1 ----------------------------------------------------------------------------
3 SimGrid (3.24.1) NOT RELEASED YET (v3.25 expected December 22. 2029, 04:19 UTC)
6 - Actor: Merge signals on_migration_start/end into on_host_change
7 - Actor: Rename migrate() into set_host()
8 - Disk: Allow users to get the read and write nominal bandwidth values
9 - Exec: Implement wait_for(timeout)
12 - Parse errors now raise a simgrid::ParseError that you may want to catch.
15 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
16 been renamed to fix a spelling error. As usual, the previous definitions are
17 kept with a deprecation warning. However, to avoid runtime errors, it is now
18 forbidden to override these deprecated methods in a derived class. Please use
19 the new names immediately if you need to override them.
22 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
23 - New module: random, providing classical random numbers generators.
25 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
26 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
27 - GH#305: Unscheduled tasks are still excuted
28 - GH#323: Crash when an actor turn off his physical host
29 - FG!19: Removing RngStream
31 ----------------------------------------------------------------------------
33 SimGrid (3.24) October 9. 2019
35 The Clean Disk Release.
38 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
39 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
40 thus becomes the default resource to execute I/O operations. Examples
41 and the file system plugin have been modified to use disks instead of
42 storage elements. Storage can still be used but is doomed to disappear.
43 Users are thus adviced to update their simulators accordingly.
44 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
45 for consistency with pthread_barrier_wait()
46 - Host::get_englobing_zone() returns the englobing netzone
47 - Actor::on_destruction is now called in the destructor
48 Actor::on_termination new signal called when the actor terminates
50 - Global signals are now part of the Engine:
51 - on_platform_creation: after config settings, before the XML parsing
52 - on_platform_created: right after the XML parsing
53 - on_time_advance: each time the clock advances
54 - on_simulation_end: after simulation, before cleanups
55 - on_deadlock: as the name implies.
57 - sg_{actor,host,link}_{data,data_set}() now all exist.
58 Use them to attach user data to the object and retrieve it.
61 - Introduce an experimental Wifi model. It sounds reasonable
62 according to the state of the art, but it still has to be properly
63 validated, at least against ns-3 if not against reality.
64 - Improved the usability of ns-3. Several bugs were ironed out.
65 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
66 It is now expressed as 'idle:epsilon:allCores' to properly model the
67 consumption of non-whole tasks on mono-core hosts. More info in the doc.
70 - convert a new set of functions to the S4U C interface and move the old MSG
71 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
74 - Fortran bindings for DVFS have been removed.
75 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
76 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
77 - SMPI can now be selected by cmake's find_module(MPI) with
78 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
79 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
80 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
81 - trace-call-location can be used with TI traces, and replayed, and is compatible
82 with smpi/comp-adjustment-file.
83 - sleep events are now correctly traced/replayed.
84 - Default for trace-call-location is now to use file names and not full paths.
85 To revert to previous behavior (in case of collision of filenames), option
86 "smpi/trace-call-use-absolute-path" can be set to yes.
89 - Use the included xxHash as an hash implem when C++14 is usable.
90 - Option model-checker/hash was removed. This is always activated now.
91 - New option smpi/buffering controls the MPI buffering in MC mode.
92 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
93 This is useful to check for MPI compliance.
96 - New section on plugins: how to define a new one, and existing ones.
99 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
100 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
103 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
104 and <mount> tags. The rationale is finish to decouple what should be handle
105 by SimGrid kernel (i.e., resources and activities) from plugins built in
106 user space (e.g., the file system plugin). This new tag currently accepts
107 only three attributes (id, read_bw, and write_bw). All the other information
108 that was declared with the storage related tags now has to be expressed as
109 properties. An example of platform using this new tag is available at
110 examples/platforms/hosts_with_disks.xml
113 - 'expect signal' can now accept more than one potential signal.
115 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
116 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
117 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
118 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
119 - FG#31: per-actor data
120 - FG#34: SG_BARRIER_SERIAL_THREAD?
121 - FG#35: model-checker does not like buster-produced binaries
122 - FG!13: MC: complete workaround in the error msg seen on modern systems
123 - FG!15: execute_flops now logs compute
124 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
125 - FG!17: ns-3: unblock the right number of communications + others issues
126 - FG!18: Improving the performance of the ns-3 bindings
127 - GH#189: Energy consumption of parallel tasks
128 - GH#219: Error in the throughput of TCP transfer
129 - GH!330: Minor polishing of English: Configuring_Simgrid
130 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
131 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
132 - GH!334: Rework energy plugin, again
134 ----------------------------------------------------------------------------
136 SimGrid (3.23.2) July 8. 2019
139 - Nicer introduction page.
140 - Migrate the "Deploy your application" page to the new doc.
141 - Move Java as a subtree of MSG.
144 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
147 - Drop xbt_dynar_sort_strings().
150 - Really fix FG#26: Turning off a link should raise NetworkFailureException
151 - FG#27: Wrong exception thrown to wait_any when link is turned off
152 - GH#328: Java: Canceling multiple tasks in a single vm/host
154 ----------------------------------------------------------------------------
156 SimGrid (3.23) June 25. 2019
158 The Exotic Solstice Release.
161 - SunOS and Haiku OS support. Because exotic platforms are fun.
162 - Stop setting random seed with srand48() at initialization.
163 - Use addr2line as a fallback for stacktraces when backtrace is not available.
164 - Build option -Denable_documentation is now OFF by default.
165 - Network model 'NS3' was renamed into 'ns-3'.
168 - Simgrid can now hopefully be installed with pip.
171 - wait_any can now be used for asynchronous executions too.
174 - New log appenders: stdout and stderr. Use stdout for xbt_help.
175 - Drop xbt_dict_dump.
178 - 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
179 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
180 - MPI interface is now const correct for input parameters
181 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
182 - Fortran bindings for async collectives.
183 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
186 - Remove option 'model-check/record': Paths are recorded in any cases now.
187 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
188 always sparse. That's an immense gain in memory, and can even be
189 faster because of cache effects. So there is no need to clutter the
190 code to allow the user to go for the unefficient mode.
193 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
194 models proved to be more accurate than these old experiments.
196 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
197 - FG#1: Broken link in error messages
198 - FG#2: missing installation documentation
199 - FG#3: missing documentation in smpirun
200 - FG#6: Python bindings not available on PyPI
201 - FG#7: simple cmake call requires doxygen
202 - FG#8: make python bindings an optional dependency
203 - FG#10: Can not use MSG_process_set_data from SMPI any more
204 - FG#11: Auto-restart actors forget their on_exit behavior
205 - FG#12: -Denable_lto=OFF doesn't disable LTO
206 - FG#13: Installs unstripped file 'bin/graphicator'
207 - FG#14: Installs the empty directory 'doc/simgrid/html'
208 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
209 - FG#17: Dead link in doc (pls_ns3)
210 - FG#20: 'tesh --help' should return 0
211 - FG#21: Documentation link on http://simgrid.org/ broken
212 - FG#22: Debian installation instruction are broken
213 - FG#26: Turning off a link should raise NetworkFailureException exceptions
214 - GH#133: Java: a process can run on a VM even if its host is off
215 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
216 - GH#326: Valgrind-detected error for join() when energy plugin is activated
218 ----------------------------------------------------------------------------
220 SimGrid (3.22) April 1. 2019
222 The Easter Christmas Release (this one is somewhat late).
225 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
226 - Not complete yet: asynchronous activities (amongst others) are still missing
227 - Still ongoing: the interface may change in the future. We need more
228 testers! Please report any glitches.
229 - No new project using Java should start now. Please switch to Python.
232 - Some of the internal cleanups may lead to speed improvements:
233 - The hard limitation on the amount of simulated actors with
234 Java+Mac was removed. Now, the available memory is the only limit.
235 - Our refcounting was tidyied, leading to 10% speedups in some cases.
236 - We are still working on making our code robust to the actor kills
237 and hosts' churn. Things are improving, but it's not perfect yet.
238 - Replaced our own code to display a backtrace (that was forking addr2line)
239 with the Boost.Stacktrace library.
240 You won't see your backtraces without this optional dependency.
241 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
242 - Stop setting random seed with srand() at initialization.
245 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
246 XML file version remains 4.2 since old files are still compatible.
249 - Process termination which was broken at version 3.21 has been repaired.
250 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
251 - Hide the examples into examples/deprecated. New users should use Python.
254 - Drop MSG_process_create_from_stdfunc() from the API.
255 This C++ function was a pimple in the C API, made necessary at some
256 point by the Java bindings. This is fixed now.
257 - Hide the examples into examples/deprecated. New users should use S4U.
258 - MSG_process_create and MSG_process_attach now crash if the host used
259 to run this process is off.
260 - Fix the protype of MSG_process_on_exit()
261 Now use (int,void*) callbacks instead of (void*,void*) ones.
262 The implementation was ways too messy to actually work, I guess.
265 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
266 too often and provide more accurate timings.
267 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
268 - MPI_Alltoallw support
269 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
270 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
271 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
272 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
274 - Basic implementation of generalized requests (SMPI doesn't
275 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
278 - Drop sg_cmdline. Please use xbt_cmdline instead.
279 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
280 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
281 OsSemaphore is implemented in a portable way with C++11 threads.
282 This should allow much more threads to be created at the same time,
283 allowing Mac and Java users to simulate many more actors.
284 - Implement the 'thread' factory with std::thread instead of xbt ones.
285 It is not possible to set the stack size with threads anymore, but
286 -fsplit-stack is the way to go nowadays when using threads.
287 - Drop the xbt_os_thread_t module (now unused)
288 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
289 - Drop xbt_str_join_array().
290 - Drop cunit, use Catch2 instead.
293 - Many cleanups in the kernel::activity namespace. This was long
294 overdue, and shall open the path to many future endeavors.
297 - #132: Java: a process can not shut down its own host
298 - #220: S4U: detached send causes exception if sender terminates soon after sending
299 - #261: Document the parameters of parallel execution's constructor
300 - #300: [s4u] BarrierPtr is missing
301 - #314: SMPI args internal cleanup
302 - #316: Fix a bug related to the CPU utilization of multi-core VM
303 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
304 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
305 - #325: Turning off a host has different behavior on sleeping actors and computing actors
307 ----------------------------------------------------------------------------
309 SimGrid (3.21) October 3. 2018
311 The Restarting Documentation (TRD) Release.
314 - Convert the user manual to Sphinx for improved usability.
315 Unstable version now hosted on https://simgrid.frama.io/simgrid
316 Many glitches remain; stay tuned.
317 - Restrict the Doxygen reference API to the installed header files.
318 (documenting all internals this way is overkill + doxygen scales badly)
319 - New tutorial on S4U (highly inspired from the old MSG one, but with
320 a git repository to fork as a starting point, and a docker image)
321 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
322 - The Developper manual is still to be converted (not compiled until then)
323 - Some parts are still missing in this conversion (such as the
324 platform and deployment sections) while others were blindly converted
325 and would need more love (such as the configuration flags).
326 Things will certainly further improve in the future releases.
329 - Fully reimplement auto_restart mechanism that was utterly broken
330 (fix #22, #131 and #281 that were all very related)
331 - Implement semaphores (fix #303 and #291)
332 - When creating an actor from a function and its parameters,
333 move-only parameters are not allowed anymore, as it would prevent
334 the actor to be restartable if its parameters are consumed on run.
335 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
336 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
337 IO operations that can be started, waited for, or canceled as a regular
339 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
343 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
344 - New host variable: 'core_count'
347 - Remove xbt_os_thread_specific features
348 - Remove portability wrapper to condition variables
349 - Remove xbt_os_thread_yield()
352 - MPICH collective selector now mimics MPICH 3.3b
353 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
354 - MPI_Init(NULL,NULL) is now allowed
355 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
356 - Command-line arguments should now be handled properly in Fortran simulations
357 - MPI Attributes and related callbacks should work in Fortran
358 - Apps using C + Fortran should now work
359 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
360 - Support MPI_Error_String
363 - Due to an internal bug, Msg.run() must now be your last line.
364 We hope to fix it in a future release, and we are sorry for the inconvenience.
367 - #22: Process autorestart seem to only work with CAS01 cpus
368 - #56: Feature request: dynamic SMPI replay
369 - #93: simgrid should not eat --help
370 - #111: How to change the loopback link characteristics?
371 - #116: Platform cluster radical
372 - #117: Platform cluster/host order
373 - #129: Replay_multiple on steroid
374 - #131: Java : autostart does not work
375 - #143: Setting a breakpoint at a given time
376 - #235: Network energy models should be integrated
377 - #264: Add ptask L07 resource tracing
378 - #271: Dynamic and manual replay of SMPI traces
379 - #279: Breakpoints option uses time deltas instead of absolute time
380 - #281: Daemonized actor no longer daemonized after an auto-restart
381 - #282: TIT convertor erase traces when using absolute path in the trace list file
382 - #285: segfault when a s4u actor kills itself with msg process tracing activated
383 - #286: Pajé traces are not exposing the number of cores
384 - #287: Command-line parsing should stop when encountering '--'
385 - #288: MPI_Init(NULL, NULL)
386 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
387 - #290: Method to check if a mailbox has a message ready to be consumed
388 - #291: [s4u] semaphores do not seem available yet
389 - #294: Alltoall collective implementation of mpich
390 - #295: Starting a Host as OFF is not supported
391 - #296: DTD too permissive
392 - #299: Add a s4u barrier example
394 ----------------------------------------------------------------------------
396 SimGrid (3.20) Released June 24. 2018
398 The proxy snake_case() release.
401 - The whole API is now using snake_case() instead of mixing it with camelCase.
402 Compatibility wrappers in place for one year.
403 - We took the opportunity to increase the consistency of the API.
404 Please report any remaining glitches.
405 - New function simgrid::s4u::Host::get_actor_count:
406 Returns the number of actors running on a specific host.
409 - SMPI is now tested with ~45 proxy apps from various sources, with none or
410 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
411 - Replay: The replay file has been re-written in C++.
412 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
413 supported. They are stored in the trace and used when replayed.
414 - Basic support of MPI_Cancel. Robustness not guaranteed.
415 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
416 (only for MPI_COMM_TYPE_SHARED).
417 - New option: smpi/privatize-libs, to add external shared libs to be privatized
418 by SMPI. They will be copied locally and loaded separately by each process.
419 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
420 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
421 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
422 with internal index optind with multiple processes. Only works if MPI_Init has
424 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
425 syntax without preprocessing tricks.
428 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
429 behavior tracing is no longer limited to MSG
430 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
431 this means that traces now need to be updated or re-obtained; both Irecv
432 and Isend lines in your traces have now in total 5 mandatory fields:
433 <rankid> <command> <to/from rankid> <tag> <size>
434 To update your traces, it suffices to add a 0 for the tag here.
435 - TIT now also supports waiting for a distinct request via MPI_Wait.
436 Wait/Test now wait for a specific request, not just the last one that was
437 issued. This unfortunately means another update, because we need to
438 identify which request you want to wait for. We do this via the
439 triplet (sender, receiver, tag), which needs to be added:
440 <rankid> <command> <sender> <receiver> <tag>
441 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
442 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
445 - The deprecation of MSG is ongoing (but this should not impact you).
446 Many MSG functions are now simple wrappers on the C API of S4U. If
447 you wish to convert your code to S4U, find the S4U counterparts of
448 your MSG calls in src/msg/msg_legacy.cpp.
449 - MSG can still be used, but won't evolve anymore.
452 - Allow to run the Link energy plugin from the command line with
453 --cfg=plugin:link_energy
454 - Rename Energy plugin into host_energy
455 - Rename Load plugin into host_load
458 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
459 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
460 - Improve performance when handling timeouts of simix synchros.
463 - Config: the C API is now deprecated (will be removed in 3.23), and
464 the C++ API has been slightly improved.
467 - Fix several build issues on OSX.
468 - Move simgrid_config.h to simgrid/config.h (old header still working)
471 - #143: Setting a breakpoint at a given time
472 - #150: Inconsistent event names in SMPI replay
473 - #258: daemonized actors hang after all non-daemonized actors have completed
474 - #267: Linker error on unit_tmgr
475 - #269: SMPI: tracing of MPI_Wait/all/any broken
476 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
478 ----------------------------------------------------------------------------
480 SimGrid (3.19.1) Released March 22 2018
482 The Fixed ABI Release.
484 As you may know, we are currently refactoring SimGrid in deep.
485 Upcoming SimGrid4 will be really different from SimGrid3: modular,
486 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
488 Our goal is to smooth this transition, with backward compatibility
489 and automatic update paths, while still progressing toward SimGrid4.
491 SimGrid remains open during works: The last pure SimGrid3 release was
492 v3.12 while all subsequent versions are usable alpha versions of
493 SimGrid4: Existing interfaces remain unchanged, but the new S4U
494 interface is budding and the internals are deeply reorganized.
496 Since 2015, we work hard to reduce the changes to public APIs. When
497 we need to rename a public library symbol in S4U, we let your
498 compiler issue an explicative warning when you use the deprecated
499 function. These messages remain for four releases, i.e. for one full
500 year, before turning into an error. Starting with v3.15, your can
501 also adapt to API changes with the SIMGRID_VERSION macro, that is
502 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
504 Starting with this v3.19.1, our commitment to reduce the changes to
505 the public interfaces is extended from the API to the ABI: a program
506 using only MSG or SimDag and compiled against a given version of
507 simgrid can probably be used with a later version of SimGrid without
508 recompilation. We will do our best... but don't expect too much of
509 it, that's a really difficult goal during such profund refactoring.
511 The difference between v3.19 and v3.19.1 is that the former was
512 accidentally breaking the ABI of MSG, while the later is restoring
515 S4U and kernel APIs will still evolve until SimGrid4, with one-year
516 deprecation warnings as currently. In fact, cleaning up these
517 interfaces and converting them to snake_case() is one release goal of
518 v3.20. But don't worry, we are working to smooth this upgrade path.
520 In summary, new projects should start with S4U to benefit of the
521 future, but old MSG projects should still be usable with no change.
523 ----------------------------------------------------------------------------
525 SimGrid (3.19) Released March 20 2018
527 The Moscovitly-cold Spring release.
529 SMPI user-visible improvements:
530 - Support of MPI_Type_create_subarray
531 - Unimplemented MPI calls now abort (instead of issuing a warning)
532 Some harmless calls around errhandlers still only issue a warning.
533 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
534 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
535 is now possible (mmap privatization was refactored to this end).
536 Adding new SMPI instances should also be technically doable now.
538 SMPI internal cleanups:
539 - Removed index notion from SMPI -- no more "getPid() - 1"!
540 In tracing files, actors are called rank-<PROCESS-ID> but these are
541 NOT the real ranks (tracing several communicators is not supported yet).
542 - Deprecate smpi_process_index() (will be removed in 3.22)
543 - Better testing on classical proxy apps, through an external project:
544 https://github.com/simgrid/SMPI-proxy-apps/
547 - Use a graphical TOC to make it easier to find the documentation you need
548 - Revamp the MSG tutorial
551 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
552 - Remove parameter of MSG_process_killall().
553 Resetting the PID was bogus anyway (several actors could have the same PID).
556 - Execution->setHost() can be called after start(), to migrate it.
557 - Comm::test_any() is now implemented.
558 - s4u::Actor now has onCreation() and onDestruction() signals.
559 - Install some kernel header files for the users' plugins and more.
560 Warning, their API is really not stable yet.
561 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
562 include/simgrid/kernel/routing/*Zone.hpp
565 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
567 Build System and other cleanups:
568 - Remove unused run-time parameter "tracing/onelink-only".
569 - SimGrid now works with intel compilers (MC still buggy)
570 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
571 - tesh kills subprocesses on timeout (but not on windows)
574 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
577 - #194: Feature request: simgrid::s4u::Comm::test_any()
578 - #245: migrating an actor does not migrate its execution
579 - #253: Feature Request: expose clusters as objects
580 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
581 - #255: Tesh broken on Windows
582 - #256: Modernize FindSimGrid.cmake
583 - #257: Fix (ab)use of CMake install
585 ----------------------------------------------------------------------------
587 SimGrid (3.18) Released December 24 2017
589 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
591 MSG IS NOW DEPRECATED. Please use S4U for new projects.
592 - Support for MSG should not be removed from SimGrid before 2020,
593 but future evolutions will be limited to the integration of
594 user-provided patches.
597 - s4u::onDeadlock() signal, run before stopping the simulation.
598 - s4u::Exec: asynchronous executions (abstraction of background
600 - s4u::Host->getLoad() returns the achieved speed in flops/s
603 - Link::name() is deprecated and will be removed in v3.21, use
604 Link::getCname() instead.
605 - Mailbox::getName() changed to return a std::string, use
606 Mailbox::getCname() instead to get a char*.
607 - Storage::getName() changed to return a std::string, use
608 Storage::getCname() instead to get a char*.
609 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
610 to have both allocation and deallocation of the map in user space.
613 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
614 The amount of flops remaining to do is a vector, not a scalar.
615 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
616 name implies on both sequential and parallel tasks.
617 - Both changes fix GitHub's #223 using PR #237 as a basis.
618 Thanks Michael Mercier.
619 - Most examples were converted to S4U and hidden elsewhere at tests.
620 Rationale: we still want MSG to work; we want newcomers to use S4U.
623 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
624 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
627 - Switch to the faster dlopen privatization mechanism by default
628 - Documentation improvments
631 - Live migration is getting moved to a plugin. Dirty page tracking is
632 the first part of this plugin. This imply that VM migration is now
633 only possible if one this function is called:
634 - C/MSG: MSG_vm_live_migration_plugin_init()
635 - C/C++: sg_vm_live_migration_plugin_init()
636 - Java: Msg.liveMigrationInit()
637 For C and C++, "simgrid/plugins/live_migration.h" also has to be
641 - Define class simgrid::xbt::Path to manage file names.
642 - Removed unused functions:
643 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
644 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
645 xbt_graph_export_graphviz()
646 - xbt/str.h: xbt_str_join()
647 - Remove unused datatypes:
648 - xbt/heap.h: use std::priority_queue or boost::heap instead
649 - xbt/swag.h: use boost::intrusive::list instead
652 - New link_energy plugin for the consumption of the links.
653 - All of the operations on files and storage contents have been
654 packaged into a plugin (src/plugins/file_system). The current
655 public interface can be found in
656 include/simgrid/plugins/file_system.h
657 To use these functions you now have to initialize the plugin by
658 calling MSG_storage_file_system_init() just after calling
659 MSG_init() or sg_storage_file_system_init() just after creating
663 - Remove the undocumented/untested tag <include>
666 - Remove viva specific tracing as the tool is no longer maintained
669 - #248: Kill of finished processes leads to segfault
670 - #240: xbt_cond_wait_timeout should gracefully return for C apps
671 - #239: please implement signal s4u::onDeadlock()
672 - #230: segfaults when exit() without run()
673 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
674 - #223: MSG_task_get_flops_amount() not working with parallel tasks
675 - #222: Actor::kill() doesn't really kill and segfaults
676 - #221: odd LMM warning when killing an actor
677 - #120: Memory leak when the processes are forcefully killed
679 ----------------------------------------------------------------------------
681 SimGrid (3.17) Released October 8 2017
683 The Drained Leaks release: (almost) no known leaks despite the tests.
685 Even very long simulations will preserve your memory: our testsuite
686 stresses SimGrid for over 45mn on fast machines for a coverage of
687 over 80%, and there is only one single known leak, of about 4kb.
691 - Comm.detach(): start and forget about asynchronous emission. A cleanup
692 handler may be given to free resources if the comm cannot be completed.
693 - this_actor::send(mailbox) is now mailbox->put()
694 - New: simgrid::s4u::Comm::wait_all()
695 - New: Host.execute() for remote executions.
698 - Unused option network/sender-gap is removed.
701 - Deprecate MSG_task_isend_with_matching(): This unused feature
702 really complicates our internals. Will be removed in v3.20.
705 - Improved context termination. It is now possible to free resources, even
706 when a process is forcibly killed.
707 - ContextBoost: add support for Boost versions above 1.61.
710 - Bring back run-time option --cfg=exception/cutpath to remove exception
712 - Removed unused functions:
713 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
714 xbt_str_rtrim(), xbt_str_trim().
715 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
718 - Removed header files obsolete since SimGrid 3.12:
719 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
721 - #3: SD_exit should be made optional
722 - #120: Memory leak when the processes are forcefully killed
723 - #159: Threading test regression in Actor refcounting
724 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
725 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
726 - #186: Actor::killAll() segfaults if some process is blocked on wait()
727 - #191: VM migration and pstate
728 - #192: Updating the energy consumptions of all hosts crashes with VMs
729 - #195: All actors have PID=0 in the logs
730 - #204: Sometimes segfault with thread contexts and mmap privatization
731 - #222: Actor::kill() doesn't really kill and segfaults
732 - #225: Actor::kill() doesn't really kill when victims are doing a join()
734 ----------------------------------------------------------------------------
736 SimGrid (3.16) Released June 22. 2017.
738 The Blooming Spring Release: developments are budding.
741 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
742 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
743 - Java JAR file should be finally fully working on Mac OSX too.
745 XML platforms: Switch to platform v4.1 format.
746 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
747 - <zone> can be used as a synonym for the now deprecated <as>
748 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
749 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
750 - <actor> can be used as a synonym for the now deprecated <process>
751 - state_file and avail_file periodicity is now easier to express
752 (check the documentation) the old behavior should still work.
753 * Storage: not backward compatible, but it seems that nobody ever used it
754 - Remove attribute 'content_type' of <storage_type>: was never used
755 - Make attribute 'model' of <storage_type> optional: for future usage
756 - Remove Bconnection model property: was never (in)validated. Replaced by
757 the maximum of the read and write bandwidth as a resource constraint.
760 - New and Backwards Compatibility break:
761 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
762 When one wants to get the list of tasks whose states have changed during a
763 simulation round, s/he has to allocate and free a dynar and use it as
764 argument to this function. The former SD_simulate (double how_long)
768 - Allow multicore VMs, with the correct sharing (unless you overcommit)
769 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
772 - New function to update the consumption of all hosts at once.
773 - Fix the model for multi-core hosts, linear in the amount of busy
774 cores with an abnormality for pIdle (see paper).
775 BUG: Still not satisfactory for ptask on mono-cores.
778 - The netzone are now available from the MSG API.
779 The old names still work, but are now deprecated.
782 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
783 - New option: smpi/keep-temps to not cleanup temp files
784 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
785 mallocs mode is used, allows to change the size of the fake file used
786 (default 1MB), to potentially limit the number of mappings for large runs.
787 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
788 - Fortran ifort and flang compilers support
789 - New RMA calls supported (experimental):
790 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
791 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
792 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
793 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
794 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
795 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
798 - New: this_actor::isMaestro() returns whether we are in kernel mode.
799 - Behind the scene, ActivityImpl are now automatically refcounted.
800 Removing this nasty bottleneck will greatly simplify our internals.
803 - Replay: New function xbt_replay_action_get():
804 Retrieve the function previously associated to an event type.
805 - DROPPED FUNCTION: xbt_str_varsubst()
806 - DROPPED MODULE: strbuff. We don't need it anymore.
807 - DROPPED MODULE: matrix. We don't need it anymore.
808 - DROPPED MODULE: lib. We don't need it anymore.
810 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
812 SimGrid (3.15) Released March 22. 2017
814 The Spring Release: continuous integration servers become green
815 We fixed even the transient bugs on all target architectures:
816 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
817 Windows, FreeBSD, NetBSD.
819 - Rename NetCards to NetPoints.
820 This was intended to help NS3 users, but that's not a netcard.
821 That's a point in the routing algorithm, let's avoid wrong simplifications.
824 - New: MSG_process_yield(). Stop and yield to other processes.
825 - New: MSG_process_daemon(). Daemon processes are automatically killed
826 when the last non-daemon process terminates
827 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
828 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
829 to make room for the upcoming network energy plugin.
830 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
833 - Ensure that an actor can kill itself with Process::exit()
834 - Kill the obscure NativeException. Nobody want to survive the issues
835 it denotes, so use JniException that is a RuntimeException (not to
836 be caught explicitly).
837 - Partial bug fix in initialization. SimGrid flags on command line were
838 consumed at C level but stayed in the original Java String[] args.
839 This could mess users' args[i] if SG flags were not put at the end of
841 The SimGrid flags are now removed from the Java arguments. However,
842 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
843 if args.length is greater than the number of YOUR OWN ARGUMENTS.
844 It might be if you have --log or --cfg flags in the command line.
845 - Fix numerous memleaks all around the place. In particular, around VMs.
849 - s4u::onPlatformCreated: right before the simulation starts
850 - s4u::onSimulationEnd: right after the main simulation loop
851 - s4u::onTimeAdvance: right after a clock change (time is discrete)
852 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
853 event from the availability_file changes the avail speed.
854 - Links are now usable from s4u
855 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
856 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
857 - New examples: The conversion of MSG examples to S4U has begun
858 - Actors: create, kill, migration, and suspend
859 - Applications: master-worker and token-ring
860 - Action replay: communications and storage
861 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
864 - Backwards Compatibility breaks
865 - The SD_route_get_list and SD_route_get_size functions have been removed.
866 They are replaced by the unique
867 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
868 The route from src to dst is built in the links dynar whose size can
869 be obtained with xbt_dynar_length.
870 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
871 sg_host_route_bandwidth and sg_host_route_latency.
872 Macros ensure the backwards compatibility, but you should fix your code
875 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
876 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
877 - Bump our claim of support from MPI 1.1 to MPI 2.2.
878 We don't support 100% of it, but it should be enough. Contact us if not.
879 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
880 naive and probably inaccurate)
881 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
882 mapped onto a unique small file using some system magic.
883 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
884 - Fortran: cleanups, fixes, support of user-added operations
885 - RMA: MPI_Accumulate are applied in correct order.
886 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
887 - Better support for MPI_IN_PLACE
888 - Support for MPI_Win attrs and keyvals.
889 - Support MPI_Comm_group_create, MPI_Type_size_x
890 - MPI ops should be properly applied to all allowed types
891 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
892 algorithm by a non failing one from mpich.
895 - Kill the fifo data container: we don't use it anymore.
898 - A <cluster> can now be created with different speed values to represent pstates.
899 see examples/platforms/energy_cluster.xml for an example.
901 SimGrid (3.14.159) Released December 28. 2016
903 The previous release was only a crude approximation of a Pi release;
904 this one should be a bit better.
906 - Revert a last minute change that broke on Mac OSX.
907 - Fix the changelog and NEWS files.
908 - Improve the documentation, in particular of the routing module.
909 - Integrate some patches flying around in Debian.
911 SimGrid (3.14) Released December 24. 2016
914 * The whole documentation was reworked and reorganized. There is
915 still much room for improvement here, but we're on it.
918 * We now need python3 for our testing infrastructure.
919 * Model checking needs libevent
922 * Now works on FreeBSD
925 * Privatization now works on FreeBSD
927 * Privatization is now activated by default in smpirun
928 You can turn it off with -no-privatize if something goes wrong.
930 * Call-location tracing for SMPI.
931 You can add the exact location (filename / linenumber) of an MPI call to
932 your trace files and slow down or speed up the simulation between two
933 consecutive calls by using an adjustment file (see the documentation).
935 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
936 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
937 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
938 values were ignored in some cases. The timings of these functions can now
939 be significantly different.
941 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
942 smpi/running-power is renamed to smpi/host-speed
944 * smpi/grow-injected-times option to enable or disable multiplication of the
945 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
946 make simulation less precise (but also much faster).
948 * smpirun script should be (much) faster for large deployments.
950 * SMPI tracing: fixed issue with poor matching of send/receives.
952 * Replay: Fix broken waitall
954 New functions and features
955 * MSG_parallel_task_execute_with_timeout, to timeout computations.
957 Dropped / renamed functions and features
958 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
959 - MSG_mailbox_is_empty() -> Mailbox::empty()
960 - MSG_mailbox_front() -> Mailbox::front()
961 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
962 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
963 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
964 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
965 * Don't pass the free_f parameter to property related functions:
966 - MSG_host_set_property_value()
967 - MSG_as_router_set_property_value()
968 - MSG_storage_set_property_value()
969 * VM properties. Since msg_vm_t are msg_host_t, just use
970 MSG_host_get_property_value() and friends
971 * VM I/O related things:
972 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
973 - Unimplemented save/restore methods
974 * MSG_as_router_get_property_value() was redundant with
975 MSG_environment_as_get_property_value().
976 - Removed MSG_as_router_*propert*() functions
977 - Added MSG_environment_as_set_property_value() for consistency
978 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
979 Well, they are still there for now, but deprecated with a warning.
980 Please switch to xbt_dict_new_homogeneous() before this is removed
982 * Task affinity. Its intended behavior (that was very badly tested
983 and probably not really working) was deceiving what most users
984 would have hoped here.
985 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
986 * Option network/coordinates is now useless and should be dropped.
989 * Added option storage/max_file_descriptors to allow more than 1024 files opened
992 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
993 * Added Dragonfly topology support
996 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
997 EiBps, EBps, Eibps, Ebps for bandwidth.
998 They may become useful to some lucky ones.
1001 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1002 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1003 * ex/async_waitAll: new example, on asynchronous communications
1006 * Memory usage should be decreased for simulations with a large number
1007 of processes. This also helps for SMPI.
1009 SimGrid (3.13) stable; urgency=low
1011 The Half Release, a.k.a. the Zealous Easter Trim.
1013 - We removed half of the lines, that were mostly experimental cruft.
1014 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1015 (not counting blanks and comments -- according to openhub.net)
1016 - The internals are now compiled in C++ (and will soon be clean C++)
1017 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1018 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1020 * Backwards Compatibility breaks
1021 - Removed Lua simulation bindings (switch to C or Java for that).
1022 Lua can still be used to describe platforms
1023 - Removed Java kernel plug-ins.
1024 Will be reintroduced after the ongoing major internals reorg.
1026 - the following functions were removed.
1027 They were too specific and should be reimplemented in a generic
1028 way, with filter function.
1029 - MSG_task_listen_from_host
1030 - MSG_mailbox_get_count_host_waiting_tasks
1031 - MSG_mailbox_put_with_timeout was removed.
1032 Please use MSG_task_send_with_timeout instead.
1034 - the SD_application_reinit function was removed. It has been a noop for a while.
1035 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1036 replaced by a more flexible mechanism.
1037 - The following functions thus do not exist anymore
1038 - SD_workstation_get_access_mode
1039 - SD_workstation_set_access_mode
1040 - SD_workstation_get_current_task
1041 - Basic estimation functions have been removed but can easily be replaced
1042 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1043 - SD_workstation_get_computation_time => amount / sg_host_speed()
1045 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1046 Use VM.getSpeed()*load/100 for the legacy behavior.
1048 - option enable_tracing was removed. It was not doing anything for a while.
1049 - In the ModelChecker:
1050 - the model-checker now ptraces the model-checked process which means
1051 you cannot use a debugger on the latter anymore (we might make this
1052 optional in the feature);
1053 - removed soft-dirty page tracking;
1054 - remove model-checked side snapshot management,
1055 MC_snapshot() and MC_compare_snapshot();
1056 - keep the MC_cut() function as a stub (it was not really working
1057 in the previous release).
1060 * All options are consistently in kebab-case. Old names are kept as alias.
1063 * Switch to platform v4 format.
1064 - Rename from 'power' to 'speed' the attributes describing the amount of
1065 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1066 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1067 - In <host> and <link>, attributes availability and state are gone.
1068 It was redundant with state and availability traces, and with peak values.
1069 - In <cluster>, attributes availability_file and state_file are gone.
1070 It was too complex and unused.
1071 - Kill <gpu>. Was not doing anything.
1072 - The DOCTYPE points to the right URL:
1073 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1074 (the file at this address now documents the changelog since its v1)
1075 - A warning is emitted for unit-less values (they are still accepted).
1076 - speed. Default: 'f' or 'flops'. Also defined:
1077 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1078 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1079 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1080 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1081 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1082 - latency. Default: 's' second. Also defined:
1083 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1085 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1086 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1089 * s4u::Host is now the preferred public interface to the Host features.
1090 sg_host_* functions are C bindings to the exact same behavior
1091 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1094 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1095 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1096 They were deprecated since a few years, and probably did not even compile anymore.
1099 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1100 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1101 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1102 encouraged to update their codes. A list of the modified functions can be found at the end of
1103 include/simgrid/simdag.h
1106 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1107 the lambda or closure passed as a parameter will run in kernel mode.
1108 All the callback functions should be rewritten to that interface at some point.
1111 * Reorganizing and cleaning the internals all around the place.
1114 * Remove old default barrier/bcast buggy algorithms (see #18407)
1115 * Various bug fixes to handle more codes
1116 * Remove the need for the --foreground option of smpirun (it is still
1117 accepted for backward compatibility).
1120 * Kill the setset data container: MC don't use it anymore.
1121 * Kill the queue data container: it made more sense with GRAS.
1122 * Kill the xbt_peer_t data type: it's useless without GRAS.
1123 * Kill rm_cb feature of config sets: it was never useful.
1124 * Kill graphxml parsing feature. It was not used.
1125 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1127 - xbt_dynar_sort_strings(), when the content is char*
1128 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1129 They throw exceptions on invalid input;
1130 * C++ support for declaring CLI flags (simgrid::config::Flag);
1131 * class for abstracting different signal backends (simgrid::xbt::signal).
1132 with no external dependencies (we need very simple signals).
1135 * refactoring and cleanup of the code;
1136 * ongoing process to cleanly separate the model-checking algorithms
1137 from the code model-checking support.
1139 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1141 SimGrid (3.12) stable; urgency=low
1143 The Facelift Release.
1146 * Require g++ v4.7 at least to not speak prehistorical C++.
1147 * Require Boost 1.48 (for signal2 component).
1148 * Java must be version 7 at least when activated.
1149 * Builds on Windows again (including Java bindings).
1150 * Tracing is now always enabled (no way to turn it out)
1151 * Remove GTNetS. It was not working anyway.
1152 * Various cleanups in the cmake scripts.
1153 * Move headers around to sort them out on installed systems:
1154 - instr/instr.h -> simgrid/instr.h
1155 - instr/jedule/* -> simgrid/jedule
1156 - simdag/datatypes.h was removed
1157 - simdag/simdag.h -> simgrid/simdag.h
1158 - msg/datatypes.h was removed
1159 - msg/msg.h -> simgrid/msg.h
1162 * Interface improvement:
1163 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1164 - Sanitize the interface in MSG_task_ module:
1165 - Merge two functions that were close enough but misleading:
1166 set_compute_duration(t) -> set_flops_amount(t)
1167 get_remaining_computation(t) -> get_flops_amount(t)
1168 - set_data_size(t) -> set_bytes_amount(t)
1169 get_data_size(t) -> get_bytes_amount(t)
1170 - Massive cleanups in the functions related to the energy
1171 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1172 - New: MSG_host_get_pstate()
1174 - msg/energy/onoff: switching hosts on and off
1176 * Interface improvement:
1177 - Rename Host.isAvail() to Host.isOn()
1178 - Rename Process.currentProcess() to Process.getCurrentProcess()
1179 - Rename Task.setDataSize() to Task.setBytesAmount()
1180 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1182 - #18874: Actually allows the GC to reclaim tasks
1186 - SIMIX_process_throw: raises an exception in a remote process
1187 * Refactoring: Separate sub-modules
1188 - libsmx: the public interface, as libc in a real system
1189 - popping: the strange dance that converts a user request into a kernel handling
1190 - smx_context_*: the virtualization mechanisms that embed the user code
1191 - smx_*: the handling of each simcalls
1192 * Interface cleanup:
1193 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1194 * Rename smx_action_t into smx_synchro_t, making explicit that these
1195 things are used to synchronize processes with their environment.
1196 For example, a communication is a sort of synchronization involving
1197 the communicating processes (that may block until the exchange) and
1198 the platform. The same can be said from computations, etc.
1200 - #18888: segfault when a process ends before its kill_time
1204 - 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.
1205 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1206 - MPI_Type_set_name, MPI_Type_get_name
1207 - MPI_*_c2f and MPI_*_f2c functions
1208 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1209 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1210 - Activate a lot of new tests from the mpich 3 testsuite
1212 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1213 - InfiniBand network model added: Based on the works of Jerome Vienne
1214 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1215 - When smpi/display_timing is set, also display global simulation time and application times
1216 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1217 * Collective communications
1218 - 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
1219 - 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/).
1220 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1222 - Replay now uses algorithms from wanted collective selector
1223 - Replay can be used with SMP-aware algorithms
1224 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1225 - 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.
1227 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1228 - Correctly use loopback on fat-tree clusters
1229 - Asynchronous small messages shouldn't trigger deadlocks anymore
1230 * Energy/DVFS cleanup and improvement
1231 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1232 - new: smpi_get_host_pstate()
1235 - "Full" network optimization flag was broken since Surf++
1236 - Better handling of precision flags in maxmin
1237 - Fix bug causing sometimes "Impossible" errors
1238 - Properly pass cluster properties to included hosts
1239 * Improvement of the Energy plugin.
1240 - Always update the consumption before returning that value
1241 - New property: watt_off to denote the dissipation when the host is off
1242 - New functions getWattMinAt and getWattMaxAt to retrieve the
1243 dissipation of pstates that we are not currently at.
1244 * Java: class NetworkLink renamed to Link
1245 * New function: simcall_process_get_kill_time()
1246 * Massive rename s/workstation/host/
1247 - That's intrusive, but that's good for the project consistency. Sorry.
1248 - Change config option "workstation/model" into "host/model"
1251 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1252 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1253 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1254 - xbt_dirname and xbt_basename for non-POSIX systems
1256 * The model checker now runs as a separate process.
1257 * The model checker runs must now be launched with the new simgrid-mc program.
1258 * Record/Replay: the MC can display a textual representation of a path in the
1259 execution graph. It can then be replayed outside of the model checker.
1261 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1263 SimGrid (3.11) stable; urgency=low
1268 * Normalizing pointers addresses tool for better diff between logs
1271 * Add cloud examples using new VMs
1272 - examples/msg/cloud/two_tasks_vm.tesh
1273 - examples/msg/cloud/simple_vm.tesh
1274 - examples/java/cloud/cloud.tesh
1275 - examples/java/cloud/migration/migration.tesh
1276 * Add java surf examples:
1277 - examples/java/surfPlugin/surf_plugin.tesh
1278 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1279 - examples/java/surfCpuModel/surf_cpu_model.tesh
1280 * Add SMPI+MSG example:
1281 - examples/smpi/smpi_msg_masterslave/
1286 - msg task destroy cancel test
1287 - msg_host on/off test
1288 * Move all tests in testsuite to teshsuite (adding tesh files)
1289 * Restructure teshsuites
1290 - one folder for each kind of test
1291 * Restructure AddTests.cmake
1293 - structure the order of tests (with sections)
1296 * Add virtual machine
1297 - creation of a VM on a PM
1298 - migration of a VM from a PM to another PM
1300 - MSG_process_join(msg_process_t process, double timeout)
1301 - msg_bar_t MSG_barrier_init(unsigned int count)
1302 - int MSG_barrier_wait(msg_bar_t barrier)
1303 - void MSG_barrier_destroy(msg_bar_t barrier)
1304 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1305 * New option "msg/debug_multiple_use" to help debugging when a task is used
1308 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1310 - Increase code coverage in test suites
1314 * Protect context stacks against stack overflow. The number of protected memory
1315 pages allocated on the top of each stack (1 by default) can be configured
1316 with the new command line option --cfg=contexts/guard_size:<value>.
1317 * Simcalls are now generated by a python script that
1318 - generates files included by SimGrid
1319 - checks that all the functions exist, and proposes prototypes
1321 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1322 host_set_data, host_get_data
1324 - simcall_process_join(smx_process_t process, double timeout)
1325 * Fix bug where sleeping processing could not be suspended.
1328 * Translate surf models from C to C++
1329 - Generic classes for all models: Model, Resource, Action
1330 - A generic interface for each kind of model (CPU, Network, Storage
1331 Workstation, WorkstationVM)
1333 * Translate surf routings from C to C++
1334 * Add callbacks using sigc++ or boost::signals2
1335 - Add callback functions for resource creation/destruction
1336 - Add callback functions for action state change
1337 - Handle Energy as a plugin
1338 * Replace swag by boost::intrusive
1339 * Add new routing models for clusters. For documentation, see
1340 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1341 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1342 parameters for cluster tag
1343 - Fat trees, with topology="FAT_TREE" and
1344 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1345 - see examples/platforms/torus_cluster.xml and
1346 examples/platforms/fat_tree_cluster.xml
1347 * More documentation
1350 * Hostfiles support host:nb_processes construct to deploy several processes on
1352 * Collective communication algorithms should not crash if used with
1353 improper number of nodes and report the error.
1354 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1355 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1356 MPI_Dims_create, MPI_Cart_sub are supported.
1357 * New interface to use SMPI programmatically (still depends on MSG for
1358 some parts, see examples/smpi/smpi_msg_masterslave):
1359 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1363 * Global variables privatization in MPI executables is now performed at runtime
1364 with the option smpi/privatize_global_variables (default:no).
1365 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1366 dynamic libraries loaded by the application are not privatized (static
1367 linking with these libraries is advised in this case)
1370 * Options defined in XML work correctly now.
1373 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1374 copied into simgrid.jar or not (ON by default). Use this option if you want
1375 to reduce the size of the installed simgrid.jar, *and* the native libraries
1376 are kept installed elsewhere.
1377 * Surf binding with SWIG (code generated in maintainer mode only):
1378 - plugin to handle callbacks
1379 - CPU model only for the moment
1382 * Supernovae build mode is definitively removed. It was used to improve
1383 inlining and inter-module optimizations. It is nowadays superseded by
1384 link-time optimizations commonly available in compilers.
1385 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1386 * Add boost dependency for surf++
1387 * Add new macro for tests
1388 - ADD_TESH(name <tesh_args>)
1389 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1393 - xbt_bar_t XBT_barrier_init(unsigned int count)
1394 - int XBT_barrier_wait(xbt_bar_t barrier)
1395 - void XBT_barrier_destroy(xbt_bar_t barrier)
1396 * Make the xbt_os_time module public
1398 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1400 SimGrid (3.10) stable; urgency=low
1402 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1405 * Reintegrate Java to the main archive as desynchronizing these
1406 package is not acceptable anymore (Java is now considered stable)
1407 * Add explicit synchronization facilities through semaphores
1408 * Bug fix: Task.setDataSize() only changed the C world, not the value
1409 cached in the Java world
1412 * Dramatically change the way files are handled. API and internals changed, but
1413 this part of MSG was not considered as production grade either.
1414 * Add explicit synchronization facilities through semaphores
1415 * Add a new function MSG_host_get_process_list()
1416 * Preliminary DVFS support (see examples/msg/energy/ for details)
1419 * SMPI is now included directly in the libsimgrid as the windows
1420 linker doesn't force us on splitting it anymore.
1421 * Improvements of the SMPI replay tool:
1422 - Most of the collective communications are now rooted in the same process as
1423 in the original application.
1424 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1425 used until now). Multiple data types can now be used in a trace.
1426 - The replay tool now supports traces produce either by TAU or a modified
1428 - Bug Fix: the compute part of the reduce action is now taken into account.
1429 - Gatherv collective is now supported
1430 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1431 of smpirun outputs time independent traces for the current run. One file
1432 is created per process. If too many processes are simulated, this behavior
1433 can be changed to one file for all processes by using the
1434 tracing/smpi/format/ti_one_file flag
1435 * smpirun generates the host file if needed (with given host count and platform)
1436 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1437 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1438 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1439 --cfg=smpi/coll_selector:(mpich/ompi)
1440 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1442 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1443 happening inside a collective SMPI call.
1444 * Fix the behavior of complex data types handling.
1445 * Make MPI_Wtime another synchronization point to take computations into
1447 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1448 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1449 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1450 (no privatization of global variables yet)
1451 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1454 * Verification of liveness properties is now available for SMPI applications
1455 (in addition to MSG applications)
1456 * Bugged examples using SMPI in examples/smpi/mc/
1457 * Add --cfg=model-check/visited option. Allows the verification of infinite
1458 programs. Detection of loops in the execution thanks to the system state
1459 comparison and reduction of the state space to explore. Can be combined with
1460 DPOR for safety properties.
1463 * Allow to change SimGrid configuration (see --help) within the code
1464 thanks to SD_config() as it can be done in MSG.
1465 * Add a new function SD_task_set_amount() upon user request.
1468 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1469 * Remove rule based routing (no more PCRE dependency)
1470 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1471 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1472 * Add a loopback_bw and loopback_lat options to cluster tag.
1473 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1474 routing. Now peers are encapsulated in an AS and have their own private
1475 router but this is transparent.
1478 * Our own implementation of getline is renamed xbt_getline, and gets
1479 used even if the OS provide a getline(). This should reduce the
1480 configuration complexity by using the same code on all platforms.
1481 * New type: xbt_cfg_elm_boolean.
1482 * Allow to use yes/no for boolean configuration options in the command line.
1483 * Allow to disable SimGrid cleanups at exit from command line option.
1484 There are situations where one may want a simulation to end with an exit.
1485 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1486 annoying when scripting around the simulator. Adding a
1487 --cfg=clean_atexit:no allows to circumvent this issue.
1490 * Lots of memory leaks were corrected in this release.
1491 There is no known memory leaks anymore, in all of our 600+ tests.
1492 * New command line option --version, to get SimGrid version information.
1493 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1495 * Supernovae builds are deprecated, and expected to be removed in the next
1498 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1500 SimGrid (3.9) stable; urgency=low
1502 The Grasgory release: GRAS is really dead now.
1504 * Complete overhaul of the internal host structures scheme.
1507 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1508 was considered as experimental and badly maintained since then.
1509 * Keeping it was thus a trap to our potential users, that could take
1510 it instead of MSG or SMPI by mistake despite is pity state.
1511 * GRAS seems to have very few users (if any), and no one volunteered
1512 to maintain it further. It also induces a lot of XBT code (for
1513 portability sake), that must be maintained too.
1514 * For all these reasons, we killed GRAS. If someone wants to revive it
1515 in the future, don't cry, our git history still remembers of GRAS.
1518 * Major overhaul. Merge our documentation again as time proved that
1519 splitting it was really not helping our users.
1520 * Further improve the developer documentation to help newcomers
1521 hacking on SimGrid itself. The user documentation (and in
1522 particular, the beginner documentation) is still in a sorry state.
1525 * Now works on Windows too!
1526 * Much more extensive test suite, from MPICH
1529 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1530 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1531 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1532 * Bug fix: let task be scheduled when the last dependency to be solved is
1533 a control dependency.
1534 * Remove SD_load_environment_script function.
1535 Use the C sg_platf function if you want to declare a platform
1539 * New function: MSG_process_get_number()
1540 * Old function documented: MSG_config()
1541 * Remove MSG_load_platform_script function
1542 Use the C sg_platf function if you want to declare a platform
1546 * Change the default value of the TCP_gamma constant (maximal size of TCP
1547 congestion window) to a more realistic 4MiB value. If you notice changes in
1548 your simulation results, you can fall back to the previous 20k tiny window
1549 by adding --cfg=network/TCP_gamma:20000 on command line.
1550 * (Hopefully) fix a bug wrt periodic availability and state traces
1551 * Bug fix: use default values at start when first event in availability/state
1552 trace is not at time 0.
1555 * remove the "new_" part of function name sg_platf_new_trace_connect
1556 (resulting in sg_platf_trace_connect), since it does not create
1560 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1561 use case where it's really mandatory, and maintaining it was a pain in
1563 * New: xbt_fifo_search(), search an item with a user-provided
1564 comparison function instead of dumb pointer comparison.
1567 * Fix the lua deployment:
1568 Use `simgrid.init_application()` before deployment instead of
1569 `simgrid.msg_register_application()` after.
1572 * Transfer the tracing files into the corresponding modules.
1574 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1576 SimGrid (3.8.1) stable; urgency=low
1578 The "we are told that some people want to also *install* the simgrid
1581 * Add missing file "tesh.1" to the archive.
1583 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1585 SimGrid (3.8) stable; urgency=low
1587 The Psssshiiiit release: SimGrid jumps into the Cloud.
1590 * Add an experimental interface to manipulate VMs. They are mainly
1591 process groups with very few intrinsic semantic, but they should
1592 allow you to build the semantic you want easily.
1593 * New function: MSG_host_set_property_value()
1594 * New function: MSG_process_on_exit(). To clean memory in all cases.
1595 * Bug fixes that made the host (and link) failures unusable.
1596 * Add a way to auto-restart process when the host in which they are
1597 executing comes back (ON_FAILURE="RESTART" on deployment file,
1598 MSG_process_auto_restart_set).
1599 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1600 please stop using the old ones, they are DEPRECATED.
1602 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1603 Please use MSG_init() instead. (reducing the amount of entry
1604 points in the library helps us).
1605 * Make it impossible to link against the wrong version of the lib
1606 * Deprecate MSG_clean(). No need to call it anymore.
1607 * Function MSG_get_host_number() is not deprecated anymore.
1610 * Split the doc into a user guide and a reference guide.
1611 * Start a developper guide to help people hacking on SimGrid.
1614 * Enable tracing by default. This modules rocks you should use it.
1615 * Remove option custom_flags. Now use environment variables CFLAGS
1617 * Use default cmake things to detect lua instead of home grown ones.
1618 * New option "enable_mallocators" to disable mallocators, for debugging
1619 purpose ("on" by default).
1622 * Bug fixes around the resource failures: don't let the processes
1623 survive the host they are running onto.
1624 * Add an interface to auto-restart processes when the host in which they are
1625 executing comes back.
1626 * Ensures that SIMIX_clean is called automatically. It's not part of
1627 the public interface anymore (bindings should be updated).
1630 * Bug fix for when SD_Simulate is called with a positive value: be careful
1631 when comparing doubles. Sometimes they are different for non significant
1633 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1634 parallel task whose initial work is distributed among host according
1635 to the Amdahl's law. Such tasks are created with a parameter alpha
1636 that corresponds to the non-parallelizable part of the computation.
1637 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1638 between two sets of workstations assuming a 1D block distribution (each
1639 workstation owns a similar share of data) on both sides.
1641 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1642 Data redistribution will be automatically scheduled once parent and child
1643 are both scheduled. The filling of computation_amount and
1644 communication_amount structures is now done seamlessly thanks to the chosen
1646 * New function SD_workstation_dump to display various information
1647 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1648 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1649 bandwidth on the route onto which the task is scheduled and the amount of
1651 To divide the nominal bandwidth by 2, the rate then has to be:
1652 rate = bandwidth/(2*amount)
1653 * Compute tasks that have failed can now be rescheduled and executed again
1654 (from their beginning)
1655 * Increasing source code coverage (src/simdag is now covered at 95.8%
1659 * Re-implement time-independent trace replay using SMPI (at the
1660 smpi_smp_* level) instead of MSG. This should replace
1661 examples/msg/actions/actions.c
1662 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1663 hindexed and structs)
1664 * Implement the exchange of non-contiguous data.
1665 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1666 * Correct behavior of smpi/sender_gap and set its default value to 0
1667 * Add option to asynchronously send small messages to allow better
1668 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1669 specifies the size in bytes under which messages will be asynchronously sent.
1670 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1671 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1674 * New C interface to define a platform: XML is now optional.
1675 For more info, please check include/simgrid/platf.h.
1676 * New interface to define random platforms from the C:
1677 For more info, please check include/simgrid/platf_generator.h and
1678 examples/msg/masterslave/masterslave_platfgen.c
1679 * Export a sg_cmdline dynar containing all the arguments we got from
1683 * Two new tracing options for adding comments to trace file so you
1684 can track your experiments (see --help-tracing for details).
1685 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1686 * Adding the SimGrid version that generated the trace file as a comment.
1687 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1688 * Fix to avoid key clashes on Paje links
1689 * Other minor fixes related to the Paje specification
1692 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1693 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1694 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1695 * Greatly improve the robustness of mmalloc to user errors (such as
1696 using an area after freeing it, or freeing it twice)
1698 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1700 SimGrid-java (3.8.1) stable; urgency=low
1702 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1703 Warning: all methods are visible, but only some of them are
1704 implemented so far. Check the source (src/jtrace.c)
1705 for further information.
1706 * New module: org.simgrid.msg.File (SimGrid File management functions)
1707 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1708 * Change the meaning of Process.restart: now restart the process from
1709 the beginning, like MSG_process_restart in C.
1710 * Add Process.setAutoRestart: handling of process restart when failed
1712 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1713 you to retrieve the properties of the processes/hosts
1714 * Deprecate Msg.clean(): you can just forget about it now.
1715 * New function Process.getCount(), that only works when compiling
1716 with the not yet released version 3.9 of the C library.
1718 * New context factory based on Coroutines. It mandates a modified JVM
1719 but then, the simulations run about five times faster, and there is
1720 no limit to the amount of processes (beside of the available memory).
1722 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1724 SimGrid (3.7.1) stable; urgency=low
1727 * Restore the prototype of MSG_process_create_with_environment() to
1728 the pre-3.7 situation by removing the kill_time argument.
1729 * Add a MSG_process_set_kill_time() function instead.
1732 * Fix weird behaviors when dealing with parallel tasks.
1735 * Simgrid is now built as a dll.
1736 * Simgrid-java now works on Windows.
1737 * Simgrid-Java is now included into Windows package.
1740 * First pre-build package for MacOSX.
1743 * Fix compilation when using MSG_USE_DEPRECATED.
1744 * Fix some compilation issues on Macs and Windows.
1745 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1746 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1748 We discovered that the Lua console is broken, but we are missing the
1749 manpower to fix it right now. The problem existed in 3.7 too, so we
1750 are not blocking the release for that. Sorry if you depended on this
1751 feature, any help would be really welcome.
1753 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1756 SimGrid-java (3.7.1) stable; urgency=low
1758 The "Java aint got to be bloated and slow" release
1761 * Various internal cleanups and performance improvement
1762 Simulations are expected to run up to twice faster or so
1763 * Make Process.kill(process) an instance method, not a static one
1764 * User processes are not java.lang.Thread subclasses.
1765 This breaks the compatibility (sorry), but previous API was
1766 brain-dead, making it impossible to have non-trivial
1767 initializations in the process constructor.
1768 * Require a full constructor per Process sub-class.
1769 Kinda breaks the compatibility (sorry), but this allows a much more
1770 efficient way to launch the processes at simulation startup.
1771 * Do not embeed our version of semaphores, java 1.5 can be considered
1772 as sufficiently prevalent for us to not dupplicate its features.
1776 * Add examples for almost every part of the API
1777 We spotted and fixed a lot of bugs in the process
1779 * New module: asynchronous communication API
1780 * New function: Process.sleep()
1781 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1782 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1783 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1785 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1787 SimGrid (3.7) stable; urgency=low
1789 The "spring cleanups (before the next Big Project kicks in)" release.
1792 * We can specify the SMPI latency/bandwidth factor with command line
1793 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1794 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1795 You can also use the "config tag" from platform file by adding this line
1796 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1797 (see "example/platforms/tag_config.xml" to use "config tag").
1798 Note that the command line supersedes the platform file configuration.
1799 * Change the correction factors used in LMM model, according to
1800 the latest experiments described in INRIA RR-7821.
1801 Accuracy should be improved this way.
1802 * Use the partial invalidation optimization by default for the
1803 network too. Should produce the exact same results, only faster.
1804 * Major cleanup in surf to merge models and split some optimization
1805 mechanisms from the core of the models. As a result you can now
1806 specify which model to use (e.g., --cfg=network/model:LV08
1807 --cfg=cpu/model:Cas01) and which optimization mode to use
1808 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1809 Incompatible combinations should err at initialization. See
1810 --help-models for the list of all models and optimization modes.
1811 * The CLM03 workstation model was dropped for simplicity because it
1812 used the deprecated CM02 network model. Use default instead.
1813 * Rename the TCP_gamma configuration option to network/TCP_gamma
1814 * Rename the coordinates configuration option to
1815 network/coordinates, and document it
1816 * Use now crosstraffic keyword instead of the terribly misleading
1817 fullduplex keyword. It is activated by default now in the current
1818 default model, use --cfg=network/crosstraffic:0 to turn it off.
1819 * Ongoing refactoring the model parsing to make XML files optional
1820 See include/simgrid/platf.h for details (still to be completed)
1823 * Major overhaul of the documentation. Almost instructive now :/
1824 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1825 functions and friends. This interface was considered as
1826 deprecated since over 2 years, it's time to inform our users that it is.
1827 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1828 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1829 use these (crappy) functions in your code.
1830 These functions will be removed soon. Stop using them now.
1831 * Deprecate MSG_get_host_{table,number}
1832 Implement MSG_hosts_as_dynar() instead.
1833 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1834 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1837 * Stabilize the parallel execution mode of user contexts
1838 * Introduce configuration variables to control parallel execution:
1839 - contexts/synchro: Synchronization mode to use when running
1840 contexts in parallel (either futex, posix or busy_wait)
1841 - contexts/parallel_threshold: Minimal number of user contexts
1842 that must be part of a scheduling round to switch to parallel
1843 execution mode (raw contexts only)
1844 * Fix bugs that prevented to use suspend/resume along with
1845 synchronization structures.
1846 * Fix bugs in process termination that lead to invalid memory access
1847 in very specific conditions.
1850 * Introduce a parallel mode for the models (controlled by surf/nthreads
1851 configuration item). In our tests, running the models in parallel
1852 never lead to any speedups because they are so fast that the gain
1853 of computing each model in parallel does not amortizes the
1854 synchronization costs, even when ultra fast futexes are used.
1855 This is released anyway because YMMV.
1858 * Performance boost by using a swag internally to compute the set of
1859 tasks that are finished and should constitute the return value of
1863 * Enable it by default now that it is considered rather stable.
1866 * Documentation of the tracing functions.
1867 * Performance gains when tracing categorized/uncategorized resource
1868 utilization by avoiding calls to get route when updating resource
1869 variables. LMM constraints are being used instead.
1870 * API changed to set task categories. Use MSG_task_set_category instead
1871 of TRACE_msg_set_task_category, and SD_task_set_category instead
1872 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1873 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1874 * Examples for link user variables added (see at examples/msg/tracing/)
1875 * Deprecated function TRACE_msg_set_process_category completely removed
1876 * Trace header updated according to the latest Paje file format
1877 * Tracing network lazy updates, no longer obligate users to use full updates
1878 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1879 * Experimental: let user code declare/set/push/pop application states for hosts
1880 * API changed to allow the manual creation of graph configuration files
1881 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1884 * Improve the API of Lua MSG bindings, using the Lua spirit.
1885 * Each simulated process now lives in its own Lua world (globals are
1886 automatically duplicated). It helps writing simulators. Will allow
1887 to run Splay programs within SimGrid in the future.
1888 * Add a Chord example in Lua, equivalent to the MSG one.
1891 * Start the implementation of a solution to express temporal
1892 properties, not only local assertions. This is still an
1893 experimental work in progress, stay clear from it to be safe.
1897 - Add new runtime parameters --help-logs and --help-log-categories
1898 to display information about supported logging parameters and
1900 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1901 don't exists anymore.
1902 * Mallocators: allow value NULL for the reset function.
1904 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1905 create homogeneous dictionaries, where all the elements share the
1906 same free function. Non homogeneous dictionaries will be
1907 deprecated in the next release.
1908 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1909 - Multi-level dictionaries are deprecated.
1911 - new function xbt_dynar_search_or_negative() that is useful when
1912 you have less than 2 million elements in your dynar and don't
1913 want of the extra complexity of catching exceptions when the
1914 element is not found.
1916 - Make xbt_os_thread module (for thread portability) public.
1917 Documentation is still to come, sorry.
1919 - Cleanups and simplifications to make it maintainable again.
1920 - Exotic features (such as memalign and valloc) were removed.
1921 - The metadata were extended and improved so that the
1922 model-checker becomes able to explore and inspect the heaps.
1923 - This may induce a performance drop when enable_model-checking is
1924 ON in cmake (even if it's not used in the simulation), but it is
1925 necessary at this point to get MC working.
1927 Turn model-checking OFF if simulation performance matters to you.
1928 Not enabling it at runtime is not enough, disable it in cmake.
1930 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1933 SimGrid (3.6.2) stable; urgency=low
1935 The "Not coding new stuff allows to polish old things" release.
1938 * New bindings to the NS3 packet level simulator (experimental)
1939 * Use the raw (efficient) execution contexts instead of the sysv
1940 (portable) ones when possible.
1941 * libpcre is now mandatory in any cases since not using it led to
1942 severe performance loss and possibly other issues
1943 * Update the XML platforms:
1944 - G5K: include the latest machine in Nancy
1945 - GridPP and LCG: new platforms
1946 * Documentation was partially updated, at least (more to come)
1948 Bug fixes, cosmetics and small improvements
1949 * Free terminated processes before the end of the simulation to avoid
1950 exhausting the memory of users having very dynamic amount of
1952 * Bug fix and cosmetics about canceling non-running tasks
1953 * Bug fix about the dot loader's issues when using libcgraph
1956 * Create an installer for windows with nsis (amd64 and win32)
1957 - Add an hello world project to illustrate simgrid project creation.
1958 - Embed libpcre into the Simgrid installer to avoid
1959 its compilation burden
1960 * The raw execution contexts should work on Apple now
1961 * Port to Windows 64 bits
1962 - Sysv contexts now have an implementation for this arch
1963 - GRAS communication features now support this arch
1964 * Drop support for borland compiler on windows
1965 - this code was not maintained, and we kinda depend on gcc nowadays
1966 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1967 * Fix portability issue on unstable ubuntu: linker became picky on
1970 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1972 SimGrid (3.6.1) stable; urgency=low
1974 The "Oops, we broke Macs too" release
1977 * Fixed contexts detection so that raw ones are used when possible
1978 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1979 leads to a strange error, with user code segfaulting sometimes when
1980 the generated code is not perfectly aligned (which is not
1981 controllable from the user side, depends on the amount of code)
1984 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1986 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1988 SimGrid-java (3.6) unstable; urgency=low
1991 * Split of every thing from simgrid v3.5 into a separate package.
1993 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1995 SimGrid (3.6) stable; urgency=medium
1997 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2000 * Bindings now constitute their own package, separated from the main one.
2001 Rationale: reduce our maintenance nightmare by reducing the module coupling
2002 They will soon be released on their own on gforge.
2004 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2005 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2007 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2008 * It's not quite deprecated for now because we have no replacement,
2009 but it may soon become the case.
2012 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2013 * Fortran: New user-level cache variable to store the rank of the running
2014 process. This improves performance by an order of magnitude.
2015 * C: New coccinelle script to automatically locate and modify global and
2016 local static variables.
2017 * Improved SMPI network model with a sender-side gap to account for multiple
2021 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2022 only say if a communication is finished, no matter whether it succeeded or
2023 failed. You can call MSG_comm_get_status() to know the status of a finished
2025 * New function MSG_task_dsend() to send a task and detach it. When a
2026 communication is detached, you are never notified of its success or failure
2027 and the memory is released automatically once it is finished. This function
2028 is useful when you don't care about the end nor the success of a
2030 * Change the prototypes of action replay. Sorry for inconvenience,
2031 but this is really more efficient this way (and to adapt your code,
2032 you just have to fix the initialization, that shouldn't be too long)
2033 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2034 ever managed to do anything useful with it.
2035 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2036 takes a dynar of communications. It returns immediately and gives the
2037 index of a finished communication (if any).
2038 * New example: a basic implementation of the Chord P2P algorithm.
2041 * New model for multi-core CPUs. You can now use the core attribute to
2042 precise the number of cores of a host. This is a basic model. Every
2043 process running on the host receives at most the power provided in
2044 the DTD (throughput<=power). Total throughput of process cannot exceed
2046 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2047 router linked by an up-link and a down-link (possibly asymmetrical).
2048 This kind of pattern allows to easily build last-mile model style platforms.
2049 Aggregating such patterns in a rule-based AS is thus the technique of
2050 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2051 * New model for Vivaldi routing. We transformed the Vivaldi network model
2052 into a Vivaldi routing model (based on the rule-based model). This allows to
2053 combine Vivaldi based latencies with last-mile platforms.
2056 * Added a check for NaN of IEEE754 infinite in the double entries of
2058 * Introduce a new context factory "raw", highly inspirated from the
2059 ucontext factory, but using manually crafted functions in assembly to
2060 do the work in an efficient manner.
2061 * Allow to change the used context factory at run time, not only at
2062 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2063 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2064 contexts at runtime (only with raw contexts or ucontexts).
2065 * Completely rewrote this module to allow parallel execution of user
2066 processes. Use --cfg=contexts/nthreads:N to execute user processes
2067 with N parallel threads (the default is 1, meaning no parallelism).
2068 * Allow to decide dynamically between sequential and parallel modes.
2069 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2070 processes in parallel only when their number is greater than or equal to P
2072 * Added a check for NaN of IEEE754 infinite in the double entries of
2076 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2077 won't output the state of processes when interrupted with Ctrl-C
2078 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2079 NULL-terminated array. This may solve backward compatibility issues
2080 due to the change to return type of SD_simulate. See also:
2081 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2082 * Add new macros with variable number of arguments.
2083 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2084 - in xbt/asserts.h: xbt_assert
2085 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2086 - in xbt/ex.h: THROWF and RETHROWF.
2087 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2089 * Change xbt_die() to accept a format string with arguments, just like printf.
2090 * New data structure: xbt_lib_t, like a dict but more general and with better
2094 * New configuration options
2095 Options triva/categorized and triva/uncategorized can be used to generate
2096 graph configuration files for Triva visualization tool.
2097 * Configuration option tracing/platform is renamed to tracing/categorized
2098 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2099 * New TRACE_user_link_variable interface:
2100 User provides the name of the link and the tracing variable to attach to it
2101 * the declaration of tracing categories must be done after the environment
2103 * simpler tracing interface, just one way to declare categories
2104 TRACE_category or TRACE_category_with_color, it is up to you
2105 * links in the trace file are again identified by their names
2106 * trace contains the full platform hierarchy exactly as declared using the ASes
2107 * Options tracing/msg/[task|process]:1 groups the process by hosts
2108 for both cases, tasks and processes must have names that are unique during the simulation
2109 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2110 * The experimental option tracing/msg/volume is deprecated
2111 its functionality may be reincorporated if needed
2113 The tracing generates a trace file with unordered timestamped events,
2114 because of the way the core simulator (surf) works. A script available
2115 at the tools directory (fix-paje-trace.sh) can be used to put the events
2116 in order. We have changed the tracing so it can generate ordered timestamped
2117 events in the final trace, but depending on the simulator (and how much time
2118 is simulated) that can lead to a huge memory utilization. It is deactivated
2119 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2121 Build Infrastructure
2122 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2123 - We are trying hard to keep the API stable, but it may happen that
2124 some things change (we're a research project after all, not a
2125 nuclear plant operating system). If such things should happen, you
2126 could rely on that macro to adapt.
2127 - current value: 30600 for 3.06.00, aka 3.6
2128 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2129 a number that can be compared with SIMGRID_VERSION.
2130 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2131 assertions and verbose/debug logging events are disabled at compile time.
2133 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2135 SimGrid (3.5) stable; urgency=medium
2138 * New feature to any SimGrid-based simulator: Model-Checking
2139 Check SIN#1 for more details.
2142 * New Model SMPI (three-interval linear regression for correction factors)
2143 See RR-7426, available at http://hal.inria.fr/inria-00527150
2144 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2145 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2146 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2147 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2148 * See also src/smpi/README
2153 - Tracing API changes: TRACE_start and TRACE_end should not be called
2154 by user-code. They are automatically called by simulators created
2155 with SimDAG, MSG and SMPI if the toolkit is compiled with
2156 tracing_enabled=ON. Categories declaration and utilization remain the
2157 same for MSG and SimDag.
2158 - A function was added to the tracing API to declare categories with
2160 - TRACE_category_with_color (char *category, char *color)
2161 where color must be in the following format
2162 "%f %f %f", red, green, blue
2163 and red, green, blue are float values in the interval [0, 1]
2164 - User can specify NULL as color parameter, or continue calling
2165 TRACE_category (cat)
2166 On that case, the tracing system will define random colors
2167 - The following command-line options are supported:
2168 --cfg=tracing/filename:msg.trace
2169 --cfg=tracing:1 (activate tracing, needed to use others)
2170 --cfg=tracing/platform:1 (categorized resource use)
2171 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2172 --cfg=tracing/msg/task:1 (task creation)
2173 --cfg=tracing/msg/process:1 (process creation, migration)
2174 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2175 --cfg=tracing/smpi:1 (SMPI interface tracing)
2176 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2177 - examples of examples/msg/tracing updated
2179 - DAXLoader and DOTLoader functions can generate tasks with categories
2180 - A new function to attribute a category to SD tasks:
2181 TRACE_sd_set_task_category (SD_task_t task, char *category)
2182 * Tracing the MPI interface implemented by SMPI
2183 - Collective operations are traced with states
2184 - Point-to-Point operations are traced with states/links
2185 - Tracing activated by a parameter "-trace filename" passed
2186 to smpirun during execution (considering that simgrid
2187 is compiled with tracing enabled)
2188 - To run the simulation with gdb, the simulator
2189 accepts --cfg=tracing/smpi:1 to trace SMPI
2190 - tesh files to check if smpi tracing is ok
2191 - See examples/smpi/NAS/DT-trace
2192 * GTNetS tracing re-worked
2193 - adaptation to the tracing system of GTNets to cope
2194 with modifications regarding the fullduplex mode
2195 - new tesh files to check if gtnets tracing is ok
2198 * Asynchronous communications through the functions:
2199 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2200 * New function: MSG_load_platform_script()
2201 to make possible using a lua script instead of XML files to set up platforms
2202 * New function: MSG_set_function
2203 to associate functions to processes, used when bypassing the parser
2204 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2206 Platforms: Add some more examples in examples/platforms
2207 * Grid'5000: see www.grid5000.fr
2208 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2211 * Change the XML format. This is a very important modification. SimGrid
2212 3.5 introduces a new hierarchical format based on the notion of
2213 Autonomous Systems. Compatibility with old format is ensured
2214 through the perl script provided in the install bin directory
2215 bin/simgrid_update_xml.
2216 It is now possible to build platforms with specific routing
2217 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2218 connect several platforms together. We will try to provide soon
2219 set of realistic platforms exploiting these properties (have a
2220 look at examples/platforms/ for the moment).
2221 * Take the opportunity of the XML format change to be a good XML citizen:
2222 rename link:ctn to link_ctn and similar changes (also dealt with by
2224 * Add a new routing scheme (rule-based) using regular expressions. It
2225 enables to have an extremely low memory footprint when the
2226 underlying routing is simple and can be compactly described. You
2227 need to have libpcre4-dev (perl regular expressions) installed if
2228 you want to use this routing scheme.
2229 * Revive the cluster TAG and allow to easily and efficiently (both in
2230 term of memory and speed) connect clusters together. Have a look
2231 at teshsuite/simdag/platforms/ to see how this can be done. With
2232 this tag, you can create clusters with thousands of tasks at no
2233 cost (have a look at examples/platforms/).
2234 Note: clusters are implemented as ASes, so there is no need for an
2235 enclosing AS tag if you have only one cluster in your platform.
2236 * Add new generic functions in the public interface that allows the user
2237 to call SURF 'create_resource' methods from your code (same
2238 functionality as the XML bypass mechanism but with a much lighter
2240 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2241 that uses a piecewise linear approximation to produce better
2242 results when exchanging small messages.
2243 * Add a new parameter to handle correctly full duplex link and account
2244 for interferences between uplink and downlink communications
2245 (activate with --cfg=fullduplex:1).
2248 * Rename the SD_READY (all dependencies are satisfied and task is
2249 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2250 dependencies are satisfied) state.
2251 This prevents a confusion between the notion of "ready to schedule"
2252 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2253 simulated" (SD_RUNNABLE) used by the simulation kernel.
2254 * Change the way a task is considered as ready. Instead of removing
2255 dependencies when a task is done, a counter is decreased. This way,
2256 it is always possible to reach ancestors thanks to the
2257 SD_taks_get_parents function (even after the end of the simulation.)
2258 * Change the return type of SD_Simulate from (SD_task_t*) into
2259 xbt_dynar_t. This function was in handling a dynar internally and
2260 converted it into a NULL terminated array for historical reasons.
2261 * New function SD_dotload(char*) to load a DAG described in dot
2262 format. This loader and the corresponding examples require the
2263 installation of the graphviz library.
2264 * Fix a bug in the management of tasks of size 0 in the surf network
2265 models. This problem was only visible with SIMDAG and you should
2266 thus disregard results produced with earlier versions if you
2267 relied on this feature (some tasks were blocked because of this).
2268 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2269 even though your applications were doing only point-to-point
2270 communications and sequential computations. Now you can really use any
2271 model you want (of course, if you create real parallel tasks, which are
2272 not implemented in most models beside ptaskL07, this will abort).
2273 * Add an example that schedules a DAX on an heterogeneous platform
2274 using a Min-Min strategy.
2275 * New function SD_workstation_get_current_task() that returns the kind
2276 of task currently running on a workstation in the sequential access
2278 * Raise some warnings when unexecuted tasks remains at the end of the
2279 simulation. This is usually caused by cycles in the DAG.
2282 * New function: SIMIX_process_set_function() called by MSG_set_function
2283 * Change the underlying waiting queue in semaphores so that a process
2284 can wait on several of them simultaneously (as in waitany).
2285 * Fix the way to handle tokens in semaphores so that all access patterns
2286 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2287 * kill the dirty pimple SIMIX_message_sizes_output()
2288 Please use (proper) visualization instead
2291 * New data container: setset (set of sets of elements)
2292 * New module: mmalloc (mapped malloc, allowing to have several
2293 independent segments of malloc)
2294 * New function: xbt_dict_cursor_set_data()
2295 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2296 * New function: xbt_dynar_is_empty()
2297 * New function: xbt_fifo_get_last_item()
2298 * Fix xbt_dynar_shrink(): use the right element size.
2299 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2300 uninitialized areas during expand.
2301 * Fix semaphores: previous implementation was severely broken.
2302 * Use library init/fini functions for our initialization.
2303 - you can use logs and other feature as soon as you want in your
2304 code (even before the xbt_init / MSG_init)
2305 - xbt_exit is now a no-op and produce a warning when used.
2308 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2309 benefit from the latest improvement to the simulation kernel.
2310 * Kill measurement sockets for now. If you rely on them, sorry. This
2311 release is not for you. This feature will be reintroduced in the
2312 future, but we cannot delay the release any further.
2313 * New function: gras_msgtype_get_name().
2314 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2315 * Fix (at last) the pmm example: it should not randomly fail anymore.
2317 Build chain: bug fixes and overall polishing
2318 * Cmake is now stable enough. Hence, we killed the autotools.
2319 * Port to windows ( TM :)
2320 * Fix the 'make install' target.
2321 No need to use 'make install-simgrid' anymore
2322 * Introduce a 'make dist' target compiling a *source* archive
2323 'make package' compiles a binary archive
2324 * Compile java files only on need
2325 * Add --cd and --setenv command line options to tesh
2326 * Out of source builds are not fully supported yet, but we are close
2327 * Enable supernovae and optimization flags by default for our users
2330 * Add layer to set up environment directly from lua, without XML.
2331 * The effect of gras_stub_generator can be achieved through
2332 lua too (check examples/gras/console/ping_generator.lua)
2334 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2336 SimGrid (3.4.1) stable; urgency=low
2338 The "Polishing easter eggs is probably a good idea" release.
2339 This is a bug fixes release only.
2343 * Fix a bug preventing the tasks from begin garbage collected.
2346 * Fix a bug occurring when a host involved in a communication fails.
2347 This was not detected properly by the other peer involved in the
2348 communication. Now, it's reported as a network error.
2351 * Warn the user about loop dependencies in data flow of DAX files
2352 * Obey the control-flow dependencies of DAX files
2355 * Add option "enable_smpi" allowing to not compile SMPI.
2356 Probably useful for the (Mac) users experiencing a build error here
2357 * Improve the detection of lua5.1 and ruby1.8
2359 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2361 SimGrid (3.4) stable; urgency=low
2363 The "Easter in Cargese" release. Also known as (major changes):
2365 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2366 ~> bindings were greatly improved
2367 ~> new tracing infrastructure for better visualization introduced
2369 * the "Welcome to configury modernity" release.
2370 ~> we switched from autotools to cmake, and improved our cdash
2373 A more detailled list of changes follow (full detail in svn log).
2375 Java Bindings: Various Cleanups
2376 * (install java-gcj-compat-dev on debian-like to use them)
2377 * Remove put/get: no need to export deprecated interface in Java
2378 Use send/receive instead.
2379 * Cleanup the examples and add a README per directory
2380 * Remove example autoDestination (that's the only way to go now)
2381 * Remove example explicitDestination (was a plain copy of basic)
2382 * Make JniException a runtime exception, so that there is no need to
2383 declare the fact that you may encounter such a beast. I guess that
2384 nobody will ever want to survive such error.
2385 * Create specific errors for each MSG case of failure:
2386 host failure, transfer failure, timeout, task cancelled
2387 * Cleanup the exceptions that may get thrown by each function
2388 * Other internal cleanups in Java bindings. Performance still bad :/
2389 Ruby and Lua Bindings: create them
2390 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2391 * That's new and great, you should try them out.
2392 Same functionalities than Java bindings, only even less polished
2394 * Kill the useless "rate" argument of SD_task_get_execution_time()
2395 Everyone used to provide -1 as a value, it was not used, and the
2396 semantic of a possible use wasn't even clear.
2397 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2398 either as comm costs or compute costs to mean that there is no
2399 such thing for that specific task.
2400 * Add a SD_task_set_name() function
2401 * Fix SD_task_unschedule() on typed tasks
2402 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2403 * In DAX loader, accept useless 'level' attributes to <job> since
2404 LIGO DAGs have them (seem to be to ease graphical representation).
2406 * Add an example masterslave_mailbox.c using send/receive and not
2407 the deprecated put/get interface.
2408 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2409 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2411 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2412 (declare MSG_USE_DEPRECATED to still have the old name)
2413 * Add a MSG_task_set_data() function
2414 * About trace replay (see examples/msg/actions):
2416 - Allow to work with splitted trace files for each process
2417 Give the specific trace file as argument of each process,
2418 and call MSG_action_trace_run(NULL)
2419 You can still have one merged file for all processes.
2420 - Fix implementation of collective operations
2421 * Allow task_execute() on 0-sized tasks (closes #10063)
2423 * This is the first release of SimGrid where SMPI is not considered
2424 beta anymore (even if some corners should still be improved)
2425 * Port over the new SIMIX_network submodule (internal refactoring)
2426 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2427 * Implement more missing elements of the standard:
2429 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2431 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2432 MPI_Scan() MPI_Get_processor_name()
2433 - Added implementation of missing case for Alltoall (warning: it's
2434 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2435 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2436 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2437 SMPI_MPI_Allgatherv()
2438 * Bug fixes include:
2439 - MPI_Waitsome() was broken
2440 - Allow relative includes in smpicc
2441 - Command line cfg argument 'reference_speed' was ignored...
2442 - Some functions did not properly lead to auto-benching of user code
2443 - smpicc passes -O2 by default (just like openmpi one)
2445 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2446 * Bug fixes about timeouts during communications
2447 * add SIMIX_message_sizes_output() as a pimple to write to file the
2448 amount of messages per size. Use gnuplot to get histogram.
2449 Pimple because that's the only user-visible function of simix,
2450 defined directly in xbt.h (irk, sorry)
2452 - Add a SIMIX_sem_get_capacity() function
2453 - Fix interactions with processe resume/suspende
2454 - release_forever() was stupidly broken
2455 - Fix SIMIX_display_process_status() for processes in a semaphore
2456 - Make SIMIX_sem_block_onto() user-visible
2457 * Refactoring context stuff:
2458 - Use pseudo-OOP for better modularity
2459 - reimplement SIMIX_process_kill() without process_schedule() so
2460 that the latter can take as invariant that it is called from
2462 - Merge context_start into context_new for sake of simplicity
2464 * Add a Vivaldi network model, coded live during SUD'10 ;)
2465 * Rename configuration variables to start a hierarchy:
2466 o cpu_model -> cpu/model
2467 o network_model -> network/model
2468 o workstation_model -> workstation/model
2469 * New configuration variables:
2470 o network/bandwidth_factor: correction to bandwidth
2471 o network/latency_factor: correction to latency
2472 o netwotk/weight_S: correction to the weight of competing streams
2473 * Add a long description to the models, that users can see with such
2474 argument on the command line: --cfg=cpu/model:help
2475 * --help-models display the long description of all known models
2477 * config: add the ability to set a default value after registration
2478 Does not override any previously set value (e.g. from cmd line)
2479 * dict: allow to have integer key and data.
2480 When so, you need to use the following functions
2481 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2482 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2483 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2484 In contrary to regular dicts, the key is not malloced before copy.
2485 Mixing scalar and regular elements in the same dict is not tested
2487 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2488 Tracing for Visualization:
2489 * SimGrid is now instrumented in order to generate a trace file for
2490 visualization analysis: to use it, need to compile SimGrid with the
2491 "tracing" option enabled, and instrument the program using SimGrid with
2492 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2493 (among other functions).
2494 * The instrumentation only traces the platform utilization for now
2495 * Documentation to use the tracing functions and how to analyze the
2496 traces with the Triva tool is written.
2497 * More information about: SimGrid FAQ (in the section Tracing Simulations
2500 * We moved to cmake as default build system. Autotools support will
2501 be dropped soon. Check the FAQ for more info about how to use it.
2502 * Greatly improved our cdash/ctest interactions
2503 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2504 * Added memory checking tests with valgrind; lot of memleak fixing.
2505 This may be the first release of simgrid with so few memory issues
2506 * Added code coverage tests.
2507 Our coverage is still improvable, but at least we see it on cdash.
2509 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2511 SimGrid (3.3.4) stable; urgency=low
2513 The "Desktop Grid needs love too" release (also called Xmas release).
2515 Models improvements:
2516 * Major speedup in the maxmin system solving by using lazy evaluation
2517 Instead of solving completely the maxmin system at each iteration,
2518 only invalidate (and recompute) the modified parts.
2519 This new feature is enabled in default models but you can try to
2520 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2521 * Cas01 IMproved as default CPU model
2522 This CPU model is the same Cas01 model, but it uses the
2523 maxmin-selective-update flag and a heap structure to manage
2524 actions on SURF kernel.
2525 It reduces the complexity to find the next action to finish and,
2526 consequently, it's faster than the old Cas01.
2527 This is the new default CPU model (Cas01).
2528 * Rename the old Cas01 model to Cas01_fullupdate
2529 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2530 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2531 * CpuTI (CPU Trace Integration)
2532 A new CPU model whose objective is simulate faster when using
2533 availability trace files.
2534 Instead of using a full featured, over engineered maxmin system for
2535 CPU modeling, this model does the pre-integration of traces files
2536 to calculate the amount of CPU power available, and so, executes
2537 faster than the old CPU models.
2538 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2539 * Use LV08 as default network model since it gives better accuracy
2540 for small messages and shouldn't change things for big ones.
2541 Use --cfg=network_model:CM02 to get the previous behavior.
2544 ******************************************
2545 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2546 ******************************************
2547 * The new CPU model may changes simulations!
2548 The point is that events occurring at the exact same timestamp
2549 are not scheduled in the same order with the old and new
2550 version. This may be enough to completely change the execution
2551 of simulations in some cases.
2552 * The new network model will change simulations!
2553 This new model is more realistic than the previous one, so you
2554 should consider redoing your old experiments with this model.
2555 Sorry for the inconvenience.
2558 * Introduce the supernovae compilation mode
2559 When compiled that way, the whole SimGrid (or almost) is put in a
2560 single compilation unit and compiled in one shoot.
2561 This is to help gcc which has difficulties to inline stuff from one
2563 The speedup seem to be above 15%, althrough more tests are needed on
2564 amd64 to confirm that gain.
2567 * Port of MSG's mailbox on top of SIMIX network
2568 The put/get mechanism was greatly simplified on the way.
2571 * New SIMIX network module. Provides:
2572 - Mailbox: rendez-vous mechanism to find with who you want to speak
2573 - Synchronous send/recv: easier and hopefully faster since the
2574 logic is handled in the maestro process directly now
2575 - Asynchronous send/recv: you dreamt of it? It's here now
2576 Too bad that nobody cared enough to propagate the change to MSG.
2577 * Add semaphores as SIMIX synchronization mechanism.
2580 * new function SD_daxload(char*) to load a DAX file
2581 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2582 * Introduce typed tasks. Specify its kind and cost at creation.
2583 At scheduling, just give where it should be placed, and the cost
2584 for each involved resource is automatically computed.
2585 Existing constructors so far (more to come of course):
2586 - SD_task_create_comm_e2e() for end-to-end communication
2587 - SD_task_create_comp_seq() for sequential computation
2588 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2589 * new function SD_task_dump() for debugging display
2590 * new function SD_task_dotty(task,FILE*) writing to file the info
2591 about the task in dotty format
2592 * SD_task_dependency_exists() can now cope with having one of its
2593 arguments NULL. If so, it tests whether the other argument has any
2595 * Add getters on list of preceding/following tasks:
2596 SD_task_get_parents(task) and SD_task_get_children(task)
2597 * Add getters on amount of workstations and list:
2598 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2599 * Add getter on task kind: SD_task_get_kind(task)
2600 * Update the start_time and finish_time of tasks on completion/failure
2601 * Bugfix: Remove task from state swags when destroyed
2604 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2607 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2608 dynar with one separate thread per value of the dynar.
2609 * Change the prototype of xbt_thread_create(), sorry.
2610 Added a boolean parameter indicating whether we want to join this
2611 thread (used in SG only for now)
2612 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2615 * GTNetS wrappers should now be usable again (and betterly tested too)
2616 * Fix a major regression from 3.2 where the timeout provided to
2617 MSG_task_put_with_timeout() was used as absolute time before which
2618 the comm should be done.
2619 * Start to fix the <cluster> tag.
2620 - Internal links should be good now (beside of the loopback, which
2621 use the private link instead)
2622 - paths to the external world is still rather broken
2623 - the <route:multi> tag is just broken. Actually that's brain-dead.
2624 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2625 to make it less stupid
2626 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2627 * Fix a source-level compatibility glitch from 3.2: after defining
2628 MSG_USE_DEPRECATED, you can use the old name
2629 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2630 * Allow to compile from the SVN with automake 1.11
2631 * Fix some problems when using the "start_time" tag in deployment XMLs.
2632 * Fix #8569: XBT/synchro.h has redundant declarations
2633 * Fix #8563: MSG return values and exceptions
2634 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2635 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2636 Thanks to Robson Peixoto.
2637 * Fix a vicious bug in dictionaries inducing that some elements were
2638 not freed on xbt_dict_free()
2640 Portability report of this version:
2641 * Main portability targets:
2642 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2643 on (amd64/i386/ia64)
2644 - mac leopard on i386
2645 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2646 but nothing critical.
2647 * Other platforms: windows, AIX and others were not tested for this release
2649 Timing report of this version:
2650 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2651 scenario parameters). From 8h to a few seconds in desktop grid settings.
2652 * Supernovae brings about 25% speedup on i386.
2654 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2656 SimGrid (3.3.3) stable; urgency=low
2658 The "Need for Speed" release.
2660 The timings done to validate the 3.3.2 were faulty.
2661 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2663 The problem was a conversion from a manually handled vector to
2664 xbt_dynar_t on the critical path.
2665 xbt_dynar_foreach calls functions, inducing stack management crap.
2667 We inlined these functions and xbt_dynar_foreach is now breath taking.
2668 We also inlined xbt_swag_belong on the way.
2670 Here are some approximate speedup measurements (on master/slaves
2671 simulations lasting between 10s and 20s each):
2672 3.3.1 -> 3.3.2: about same performance
2673 3.3.2 -> 3.3.3: 40% speedup
2674 3.3.1 -> 3.3.3: 40% speedup
2675 3.3.1 with inline patch -> 3.3.3: 30% speedup
2677 Our reading is that the refactoring which occurred in 3.3.2 made us
2678 suffer much more from the xbt_dynar_foreach low performance, but
2679 once we solved this, this refactoring proved to be very performance
2680 effective. From the 40% speedup, somehow, 10% are due to the
2681 inlining and 30% to the refactoring.
2683 That's a pitty that gcc cannot inline functions placed in other files
2684 alone. We have to choose between:
2685 - break the encapsulation (by putting private data structures and
2686 accessors in headers files to help gcc)
2687 - live with low performance
2688 - switch to a decent compiler such as icc (not quite possible).
2690 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2692 SimGrid (3.3.2) stable; urgency=low
2694 The "Simplicity does not preceed complexity, but follows it" release.
2696 The main contributors of this release were (lexical order):
2697 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2700 * Extract the routing logic into its own object.
2701 (was dupplicated in network.c and workstation_LV07.c;
2702 Allows to implement other ways of storing that info)
2703 => kill now useless network_card concept
2704 - Use dynar to represent routes (instead of void** + int*)
2705 - kill link_set (use surf_network_model->resource_set instead)
2706 - Add a command-line option to choose the routing schema to use
2707 - Add three new models:
2708 * Floyd (shortest path computed at initialization)
2709 * Dijikstra (shortest path recomputed all the time)
2710 * Cached Dijikstra (shortest path computed on need)
2711 All these models where contributed by Silas De Munck, and are
2712 described in his ICCS09 paper.
2714 * Simplify model declaration
2715 (less redirections, less function to write when defining a model)
2716 - Factorize stuff between models:
2719 surf_model_resource_set(model)
2720 surf_model_resource_by_name(model, name)
2721 - Unify the types of models in s_surf_model_t (using an union)
2722 - Embeed fields of common_public directly into s_surf_model_t
2723 - Rename model methods:
2724 action_free ~> action_unref
2725 action_change_state ~> action_state_set
2726 action_get_state ~> action_state_get
2727 - Change model methods into functions:
2728 (model)->common_public->action_use ~> surf_action_ref
2730 * Implement a generic resource; use it as ancestor to specific ones
2731 (allows to kill duplicated code in models)
2732 Drawback: timer command don't need no name nor properties;
2733 workstation_CLM03 don't need no properties
2734 (but I guess we can live with those few bytes wasted)
2736 * Improve the action object model
2737 - implement a constructor avoiding dupplicated code about field
2738 initialization in generic_action part.
2740 * Kill the SDP model: it has an external dependency, is deprecated
2741 in flavor of modern lmm models, and didn't compile since a while
2744 * Relocation of the context module from XBT to SIMIX.
2745 (the context were decoupled from the simix processes, duplicating a lot of code)
2746 => a lot of code was factorized
2747 - less overhead is introduced during scheduling
2748 - simpler API for the context factory
2749 - the logic for process creation,destruction and manipulation was simplified
2750 * Simplification of the s_smx_process_t data structure.
2751 => accesing the simix level data associated to a process is faster now,
2752 and the code is a lot more readable.
2755 * Implement some more MPI primitives:
2756 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2757 -implementation: Bcast: flat or 2-ary tree (default),
2758 Barrier: 4-ary tree,
2760 Allreduce: Reduce then Bcast
2761 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2762 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2763 Alltoallv: flat tree, like ompi
2765 * Add support for optimized collectives (Bcast is now binomial by default)
2766 * Port smpirun and smpicc to OS X
2769 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2770 Shout out if you used it.
2773 * Display the list of still queued messages in SG mode when existing
2777 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2778 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2779 * Add xbt_str_from_file(FILE*)
2780 * Add xbt_dict_get_key achieving a linear reverse search
2781 * Remove the context module
2783 Portability report of this version:
2784 * Main portability targets:
2785 - Linux(debian)/x86/context
2786 - Linux(debian)/x86/pthreads
2787 - Linux(debian)/amd64/context
2788 - Linux(debian)/amd64/pthreads
2789 On these, we still have the eratic breakages of gras/pmm and
2790 amok/saturate_sg reported in previous version. We still think
2791 that the tests are the cause of the fault, not the tested code.
2793 - Mac OSX Leopard/x86/context
2794 Still false negative in tesh autotesting.
2795 Smpi still fails, but this time because readlink does not accept -f
2796 Everything seems to work properly beside of that.
2799 - AIX version 5.3 (only tested contexts this time)
2800 Smpi still fails there because mktemp is not installed.
2801 Everything seems to work properly beside of that.
2803 I managed to compile it for the first time, but several breakages.
2804 Won't delay the release for this exotic platform.
2806 * Windows: it's still lagging behind. If you want to help, please
2809 Timing report of this version:
2810 This version seem to be more than 5% faster than 3.3.1 (on linux
2811 64bits with contextes). The gain is less than expected, we are
2812 investigating this for next release.
2814 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2816 SimGrid (3.3.1) stable; urgency=low
2819 * Implement a --cfg-help to show existing configuration variables
2820 * Build chain do not require doxygen in maintainer mode
2823 * fix a bug on struct sizeof computation, which prevented the
2824 exchange of arrays of structs in some conditions
2825 - added a regression test about this in datadesc_usage
2826 * Allow the exchange of 0-long dynamic vectors.
2827 - for that, use -1 as indicator of dynamic size instead of 0
2828 - This implied to change any size from unsigned long to long,
2829 reducing a bit communication abilities, but I guess that with
2830 64bits being quite common, this is more than enough.
2831 - This also induce a protocol change, thus bumping network protocol
2832 version from 0 to 1 (if we have external users, we have to get
2833 clean on that point too ;)
2834 - added two regression tests about this in datadesc_usage
2835 * Be more verbose when propagating local exceptions
2836 This helps debugging.
2837 * Display the status of simulated processes when receiving SIGINT in
2841 * Allow to control the simulation from a trace file.
2842 New functions MSG_action_register() and MSG_action_trace_run()
2843 The first one allows to associate a function execution to each
2844 kind of action while the second one parses a trace file and
2845 triggers the corresponding actions within the system.
2846 For now, only a toy example is provided in examples/msg/actions
2847 * Add an example of process migration in examples/msg/migration
2848 * Fix a bug in task exchange which broke MSG_task_get_sender()
2849 Add a teshsuite regression test for that.
2850 [Bug: if MSG_task_get_sender() is called after sender exit,
2852 * Fix a bug which prevented suspend/resume to work properly
2853 * Display the status of simulated processes when receiving SIGINT
2854 This fixes a regression of v3.3. due to the introduction of SIMIX
2855 * Bug fixing in failure management:
2856 - trace could not start by a failure at time 0
2857 - failure during communications were not working
2860 * Add SIMIX_process_set_name() to change the name of the current
2861 process in the log messages.
2862 * Store smx_hosts in a dict since we only retrieve them by name
2863 * Move the configuration infrastructure to surf
2866 * Move the configuration infrastructure to surf
2869 * Massive internal cleanups:
2870 - Store internal structures on processes instead of hosts (allows
2871 to have more than one process per host, in addition of being more
2873 - Cleanup the initialization/finalization process
2874 - Kill a whole bunch of unneeded synchronization:
2875 processes run in exclusive manner within the simulator
2876 - Move queues from global tables to process data fields
2878 - now accept -platform and -hostfile arguments
2879 - Pass the right rank value to processes according to the hostfile
2880 * Compile the examples by default, and use them as regression tests
2881 * Implement MPI_Wtime()
2882 * Change the reference speed to a command line option
2885 * TCP_gamma can now be specified as command line option using
2886 --cfg=TCP_gamma:10000000.0
2887 * Change the --surf-path cmd line option into --cfg=path:
2890 * Also include strbuff from xbt.h public header
2891 * xbt_ex_display(): do not free the exception after displaying
2892 This allows to do more with the given exception afterward.
2893 Users should call xbt_ex_free() themselves.
2897 Portability report of this version:
2898 * Main portability targets:
2899 - Linux(debian)/x86/context
2900 - Linux(debian)/x86/pthreads
2901 - Linux(debian)/amd64/context
2902 - Linux(debian)/amd64/pthreads
2903 These targets fail about 1/10 of times on gras/pmm, but we believe
2904 that this is because of the test, not because of simgrid.
2905 amok/saturate_sg fails even more rarely, and the test may not be
2908 - Mac OSX Leopard/x86/context
2909 The test suite still spits tons of errors because some obscure
2910 force prevents us from removing the temporary directories
2911 arguing that they still contain some metadata I've never heard of.
2912 Smpi fails because seq is not installed.
2913 Everything seems to work properly beside of that.
2916 - AIX version 5.3 (both contexts and pthread)
2917 Smpi still fails there because mktemp is not installed.
2918 XML inclusions seems rosty on AIX.
2920 * Windows: it's still lagging behind. If you want to help, please
2923 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2925 SimGrid (3.3) stable; urgency=high
2929 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2930 [Malek Cherier & Mt]
2932 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2935 Until now, GRAS were using MSG as an interface to SURF. It was
2936 quite difficult because both interface have several differences
2937 (MSG channels vs GRAS sockets were the most notable point).
2939 This also opens the gate to SMPI (which should occur soon) and speed
2940 up simulations by to 40% (even if it were not the main goal).
2942 **************************************
2943 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2944 **************************************
2945 The point is that events occurring at the exact same timestamp are
2946 not scheduled in the same order with the old and new version. This
2947 may be enough to completely change the execution of simulations in
2948 some cases. Sorry for the inconvenience.
2950 * Cleanup and upgrade the XML format to push further scalability
2951 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2953 * Improve the testing infrastructure with tesh. Now a very large part of
2954 the code is tested not only by being run but also by checking that the
2955 output match an expected output [Mt].
2957 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2958 is really less memory-demanding, which should allow you to use
2959 larger files in SimGrid [AL].
2961 * Inform valgrind about our contextes, so that it becomes usable
2962 with the default (and more effecient) version of SimGrid
2963 [contributed by Sékou Diakite, many thanks]
2966 * Introduce a listener thread in charge of receiving incoming
2967 messages from the network. It allows to overlap communication and
2968 computation but most notably, it removes some stupid deadlocks due
2969 to the fact that so far, a process could not send and receive at
2970 the same time. This made most non trivial communication schema
2972 * Convert the PIDs from long int to int to match the MSG ones (and
2973 linux ones too) [Mt]
2974 * New function: gras_agent_spawn() to launch a new process on
2975 current host. Only working in simulation for now. [Mt]
2976 * New function: gras_os_hostport() returning a constant form (ie,
2977 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2980 * Make the backtrace of exceptions more human readable [Mt]
2981 * New module: xbt/str [Mt]
2982 a ton of string utility functions (split, join, printf to a newly
2983 allocated buffer, trim, etc)
2984 * New module: xbt/hash [Mt]
2985 SHA1 hashing algorithm (more to come if needed)
2986 * New module: xbt/synchro [Mt]
2987 synchronization tools (mutex and conditions) working the same way
2988 in simulation and in real life (mainly useful for GRAS, but not
2990 * New module: xbt/queue [Mt]
2991 classical producer/consumer synchronization scheme
2992 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2993 (using the classical functions will get serialized) [Mt]
2994 * Make dictionary internal table dynamic. No need to specify its size
2995 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2997 * Make sure the log channels are organized as a tree under windows
2998 (because of ANSI C compatibility issue, any channel were child of
2999 root directly) [Mt].
3002 * Cleaned many thing in surf and fixed a few bugs [AL].
3003 * Add a nice command line configuration mechanism to compose models [AL].
3004 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3005 the previous one (KCCFLN05). It relies on something that looks like
3006 a max-min sharing mechanism but cannot be written as such. A new solver
3007 was thus designed [AL].
3008 * Add a new solver to lmm. Based on Lagrange optimization and
3009 gradient-based descent, it enables to efficiently maximise systems s.a
3011 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3013 This solver enables to propose two new network models for TCP Reno and
3014 TCP Vegas based on Low's work. These models still need to be fully
3015 tested though [Pedro Velho].
3018 * Bug fix in SD_simulate. Now the time bound given as argument is
3020 * Use the new parallel task model (ptask_L07) as default.
3021 * Use the SURF command line configuration mechanism.
3022 * 0-size tasks (for synchronization) should now work.
3024 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3026 SimGrid (3.2) stable; urgency=high
3030 We still experience issues on this platform, but we believe that at
3031 least MSG is usable.
3033 GRAS API BREAKAGE (for simplification purpose, sorry):
3034 * the gras_msgtype_by_name is not used anymore. Instead of
3035 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3036 you can write (and must)
3037 gras_msg_send(toserver, "request", &request);
3038 - If you still want to pass a gras_msgtype_t to the function (to cache
3039 the type and avoid the lookup time), use the gras_msg_send_() variant.
3040 - Impacted functions:
3041 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3042 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3043 * The callbacks are now expected to return 0 when everything went well
3044 (just like the main() function)
3046 GRAS new features and improvements:
3047 * New module mechanism where user code can use per process globals [Mt]
3048 This is similar to gras_userdata_*() functions, but for libraries. It
3049 factorize some code developped over and over in the examples and AMOK.
3050 It has still to be documented and used (only amok/peermanagement is
3052 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3053 loss when they were small enough to fit into the buffer and sent quickly
3054 enough so that they can all get received in one shoot.
3055 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3056 if not found. Use the *_or_null() variant for the old semantic.
3057 * In gras_msg_handle, do not discard messages without callback.
3058 They are probably messages to be explicitly awaited later (ie, proofs of
3059 mis-synchronization in userland since they are sent before being awaited)
3060 No big deal usually.
3061 * gras_socket_meas_send/recv: semantic changed!
3062 The numerical arguments used to be (1) the total amount of data to send
3063 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3064 messages. This was need for the fool willing to send more than MAXINT
3065 bytes on quite fat pipes.
3068 * Do really rename the hostmanagement module to peermanagement. [Mt]
3069 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3070 but this is rather new and this was documented in the module
3071 documentation (poor excuses, I admit)
3072 * Bandwidth measurement semantic changed! This follows the changes to
3073 gras_socket_meas_send/recv explained above.
3076 * A sequential mode has been added to the workstations. When a workstation
3077 is in sequential mode, it can execute only one task, and the other tasks
3078 are waiting in a FIFO. [Christophe Thiery]
3081 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3082 model for SIMDAG. [Christophe Thiery]
3083 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3084 equal to 0, which caused some bad behaviors in
3085 saturated_constraint_set_update. I now use a threshold mechanism like in
3089 * When running manually src/testall, you select specific units [Mt]
3090 testall is the result of our cunit mechanism, and should replace all
3091 the scripty thingy around since bash don't run easily on billware.
3093 * A mallocator system has been added. [Christophe Thiery]
3094 Mallocators allow you to recycle your unused objects instead of freeing them
3095 and allocating new ones.
3097 Documentation update:
3098 * FAQ reworking + New FAQs:
3099 - "Valgrind spits tons of errors!" [Mt]
3100 - "How to repport bugs" [Mt]
3101 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3102 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3103 - Communication time measurement within MSG [AL]
3104 - I experience weird communication times when I change the latency [AL]
3105 * GRAS tutorial [Mt]
3107 - an introduction to the framework and to the used communication model
3108 - an initiatic tour introducing the most proheminent features:
3110 . Lesson 0: Installing GRAS
3111 . Lesson 1: Setting up your own project
3112 o Part 2: Message passing
3113 . Lesson 2: Exchanging simple messages
3114 . Lesson 3: Passing arguments to the processes (in SG)
3115 . Lesson 4: Attaching callbacks to messages
3116 . Lesson 5: Using globals in processes
3117 . Lesson 6: Logging information properly
3118 . Lesson 7: Using internal timers
3119 . Lesson 8: Handling errors through exceptions
3120 . Lesson 9: Exchanging simple data
3121 . Lesson 10: Remote Procedure Calling (RPC)
3122 . Lesson 11: Explicitly waiting for messages
3123 . Recapping of message passing features in GRAS
3124 - A HOWTO section containing:
3125 o HOWTO design a GRAS application
3126 More are due, of course. They will come latter. In the meanwhile, you can
3127 check the examples which are still here.
3129 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3131 SimGrid (3.1) stable; urgency=high
3135 There was a stack corruption somewhere, visible only when optimizing
3136 with these versions. [Vince]
3139 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3140 module that enabled to play with Directed Acyclic Graphs. It is built
3141 directly on top of SURF and provides an API rather close to the old
3142 SG. Some old codes using SG are currently under rewrite to check that
3143 all needful functions are provided. [Christophe Thiery]
3146 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3147 extension of the classical CLM03 model that gracefully handles
3148 failures. This is now the default model for MSG and GRAS. It doesn't
3149 handle parallel tasks yet though. [AL]
3150 * Bug fix: Weights were not correctly set in the network part.
3151 WARNING: This may have resulted in incorrect results with simulations
3152 where there are more than one flow on a given link. [AL]
3155 * After a (long ?) discussion on simgrid-devel, we have decided that the
3156 convention we had on units was stupid. That is why it has been decided
3157 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3158 WARNING: This means that all previous platform files will not work as
3159 such with this version! A warning is issued to ask users to update
3161 A conversion script can be found in the contrib module of the CVS, under
3162 the name contrib/platform_generation/surfxml_update.pl [MQ]
3165 * Bug fix: Processes were started in reverse order, wrt deployment file.
3166 WARNING: if your code relies on this bug, please fix it. [AL]
3167 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3168 being executed on two different locations. [AL]
3169 * Bug fix: Failures are now better supported thanks to Derrick's tests
3170 (there was many failure situations I hadn't thought of and that weren't
3171 correctly handled). [AL]
3172 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3176 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3177 Exception occurring on server-side are propagated back to client (!).
3179 API CHANGE: the callback changed their prototype. Change:
3180 int my_handler(gras_socket_t expeditor, void *payload_data) {
3182 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3183 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3185 * New! function: gras_msg_handleall to deal with all messages arriving
3186 within a given period.
3187 * New! function: gras_socket_server_range to get a server socket in a
3188 range of port numbers (ease to avoid port number conflicts) [MQ]
3189 * New! gras processes display their backtrace when they get a SIGUSR1
3190 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3191 Sweet to debug RL processes [MQ]
3195 - Do not force experiment sizes to be expressed in kb, or it becomes
3196 impossible to measure the latency this way (needs one byte-long tests)
3197 WARNING: this changes the amok_bw_* function semantic. [MQ]
3198 - Implements the link saturation stuff. [MQ]
3199 * Peer management module:
3200 New! module factorizing code that we wrote over and over [MQ].
3203 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3204 - Test units are placed directly into the library code, they get extracted
3205 automatically and placed into the src/testall binary.
3206 - Convert most of the XBT tests to this system.
3207 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3208 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3209 live in the same C code. This required to change a little bit the API
3210 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3211 * New module: sparse graph structure with basic algorithms (this is work
3212 in progress and the API is not considered to be frozen yet). [AL]
3213 * Display more information on backtraces: source line & function names are
3214 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3215 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3216 * Reworked a little bit some #include statements to load only required
3217 headers. Some user code that relied on SimGrid to include stdlib or
3218 stdio may need to include it by themselves. [AL]
3219 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3221 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3222 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3223 but I guess that almost nobody used those parts. [MQ]
3225 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3227 SimGrid (3.0.1) stable; urgency=low
3230 * Unfortunately, I had missed 5 misnamed functions:
3231 xbt_fifo_item_t xbt_fifo_newitem(void);
3232 void xbt_fifo_freeitem(xbt_fifo_item_t);
3233 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3234 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3235 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3236 They're now deprecated. Please use their new versions:
3237 xbt_fifo_item_t xbt_fifo_new_item(void);
3238 void xbt_fifo_free_item(xbt_fifo_item_t);
3239 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3240 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3241 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3243 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3244 * Documentation: xbt_log module unmercifully reworked [MQ]
3245 * Bugfix: there was a problem with the ending of contexts with
3246 the pthread backend. It caused some weird deadlock or behavior
3247 depending on the pthread implementation. [AL]
3248 * Bugfix: get the exceptions raised in the simulator repport where
3249 and why they come from when they are not caught in time [AL, MQ]
3252 * Bugfix: Do repport the error when two non-connected hosts try to
3253 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3256 * Add additionnal checkings on communications. Assert that two
3257 communicating hosts are connected by a set of links... [AL]
3260 * Add additionnal checkings on channel values in communication [AL]
3261 * New: MSG_task_get_source to see on which host a task was generated [HC]
3262 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3263 the number of tasks waiting to be received on channel and sent
3265 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3266 waits for the first task coming from a given host.. [AL]
3268 GRAS new functionnalities: [MQ]
3269 * Enhance the parsing macro to allow the size of multidimentional objects
3270 to be given thru annotations.
3271 * New example (and documentation): Matrix Multiplication a la RPC
3272 (as when I was young!) and fix a bunch of bugs found on the way.
3274 GRAS performance improvements: [MQ]
3276 * Reduce the amount of cbps creation/destruction by making it static to
3277 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3280 * Change libdata to a set so that we can search for stuff by ID (and thus
3281 reduce the insane amount of dict lookups)
3284 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3285 RL communication on top of SG since it's so uneffective.
3286 It may also be used for inter-thread communication in RL, one day.
3287 * Use gras_datadesc_copy() to exchange messages on top of SG
3289 - improve message exchange performance on top of SG
3290 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3291 * Don't exchange on the network the size of the used part of buffer,
3292 instead, specify the possible buffer size to read().
3294 - reduces the amount of read/write calls (one pair per exchange)
3295 - reduces the amount of exchanged data (the size)
3296 - allows to retrieve all arrived data on receiver side, if we don't need
3297 it right now (subsequent read will peek the buffer)
3298 - allows the receiver to proceed with the beginning of the stream before
3299 everything is arrived
3300 - make it possible to build an iov transport (using readv/writev)
3302 - take care of the data with non-stable storage (like stacked data),
3304 * If possible, TCP send uses vector I/O (when writev() is here)
3305 - Don't use it for receive since we send data sizes and data on the
3306 same stream, so we wouldn't be able to chain large amount of chunks
3307 before having to flush the stuff to read the size.
3308 * Rework the transport plugin mechanism to simplify it and reduce the
3309 amount of pointer dereferencement when searching for the right function
3312 * I guess that now, we do almost as few system calls as possible while
3313 doing as few data copy as possible.
3315 To improve it further, we could try to send all the sizes first and then
3316 all the data (to use iov on receiving size), but it's only a partial
3317 solution: when you have 2 dimensional data, the sizes of the second
3318 dimension is data of the first dimension, so you need 3 streams.
3320 I'm not sure the potential performance gains justify the coding burden.
3322 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3324 SimGrid (3.00) stable; urgency=high
3327 * New! Give the possibility to hijack the surf parser and thus bypass
3328 MSG_create_environment and MSG_launch_application. Have a look at
3329 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3331 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3333 SimGrid (2.96) unstable; urgency=low
3338 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3339 This deprecates the xbt_error_t mechanisms.
3340 It modifies (simplifies) all XBT and GRAS API.
3341 MSG API keeps unchanged (exceptions raised by XBT are caught from
3342 within MSG and masked with existing error handling facilities)
3345 * New! Add a FATPIPE model. [AL]
3346 * New! Add a parallel task model. [AL]
3347 * New! Add automatically a loopback interface (in the default
3348 network model) if none was precised.
3351 * Bugfix: MSG_process_resume now works with the current running process.
3353 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3354 * Modification of MSG_task_get_compute_duration. Once a task has been
3355 processed, the value returned by this function is now equal to 0. [AL]
3356 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3357 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3358 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3359 of a m_task. For now, MSG_task_cancel only works with computation
3361 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3362 of the processor (in Mflop/s) regardless of the current load on the
3364 * API Change: use proper naming convention for MSG_getClock and
3365 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3367 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3368 This function changes the priority of a computation task. This priority
3369 doesn't affect the transfer rate. A priority of 2 will make a task
3370 receive two times more cpu power than the other ones. This function
3371 has been added to suit the needs of Nguyen The Loc and hasn't been that
3372 much tested yet. So if it fails, please report it and send me your code.
3374 * API Change: removed all functions and types that were marked "deprecated"
3375 since many months. Renamed MSG_global_init_args to MSG_global_init.
3377 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3379 SimGrid (2.95) unstable; urgency=low
3382 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3383 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3384 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3387 * Bugfix: Allow absolute paths for platform description files [MQ]
3388 * Bugfix: do free the variables after use. Leads to drastic performance
3390 * Implement max_duration (ie, timeouts) on resources [AL]
3393 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3395 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3396 the needed support (timeouts) [AL]
3399 * Implement measurement sockets. You can now get the bandwidth between two
3400 hosts thanks to AMOK (see below). [MQ]
3401 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3402 dynar over the network (yeah) [MQ]
3403 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3404 * simple/static token-ring example. [Alexandre Colucci and MQ]
3405 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3406 in place before [MQ]
3409 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3410 type parsing. "short" and "long" modifiers were also ignored; other
3411 modifier (such as reference level) are still ignored. [MQ]
3412 * Embeed the buffer size within the buffer itself on SG. [MQ]
3413 That way, send() are atomic and cannot get intermixed anymore (at least
3414 the ones which are less than 100k; bigger messages still have the issue)
3415 * Array size pushed by the field, not by the field type (or each
3416 and every long int will push stuff to the cbps) [MQ]
3417 * use select() to sleep since it allows to portably sleep less than one
3420 GRAS (minor cleanups)
3421 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3424 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3425 * type category 'ignored' killed as it was never used and were difficult
3427 * whether a type can cycle or not is now a flag, leaving room for more
3428 flags (as "ignored", if we feel the need one day ;)
3429 * Rename raw sockets to measurement sockets since "raw" has another
3430 meaning in networking community.
3433 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3434 and offers features not belonging to GRAS but that most applications
3435 need. One day, it may be a set of plugins loadable at runtime.
3436 * New module: bandwidth
3437 bandwidth measurement between arbitrary nodes running this module. [MQ]
3439 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3441 SimGrid (2.94) unstable; urgency=low
3443 The first beta release of SimGrid 3 !
3447 * Update the main page and the FAQ. Adding references to gforge.
3450 * Add a gras_os_getpid function.
3453 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3454 * Extend the logs so that they also print PID, hostname, date, ... if
3456 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3457 and kill the old version which were in testsuite/
3458 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3462 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3463 for users. Instead, each programming environment (GRAS, MSG,...) use it
3464 when needed to provide such a feature to users.
3465 Don't shortcut the mechanism or you will also shortcut the virtualization
3466 you need on the simulator.
3470 * Cleanups in configury with regard to compile optimization/warning flags.
3471 Also add -fno-loop-optimize to any powerpc since it's the optimization
3472 killing gcc (< 3.4.0).
3473 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3474 needed by MSG examples complications
3475 * Borrow configury beautifications from PHP
3478 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3479 warning (thanks loris for stumbling into this one).
3480 * Bugfix: stop loading private headers (gras_config.h) from the public
3484 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3485 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3486 RL would imply a DNS resolver, which is *hard* to do in a portable way
3487 (and therefore delayed).
3488 * Implement a real timer mechanism and use it in timing macros. This allows
3489 to avoid rounding errors and get a 0.000005 sec precision in timing
3490 macros. While I was at it, various cleanups:
3491 - allow to declare more than one timed section per file (fix a stupid bug)
3492 - move some private declaration to the right place
3493 - merge conditional execution and timing macros into emulation module
3494 - document the module
3495 - make sure the module cleanups its mess on gras_exit
3496 * Documentation improvements:
3497 - (new) how to compile applications using GRAS
3498 - (new) emulation support (timing macros)
3500 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3502 SimGrid (2.93) unstable; urgency=low
3504 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3507 - Use Paje properly where used. Still to be sanitized properly.
3508 - Portability fix: Add an implementation of the contexts using pthread
3512 - Add xbt_procname(): returns the name of the current process.
3513 Use it to show the current process's name in all logging.
3515 - fix detection of older flex version and the reaction, since we do
3516 depend on modern ones (we use lex_destroy)
3517 - Better separation of SG and RL in the libs: remove all simulation code
3518 from libgras. As a result, this lib is now only 200k when stripped.
3519 Some of the xbt modules may also be duplicated (two sets and such) and
3520 should be cleaned/killed before SG3.
3521 - Insist on using xlC on AIX because of weird problems involving gcc there.
3522 - Cleanup the make remote stuff. This is now done by scripts
3523 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3524 mainly for our private use, but we're working on changing them to user
3527 - Bugfix: flush the socket on close only if there is some *output*.
3528 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3530 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3533 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3535 SimGrid (2.92) unstable; urgency=low
3537 Alpha 3 on the path to SimGrid 3
3541 - New! First try of benchmarking macros.
3542 - New! First try so that gras_stub_generator generate deployment and
3543 remote compilation helpers.
3545 - Bugfix: Initialization fix in msg_test.
3549 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3551 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3553 - API Change: gras_os_sleep to take the amount of seconds as a double.
3554 Accepting an int was error prone since it was the only location where
3555 seconds were coded as such. It leaded to damn rounding errors.
3556 - Bugfix: Hard to belive that timers ever worked before this.
3558 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3560 SimGrid (2.91) unstable; urgency=low
3562 Alpha 2 on the path to SimGrid 3
3566 - Bug fix in the lmm_solver.
3568 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3569 through the function MSG_paje_output.
3570 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3571 - It is possible to bound the rate of a communication in MSG with
3572 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3573 to write it in the changelog).
3574 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3579 - Various cleanups to the autotools stuff
3580 - Begin to move Gras examples to examples/gras/
3581 - Let make distcheck work again (yeah!)
3583 - documentation overhauled using doxygen.
3584 gtk-doc-tools is dead in SimGrid now.
3585 - Automatically extract all existing logging categories, and add the list
3586 to the documentation (long standing one, to say the less)
3588 - Cleanup the known architecture table. Reorder the entries to group what
3589 should be, and use a more consistent naming scheme.
3590 (some of the test dataset are still to be regenerated)
3591 - New! Allow library to register globals on each process just as userdata
3593 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3594 lookup time (for now).
3595 Use it in msg and trp.
3596 This cleans a lot the internals and helps enforcing privacy of the
3597 headers between the gras components.
3598 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3599 - Bugfix: gras_os_time was delirious in RL.
3600 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3602 - Reenable GRAS now that it works.
3604 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3606 SimGrid (2.90) unstable; urgency=low
3608 Alpha 1 on the path to SimGrid 3
3610 * It is a long time since the last release of SimGrid. I'm sorry about
3611 that but as I had told you, I was rewriting a lot of things. I apologize
3612 to those who had been reporting bugs to me and that I had not answered.
3613 If your bug is still in the new version, please tell me. Here is a
3614 summary of the main changes.
3616 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3617 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3618 portability, speed, and a lot more but you'll figure it out later.
3619 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3620 I wanted to release the new MSG as soon as possible and I have
3621 broken GRAS, which is the reason why, for now, only MSG is fully
3622 functional. A laconic description of these projects is available
3623 in the documentation.
3625 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3626 kernel optimized for our needs (called SURF but only the developers
3627 should use it). Hence, MSG is now roughly 30 times faster and I think
3628 that by rewriting a little bit MSG, I could event speed it up a little
3629 bit more. Beside the gain in speed, it is also much easier to encode a
3630 new platform model with SURF than it was with SG. More to come...
3632 * REVOLUTION 3: I have tried to change a little as possible the API of
3633 MSG but a few things really had to disappear. The main differences
3634 with the previous version are:
3635 1) no more m_links_t and the corresponding functions. Platforms are
3636 directly read from a XML description and cannot be hard-coded
3637 anymore. The same format is used for application deployment
3638 description. The new format is described in the documentation.
3639 Have a look in tools/platform_generation. There is a tiny script
3640 that converts from the old platform format to the new one. Concerning
3641 the application deployment format, parsing the old one is tricky.
3642 I think most of you should however be able to convert your files. If
3643 it is really an issue, I can write a C code that does the conversion.
3645 2) the toolbox tbx does not exist anymore. We now have a library
3646 with much more data-structures but without the hash-tables (we have
3647 dictionaries that are much faster).
3649 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3651 *****************************************************************************
3652 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3653 * don't feel like converting the oldies. (Mt) *
3654 *****************************************************************************
3657 Version 2.90: "the long awaited one"
3658 - Finished rewriting and debugging MSG. Rewrote the documentation.
3659 - disable GRAS for now since it needs to be ported to the newest SG
3662 - Finish the port to windows (using mingw32 for cross-compile)
3665 - Main loop and datastructures of SURF. A cpu resource object is
3666 functional. Surf can thus be used to create cpu's with variable
3667 performance on which you can execute some actions.
3669 2004-11-15 Martin Quinson
3670 - Port to ARM. Simply added the alignment and size descriptions. Should
3671 work, but the ARM machines are so slow that I didn't had the opportunity
3672 to 'make check' over there yet.
3674 2004-11-15 Arnaud Legrand
3675 - Trace manager now written. It uses a heap structure and is therefore
3676 expected to be efficient. It may however be speeded up (particularly
3677 when many events occur at the same date) by using red and black
3678 trees. One day maybe...
3679 - Max-min linear system solver written. It uses a sparse matrix
3680 structure taking advantage of its expected use. Most operations are
3681 O(1) and free/calloc are called as few as possible. The computation of
3682 the minimum could however be improved by using a red and black tree
3685 2004-11-03 Arnaud Legrand
3686 - Rename every gras_* function that was in xbt/ to its xbt_
3688 - Add a heap and a doubly-linked list to xbt
3689 - Added a dichotomy to the dictionaries. make check works as well before
3690 so I assume that the patch is correct. I do not know however if things
3691 run effectively faster than before now. :)
3693 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3694 SimGrid, and the version number is bumped to 2.x
3696 2004-10-29 Martin Quinson
3697 - Introduction of the remote errors.
3698 They are the result of a RMI/RPC on the remote machine.
3699 ErrCodes being scalar values, you can't get the host on which those
3700 errors did happen. Extending the error mechanism as in Gnome is possible.
3701 No idea yet whether it is a good idea.
3703 2004-10-28 Martin Quinson
3704 - Interface revolution: the Starred Structure Eradication.
3705 I used to do typedef struct {} toto_t; and then handle *toto_t.
3706 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3707 - ???_t is a valid type (builded with typedef)
3708 - s_toto_t is a structure (access to fields with .)
3709 - s_toto is a structure needing 'struct' keyword to be used
3710 - e_toto_t is an enum
3711 - toto_t is an 'object' (struct*)
3713 typedef struct s_toto {} s_toto_t, *toto_t;
3714 typedef enum {} e_toto_t;
3715 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3716 s_toto_t) is private.
3718 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3719 so that it can set the variable to NULL. It was so for dicts and sets,
3720 it changed for dynars.
3722 - Fix a bunch of memleaks in dict_remove
3723 - Fix a bug in sg/server_socket opening: it failed all the time.
3725 2004-10-07 Martin Quinson
3726 - Speed up dynar lookup operation a bit.
3728 gras_dynar_get is dead.
3730 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3731 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3732 (which gives you the address of the stored data).
3734 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3735 copy of the data using an affectation to do the job and not a memcpy.
3737 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3738 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3740 It does not really speedup the dynar test because they are
3741 setting elements all the time (and look them seldom). But the dict does
3742 far more lookup than setting.
3744 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3746 2004-10-05 Martin Quinson
3747 - Allow to (en/dis)able the cycle detection at run time.
3749 Whether we should check for cycle or not is now a property of each
3750 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3751 datadesc_cycle_unset allow to remove this property when previously set.
3753 Note that the cycle detection is off by default since it impacts the
3754 performance. Watch the data you feed GRAS with ;)
3756 This property is hereditary. Any element embedded in a structure having it
3757 set have it set for the time of this data exchange.
3759 You should set it both on sender and receiver side. If you don't set it on
3760 sender side, it will enter an endless loop. If you forget on receiver
3761 side, the cycles won't be recreated after communication.
3763 - Header reorganization.
3764 Kill gras_private.h, each submodule must load the headers it needs.
3766 2004-10-04 Martin Quinson
3767 - Interface revolution: do not try to survive to malloc failure.
3769 Now, gras_malloc and friends call gras_abort() on failure.
3770 As a conclusion, malloc_error is not a valid error anymore, and all
3771 functions for which it was the only gras_error_t return value are
3772 changed. They now return void, or there result directly.
3773 This simplify the API a lot.
3775 2004-09-29 Martin Quinson
3776 - Re-enable raw sockets.
3777 Created by gras_socket_{client,server}_ext;
3778 Used with gras_raw_{send,recv}
3781 It should allow to kill the last bits of gras first version soon.
3783 This is not completely satisfactory yet (duplicate code with
3784 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3787 - Simplify transport plugin (internal) interface by not passing any
3788 argument to _server and _client, but embedding them in the socket
3791 2004-09-28 Martin Quinson
3792 - Finish the port to AIX.
3793 autoconf was my problem (segfault within the malloc replacement
3794 function. No idea why)
3796 2004-09-16 Martin Quinson
3797 - Fix some size_t madness on 64bit architectures.
3799 2004-09-08 Martin Quinson
3800 - Reduce the number of system headers loaded, overload some more system
3801 calls (such as malloc to cast the result of the system one, and work
3803 - Fix and reintroduce the config support
3805 2004-09-07 Martin Quinson
3806 - Source code reorganization to allow Arnaud to surf all over there.
3807 - Allow to document the logging categories.
3808 - Remove all uppercase from logging categories and useless cleanup in names.
3810 2004-08-18 Martin Quinson
3811 Version 0.6.2 (protocol not changed; API changed)
3812 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3813 gras_error_t), and NULL when not found. Functions expecting a msgtype
3814 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3815 hopefully usefull message.
3816 - Portability to prehistoric sparcs again
3818 2004-08-17 Martin Quinson
3819 Version 0.6.1 (protocol not changed; ABI not changed)
3820 - prealloc some buffers to speed things up
3822 2004-08-11 Martin Quinson
3823 Version 0.6 (protocol not changed; ABI expended)
3824 - The parsing macro can deal with the references, provided that you add
3825 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3827 2004-08-09 Martin Quinson
3828 Version 0.5 (protocol not changed; ABI changed)
3829 - Allow to off turn the cycle detection code in data exchange at
3830 compilation time. It should be at run time, but I'm short of time (and
3831 the config stuff is still broken). That way, we keep dict out of the
3832 critical path, which is good because the performance is poor:
3833 - search not dichotomial yet
3834 - dynar give no way to access their content and memcpy everytime
3835 - In composed data description (struct, ref and so on), stop foolness of
3836 keeping the subtype's ID, but store the type itself. This keeps sets out
3837 of the critical path, which is good since they rely on dynar and
3838 dictionaries. The only loose of that is that we cannot detect the
3839 redeclaration of a structure/union with another content (but I'm not sure
3840 the code detected well this error before anyway). We still can detect
3841 the redefinition discrepancy for the other types.
3842 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3843 it breaks the code because of type-punning used all over the place).
3844 This breaks on all non-gcc architectures (for now).
3846 All those changes (plus the buffer of last time) allow me to gain 2 order
3847 of magnitude on cruel tests consisting of 800000 array of integers on two
3848 level of a hierarchical structure (200 secondes -> 4 secondes)
3851 - the selector of reference must now return the type it points to, not
3852 the ID of this type.
3854 2004-08-06 Martin Quinson
3855 Version 0.4 (protocol changed; ABI not changed)
3856 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3857 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3858 - Implement a buffer transport, and use it by default (it relies on tcp in
3859 real life and on sg in simulation).
3860 That's a bit hackish since I had a new field to the structure to store
3861 its data without interfering with the subtype ones. Inheritance
3862 is tricky in C. And that's a kind of reverse inheritance with one class
3863 derivating two classes. Or maybe a game with java interfaces. Anyway,
3864 that's damn hard in C (at least).
3865 Moreover, I got tired while trying to ensure plugin separation and
3866 genericity in SG mode. MSG wants me to do weird things, so let's go for
3867 cruel hacks (temporarily of course ;).
3868 See comment in transport_private.h:71
3869 - do not include all the _interface headers in private but in the files
3870 which really need them (to cut the compilation time when they are
3873 2004-07-26 Martin Quinson
3874 Version 0.3 (protocol not changed; ABI changed)
3875 - Major overhault of the datadesc interface to simplify it:
3876 - shorted the function names:
3877 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3878 - add a trivial way to push/pop integers into the cbps without malloc.
3879 This allows to make really generic sub_type description, which simply
3880 pop their size of the stack.
3881 - add a function gras_datadesc_ref_pop_arr() which does what users want
3882 most of the time: Declare a dynamic array (which pops its size of the
3883 stack) and declare a reference to it. Poor name, but anyway.
3884 - kill the post-send callback, add a post-receive one
3886 2004-07-23 Martin Quinson
3887 Version 0.2 (protocol changed; ABI changed)
3888 - add some testing for cpbs in the test cases, and fix some more bugs.
3889 This invalidate again the little64 data file, since I cannot regenerate
3891 - remove an awfull optimization in the logging stuff, allowing me to:
3892 - understand it again
3893 - learn gcc how to check that the argument match the provided format
3894 - fix all errors revealed by gcc after that
3895 - internal keys of dict are not \0 terminated. Deal with it properly in
3896 loggings instead of segfaulting when the user want to see the logs :-/
3898 2004-07-22 Martin Quinson
3899 - Fix some stupid bug preventing cbps (callback postit) from working
3901 2004-07-21 Martin Quinson
3902 - Some documentation cleanups
3903 - remove the useless last argument of msgtype_declare
3904 - rename the Virtu functions to fit into the 'os' namespace
3905 - move headers src/include -> src/include/gras/ and stop fooling with
3906 gras -> . symbolic link
3907 - make distcheck is now successful
3909 2004-07-19 Martin Quinson
3911 - Build shared library also
3912 - Install html doc to the right location
3913 - stop removing maintainer files in make clean
3914 - build tests only on make check
3916 2004-07-13 Martin Quinson
3918 - No major issue in previous version => change versioning schema
3919 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3920 corresponding dataset.
3922 2004-07-11 Martin Quinson
3924 - Get it working with any kind of structure (we can compute the padding
3925 bytes remotely for all the architectures I have access to)
3926 - Implement the structure parsing macro (still not quite robust/complete)
3927 - Improvement to the remote testing toysuite
3929 2004-07-10 Martin Quinson
3930 [autoconf mechanism]
3931 - get ride of a bunch of deprecated macros
3932 - actually run the test for two-compliment, not only compile it :-/
3933 - test whether the structures get packed (and bail out if yes. Damn.
3934 Alignment is a serious matter)
3935 - test whether the structures get compacted (but respecting the alignment
3936 constraints of each types)
3937 - test whether the array fields of structures can straddle alignment boundaries
3939 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3940 (took me 2 hours to find that bug, looking at the wrong place)
3941 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3942 before sending/receiving and are used to compute the offsets of fields
3943 - Given that padding size depend even on compiler options, keep track of
3944 alignment and aligned_size only for the current architecture. Not a big
3945 deal since we send structure fields one after the other (seems
3947 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3948 program, let it work on linux/gcc/little32. portability todo.
3950 2004-07-08 Martin Quinson
3951 - import and improve remote compilation support from FAST
3952 - make sure make check works on half a dozen of machines out there
3954 2004-07-07 Martin Quinson
3955 Let's say it's version 0.0.3 ;)
3956 - Implement conversions (yuhu!)
3957 - Let it work on solaris (beside conversion, of course)
3958 - Stupid me, using rand() to generate the conversion datatests in not wise.
3960 2004-07-06 Martin Quinson
3961 - Let make dist work, since I'm gonna need it to compile on remote hosts
3962 - Let Tests/datadesc_usage write the architecture on which the file was
3963 generated as first byte.
3964 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3965 architecture descriptions.
3966 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3967 on those architectures.
3968 - Optimization: send/recv array of scalar in one shoot
3970 2004-07-05 Martin Quinson
3971 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3973 - Plug a whole bunch of memleaks
3974 - each process now have to call gras_{init,exit}. One day, their log
3975 settings will be separated
3976 - Continue the code factorisation between SG, RL and common in Transport.
3978 2004-07-04 Martin Quinson
3980 - Redistribution between SG and RL.
3981 We wanna have to accept in SG, so move accepted related parts of RL in
3982 the common part. (more precisely, the dynar of all known sockets is no
3983 more a static in transport.c, but part of the process_data)
3985 [gras_stub_generator]
3986 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3988 2004-07-03 Martin Quinson
3989 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3990 them all up in one shot)
3992 - Fix the ugly case of reference to dynamic array.
3993 - New (semi-public) function gras_datadesc_size to allow the messaging
3994 layer to malloc the needed space for the buffer.
3996 - gras_socket_close now expect the socket to close (and not its address to
3997 put NULL in it after it). This is because the socket passed to handlers
3998 is one of their argument (=> not writable).
4000 - propagate the interface cleanup from last week in datadesc, ie remove a
4001 superfluous level of indirection. User pass adress of variable
4002 containing data (both when sending and receiving), and not of a variable
4003 being a pointer to the data. Let's say that I like it better ;)
4004 The price for that is constructs like "int msg=*(int*)payload" in
4005 handlers, but it's a fine price, IMHO.
4007 - Let it work in RL (yuhu)
4009 2004-06-21 Martin Quinson
4011 - porting SG plugin and SG select to new standards (works almost).
4012 - plug memleaks and fix bugs around.
4015 - cleanup the prototype of data recv and force users to specify when they
4016 want to handle references to objects. Test case working even for cycles.
4017 - plug memleaks. Valgrind is perfectly ok with this.
4019 2004-06-12 Martin Quinson
4021 - cleanup the separation between plugin and main code in plugin creation
4023 2004-06-11 Martin Quinson
4025 - Reput hook for raw sockets, needed for BW experiments
4026 - kill a few lines of dead code
4027 [Data description] Interface cleanup
4028 - gras_datadesc_by_name returns the searched type or NULL.
4029 That way, no variable is needed to use a type desc once, which makes
4031 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4032 two parameters were strings (field name, type name), leading to
4034 [Dicos] Interface cleanup
4035 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4036 This is consistant with the dynar API.
4038 2004-04-21 Martin Quinson
4040 - Porting to new standards.
4042 - interface cleanup.
4043 There is no bag anymore, no need to take extra provision to mask the
4044 pointers behind "ID".
4045 Better splitup of functions between files create/exchange/convert.
4046 This is still a bit artificial since convert and receive are so
4047 interleaved, but anyway.
4049 - add a queued message list to procdata (the ones not matching criteria
4051 - factorize some more code between SG and RL wrt procdata
4053 - use gras_exit in example to track memleaks
4054 - get rid of gs_example now that GS is properly integrated into gras
4055 - update run_test to integrate the latest tests (datadesc)
4057 - rename WARNINGn macros to WARNn since it prooved error-prone
4059 2004-04-19 Martin Quinson
4061 - register init/exit functions within gras module mechanism
4062 - send/receive function.
4063 Convertion is not implemented, but short-cutted if not needed.
4064 struct/array elements are sent one by one (instead of block-wise), but
4065 nobody really cares (yet). Get a prototype before optimizing.
4066 - tests (using a file socket) for DD send/receive on:
4067 - base types: int, float
4068 - array: fixed size, string (ie ref to dynamic string)
4069 - structure: homogeneous, heterogeneous
4070 - chained list, graph with cycle
4071 Believe it or not, valgrind is not too unhappy with the results. The
4072 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4076 - Bugfix when using a filename explicitly (instead of '-')
4078 2004-04-09 Martin Quinson
4080 - factorize more code between RL and SG in socket creation
4081 - Complete the implementation and tests of:
4083 o file (only in RL, and mainly for debugging)
4085 I lost 3 days to design a portable address resolver, and then decided
4086 that the prototype mainly have to run on my box.
4087 Addressing portability too early may be like optimizing too early :-/
4089 - use gras_init in the Tests instead of the crappy parse_log_opt
4090 (the latter function is removed)
4091 [Conditional execution]
4092 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4093 [Code reorganisation]
4094 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4095 Build examples against the RL library, since there is no way to disable
4096 its creation for now.
4098 For information, the beginning of coding on GRAS was back in june
4099 2003. I guess that every line has been rewritten at least twice since