1 ----------------------------------------------------------------------------
3 SimGrid (3.24.1) NOT RELEASED YET (v3.25 expected December 22. 2029, 04:19 UTC)
6 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
7 - Drop support for Viva/Triva (old visualization tools), hence removed
8 TRACE_get_node_types() and TRACE_get_edge_types() functions.
11 - Actor: Merge signals on_migration_start/end into on_host_change
12 - Actor: Rename migrate() into set_host()
13 - Disk: Allow users to get the read and write nominal bandwidth values
14 - Exec: Implement wait_for(timeout)
15 - Io: Implement wait_for(timeout)
18 - Parse errors now raise a simgrid::ParseError that you may want to catch.
21 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
22 been renamed to fix a spelling error. As usual, the previous definitions are
23 kept with a deprecation warning. However, to avoid runtime errors, it is now
24 forbidden to override these deprecated methods in a derived class. Please use
25 the new names immediately if you need to override them.
28 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
29 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
30 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
31 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
32 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
33 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
34 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
35 - New module: random, providing classical random numbers generators.
38 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
39 allocations' size above which they are considered "shared" by default (as if
40 they were performed through SMPI_SHARED_MALLOC macros).
41 Default = 0 = disabled feature.
42 Note : malloc, calloc and free are now overriden by smpicc/cxx by default.
43 This can cause some troubles if codes are already overriding these. If this
44 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
45 help, but will make this feature unusable.
46 - Input check is now performed much more consistently and outputs precise
47 warnings for each error found
48 - New supported calls : MPI_File_set_view, MPI_File_get_view
49 - MPI I/O now only supports disks, as storage is discontinued.
50 - MPI I/O allows opening files without specifying absolute path
51 - Note : SMPI configuration options are now initialized by smpi_init_options()
52 call, which should be called if SMPI is not being used through smpirun.
54 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
55 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
56 - GH#305: Unscheduled tasks are still excuted
57 - GH#313: smpirun: manual outdated w.r.t. --help
58 - GH#321: [S4U] Get task remaining work ratio
59 - GH#323: Crash when an actor turn off his physical host
60 - GH#335: Missing links on dragonfly example svg
61 - FG!19: Removing RngStream
62 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
63 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
65 ----------------------------------------------------------------------------
67 SimGrid (3.24) October 9. 2019
69 The Clean Disk Release.
72 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
73 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
74 thus becomes the default resource to execute I/O operations. Examples
75 and the file system plugin have been modified to use disks instead of
76 storage elements. Storage can still be used but is doomed to disappear.
77 Users are thus adviced to update their simulators accordingly.
78 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
79 for consistency with pthread_barrier_wait()
80 - Host::get_englobing_zone() returns the englobing netzone
81 - Actor::on_destruction is now called in the destructor
82 Actor::on_termination new signal called when the actor terminates
84 - Global signals are now part of the Engine:
85 - on_platform_creation: after config settings, before the XML parsing
86 - on_platform_created: right after the XML parsing
87 - on_time_advance: each time the clock advances
88 - on_simulation_end: after simulation, before cleanups
89 - on_deadlock: as the name implies.
91 - sg_{actor,host,link}_{data,data_set}() now all exist.
92 Use them to attach user data to the object and retrieve it.
95 - Introduce an experimental Wifi model. It sounds reasonable
96 according to the state of the art, but it still has to be properly
97 validated, at least against ns-3 if not against reality.
98 - Improved the usability of ns-3. Several bugs were ironed out.
99 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
100 It is now expressed as 'idle:epsilon:allCores' to properly model the
101 consumption of non-whole tasks on mono-core hosts. More info in the doc.
104 - convert a new set of functions to the S4U C interface and move the old MSG
105 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
108 - Fortran bindings for DVFS have been removed.
109 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
110 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
111 - SMPI can now be selected by cmake's find_module(MPI) with
112 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
113 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
114 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
115 - trace-call-location can be used with TI traces, and replayed, and is compatible
116 with smpi/comp-adjustment-file.
117 - sleep events are now correctly traced/replayed.
118 - Default for trace-call-location is now to use file names and not full paths.
119 To revert to previous behavior (in case of collision of filenames), option
120 "smpi/trace-call-use-absolute-path" can be set to yes.
123 - Use the included xxHash as an hash implem when C++14 is usable.
124 - Option model-checker/hash was removed. This is always activated now.
125 - New option smpi/buffering controls the MPI buffering in MC mode.
126 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
127 This is useful to check for MPI compliance.
130 - New section on plugins: how to define a new one, and existing ones.
133 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
134 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
137 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
138 and <mount> tags. The rationale is finish to decouple what should be handle
139 by SimGrid kernel (i.e., resources and activities) from plugins built in
140 user space (e.g., the file system plugin). This new tag currently accepts
141 only three attributes (id, read_bw, and write_bw). All the other information
142 that was declared with the storage related tags now has to be expressed as
143 properties. An example of platform using this new tag is available at
144 examples/platforms/hosts_with_disks.xml
147 - 'expect signal' can now accept more than one potential signal.
149 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
150 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
151 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
152 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
153 - FG#31: per-actor data
154 - FG#34: SG_BARRIER_SERIAL_THREAD?
155 - FG#35: model-checker does not like buster-produced binaries
156 - FG!13: MC: complete workaround in the error msg seen on modern systems
157 - FG!15: execute_flops now logs compute
158 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
159 - FG!17: ns-3: unblock the right number of communications + others issues
160 - FG!18: Improving the performance of the ns-3 bindings
161 - GH#189: Energy consumption of parallel tasks
162 - GH#219: Error in the throughput of TCP transfer
163 - GH!330: Minor polishing of English: Configuring_Simgrid
164 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
165 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
166 - GH!334: Rework energy plugin, again
168 ----------------------------------------------------------------------------
170 SimGrid (3.23.2) July 8. 2019
173 - Nicer introduction page.
174 - Migrate the "Deploy your application" page to the new doc.
175 - Move Java as a subtree of MSG.
178 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
181 - Drop xbt_dynar_sort_strings().
184 - Really fix FG#26: Turning off a link should raise NetworkFailureException
185 - FG#27: Wrong exception thrown to wait_any when link is turned off
186 - GH#328: Java: Canceling multiple tasks in a single vm/host
188 ----------------------------------------------------------------------------
190 SimGrid (3.23) June 25. 2019
192 The Exotic Solstice Release.
195 - SunOS and Haiku OS support. Because exotic platforms are fun.
196 - Stop setting random seed with srand48() at initialization.
197 - Use addr2line as a fallback for stacktraces when backtrace is not available.
198 - Build option -Denable_documentation is now OFF by default.
199 - Network model 'NS3' was renamed into 'ns-3'.
202 - Simgrid can now hopefully be installed with pip.
205 - wait_any can now be used for asynchronous executions too.
208 - New log appenders: stdout and stderr. Use stdout for xbt_help.
209 - Drop xbt_dict_dump.
212 - 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
213 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
214 - MPI interface is now const correct for input parameters
215 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
216 - Fortran bindings for async collectives.
217 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
220 - Remove option 'model-check/record': Paths are recorded in any cases now.
221 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
222 always sparse. That's an immense gain in memory, and can even be
223 faster because of cache effects. So there is no need to clutter the
224 code to allow the user to go for the unefficient mode.
227 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
228 models proved to be more accurate than these old experiments.
230 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
231 - FG#1: Broken link in error messages
232 - FG#2: missing installation documentation
233 - FG#3: missing documentation in smpirun
234 - FG#6: Python bindings not available on PyPI
235 - FG#7: simple cmake call requires doxygen
236 - FG#8: make python bindings an optional dependency
237 - FG#10: Can not use MSG_process_set_data from SMPI any more
238 - FG#11: Auto-restart actors forget their on_exit behavior
239 - FG#12: -Denable_lto=OFF doesn't disable LTO
240 - FG#13: Installs unstripped file 'bin/graphicator'
241 - FG#14: Installs the empty directory 'doc/simgrid/html'
242 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
243 - FG#17: Dead link in doc (pls_ns3)
244 - FG#20: 'tesh --help' should return 0
245 - FG#21: Documentation link on http://simgrid.org/ broken
246 - FG#22: Debian installation instruction are broken
247 - FG#26: Turning off a link should raise NetworkFailureException exceptions
248 - GH#133: Java: a process can run on a VM even if its host is off
249 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
250 - GH#326: Valgrind-detected error for join() when energy plugin is activated
252 ----------------------------------------------------------------------------
254 SimGrid (3.22) April 1. 2019
256 The Easter Christmas Release (this one is somewhat late).
259 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
260 - Not complete yet: asynchronous activities (amongst others) are still missing
261 - Still ongoing: the interface may change in the future. We need more
262 testers! Please report any glitches.
263 - No new project using Java should start now. Please switch to Python.
266 - Some of the internal cleanups may lead to speed improvements:
267 - The hard limitation on the amount of simulated actors with
268 Java+Mac was removed. Now, the available memory is the only limit.
269 - Our refcounting was tidyied, leading to 10% speedups in some cases.
270 - We are still working on making our code robust to the actor kills
271 and hosts' churn. Things are improving, but it's not perfect yet.
272 - Replaced our own code to display a backtrace (that was forking addr2line)
273 with the Boost.Stacktrace library.
274 You won't see your backtraces without this optional dependency.
275 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
276 - Stop setting random seed with srand() at initialization.
279 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
280 XML file version remains 4.2 since old files are still compatible.
283 - Process termination which was broken at version 3.21 has been repaired.
284 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
285 - Hide the examples into examples/deprecated. New users should use Python.
288 - Drop MSG_process_create_from_stdfunc() from the API.
289 This C++ function was a pimple in the C API, made necessary at some
290 point by the Java bindings. This is fixed now.
291 - Hide the examples into examples/deprecated. New users should use S4U.
292 - MSG_process_create and MSG_process_attach now crash if the host used
293 to run this process is off.
294 - Fix the protype of MSG_process_on_exit()
295 Now use (int,void*) callbacks instead of (void*,void*) ones.
296 The implementation was ways too messy to actually work, I guess.
299 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
300 too often and provide more accurate timings.
301 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
302 - MPI_Alltoallw support
303 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
304 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
305 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
306 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
308 - Basic implementation of generalized requests (SMPI doesn't
309 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
312 - Drop sg_cmdline. Please use xbt_cmdline instead.
313 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
314 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
315 OsSemaphore is implemented in a portable way with C++11 threads.
316 This should allow much more threads to be created at the same time,
317 allowing Mac and Java users to simulate many more actors.
318 - Implement the 'thread' factory with std::thread instead of xbt ones.
319 It is not possible to set the stack size with threads anymore, but
320 -fsplit-stack is the way to go nowadays when using threads.
321 - Drop the xbt_os_thread_t module (now unused)
322 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
323 - Drop xbt_str_join_array().
324 - Drop cunit, use Catch2 instead.
327 - Many cleanups in the kernel::activity namespace. This was long
328 overdue, and shall open the path to many future endeavors.
331 - #132: Java: a process can not shut down its own host
332 - #220: S4U: detached send causes exception if sender terminates soon after sending
333 - #261: Document the parameters of parallel execution's constructor
334 - #300: [s4u] BarrierPtr is missing
335 - #314: SMPI args internal cleanup
336 - #316: Fix a bug related to the CPU utilization of multi-core VM
337 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
338 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
339 - #325: Turning off a host has different behavior on sleeping actors and computing actors
341 ----------------------------------------------------------------------------
343 SimGrid (3.21) October 3. 2018
345 The Restarting Documentation (TRD) Release.
348 - Convert the user manual to Sphinx for improved usability.
349 Unstable version now hosted on https://simgrid.frama.io/simgrid
350 Many glitches remain; stay tuned.
351 - Restrict the Doxygen reference API to the installed header files.
352 (documenting all internals this way is overkill + doxygen scales badly)
353 - New tutorial on S4U (highly inspired from the old MSG one, but with
354 a git repository to fork as a starting point, and a docker image)
355 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
356 - The Developper manual is still to be converted (not compiled until then)
357 - Some parts are still missing in this conversion (such as the
358 platform and deployment sections) while others were blindly converted
359 and would need more love (such as the configuration flags).
360 Things will certainly further improve in the future releases.
363 - Fully reimplement auto_restart mechanism that was utterly broken
364 (fix #22, #131 and #281 that were all very related)
365 - Implement semaphores (fix #303 and #291)
366 - When creating an actor from a function and its parameters,
367 move-only parameters are not allowed anymore, as it would prevent
368 the actor to be restartable if its parameters are consumed on run.
369 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
370 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
371 IO operations that can be started, waited for, or canceled as a regular
373 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
377 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
378 - New host variable: 'core_count'
381 - Remove xbt_os_thread_specific features
382 - Remove portability wrapper to condition variables
383 - Remove xbt_os_thread_yield()
386 - MPICH collective selector now mimics MPICH 3.3b
387 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
388 - MPI_Init(NULL,NULL) is now allowed
389 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
390 - Command-line arguments should now be handled properly in Fortran simulations
391 - MPI Attributes and related callbacks should work in Fortran
392 - Apps using C + Fortran should now work
393 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
394 - Support MPI_Error_String
397 - Due to an internal bug, Msg.run() must now be your last line.
398 We hope to fix it in a future release, and we are sorry for the inconvenience.
401 - #22: Process autorestart seem to only work with CAS01 cpus
402 - #56: Feature request: dynamic SMPI replay
403 - #93: simgrid should not eat --help
404 - #111: How to change the loopback link characteristics?
405 - #116: Platform cluster radical
406 - #117: Platform cluster/host order
407 - #129: Replay_multiple on steroid
408 - #131: Java : autostart does not work
409 - #143: Setting a breakpoint at a given time
410 - #235: Network energy models should be integrated
411 - #264: Add ptask L07 resource tracing
412 - #271: Dynamic and manual replay of SMPI traces
413 - #279: Breakpoints option uses time deltas instead of absolute time
414 - #281: Daemonized actor no longer daemonized after an auto-restart
415 - #282: TIT convertor erase traces when using absolute path in the trace list file
416 - #285: segfault when a s4u actor kills itself with msg process tracing activated
417 - #286: Pajé traces are not exposing the number of cores
418 - #287: Command-line parsing should stop when encountering '--'
419 - #288: MPI_Init(NULL, NULL)
420 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
421 - #290: Method to check if a mailbox has a message ready to be consumed
422 - #291: [s4u] semaphores do not seem available yet
423 - #294: Alltoall collective implementation of mpich
424 - #295: Starting a Host as OFF is not supported
425 - #296: DTD too permissive
426 - #299: Add a s4u barrier example
428 ----------------------------------------------------------------------------
430 SimGrid (3.20) Released June 24. 2018
432 The proxy snake_case() release.
435 - The whole API is now using snake_case() instead of mixing it with camelCase.
436 Compatibility wrappers in place for one year.
437 - We took the opportunity to increase the consistency of the API.
438 Please report any remaining glitches.
439 - New function simgrid::s4u::Host::get_actor_count:
440 Returns the number of actors running on a specific host.
443 - SMPI is now tested with ~45 proxy apps from various sources, with none or
444 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
445 - Replay: The replay file has been re-written in C++.
446 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
447 supported. They are stored in the trace and used when replayed.
448 - Basic support of MPI_Cancel. Robustness not guaranteed.
449 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
450 (only for MPI_COMM_TYPE_SHARED).
451 - New option: smpi/privatize-libs, to add external shared libs to be privatized
452 by SMPI. They will be copied locally and loaded separately by each process.
453 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
454 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
455 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
456 with internal index optind with multiple processes. Only works if MPI_Init has
458 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
459 syntax without preprocessing tricks.
462 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
463 behavior tracing is no longer limited to MSG
464 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
465 this means that traces now need to be updated or re-obtained; both Irecv
466 and Isend lines in your traces have now in total 5 mandatory fields:
467 <rankid> <command> <to/from rankid> <tag> <size>
468 To update your traces, it suffices to add a 0 for the tag here.
469 - TIT now also supports waiting for a distinct request via MPI_Wait.
470 Wait/Test now wait for a specific request, not just the last one that was
471 issued. This unfortunately means another update, because we need to
472 identify which request you want to wait for. We do this via the
473 triplet (sender, receiver, tag), which needs to be added:
474 <rankid> <command> <sender> <receiver> <tag>
475 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
476 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
479 - The deprecation of MSG is ongoing (but this should not impact you).
480 Many MSG functions are now simple wrappers on the C API of S4U. If
481 you wish to convert your code to S4U, find the S4U counterparts of
482 your MSG calls in src/msg/msg_legacy.cpp.
483 - MSG can still be used, but won't evolve anymore.
486 - Allow to run the Link energy plugin from the command line with
487 --cfg=plugin:link_energy
488 - Rename Energy plugin into host_energy
489 - Rename Load plugin into host_load
492 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
493 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
494 - Improve performance when handling timeouts of simix synchros.
497 - Config: the C API is now deprecated (will be removed in 3.23), and
498 the C++ API has been slightly improved.
501 - Fix several build issues on OSX.
502 - Move simgrid_config.h to simgrid/config.h (old header still working)
505 - #143: Setting a breakpoint at a given time
506 - #150: Inconsistent event names in SMPI replay
507 - #258: daemonized actors hang after all non-daemonized actors have completed
508 - #267: Linker error on unit_tmgr
509 - #269: SMPI: tracing of MPI_Wait/all/any broken
510 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
512 ----------------------------------------------------------------------------
514 SimGrid (3.19.1) Released March 22 2018
516 The Fixed ABI Release.
518 As you may know, we are currently refactoring SimGrid in deep.
519 Upcoming SimGrid4 will be really different from SimGrid3: modular,
520 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
522 Our goal is to smooth this transition, with backward compatibility
523 and automatic update paths, while still progressing toward SimGrid4.
525 SimGrid remains open during works: The last pure SimGrid3 release was
526 v3.12 while all subsequent versions are usable alpha versions of
527 SimGrid4: Existing interfaces remain unchanged, but the new S4U
528 interface is budding and the internals are deeply reorganized.
530 Since 2015, we work hard to reduce the changes to public APIs. When
531 we need to rename a public library symbol in S4U, we let your
532 compiler issue an explicative warning when you use the deprecated
533 function. These messages remain for four releases, i.e. for one full
534 year, before turning into an error. Starting with v3.15, your can
535 also adapt to API changes with the SIMGRID_VERSION macro, that is
536 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
538 Starting with this v3.19.1, our commitment to reduce the changes to
539 the public interfaces is extended from the API to the ABI: a program
540 using only MSG or SimDag and compiled against a given version of
541 simgrid can probably be used with a later version of SimGrid without
542 recompilation. We will do our best... but don't expect too much of
543 it, that's a really difficult goal during such profund refactoring.
545 The difference between v3.19 and v3.19.1 is that the former was
546 accidentally breaking the ABI of MSG, while the later is restoring
549 S4U and kernel APIs will still evolve until SimGrid4, with one-year
550 deprecation warnings as currently. In fact, cleaning up these
551 interfaces and converting them to snake_case() is one release goal of
552 v3.20. But don't worry, we are working to smooth this upgrade path.
554 In summary, new projects should start with S4U to benefit of the
555 future, but old MSG projects should still be usable with no change.
557 ----------------------------------------------------------------------------
559 SimGrid (3.19) Released March 20 2018
561 The Moscovitly-cold Spring release.
563 SMPI user-visible improvements:
564 - Support of MPI_Type_create_subarray
565 - Unimplemented MPI calls now abort (instead of issuing a warning)
566 Some harmless calls around errhandlers still only issue a warning.
567 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
568 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
569 is now possible (mmap privatization was refactored to this end).
570 Adding new SMPI instances should also be technically doable now.
572 SMPI internal cleanups:
573 - Removed index notion from SMPI -- no more "getPid() - 1"!
574 In tracing files, actors are called rank-<PROCESS-ID> but these are
575 NOT the real ranks (tracing several communicators is not supported yet).
576 - Deprecate smpi_process_index() (will be removed in 3.22)
577 - Better testing on classical proxy apps, through an external project:
578 https://github.com/simgrid/SMPI-proxy-apps/
581 - Use a graphical TOC to make it easier to find the documentation you need
582 - Revamp the MSG tutorial
585 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
586 - Remove parameter of MSG_process_killall().
587 Resetting the PID was bogus anyway (several actors could have the same PID).
590 - Execution->setHost() can be called after start(), to migrate it.
591 - Comm::test_any() is now implemented.
592 - s4u::Actor now has onCreation() and onDestruction() signals.
593 - Install some kernel header files for the users' plugins and more.
594 Warning, their API is really not stable yet.
595 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
596 include/simgrid/kernel/routing/*Zone.hpp
599 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
601 Build System and other cleanups:
602 - Remove unused run-time parameter "tracing/onelink-only".
603 - SimGrid now works with intel compilers (MC still buggy)
604 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
605 - tesh kills subprocesses on timeout (but not on windows)
608 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
611 - #194: Feature request: simgrid::s4u::Comm::test_any()
612 - #245: migrating an actor does not migrate its execution
613 - #253: Feature Request: expose clusters as objects
614 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
615 - #255: Tesh broken on Windows
616 - #256: Modernize FindSimGrid.cmake
617 - #257: Fix (ab)use of CMake install
619 ----------------------------------------------------------------------------
621 SimGrid (3.18) Released December 24 2017
623 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
625 MSG IS NOW DEPRECATED. Please use S4U for new projects.
626 - Support for MSG should not be removed from SimGrid before 2020,
627 but future evolutions will be limited to the integration of
628 user-provided patches.
631 - s4u::onDeadlock() signal, run before stopping the simulation.
632 - s4u::Exec: asynchronous executions (abstraction of background
634 - s4u::Host->getLoad() returns the achieved speed in flops/s
637 - Link::name() is deprecated and will be removed in v3.21, use
638 Link::getCname() instead.
639 - Mailbox::getName() changed to return a std::string, use
640 Mailbox::getCname() instead to get a char*.
641 - Storage::getName() changed to return a std::string, use
642 Storage::getCname() instead to get a char*.
643 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
644 to have both allocation and deallocation of the map in user space.
647 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
648 The amount of flops remaining to do is a vector, not a scalar.
649 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
650 name implies on both sequential and parallel tasks.
651 - Both changes fix GitHub's #223 using PR #237 as a basis.
652 Thanks Michael Mercier.
653 - Most examples were converted to S4U and hidden elsewhere at tests.
654 Rationale: we still want MSG to work; we want newcomers to use S4U.
657 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
658 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
661 - Switch to the faster dlopen privatization mechanism by default
662 - Documentation improvments
665 - Live migration is getting moved to a plugin. Dirty page tracking is
666 the first part of this plugin. This imply that VM migration is now
667 only possible if one this function is called:
668 - C/MSG: MSG_vm_live_migration_plugin_init()
669 - C/C++: sg_vm_live_migration_plugin_init()
670 - Java: Msg.liveMigrationInit()
671 For C and C++, "simgrid/plugins/live_migration.h" also has to be
675 - Define class simgrid::xbt::Path to manage file names.
676 - Removed unused functions:
677 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
678 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
679 xbt_graph_export_graphviz()
680 - xbt/str.h: xbt_str_join()
681 - Remove unused datatypes:
682 - xbt/heap.h: use std::priority_queue or boost::heap instead
683 - xbt/swag.h: use boost::intrusive::list instead
686 - New link_energy plugin for the consumption of the links.
687 - All of the operations on files and storage contents have been
688 packaged into a plugin (src/plugins/file_system). The current
689 public interface can be found in
690 include/simgrid/plugins/file_system.h
691 To use these functions you now have to initialize the plugin by
692 calling MSG_storage_file_system_init() just after calling
693 MSG_init() or sg_storage_file_system_init() just after creating
697 - Remove the undocumented/untested tag <include>
700 - Remove viva specific tracing as the tool is no longer maintained
703 - #248: Kill of finished processes leads to segfault
704 - #240: xbt_cond_wait_timeout should gracefully return for C apps
705 - #239: please implement signal s4u::onDeadlock()
706 - #230: segfaults when exit() without run()
707 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
708 - #223: MSG_task_get_flops_amount() not working with parallel tasks
709 - #222: Actor::kill() doesn't really kill and segfaults
710 - #221: odd LMM warning when killing an actor
711 - #120: Memory leak when the processes are forcefully killed
713 ----------------------------------------------------------------------------
715 SimGrid (3.17) Released October 8 2017
717 The Drained Leaks release: (almost) no known leaks despite the tests.
719 Even very long simulations will preserve your memory: our testsuite
720 stresses SimGrid for over 45mn on fast machines for a coverage of
721 over 80%, and there is only one single known leak, of about 4kb.
725 - Comm.detach(): start and forget about asynchronous emission. A cleanup
726 handler may be given to free resources if the comm cannot be completed.
727 - this_actor::send(mailbox) is now mailbox->put()
728 - New: simgrid::s4u::Comm::wait_all()
729 - New: Host.execute() for remote executions.
732 - Unused option network/sender-gap is removed.
735 - Deprecate MSG_task_isend_with_matching(): This unused feature
736 really complicates our internals. Will be removed in v3.20.
739 - Improved context termination. It is now possible to free resources, even
740 when a process is forcibly killed.
741 - ContextBoost: add support for Boost versions above 1.61.
744 - Bring back run-time option --cfg=exception/cutpath to remove exception
746 - Removed unused functions:
747 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
748 xbt_str_rtrim(), xbt_str_trim().
749 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
752 - Removed header files obsolete since SimGrid 3.12:
753 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
755 - #3: SD_exit should be made optional
756 - #120: Memory leak when the processes are forcefully killed
757 - #159: Threading test regression in Actor refcounting
758 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
759 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
760 - #186: Actor::killAll() segfaults if some process is blocked on wait()
761 - #191: VM migration and pstate
762 - #192: Updating the energy consumptions of all hosts crashes with VMs
763 - #195: All actors have PID=0 in the logs
764 - #204: Sometimes segfault with thread contexts and mmap privatization
765 - #222: Actor::kill() doesn't really kill and segfaults
766 - #225: Actor::kill() doesn't really kill when victims are doing a join()
768 ----------------------------------------------------------------------------
770 SimGrid (3.16) Released June 22. 2017.
772 The Blooming Spring Release: developments are budding.
775 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
776 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
777 - Java JAR file should be finally fully working on Mac OSX too.
779 XML platforms: Switch to platform v4.1 format.
780 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
781 - <zone> can be used as a synonym for the now deprecated <as>
782 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
783 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
784 - <actor> can be used as a synonym for the now deprecated <process>
785 - state_file and avail_file periodicity is now easier to express
786 (check the documentation) the old behavior should still work.
787 * Storage: not backward compatible, but it seems that nobody ever used it
788 - Remove attribute 'content_type' of <storage_type>: was never used
789 - Make attribute 'model' of <storage_type> optional: for future usage
790 - Remove Bconnection model property: was never (in)validated. Replaced by
791 the maximum of the read and write bandwidth as a resource constraint.
794 - New and Backwards Compatibility break:
795 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
796 When one wants to get the list of tasks whose states have changed during a
797 simulation round, s/he has to allocate and free a dynar and use it as
798 argument to this function. The former SD_simulate (double how_long)
802 - Allow multicore VMs, with the correct sharing (unless you overcommit)
803 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
806 - New function to update the consumption of all hosts at once.
807 - Fix the model for multi-core hosts, linear in the amount of busy
808 cores with an abnormality for pIdle (see paper).
809 BUG: Still not satisfactory for ptask on mono-cores.
812 - The netzone are now available from the MSG API.
813 The old names still work, but are now deprecated.
816 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
817 - New option: smpi/keep-temps to not cleanup temp files
818 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
819 mallocs mode is used, allows to change the size of the fake file used
820 (default 1MB), to potentially limit the number of mappings for large runs.
821 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
822 - Fortran ifort and flang compilers support
823 - New RMA calls supported (experimental):
824 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
825 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
826 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
827 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
828 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
829 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
832 - New: this_actor::isMaestro() returns whether we are in kernel mode.
833 - Behind the scene, ActivityImpl are now automatically refcounted.
834 Removing this nasty bottleneck will greatly simplify our internals.
837 - Replay: New function xbt_replay_action_get():
838 Retrieve the function previously associated to an event type.
839 - DROPPED FUNCTION: xbt_str_varsubst()
840 - DROPPED MODULE: strbuff. We don't need it anymore.
841 - DROPPED MODULE: matrix. We don't need it anymore.
842 - DROPPED MODULE: lib. We don't need it anymore.
844 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
846 SimGrid (3.15) Released March 22. 2017
848 The Spring Release: continuous integration servers become green
849 We fixed even the transient bugs on all target architectures:
850 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
851 Windows, FreeBSD, NetBSD.
853 - Rename NetCards to NetPoints.
854 This was intended to help NS3 users, but that's not a netcard.
855 That's a point in the routing algorithm, let's avoid wrong simplifications.
858 - New: MSG_process_yield(). Stop and yield to other processes.
859 - New: MSG_process_daemon(). Daemon processes are automatically killed
860 when the last non-daemon process terminates
861 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
862 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
863 to make room for the upcoming network energy plugin.
864 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
867 - Ensure that an actor can kill itself with Process::exit()
868 - Kill the obscure NativeException. Nobody want to survive the issues
869 it denotes, so use JniException that is a RuntimeException (not to
870 be caught explicitly).
871 - Partial bug fix in initialization. SimGrid flags on command line were
872 consumed at C level but stayed in the original Java String[] args.
873 This could mess users' args[i] if SG flags were not put at the end of
875 The SimGrid flags are now removed from the Java arguments. However,
876 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
877 if args.length is greater than the number of YOUR OWN ARGUMENTS.
878 It might be if you have --log or --cfg flags in the command line.
879 - Fix numerous memleaks all around the place. In particular, around VMs.
883 - s4u::onPlatformCreated: right before the simulation starts
884 - s4u::onSimulationEnd: right after the main simulation loop
885 - s4u::onTimeAdvance: right after a clock change (time is discrete)
886 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
887 event from the availability_file changes the avail speed.
888 - Links are now usable from s4u
889 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
890 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
891 - New examples: The conversion of MSG examples to S4U has begun
892 - Actors: create, kill, migration, and suspend
893 - Applications: master-worker and token-ring
894 - Action replay: communications and storage
895 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
898 - Backwards Compatibility breaks
899 - The SD_route_get_list and SD_route_get_size functions have been removed.
900 They are replaced by the unique
901 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
902 The route from src to dst is built in the links dynar whose size can
903 be obtained with xbt_dynar_length.
904 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
905 sg_host_route_bandwidth and sg_host_route_latency.
906 Macros ensure the backwards compatibility, but you should fix your code
909 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
910 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
911 - Bump our claim of support from MPI 1.1 to MPI 2.2.
912 We don't support 100% of it, but it should be enough. Contact us if not.
913 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
914 naive and probably inaccurate)
915 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
916 mapped onto a unique small file using some system magic.
917 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
918 - Fortran: cleanups, fixes, support of user-added operations
919 - RMA: MPI_Accumulate are applied in correct order.
920 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
921 - Better support for MPI_IN_PLACE
922 - Support for MPI_Win attrs and keyvals.
923 - Support MPI_Comm_group_create, MPI_Type_size_x
924 - MPI ops should be properly applied to all allowed types
925 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
926 algorithm by a non failing one from mpich.
929 - Kill the fifo data container: we don't use it anymore.
932 - A <cluster> can now be created with different speed values to represent pstates.
933 see examples/platforms/energy_cluster.xml for an example.
935 SimGrid (3.14.159) Released December 28. 2016
937 The previous release was only a crude approximation of a Pi release;
938 this one should be a bit better.
940 - Revert a last minute change that broke on Mac OSX.
941 - Fix the changelog and NEWS files.
942 - Improve the documentation, in particular of the routing module.
943 - Integrate some patches flying around in Debian.
945 SimGrid (3.14) Released December 24. 2016
948 * The whole documentation was reworked and reorganized. There is
949 still much room for improvement here, but we're on it.
952 * We now need python3 for our testing infrastructure.
953 * Model checking needs libevent
956 * Now works on FreeBSD
959 * Privatization now works on FreeBSD
961 * Privatization is now activated by default in smpirun
962 You can turn it off with -no-privatize if something goes wrong.
964 * Call-location tracing for SMPI.
965 You can add the exact location (filename / linenumber) of an MPI call to
966 your trace files and slow down or speed up the simulation between two
967 consecutive calls by using an adjustment file (see the documentation).
969 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
970 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
971 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
972 values were ignored in some cases. The timings of these functions can now
973 be significantly different.
975 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
976 smpi/running-power is renamed to smpi/host-speed
978 * smpi/grow-injected-times option to enable or disable multiplication of the
979 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
980 make simulation less precise (but also much faster).
982 * smpirun script should be (much) faster for large deployments.
984 * SMPI tracing: fixed issue with poor matching of send/receives.
986 * Replay: Fix broken waitall
988 New functions and features
989 * MSG_parallel_task_execute_with_timeout, to timeout computations.
991 Dropped / renamed functions and features
992 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
993 - MSG_mailbox_is_empty() -> Mailbox::empty()
994 - MSG_mailbox_front() -> Mailbox::front()
995 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
996 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
997 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
998 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
999 * Don't pass the free_f parameter to property related functions:
1000 - MSG_host_set_property_value()
1001 - MSG_as_router_set_property_value()
1002 - MSG_storage_set_property_value()
1003 * VM properties. Since msg_vm_t are msg_host_t, just use
1004 MSG_host_get_property_value() and friends
1005 * VM I/O related things:
1006 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1007 - Unimplemented save/restore methods
1008 * MSG_as_router_get_property_value() was redundant with
1009 MSG_environment_as_get_property_value().
1010 - Removed MSG_as_router_*propert*() functions
1011 - Added MSG_environment_as_set_property_value() for consistency
1012 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1013 Well, they are still there for now, but deprecated with a warning.
1014 Please switch to xbt_dict_new_homogeneous() before this is removed
1016 * Task affinity. Its intended behavior (that was very badly tested
1017 and probably not really working) was deceiving what most users
1018 would have hoped here.
1019 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1020 * Option network/coordinates is now useless and should be dropped.
1023 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1026 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1027 * Added Dragonfly topology support
1030 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1031 EiBps, EBps, Eibps, Ebps for bandwidth.
1032 They may become useful to some lucky ones.
1035 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1036 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1037 * ex/async_waitAll: new example, on asynchronous communications
1040 * Memory usage should be decreased for simulations with a large number
1041 of processes. This also helps for SMPI.
1043 SimGrid (3.13) stable; urgency=low
1045 The Half Release, a.k.a. the Zealous Easter Trim.
1047 - We removed half of the lines, that were mostly experimental cruft.
1048 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1049 (not counting blanks and comments -- according to openhub.net)
1050 - The internals are now compiled in C++ (and will soon be clean C++)
1051 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1052 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1054 * Backwards Compatibility breaks
1055 - Removed Lua simulation bindings (switch to C or Java for that).
1056 Lua can still be used to describe platforms
1057 - Removed Java kernel plug-ins.
1058 Will be reintroduced after the ongoing major internals reorg.
1060 - the following functions were removed.
1061 They were too specific and should be reimplemented in a generic
1062 way, with filter function.
1063 - MSG_task_listen_from_host
1064 - MSG_mailbox_get_count_host_waiting_tasks
1065 - MSG_mailbox_put_with_timeout was removed.
1066 Please use MSG_task_send_with_timeout instead.
1068 - the SD_application_reinit function was removed. It has been a noop for a while.
1069 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1070 replaced by a more flexible mechanism.
1071 - The following functions thus do not exist anymore
1072 - SD_workstation_get_access_mode
1073 - SD_workstation_set_access_mode
1074 - SD_workstation_get_current_task
1075 - Basic estimation functions have been removed but can easily be replaced
1076 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1077 - SD_workstation_get_computation_time => amount / sg_host_speed()
1079 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1080 Use VM.getSpeed()*load/100 for the legacy behavior.
1082 - option enable_tracing was removed. It was not doing anything for a while.
1083 - In the ModelChecker:
1084 - the model-checker now ptraces the model-checked process which means
1085 you cannot use a debugger on the latter anymore (we might make this
1086 optional in the feature);
1087 - removed soft-dirty page tracking;
1088 - remove model-checked side snapshot management,
1089 MC_snapshot() and MC_compare_snapshot();
1090 - keep the MC_cut() function as a stub (it was not really working
1091 in the previous release).
1094 * All options are consistently in kebab-case. Old names are kept as alias.
1097 * Switch to platform v4 format.
1098 - Rename from 'power' to 'speed' the attributes describing the amount of
1099 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1100 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1101 - In <host> and <link>, attributes availability and state are gone.
1102 It was redundant with state and availability traces, and with peak values.
1103 - In <cluster>, attributes availability_file and state_file are gone.
1104 It was too complex and unused.
1105 - Kill <gpu>. Was not doing anything.
1106 - The DOCTYPE points to the right URL:
1107 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1108 (the file at this address now documents the changelog since its v1)
1109 - A warning is emitted for unit-less values (they are still accepted).
1110 - speed. Default: 'f' or 'flops'. Also defined:
1111 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1112 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1113 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1114 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1115 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1116 - latency. Default: 's' second. Also defined:
1117 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1119 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1120 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1123 * s4u::Host is now the preferred public interface to the Host features.
1124 sg_host_* functions are C bindings to the exact same behavior
1125 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1128 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1129 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1130 They were deprecated since a few years, and probably did not even compile anymore.
1133 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1134 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1135 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1136 encouraged to update their codes. A list of the modified functions can be found at the end of
1137 include/simgrid/simdag.h
1140 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1141 the lambda or closure passed as a parameter will run in kernel mode.
1142 All the callback functions should be rewritten to that interface at some point.
1145 * Reorganizing and cleaning the internals all around the place.
1148 * Remove old default barrier/bcast buggy algorithms (see #18407)
1149 * Various bug fixes to handle more codes
1150 * Remove the need for the --foreground option of smpirun (it is still
1151 accepted for backward compatibility).
1154 * Kill the setset data container: MC don't use it anymore.
1155 * Kill the queue data container: it made more sense with GRAS.
1156 * Kill the xbt_peer_t data type: it's useless without GRAS.
1157 * Kill rm_cb feature of config sets: it was never useful.
1158 * Kill graphxml parsing feature. It was not used.
1159 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1161 - xbt_dynar_sort_strings(), when the content is char*
1162 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1163 They throw exceptions on invalid input;
1164 * C++ support for declaring CLI flags (simgrid::config::Flag);
1165 * class for abstracting different signal backends (simgrid::xbt::signal).
1166 with no external dependencies (we need very simple signals).
1169 * refactoring and cleanup of the code;
1170 * ongoing process to cleanly separate the model-checking algorithms
1171 from the code model-checking support.
1173 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1175 SimGrid (3.12) stable; urgency=low
1177 The Facelift Release.
1180 * Require g++ v4.7 at least to not speak prehistorical C++.
1181 * Require Boost 1.48 (for signal2 component).
1182 * Java must be version 7 at least when activated.
1183 * Builds on Windows again (including Java bindings).
1184 * Tracing is now always enabled (no way to turn it out)
1185 * Remove GTNetS. It was not working anyway.
1186 * Various cleanups in the cmake scripts.
1187 * Move headers around to sort them out on installed systems:
1188 - instr/instr.h -> simgrid/instr.h
1189 - instr/jedule/* -> simgrid/jedule
1190 - simdag/datatypes.h was removed
1191 - simdag/simdag.h -> simgrid/simdag.h
1192 - msg/datatypes.h was removed
1193 - msg/msg.h -> simgrid/msg.h
1196 * Interface improvement:
1197 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1198 - Sanitize the interface in MSG_task_ module:
1199 - Merge two functions that were close enough but misleading:
1200 set_compute_duration(t) -> set_flops_amount(t)
1201 get_remaining_computation(t) -> get_flops_amount(t)
1202 - set_data_size(t) -> set_bytes_amount(t)
1203 get_data_size(t) -> get_bytes_amount(t)
1204 - Massive cleanups in the functions related to the energy
1205 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1206 - New: MSG_host_get_pstate()
1208 - msg/energy/onoff: switching hosts on and off
1210 * Interface improvement:
1211 - Rename Host.isAvail() to Host.isOn()
1212 - Rename Process.currentProcess() to Process.getCurrentProcess()
1213 - Rename Task.setDataSize() to Task.setBytesAmount()
1214 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1216 - #18874: Actually allows the GC to reclaim tasks
1220 - SIMIX_process_throw: raises an exception in a remote process
1221 * Refactoring: Separate sub-modules
1222 - libsmx: the public interface, as libc in a real system
1223 - popping: the strange dance that converts a user request into a kernel handling
1224 - smx_context_*: the virtualization mechanisms that embed the user code
1225 - smx_*: the handling of each simcalls
1226 * Interface cleanup:
1227 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1228 * Rename smx_action_t into smx_synchro_t, making explicit that these
1229 things are used to synchronize processes with their environment.
1230 For example, a communication is a sort of synchronization involving
1231 the communicating processes (that may block until the exchange) and
1232 the platform. The same can be said from computations, etc.
1234 - #18888: segfault when a process ends before its kill_time
1238 - 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.
1239 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1240 - MPI_Type_set_name, MPI_Type_get_name
1241 - MPI_*_c2f and MPI_*_f2c functions
1242 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1243 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1244 - Activate a lot of new tests from the mpich 3 testsuite
1246 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1247 - InfiniBand network model added: Based on the works of Jerome Vienne
1248 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1249 - When smpi/display_timing is set, also display global simulation time and application times
1250 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1251 * Collective communications
1252 - 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
1253 - 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/).
1254 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1256 - Replay now uses algorithms from wanted collective selector
1257 - Replay can be used with SMP-aware algorithms
1258 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1259 - 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.
1261 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1262 - Correctly use loopback on fat-tree clusters
1263 - Asynchronous small messages shouldn't trigger deadlocks anymore
1264 * Energy/DVFS cleanup and improvement
1265 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1266 - new: smpi_get_host_pstate()
1269 - "Full" network optimization flag was broken since Surf++
1270 - Better handling of precision flags in maxmin
1271 - Fix bug causing sometimes "Impossible" errors
1272 - Properly pass cluster properties to included hosts
1273 * Improvement of the Energy plugin.
1274 - Always update the consumption before returning that value
1275 - New property: watt_off to denote the dissipation when the host is off
1276 - New functions getWattMinAt and getWattMaxAt to retrieve the
1277 dissipation of pstates that we are not currently at.
1278 * Java: class NetworkLink renamed to Link
1279 * New function: simcall_process_get_kill_time()
1280 * Massive rename s/workstation/host/
1281 - That's intrusive, but that's good for the project consistency. Sorry.
1282 - Change config option "workstation/model" into "host/model"
1285 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1286 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1287 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1288 - xbt_dirname and xbt_basename for non-POSIX systems
1290 * The model checker now runs as a separate process.
1291 * The model checker runs must now be launched with the new simgrid-mc program.
1292 * Record/Replay: the MC can display a textual representation of a path in the
1293 execution graph. It can then be replayed outside of the model checker.
1295 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1297 SimGrid (3.11) stable; urgency=low
1302 * Normalizing pointers addresses tool for better diff between logs
1305 * Add cloud examples using new VMs
1306 - examples/msg/cloud/two_tasks_vm.tesh
1307 - examples/msg/cloud/simple_vm.tesh
1308 - examples/java/cloud/cloud.tesh
1309 - examples/java/cloud/migration/migration.tesh
1310 * Add java surf examples:
1311 - examples/java/surfPlugin/surf_plugin.tesh
1312 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1313 - examples/java/surfCpuModel/surf_cpu_model.tesh
1314 * Add SMPI+MSG example:
1315 - examples/smpi/smpi_msg_masterslave/
1320 - msg task destroy cancel test
1321 - msg_host on/off test
1322 * Move all tests in testsuite to teshsuite (adding tesh files)
1323 * Restructure teshsuites
1324 - one folder for each kind of test
1325 * Restructure AddTests.cmake
1327 - structure the order of tests (with sections)
1330 * Add virtual machine
1331 - creation of a VM on a PM
1332 - migration of a VM from a PM to another PM
1334 - MSG_process_join(msg_process_t process, double timeout)
1335 - msg_bar_t MSG_barrier_init(unsigned int count)
1336 - int MSG_barrier_wait(msg_bar_t barrier)
1337 - void MSG_barrier_destroy(msg_bar_t barrier)
1338 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1339 * New option "msg/debug_multiple_use" to help debugging when a task is used
1342 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1344 - Increase code coverage in test suites
1348 * Protect context stacks against stack overflow. The number of protected memory
1349 pages allocated on the top of each stack (1 by default) can be configured
1350 with the new command line option --cfg=contexts/guard_size:<value>.
1351 * Simcalls are now generated by a python script that
1352 - generates files included by SimGrid
1353 - checks that all the functions exist, and proposes prototypes
1355 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1356 host_set_data, host_get_data
1358 - simcall_process_join(smx_process_t process, double timeout)
1359 * Fix bug where sleeping processing could not be suspended.
1362 * Translate surf models from C to C++
1363 - Generic classes for all models: Model, Resource, Action
1364 - A generic interface for each kind of model (CPU, Network, Storage
1365 Workstation, WorkstationVM)
1367 * Translate surf routings from C to C++
1368 * Add callbacks using sigc++ or boost::signals2
1369 - Add callback functions for resource creation/destruction
1370 - Add callback functions for action state change
1371 - Handle Energy as a plugin
1372 * Replace swag by boost::intrusive
1373 * Add new routing models for clusters. For documentation, see
1374 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1375 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1376 parameters for cluster tag
1377 - Fat trees, with topology="FAT_TREE" and
1378 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1379 - see examples/platforms/torus_cluster.xml and
1380 examples/platforms/fat_tree_cluster.xml
1381 * More documentation
1384 * Hostfiles support host:nb_processes construct to deploy several processes on
1386 * Collective communication algorithms should not crash if used with
1387 improper number of nodes and report the error.
1388 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1389 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1390 MPI_Dims_create, MPI_Cart_sub are supported.
1391 * New interface to use SMPI programmatically (still depends on MSG for
1392 some parts, see examples/smpi/smpi_msg_masterslave):
1393 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1397 * Global variables privatization in MPI executables is now performed at runtime
1398 with the option smpi/privatize_global_variables (default:no).
1399 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1400 dynamic libraries loaded by the application are not privatized (static
1401 linking with these libraries is advised in this case)
1404 * Options defined in XML work correctly now.
1407 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1408 copied into simgrid.jar or not (ON by default). Use this option if you want
1409 to reduce the size of the installed simgrid.jar, *and* the native libraries
1410 are kept installed elsewhere.
1411 * Surf binding with SWIG (code generated in maintainer mode only):
1412 - plugin to handle callbacks
1413 - CPU model only for the moment
1416 * Supernovae build mode is definitively removed. It was used to improve
1417 inlining and inter-module optimizations. It is nowadays superseded by
1418 link-time optimizations commonly available in compilers.
1419 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1420 * Add boost dependency for surf++
1421 * Add new macro for tests
1422 - ADD_TESH(name <tesh_args>)
1423 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1427 - xbt_bar_t XBT_barrier_init(unsigned int count)
1428 - int XBT_barrier_wait(xbt_bar_t barrier)
1429 - void XBT_barrier_destroy(xbt_bar_t barrier)
1430 * Make the xbt_os_time module public
1432 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1434 SimGrid (3.10) stable; urgency=low
1436 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1439 * Reintegrate Java to the main archive as desynchronizing these
1440 package is not acceptable anymore (Java is now considered stable)
1441 * Add explicit synchronization facilities through semaphores
1442 * Bug fix: Task.setDataSize() only changed the C world, not the value
1443 cached in the Java world
1446 * Dramatically change the way files are handled. API and internals changed, but
1447 this part of MSG was not considered as production grade either.
1448 * Add explicit synchronization facilities through semaphores
1449 * Add a new function MSG_host_get_process_list()
1450 * Preliminary DVFS support (see examples/msg/energy/ for details)
1453 * SMPI is now included directly in the libsimgrid as the windows
1454 linker doesn't force us on splitting it anymore.
1455 * Improvements of the SMPI replay tool:
1456 - Most of the collective communications are now rooted in the same process as
1457 in the original application.
1458 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1459 used until now). Multiple data types can now be used in a trace.
1460 - The replay tool now supports traces produce either by TAU or a modified
1462 - Bug Fix: the compute part of the reduce action is now taken into account.
1463 - Gatherv collective is now supported
1464 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1465 of smpirun outputs time independent traces for the current run. One file
1466 is created per process. If too many processes are simulated, this behavior
1467 can be changed to one file for all processes by using the
1468 tracing/smpi/format/ti_one_file flag
1469 * smpirun generates the host file if needed (with given host count and platform)
1470 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1471 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1472 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1473 --cfg=smpi/coll_selector:(mpich/ompi)
1474 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1476 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1477 happening inside a collective SMPI call.
1478 * Fix the behavior of complex data types handling.
1479 * Make MPI_Wtime another synchronization point to take computations into
1481 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1482 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1483 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1484 (no privatization of global variables yet)
1485 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1488 * Verification of liveness properties is now available for SMPI applications
1489 (in addition to MSG applications)
1490 * Bugged examples using SMPI in examples/smpi/mc/
1491 * Add --cfg=model-check/visited option. Allows the verification of infinite
1492 programs. Detection of loops in the execution thanks to the system state
1493 comparison and reduction of the state space to explore. Can be combined with
1494 DPOR for safety properties.
1497 * Allow to change SimGrid configuration (see --help) within the code
1498 thanks to SD_config() as it can be done in MSG.
1499 * Add a new function SD_task_set_amount() upon user request.
1502 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1503 * Remove rule based routing (no more PCRE dependency)
1504 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1505 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1506 * Add a loopback_bw and loopback_lat options to cluster tag.
1507 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1508 routing. Now peers are encapsulated in an AS and have their own private
1509 router but this is transparent.
1512 * Our own implementation of getline is renamed xbt_getline, and gets
1513 used even if the OS provide a getline(). This should reduce the
1514 configuration complexity by using the same code on all platforms.
1515 * New type: xbt_cfg_elm_boolean.
1516 * Allow to use yes/no for boolean configuration options in the command line.
1517 * Allow to disable SimGrid cleanups at exit from command line option.
1518 There are situations where one may want a simulation to end with an exit.
1519 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1520 annoying when scripting around the simulator. Adding a
1521 --cfg=clean_atexit:no allows to circumvent this issue.
1524 * Lots of memory leaks were corrected in this release.
1525 There is no known memory leaks anymore, in all of our 600+ tests.
1526 * New command line option --version, to get SimGrid version information.
1527 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1529 * Supernovae builds are deprecated, and expected to be removed in the next
1532 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1534 SimGrid (3.9) stable; urgency=low
1536 The Grasgory release: GRAS is really dead now.
1538 * Complete overhaul of the internal host structures scheme.
1541 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1542 was considered as experimental and badly maintained since then.
1543 * Keeping it was thus a trap to our potential users, that could take
1544 it instead of MSG or SMPI by mistake despite is pity state.
1545 * GRAS seems to have very few users (if any), and no one volunteered
1546 to maintain it further. It also induces a lot of XBT code (for
1547 portability sake), that must be maintained too.
1548 * For all these reasons, we killed GRAS. If someone wants to revive it
1549 in the future, don't cry, our git history still remembers of GRAS.
1552 * Major overhaul. Merge our documentation again as time proved that
1553 splitting it was really not helping our users.
1554 * Further improve the developer documentation to help newcomers
1555 hacking on SimGrid itself. The user documentation (and in
1556 particular, the beginner documentation) is still in a sorry state.
1559 * Now works on Windows too!
1560 * Much more extensive test suite, from MPICH
1563 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1564 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1565 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1566 * Bug fix: let task be scheduled when the last dependency to be solved is
1567 a control dependency.
1568 * Remove SD_load_environment_script function.
1569 Use the C sg_platf function if you want to declare a platform
1573 * New function: MSG_process_get_number()
1574 * Old function documented: MSG_config()
1575 * Remove MSG_load_platform_script function
1576 Use the C sg_platf function if you want to declare a platform
1580 * Change the default value of the TCP_gamma constant (maximal size of TCP
1581 congestion window) to a more realistic 4MiB value. If you notice changes in
1582 your simulation results, you can fall back to the previous 20k tiny window
1583 by adding --cfg=network/TCP_gamma:20000 on command line.
1584 * (Hopefully) fix a bug wrt periodic availability and state traces
1585 * Bug fix: use default values at start when first event in availability/state
1586 trace is not at time 0.
1589 * remove the "new_" part of function name sg_platf_new_trace_connect
1590 (resulting in sg_platf_trace_connect), since it does not create
1594 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1595 use case where it's really mandatory, and maintaining it was a pain in
1597 * New: xbt_fifo_search(), search an item with a user-provided
1598 comparison function instead of dumb pointer comparison.
1601 * Fix the lua deployment:
1602 Use `simgrid.init_application()` before deployment instead of
1603 `simgrid.msg_register_application()` after.
1606 * Transfer the tracing files into the corresponding modules.
1608 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1610 SimGrid (3.8.1) stable; urgency=low
1612 The "we are told that some people want to also *install* the simgrid
1615 * Add missing file "tesh.1" to the archive.
1617 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1619 SimGrid (3.8) stable; urgency=low
1621 The Psssshiiiit release: SimGrid jumps into the Cloud.
1624 * Add an experimental interface to manipulate VMs. They are mainly
1625 process groups with very few intrinsic semantic, but they should
1626 allow you to build the semantic you want easily.
1627 * New function: MSG_host_set_property_value()
1628 * New function: MSG_process_on_exit(). To clean memory in all cases.
1629 * Bug fixes that made the host (and link) failures unusable.
1630 * Add a way to auto-restart process when the host in which they are
1631 executing comes back (ON_FAILURE="RESTART" on deployment file,
1632 MSG_process_auto_restart_set).
1633 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1634 please stop using the old ones, they are DEPRECATED.
1636 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1637 Please use MSG_init() instead. (reducing the amount of entry
1638 points in the library helps us).
1639 * Make it impossible to link against the wrong version of the lib
1640 * Deprecate MSG_clean(). No need to call it anymore.
1641 * Function MSG_get_host_number() is not deprecated anymore.
1644 * Split the doc into a user guide and a reference guide.
1645 * Start a developper guide to help people hacking on SimGrid.
1648 * Enable tracing by default. This modules rocks you should use it.
1649 * Remove option custom_flags. Now use environment variables CFLAGS
1651 * Use default cmake things to detect lua instead of home grown ones.
1652 * New option "enable_mallocators" to disable mallocators, for debugging
1653 purpose ("on" by default).
1656 * Bug fixes around the resource failures: don't let the processes
1657 survive the host they are running onto.
1658 * Add an interface to auto-restart processes when the host in which they are
1659 executing comes back.
1660 * Ensures that SIMIX_clean is called automatically. It's not part of
1661 the public interface anymore (bindings should be updated).
1664 * Bug fix for when SD_Simulate is called with a positive value: be careful
1665 when comparing doubles. Sometimes they are different for non significant
1667 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1668 parallel task whose initial work is distributed among host according
1669 to the Amdahl's law. Such tasks are created with a parameter alpha
1670 that corresponds to the non-parallelizable part of the computation.
1671 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1672 between two sets of workstations assuming a 1D block distribution (each
1673 workstation owns a similar share of data) on both sides.
1675 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1676 Data redistribution will be automatically scheduled once parent and child
1677 are both scheduled. The filling of computation_amount and
1678 communication_amount structures is now done seamlessly thanks to the chosen
1680 * New function SD_workstation_dump to display various information
1681 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1682 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1683 bandwidth on the route onto which the task is scheduled and the amount of
1685 To divide the nominal bandwidth by 2, the rate then has to be:
1686 rate = bandwidth/(2*amount)
1687 * Compute tasks that have failed can now be rescheduled and executed again
1688 (from their beginning)
1689 * Increasing source code coverage (src/simdag is now covered at 95.8%
1693 * Re-implement time-independent trace replay using SMPI (at the
1694 smpi_smp_* level) instead of MSG. This should replace
1695 examples/msg/actions/actions.c
1696 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1697 hindexed and structs)
1698 * Implement the exchange of non-contiguous data.
1699 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1700 * Correct behavior of smpi/sender_gap and set its default value to 0
1701 * Add option to asynchronously send small messages to allow better
1702 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1703 specifies the size in bytes under which messages will be asynchronously sent.
1704 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1705 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1708 * New C interface to define a platform: XML is now optional.
1709 For more info, please check include/simgrid/platf.h.
1710 * New interface to define random platforms from the C:
1711 For more info, please check include/simgrid/platf_generator.h and
1712 examples/msg/masterslave/masterslave_platfgen.c
1713 * Export a sg_cmdline dynar containing all the arguments we got from
1717 * Two new tracing options for adding comments to trace file so you
1718 can track your experiments (see --help-tracing for details).
1719 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1720 * Adding the SimGrid version that generated the trace file as a comment.
1721 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1722 * Fix to avoid key clashes on Paje links
1723 * Other minor fixes related to the Paje specification
1726 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1727 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1728 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1729 * Greatly improve the robustness of mmalloc to user errors (such as
1730 using an area after freeing it, or freeing it twice)
1732 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1734 SimGrid-java (3.8.1) stable; urgency=low
1736 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1737 Warning: all methods are visible, but only some of them are
1738 implemented so far. Check the source (src/jtrace.c)
1739 for further information.
1740 * New module: org.simgrid.msg.File (SimGrid File management functions)
1741 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1742 * Change the meaning of Process.restart: now restart the process from
1743 the beginning, like MSG_process_restart in C.
1744 * Add Process.setAutoRestart: handling of process restart when failed
1746 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1747 you to retrieve the properties of the processes/hosts
1748 * Deprecate Msg.clean(): you can just forget about it now.
1749 * New function Process.getCount(), that only works when compiling
1750 with the not yet released version 3.9 of the C library.
1752 * New context factory based on Coroutines. It mandates a modified JVM
1753 but then, the simulations run about five times faster, and there is
1754 no limit to the amount of processes (beside of the available memory).
1756 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1758 SimGrid (3.7.1) stable; urgency=low
1761 * Restore the prototype of MSG_process_create_with_environment() to
1762 the pre-3.7 situation by removing the kill_time argument.
1763 * Add a MSG_process_set_kill_time() function instead.
1766 * Fix weird behaviors when dealing with parallel tasks.
1769 * Simgrid is now built as a dll.
1770 * Simgrid-java now works on Windows.
1771 * Simgrid-Java is now included into Windows package.
1774 * First pre-build package for MacOSX.
1777 * Fix compilation when using MSG_USE_DEPRECATED.
1778 * Fix some compilation issues on Macs and Windows.
1779 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1780 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1782 We discovered that the Lua console is broken, but we are missing the
1783 manpower to fix it right now. The problem existed in 3.7 too, so we
1784 are not blocking the release for that. Sorry if you depended on this
1785 feature, any help would be really welcome.
1787 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1790 SimGrid-java (3.7.1) stable; urgency=low
1792 The "Java aint got to be bloated and slow" release
1795 * Various internal cleanups and performance improvement
1796 Simulations are expected to run up to twice faster or so
1797 * Make Process.kill(process) an instance method, not a static one
1798 * User processes are not java.lang.Thread subclasses.
1799 This breaks the compatibility (sorry), but previous API was
1800 brain-dead, making it impossible to have non-trivial
1801 initializations in the process constructor.
1802 * Require a full constructor per Process sub-class.
1803 Kinda breaks the compatibility (sorry), but this allows a much more
1804 efficient way to launch the processes at simulation startup.
1805 * Do not embeed our version of semaphores, java 1.5 can be considered
1806 as sufficiently prevalent for us to not dupplicate its features.
1810 * Add examples for almost every part of the API
1811 We spotted and fixed a lot of bugs in the process
1813 * New module: asynchronous communication API
1814 * New function: Process.sleep()
1815 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1816 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1817 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1819 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1821 SimGrid (3.7) stable; urgency=low
1823 The "spring cleanups (before the next Big Project kicks in)" release.
1826 * We can specify the SMPI latency/bandwidth factor with command line
1827 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1828 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1829 You can also use the "config tag" from platform file by adding this line
1830 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1831 (see "example/platforms/tag_config.xml" to use "config tag").
1832 Note that the command line supersedes the platform file configuration.
1833 * Change the correction factors used in LMM model, according to
1834 the latest experiments described in INRIA RR-7821.
1835 Accuracy should be improved this way.
1836 * Use the partial invalidation optimization by default for the
1837 network too. Should produce the exact same results, only faster.
1838 * Major cleanup in surf to merge models and split some optimization
1839 mechanisms from the core of the models. As a result you can now
1840 specify which model to use (e.g., --cfg=network/model:LV08
1841 --cfg=cpu/model:Cas01) and which optimization mode to use
1842 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1843 Incompatible combinations should err at initialization. See
1844 --help-models for the list of all models and optimization modes.
1845 * The CLM03 workstation model was dropped for simplicity because it
1846 used the deprecated CM02 network model. Use default instead.
1847 * Rename the TCP_gamma configuration option to network/TCP_gamma
1848 * Rename the coordinates configuration option to
1849 network/coordinates, and document it
1850 * Use now crosstraffic keyword instead of the terribly misleading
1851 fullduplex keyword. It is activated by default now in the current
1852 default model, use --cfg=network/crosstraffic:0 to turn it off.
1853 * Ongoing refactoring the model parsing to make XML files optional
1854 See include/simgrid/platf.h for details (still to be completed)
1857 * Major overhaul of the documentation. Almost instructive now :/
1858 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1859 functions and friends. This interface was considered as
1860 deprecated since over 2 years, it's time to inform our users that it is.
1861 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1862 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1863 use these (crappy) functions in your code.
1864 These functions will be removed soon. Stop using them now.
1865 * Deprecate MSG_get_host_{table,number}
1866 Implement MSG_hosts_as_dynar() instead.
1867 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1868 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1871 * Stabilize the parallel execution mode of user contexts
1872 * Introduce configuration variables to control parallel execution:
1873 - contexts/synchro: Synchronization mode to use when running
1874 contexts in parallel (either futex, posix or busy_wait)
1875 - contexts/parallel_threshold: Minimal number of user contexts
1876 that must be part of a scheduling round to switch to parallel
1877 execution mode (raw contexts only)
1878 * Fix bugs that prevented to use suspend/resume along with
1879 synchronization structures.
1880 * Fix bugs in process termination that lead to invalid memory access
1881 in very specific conditions.
1884 * Introduce a parallel mode for the models (controlled by surf/nthreads
1885 configuration item). In our tests, running the models in parallel
1886 never lead to any speedups because they are so fast that the gain
1887 of computing each model in parallel does not amortizes the
1888 synchronization costs, even when ultra fast futexes are used.
1889 This is released anyway because YMMV.
1892 * Performance boost by using a swag internally to compute the set of
1893 tasks that are finished and should constitute the return value of
1897 * Enable it by default now that it is considered rather stable.
1900 * Documentation of the tracing functions.
1901 * Performance gains when tracing categorized/uncategorized resource
1902 utilization by avoiding calls to get route when updating resource
1903 variables. LMM constraints are being used instead.
1904 * API changed to set task categories. Use MSG_task_set_category instead
1905 of TRACE_msg_set_task_category, and SD_task_set_category instead
1906 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1907 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1908 * Examples for link user variables added (see at examples/msg/tracing/)
1909 * Deprecated function TRACE_msg_set_process_category completely removed
1910 * Trace header updated according to the latest Paje file format
1911 * Tracing network lazy updates, no longer obligate users to use full updates
1912 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1913 * Experimental: let user code declare/set/push/pop application states for hosts
1914 * API changed to allow the manual creation of graph configuration files
1915 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1918 * Improve the API of Lua MSG bindings, using the Lua spirit.
1919 * Each simulated process now lives in its own Lua world (globals are
1920 automatically duplicated). It helps writing simulators. Will allow
1921 to run Splay programs within SimGrid in the future.
1922 * Add a Chord example in Lua, equivalent to the MSG one.
1925 * Start the implementation of a solution to express temporal
1926 properties, not only local assertions. This is still an
1927 experimental work in progress, stay clear from it to be safe.
1931 - Add new runtime parameters --help-logs and --help-log-categories
1932 to display information about supported logging parameters and
1934 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1935 don't exists anymore.
1936 * Mallocators: allow value NULL for the reset function.
1938 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1939 create homogeneous dictionaries, where all the elements share the
1940 same free function. Non homogeneous dictionaries will be
1941 deprecated in the next release.
1942 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1943 - Multi-level dictionaries are deprecated.
1945 - new function xbt_dynar_search_or_negative() that is useful when
1946 you have less than 2 million elements in your dynar and don't
1947 want of the extra complexity of catching exceptions when the
1948 element is not found.
1950 - Make xbt_os_thread module (for thread portability) public.
1951 Documentation is still to come, sorry.
1953 - Cleanups and simplifications to make it maintainable again.
1954 - Exotic features (such as memalign and valloc) were removed.
1955 - The metadata were extended and improved so that the
1956 model-checker becomes able to explore and inspect the heaps.
1957 - This may induce a performance drop when enable_model-checking is
1958 ON in cmake (even if it's not used in the simulation), but it is
1959 necessary at this point to get MC working.
1961 Turn model-checking OFF if simulation performance matters to you.
1962 Not enabling it at runtime is not enough, disable it in cmake.
1964 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1967 SimGrid (3.6.2) stable; urgency=low
1969 The "Not coding new stuff allows to polish old things" release.
1972 * New bindings to the NS3 packet level simulator (experimental)
1973 * Use the raw (efficient) execution contexts instead of the sysv
1974 (portable) ones when possible.
1975 * libpcre is now mandatory in any cases since not using it led to
1976 severe performance loss and possibly other issues
1977 * Update the XML platforms:
1978 - G5K: include the latest machine in Nancy
1979 - GridPP and LCG: new platforms
1980 * Documentation was partially updated, at least (more to come)
1982 Bug fixes, cosmetics and small improvements
1983 * Free terminated processes before the end of the simulation to avoid
1984 exhausting the memory of users having very dynamic amount of
1986 * Bug fix and cosmetics about canceling non-running tasks
1987 * Bug fix about the dot loader's issues when using libcgraph
1990 * Create an installer for windows with nsis (amd64 and win32)
1991 - Add an hello world project to illustrate simgrid project creation.
1992 - Embed libpcre into the Simgrid installer to avoid
1993 its compilation burden
1994 * The raw execution contexts should work on Apple now
1995 * Port to Windows 64 bits
1996 - Sysv contexts now have an implementation for this arch
1997 - GRAS communication features now support this arch
1998 * Drop support for borland compiler on windows
1999 - this code was not maintained, and we kinda depend on gcc nowadays
2000 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2001 * Fix portability issue on unstable ubuntu: linker became picky on
2004 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2006 SimGrid (3.6.1) stable; urgency=low
2008 The "Oops, we broke Macs too" release
2011 * Fixed contexts detection so that raw ones are used when possible
2012 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2013 leads to a strange error, with user code segfaulting sometimes when
2014 the generated code is not perfectly aligned (which is not
2015 controllable from the user side, depends on the amount of code)
2018 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2020 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2022 SimGrid-java (3.6) unstable; urgency=low
2025 * Split of every thing from simgrid v3.5 into a separate package.
2027 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2029 SimGrid (3.6) stable; urgency=medium
2031 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2034 * Bindings now constitute their own package, separated from the main one.
2035 Rationale: reduce our maintenance nightmare by reducing the module coupling
2036 They will soon be released on their own on gforge.
2038 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2039 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2041 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2042 * It's not quite deprecated for now because we have no replacement,
2043 but it may soon become the case.
2046 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2047 * Fortran: New user-level cache variable to store the rank of the running
2048 process. This improves performance by an order of magnitude.
2049 * C: New coccinelle script to automatically locate and modify global and
2050 local static variables.
2051 * Improved SMPI network model with a sender-side gap to account for multiple
2055 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2056 only say if a communication is finished, no matter whether it succeeded or
2057 failed. You can call MSG_comm_get_status() to know the status of a finished
2059 * New function MSG_task_dsend() to send a task and detach it. When a
2060 communication is detached, you are never notified of its success or failure
2061 and the memory is released automatically once it is finished. This function
2062 is useful when you don't care about the end nor the success of a
2064 * Change the prototypes of action replay. Sorry for inconvenience,
2065 but this is really more efficient this way (and to adapt your code,
2066 you just have to fix the initialization, that shouldn't be too long)
2067 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2068 ever managed to do anything useful with it.
2069 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2070 takes a dynar of communications. It returns immediately and gives the
2071 index of a finished communication (if any).
2072 * New example: a basic implementation of the Chord P2P algorithm.
2075 * New model for multi-core CPUs. You can now use the core attribute to
2076 precise the number of cores of a host. This is a basic model. Every
2077 process running on the host receives at most the power provided in
2078 the DTD (throughput<=power). Total throughput of process cannot exceed
2080 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2081 router linked by an up-link and a down-link (possibly asymmetrical).
2082 This kind of pattern allows to easily build last-mile model style platforms.
2083 Aggregating such patterns in a rule-based AS is thus the technique of
2084 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2085 * New model for Vivaldi routing. We transformed the Vivaldi network model
2086 into a Vivaldi routing model (based on the rule-based model). This allows to
2087 combine Vivaldi based latencies with last-mile platforms.
2090 * Added a check for NaN of IEEE754 infinite in the double entries of
2092 * Introduce a new context factory "raw", highly inspirated from the
2093 ucontext factory, but using manually crafted functions in assembly to
2094 do the work in an efficient manner.
2095 * Allow to change the used context factory at run time, not only at
2096 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2097 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2098 contexts at runtime (only with raw contexts or ucontexts).
2099 * Completely rewrote this module to allow parallel execution of user
2100 processes. Use --cfg=contexts/nthreads:N to execute user processes
2101 with N parallel threads (the default is 1, meaning no parallelism).
2102 * Allow to decide dynamically between sequential and parallel modes.
2103 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2104 processes in parallel only when their number is greater than or equal to P
2106 * Added a check for NaN of IEEE754 infinite in the double entries of
2110 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2111 won't output the state of processes when interrupted with Ctrl-C
2112 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2113 NULL-terminated array. This may solve backward compatibility issues
2114 due to the change to return type of SD_simulate. See also:
2115 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2116 * Add new macros with variable number of arguments.
2117 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2118 - in xbt/asserts.h: xbt_assert
2119 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2120 - in xbt/ex.h: THROWF and RETHROWF.
2121 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2123 * Change xbt_die() to accept a format string with arguments, just like printf.
2124 * New data structure: xbt_lib_t, like a dict but more general and with better
2128 * New configuration options
2129 Options triva/categorized and triva/uncategorized can be used to generate
2130 graph configuration files for Triva visualization tool.
2131 * Configuration option tracing/platform is renamed to tracing/categorized
2132 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2133 * New TRACE_user_link_variable interface:
2134 User provides the name of the link and the tracing variable to attach to it
2135 * the declaration of tracing categories must be done after the environment
2137 * simpler tracing interface, just one way to declare categories
2138 TRACE_category or TRACE_category_with_color, it is up to you
2139 * links in the trace file are again identified by their names
2140 * trace contains the full platform hierarchy exactly as declared using the ASes
2141 * Options tracing/msg/[task|process]:1 groups the process by hosts
2142 for both cases, tasks and processes must have names that are unique during the simulation
2143 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2144 * The experimental option tracing/msg/volume is deprecated
2145 its functionality may be reincorporated if needed
2147 The tracing generates a trace file with unordered timestamped events,
2148 because of the way the core simulator (surf) works. A script available
2149 at the tools directory (fix-paje-trace.sh) can be used to put the events
2150 in order. We have changed the tracing so it can generate ordered timestamped
2151 events in the final trace, but depending on the simulator (and how much time
2152 is simulated) that can lead to a huge memory utilization. It is deactivated
2153 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2155 Build Infrastructure
2156 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2157 - We are trying hard to keep the API stable, but it may happen that
2158 some things change (we're a research project after all, not a
2159 nuclear plant operating system). If such things should happen, you
2160 could rely on that macro to adapt.
2161 - current value: 30600 for 3.06.00, aka 3.6
2162 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2163 a number that can be compared with SIMGRID_VERSION.
2164 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2165 assertions and verbose/debug logging events are disabled at compile time.
2167 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2169 SimGrid (3.5) stable; urgency=medium
2172 * New feature to any SimGrid-based simulator: Model-Checking
2173 Check SIN#1 for more details.
2176 * New Model SMPI (three-interval linear regression for correction factors)
2177 See RR-7426, available at http://hal.inria.fr/inria-00527150
2178 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2179 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2180 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2181 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2182 * See also src/smpi/README
2187 - Tracing API changes: TRACE_start and TRACE_end should not be called
2188 by user-code. They are automatically called by simulators created
2189 with SimDAG, MSG and SMPI if the toolkit is compiled with
2190 tracing_enabled=ON. Categories declaration and utilization remain the
2191 same for MSG and SimDag.
2192 - A function was added to the tracing API to declare categories with
2194 - TRACE_category_with_color (char *category, char *color)
2195 where color must be in the following format
2196 "%f %f %f", red, green, blue
2197 and red, green, blue are float values in the interval [0, 1]
2198 - User can specify NULL as color parameter, or continue calling
2199 TRACE_category (cat)
2200 On that case, the tracing system will define random colors
2201 - The following command-line options are supported:
2202 --cfg=tracing/filename:msg.trace
2203 --cfg=tracing:1 (activate tracing, needed to use others)
2204 --cfg=tracing/platform:1 (categorized resource use)
2205 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2206 --cfg=tracing/msg/task:1 (task creation)
2207 --cfg=tracing/msg/process:1 (process creation, migration)
2208 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2209 --cfg=tracing/smpi:1 (SMPI interface tracing)
2210 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2211 - examples of examples/msg/tracing updated
2213 - DAXLoader and DOTLoader functions can generate tasks with categories
2214 - A new function to attribute a category to SD tasks:
2215 TRACE_sd_set_task_category (SD_task_t task, char *category)
2216 * Tracing the MPI interface implemented by SMPI
2217 - Collective operations are traced with states
2218 - Point-to-Point operations are traced with states/links
2219 - Tracing activated by a parameter "-trace filename" passed
2220 to smpirun during execution (considering that simgrid
2221 is compiled with tracing enabled)
2222 - To run the simulation with gdb, the simulator
2223 accepts --cfg=tracing/smpi:1 to trace SMPI
2224 - tesh files to check if smpi tracing is ok
2225 - See examples/smpi/NAS/DT-trace
2226 * GTNetS tracing re-worked
2227 - adaptation to the tracing system of GTNets to cope
2228 with modifications regarding the fullduplex mode
2229 - new tesh files to check if gtnets tracing is ok
2232 * Asynchronous communications through the functions:
2233 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2234 * New function: MSG_load_platform_script()
2235 to make possible using a lua script instead of XML files to set up platforms
2236 * New function: MSG_set_function
2237 to associate functions to processes, used when bypassing the parser
2238 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2240 Platforms: Add some more examples in examples/platforms
2241 * Grid'5000: see www.grid5000.fr
2242 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2245 * Change the XML format. This is a very important modification. SimGrid
2246 3.5 introduces a new hierarchical format based on the notion of
2247 Autonomous Systems. Compatibility with old format is ensured
2248 through the perl script provided in the install bin directory
2249 bin/simgrid_update_xml.
2250 It is now possible to build platforms with specific routing
2251 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2252 connect several platforms together. We will try to provide soon
2253 set of realistic platforms exploiting these properties (have a
2254 look at examples/platforms/ for the moment).
2255 * Take the opportunity of the XML format change to be a good XML citizen:
2256 rename link:ctn to link_ctn and similar changes (also dealt with by
2258 * Add a new routing scheme (rule-based) using regular expressions. It
2259 enables to have an extremely low memory footprint when the
2260 underlying routing is simple and can be compactly described. You
2261 need to have libpcre4-dev (perl regular expressions) installed if
2262 you want to use this routing scheme.
2263 * Revive the cluster TAG and allow to easily and efficiently (both in
2264 term of memory and speed) connect clusters together. Have a look
2265 at teshsuite/simdag/platforms/ to see how this can be done. With
2266 this tag, you can create clusters with thousands of tasks at no
2267 cost (have a look at examples/platforms/).
2268 Note: clusters are implemented as ASes, so there is no need for an
2269 enclosing AS tag if you have only one cluster in your platform.
2270 * Add new generic functions in the public interface that allows the user
2271 to call SURF 'create_resource' methods from your code (same
2272 functionality as the XML bypass mechanism but with a much lighter
2274 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2275 that uses a piecewise linear approximation to produce better
2276 results when exchanging small messages.
2277 * Add a new parameter to handle correctly full duplex link and account
2278 for interferences between uplink and downlink communications
2279 (activate with --cfg=fullduplex:1).
2282 * Rename the SD_READY (all dependencies are satisfied and task is
2283 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2284 dependencies are satisfied) state.
2285 This prevents a confusion between the notion of "ready to schedule"
2286 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2287 simulated" (SD_RUNNABLE) used by the simulation kernel.
2288 * Change the way a task is considered as ready. Instead of removing
2289 dependencies when a task is done, a counter is decreased. This way,
2290 it is always possible to reach ancestors thanks to the
2291 SD_taks_get_parents function (even after the end of the simulation.)
2292 * Change the return type of SD_Simulate from (SD_task_t*) into
2293 xbt_dynar_t. This function was in handling a dynar internally and
2294 converted it into a NULL terminated array for historical reasons.
2295 * New function SD_dotload(char*) to load a DAG described in dot
2296 format. This loader and the corresponding examples require the
2297 installation of the graphviz library.
2298 * Fix a bug in the management of tasks of size 0 in the surf network
2299 models. This problem was only visible with SIMDAG and you should
2300 thus disregard results produced with earlier versions if you
2301 relied on this feature (some tasks were blocked because of this).
2302 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2303 even though your applications were doing only point-to-point
2304 communications and sequential computations. Now you can really use any
2305 model you want (of course, if you create real parallel tasks, which are
2306 not implemented in most models beside ptaskL07, this will abort).
2307 * Add an example that schedules a DAX on an heterogeneous platform
2308 using a Min-Min strategy.
2309 * New function SD_workstation_get_current_task() that returns the kind
2310 of task currently running on a workstation in the sequential access
2312 * Raise some warnings when unexecuted tasks remains at the end of the
2313 simulation. This is usually caused by cycles in the DAG.
2316 * New function: SIMIX_process_set_function() called by MSG_set_function
2317 * Change the underlying waiting queue in semaphores so that a process
2318 can wait on several of them simultaneously (as in waitany).
2319 * Fix the way to handle tokens in semaphores so that all access patterns
2320 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2321 * kill the dirty pimple SIMIX_message_sizes_output()
2322 Please use (proper) visualization instead
2325 * New data container: setset (set of sets of elements)
2326 * New module: mmalloc (mapped malloc, allowing to have several
2327 independent segments of malloc)
2328 * New function: xbt_dict_cursor_set_data()
2329 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2330 * New function: xbt_dynar_is_empty()
2331 * New function: xbt_fifo_get_last_item()
2332 * Fix xbt_dynar_shrink(): use the right element size.
2333 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2334 uninitialized areas during expand.
2335 * Fix semaphores: previous implementation was severely broken.
2336 * Use library init/fini functions for our initialization.
2337 - you can use logs and other feature as soon as you want in your
2338 code (even before the xbt_init / MSG_init)
2339 - xbt_exit is now a no-op and produce a warning when used.
2342 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2343 benefit from the latest improvement to the simulation kernel.
2344 * Kill measurement sockets for now. If you rely on them, sorry. This
2345 release is not for you. This feature will be reintroduced in the
2346 future, but we cannot delay the release any further.
2347 * New function: gras_msgtype_get_name().
2348 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2349 * Fix (at last) the pmm example: it should not randomly fail anymore.
2351 Build chain: bug fixes and overall polishing
2352 * Cmake is now stable enough. Hence, we killed the autotools.
2353 * Port to windows ( TM :)
2354 * Fix the 'make install' target.
2355 No need to use 'make install-simgrid' anymore
2356 * Introduce a 'make dist' target compiling a *source* archive
2357 'make package' compiles a binary archive
2358 * Compile java files only on need
2359 * Add --cd and --setenv command line options to tesh
2360 * Out of source builds are not fully supported yet, but we are close
2361 * Enable supernovae and optimization flags by default for our users
2364 * Add layer to set up environment directly from lua, without XML.
2365 * The effect of gras_stub_generator can be achieved through
2366 lua too (check examples/gras/console/ping_generator.lua)
2368 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2370 SimGrid (3.4.1) stable; urgency=low
2372 The "Polishing easter eggs is probably a good idea" release.
2373 This is a bug fixes release only.
2377 * Fix a bug preventing the tasks from begin garbage collected.
2380 * Fix a bug occurring when a host involved in a communication fails.
2381 This was not detected properly by the other peer involved in the
2382 communication. Now, it's reported as a network error.
2385 * Warn the user about loop dependencies in data flow of DAX files
2386 * Obey the control-flow dependencies of DAX files
2389 * Add option "enable_smpi" allowing to not compile SMPI.
2390 Probably useful for the (Mac) users experiencing a build error here
2391 * Improve the detection of lua5.1 and ruby1.8
2393 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2395 SimGrid (3.4) stable; urgency=low
2397 The "Easter in Cargese" release. Also known as (major changes):
2399 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2400 ~> bindings were greatly improved
2401 ~> new tracing infrastructure for better visualization introduced
2403 * the "Welcome to configury modernity" release.
2404 ~> we switched from autotools to cmake, and improved our cdash
2407 A more detailled list of changes follow (full detail in svn log).
2409 Java Bindings: Various Cleanups
2410 * (install java-gcj-compat-dev on debian-like to use them)
2411 * Remove put/get: no need to export deprecated interface in Java
2412 Use send/receive instead.
2413 * Cleanup the examples and add a README per directory
2414 * Remove example autoDestination (that's the only way to go now)
2415 * Remove example explicitDestination (was a plain copy of basic)
2416 * Make JniException a runtime exception, so that there is no need to
2417 declare the fact that you may encounter such a beast. I guess that
2418 nobody will ever want to survive such error.
2419 * Create specific errors for each MSG case of failure:
2420 host failure, transfer failure, timeout, task cancelled
2421 * Cleanup the exceptions that may get thrown by each function
2422 * Other internal cleanups in Java bindings. Performance still bad :/
2423 Ruby and Lua Bindings: create them
2424 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2425 * That's new and great, you should try them out.
2426 Same functionalities than Java bindings, only even less polished
2428 * Kill the useless "rate" argument of SD_task_get_execution_time()
2429 Everyone used to provide -1 as a value, it was not used, and the
2430 semantic of a possible use wasn't even clear.
2431 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2432 either as comm costs or compute costs to mean that there is no
2433 such thing for that specific task.
2434 * Add a SD_task_set_name() function
2435 * Fix SD_task_unschedule() on typed tasks
2436 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2437 * In DAX loader, accept useless 'level' attributes to <job> since
2438 LIGO DAGs have them (seem to be to ease graphical representation).
2440 * Add an example masterslave_mailbox.c using send/receive and not
2441 the deprecated put/get interface.
2442 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2443 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2445 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2446 (declare MSG_USE_DEPRECATED to still have the old name)
2447 * Add a MSG_task_set_data() function
2448 * About trace replay (see examples/msg/actions):
2450 - Allow to work with splitted trace files for each process
2451 Give the specific trace file as argument of each process,
2452 and call MSG_action_trace_run(NULL)
2453 You can still have one merged file for all processes.
2454 - Fix implementation of collective operations
2455 * Allow task_execute() on 0-sized tasks (closes #10063)
2457 * This is the first release of SimGrid where SMPI is not considered
2458 beta anymore (even if some corners should still be improved)
2459 * Port over the new SIMIX_network submodule (internal refactoring)
2460 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2461 * Implement more missing elements of the standard:
2463 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2465 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2466 MPI_Scan() MPI_Get_processor_name()
2467 - Added implementation of missing case for Alltoall (warning: it's
2468 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2469 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2470 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2471 SMPI_MPI_Allgatherv()
2472 * Bug fixes include:
2473 - MPI_Waitsome() was broken
2474 - Allow relative includes in smpicc
2475 - Command line cfg argument 'reference_speed' was ignored...
2476 - Some functions did not properly lead to auto-benching of user code
2477 - smpicc passes -O2 by default (just like openmpi one)
2479 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2480 * Bug fixes about timeouts during communications
2481 * add SIMIX_message_sizes_output() as a pimple to write to file the
2482 amount of messages per size. Use gnuplot to get histogram.
2483 Pimple because that's the only user-visible function of simix,
2484 defined directly in xbt.h (irk, sorry)
2486 - Add a SIMIX_sem_get_capacity() function
2487 - Fix interactions with processe resume/suspende
2488 - release_forever() was stupidly broken
2489 - Fix SIMIX_display_process_status() for processes in a semaphore
2490 - Make SIMIX_sem_block_onto() user-visible
2491 * Refactoring context stuff:
2492 - Use pseudo-OOP for better modularity
2493 - reimplement SIMIX_process_kill() without process_schedule() so
2494 that the latter can take as invariant that it is called from
2496 - Merge context_start into context_new for sake of simplicity
2498 * Add a Vivaldi network model, coded live during SUD'10 ;)
2499 * Rename configuration variables to start a hierarchy:
2500 o cpu_model -> cpu/model
2501 o network_model -> network/model
2502 o workstation_model -> workstation/model
2503 * New configuration variables:
2504 o network/bandwidth_factor: correction to bandwidth
2505 o network/latency_factor: correction to latency
2506 o netwotk/weight_S: correction to the weight of competing streams
2507 * Add a long description to the models, that users can see with such
2508 argument on the command line: --cfg=cpu/model:help
2509 * --help-models display the long description of all known models
2511 * config: add the ability to set a default value after registration
2512 Does not override any previously set value (e.g. from cmd line)
2513 * dict: allow to have integer key and data.
2514 When so, you need to use the following functions
2515 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2516 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2517 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2518 In contrary to regular dicts, the key is not malloced before copy.
2519 Mixing scalar and regular elements in the same dict is not tested
2521 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2522 Tracing for Visualization:
2523 * SimGrid is now instrumented in order to generate a trace file for
2524 visualization analysis: to use it, need to compile SimGrid with the
2525 "tracing" option enabled, and instrument the program using SimGrid with
2526 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2527 (among other functions).
2528 * The instrumentation only traces the platform utilization for now
2529 * Documentation to use the tracing functions and how to analyze the
2530 traces with the Triva tool is written.
2531 * More information about: SimGrid FAQ (in the section Tracing Simulations
2534 * We moved to cmake as default build system. Autotools support will
2535 be dropped soon. Check the FAQ for more info about how to use it.
2536 * Greatly improved our cdash/ctest interactions
2537 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2538 * Added memory checking tests with valgrind; lot of memleak fixing.
2539 This may be the first release of simgrid with so few memory issues
2540 * Added code coverage tests.
2541 Our coverage is still improvable, but at least we see it on cdash.
2543 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2545 SimGrid (3.3.4) stable; urgency=low
2547 The "Desktop Grid needs love too" release (also called Xmas release).
2549 Models improvements:
2550 * Major speedup in the maxmin system solving by using lazy evaluation
2551 Instead of solving completely the maxmin system at each iteration,
2552 only invalidate (and recompute) the modified parts.
2553 This new feature is enabled in default models but you can try to
2554 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2555 * Cas01 IMproved as default CPU model
2556 This CPU model is the same Cas01 model, but it uses the
2557 maxmin-selective-update flag and a heap structure to manage
2558 actions on SURF kernel.
2559 It reduces the complexity to find the next action to finish and,
2560 consequently, it's faster than the old Cas01.
2561 This is the new default CPU model (Cas01).
2562 * Rename the old Cas01 model to Cas01_fullupdate
2563 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2564 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2565 * CpuTI (CPU Trace Integration)
2566 A new CPU model whose objective is simulate faster when using
2567 availability trace files.
2568 Instead of using a full featured, over engineered maxmin system for
2569 CPU modeling, this model does the pre-integration of traces files
2570 to calculate the amount of CPU power available, and so, executes
2571 faster than the old CPU models.
2572 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2573 * Use LV08 as default network model since it gives better accuracy
2574 for small messages and shouldn't change things for big ones.
2575 Use --cfg=network_model:CM02 to get the previous behavior.
2578 ******************************************
2579 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2580 ******************************************
2581 * The new CPU model may changes simulations!
2582 The point is that events occurring at the exact same timestamp
2583 are not scheduled in the same order with the old and new
2584 version. This may be enough to completely change the execution
2585 of simulations in some cases.
2586 * The new network model will change simulations!
2587 This new model is more realistic than the previous one, so you
2588 should consider redoing your old experiments with this model.
2589 Sorry for the inconvenience.
2592 * Introduce the supernovae compilation mode
2593 When compiled that way, the whole SimGrid (or almost) is put in a
2594 single compilation unit and compiled in one shoot.
2595 This is to help gcc which has difficulties to inline stuff from one
2597 The speedup seem to be above 15%, althrough more tests are needed on
2598 amd64 to confirm that gain.
2601 * Port of MSG's mailbox on top of SIMIX network
2602 The put/get mechanism was greatly simplified on the way.
2605 * New SIMIX network module. Provides:
2606 - Mailbox: rendez-vous mechanism to find with who you want to speak
2607 - Synchronous send/recv: easier and hopefully faster since the
2608 logic is handled in the maestro process directly now
2609 - Asynchronous send/recv: you dreamt of it? It's here now
2610 Too bad that nobody cared enough to propagate the change to MSG.
2611 * Add semaphores as SIMIX synchronization mechanism.
2614 * new function SD_daxload(char*) to load a DAX file
2615 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2616 * Introduce typed tasks. Specify its kind and cost at creation.
2617 At scheduling, just give where it should be placed, and the cost
2618 for each involved resource is automatically computed.
2619 Existing constructors so far (more to come of course):
2620 - SD_task_create_comm_e2e() for end-to-end communication
2621 - SD_task_create_comp_seq() for sequential computation
2622 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2623 * new function SD_task_dump() for debugging display
2624 * new function SD_task_dotty(task,FILE*) writing to file the info
2625 about the task in dotty format
2626 * SD_task_dependency_exists() can now cope with having one of its
2627 arguments NULL. If so, it tests whether the other argument has any
2629 * Add getters on list of preceding/following tasks:
2630 SD_task_get_parents(task) and SD_task_get_children(task)
2631 * Add getters on amount of workstations and list:
2632 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2633 * Add getter on task kind: SD_task_get_kind(task)
2634 * Update the start_time and finish_time of tasks on completion/failure
2635 * Bugfix: Remove task from state swags when destroyed
2638 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2641 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2642 dynar with one separate thread per value of the dynar.
2643 * Change the prototype of xbt_thread_create(), sorry.
2644 Added a boolean parameter indicating whether we want to join this
2645 thread (used in SG only for now)
2646 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2649 * GTNetS wrappers should now be usable again (and betterly tested too)
2650 * Fix a major regression from 3.2 where the timeout provided to
2651 MSG_task_put_with_timeout() was used as absolute time before which
2652 the comm should be done.
2653 * Start to fix the <cluster> tag.
2654 - Internal links should be good now (beside of the loopback, which
2655 use the private link instead)
2656 - paths to the external world is still rather broken
2657 - the <route:multi> tag is just broken. Actually that's brain-dead.
2658 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2659 to make it less stupid
2660 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2661 * Fix a source-level compatibility glitch from 3.2: after defining
2662 MSG_USE_DEPRECATED, you can use the old name
2663 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2664 * Allow to compile from the SVN with automake 1.11
2665 * Fix some problems when using the "start_time" tag in deployment XMLs.
2666 * Fix #8569: XBT/synchro.h has redundant declarations
2667 * Fix #8563: MSG return values and exceptions
2668 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2669 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2670 Thanks to Robson Peixoto.
2671 * Fix a vicious bug in dictionaries inducing that some elements were
2672 not freed on xbt_dict_free()
2674 Portability report of this version:
2675 * Main portability targets:
2676 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2677 on (amd64/i386/ia64)
2678 - mac leopard on i386
2679 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2680 but nothing critical.
2681 * Other platforms: windows, AIX and others were not tested for this release
2683 Timing report of this version:
2684 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2685 scenario parameters). From 8h to a few seconds in desktop grid settings.
2686 * Supernovae brings about 25% speedup on i386.
2688 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2690 SimGrid (3.3.3) stable; urgency=low
2692 The "Need for Speed" release.
2694 The timings done to validate the 3.3.2 were faulty.
2695 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2697 The problem was a conversion from a manually handled vector to
2698 xbt_dynar_t on the critical path.
2699 xbt_dynar_foreach calls functions, inducing stack management crap.
2701 We inlined these functions and xbt_dynar_foreach is now breath taking.
2702 We also inlined xbt_swag_belong on the way.
2704 Here are some approximate speedup measurements (on master/slaves
2705 simulations lasting between 10s and 20s each):
2706 3.3.1 -> 3.3.2: about same performance
2707 3.3.2 -> 3.3.3: 40% speedup
2708 3.3.1 -> 3.3.3: 40% speedup
2709 3.3.1 with inline patch -> 3.3.3: 30% speedup
2711 Our reading is that the refactoring which occurred in 3.3.2 made us
2712 suffer much more from the xbt_dynar_foreach low performance, but
2713 once we solved this, this refactoring proved to be very performance
2714 effective. From the 40% speedup, somehow, 10% are due to the
2715 inlining and 30% to the refactoring.
2717 That's a pitty that gcc cannot inline functions placed in other files
2718 alone. We have to choose between:
2719 - break the encapsulation (by putting private data structures and
2720 accessors in headers files to help gcc)
2721 - live with low performance
2722 - switch to a decent compiler such as icc (not quite possible).
2724 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2726 SimGrid (3.3.2) stable; urgency=low
2728 The "Simplicity does not preceed complexity, but follows it" release.
2730 The main contributors of this release were (lexical order):
2731 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2734 * Extract the routing logic into its own object.
2735 (was dupplicated in network.c and workstation_LV07.c;
2736 Allows to implement other ways of storing that info)
2737 => kill now useless network_card concept
2738 - Use dynar to represent routes (instead of void** + int*)
2739 - kill link_set (use surf_network_model->resource_set instead)
2740 - Add a command-line option to choose the routing schema to use
2741 - Add three new models:
2742 * Floyd (shortest path computed at initialization)
2743 * Dijikstra (shortest path recomputed all the time)
2744 * Cached Dijikstra (shortest path computed on need)
2745 All these models where contributed by Silas De Munck, and are
2746 described in his ICCS09 paper.
2748 * Simplify model declaration
2749 (less redirections, less function to write when defining a model)
2750 - Factorize stuff between models:
2753 surf_model_resource_set(model)
2754 surf_model_resource_by_name(model, name)
2755 - Unify the types of models in s_surf_model_t (using an union)
2756 - Embeed fields of common_public directly into s_surf_model_t
2757 - Rename model methods:
2758 action_free ~> action_unref
2759 action_change_state ~> action_state_set
2760 action_get_state ~> action_state_get
2761 - Change model methods into functions:
2762 (model)->common_public->action_use ~> surf_action_ref
2764 * Implement a generic resource; use it as ancestor to specific ones
2765 (allows to kill duplicated code in models)
2766 Drawback: timer command don't need no name nor properties;
2767 workstation_CLM03 don't need no properties
2768 (but I guess we can live with those few bytes wasted)
2770 * Improve the action object model
2771 - implement a constructor avoiding dupplicated code about field
2772 initialization in generic_action part.
2774 * Kill the SDP model: it has an external dependency, is deprecated
2775 in flavor of modern lmm models, and didn't compile since a while
2778 * Relocation of the context module from XBT to SIMIX.
2779 (the context were decoupled from the simix processes, duplicating a lot of code)
2780 => a lot of code was factorized
2781 - less overhead is introduced during scheduling
2782 - simpler API for the context factory
2783 - the logic for process creation,destruction and manipulation was simplified
2784 * Simplification of the s_smx_process_t data structure.
2785 => accesing the simix level data associated to a process is faster now,
2786 and the code is a lot more readable.
2789 * Implement some more MPI primitives:
2790 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2791 -implementation: Bcast: flat or 2-ary tree (default),
2792 Barrier: 4-ary tree,
2794 Allreduce: Reduce then Bcast
2795 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2796 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2797 Alltoallv: flat tree, like ompi
2799 * Add support for optimized collectives (Bcast is now binomial by default)
2800 * Port smpirun and smpicc to OS X
2803 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2804 Shout out if you used it.
2807 * Display the list of still queued messages in SG mode when existing
2811 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2812 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2813 * Add xbt_str_from_file(FILE*)
2814 * Add xbt_dict_get_key achieving a linear reverse search
2815 * Remove the context module
2817 Portability report of this version:
2818 * Main portability targets:
2819 - Linux(debian)/x86/context
2820 - Linux(debian)/x86/pthreads
2821 - Linux(debian)/amd64/context
2822 - Linux(debian)/amd64/pthreads
2823 On these, we still have the eratic breakages of gras/pmm and
2824 amok/saturate_sg reported in previous version. We still think
2825 that the tests are the cause of the fault, not the tested code.
2827 - Mac OSX Leopard/x86/context
2828 Still false negative in tesh autotesting.
2829 Smpi still fails, but this time because readlink does not accept -f
2830 Everything seems to work properly beside of that.
2833 - AIX version 5.3 (only tested contexts this time)
2834 Smpi still fails there because mktemp is not installed.
2835 Everything seems to work properly beside of that.
2837 I managed to compile it for the first time, but several breakages.
2838 Won't delay the release for this exotic platform.
2840 * Windows: it's still lagging behind. If you want to help, please
2843 Timing report of this version:
2844 This version seem to be more than 5% faster than 3.3.1 (on linux
2845 64bits with contextes). The gain is less than expected, we are
2846 investigating this for next release.
2848 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2850 SimGrid (3.3.1) stable; urgency=low
2853 * Implement a --cfg-help to show existing configuration variables
2854 * Build chain do not require doxygen in maintainer mode
2857 * fix a bug on struct sizeof computation, which prevented the
2858 exchange of arrays of structs in some conditions
2859 - added a regression test about this in datadesc_usage
2860 * Allow the exchange of 0-long dynamic vectors.
2861 - for that, use -1 as indicator of dynamic size instead of 0
2862 - This implied to change any size from unsigned long to long,
2863 reducing a bit communication abilities, but I guess that with
2864 64bits being quite common, this is more than enough.
2865 - This also induce a protocol change, thus bumping network protocol
2866 version from 0 to 1 (if we have external users, we have to get
2867 clean on that point too ;)
2868 - added two regression tests about this in datadesc_usage
2869 * Be more verbose when propagating local exceptions
2870 This helps debugging.
2871 * Display the status of simulated processes when receiving SIGINT in
2875 * Allow to control the simulation from a trace file.
2876 New functions MSG_action_register() and MSG_action_trace_run()
2877 The first one allows to associate a function execution to each
2878 kind of action while the second one parses a trace file and
2879 triggers the corresponding actions within the system.
2880 For now, only a toy example is provided in examples/msg/actions
2881 * Add an example of process migration in examples/msg/migration
2882 * Fix a bug in task exchange which broke MSG_task_get_sender()
2883 Add a teshsuite regression test for that.
2884 [Bug: if MSG_task_get_sender() is called after sender exit,
2886 * Fix a bug which prevented suspend/resume to work properly
2887 * Display the status of simulated processes when receiving SIGINT
2888 This fixes a regression of v3.3. due to the introduction of SIMIX
2889 * Bug fixing in failure management:
2890 - trace could not start by a failure at time 0
2891 - failure during communications were not working
2894 * Add SIMIX_process_set_name() to change the name of the current
2895 process in the log messages.
2896 * Store smx_hosts in a dict since we only retrieve them by name
2897 * Move the configuration infrastructure to surf
2900 * Move the configuration infrastructure to surf
2903 * Massive internal cleanups:
2904 - Store internal structures on processes instead of hosts (allows
2905 to have more than one process per host, in addition of being more
2907 - Cleanup the initialization/finalization process
2908 - Kill a whole bunch of unneeded synchronization:
2909 processes run in exclusive manner within the simulator
2910 - Move queues from global tables to process data fields
2912 - now accept -platform and -hostfile arguments
2913 - Pass the right rank value to processes according to the hostfile
2914 * Compile the examples by default, and use them as regression tests
2915 * Implement MPI_Wtime()
2916 * Change the reference speed to a command line option
2919 * TCP_gamma can now be specified as command line option using
2920 --cfg=TCP_gamma:10000000.0
2921 * Change the --surf-path cmd line option into --cfg=path:
2924 * Also include strbuff from xbt.h public header
2925 * xbt_ex_display(): do not free the exception after displaying
2926 This allows to do more with the given exception afterward.
2927 Users should call xbt_ex_free() themselves.
2931 Portability report of this version:
2932 * Main portability targets:
2933 - Linux(debian)/x86/context
2934 - Linux(debian)/x86/pthreads
2935 - Linux(debian)/amd64/context
2936 - Linux(debian)/amd64/pthreads
2937 These targets fail about 1/10 of times on gras/pmm, but we believe
2938 that this is because of the test, not because of simgrid.
2939 amok/saturate_sg fails even more rarely, and the test may not be
2942 - Mac OSX Leopard/x86/context
2943 The test suite still spits tons of errors because some obscure
2944 force prevents us from removing the temporary directories
2945 arguing that they still contain some metadata I've never heard of.
2946 Smpi fails because seq is not installed.
2947 Everything seems to work properly beside of that.
2950 - AIX version 5.3 (both contexts and pthread)
2951 Smpi still fails there because mktemp is not installed.
2952 XML inclusions seems rosty on AIX.
2954 * Windows: it's still lagging behind. If you want to help, please
2957 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2959 SimGrid (3.3) stable; urgency=high
2963 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2964 [Malek Cherier & Mt]
2966 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2969 Until now, GRAS were using MSG as an interface to SURF. It was
2970 quite difficult because both interface have several differences
2971 (MSG channels vs GRAS sockets were the most notable point).
2973 This also opens the gate to SMPI (which should occur soon) and speed
2974 up simulations by to 40% (even if it were not the main goal).
2976 **************************************
2977 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2978 **************************************
2979 The point is that events occurring at the exact same timestamp are
2980 not scheduled in the same order with the old and new version. This
2981 may be enough to completely change the execution of simulations in
2982 some cases. Sorry for the inconvenience.
2984 * Cleanup and upgrade the XML format to push further scalability
2985 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2987 * Improve the testing infrastructure with tesh. Now a very large part of
2988 the code is tested not only by being run but also by checking that the
2989 output match an expected output [Mt].
2991 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2992 is really less memory-demanding, which should allow you to use
2993 larger files in SimGrid [AL].
2995 * Inform valgrind about our contextes, so that it becomes usable
2996 with the default (and more effecient) version of SimGrid
2997 [contributed by Sékou Diakite, many thanks]
3000 * Introduce a listener thread in charge of receiving incoming
3001 messages from the network. It allows to overlap communication and
3002 computation but most notably, it removes some stupid deadlocks due
3003 to the fact that so far, a process could not send and receive at
3004 the same time. This made most non trivial communication schema
3006 * Convert the PIDs from long int to int to match the MSG ones (and
3007 linux ones too) [Mt]
3008 * New function: gras_agent_spawn() to launch a new process on
3009 current host. Only working in simulation for now. [Mt]
3010 * New function: gras_os_hostport() returning a constant form (ie,
3011 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3014 * Make the backtrace of exceptions more human readable [Mt]
3015 * New module: xbt/str [Mt]
3016 a ton of string utility functions (split, join, printf to a newly
3017 allocated buffer, trim, etc)
3018 * New module: xbt/hash [Mt]
3019 SHA1 hashing algorithm (more to come if needed)
3020 * New module: xbt/synchro [Mt]
3021 synchronization tools (mutex and conditions) working the same way
3022 in simulation and in real life (mainly useful for GRAS, but not
3024 * New module: xbt/queue [Mt]
3025 classical producer/consumer synchronization scheme
3026 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3027 (using the classical functions will get serialized) [Mt]
3028 * Make dictionary internal table dynamic. No need to specify its size
3029 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3031 * Make sure the log channels are organized as a tree under windows
3032 (because of ANSI C compatibility issue, any channel were child of
3033 root directly) [Mt].
3036 * Cleaned many thing in surf and fixed a few bugs [AL].
3037 * Add a nice command line configuration mechanism to compose models [AL].
3038 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3039 the previous one (KCCFLN05). It relies on something that looks like
3040 a max-min sharing mechanism but cannot be written as such. A new solver
3041 was thus designed [AL].
3042 * Add a new solver to lmm. Based on Lagrange optimization and
3043 gradient-based descent, it enables to efficiently maximise systems s.a
3045 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3047 This solver enables to propose two new network models for TCP Reno and
3048 TCP Vegas based on Low's work. These models still need to be fully
3049 tested though [Pedro Velho].
3052 * Bug fix in SD_simulate. Now the time bound given as argument is
3054 * Use the new parallel task model (ptask_L07) as default.
3055 * Use the SURF command line configuration mechanism.
3056 * 0-size tasks (for synchronization) should now work.
3058 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3060 SimGrid (3.2) stable; urgency=high
3064 We still experience issues on this platform, but we believe that at
3065 least MSG is usable.
3067 GRAS API BREAKAGE (for simplification purpose, sorry):
3068 * the gras_msgtype_by_name is not used anymore. Instead of
3069 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3070 you can write (and must)
3071 gras_msg_send(toserver, "request", &request);
3072 - If you still want to pass a gras_msgtype_t to the function (to cache
3073 the type and avoid the lookup time), use the gras_msg_send_() variant.
3074 - Impacted functions:
3075 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3076 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3077 * The callbacks are now expected to return 0 when everything went well
3078 (just like the main() function)
3080 GRAS new features and improvements:
3081 * New module mechanism where user code can use per process globals [Mt]
3082 This is similar to gras_userdata_*() functions, but for libraries. It
3083 factorize some code developped over and over in the examples and AMOK.
3084 It has still to be documented and used (only amok/peermanagement is
3086 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3087 loss when they were small enough to fit into the buffer and sent quickly
3088 enough so that they can all get received in one shoot.
3089 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3090 if not found. Use the *_or_null() variant for the old semantic.
3091 * In gras_msg_handle, do not discard messages without callback.
3092 They are probably messages to be explicitly awaited later (ie, proofs of
3093 mis-synchronization in userland since they are sent before being awaited)
3094 No big deal usually.
3095 * gras_socket_meas_send/recv: semantic changed!
3096 The numerical arguments used to be (1) the total amount of data to send
3097 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3098 messages. This was need for the fool willing to send more than MAXINT
3099 bytes on quite fat pipes.
3102 * Do really rename the hostmanagement module to peermanagement. [Mt]
3103 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3104 but this is rather new and this was documented in the module
3105 documentation (poor excuses, I admit)
3106 * Bandwidth measurement semantic changed! This follows the changes to
3107 gras_socket_meas_send/recv explained above.
3110 * A sequential mode has been added to the workstations. When a workstation
3111 is in sequential mode, it can execute only one task, and the other tasks
3112 are waiting in a FIFO. [Christophe Thiery]
3115 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3116 model for SIMDAG. [Christophe Thiery]
3117 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3118 equal to 0, which caused some bad behaviors in
3119 saturated_constraint_set_update. I now use a threshold mechanism like in
3123 * When running manually src/testall, you select specific units [Mt]
3124 testall is the result of our cunit mechanism, and should replace all
3125 the scripty thingy around since bash don't run easily on billware.
3127 * A mallocator system has been added. [Christophe Thiery]
3128 Mallocators allow you to recycle your unused objects instead of freeing them
3129 and allocating new ones.
3131 Documentation update:
3132 * FAQ reworking + New FAQs:
3133 - "Valgrind spits tons of errors!" [Mt]
3134 - "How to repport bugs" [Mt]
3135 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3136 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3137 - Communication time measurement within MSG [AL]
3138 - I experience weird communication times when I change the latency [AL]
3139 * GRAS tutorial [Mt]
3141 - an introduction to the framework and to the used communication model
3142 - an initiatic tour introducing the most proheminent features:
3144 . Lesson 0: Installing GRAS
3145 . Lesson 1: Setting up your own project
3146 o Part 2: Message passing
3147 . Lesson 2: Exchanging simple messages
3148 . Lesson 3: Passing arguments to the processes (in SG)
3149 . Lesson 4: Attaching callbacks to messages
3150 . Lesson 5: Using globals in processes
3151 . Lesson 6: Logging information properly
3152 . Lesson 7: Using internal timers
3153 . Lesson 8: Handling errors through exceptions
3154 . Lesson 9: Exchanging simple data
3155 . Lesson 10: Remote Procedure Calling (RPC)
3156 . Lesson 11: Explicitly waiting for messages
3157 . Recapping of message passing features in GRAS
3158 - A HOWTO section containing:
3159 o HOWTO design a GRAS application
3160 More are due, of course. They will come latter. In the meanwhile, you can
3161 check the examples which are still here.
3163 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3165 SimGrid (3.1) stable; urgency=high
3169 There was a stack corruption somewhere, visible only when optimizing
3170 with these versions. [Vince]
3173 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3174 module that enabled to play with Directed Acyclic Graphs. It is built
3175 directly on top of SURF and provides an API rather close to the old
3176 SG. Some old codes using SG are currently under rewrite to check that
3177 all needful functions are provided. [Christophe Thiery]
3180 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3181 extension of the classical CLM03 model that gracefully handles
3182 failures. This is now the default model for MSG and GRAS. It doesn't
3183 handle parallel tasks yet though. [AL]
3184 * Bug fix: Weights were not correctly set in the network part.
3185 WARNING: This may have resulted in incorrect results with simulations
3186 where there are more than one flow on a given link. [AL]
3189 * After a (long ?) discussion on simgrid-devel, we have decided that the
3190 convention we had on units was stupid. That is why it has been decided
3191 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3192 WARNING: This means that all previous platform files will not work as
3193 such with this version! A warning is issued to ask users to update
3195 A conversion script can be found in the contrib module of the CVS, under
3196 the name contrib/platform_generation/surfxml_update.pl [MQ]
3199 * Bug fix: Processes were started in reverse order, wrt deployment file.
3200 WARNING: if your code relies on this bug, please fix it. [AL]
3201 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3202 being executed on two different locations. [AL]
3203 * Bug fix: Failures are now better supported thanks to Derrick's tests
3204 (there was many failure situations I hadn't thought of and that weren't
3205 correctly handled). [AL]
3206 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3210 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3211 Exception occurring on server-side are propagated back to client (!).
3213 API CHANGE: the callback changed their prototype. Change:
3214 int my_handler(gras_socket_t expeditor, void *payload_data) {
3216 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3217 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3219 * New! function: gras_msg_handleall to deal with all messages arriving
3220 within a given period.
3221 * New! function: gras_socket_server_range to get a server socket in a
3222 range of port numbers (ease to avoid port number conflicts) [MQ]
3223 * New! gras processes display their backtrace when they get a SIGUSR1
3224 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3225 Sweet to debug RL processes [MQ]
3229 - Do not force experiment sizes to be expressed in kb, or it becomes
3230 impossible to measure the latency this way (needs one byte-long tests)
3231 WARNING: this changes the amok_bw_* function semantic. [MQ]
3232 - Implements the link saturation stuff. [MQ]
3233 * Peer management module:
3234 New! module factorizing code that we wrote over and over [MQ].
3237 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3238 - Test units are placed directly into the library code, they get extracted
3239 automatically and placed into the src/testall binary.
3240 - Convert most of the XBT tests to this system.
3241 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3242 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3243 live in the same C code. This required to change a little bit the API
3244 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3245 * New module: sparse graph structure with basic algorithms (this is work
3246 in progress and the API is not considered to be frozen yet). [AL]
3247 * Display more information on backtraces: source line & function names are
3248 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3249 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3250 * Reworked a little bit some #include statements to load only required
3251 headers. Some user code that relied on SimGrid to include stdlib or
3252 stdio may need to include it by themselves. [AL]
3253 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3255 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3256 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3257 but I guess that almost nobody used those parts. [MQ]
3259 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3261 SimGrid (3.0.1) stable; urgency=low
3264 * Unfortunately, I had missed 5 misnamed functions:
3265 xbt_fifo_item_t xbt_fifo_newitem(void);
3266 void xbt_fifo_freeitem(xbt_fifo_item_t);
3267 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3268 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3269 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3270 They're now deprecated. Please use their new versions:
3271 xbt_fifo_item_t xbt_fifo_new_item(void);
3272 void xbt_fifo_free_item(xbt_fifo_item_t);
3273 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3274 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3275 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3277 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3278 * Documentation: xbt_log module unmercifully reworked [MQ]
3279 * Bugfix: there was a problem with the ending of contexts with
3280 the pthread backend. It caused some weird deadlock or behavior
3281 depending on the pthread implementation. [AL]
3282 * Bugfix: get the exceptions raised in the simulator repport where
3283 and why they come from when they are not caught in time [AL, MQ]
3286 * Bugfix: Do repport the error when two non-connected hosts try to
3287 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3290 * Add additionnal checkings on communications. Assert that two
3291 communicating hosts are connected by a set of links... [AL]
3294 * Add additionnal checkings on channel values in communication [AL]
3295 * New: MSG_task_get_source to see on which host a task was generated [HC]
3296 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3297 the number of tasks waiting to be received on channel and sent
3299 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3300 waits for the first task coming from a given host.. [AL]
3302 GRAS new functionnalities: [MQ]
3303 * Enhance the parsing macro to allow the size of multidimentional objects
3304 to be given thru annotations.
3305 * New example (and documentation): Matrix Multiplication a la RPC
3306 (as when I was young!) and fix a bunch of bugs found on the way.
3308 GRAS performance improvements: [MQ]
3310 * Reduce the amount of cbps creation/destruction by making it static to
3311 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3314 * Change libdata to a set so that we can search for stuff by ID (and thus
3315 reduce the insane amount of dict lookups)
3318 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3319 RL communication on top of SG since it's so uneffective.
3320 It may also be used for inter-thread communication in RL, one day.
3321 * Use gras_datadesc_copy() to exchange messages on top of SG
3323 - improve message exchange performance on top of SG
3324 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3325 * Don't exchange on the network the size of the used part of buffer,
3326 instead, specify the possible buffer size to read().
3328 - reduces the amount of read/write calls (one pair per exchange)
3329 - reduces the amount of exchanged data (the size)
3330 - allows to retrieve all arrived data on receiver side, if we don't need
3331 it right now (subsequent read will peek the buffer)
3332 - allows the receiver to proceed with the beginning of the stream before
3333 everything is arrived
3334 - make it possible to build an iov transport (using readv/writev)
3336 - take care of the data with non-stable storage (like stacked data),
3338 * If possible, TCP send uses vector I/O (when writev() is here)
3339 - Don't use it for receive since we send data sizes and data on the
3340 same stream, so we wouldn't be able to chain large amount of chunks
3341 before having to flush the stuff to read the size.
3342 * Rework the transport plugin mechanism to simplify it and reduce the
3343 amount of pointer dereferencement when searching for the right function
3346 * I guess that now, we do almost as few system calls as possible while
3347 doing as few data copy as possible.
3349 To improve it further, we could try to send all the sizes first and then
3350 all the data (to use iov on receiving size), but it's only a partial
3351 solution: when you have 2 dimensional data, the sizes of the second
3352 dimension is data of the first dimension, so you need 3 streams.
3354 I'm not sure the potential performance gains justify the coding burden.
3356 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3358 SimGrid (3.00) stable; urgency=high
3361 * New! Give the possibility to hijack the surf parser and thus bypass
3362 MSG_create_environment and MSG_launch_application. Have a look at
3363 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3365 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3367 SimGrid (2.96) unstable; urgency=low
3372 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3373 This deprecates the xbt_error_t mechanisms.
3374 It modifies (simplifies) all XBT and GRAS API.
3375 MSG API keeps unchanged (exceptions raised by XBT are caught from
3376 within MSG and masked with existing error handling facilities)
3379 * New! Add a FATPIPE model. [AL]
3380 * New! Add a parallel task model. [AL]
3381 * New! Add automatically a loopback interface (in the default
3382 network model) if none was precised.
3385 * Bugfix: MSG_process_resume now works with the current running process.
3387 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3388 * Modification of MSG_task_get_compute_duration. Once a task has been
3389 processed, the value returned by this function is now equal to 0. [AL]
3390 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3391 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3392 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3393 of a m_task. For now, MSG_task_cancel only works with computation
3395 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3396 of the processor (in Mflop/s) regardless of the current load on the
3398 * API Change: use proper naming convention for MSG_getClock and
3399 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3401 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3402 This function changes the priority of a computation task. This priority
3403 doesn't affect the transfer rate. A priority of 2 will make a task
3404 receive two times more cpu power than the other ones. This function
3405 has been added to suit the needs of Nguyen The Loc and hasn't been that
3406 much tested yet. So if it fails, please report it and send me your code.
3408 * API Change: removed all functions and types that were marked "deprecated"
3409 since many months. Renamed MSG_global_init_args to MSG_global_init.
3411 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3413 SimGrid (2.95) unstable; urgency=low
3416 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3417 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3418 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3421 * Bugfix: Allow absolute paths for platform description files [MQ]
3422 * Bugfix: do free the variables after use. Leads to drastic performance
3424 * Implement max_duration (ie, timeouts) on resources [AL]
3427 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3429 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3430 the needed support (timeouts) [AL]
3433 * Implement measurement sockets. You can now get the bandwidth between two
3434 hosts thanks to AMOK (see below). [MQ]
3435 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3436 dynar over the network (yeah) [MQ]
3437 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3438 * simple/static token-ring example. [Alexandre Colucci and MQ]
3439 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3440 in place before [MQ]
3443 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3444 type parsing. "short" and "long" modifiers were also ignored; other
3445 modifier (such as reference level) are still ignored. [MQ]
3446 * Embeed the buffer size within the buffer itself on SG. [MQ]
3447 That way, send() are atomic and cannot get intermixed anymore (at least
3448 the ones which are less than 100k; bigger messages still have the issue)
3449 * Array size pushed by the field, not by the field type (or each
3450 and every long int will push stuff to the cbps) [MQ]
3451 * use select() to sleep since it allows to portably sleep less than one
3454 GRAS (minor cleanups)
3455 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3458 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3459 * type category 'ignored' killed as it was never used and were difficult
3461 * whether a type can cycle or not is now a flag, leaving room for more
3462 flags (as "ignored", if we feel the need one day ;)
3463 * Rename raw sockets to measurement sockets since "raw" has another
3464 meaning in networking community.
3467 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3468 and offers features not belonging to GRAS but that most applications
3469 need. One day, it may be a set of plugins loadable at runtime.
3470 * New module: bandwidth
3471 bandwidth measurement between arbitrary nodes running this module. [MQ]
3473 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3475 SimGrid (2.94) unstable; urgency=low
3477 The first beta release of SimGrid 3 !
3481 * Update the main page and the FAQ. Adding references to gforge.
3484 * Add a gras_os_getpid function.
3487 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3488 * Extend the logs so that they also print PID, hostname, date, ... if
3490 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3491 and kill the old version which were in testsuite/
3492 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3496 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3497 for users. Instead, each programming environment (GRAS, MSG,...) use it
3498 when needed to provide such a feature to users.
3499 Don't shortcut the mechanism or you will also shortcut the virtualization
3500 you need on the simulator.
3504 * Cleanups in configury with regard to compile optimization/warning flags.
3505 Also add -fno-loop-optimize to any powerpc since it's the optimization
3506 killing gcc (< 3.4.0).
3507 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3508 needed by MSG examples complications
3509 * Borrow configury beautifications from PHP
3512 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3513 warning (thanks loris for stumbling into this one).
3514 * Bugfix: stop loading private headers (gras_config.h) from the public
3518 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3519 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3520 RL would imply a DNS resolver, which is *hard* to do in a portable way
3521 (and therefore delayed).
3522 * Implement a real timer mechanism and use it in timing macros. This allows
3523 to avoid rounding errors and get a 0.000005 sec precision in timing
3524 macros. While I was at it, various cleanups:
3525 - allow to declare more than one timed section per file (fix a stupid bug)
3526 - move some private declaration to the right place
3527 - merge conditional execution and timing macros into emulation module
3528 - document the module
3529 - make sure the module cleanups its mess on gras_exit
3530 * Documentation improvements:
3531 - (new) how to compile applications using GRAS
3532 - (new) emulation support (timing macros)
3534 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3536 SimGrid (2.93) unstable; urgency=low
3538 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3541 - Use Paje properly where used. Still to be sanitized properly.
3542 - Portability fix: Add an implementation of the contexts using pthread
3546 - Add xbt_procname(): returns the name of the current process.
3547 Use it to show the current process's name in all logging.
3549 - fix detection of older flex version and the reaction, since we do
3550 depend on modern ones (we use lex_destroy)
3551 - Better separation of SG and RL in the libs: remove all simulation code
3552 from libgras. As a result, this lib is now only 200k when stripped.
3553 Some of the xbt modules may also be duplicated (two sets and such) and
3554 should be cleaned/killed before SG3.
3555 - Insist on using xlC on AIX because of weird problems involving gcc there.
3556 - Cleanup the make remote stuff. This is now done by scripts
3557 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3558 mainly for our private use, but we're working on changing them to user
3561 - Bugfix: flush the socket on close only if there is some *output*.
3562 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3564 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3567 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3569 SimGrid (2.92) unstable; urgency=low
3571 Alpha 3 on the path to SimGrid 3
3575 - New! First try of benchmarking macros.
3576 - New! First try so that gras_stub_generator generate deployment and
3577 remote compilation helpers.
3579 - Bugfix: Initialization fix in msg_test.
3583 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3585 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3587 - API Change: gras_os_sleep to take the amount of seconds as a double.
3588 Accepting an int was error prone since it was the only location where
3589 seconds were coded as such. It leaded to damn rounding errors.
3590 - Bugfix: Hard to belive that timers ever worked before this.
3592 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3594 SimGrid (2.91) unstable; urgency=low
3596 Alpha 2 on the path to SimGrid 3
3600 - Bug fix in the lmm_solver.
3602 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3603 through the function MSG_paje_output.
3604 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3605 - It is possible to bound the rate of a communication in MSG with
3606 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3607 to write it in the changelog).
3608 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3613 - Various cleanups to the autotools stuff
3614 - Begin to move Gras examples to examples/gras/
3615 - Let make distcheck work again (yeah!)
3617 - documentation overhauled using doxygen.
3618 gtk-doc-tools is dead in SimGrid now.
3619 - Automatically extract all existing logging categories, and add the list
3620 to the documentation (long standing one, to say the less)
3622 - Cleanup the known architecture table. Reorder the entries to group what
3623 should be, and use a more consistent naming scheme.
3624 (some of the test dataset are still to be regenerated)
3625 - New! Allow library to register globals on each process just as userdata
3627 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3628 lookup time (for now).
3629 Use it in msg and trp.
3630 This cleans a lot the internals and helps enforcing privacy of the
3631 headers between the gras components.
3632 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3633 - Bugfix: gras_os_time was delirious in RL.
3634 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3636 - Reenable GRAS now that it works.
3638 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3640 SimGrid (2.90) unstable; urgency=low
3642 Alpha 1 on the path to SimGrid 3
3644 * It is a long time since the last release of SimGrid. I'm sorry about
3645 that but as I had told you, I was rewriting a lot of things. I apologize
3646 to those who had been reporting bugs to me and that I had not answered.
3647 If your bug is still in the new version, please tell me. Here is a
3648 summary of the main changes.
3650 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3651 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3652 portability, speed, and a lot more but you'll figure it out later.
3653 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3654 I wanted to release the new MSG as soon as possible and I have
3655 broken GRAS, which is the reason why, for now, only MSG is fully
3656 functional. A laconic description of these projects is available
3657 in the documentation.
3659 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3660 kernel optimized for our needs (called SURF but only the developers
3661 should use it). Hence, MSG is now roughly 30 times faster and I think
3662 that by rewriting a little bit MSG, I could event speed it up a little
3663 bit more. Beside the gain in speed, it is also much easier to encode a
3664 new platform model with SURF than it was with SG. More to come...
3666 * REVOLUTION 3: I have tried to change a little as possible the API of
3667 MSG but a few things really had to disappear. The main differences
3668 with the previous version are:
3669 1) no more m_links_t and the corresponding functions. Platforms are
3670 directly read from a XML description and cannot be hard-coded
3671 anymore. The same format is used for application deployment
3672 description. The new format is described in the documentation.
3673 Have a look in tools/platform_generation. There is a tiny script
3674 that converts from the old platform format to the new one. Concerning
3675 the application deployment format, parsing the old one is tricky.
3676 I think most of you should however be able to convert your files. If
3677 it is really an issue, I can write a C code that does the conversion.
3679 2) the toolbox tbx does not exist anymore. We now have a library
3680 with much more data-structures but without the hash-tables (we have
3681 dictionaries that are much faster).
3683 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3685 *****************************************************************************
3686 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3687 * don't feel like converting the oldies. (Mt) *
3688 *****************************************************************************
3691 Version 2.90: "the long awaited one"
3692 - Finished rewriting and debugging MSG. Rewrote the documentation.
3693 - disable GRAS for now since it needs to be ported to the newest SG
3696 - Finish the port to windows (using mingw32 for cross-compile)
3699 - Main loop and datastructures of SURF. A cpu resource object is
3700 functional. Surf can thus be used to create cpu's with variable
3701 performance on which you can execute some actions.
3703 2004-11-15 Martin Quinson
3704 - Port to ARM. Simply added the alignment and size descriptions. Should
3705 work, but the ARM machines are so slow that I didn't had the opportunity
3706 to 'make check' over there yet.
3708 2004-11-15 Arnaud Legrand
3709 - Trace manager now written. It uses a heap structure and is therefore
3710 expected to be efficient. It may however be speeded up (particularly
3711 when many events occur at the same date) by using red and black
3712 trees. One day maybe...
3713 - Max-min linear system solver written. It uses a sparse matrix
3714 structure taking advantage of its expected use. Most operations are
3715 O(1) and free/calloc are called as few as possible. The computation of
3716 the minimum could however be improved by using a red and black tree
3719 2004-11-03 Arnaud Legrand
3720 - Rename every gras_* function that was in xbt/ to its xbt_
3722 - Add a heap and a doubly-linked list to xbt
3723 - Added a dichotomy to the dictionaries. make check works as well before
3724 so I assume that the patch is correct. I do not know however if things
3725 run effectively faster than before now. :)
3727 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3728 SimGrid, and the version number is bumped to 2.x
3730 2004-10-29 Martin Quinson
3731 - Introduction of the remote errors.
3732 They are the result of a RMI/RPC on the remote machine.
3733 ErrCodes being scalar values, you can't get the host on which those
3734 errors did happen. Extending the error mechanism as in Gnome is possible.
3735 No idea yet whether it is a good idea.
3737 2004-10-28 Martin Quinson
3738 - Interface revolution: the Starred Structure Eradication.
3739 I used to do typedef struct {} toto_t; and then handle *toto_t.
3740 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3741 - ???_t is a valid type (builded with typedef)
3742 - s_toto_t is a structure (access to fields with .)
3743 - s_toto is a structure needing 'struct' keyword to be used
3744 - e_toto_t is an enum
3745 - toto_t is an 'object' (struct*)
3747 typedef struct s_toto {} s_toto_t, *toto_t;
3748 typedef enum {} e_toto_t;
3749 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3750 s_toto_t) is private.
3752 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3753 so that it can set the variable to NULL. It was so for dicts and sets,
3754 it changed for dynars.
3756 - Fix a bunch of memleaks in dict_remove
3757 - Fix a bug in sg/server_socket opening: it failed all the time.
3759 2004-10-07 Martin Quinson
3760 - Speed up dynar lookup operation a bit.
3762 gras_dynar_get is dead.
3764 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3765 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3766 (which gives you the address of the stored data).
3768 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3769 copy of the data using an affectation to do the job and not a memcpy.
3771 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3772 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3774 It does not really speedup the dynar test because they are
3775 setting elements all the time (and look them seldom). But the dict does
3776 far more lookup than setting.
3778 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3780 2004-10-05 Martin Quinson
3781 - Allow to (en/dis)able the cycle detection at run time.
3783 Whether we should check for cycle or not is now a property of each
3784 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3785 datadesc_cycle_unset allow to remove this property when previously set.
3787 Note that the cycle detection is off by default since it impacts the
3788 performance. Watch the data you feed GRAS with ;)
3790 This property is hereditary. Any element embedded in a structure having it
3791 set have it set for the time of this data exchange.
3793 You should set it both on sender and receiver side. If you don't set it on
3794 sender side, it will enter an endless loop. If you forget on receiver
3795 side, the cycles won't be recreated after communication.
3797 - Header reorganization.
3798 Kill gras_private.h, each submodule must load the headers it needs.
3800 2004-10-04 Martin Quinson
3801 - Interface revolution: do not try to survive to malloc failure.
3803 Now, gras_malloc and friends call gras_abort() on failure.
3804 As a conclusion, malloc_error is not a valid error anymore, and all
3805 functions for which it was the only gras_error_t return value are
3806 changed. They now return void, or there result directly.
3807 This simplify the API a lot.
3809 2004-09-29 Martin Quinson
3810 - Re-enable raw sockets.
3811 Created by gras_socket_{client,server}_ext;
3812 Used with gras_raw_{send,recv}
3815 It should allow to kill the last bits of gras first version soon.
3817 This is not completely satisfactory yet (duplicate code with
3818 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3821 - Simplify transport plugin (internal) interface by not passing any
3822 argument to _server and _client, but embedding them in the socket
3825 2004-09-28 Martin Quinson
3826 - Finish the port to AIX.
3827 autoconf was my problem (segfault within the malloc replacement
3828 function. No idea why)
3830 2004-09-16 Martin Quinson
3831 - Fix some size_t madness on 64bit architectures.
3833 2004-09-08 Martin Quinson
3834 - Reduce the number of system headers loaded, overload some more system
3835 calls (such as malloc to cast the result of the system one, and work
3837 - Fix and reintroduce the config support
3839 2004-09-07 Martin Quinson
3840 - Source code reorganization to allow Arnaud to surf all over there.
3841 - Allow to document the logging categories.
3842 - Remove all uppercase from logging categories and useless cleanup in names.
3844 2004-08-18 Martin Quinson
3845 Version 0.6.2 (protocol not changed; API changed)
3846 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3847 gras_error_t), and NULL when not found. Functions expecting a msgtype
3848 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3849 hopefully usefull message.
3850 - Portability to prehistoric sparcs again
3852 2004-08-17 Martin Quinson
3853 Version 0.6.1 (protocol not changed; ABI not changed)
3854 - prealloc some buffers to speed things up
3856 2004-08-11 Martin Quinson
3857 Version 0.6 (protocol not changed; ABI expended)
3858 - The parsing macro can deal with the references, provided that you add
3859 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3861 2004-08-09 Martin Quinson
3862 Version 0.5 (protocol not changed; ABI changed)
3863 - Allow to off turn the cycle detection code in data exchange at
3864 compilation time. It should be at run time, but I'm short of time (and
3865 the config stuff is still broken). That way, we keep dict out of the
3866 critical path, which is good because the performance is poor:
3867 - search not dichotomial yet
3868 - dynar give no way to access their content and memcpy everytime
3869 - In composed data description (struct, ref and so on), stop foolness of
3870 keeping the subtype's ID, but store the type itself. This keeps sets out
3871 of the critical path, which is good since they rely on dynar and
3872 dictionaries. The only loose of that is that we cannot detect the
3873 redeclaration of a structure/union with another content (but I'm not sure
3874 the code detected well this error before anyway). We still can detect
3875 the redefinition discrepancy for the other types.
3876 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3877 it breaks the code because of type-punning used all over the place).
3878 This breaks on all non-gcc architectures (for now).
3880 All those changes (plus the buffer of last time) allow me to gain 2 order
3881 of magnitude on cruel tests consisting of 800000 array of integers on two
3882 level of a hierarchical structure (200 secondes -> 4 secondes)
3885 - the selector of reference must now return the type it points to, not
3886 the ID of this type.
3888 2004-08-06 Martin Quinson
3889 Version 0.4 (protocol changed; ABI not changed)
3890 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3891 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3892 - Implement a buffer transport, and use it by default (it relies on tcp in
3893 real life and on sg in simulation).
3894 That's a bit hackish since I had a new field to the structure to store
3895 its data without interfering with the subtype ones. Inheritance
3896 is tricky in C. And that's a kind of reverse inheritance with one class
3897 derivating two classes. Or maybe a game with java interfaces. Anyway,
3898 that's damn hard in C (at least).
3899 Moreover, I got tired while trying to ensure plugin separation and
3900 genericity in SG mode. MSG wants me to do weird things, so let's go for
3901 cruel hacks (temporarily of course ;).
3902 See comment in transport_private.h:71
3903 - do not include all the _interface headers in private but in the files
3904 which really need them (to cut the compilation time when they are
3907 2004-07-26 Martin Quinson
3908 Version 0.3 (protocol not changed; ABI changed)
3909 - Major overhault of the datadesc interface to simplify it:
3910 - shorted the function names:
3911 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3912 - add a trivial way to push/pop integers into the cbps without malloc.
3913 This allows to make really generic sub_type description, which simply
3914 pop their size of the stack.
3915 - add a function gras_datadesc_ref_pop_arr() which does what users want
3916 most of the time: Declare a dynamic array (which pops its size of the
3917 stack) and declare a reference to it. Poor name, but anyway.
3918 - kill the post-send callback, add a post-receive one
3920 2004-07-23 Martin Quinson
3921 Version 0.2 (protocol changed; ABI changed)
3922 - add some testing for cpbs in the test cases, and fix some more bugs.
3923 This invalidate again the little64 data file, since I cannot regenerate
3925 - remove an awfull optimization in the logging stuff, allowing me to:
3926 - understand it again
3927 - learn gcc how to check that the argument match the provided format
3928 - fix all errors revealed by gcc after that
3929 - internal keys of dict are not \0 terminated. Deal with it properly in
3930 loggings instead of segfaulting when the user want to see the logs :-/
3932 2004-07-22 Martin Quinson
3933 - Fix some stupid bug preventing cbps (callback postit) from working
3935 2004-07-21 Martin Quinson
3936 - Some documentation cleanups
3937 - remove the useless last argument of msgtype_declare
3938 - rename the Virtu functions to fit into the 'os' namespace
3939 - move headers src/include -> src/include/gras/ and stop fooling with
3940 gras -> . symbolic link
3941 - make distcheck is now successful
3943 2004-07-19 Martin Quinson
3945 - Build shared library also
3946 - Install html doc to the right location
3947 - stop removing maintainer files in make clean
3948 - build tests only on make check
3950 2004-07-13 Martin Quinson
3952 - No major issue in previous version => change versioning schema
3953 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3954 corresponding dataset.
3956 2004-07-11 Martin Quinson
3958 - Get it working with any kind of structure (we can compute the padding
3959 bytes remotely for all the architectures I have access to)
3960 - Implement the structure parsing macro (still not quite robust/complete)
3961 - Improvement to the remote testing toysuite
3963 2004-07-10 Martin Quinson
3964 [autoconf mechanism]
3965 - get ride of a bunch of deprecated macros
3966 - actually run the test for two-compliment, not only compile it :-/
3967 - test whether the structures get packed (and bail out if yes. Damn.
3968 Alignment is a serious matter)
3969 - test whether the structures get compacted (but respecting the alignment
3970 constraints of each types)
3971 - test whether the array fields of structures can straddle alignment boundaries
3973 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3974 (took me 2 hours to find that bug, looking at the wrong place)
3975 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3976 before sending/receiving and are used to compute the offsets of fields
3977 - Given that padding size depend even on compiler options, keep track of
3978 alignment and aligned_size only for the current architecture. Not a big
3979 deal since we send structure fields one after the other (seems
3981 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3982 program, let it work on linux/gcc/little32. portability todo.
3984 2004-07-08 Martin Quinson
3985 - import and improve remote compilation support from FAST
3986 - make sure make check works on half a dozen of machines out there
3988 2004-07-07 Martin Quinson
3989 Let's say it's version 0.0.3 ;)
3990 - Implement conversions (yuhu!)
3991 - Let it work on solaris (beside conversion, of course)
3992 - Stupid me, using rand() to generate the conversion datatests in not wise.
3994 2004-07-06 Martin Quinson
3995 - Let make dist work, since I'm gonna need it to compile on remote hosts
3996 - Let Tests/datadesc_usage write the architecture on which the file was
3997 generated as first byte.
3998 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3999 architecture descriptions.
4000 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4001 on those architectures.
4002 - Optimization: send/recv array of scalar in one shoot
4004 2004-07-05 Martin Quinson
4005 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4007 - Plug a whole bunch of memleaks
4008 - each process now have to call gras_{init,exit}. One day, their log
4009 settings will be separated
4010 - Continue the code factorisation between SG, RL and common in Transport.
4012 2004-07-04 Martin Quinson
4014 - Redistribution between SG and RL.
4015 We wanna have to accept in SG, so move accepted related parts of RL in
4016 the common part. (more precisely, the dynar of all known sockets is no
4017 more a static in transport.c, but part of the process_data)
4019 [gras_stub_generator]
4020 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4022 2004-07-03 Martin Quinson
4023 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4024 them all up in one shot)
4026 - Fix the ugly case of reference to dynamic array.
4027 - New (semi-public) function gras_datadesc_size to allow the messaging
4028 layer to malloc the needed space for the buffer.
4030 - gras_socket_close now expect the socket to close (and not its address to
4031 put NULL in it after it). This is because the socket passed to handlers
4032 is one of their argument (=> not writable).
4034 - propagate the interface cleanup from last week in datadesc, ie remove a
4035 superfluous level of indirection. User pass adress of variable
4036 containing data (both when sending and receiving), and not of a variable
4037 being a pointer to the data. Let's say that I like it better ;)
4038 The price for that is constructs like "int msg=*(int*)payload" in
4039 handlers, but it's a fine price, IMHO.
4041 - Let it work in RL (yuhu)
4043 2004-06-21 Martin Quinson
4045 - porting SG plugin and SG select to new standards (works almost).
4046 - plug memleaks and fix bugs around.
4049 - cleanup the prototype of data recv and force users to specify when they
4050 want to handle references to objects. Test case working even for cycles.
4051 - plug memleaks. Valgrind is perfectly ok with this.
4053 2004-06-12 Martin Quinson
4055 - cleanup the separation between plugin and main code in plugin creation
4057 2004-06-11 Martin Quinson
4059 - Reput hook for raw sockets, needed for BW experiments
4060 - kill a few lines of dead code
4061 [Data description] Interface cleanup
4062 - gras_datadesc_by_name returns the searched type or NULL.
4063 That way, no variable is needed to use a type desc once, which makes
4065 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4066 two parameters were strings (field name, type name), leading to
4068 [Dicos] Interface cleanup
4069 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4070 This is consistant with the dynar API.
4072 2004-04-21 Martin Quinson
4074 - Porting to new standards.
4076 - interface cleanup.
4077 There is no bag anymore, no need to take extra provision to mask the
4078 pointers behind "ID".
4079 Better splitup of functions between files create/exchange/convert.
4080 This is still a bit artificial since convert and receive are so
4081 interleaved, but anyway.
4083 - add a queued message list to procdata (the ones not matching criteria
4085 - factorize some more code between SG and RL wrt procdata
4087 - use gras_exit in example to track memleaks
4088 - get rid of gs_example now that GS is properly integrated into gras
4089 - update run_test to integrate the latest tests (datadesc)
4091 - rename WARNINGn macros to WARNn since it prooved error-prone
4093 2004-04-19 Martin Quinson
4095 - register init/exit functions within gras module mechanism
4096 - send/receive function.
4097 Convertion is not implemented, but short-cutted if not needed.
4098 struct/array elements are sent one by one (instead of block-wise), but
4099 nobody really cares (yet). Get a prototype before optimizing.
4100 - tests (using a file socket) for DD send/receive on:
4101 - base types: int, float
4102 - array: fixed size, string (ie ref to dynamic string)
4103 - structure: homogeneous, heterogeneous
4104 - chained list, graph with cycle
4105 Believe it or not, valgrind is not too unhappy with the results. The
4106 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4110 - Bugfix when using a filename explicitly (instead of '-')
4112 2004-04-09 Martin Quinson
4114 - factorize more code between RL and SG in socket creation
4115 - Complete the implementation and tests of:
4117 o file (only in RL, and mainly for debugging)
4119 I lost 3 days to design a portable address resolver, and then decided
4120 that the prototype mainly have to run on my box.
4121 Addressing portability too early may be like optimizing too early :-/
4123 - use gras_init in the Tests instead of the crappy parse_log_opt
4124 (the latter function is removed)
4125 [Conditional execution]
4126 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4127 [Code reorganisation]
4128 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4129 Build examples against the RL library, since there is no way to disable
4130 its creation for now.
4132 For information, the beginning of coding on GRAS was back in june
4133 2003. I guess that every line has been rewritten at least twice since