely, and the test may not be
2859 - Mac OSX Leopard/x86/context
2860 The test suite still spits tons of errors because some obscure
2861 force prevents us from removing the temporary directories
2862 arguing that they still contain some metadata I've never heard of.
2863 Smpi fails because seq is not installed.
2864 Everything seems to work properly beside of that.
2867 - AIX version 5.3 (both contexts and pthread)
2868 Smpi still fails there because mktemp is not installed.
2869 XML inclusions seems rosty on AIX.
2871 * Windows: it's still lagging behind. If you want to help, please
2874 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2876 SimGrid (3.3) stable; urgency=high
2880 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2881 [Malek Cherier & Mt]
2883 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2886 Until now, GRAS were using MSG as an interface to SURF. It was
2887 quite difficult because both interface have several differences
2888 (MSG channels vs GRAS sockets were the most notable point).
2890 This also opens the gate to SMPI (which should occur soon) and speed
2891 up simulations by to 40% (even if it were not the main goal).
2893 **************************************
2894 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2895 **************************************
2896 The point is that events occuring at the exact same timestamp are
2897 not scheduled in the same order with the old and new version. This
2898 may be enough to completely change the execution of simulations in
2899 some cases. Sorry for the inconvenience.
2901 * Cleanup and upgrade the XML format to push further scalability
2902 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2904 * Improve the testing infrastructure with tesh. Now a very large part of
2905 the code is tested not only by being run but also by checking that the
2906 output match an expected output [Mt].
2908 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2909 is really less memory-demanding, which should allow you to use
2910 larger files in SimGrid [AL].
2912 * Inform valgrind about our contextes, so that it becomes usable
2913 with the default (and more effecient) version of SimGrid
2914 [contributed by Sékou Diakite, many thanks]
2917 * Introduce a listener thread in charge of receiving incoming
2918 messages from the network. It allows to overlap communication and
2919 computation but most notably, it removes some stupid deadlocks due
2920 to the fact that so far, a process could not send and receive at
2921 the same time. This made most non trivial communication schema
2923 * Convert the PIDs from long int to int to match the MSG ones (and
2924 linux ones too) [Mt]
2925 * New function: gras_agent_spawn() to launch a new process on
2926 current host. Only working in simulation for now. [Mt]
2927 * New function: gras_os_hostport() returning a constant form (ie,
2928 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2931 * Make the backtrace of exceptions more human readable [Mt]
2932 * New module: xbt/str [Mt]
2933 a ton of string utility functions (split, join, printf to a newly
2934 allocated buffer, trim, etc)
2935 * New module: xbt/hash [Mt]
2936 SHA1 hashing algorithm (more to come if needed)
2937 * New module: xbt/synchro [Mt]
2938 synchronization tools (mutex and conditions) working the same way
2939 in simulation and in real life (mainly useful for GRAS, but not
2941 * New module: xbt/queue [Mt]
2942 classical producer/consumer synchronization scheme
2943 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2944 (using the classical functions will get serialized) [Mt]
2945 * Make dictionary internal table dynamic. No need to specify its size
2946 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2948 * Make sure the log channels are organized as a tree under windows
2949 (because of ANSI C compatibility issue, any channel were child of
2950 root directly) [Mt].
2953 * Cleaned many thing in surf and fixed a few bugs [AL].
2954 * Add a nice command line configuration mechanism to compose models [AL].
2955 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2956 the previous one (KCCFLN05). It relies on something that looks like
2957 a max-min sharing mechanism but cannot be written as such. A new solver
2958 was thus designed [AL].
2959 * Add a new solver to lmm. Based on Lagrange optimization and
2960 gradient-based descent, it enables to efficiently maximise systems s.a
2962 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2964 This solver enables to propose two new network models for TCP Reno and
2965 TCP Vegas based on Low's work. These models still need to be fully
2966 tested though [Pedro Velho].
2969 * Bug fix in SD_simulate. Now the time bound given as argument is
2971 * Use the new parallel task model (ptask_L07) as default.
2972 * Use the SURF command line configuration mechanism.
2973 * 0-size tasks (for synchronization) should now work.
2975 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2977 SimGrid (3.2) stable; urgency=high
2981 We still experience issues on this platform, but we believe that at
2982 least MSG is usable.
2984 GRAS API BREAKAGE (for simplification purpose, sorry):
2985 * the gras_msgtype_by_name is not used anymore. Instead of
2986 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2987 you can write (and must)
2988 gras_msg_send(toserver, "request", &request);
2989 - If you still want to pass a gras_msgtype_t to the function (to cache
2990 the type and avoid the lookup time), use the gras_msg_send_() variant.
2991 - Impacted functions:
2992 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2993 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2994 * The callbacks are now expected to return 0 when everything went well
2995 (just like the main() function)
2997 GRAS new features and improvements:
2998 * New module mechanism where user code can use per process globals [Mt]
2999 This is similar to gras_userdata_*() functions, but for libraries. It
3000 factorize some code developped over and over in the examples and AMOK.
3001 It has still to be documented and used (only amok/peermanagement is
3003 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3004 loss when they were small enough to fit into the buffer and sent quickly
3005 enough so that they can all get received in one shoot.
3006 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3007 if not found. Use the *_or_null() variant for the old semantic.
3008 * In gras_msg_handle, do not discard messages without callback.
3009 They are probably messages to be explicitly awaited later (ie, proofs of
3010 mis-synchronization in userland since they are sent before being awaited)
3011 No big deal usually.
3012 * gras_socket_meas_send/recv: semantic changed!
3013 The numerical arguments used to be (1) the total amount of data to send
3014 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3015 messages. This was need for the fool willing to send more than MAXINT
3016 bytes on quite fat pipes.
3019 * Do really rename the hostmanagement module to peermanagement. [Mt]
3020 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3021 but this is rather new and this was documented in the module
3022 documentation (poor excuses, I admit)
3023 * Bandwidth measurement semantic changed! This follows the changes to
3024 gras_socket_meas_send/recv explained above.
3027 * A sequential mode has been added to the workstations. When a workstation
3028 is in sequential mode, it can execute only one task, and the other tasks
3029 are waiting in a FIFO. [Christophe Thiery]
3032 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3033 model for SIMDAG. [Christophe Thiery]
3034 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3035 equal to 0, which caused some bad behaviors in
3036 saturated_constraint_set_update. I now use a threshold mechanism like in
3040 * When running manually src/testall, you select specific units [Mt]
3041 testall is the result of our cunit mechanism, and should replace all
3042 the scripty thingy around since bash don't run easily on billware.
3044 * A mallocator system has been added. [Christophe Thiery]
3045 Mallocators allow you to recycle your unused objects instead of freeing them
3046 and allocating new ones.
3048 Documentation update:
3049 * FAQ reworking + New FAQs:
3050 - "Valgrind spits tons of errors!" [Mt]
3051 - "How to repport bugs" [Mt]
3052 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3053 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3054 - Communication time measurement within MSG [AL]
3055 - I experience weird communication times when I change the latency [AL]
3056 * GRAS tutorial [Mt]
3058 - an introduction to the framework and to the used communication model
3059 - an initiatic tour introducing the most proheminent features:
3061 . Lesson 0: Installing GRAS
3062 . Lesson 1: Setting up your own project
3063 o Part 2: Message passing
3064 . Lesson 2: Exchanging simple messages
3065 . Lesson 3: Passing arguments to the processes (in SG)
3066 . Lesson 4: Attaching callbacks to messages
3067 . Lesson 5: Using globals in processes
3068 . Lesson 6: Logging information properly
3069 . Lesson 7: Using internal timers
3070 . Lesson 8: Handling errors through exceptions
3071 . Lesson 9: Exchanging simple data
3072 . Lesson 10: Remote Procedure Calling (RPC)
3073 . Lesson 11: Explicitely waiting for messages
3074 . Recapping of message passing features in GRAS
3075 - A HOWTO section containing:
3076 o HOWTO design a GRAS application
3077 More are due, of course. They will come latter. In the meanwhile, you can
3078 check the examples which are still here.
3080 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3082 SimGrid (3.1) stable; urgency=high
3086 There was a stack corruption somewhere, visible only when optimizing
3087 with these versions. [Vince]
3090 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3091 module that enabled to play with Directed Acyclic Graphs. It is built
3092 directly on top of SURF and provides an API rather close to the old
3093 SG. Some old codes using SG are currently under rewrite to check that
3094 all needful functions are provided. [Christophe Thiery]
3097 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3098 extension of the classical CLM03 model that gracefully handles
3099 failures. This is now the default model for MSG and GRAS. It doesn't
3100 handle parallel tasks yet though. [AL]
3101 * Bug fix: Weights were not correctly set in the network part.
3102 WARNING: This may have resulted in incorrect results with simulations
3103 where there are more than one flow on a given link. [AL]
3106 * After a (long ?) discussion on simgrid-devel, we have decided that the
3107 convention we had on units was stupid. That is why it has been decided
3108 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3109 WARNING: This means that all previous platform files will not work as
3110 such with this version! A warning is issued to ask users to update
3112 A conversion script can be found in the contrib module of the CVS, under
3113 the name contrib/platform_generation/surfxml_update.pl [MQ]
3116 * Bug fix: Processes were started in reverse order, wrt deployment file.
3117 WARNING: if your code relies on this bug, please fix it. [AL]
3118 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3119 being executed on two different locations. [AL]
3120 * Bug fix: Failures are now better supported thanks to Derrick's tests
3121 (there was many failure situations I hadn't thought of and that weren't
3122 correctly handled). [AL]
3123 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3127 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3128 Exception occurring on server-side are propagated back to client (!).
3130 API CHANGE: the callback changed their prototype. Change:
3131 int my_handler(gras_socket_t expeditor, void *payload_data) {
3133 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3134 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3136 * New! function: gras_msg_handleall to deal with all messages arriving
3137 within a given period.
3138 * New! function: gras_socket_server_range to get a server socket in a
3139 range of port numbers (ease to avoid port number conflicts) [MQ]
3140 * New! gras processes display their backtrace when they get a SIGUSR1
3141 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3142 Sweet to debug RL processes [MQ]
3146 - Do not force experiment sizes to be expressed in kb, or it becomes
3147 impossible to measure the latency this way (needs one byte-long tests)
3148 WARNING: this changes the amok_bw_* function semantic. [MQ]
3149 - Implements the link saturation stuff. [MQ]
3150 * Peer management module:
3151 New! module factorizing code that we wrote over and over [MQ].
3154 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3155 - Test units are placed directly into the library code, they get extracted
3156 automatically and placed into the src/testall binary.
3157 - Convert most of the XBT tests to this system.
3158 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3159 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3160 live in the same C code. This required to change a little bit the API
3161 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3162 * New module: sparse graph structure with basic algorithms (this is work
3163 in progress and the API is not considered to be frozen yet). [AL]
3164 * Display more information on backtraces: source line & function names are
3165 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3166 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
3167 * Reworked a little bit some #include statements to load only required
3168 headers. Some user code that relied on SimGrid to include stdlib or
3169 stdio may need to include it by themselves. [AL]
3170 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3172 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3173 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3174 but I guess that almost nobody used those parts. [MQ]
3176 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3178 SimGrid (3.0.1) stable; urgency=low
3181 * Unfortunately, I had missed 5 misnamed functions:
3182 xbt_fifo_item_t xbt_fifo_newitem(void);
3183 void xbt_fifo_freeitem(xbt_fifo_item_t);
3184 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3185 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3186 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3187 They're now deprecated. Please use their new versions:
3188 xbt_fifo_item_t xbt_fifo_new_item(void);
3189 void xbt_fifo_free_item(xbt_fifo_item_t);
3190 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3191 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3192 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3194 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3195 * Documentation: xbt_log module unmercifully reworked [MQ]
3196 * Bugfix: there was a problem with the ending of contexts with
3197 the pthread backend. It caused some weird deadlock or behavior
3198 depending on the pthread implementation. [AL]
3199 * Bugfix: get the exceptions raised in the simulator repport where
3200 and why they come from when they are not caught in time [AL, MQ]
3203 * Bugfix: Do repport the error when two non-connected hosts try to
3204 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3207 * Add additionnal checkings on communications. Assert that two
3208 communicating hosts are connected by a set of links... [AL]
3211 * Add additionnal checkings on channel values in communication [AL]
3212 * New: MSG_task_get_source to see on which host a task was generated [HC]
3213 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3214 the number of tasks waiting to be received on channel and sent
3216 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3217 waits for the first task coming from a given host.. [AL]
3219 GRAS new functionnalities: [MQ]
3220 * Enhance the parsing macro to allow the size of multidimentional objects
3221 to be given thru annotations.
3222 * New example (and documentation): Matrix Multiplication a la RPC
3223 (as when I was young!) and fix a bunch of bugs found on the way.
3225 GRAS performance improvements: [MQ]
3227 * Reduce the amount of cbps creation/destruction by making it static to
3228 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3231 * Change libdata to a set so that we can search for stuff by ID (and thus
3232 reduce the insane amount of dict lookups)
3235 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3236 RL communication on top of SG since it's so uneffective.
3237 It may also be used for inter-thread communication in RL, one day.
3238 * Use gras_datadesc_copy() to exchange messages on top of SG
3240 - improve message exchange performance on top of SG
3241 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3242 * Don't exchange on the network the size of the used part of buffer,
3243 instead, specify the possible buffer size to read().
3245 - reduces the amount of read/write calls (one pair per exchange)
3246 - reduces the amount of exchanged data (the size)
3247 - allows to retrieve all arrived data on receiver side, if we don't need
3248 it right now (subsequent read will peek the buffer)
3249 - allows the receiver to proceed with the begining of the stream before
3250 everything is arrived
3251 - make it possible to build an iov transport (using readv/writev)
3253 - take care of the data with non-stable storage (like stacked data),
3255 * If possible, TCP send uses vector I/O (when writev() is here)
3256 - Don't use it for receive since we send data sizes and data on the
3257 same stream, so we wouldn't be able to chain large amount of chunks
3258 before having to flush the stuff to read the size.
3259 * Rework the transport plugin mechanism to simplify it and reduce the
3260 amount of pointer dereferencement when searching for the right function
3263 * I guess that now, we do almost as few system calls as possible while
3264 doing as few data copy as possible.
3266 To improve it further, we could try to send all the sizes first and then
3267 all the data (to use iov on receiving size), but it's only a partial
3268 solution: when you have 2 dimensional data, the sizes of the second
3269 dimension is data of the first dimension, so you need 3 streams.
3271 I'm not sure the potential performance gains justify the coding burden.
3273 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3275 SimGrid (3.00) stable; urgency=high
3278 * New! Give the possibility to hijack the surf parser and thus bypass
3279 MSG_create_environment and MSG_launch_application. Have a look at
3280 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3282 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3284 SimGrid (2.96) unstable; urgency=low
3289 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3290 This deprecates the xbt_error_t mechanisms.
3291 It modifies (simplifies) all XBT and GRAS API.
3292 MSG API keeps unchanged (exceptions raised by XBT are caught from
3293 within MSG and masked with existing error handling facilities)
3296 * New! Add a FATPIPE model. [AL]
3297 * New! Add a parallel task model. [AL]
3298 * New! Add automatically a loopback interface (in the default
3299 network model) if none was precised.
3302 * Bugfix: MSG_process_resume now works with the current running process.
3304 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3305 * Modification of MSG_task_get_compute_duration. Once a task has been
3306 processed, the value returned by this function is now equal to 0. [AL]
3307 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3308 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3309 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3310 of a m_task. For now, MSG_task_cancel only works with computation
3312 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3313 of the processor (in Mflop/s) regardless of the current load on the
3315 * API Change: use proper naming convention for MSG_getClock and
3316 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3318 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3319 This function changes the priority of a computation task. This priority
3320 doesn't affect the transfer rate. A priority of 2 will make a task
3321 receive two times more cpu power than the other ones. This function
3322 has been added to suit the needs of Nguyen The Loc and hasn't been that
3323 much tested yet. So if it fails, please report it and send me your code.
3325 * API Change: removed all functions and types that were marked "deprecated"
3326 since many months. Renamed MSG_global_init_args to MSG_global_init.
3328 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3330 SimGrid (2.95) unstable; urgency=low
3333 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3334 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3335 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3338 * Bugfix: Allow absolute paths for platform description files [MQ]
3339 * Bugfix: do free the variables after use. Leads to drastic performance
3341 * Implement max_duration (ie, timeouts) on resources [AL]
3344 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3346 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3347 the needed support (timeouts) [AL]
3350 * Implement measurement sockets. You can now get the bandwidth between two
3351 hosts thanks to AMOK (see below). [MQ]
3352 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3353 dynar over the network (yeah) [MQ]
3354 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3355 * simple/static token-ring example. [Alexandre Colucci and MQ]
3356 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3357 in place before [MQ]
3360 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3361 type parsing. "short" and "long" modifiers were also ignored; other
3362 modifier (such as reference level) are still ignored. [MQ]
3363 * Embeed the buffer size within the buffer itself on SG. [MQ]
3364 That way, send() are atomic and cannot get intermixed anymore (at least
3365 the ones which are less than 100k; bigger messages still have the issue)
3366 * Array size pushed by the field, not by the field type (or each
3367 and every long int will push stuff to the cbps) [MQ]
3368 * use select() to sleep since it allows to portably sleep less than one
3371 GRAS (minor cleanups)
3372 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3375 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3376 * type category 'ignored' killed as it was never used and were difficult
3378 * whether a type can cycle or not is now a flag, leaving room for more
3379 flags (as "ignored", if we feel the need one day ;)
3380 * Rename raw sockets to measurement sockets since "raw" has another
3381 meaning in networking community.
3384 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3385 and offers features not belonging to GRAS but that most applications
3386 need. One day, it may be a set of plugins loadable at runtime.
3387 * New module: bandwidth
3388 bandwidth measurement between arbitrary nodes running this module. [MQ]
3390 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3392 SimGrid (2.94) unstable; urgency=low
3394 The first beta release of SimGrid 3 !
3398 * Update the main page and the FAQ. Adding references to gforge.
3401 * Add a gras_os_getpid function.
3404 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3405 * Extend the logs so that they also print PID, hostname, date, ... if
3407 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3408 and kill the old version which were in testsuite/
3409 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3413 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3414 for users. Instead, each programming environment (GRAS, MSG,...) use it
3415 when needed to provide such a feature to users.
3416 Don't shortcut the mechanism or you will also shortcut the virtualization
3417 you need on the simulator.
3421 * Cleanups in configury with regard to compile optimization/warning flags.
3422 Also add -fno-loop-optimize to any powerpc since it's the optimization
3423 killing gcc (< 3.4.0).
3424 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3425 needed by MSG examples complications
3426 * Borrow configury beautifications from PHP
3429 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3430 warning (thanks loris for stumbling into this one).
3431 * Bugfix: stop loading private headers (gras_config.h) from the public
3435 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3436 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3437 RL would imply a DNS resolver, which is *hard* to do in a portable way
3438 (and therefore delayed).
3439 * Implement a real timer mechanism and use it in timing macros. This allows
3440 to avoid rounding errors and get a 0.000005 sec precision in timing
3441 macros. While I was at it, various cleanups:
3442 - allow to declare more than one timed section per file (fix a stupid bug)
3443 - move some private declaration to the right place
3444 - merge conditional execution and timing macros into emulation module
3445 - document the module
3446 - make sure the module cleanups its mess on gras_exit
3447 * Documentation improvements:
3448 - (new) how to compile applications using GRAS
3449 - (new) emulation support (timing macros)
3451 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3453 SimGrid (2.93) unstable; urgency=low
3455 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3458 - Use Paje properly where used. Still to be sanitized properly.
3459 - Portability fix: Add an implementation of the contexts using pthread
3463 - Add xbt_procname(): returns the name of the current process.
3464 Use it to show the current process's name in all logging.
3466 - fix detection of older flex version and the reaction, since we do
3467 depend on modern ones (we use lex_destroy)
3468 - Better separation of SG and RL in the libs: remove all simulation code
3469 from libgras. As a result, this lib is now only 200k when stripped.
3470 Some of the xbt modules may also be duplicated (two sets and such) and
3471 should be cleaned/killed before SG3.
3472 - Insist on using xlC on AIX because of weird problems involving gcc there.
3473 - Cleanup the make remote stuff. This is now done by scripts
3474 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3475 mainly for our private use, but we're working on changing them to user
3478 - Bugfix: flush the socket on close only if there is some *output*.
3479 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3481 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3484 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3486 SimGrid (2.92) unstable; urgency=low
3488 Alpha 3 on the path to SimGrid 3
3492 - New! First try of benchmarking macros.
3493 - New! First try so that gras_stub_generator generate deployment and
3494 remote compilation helpers.
3496 - Bugfix: Initialization fix in msg_test.
3500 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3502 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3504 - API Change: gras_os_sleep to take the amount of seconds as a double.
3505 Accepting an int was error prone since it was the only location where
3506 seconds were coded as such. It leaded to damn rounding errors.
3507 - Bugfix: Hard to belive that timers ever worked before this.
3509 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3511 SimGrid (2.91) unstable; urgency=low
3513 Alpha 2 on the path to SimGrid 3
3517 - Bug fix in the lmm_solver.
3519 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3520 through the function MSG_paje_output.
3521 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3522 - It is possible to bound the rate of a communication in MSG with
3523 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3524 to write it in the changelog).
3525 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3530 - Various cleanups to the autotools stuff
3531 - Begin to move Gras examples to examples/gras/
3532 - Let make distcheck work again (yeah!)
3534 - documentation overhauled using doxygen.
3535 gtk-doc-tools is dead in SimGrid now.
3536 - Automatically extract all existing logging categories, and add the list
3537 to the documentation (long standing one, to say the less)
3539 - Cleanup the known architecture table. Reorder the entries to group what
3540 should be, and use a more consistent naming scheme.
3541 (some of the test dataset are still to be regenerated)
3542 - New! Allow library to register globals on each process just as userdata
3544 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3545 lookup time (for now).
3546 Use it in msg and trp.
3547 This cleans a lot the internals and helps enforcing privacy of the
3548 headers between the gras components.
3549 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3550 - Bugfix: gras_os_time was delirious in RL.
3551 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3553 - Reenable GRAS now that it works.
3555 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3557 SimGrid (2.90) unstable; urgency=low
3559 Alpha 1 on the path to SimGrid 3
3561 * It is a long time since the last release of SimGrid. I'm sorry about
3562 that but as I had told you, I was rewriting a lot of things. I apologize
3563 to those who had been reporting bugs to me and that I had not answered.
3564 If your bug is still in the new version, please tell me. Here is a
3565 summary of the main changes.
3567 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3568 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3569 portability, speed, and a lot more but you'll figure it out later.
3570 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3571 I wanted to release the new MSG as soon as possible and I have
3572 broken GRAS, which is the reason why, for now, only MSG is fully
3573 functional. A laconic description of these projects is available
3574 in the documentation.
3576 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3577 kernel optimized for our needs (called SURF but only the developers
3578 should use it). Hence, MSG is now roughly 30 times faster and I think
3579 that by rewriting a little bit MSG, I could event speed it up a little
3580 bit more. Beside the gain in speed, it is also much easier to encode a
3581 new platform model with SURF than it was with SG. More to come...
3583 * REVOLUTION 3: I have tried to change a little as possible the API of
3584 MSG but a few things really had to disappear. The main differences
3585 with the previous version are:
3586 1) no more m_links_t and the corresponding functions. Platforms are
3587 directly read from a XML description and cannot be hard-coded
3588 anymore. The same format is used for application deployment
3589 description. The new format is described in the documentation.
3590 Have a look in tools/platform_generation. There is a tiny script
3591 that converts from the old platform format to the new one. Concerning
3592 the application deployment format, parsing the old one is tricky.
3593 I think most of you should however be able to convert your files. If
3594 it is really an issue, I can write a C code that does the conversion.
3596 2) the toolbox tbx does not exist anymore. We now have a library
3597 with much more data-structures but without the hash-tables (we have
3598 dictionaries that are much faster).
3600 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3602 *****************************************************************************
3603 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3604 * don't feel like converting the oldies. (Mt) *
3605 *****************************************************************************
3608 Version 2.90: "the long awaited one"
3609 - Finished rewriting and debugging MSG. Rewrote the documentation.
3610 - disable GRAS for now since it needs to be ported to the newest SG
3613 - Finish the port to windows (using mingw32 for cross-compile)
3616 - Main loop and datastructures of SURF. A cpu resource object is
3617 functional. Surf can thus be used to create cpu's with variable
3618 performance on which you can execute some actions.
3620 2004-11-15 Martin Quinson
3621 - Port to ARM. Simply added the alignment and size descriptions. Should
3622 work, but the ARM machines are so slow that I didn't had the opportunity
3623 to 'make check' over there yet.
3625 2004-11-15 Arnaud Legrand
3626 - Trace manager now written. It uses a heap structure and is therefore
3627 expected to be efficient. It may however be speeded up (particularly
3628 when many events occur at the same date) by using red and black
3629 trees. One day maybe...
3630 - Max-min linear system solver written. It uses a sparse matrix
3631 structure taking advantage of its expected use. Most operations are
3632 O(1) and free/calloc are called as few as possible. The computation of
3633 the minimum could however be improved by using a red and black tree
3636 2004-11-03 Arnaud Legrand
3637 - Rename every gras_* function that was in xbt/ to its xbt_
3639 - Add a heap and a doubly-linked list to xbt
3640 - Added a dichotomy to the dictionaries. make check works as well before
3641 so I assume that the patch is correct. I do not know however if things
3642 run effectively faster than before now. :)
3644 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3645 SimGrid, and the version number is bumped to 2.x
3647 2004-10-29 Martin Quinson
3648 - Introduction of the remote errors.
3649 They are the result of a RMI/RPC on the remote machine.
3650 ErrCodes being scalar values, you can't get the host on which those
3651 errors did happen. Extending the error mechanism as in Gnome is possible.
3652 No idea yet whether it is a good idea.
3654 2004-10-28 Martin Quinson
3655 - Interface revolution: the Starred Structure Eradication.
3656 I used to do typedef struct {} toto_t; and then handle *toto_t.
3657 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3658 - ???_t is a valid type (builded with typedef)
3659 - s_toto_t is a structure (access to fields with .)
3660 - s_toto is a structure needing 'struct' keyword to be used
3661 - e_toto_t is an enum
3662 - toto_t is an 'object' (struct*)
3664 typedef struct s_toto {} s_toto_t, *toto_t;
3665 typedef enum {} e_toto_t;
3666 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3667 s_toto_t) is private.
3669 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3670 so that it can set the variable to NULL. It was so for dicts and sets,
3671 it changed for dynars.
3673 - Fix a bunch of memleaks in dict_remove
3674 - Fix a bug in sg/server_socket opening: it failed all the time.
3676 2004-10-07 Martin Quinson
3677 - Speed up dynar lookup operation a bit.
3679 gras_dynar_get is dead.
3681 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3682 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3683 (which gives you the address of the stored data).
3685 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3686 copy of the data using an affectation to do the job and not a memcpy.
3688 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3689 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3691 It does not really speedup the dynar test because they are
3692 setting elements all the time (and look them seldom). But the dict does
3693 far more lookup than setting.
3695 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3697 2004-10-05 Martin Quinson
3698 - Allow to (en/dis)able the cycle detection at run time.
3700 Whether we should check for cycle or not is now a property of each
3701 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3702 datadesc_cycle_unset allow to remove this property when previously set.
3704 Note that the cycle detection is off by default since it impacts the
3705 performance. Watch the data you feed GRAS with ;)
3707 This property is hereditary. Any element embedded in a structure having it
3708 set have it set for the time of this data exchange.
3710 You should set it both on sender and receiver side. If you don't set it on
3711 sender side, it will enter an endless loop. If you forget on receiver
3712 side, the cycles won't be recreated after communication.
3714 - Header reorganization.
3715 Kill gras_private.h, each submodule must load the headers it needs.
3717 2004-10-04 Martin Quinson
3718 - Interface revolution: do not try to survive to malloc failure.
3720 Now, gras_malloc and friends call gras_abort() on failure.
3721 As a conclusion, malloc_error is not a valid error anymore, and all
3722 functions for which it was the only gras_error_t return value are
3723 changed. They now return void, or there result directly.
3724 This simplify the API a lot.
3726 2004-09-29 Martin Quinson
3727 - Re-enable raw sockets.
3728 Created by gras_socket_{client,server}_ext;
3729 Used with gras_raw_{send,recv}
3732 It should allow to kill the last bits of gras first version soon.
3734 This is not completely satisfactory yet (duplicate code with
3735 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3738 - Simplify transport plugin (internal) interface by not passing any
3739 argument to _server and _client, but embedding them in the socket
3742 2004-09-28 Martin Quinson
3743 - Finish the port to AIX.
3744 autoconf was my problem (segfault within the malloc replacement
3745 function. No idea why)
3747 2004-09-16 Martin Quinson
3748 - Fix some size_t madness on 64bit architectures.
3750 2004-09-08 Martin Quinson
3751 - Reduce the number of system headers loaded, overload some more system
3752 calls (such as malloc to cast the result of the system one, and work
3754 - Fix and reintroduce the config support
3756 2004-09-07 Martin Quinson
3757 - Source code reorganization to allow Arnaud to surf all over there.
3758 - Allow to document the logging categories.
3759 - Remove all uppercase from logging categories and useless cleanup in names.
3761 2004-08-18 Martin Quinson
3762 Version 0.6.2 (protocol not changed; API changed)
3763 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3764 gras_error_t), and NULL when not found. Functions expecting a msgtype
3765 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3766 hopefully usefull message.
3767 - Portability to prehistoric sparcs again
3769 2004-08-17 Martin Quinson
3770 Version 0.6.1 (protocol not changed; ABI not changed)
3771 - prealloc some buffers to speed things up
3773 2004-08-11 Martin Quinson
3774 Version 0.6 (protocol not changed; ABI expended)
3775 - The parsing macro can deal with the references, provided that you add
3776 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3778 2004-08-09 Martin Quinson
3779 Version 0.5 (protocol not changed; ABI changed)
3780 - Allow to off turn the cycle detection code in data exchange at
3781 compilation time. It should be at run time, but I'm short of time (and
3782 the config stuff is still broken). That way, we keep dict out of the
3783 critical path, which is good because the performance is poor:
3784 - search not dichotomial yet
3785 - dynar give no way to access their content and memcpy everytime
3786 - In composed data description (struct, ref and so on), stop foolness of
3787 keeping the subtype's ID, but store the type itself. This keeps sets out
3788 of the critical path, which is good since they rely on dynar and
3789 dictionnaries. The only loose of that is that we cannot detect the
3790 redeclaration of a structure/union with another content (but I'm not sure
3791 the code detected well this error before anyway). We still can detect
3792 the redefinition discrepancy for the other types.
3793 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3794 it breaks the code because of type-punning used all over the place).
3795 This breaks on all non-gcc architectures (for now).
3797 All those changes (plus the buffer of last time) allow me to gain 2 order
3798 of magnitude on cruel tests consisting of 800000 array of integers on two
3799 level of a hierarchical structure (200 secondes -> 4 secondes)
3802 - the selector of reference must now return the type it points to, not
3803 the ID of this type.
3805 2004-08-06 Martin Quinson
3806 Version 0.4 (protocol changed; ABI not changed)
3807 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3808 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3809 - Implement a buffer transport, and use it by default (it relies on tcp in
3810 real life and on sg in simulation).
3811 That's a bit hackish since I had a new field to the structure to store
3812 its data without interfering with the subtype ones. Inheritance
3813 is tricky in C. And that's a kind of reverse inheritance with one class
3814 derivating two classes. Or maybe a game with java interfaces. Anyway,
3815 that's damn hard in C (at least).
3816 Moreover, I got tired while trying to ensure plugin separation and
3817 genericity in SG mode. MSG wants me to do weird things, so let's go for
3818 cruel hacks (temporarily of course ;).
3819 See comment in transport_private.h:71
3820 - do not include all the _interface headers in private but in the files
3821 which really need them (to cut the compilation time when they are
3824 2004-07-26 Martin Quinson
3825 Version 0.3 (protocol not changed; ABI changed)
3826 - Major overhault of the datadesc interface to simplify it:
3827 - shorted the function names:
3828 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3829 - add a trivial way to push/pop integers into the cbps without malloc.
3830 This allows to make really generic sub_type description, which simply
3831 pop their size of the stack.
3832 - add a function gras_datadesc_ref_pop_arr() which does what users want
3833 most of the time: Declare a dynamic array (which pops its size of the
3834 stack) and declare a reference to it. Poor name, but anyway.
3835 - kill the post-send callback, add a post-receive one
3837 2004-07-23 Martin Quinson
3838 Version 0.2 (protocol changed; ABI changed)
3839 - add some testing for cpbs in the test cases, and fix some more bugs.
3840 This invalidate again the little64 data file, since I cannot regenerate
3842 - remove an awfull optimization in the logging stuff, allowing me to:
3843 - understand it again
3844 - learn gcc how to check that the argument match the provided format
3845 - fix all errors revealed by gcc after that
3846 - internal keys of dict are not \0 terminated. Deal with it properly in
3847 loggings instead of segfaulting when the user want to see the logs :-/
3849 2004-07-22 Martin Quinson
3850 - Fix some stupid bug preventing cbps (callback postit) from working
3852 2004-07-21 Martin Quinson
3853 - Some documentation cleanups
3854 - remove the useless last argument of msgtype_declare
3855 - rename the Virtu functions to fit into the 'os' namespace
3856 - move headers src/include -> src/include/gras/ and stop fooling with
3857 gras -> . symbolic link
3858 - make distcheck is now successful
3860 2004-07-19 Martin Quinson
3862 - Build shared library also
3863 - Install html doc to the right location
3864 - stop removing maintainer files in make clean
3865 - build tests only on make check
3867 2004-07-13 Martin Quinson
3869 - No major issue in previous version => change versioning schema
3870 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3871 corresponding dataset.
3873 2004-07-11 Martin Quinson
3875 - Get it working with any kind of structure (we can compute the padding
3876 bytes remotely for all the architectures I have access to)
3877 - Implement the structure parsing macro (still not quite robust/complete)
3878 - Improvement to the remote testing toysuite
3880 2004-07-10 Martin Quinson
3881 [autoconf mechanism]
3882 - get ride of a bunch of deprecated macros
3883 - actually run the test for two-compliment, not only compile it :-/
3884 - test whether the structures get packed (and bail out if yes. Damn.
3885 Alignment is a serious matter)
3886 - test whether the structures get compacted (but respecting the alignment
3887 constraints of each types)
3888 - test whether the array fields of structures can straddle alignment boundaries
3890 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3891 (took me 2 hours to find that bug, looking at the wrong place)
3892 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3893 before sending/receiving and are used to compute the offsets of fields
3894 - Given that padding size depend even on compiler options, keep track of
3895 alignment and aligned_size only for the current architecture. Not a big
3896 deal since we send structure fields one after the other (seems
3898 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3899 program, let it work on linux/gcc/little32. portability todo.
3901 2004-07-08 Martin Quinson
3902 - import and improve remote compilation support from FAST
3903 - make sure make check works on half a dozen of machines out there
3905 2004-07-07 Martin Quinson
3906 Let's say it's version 0.0.3 ;)
3907 - Implement conversions (yuhu!)
3908 - Let it work on solaris (beside conversion, of course)
3909 - Stupid me, using rand() to generate the conversion datatests in not wise.
3911 2004-07-06 Martin Quinson
3912 - Let make dist work, since I'm gonna need it to compile on remote hosts
3913 - Let Tests/datadesc_usage write the architecture on which the file was
3914 generated as first byte.
3915 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3916 architecture descriptions.
3917 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3918 on those architectures.
3919 - Optimization: send/recv array of scalar in one shoot
3921 2004-07-05 Martin Quinson
3922 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3924 - Plug a whole bunch of memleaks
3925 - each process now have to call gras_{init,exit}. One day, their log
3926 settings will be separated
3927 - Continue the code factorisation between SG, RL and common in Transport.
3929 2004-07-04 Martin Quinson
3931 - Redistribution between SG and RL.
3932 We wanna have to accept in SG, so move accepted related parts of RL in
3933 the common part. (more precisely, the dynar of all known sockets is no
3934 more a static in transport.c, but part of the process_data)
3936 [gras_stub_generator]
3937 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3939 2004-07-03 Martin Quinson
3940 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3941 them all up in one shot)
3943 - Fix the ugly case of reference to dynamic array.
3944 - New (semi-public) function gras_datadesc_size to allow the messaging
3945 layer to malloc the needed space for the buffer.
3947 - gras_socket_close now expect the socket to close (and not its address to
3948 put NULL in it after it). This is because the socket passed to handlers
3949 is one of their argument (=> not writable).
3951 - propagate the interface cleanup from last week in datadesc, ie remove a
3952 superfluous level of indirection. User pass adress of variable
3953 containing data (both when sending and receiving), and not of a variable
3954 being a pointer to the data. Let's say that I like it better ;)
3955 The price for that is constructs like "int msg=*(int*)payload" in
3956 handlers, but it's a fine price, IMHO.
3958 - Let it work in RL (yuhu)
3960 2004-06-21 Martin Quinson
3962 - porting SG plugin and SG select to new standards (works almost).
3963 - plug memleaks and fix bugs around.
3966 - cleanup the prototype of data recv and force users to specify when they
3967 want to handle references to objects. Test case working even for cycles.
3968 - plug memleaks. Valgrind is perfectly ok with this.
3970 2004-06-12 Martin Quinson
3972 - cleanup the separation between plugin and main code in plugin creation
3974 2004-06-11 Martin Quinson
3976 - Reput hook for raw sockets, needed for BW experiments
3977 - kill a few lines of dead code
3978 [Data description] Interface cleanup
3979 - gras_datadesc_by_name returns the searched type or NULL.
3980 That way, no variable is needed to use a type desc once, which makes
3982 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3983 two parameters were strings (field name, type name), leading to
3985 [Dicos] Interface cleanup
3986 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3987 This is consistant with the dynar API.
3989 2004-04-21 Martin Quinson
3991 - Porting to new standards.
3993 - interface cleanup.
3994 There is no bag anymore, no need to take extra provision to mask the
3995 pointers behind "ID".
3996 Better splitup of functions between files create/exchange/convert.
3997 This is still a bit artificial since convert and receive are so
3998 interleaved, but anyway.
4000 - add a queued message list to procdata (the ones not matching criteria
4002 - factorize some more code between SG and RL wrt procdata
4004 - use gras_exit in example to track memleaks
4005 - get rid of gs_example now that GS is properly integrated into gras
4006 - update run_test to integrate the latest tests (datadesc)
4008 - rename WARNINGn macros to WARNn since it prooved error-prone
4010 2004-04-19 Martin Quinson
4012 - register init/exit functions within gras module mechanism
4013 - send/receive function.
4014 Convertion is not implemented, but short-cutted if not needed.
4015 struct/array elements are sent one by one (instead of block-wise), but
4016 nobody really cares (yet). Get a prototype before optimizing.
4017 - tests (using a file socket) for DD send/receive on:
4018 - base types: int, float
4019 - array: fixed size, string (ie ref to dynamic string)
4020 - structure: homogeneous, heterogeneous
4021 - chained list, graph with cycle
4022 Believe it or not, valgrind is not too unhappy with the results. The
4023 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4027 - Bugfix when using a filename explicitely (instead of '-')
4029 2004-04-09 Martin Quinson
4031 - factorize more code between RL and SG in socket creation
4032 - Complete the implementation and tests of:
4034 o file (only in RL, and mainly for debugging)
4036 I lost 3 days to design a portable address resolver, and then decided
4037 that the prototype mainly have to run on my box.
4038 Addressing portability too early may be like optimizing too early :-/
4040 - use gras_init in the Tests instead of the crappy parse_log_opt
4041 (the latter function is removed)
4042 [Conditional execution]
4043 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4044 [Code reorganisation]
4045 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4046 Build examples against the RL library, since there is no way to disable
4047 its creation for now.
4049 For information, the beginning of coding on GRAS was back in june
4050 2003. I guess that every line has been rewritten at least twice since