1 ----------------------------------------------------------------------------
3 SimGrid (3.27.1) NOT RELEASED YET (v3.28 expected June 21. 2021, 03:32 UTC)
6 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
8 ----------------------------------------------------------------------------
10 SimGrid (3.27) March 29. 2021
12 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
14 Main user-visible changes:
15 - All functions marked as deprecated for one year were removed.
16 - Removed all that was related to the "Storage" type in all APIs and bindings.
17 "Disks" have been introduced more than a year ago to replace them for the better.
18 It's time to finish this replacement.
19 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
20 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
21 avoiding "initializer element is not constant" errors with some codes. Please be
22 careful to recompile your MPI codes if switching to this new SimGrid version.
23 - Network model ns-3 now requires version ns-3.28 or higher.
24 - Removed some option aliases that were deprecated since v3.23 (June 2019)
25 - 'verbose-exit' is now 'debug/verbose-exit'
26 - 'simix/breakpoint' is now 'debug/breakpoint'
27 - 'clean-atexit' is now 'debug/clean-atexit'
30 - New functions: Comm::sendto_{init,async} to initiate a communication
31 on between two (possibly remote) hosts.
32 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
33 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
34 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
35 or Mailbox::get_async<void>() if you really want to play with void*.
36 - Unify the interface of Activity::on_{start/activity}
37 - New function: Comm::get_dst_data()
38 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
41 - ptask should now work better with multicore hosts (FG#37).
42 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
43 - Internal cleanups and refactoring around models, to prepare for programmatic
45 - Modernize our tests of the performance models
49 - New section: Configuring the logs from the command line.
50 - New section: Programming API of logs.
51 - New section: Getting realistic results
52 - New section: Examples of SMPI Usage
53 - New part: SimGrid's Internals, with a section on Design goals
54 - Move back to the classical Breathe tool chain, and fix many small glitches.
57 - Handle leak checking: SMPI will now display a message if MPI handles are still
58 present at the end of a simulation. Option smpi/list-leaks:n will display the n
59 first leaked pointers and their types.
60 - smpi/display-allocs option : Display Memory allocations analysis, with hints
61 about the largest allocations. This only works for allocations with malloc/calloc
62 and not compiled with SMPI_NO_OVERRIDE_MALLOC
63 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
64 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
65 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
66 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
67 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
68 - Replay now opens and parses the trace file only once for all ranks when possible.
70 - better handling of MPI_PROC_NULL
71 - MPI_Group_incl was broken in some cases
72 - PAPI tracing now fixed
73 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
74 - SMPI will now fail properly when receive size < sent size, instead of
75 just setting MPI_ERR_TRUNCATE and keep going.
76 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
77 - Remove useless build patches and split build/run patches on the page
78 - Build in parallel when possible
79 - Test on yum-based systems
80 - New nightly tested Proxy-apps:
81 - OpenMPI mpi-test-suite
82 - Coral MCB, LSMS, Qbox, CLOMP
83 - ECP/Mantevo Cloverleaf3D, Ember
84 - Codevault bhtree_mpi
87 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
88 - Start to overhaul the implementation to modernize it (still ongoing).
91 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
92 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
93 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
94 - Deprecate xbt_dynar_to_array().
95 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
97 C binding and interface:
98 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
99 - Constness requirement has been lowered for parameter argv of sg_actor_start()
100 and sg_actor_create() It's now (char* const*). The previous version is still
101 available as sg_actor_start_() and sg_actor_create_().
103 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
104 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
105 - FG#37: Parallel tasks are limited to 1 core per host
106 - FG#62: Running "smpirun -replay" on large networks
107 - FG!46: Fix a few potential memory leaks in SMPI colls
108 - GH#343: Wrong consumed energy when an exec timeouts
110 ----------------------------------------------------------------------------
112 SimGrid (3.26) December 15. 2020
114 The Release release (the French lockdown was eased today).
116 Important user-visible changes:
117 - SimGrid now requires a compiler with C++14 support.
118 Sibling projects should upgrade their FindSimgrid.cmake
119 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
120 several users had difficulties to understand issues when using high bandwidth or
121 small latency events. The new value was already the default for SMPI and
122 should not cause too much performance hit. It can change some simulation
123 timings if simulation was skipping a lot of small events (which was actually
124 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
125 - On some network models (L07, CM02, LV08), an implicit loopback link is created
126 for local communication. Default bandwidth was 500MBps and latency was 15µs,
127 which was rather bad. Default values have been changed to 10GBps and 0 latency.
128 Options network/loopback-lat and network/loopback-bw have been added to adapt
129 these values if needed. Network and routing models already providing separate
130 loopbacks (clusters and such) are not impacted by this change.
131 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
132 metadata that are added to each MPI message by MPI libraries (tag,
133 src/dest, and communicator information). This may change simulation times
134 for very small message sizes, but should be more realistic.
137 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
138 - New plugins: wifi_energy and link_load. See the associated examples.
139 - New examples: wifi networks, and communication suspend.
142 - LTO is now enabled for Intel/clang compilers.
143 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
144 Setting it to "auto" will use all available cores, while setting it to n will
145 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
146 - Remove obsolete runtime option 'contexts/parallel-threshold'.
147 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
148 has been kept for compatibility.
149 - Finally remove obsolete snake_case() aliases for runtime options.
150 - Further improve the documentation.
153 - the on_exit() of each actor is also executed when the simulation deadlocks.
154 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
155 An example is provided for s4u::Comm but it should work with Exec and Io.
158 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
159 https://framagit.org/simgrid/SMPI-proxy-apps
160 - MPI/IO: supports names starting with "./"
161 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
163 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
164 building shared libraries, this may cause issues, so environment variable
165 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
166 - most temporary files should now be created in /tmp dir (or equivalent).
167 If this one does not allow execution of code (noexec flag), this may cause issues.
168 Please use another tmp directory (using TMPDIR or equivalent system variable)
172 - A large refactoring is ongoing, but this should be transparent for users so far.
175 - Drop xbt_str_split().
177 C binding and interface:
178 - The return type of the sg_actor_on_exit() callbacks is now 'void'
179 (int value was previously ignored)
180 - Many C functions were renamed to follow a common naming style
181 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
182 names are kept until version 3.30.
183 - Many MSG tests were converted to the new S4U's interface in C, that
184 was extended for that.
186 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
187 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
188 - FG#41: Add sg_actor_create C interface
189 - FG#43: xbt::random needs some care
190 - FG#48: The Impossible Did Happen (yet again)
191 - FG#50: Suspending an actor executed at the current timestamp fails
192 - FG#52: Zero-seconds timeout : "That's in the past already"
193 - FG#53: Crash while using ns-3 network model
194 - FG#54: How to suspend a comm?
195 - FG!22: Stochastic Profiles
196 - FG!24: Documentation and fix for xbt/random
197 - FG!35: Add a modeling hint for parallel links in doc
198 - FG!36: [xbt/random] Read/Write the state of the RNG
199 - FG!42: Link cumulated load plugin
200 - FG!44: Test condition variable wait_until+notify_all
201 - GF#18137: Allow different stack sizes?
202 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
203 - GH#139: Allow pthread creation in SMPI
204 - GH#336: Packet-level simulation using SMPI?
205 - GH#345: Error in the LMM while migrating a VM
206 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
207 - GH#352: pip install / python setup.py install fail to find pybind11
208 - GH!337: Fix link_energy plugin for wifi platforms
209 - GH!339: Add Mailbox set_receiver method to python binding
210 - GH!344: Cast hugepages macros parameters to int64
212 ----------------------------------------------------------------------------
214 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
216 The Palindrom Day Release.
218 Important user-visible changes:
219 - Improve the Python usability (stability and documentation).
220 - A nasty synchronization bug was ironed out, see also below.
221 - Python's doc was organized as a separate tree, now integrated with C++.
222 - C bindings of S4U were not part of the doc.
223 - The C++ doc was also improved as methods are now split by theme.
224 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
225 - This is mandatory to use the Java bindings.
226 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
227 - SimDAG++: Automatic dependencies on S4U activities (experimental)
228 - Some features are already implemented but not all of them
229 - Cannot block an activity until it's scheduled on a resource
230 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
231 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
234 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
235 - Drop support for Viva/Triva (old visualization tools), hence removed
236 TRACE_get_node_types() and TRACE_get_edge_types() functions.
239 - Require pybind11 version 2.4.
240 - Greatly improve locking during context switches with Python.
241 - Declaring Python callbacks to this_actor::on_exit is now working.
242 - No more tests are failing currently, even on win32.
245 - Actor: Merge signals on_migration_start/end into on_host_change
246 - Actor: Rename migrate() into set_host()
247 - Disk: Allow users to get the read and write nominal bandwidth values
248 - Exec: Implement wait_for(timeout)
249 - Io: Implement wait_for(timeout)
252 - Parse errors now raise a simgrid::ParseError that you may want to catch.
255 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
256 been renamed to fix a spelling error. As usual, the previous definitions are
257 kept with a deprecation warning. However, to avoid runtime errors, it is now
258 forbidden to override these deprecated methods in a derived class. Please use
259 the new names immediately if you need to override them.
262 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
263 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
264 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
265 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
266 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
267 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
268 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
269 - New module: random, providing classical random numbers generators.
272 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
273 allocations' size above which they are considered "shared" by default (as if
274 they were performed through SMPI_SHARED_MALLOC macros).
275 Default = 0 = disabled feature.
276 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
277 This can cause some troubles if codes are already overriding these. If this
278 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
279 help, but will make this feature unusable.
280 - Input check is now performed much more consistently and outputs precise
281 warnings for each error found
282 - New supported calls : MPI_File_set_view, MPI_File_get_view
283 - MPI I/O now only supports disks, as storage is discontinued.
284 - MPI I/O allows opening files without specifying absolute path
285 - Note : SMPI configuration options are now initialized by smpi_init_options()
286 call, which should be called if SMPI is not being used through smpirun.
288 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
289 - FG#9: Python bindings crashing
290 - FG#39: Missing s4u::Comm::wait_any_for example
291 - FG#42: Add support for ThreadSanitizer (TSan)
292 - FG!19: Removing RngStream
293 - FG!20: A module for RNG calls
294 - FG!21: Choice between ad-hoc and standard distributions implementations
296 - FG!25: Fix link in Mutex doc
297 - FG!26: Fix links in SMPI interface doc
298 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
299 - GH#305: Unscheduled tasks are still excuted
300 - GH#313: smpirun: manual outdated w.r.t. --help
301 - GH#321: [S4U] Get task remaining work ratio
302 - GH#323: Crash when an actor turn off his physical host
303 - GH#335: Missing links on dragonfly example svg
304 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
305 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
307 ----------------------------------------------------------------------------
309 SimGrid (3.24) October 9. 2019
311 The Clean Disk Release.
314 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
315 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
316 thus becomes the default resource to execute I/O operations. Examples
317 and the file system plugin have been modified to use disks instead of
318 storage elements. Storage can still be used but is doomed to disappear.
319 Users are thus adviced to update their simulators accordingly.
320 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
321 for consistency with pthread_barrier_wait()
322 - Host::get_englobing_zone() returns the englobing netzone
323 - Actor::on_destruction is now called in the destructor
324 Actor::on_termination new signal called when the actor terminates
326 - Global signals are now part of the Engine:
327 - on_platform_creation: after config settings, before the XML parsing
328 - on_platform_created: right after the XML parsing
329 - on_time_advance: each time the clock advances
330 - on_simulation_end: after simulation, before cleanups
331 - on_deadlock: as the name implies.
333 - sg_{actor,host,link}_{data,data_set}() now all exist.
334 Use them to attach user data to the object and retrieve it.
337 - Introduce an experimental Wifi model. It sounds reasonable
338 according to the state of the art, but it still has to be properly
339 validated, at least against ns-3 if not against reality.
340 - Improved the usability of ns-3. Several bugs were ironed out.
341 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
342 It is now expressed as 'idle:epsilon:allCores' to properly model the
343 consumption of non-whole tasks on mono-core hosts. More info in the doc.
346 - convert a new set of functions to the S4U C interface and move the old MSG
347 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
350 - Fortran bindings for DVFS have been removed.
351 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
352 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
353 - SMPI can now be selected by cmake's find_module(MPI) with
354 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
355 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
356 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
357 - trace-call-location can be used with TI traces, and replayed, and is compatible
358 with smpi/comp-adjustment-file.
359 - sleep events are now correctly traced/replayed.
360 - Default for trace-call-location is now to use file names and not full paths.
361 To revert to previous behavior (in case of collision of filenames), option
362 "smpi/trace-call-use-absolute-path" can be set to yes.
365 - Use the included xxHash as an hash implem when C++14 is usable.
366 - Option model-checker/hash was removed. This is always activated now.
367 - New option smpi/buffering controls the MPI buffering in MC mode.
368 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
369 This is useful to check for MPI compliance.
372 - New section on plugins: how to define a new one, and existing ones.
375 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
376 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
379 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
380 and <mount> tags. The rationale is finish to decouple what should be handle
381 by SimGrid kernel (i.e., resources and activities) from plugins built in
382 user space (e.g., the file system plugin). This new tag currently accepts
383 only three attributes (id, read_bw, and write_bw). All the other information
384 that was declared with the storage related tags now has to be expressed as
385 properties. An example of platform using this new tag is available at
386 examples/platforms/hosts_with_disks.xml
389 - 'expect signal' can now accept more than one potential signal.
391 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
392 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
393 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
394 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
395 - FG#31: per-actor data
396 - FG#34: SG_BARRIER_SERIAL_THREAD?
397 - FG#35: model-checker does not like buster-produced binaries
398 - FG!13: MC: complete workaround in the error msg seen on modern systems
399 - FG!15: execute_flops now logs compute
400 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
401 - FG!17: ns-3: unblock the right number of communications + others issues
402 - FG!18: Improving the performance of the ns-3 bindings
403 - GH#189: Energy consumption of parallel tasks
404 - GH#219: Error in the throughput of TCP transfer
405 - GH!330: Minor polishing of English: Configuring_Simgrid
406 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
407 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
408 - GH!334: Rework energy plugin, again
410 ----------------------------------------------------------------------------
412 SimGrid (3.23.2) July 8. 2019
415 - Nicer introduction page.
416 - Migrate the "Deploy your application" page to the new doc.
417 - Move Java as a subtree of MSG.
420 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
423 - Drop xbt_dynar_sort_strings().
426 - Really fix FG#26: Turning off a link should raise NetworkFailureException
427 - FG#27: Wrong exception thrown to wait_any when link is turned off
428 - GH#328: Java: Canceling multiple tasks in a single vm/host
430 ----------------------------------------------------------------------------
432 SimGrid (3.23) June 25. 2019
434 The Exotic Solstice Release.
437 - SunOS and Haiku OS support. Because exotic platforms are fun.
438 - Stop setting random seed with srand48() at initialization.
439 - Use addr2line as a fallback for stacktraces when backtrace is not available.
440 - Build option -Denable_documentation is now OFF by default.
441 - Network model 'NS3' was renamed into 'ns-3'.
444 - Simgrid can now hopefully be installed with pip.
447 - wait_any can now be used for asynchronous executions too.
450 - New log appenders: stdout and stderr. Use stdout for xbt_help.
451 - Drop xbt_dict_dump.
454 - 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
455 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
456 - MPI interface is now const correct for input parameters
457 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
458 - Fortran bindings for async collectives.
459 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
462 - Remove option 'model-check/record': Paths are recorded in any cases now.
463 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
464 always sparse. That's an immense gain in memory, and can even be
465 faster because of cache effects. So there is no need to clutter the
466 code to allow the user to go for the unefficient mode.
469 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
470 models proved to be more accurate than these old experiments.
472 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
473 - FG#1: Broken link in error messages
474 - FG#2: missing installation documentation
475 - FG#3: missing documentation in smpirun
476 - FG#6: Python bindings not available on PyPI
477 - FG#7: simple cmake call requires doxygen
478 - FG#8: make python bindings an optional dependency
479 - FG#10: Can not use MSG_process_set_data from SMPI any more
480 - FG#11: Auto-restart actors forget their on_exit behavior
481 - FG#12: -Denable_lto=OFF doesn't disable LTO
482 - FG#13: Installs unstripped file 'bin/graphicator'
483 - FG#14: Installs the empty directory 'doc/simgrid/html'
484 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
485 - FG#17: Dead link in doc (pls_ns3)
486 - FG#20: 'tesh --help' should return 0
487 - FG#21: Documentation link on http://simgrid.org/ broken
488 - FG#22: Debian installation instruction are broken
489 - FG#26: Turning off a link should raise NetworkFailureException exceptions
490 - GH#133: Java: a process can run on a VM even if its host is off
491 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
492 - GH#326: Valgrind-detected error for join() when energy plugin is activated
494 ----------------------------------------------------------------------------
496 SimGrid (3.22) April 1. 2019
498 The Easter Christmas Release (this one is somewhat late).
501 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
502 - Not complete yet: asynchronous activities (amongst others) are still missing
503 - Still ongoing: the interface may change in the future. We need more
504 testers! Please report any glitches.
505 - No new project using Java should start now. Please switch to Python.
508 - Some of the internal cleanups may lead to speed improvements:
509 - The hard limitation on the amount of simulated actors with
510 Java+Mac was removed. Now, the available memory is the only limit.
511 - Our refcounting was tidyied, leading to 10% speedups in some cases.
512 - We are still working on making our code robust to the actor kills
513 and hosts' churn. Things are improving, but it's not perfect yet.
514 - Replaced our own code to display a backtrace (that was forking addr2line)
515 with the Boost.Stacktrace library.
516 You won't see your backtraces without this optional dependency.
517 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
518 - Stop setting random seed with srand() at initialization.
521 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
522 XML file version remains 4.2 since old files are still compatible.
525 - Process termination which was broken at version 3.21 has been repaired.
526 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
527 - Hide the examples into examples/deprecated. New users should use Python.
530 - Drop MSG_process_create_from_stdfunc() from the API.
531 This C++ function was a pimple in the C API, made necessary at some
532 point by the Java bindings. This is fixed now.
533 - Hide the examples into examples/deprecated. New users should use S4U.
534 - MSG_process_create and MSG_process_attach now crash if the host used
535 to run this process is off.
536 - Fix the protype of MSG_process_on_exit()
537 Now use (int,void*) callbacks instead of (void*,void*) ones.
538 The implementation was ways too messy to actually work, I guess.
541 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
542 too often and provide more accurate timings.
543 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
544 - MPI_Alltoallw support
545 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
546 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
547 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
548 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
550 - Basic implementation of generalized requests (SMPI doesn't
551 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
554 - Drop sg_cmdline. Please use xbt_cmdline instead.
555 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
556 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
557 OsSemaphore is implemented in a portable way with C++11 threads.
558 This should allow much more threads to be created at the same time,
559 allowing Mac and Java users to simulate many more actors.
560 - Implement the 'thread' factory with std::thread instead of xbt ones.
561 It is not possible to set the stack size with threads anymore, but
562 -fsplit-stack is the way to go nowadays when using threads.
563 - Drop the xbt_os_thread_t module (now unused)
564 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
565 - Drop xbt_str_join_array().
566 - Drop cunit, use Catch2 instead.
569 - Many cleanups in the kernel::activity namespace. This was long
570 overdue, and shall open the path to many future endeavors.
573 - #132: Java: a process can not shut down its own host
574 - #220: S4U: detached send causes exception if sender terminates soon after sending
575 - #261: Document the parameters of parallel execution's constructor
576 - #300: [s4u] BarrierPtr is missing
577 - #314: SMPI args internal cleanup
578 - #316: Fix a bug related to the CPU utilization of multi-core VM
579 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
580 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
581 - #325: Turning off a host has different behavior on sleeping actors and computing actors
583 ----------------------------------------------------------------------------
585 SimGrid (3.21) October 3. 2018
587 The Restarting Documentation (TRD) Release.
590 - Convert the user manual to Sphinx for improved usability.
591 Unstable version now hosted on https://simgrid.frama.io/simgrid
592 Many glitches remain; stay tuned.
593 - Restrict the Doxygen reference API to the installed header files.
594 (documenting all internals this way is overkill + doxygen scales badly)
595 - New tutorial on S4U (highly inspired from the old MSG one, but with
596 a git repository to fork as a starting point, and a docker image)
597 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
598 - The Developper manual is still to be converted (not compiled until then)
599 - Some parts are still missing in this conversion (such as the
600 platform and deployment sections) while others were blindly converted
601 and would need more love (such as the configuration flags).
602 Things will certainly further improve in the future releases.
605 - Fully reimplement auto_restart mechanism that was utterly broken
606 (fix #22, #131 and #281 that were all very related)
607 - Implement semaphores (fix #303 and #291)
608 - When creating an actor from a function and its parameters,
609 move-only parameters are not allowed anymore, as it would prevent
610 the actor to be restartable if its parameters are consumed on run.
611 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
612 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
613 IO operations that can be started, waited for, or canceled as a regular
615 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
619 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
620 - New host variable: 'core_count'
623 - Remove xbt_os_thread_specific features
624 - Remove portability wrapper to condition variables
625 - Remove xbt_os_thread_yield()
628 - MPICH collective selector now mimics MPICH 3.3b
629 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
630 - MPI_Init(NULL,NULL) is now allowed
631 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
632 - Command-line arguments should now be handled properly in Fortran simulations
633 - MPI Attributes and related callbacks should work in Fortran
634 - Apps using C + Fortran should now work
635 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
636 - Support MPI_Error_String
639 - Due to an internal bug, Msg.run() must now be your last line.
640 We hope to fix it in a future release, and we are sorry for the inconvenience.
643 - #22: Process autorestart seem to only work with CAS01 cpus
644 - #56: Feature request: dynamic SMPI replay
645 - #93: simgrid should not eat --help
646 - #111: How to change the loopback link characteristics?
647 - #116: Platform cluster radical
648 - #117: Platform cluster/host order
649 - #129: Replay_multiple on steroid
650 - #131: Java : autostart does not work
651 - #143: Setting a breakpoint at a given time
652 - #235: Network energy models should be integrated
653 - #264: Add ptask L07 resource tracing
654 - #271: Dynamic and manual replay of SMPI traces
655 - #279: Breakpoints option uses time deltas instead of absolute time
656 - #281: Daemonized actor no longer daemonized after an auto-restart
657 - #282: TIT convertor erase traces when using absolute path in the trace list file
658 - #285: segfault when a s4u actor kills itself with msg process tracing activated
659 - #286: Pajé traces are not exposing the number of cores
660 - #287: Command-line parsing should stop when encountering '--'
661 - #288: MPI_Init(NULL, NULL)
662 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
663 - #290: Method to check if a mailbox has a message ready to be consumed
664 - #291: [s4u] semaphores do not seem available yet
665 - #294: Alltoall collective implementation of mpich
666 - #295: Starting a Host as OFF is not supported
667 - #296: DTD too permissive
668 - #299: Add a s4u barrier example
670 ----------------------------------------------------------------------------
672 SimGrid (3.20) Released June 24. 2018
674 The proxy snake_case() release.
677 - The whole API is now using snake_case() instead of mixing it with camelCase.
678 Compatibility wrappers in place for one year.
679 - We took the opportunity to increase the consistency of the API.
680 Please report any remaining glitches.
681 - New function simgrid::s4u::Host::get_actor_count:
682 Returns the number of actors running on a specific host.
685 - SMPI is now tested with ~45 proxy apps from various sources, with none or
686 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
687 - Replay: The replay file has been re-written in C++.
688 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
689 supported. They are stored in the trace and used when replayed.
690 - Basic support of MPI_Cancel. Robustness not guaranteed.
691 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
692 (only for MPI_COMM_TYPE_SHARED).
693 - New option: smpi/privatize-libs, to add external shared libs to be privatized
694 by SMPI. They will be copied locally and loaded separately by each process.
695 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
696 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
697 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
698 with internal index optind with multiple processes. Only works if MPI_Init has
700 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
701 syntax without preprocessing tricks.
704 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
705 behavior tracing is no longer limited to MSG
706 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
707 this means that traces now need to be updated or re-obtained; both Irecv
708 and Isend lines in your traces have now in total 5 mandatory fields:
709 <rankid> <command> <to/from rankid> <tag> <size>
710 To update your traces, it suffices to add a 0 for the tag here.
711 - TIT now also supports waiting for a distinct request via MPI_Wait.
712 Wait/Test now wait for a specific request, not just the last one that was
713 issued. This unfortunately means another update, because we need to
714 identify which request you want to wait for. We do this via the
715 triplet (sender, receiver, tag), which needs to be added:
716 <rankid> <command> <sender> <receiver> <tag>
717 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
718 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
721 - The deprecation of MSG is ongoing (but this should not impact you).
722 Many MSG functions are now simple wrappers on the C API of S4U. If
723 you wish to convert your code to S4U, find the S4U counterparts of
724 your MSG calls in src/msg/msg_legacy.cpp.
725 - MSG can still be used, but won't evolve anymore.
728 - Allow to run the Link energy plugin from the command line with
729 --cfg=plugin:link_energy
730 - Rename Energy plugin into host_energy
731 - Rename Load plugin into host_load
734 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
735 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
736 - Improve performance when handling timeouts of simix synchros.
739 - Config: the C API is now deprecated (will be removed in 3.23), and
740 the C++ API has been slightly improved.
743 - Fix several build issues on OSX.
744 - Move simgrid_config.h to simgrid/config.h (old header still working)
747 - #143: Setting a breakpoint at a given time
748 - #150: Inconsistent event names in SMPI replay
749 - #258: daemonized actors hang after all non-daemonized actors have completed
750 - #267: Linker error on unit_tmgr
751 - #269: SMPI: tracing of MPI_Wait/all/any broken
752 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
754 ----------------------------------------------------------------------------
756 SimGrid (3.19.1) Released March 22 2018
758 The Fixed ABI Release.
760 As you may know, we are currently refactoring SimGrid in deep.
761 Upcoming SimGrid4 will be really different from SimGrid3: modular,
762 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
764 Our goal is to smooth this transition, with backward compatibility
765 and automatic update paths, while still progressing toward SimGrid4.
767 SimGrid remains open during works: The last pure SimGrid3 release was
768 v3.12 while all subsequent versions are usable alpha versions of
769 SimGrid4: Existing interfaces remain unchanged, but the new S4U
770 interface is budding and the internals are deeply reorganized.
772 Since 2015, we work hard to reduce the changes to public APIs. When
773 we need to rename a public library symbol in S4U, we let your
774 compiler issue an explicative warning when you use the deprecated
775 function. These messages remain for four releases, i.e. for one full
776 year, before turning into an error. Starting with v3.15, your can
777 also adapt to API changes with the SIMGRID_VERSION macro, that is
778 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
780 Starting with this v3.19.1, our commitment to reduce the changes to
781 the public interfaces is extended from the API to the ABI: a program
782 using only MSG or SimDag and compiled against a given version of
783 simgrid can probably be used with a later version of SimGrid without
784 recompilation. We will do our best... but don't expect too much of
785 it, that's a really difficult goal during such profund refactoring.
787 The difference between v3.19 and v3.19.1 is that the former was
788 accidentally breaking the ABI of MSG, while the later is restoring
791 S4U and kernel APIs will still evolve until SimGrid4, with one-year
792 deprecation warnings as currently. In fact, cleaning up these
793 interfaces and converting them to snake_case() is one release goal of
794 v3.20. But don't worry, we are working to smooth this upgrade path.
796 In summary, new projects should start with S4U to benefit of the
797 future, but old MSG projects should still be usable with no change.
799 ----------------------------------------------------------------------------
801 SimGrid (3.19) Released March 20 2018
803 The Moscovitly-cold Spring release.
805 SMPI user-visible improvements:
806 - Support of MPI_Type_create_subarray
807 - Unimplemented MPI calls now abort (instead of issuing a warning)
808 Some harmless calls around errhandlers still only issue a warning.
809 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
810 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
811 is now possible (mmap privatization was refactored to this end).
812 Adding new SMPI instances should also be technically doable now.
814 SMPI internal cleanups:
815 - Removed index notion from SMPI -- no more "getPid() - 1"!
816 In tracing files, actors are called rank-<PROCESS-ID> but these are
817 NOT the real ranks (tracing several communicators is not supported yet).
818 - Deprecate smpi_process_index() (will be removed in 3.22)
819 - Better testing on classical proxy apps, through an external project:
820 https://github.com/simgrid/SMPI-proxy-apps/
823 - Use a graphical TOC to make it easier to find the documentation you need
824 - Revamp the MSG tutorial
827 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
828 - Remove parameter of MSG_process_killall().
829 Resetting the PID was bogus anyway (several actors could have the same PID).
832 - Execution->setHost() can be called after start(), to migrate it.
833 - Comm::test_any() is now implemented.
834 - s4u::Actor now has onCreation() and onDestruction() signals.
835 - Install some kernel header files for the users' plugins and more.
836 Warning, their API is really not stable yet.
837 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
838 include/simgrid/kernel/routing/*Zone.hpp
841 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
843 Build System and other cleanups:
844 - Remove unused run-time parameter "tracing/onelink-only".
845 - SimGrid now works with intel compilers (MC still buggy)
846 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
847 - tesh kills subprocesses on timeout (but not on windows)
850 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
853 - #194: Feature request: simgrid::s4u::Comm::test_any()
854 - #245: migrating an actor does not migrate its execution
855 - #253: Feature Request: expose clusters as objects
856 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
857 - #255: Tesh broken on Windows
858 - #256: Modernize FindSimGrid.cmake
859 - #257: Fix (ab)use of CMake install
861 ----------------------------------------------------------------------------
863 SimGrid (3.18) Released December 24 2017
865 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
867 MSG IS NOW DEPRECATED. Please use S4U for new projects.
868 - Support for MSG should not be removed from SimGrid before 2020,
869 but future evolutions will be limited to the integration of
870 user-provided patches.
873 - s4u::onDeadlock() signal, run before stopping the simulation.
874 - s4u::Exec: asynchronous executions (abstraction of background
876 - s4u::Host->getLoad() returns the achieved speed in flops/s
879 - Link::name() is deprecated and will be removed in v3.21, use
880 Link::getCname() instead.
881 - Mailbox::getName() changed to return a std::string, use
882 Mailbox::getCname() instead to get a char*.
883 - Storage::getName() changed to return a std::string, use
884 Storage::getCname() instead to get a char*.
885 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
886 to have both allocation and deallocation of the map in user space.
889 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
890 The amount of flops remaining to do is a vector, not a scalar.
891 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
892 name implies on both sequential and parallel tasks.
893 - Both changes fix GitHub's #223 using PR #237 as a basis.
894 Thanks Michael Mercier.
895 - Most examples were converted to S4U and hidden elsewhere at tests.
896 Rationale: we still want MSG to work; we want newcomers to use S4U.
899 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
900 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
903 - Switch to the faster dlopen privatization mechanism by default
904 - Documentation improvments
907 - Live migration is getting moved to a plugin. Dirty page tracking is
908 the first part of this plugin. This imply that VM migration is now
909 only possible if one this function is called:
910 - C/MSG: MSG_vm_live_migration_plugin_init()
911 - C/C++: sg_vm_live_migration_plugin_init()
912 - Java: Msg.liveMigrationInit()
913 For C and C++, "simgrid/plugins/live_migration.h" also has to be
917 - Define class simgrid::xbt::Path to manage file names.
918 - Removed unused functions:
919 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
920 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
921 xbt_graph_export_graphviz()
922 - xbt/str.h: xbt_str_join()
923 - Remove unused datatypes:
924 - xbt/heap.h: use std::priority_queue or boost::heap instead
925 - xbt/swag.h: use boost::intrusive::list instead
928 - New link_energy plugin for the consumption of the links.
929 - All of the operations on files and storage contents have been
930 packaged into a plugin (src/plugins/file_system). The current
931 public interface can be found in
932 include/simgrid/plugins/file_system.h
933 To use these functions you now have to initialize the plugin by
934 calling MSG_storage_file_system_init() just after calling
935 MSG_init() or sg_storage_file_system_init() just after creating
939 - Remove the undocumented/untested tag <include>
942 - Remove viva specific tracing as the tool is no longer maintained
945 - #248: Kill of finished processes leads to segfault
946 - #240: xbt_cond_wait_timeout should gracefully return for C apps
947 - #239: please implement signal s4u::onDeadlock()
948 - #230: segfaults when exit() without run()
949 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
950 - #223: MSG_task_get_flops_amount() not working with parallel tasks
951 - #222: Actor::kill() doesn't really kill and segfaults
952 - #221: odd LMM warning when killing an actor
953 - #120: Memory leak when the processes are forcefully killed
955 ----------------------------------------------------------------------------
957 SimGrid (3.17) Released October 8 2017
959 The Drained Leaks release: (almost) no known leaks despite the tests.
961 Even very long simulations will preserve your memory: our testsuite
962 stresses SimGrid for over 45mn on fast machines for a coverage of
963 over 80%, and there is only one single known leak, of about 4kb.
967 - Comm.detach(): start and forget about asynchronous emission. A cleanup
968 handler may be given to free resources if the comm cannot be completed.
969 - this_actor::send(mailbox) is now mailbox->put()
970 - New: simgrid::s4u::Comm::wait_all()
971 - New: Host.execute() for remote executions.
974 - Unused option network/sender-gap is removed.
977 - Deprecate MSG_task_isend_with_matching(): This unused feature
978 really complicates our internals. Will be removed in v3.20.
981 - Improved context termination. It is now possible to free resources, even
982 when a process is forcibly killed.
983 - ContextBoost: add support for Boost versions above 1.61.
986 - Bring back run-time option --cfg=exception/cutpath to remove exception
988 - Removed unused functions:
989 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
990 xbt_str_rtrim(), xbt_str_trim().
991 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
994 - Removed header files obsolete since SimGrid 3.12:
995 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
997 - #3: SD_exit should be made optional
998 - #120: Memory leak when the processes are forcefully killed
999 - #159: Threading test regression in Actor refcounting
1000 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1001 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1002 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1003 - #191: VM migration and pstate
1004 - #192: Updating the energy consumptions of all hosts crashes with VMs
1005 - #195: All actors have PID=0 in the logs
1006 - #204: Sometimes segfault with thread contexts and mmap privatization
1007 - #222: Actor::kill() doesn't really kill and segfaults
1008 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1010 ----------------------------------------------------------------------------
1012 SimGrid (3.16) Released June 22. 2017.
1014 The Blooming Spring Release: developments are budding.
1017 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1018 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1019 - Java JAR file should be finally fully working on Mac OSX too.
1021 XML platforms: Switch to platform v4.1 format.
1022 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1023 - <zone> can be used as a synonym for the now deprecated <as>
1024 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1025 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1026 - <actor> can be used as a synonym for the now deprecated <process>
1027 - state_file and avail_file periodicity is now easier to express
1028 (check the documentation) the old behavior should still work.
1029 * Storage: not backward compatible, but it seems that nobody ever used it
1030 - Remove attribute 'content_type' of <storage_type>: was never used
1031 - Make attribute 'model' of <storage_type> optional: for future usage
1032 - Remove Bconnection model property: was never (in)validated. Replaced by
1033 the maximum of the read and write bandwidth as a resource constraint.
1036 - New and Backwards Compatibility break:
1037 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1038 When one wants to get the list of tasks whose states have changed during a
1039 simulation round, s/he has to allocate and free a dynar and use it as
1040 argument to this function. The former SD_simulate (double how_long)
1044 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1045 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1048 - New function to update the consumption of all hosts at once.
1049 - Fix the model for multi-core hosts, linear in the amount of busy
1050 cores with an abnormality for pIdle (see paper).
1051 BUG: Still not satisfactory for ptask on mono-cores.
1054 - The netzone are now available from the MSG API.
1055 The old names still work, but are now deprecated.
1058 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1059 - New option: smpi/keep-temps to not cleanup temp files
1060 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1061 mallocs mode is used, allows to change the size of the fake file used
1062 (default 1MB), to potentially limit the number of mappings for large runs.
1063 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1064 - Fortran ifort and flang compilers support
1065 - New RMA calls supported (experimental):
1066 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1067 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1068 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1069 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1070 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1071 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1074 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1075 - Behind the scene, ActivityImpl are now automatically refcounted.
1076 Removing this nasty bottleneck will greatly simplify our internals.
1079 - Replay: New function xbt_replay_action_get():
1080 Retrieve the function previously associated to an event type.
1081 - DROPPED FUNCTION: xbt_str_varsubst()
1082 - DROPPED MODULE: strbuff. We don't need it anymore.
1083 - DROPPED MODULE: matrix. We don't need it anymore.
1084 - DROPPED MODULE: lib. We don't need it anymore.
1086 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1088 SimGrid (3.15) Released March 22. 2017
1090 The Spring Release: continuous integration servers become green
1091 We fixed even the transient bugs on all target architectures:
1092 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1093 Windows, FreeBSD, NetBSD.
1095 - Rename NetCards to NetPoints.
1096 This was intended to help NS3 users, but that's not a netcard.
1097 That's a point in the routing algorithm, let's avoid wrong simplifications.
1100 - New: MSG_process_yield(). Stop and yield to other processes.
1101 - New: MSG_process_daemon(). Daemon processes are automatically killed
1102 when the last non-daemon process terminates
1103 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1104 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1105 to make room for the upcoming network energy plugin.
1106 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1109 - Ensure that an actor can kill itself with Process::exit()
1110 - Kill the obscure NativeException. Nobody want to survive the issues
1111 it denotes, so use JniException that is a RuntimeException (not to
1112 be caught explicitly).
1113 - Partial bug fix in initialization. SimGrid flags on command line were
1114 consumed at C level but stayed in the original Java String[] args.
1115 This could mess users' args[i] if SG flags were not put at the end of
1117 The SimGrid flags are now removed from the Java arguments. However,
1118 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1119 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1120 It might be if you have --log or --cfg flags in the command line.
1121 - Fix numerous memleaks all around the place. In particular, around VMs.
1125 - s4u::onPlatformCreated: right before the simulation starts
1126 - s4u::onSimulationEnd: right after the main simulation loop
1127 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1128 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1129 event from the availability_file changes the avail speed.
1130 - Links are now usable from s4u
1131 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1132 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1133 - New examples: The conversion of MSG examples to S4U has begun
1134 - Actors: create, kill, migration, and suspend
1135 - Applications: master-worker and token-ring
1136 - Action replay: communications and storage
1137 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1140 - Backwards Compatibility breaks
1141 - The SD_route_get_list and SD_route_get_size functions have been removed.
1142 They are replaced by the unique
1143 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1144 The route from src to dst is built in the links dynar whose size can
1145 be obtained with xbt_dynar_length.
1146 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1147 sg_host_route_bandwidth and sg_host_route_latency.
1148 Macros ensure the backwards compatibility, but you should fix your code
1151 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1152 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1153 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1154 We don't support 100% of it, but it should be enough. Contact us if not.
1155 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1156 naive and probably inaccurate)
1157 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1158 mapped onto a unique small file using some system magic.
1159 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1160 - Fortran: cleanups, fixes, support of user-added operations
1161 - RMA: MPI_Accumulate are applied in correct order.
1162 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1163 - Better support for MPI_IN_PLACE
1164 - Support for MPI_Win attrs and keyvals.
1165 - Support MPI_Comm_group_create, MPI_Type_size_x
1166 - MPI ops should be properly applied to all allowed types
1167 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1168 algorithm by a non failing one from mpich.
1171 - Kill the fifo data container: we don't use it anymore.
1174 - A <cluster> can now be created with different speed values to represent pstates.
1175 see examples/platforms/energy_cluster.xml for an example.
1177 SimGrid (3.14.159) Released December 28. 2016
1179 The previous release was only a crude approximation of a Pi release;
1180 this one should be a bit better.
1182 - Revert a last minute change that broke on Mac OSX.
1183 - Fix the changelog and NEWS files.
1184 - Improve the documentation, in particular of the routing module.
1185 - Integrate some patches flying around in Debian.
1187 SimGrid (3.14) Released December 24. 2016
1190 * The whole documentation was reworked and reorganized. There is
1191 still much room for improvement here, but we're on it.
1194 * We now need python3 for our testing infrastructure.
1195 * Model checking needs libevent
1198 * Now works on FreeBSD
1201 * Privatization now works on FreeBSD
1203 * Privatization is now activated by default in smpirun
1204 You can turn it off with -no-privatize if something goes wrong.
1206 * Call-location tracing for SMPI.
1207 You can add the exact location (filename / linenumber) of an MPI call to
1208 your trace files and slow down or speed up the simulation between two
1209 consecutive calls by using an adjustment file (see the documentation).
1211 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1212 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1213 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1214 values were ignored in some cases. The timings of these functions can now
1215 be significantly different.
1217 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1218 smpi/running-power is renamed to smpi/host-speed
1220 * smpi/grow-injected-times option to enable or disable multiplication of the
1221 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1222 make simulation less precise (but also much faster).
1224 * smpirun script should be (much) faster for large deployments.
1226 * SMPI tracing: fixed issue with poor matching of send/receives.
1228 * Replay: Fix broken waitall
1230 New functions and features
1231 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1233 Dropped / renamed functions and features
1234 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1235 - MSG_mailbox_is_empty() -> Mailbox::empty()
1236 - MSG_mailbox_front() -> Mailbox::front()
1237 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1238 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1239 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1240 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1241 * Don't pass the free_f parameter to property related functions:
1242 - MSG_host_set_property_value()
1243 - MSG_as_router_set_property_value()
1244 - MSG_storage_set_property_value()
1245 * VM properties. Since msg_vm_t are msg_host_t, just use
1246 MSG_host_get_property_value() and friends
1247 * VM I/O related things:
1248 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1249 - Unimplemented save/restore methods
1250 * MSG_as_router_get_property_value() was redundant with
1251 MSG_environment_as_get_property_value().
1252 - Removed MSG_as_router_*propert*() functions
1253 - Added MSG_environment_as_set_property_value() for consistency
1254 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1255 Well, they are still there for now, but deprecated with a warning.
1256 Please switch to xbt_dict_new_homogeneous() before this is removed
1258 * Task affinity. Its intended behavior (that was very badly tested
1259 and probably not really working) was deceiving what most users
1260 would have hoped here.
1261 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1262 * Option network/coordinates is now useless and should be dropped.
1265 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1268 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1269 * Added Dragonfly topology support
1272 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1273 EiBps, EBps, Eibps, Ebps for bandwidth.
1274 They may become useful to some lucky ones.
1277 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1278 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1279 * ex/async_waitAll: new example, on asynchronous communications
1282 * Memory usage should be decreased for simulations with a large number
1283 of processes. This also helps for SMPI.
1285 SimGrid (3.13) stable; urgency=low
1287 The Half Release, a.k.a. the Zealous Easter Trim.
1289 - We removed half of the lines, that were mostly experimental cruft.
1290 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1291 (not counting blanks and comments -- according to openhub.net)
1292 - The internals are now compiled in C++ (and will soon be clean C++)
1293 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1294 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1296 * Backwards Compatibility breaks
1297 - Removed Lua simulation bindings (switch to C or Java for that).
1298 Lua can still be used to describe platforms
1299 - Removed Java kernel plug-ins.
1300 Will be reintroduced after the ongoing major internals reorg.
1302 - the following functions were removed.
1303 They were too specific and should be reimplemented in a generic
1304 way, with filter function.
1305 - MSG_task_listen_from_host
1306 - MSG_mailbox_get_count_host_waiting_tasks
1307 - MSG_mailbox_put_with_timeout was removed.
1308 Please use MSG_task_send_with_timeout instead.
1310 - the SD_application_reinit function was removed. It has been a noop for a while.
1311 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1312 replaced by a more flexible mechanism.
1313 - The following functions thus do not exist anymore
1314 - SD_workstation_get_access_mode
1315 - SD_workstation_set_access_mode
1316 - SD_workstation_get_current_task
1317 - Basic estimation functions have been removed but can easily be replaced
1318 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1319 - SD_workstation_get_computation_time => amount / sg_host_speed()
1321 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1322 Use VM.getSpeed()*load/100 for the legacy behavior.
1324 - option enable_tracing was removed. It was not doing anything for a while.
1325 - In the ModelChecker:
1326 - the model-checker now ptraces the model-checked process which means
1327 you cannot use a debugger on the latter anymore (we might make this
1328 optional in the feature);
1329 - removed soft-dirty page tracking;
1330 - remove model-checked side snapshot management,
1331 MC_snapshot() and MC_compare_snapshot();
1332 - keep the MC_cut() function as a stub (it was not really working
1333 in the previous release).
1336 * All options are consistently in kebab-case. Old names are kept as alias.
1339 * Switch to platform v4 format.
1340 - Rename from 'power' to 'speed' the attributes describing the amount of
1341 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1342 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1343 - In <host> and <link>, attributes availability and state are gone.
1344 It was redundant with state and availability traces, and with peak values.
1345 - In <cluster>, attributes availability_file and state_file are gone.
1346 It was too complex and unused.
1347 - Kill <gpu>. Was not doing anything.
1348 - The DOCTYPE points to the right URL:
1349 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1350 (the file at this address now documents the changelog since its v1)
1351 - A warning is emitted for unit-less values (they are still accepted).
1352 - speed. Default: 'f' or 'flops'. Also defined:
1353 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1354 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1355 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1356 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1357 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1358 - latency. Default: 's' second. Also defined:
1359 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1361 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1362 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1365 * s4u::Host is now the preferred public interface to the Host features.
1366 sg_host_* functions are C bindings to the exact same behavior
1367 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1370 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1371 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1372 They were deprecated since a few years, and probably did not even compile anymore.
1375 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1376 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1377 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1378 encouraged to update their codes. A list of the modified functions can be found at the end of
1379 include/simgrid/simdag.h
1382 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1383 the lambda or closure passed as a parameter will run in kernel mode.
1384 All the callback functions should be rewritten to that interface at some point.
1387 * Reorganizing and cleaning the internals all around the place.
1390 * Remove old default barrier/bcast buggy algorithms (see #18407)
1391 * Various bug fixes to handle more codes
1392 * Remove the need for the --foreground option of smpirun (it is still
1393 accepted for backward compatibility).
1396 * Kill the setset data container: MC don't use it anymore.
1397 * Kill the queue data container: it made more sense with GRAS.
1398 * Kill the xbt_peer_t data type: it's useless without GRAS.
1399 * Kill rm_cb feature of config sets: it was never useful.
1400 * Kill graphxml parsing feature. It was not used.
1401 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1403 - xbt_dynar_sort_strings(), when the content is char*
1404 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1405 They throw exceptions on invalid input;
1406 * C++ support for declaring CLI flags (simgrid::config::Flag);
1407 * class for abstracting different signal backends (simgrid::xbt::signal).
1408 with no external dependencies (we need very simple signals).
1411 * refactoring and cleanup of the code;
1412 * ongoing process to cleanly separate the model-checking algorithms
1413 from the code model-checking support.
1415 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1417 SimGrid (3.12) stable; urgency=low
1419 The Facelift Release.
1422 * Require g++ v4.7 at least to not speak prehistorical C++.
1423 * Require Boost 1.48 (for signal2 component).
1424 * Java must be version 7 at least when activated.
1425 * Builds on Windows again (including Java bindings).
1426 * Tracing is now always enabled (no way to turn it out)
1427 * Remove GTNetS. It was not working anyway.
1428 * Various cleanups in the cmake scripts.
1429 * Move headers around to sort them out on installed systems:
1430 - instr/instr.h -> simgrid/instr.h
1431 - instr/jedule/* -> simgrid/jedule
1432 - simdag/datatypes.h was removed
1433 - simdag/simdag.h -> simgrid/simdag.h
1434 - msg/datatypes.h was removed
1435 - msg/msg.h -> simgrid/msg.h
1438 * Interface improvement:
1439 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1440 - Sanitize the interface in MSG_task_ module:
1441 - Merge two functions that were close enough but misleading:
1442 set_compute_duration(t) -> set_flops_amount(t)
1443 get_remaining_computation(t) -> get_flops_amount(t)
1444 - set_data_size(t) -> set_bytes_amount(t)
1445 get_data_size(t) -> get_bytes_amount(t)
1446 - Massive cleanups in the functions related to the energy
1447 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1448 - New: MSG_host_get_pstate()
1450 - msg/energy/onoff: switching hosts on and off
1452 * Interface improvement:
1453 - Rename Host.isAvail() to Host.isOn()
1454 - Rename Process.currentProcess() to Process.getCurrentProcess()
1455 - Rename Task.setDataSize() to Task.setBytesAmount()
1456 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1458 - #18874: Actually allows the GC to reclaim tasks
1462 - SIMIX_process_throw: raises an exception in a remote process
1463 * Refactoring: Separate sub-modules
1464 - libsmx: the public interface, as libc in a real system
1465 - popping: the strange dance that converts a user request into a kernel handling
1466 - smx_context_*: the virtualization mechanisms that embed the user code
1467 - smx_*: the handling of each simcalls
1468 * Interface cleanup:
1469 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1470 * Rename smx_action_t into smx_synchro_t, making explicit that these
1471 things are used to synchronize processes with their environment.
1472 For example, a communication is a sort of synchronization involving
1473 the communicating processes (that may block until the exchange) and
1474 the platform. The same can be said from computations, etc.
1476 - #18888: segfault when a process ends before its kill_time
1480 - 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.
1481 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1482 - MPI_Type_set_name, MPI_Type_get_name
1483 - MPI_*_c2f and MPI_*_f2c functions
1484 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1485 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1486 - Activate a lot of new tests from the mpich 3 testsuite
1488 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1489 - InfiniBand network model added: Based on the works of Jerome Vienne
1490 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1491 - When smpi/display_timing is set, also display global simulation time and application times
1492 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1493 * Collective communications
1494 - 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
1495 - 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/).
1496 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1498 - Replay now uses algorithms from wanted collective selector
1499 - Replay can be used with SMP-aware algorithms
1500 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1501 - 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.
1503 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1504 - Correctly use loopback on fat-tree clusters
1505 - Asynchronous small messages shouldn't trigger deadlocks anymore
1506 * Energy/DVFS cleanup and improvement
1507 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1508 - new: smpi_get_host_pstate()
1511 - "Full" network optimization flag was broken since Surf++
1512 - Better handling of precision flags in maxmin
1513 - Fix bug causing sometimes "Impossible" errors
1514 - Properly pass cluster properties to included hosts
1515 * Improvement of the Energy plugin.
1516 - Always update the consumption before returning that value
1517 - New property: watt_off to denote the dissipation when the host is off
1518 - New functions getWattMinAt and getWattMaxAt to retrieve the
1519 dissipation of pstates that we are not currently at.
1520 * Java: class NetworkLink renamed to Link
1521 * New function: simcall_process_get_kill_time()
1522 * Massive rename s/workstation/host/
1523 - That's intrusive, but that's good for the project consistency. Sorry.
1524 - Change config option "workstation/model" into "host/model"
1527 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1528 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1529 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1530 - xbt_dirname and xbt_basename for non-POSIX systems
1532 * The model checker now runs as a separate process.
1533 * The model checker runs must now be launched with the new simgrid-mc program.
1534 * Record/Replay: the MC can display a textual representation of a path in the
1535 execution graph. It can then be replayed outside of the model checker.
1537 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1539 SimGrid (3.11) stable; urgency=low
1544 * Normalizing pointers addresses tool for better diff between logs
1547 * Add cloud examples using new VMs
1548 - examples/msg/cloud/two_tasks_vm.tesh
1549 - examples/msg/cloud/simple_vm.tesh
1550 - examples/java/cloud/cloud.tesh
1551 - examples/java/cloud/migration/migration.tesh
1552 * Add java surf examples:
1553 - examples/java/surfPlugin/surf_plugin.tesh
1554 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1555 - examples/java/surfCpuModel/surf_cpu_model.tesh
1556 * Add SMPI+MSG example:
1557 - examples/smpi/smpi_msg_masterslave/
1562 - msg task destroy cancel test
1563 - msg_host on/off test
1564 * Move all tests in testsuite to teshsuite (adding tesh files)
1565 * Restructure teshsuites
1566 - one folder for each kind of test
1567 * Restructure AddTests.cmake
1569 - structure the order of tests (with sections)
1572 * Add virtual machine
1573 - creation of a VM on a PM
1574 - migration of a VM from a PM to another PM
1576 - MSG_process_join(msg_process_t process, double timeout)
1577 - msg_bar_t MSG_barrier_init(unsigned int count)
1578 - int MSG_barrier_wait(msg_bar_t barrier)
1579 - void MSG_barrier_destroy(msg_bar_t barrier)
1580 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1581 * New option "msg/debug_multiple_use" to help debugging when a task is used
1584 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1586 - Increase code coverage in test suites
1590 * Protect context stacks against stack overflow. The number of protected memory
1591 pages allocated on the top of each stack (1 by default) can be configured
1592 with the new command line option --cfg=contexts/guard_size:<value>.
1593 * Simcalls are now generated by a python script that
1594 - generates files included by SimGrid
1595 - checks that all the functions exist, and proposes prototypes
1597 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1598 host_set_data, host_get_data
1600 - simcall_process_join(smx_process_t process, double timeout)
1601 * Fix bug where sleeping processing could not be suspended.
1604 * Translate surf models from C to C++
1605 - Generic classes for all models: Model, Resource, Action
1606 - A generic interface for each kind of model (CPU, Network, Storage
1607 Workstation, WorkstationVM)
1609 * Translate surf routings from C to C++
1610 * Add callbacks using sigc++ or boost::signals2
1611 - Add callback functions for resource creation/destruction
1612 - Add callback functions for action state change
1613 - Handle Energy as a plugin
1614 * Replace swag by boost::intrusive
1615 * Add new routing models for clusters. For documentation, see
1616 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1617 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1618 parameters for cluster tag
1619 - Fat trees, with topology="FAT_TREE" and
1620 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1621 - see examples/platforms/torus_cluster.xml and
1622 examples/platforms/fat_tree_cluster.xml
1623 * More documentation
1626 * Hostfiles support host:nb_processes construct to deploy several processes on
1628 * Collective communication algorithms should not crash if used with
1629 improper number of nodes and report the error.
1630 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1631 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1632 MPI_Dims_create, MPI_Cart_sub are supported.
1633 * New interface to use SMPI programmatically (still depends on MSG for
1634 some parts, see examples/smpi/smpi_msg_masterslave):
1635 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1639 * Global variables privatization in MPI executables is now performed at runtime
1640 with the option smpi/privatize_global_variables (default:no).
1641 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1642 dynamic libraries loaded by the application are not privatized (static
1643 linking with these libraries is advised in this case)
1646 * Options defined in XML work correctly now.
1649 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1650 copied into simgrid.jar or not (ON by default). Use this option if you want
1651 to reduce the size of the installed simgrid.jar, *and* the native libraries
1652 are kept installed elsewhere.
1653 * Surf binding with SWIG (code generated in maintainer mode only):
1654 - plugin to handle callbacks
1655 - CPU model only for the moment
1658 * Supernovae build mode is definitively removed. It was used to improve
1659 inlining and inter-module optimizations. It is nowadays superseded by
1660 link-time optimizations commonly available in compilers.
1661 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1662 * Add boost dependency for surf++
1663 * Add new macro for tests
1664 - ADD_TESH(name <tesh_args>)
1665 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1669 - xbt_bar_t XBT_barrier_init(unsigned int count)
1670 - int XBT_barrier_wait(xbt_bar_t barrier)
1671 - void XBT_barrier_destroy(xbt_bar_t barrier)
1672 * Make the xbt_os_time module public
1674 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1676 SimGrid (3.10) stable; urgency=low
1678 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1681 * Reintegrate Java to the main archive as desynchronizing these
1682 package is not acceptable anymore (Java is now considered stable)
1683 * Add explicit synchronization facilities through semaphores
1684 * Bug fix: Task.setDataSize() only changed the C world, not the value
1685 cached in the Java world
1688 * Dramatically change the way files are handled. API and internals changed, but
1689 this part of MSG was not considered as production grade either.
1690 * Add explicit synchronization facilities through semaphores
1691 * Add a new function MSG_host_get_process_list()
1692 * Preliminary DVFS support (see examples/msg/energy/ for details)
1695 * SMPI is now included directly in the libsimgrid as the windows
1696 linker doesn't force us on splitting it anymore.
1697 * Improvements of the SMPI replay tool:
1698 - Most of the collective communications are now rooted in the same process as
1699 in the original application.
1700 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1701 used until now). Multiple data types can now be used in a trace.
1702 - The replay tool now supports traces produce either by TAU or a modified
1704 - Bug Fix: the compute part of the reduce action is now taken into account.
1705 - Gatherv collective is now supported
1706 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1707 of smpirun outputs time independent traces for the current run. One file
1708 is created per process. If too many processes are simulated, this behavior
1709 can be changed to one file for all processes by using the
1710 tracing/smpi/format/ti_one_file flag
1711 * smpirun generates the host file if needed (with given host count and platform)
1712 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1713 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1714 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1715 --cfg=smpi/coll_selector:(mpich/ompi)
1716 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1718 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1719 happening inside a collective SMPI call.
1720 * Fix the behavior of complex data types handling.
1721 * Make MPI_Wtime another synchronization point to take computations into
1723 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1724 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1725 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1726 (no privatization of global variables yet)
1727 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1730 * Verification of liveness properties is now available for SMPI applications
1731 (in addition to MSG applications)
1732 * Bugged examples using SMPI in examples/smpi/mc/
1733 * Add --cfg=model-check/visited option. Allows the verification of infinite
1734 programs. Detection of loops in the execution thanks to the system state
1735 comparison and reduction of the state space to explore. Can be combined with
1736 DPOR for safety properties.
1739 * Allow to change SimGrid configuration (see --help) within the code
1740 thanks to SD_config() as it can be done in MSG.
1741 * Add a new function SD_task_set_amount() upon user request.
1744 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1745 * Remove rule based routing (no more PCRE dependency)
1746 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1747 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1748 * Add a loopback_bw and loopback_lat options to cluster tag.
1749 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1750 routing. Now peers are encapsulated in an AS and have their own private
1751 router but this is transparent.
1754 * Our own implementation of getline is renamed xbt_getline, and gets
1755 used even if the OS provide a getline(). This should reduce the
1756 configuration complexity by using the same code on all platforms.
1757 * New type: xbt_cfg_elm_boolean.
1758 * Allow to use yes/no for boolean configuration options in the command line.
1759 * Allow to disable SimGrid cleanups at exit from command line option.
1760 There are situations where one may want a simulation to end with an exit.
1761 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1762 annoying when scripting around the simulator. Adding a
1763 --cfg=clean_atexit:no allows to circumvent this issue.
1766 * Lots of memory leaks were corrected in this release.
1767 There is no known memory leaks anymore, in all of our 600+ tests.
1768 * New command line option --version, to get SimGrid version information.
1769 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1771 * Supernovae builds are deprecated, and expected to be removed in the next
1774 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1776 SimGrid (3.9) stable; urgency=low
1778 The Grasgory release: GRAS is really dead now.
1780 * Complete overhaul of the internal host structures scheme.
1783 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1784 was considered as experimental and badly maintained since then.
1785 * Keeping it was thus a trap to our potential users, that could take
1786 it instead of MSG or SMPI by mistake despite is pity state.
1787 * GRAS seems to have very few users (if any), and no one volunteered
1788 to maintain it further. It also induces a lot of XBT code (for
1789 portability sake), that must be maintained too.
1790 * For all these reasons, we killed GRAS. If someone wants to revive it
1791 in the future, don't cry, our git history still remembers of GRAS.
1794 * Major overhaul. Merge our documentation again as time proved that
1795 splitting it was really not helping our users.
1796 * Further improve the developer documentation to help newcomers
1797 hacking on SimGrid itself. The user documentation (and in
1798 particular, the beginner documentation) is still in a sorry state.
1801 * Now works on Windows too!
1802 * Much more extensive test suite, from MPICH
1805 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1806 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1807 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1808 * Bug fix: let task be scheduled when the last dependency to be solved is
1809 a control dependency.
1810 * Remove SD_load_environment_script function.
1811 Use the C sg_platf function if you want to declare a platform
1815 * New function: MSG_process_get_number()
1816 * Old function documented: MSG_config()
1817 * Remove MSG_load_platform_script function
1818 Use the C sg_platf function if you want to declare a platform
1822 * Change the default value of the TCP_gamma constant (maximal size of TCP
1823 congestion window) to a more realistic 4MiB value. If you notice changes in
1824 your simulation results, you can fall back to the previous 20k tiny window
1825 by adding --cfg=network/TCP_gamma:20000 on command line.
1826 * (Hopefully) fix a bug wrt periodic availability and state traces
1827 * Bug fix: use default values at start when first event in availability/state
1828 trace is not at time 0.
1831 * remove the "new_" part of function name sg_platf_new_trace_connect
1832 (resulting in sg_platf_trace_connect), since it does not create
1836 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1837 use case where it's really mandatory, and maintaining it was a pain in
1839 * New: xbt_fifo_search(), search an item with a user-provided
1840 comparison function instead of dumb pointer comparison.
1843 * Fix the lua deployment:
1844 Use `simgrid.init_application()` before deployment instead of
1845 `simgrid.msg_register_application()` after.
1848 * Transfer the tracing files into the corresponding modules.
1850 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1852 SimGrid (3.8.1) stable; urgency=low
1854 The "we are told that some people want to also *install* the simgrid
1857 * Add missing file "tesh.1" to the archive.
1859 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1861 SimGrid (3.8) stable; urgency=low
1863 The Psssshiiiit release: SimGrid jumps into the Cloud.
1866 * Add an experimental interface to manipulate VMs. They are mainly
1867 process groups with very few intrinsic semantic, but they should
1868 allow you to build the semantic you want easily.
1869 * New function: MSG_host_set_property_value()
1870 * New function: MSG_process_on_exit(). To clean memory in all cases.
1871 * Bug fixes that made the host (and link) failures unusable.
1872 * Add a way to auto-restart process when the host in which they are
1873 executing comes back (ON_FAILURE="RESTART" on deployment file,
1874 MSG_process_auto_restart_set).
1875 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1876 please stop using the old ones, they are DEPRECATED.
1878 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1879 Please use MSG_init() instead. (reducing the amount of entry
1880 points in the library helps us).
1881 * Make it impossible to link against the wrong version of the lib
1882 * Deprecate MSG_clean(). No need to call it anymore.
1883 * Function MSG_get_host_number() is not deprecated anymore.
1886 * Split the doc into a user guide and a reference guide.
1887 * Start a developper guide to help people hacking on SimGrid.
1890 * Enable tracing by default. This modules rocks you should use it.
1891 * Remove option custom_flags. Now use environment variables CFLAGS
1893 * Use default cmake things to detect lua instead of home grown ones.
1894 * New option "enable_mallocators" to disable mallocators, for debugging
1895 purpose ("on" by default).
1898 * Bug fixes around the resource failures: don't let the processes
1899 survive the host they are running onto.
1900 * Add an interface to auto-restart processes when the host in which they are
1901 executing comes back.
1902 * Ensures that SIMIX_clean is called automatically. It's not part of
1903 the public interface anymore (bindings should be updated).
1906 * Bug fix for when SD_Simulate is called with a positive value: be careful
1907 when comparing doubles. Sometimes they are different for non significant
1909 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1910 parallel task whose initial work is distributed among host according
1911 to the Amdahl's law. Such tasks are created with a parameter alpha
1912 that corresponds to the non-parallelizable part of the computation.
1913 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1914 between two sets of workstations assuming a 1D block distribution (each
1915 workstation owns a similar share of data) on both sides.
1917 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1918 Data redistribution will be automatically scheduled once parent and child
1919 are both scheduled. The filling of computation_amount and
1920 communication_amount structures is now done seamlessly thanks to the chosen
1922 * New function SD_workstation_dump to display various information
1923 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1924 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1925 bandwidth on the route onto which the task is scheduled and the amount of
1927 To divide the nominal bandwidth by 2, the rate then has to be:
1928 rate = bandwidth/(2*amount)
1929 * Compute tasks that have failed can now be rescheduled and executed again
1930 (from their beginning)
1931 * Increasing source code coverage (src/simdag is now covered at 95.8%
1935 * Re-implement time-independent trace replay using SMPI (at the
1936 smpi_smp_* level) instead of MSG. This should replace
1937 examples/msg/actions/actions.c
1938 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1939 hindexed and structs)
1940 * Implement the exchange of non-contiguous data.
1941 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1942 * Correct behavior of smpi/sender_gap and set its default value to 0
1943 * Add option to asynchronously send small messages to allow better
1944 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1945 specifies the size in bytes under which messages will be asynchronously sent.
1946 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1947 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1950 * New C interface to define a platform: XML is now optional.
1951 For more info, please check include/simgrid/platf.h.
1952 * New interface to define random platforms from the C:
1953 For more info, please check include/simgrid/platf_generator.h and
1954 examples/msg/masterslave/masterslave_platfgen.c
1955 * Export a sg_cmdline dynar containing all the arguments we got from
1959 * Two new tracing options for adding comments to trace file so you
1960 can track your experiments (see --help-tracing for details).
1961 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
1962 * Adding the SimGrid version that generated the trace file as a comment.
1963 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1964 * Fix to avoid key clashes on Paje links
1965 * Other minor fixes related to the Paje specification
1968 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1969 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1970 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1971 * Greatly improve the robustness of mmalloc to user errors (such as
1972 using an area after freeing it, or freeing it twice)
1974 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1976 SimGrid-java (3.8.1) stable; urgency=low
1978 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1979 Warning: all methods are visible, but only some of them are
1980 implemented so far. Check the source (src/jtrace.c)
1981 for further information.
1982 * New module: org.simgrid.msg.File (SimGrid File management functions)
1983 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1984 * Change the meaning of Process.restart: now restart the process from
1985 the beginning, like MSG_process_restart in C.
1986 * Add Process.setAutoRestart: handling of process restart when failed
1988 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1989 you to retrieve the properties of the processes/hosts
1990 * Deprecate Msg.clean(): you can just forget about it now.
1991 * New function Process.getCount(), that only works when compiling
1992 with the not yet released version 3.9 of the C library.
1994 * New context factory based on Coroutines. It mandates a modified JVM
1995 but then, the simulations run about five times faster, and there is
1996 no limit to the amount of processes (beside of the available memory).
1998 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2000 SimGrid (3.7.1) stable; urgency=low
2003 * Restore the prototype of MSG_process_create_with_environment() to
2004 the pre-3.7 situation by removing the kill_time argument.
2005 * Add a MSG_process_set_kill_time() function instead.
2008 * Fix weird behaviors when dealing with parallel tasks.
2011 * Simgrid is now built as a dll.
2012 * Simgrid-java now works on Windows.
2013 * Simgrid-Java is now included into Windows package.
2016 * First pre-build package for MacOSX.
2019 * Fix compilation when using MSG_USE_DEPRECATED.
2020 * Fix some compilation issues on Macs and Windows.
2021 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2022 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2024 We discovered that the Lua console is broken, but we are missing the
2025 manpower to fix it right now. The problem existed in 3.7 too, so we
2026 are not blocking the release for that. Sorry if you depended on this
2027 feature, any help would be really welcome.
2029 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2032 SimGrid-java (3.7.1) stable; urgency=low
2034 The "Java aint got to be bloated and slow" release
2037 * Various internal cleanups and performance improvement
2038 Simulations are expected to run up to twice faster or so
2039 * Make Process.kill(process) an instance method, not a static one
2040 * User processes are not java.lang.Thread subclasses.
2041 This breaks the compatibility (sorry), but previous API was
2042 brain-dead, making it impossible to have non-trivial
2043 initializations in the process constructor.
2044 * Require a full constructor per Process sub-class.
2045 Kinda breaks the compatibility (sorry), but this allows a much more
2046 efficient way to launch the processes at simulation startup.
2047 * Do not embeed our version of semaphores, java 1.5 can be considered
2048 as sufficiently prevalent for us to not dupplicate its features.
2052 * Add examples for almost every part of the API
2053 We spotted and fixed a lot of bugs in the process
2055 * New module: asynchronous communication API
2056 * New function: Process.sleep()
2057 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2058 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2059 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2061 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2063 SimGrid (3.7) stable; urgency=low
2065 The "spring cleanups (before the next Big Project kicks in)" release.
2068 * We can specify the SMPI latency/bandwidth factor with command line
2069 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2070 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2071 You can also use the "config tag" from platform file by adding this line
2072 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2073 (see "example/platforms/tag_config.xml" to use "config tag").
2074 Note that the command line supersedes the platform file configuration.
2075 * Change the correction factors used in LMM model, according to
2076 the latest experiments described in INRIA RR-7821.
2077 Accuracy should be improved this way.
2078 * Use the partial invalidation optimization by default for the
2079 network too. Should produce the exact same results, only faster.
2080 * Major cleanup in surf to merge models and split some optimization
2081 mechanisms from the core of the models. As a result you can now
2082 specify which model to use (e.g., --cfg=network/model:LV08
2083 --cfg=cpu/model:Cas01) and which optimization mode to use
2084 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2085 Incompatible combinations should err at initialization. See
2086 --help-models for the list of all models and optimization modes.
2087 * The CLM03 workstation model was dropped for simplicity because it
2088 used the deprecated CM02 network model. Use default instead.
2089 * Rename the TCP_gamma configuration option to network/TCP_gamma
2090 * Rename the coordinates configuration option to
2091 network/coordinates, and document it
2092 * Use now crosstraffic keyword instead of the terribly misleading
2093 fullduplex keyword. It is activated by default now in the current
2094 default model, use --cfg=network/crosstraffic:0 to turn it off.
2095 * Ongoing refactoring the model parsing to make XML files optional
2096 See include/simgrid/platf.h for details (still to be completed)
2099 * Major overhaul of the documentation. Almost instructive now :/
2100 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2101 functions and friends. This interface was considered as
2102 deprecated since over 2 years, it's time to inform our users that it is.
2103 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2104 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2105 use these (crappy) functions in your code.
2106 These functions will be removed soon. Stop using them now.
2107 * Deprecate MSG_get_host_{table,number}
2108 Implement MSG_hosts_as_dynar() instead.
2109 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2110 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2113 * Stabilize the parallel execution mode of user contexts
2114 * Introduce configuration variables to control parallel execution:
2115 - contexts/synchro: Synchronization mode to use when running
2116 contexts in parallel (either futex, posix or busy_wait)
2117 - contexts/parallel_threshold: Minimal number of user contexts
2118 that must be part of a scheduling round to switch to parallel
2119 execution mode (raw contexts only)
2120 * Fix bugs that prevented to use suspend/resume along with
2121 synchronization structures.
2122 * Fix bugs in process termination that lead to invalid memory access
2123 in very specific conditions.
2126 * Introduce a parallel mode for the models (controlled by surf/nthreads
2127 configuration item). In our tests, running the models in parallel
2128 never lead to any speedups because they are so fast that the gain
2129 of computing each model in parallel does not amortizes the
2130 synchronization costs, even when ultra fast futexes are used.
2131 This is released anyway because YMMV.
2134 * Performance boost by using a swag internally to compute the set of
2135 tasks that are finished and should constitute the return value of
2139 * Enable it by default now that it is considered rather stable.
2142 * Documentation of the tracing functions.
2143 * Performance gains when tracing categorized/uncategorized resource
2144 utilization by avoiding calls to get route when updating resource
2145 variables. LMM constraints are being used instead.
2146 * API changed to set task categories. Use MSG_task_set_category instead
2147 of TRACE_msg_set_task_category, and SD_task_set_category instead
2148 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2149 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2150 * Examples for link user variables added (see at examples/msg/tracing/)
2151 * Deprecated function TRACE_msg_set_process_category completely removed
2152 * Trace header updated according to the latest Paje file format
2153 * Tracing network lazy updates, no longer obligate users to use full updates
2154 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2155 * Experimental: let user code declare/set/push/pop application states for hosts
2156 * API changed to allow the manual creation of graph configuration files
2157 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2160 * Improve the API of Lua MSG bindings, using the Lua spirit.
2161 * Each simulated process now lives in its own Lua world (globals are
2162 automatically duplicated). It helps writing simulators. Will allow
2163 to run Splay programs within SimGrid in the future.
2164 * Add a Chord example in Lua, equivalent to the MSG one.
2167 * Start the implementation of a solution to express temporal
2168 properties, not only local assertions. This is still an
2169 experimental work in progress, stay clear from it to be safe.
2173 - Add new runtime parameters --help-logs and --help-log-categories
2174 to display information about supported logging parameters and
2176 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2177 don't exists anymore.
2178 * Mallocators: allow value NULL for the reset function.
2180 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2181 create homogeneous dictionaries, where all the elements share the
2182 same free function. Non homogeneous dictionaries will be
2183 deprecated in the next release.
2184 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2185 - Multi-level dictionaries are deprecated.
2187 - new function xbt_dynar_search_or_negative() that is useful when
2188 you have less than 2 million elements in your dynar and don't
2189 want of the extra complexity of catching exceptions when the
2190 element is not found.
2192 - Make xbt_os_thread module (for thread portability) public.
2193 Documentation is still to come, sorry.
2195 - Cleanups and simplifications to make it maintainable again.
2196 - Exotic features (such as memalign and valloc) were removed.
2197 - The metadata were extended and improved so that the
2198 model-checker becomes able to explore and inspect the heaps.
2199 - This may induce a performance drop when enable_model-checking is
2200 ON in cmake (even if it's not used in the simulation), but it is
2201 necessary at this point to get MC working.
2203 Turn model-checking OFF if simulation performance matters to you.
2204 Not enabling it at runtime is not enough, disable it in cmake.
2206 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2209 SimGrid (3.6.2) stable; urgency=low
2211 The "Not coding new stuff allows to polish old things" release.
2214 * New bindings to the NS3 packet level simulator (experimental)
2215 * Use the raw (efficient) execution contexts instead of the sysv
2216 (portable) ones when possible.
2217 * libpcre is now mandatory in any cases since not using it led to
2218 severe performance loss and possibly other issues
2219 * Update the XML platforms:
2220 - G5K: include the latest machine in Nancy
2221 - GridPP and LCG: new platforms
2222 * Documentation was partially updated, at least (more to come)
2224 Bug fixes, cosmetics and small improvements
2225 * Free terminated processes before the end of the simulation to avoid
2226 exhausting the memory of users having very dynamic amount of
2228 * Bug fix and cosmetics about canceling non-running tasks
2229 * Bug fix about the dot loader's issues when using libcgraph
2232 * Create an installer for windows with nsis (amd64 and win32)
2233 - Add an hello world project to illustrate simgrid project creation.
2234 - Embed libpcre into the Simgrid installer to avoid
2235 its compilation burden
2236 * The raw execution contexts should work on Apple now
2237 * Port to Windows 64 bits
2238 - Sysv contexts now have an implementation for this arch
2239 - GRAS communication features now support this arch
2240 * Drop support for borland compiler on windows
2241 - this code was not maintained, and we kinda depend on gcc nowadays
2242 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2243 * Fix portability issue on unstable ubuntu: linker became picky on
2246 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2248 SimGrid (3.6.1) stable; urgency=low
2250 The "Oops, we broke Macs too" release
2253 * Fixed contexts detection so that raw ones are used when possible
2254 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2255 leads to a strange error, with user code segfaulting sometimes when
2256 the generated code is not perfectly aligned (which is not
2257 controllable from the user side, depends on the amount of code)
2260 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2262 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2264 SimGrid-java (3.6) unstable; urgency=low
2267 * Split of every thing from simgrid v3.5 into a separate package.
2269 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2271 SimGrid (3.6) stable; urgency=medium
2273 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2276 * Bindings now constitute their own package, separated from the main one.
2277 Rationale: reduce our maintenance nightmare by reducing the module coupling
2278 They will soon be released on their own on gforge.
2280 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2281 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2283 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2284 * It's not quite deprecated for now because we have no replacement,
2285 but it may soon become the case.
2288 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2289 * Fortran: New user-level cache variable to store the rank of the running
2290 process. This improves performance by an order of magnitude.
2291 * C: New coccinelle script to automatically locate and modify global and
2292 local static variables.
2293 * Improved SMPI network model with a sender-side gap to account for multiple
2297 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2298 only say if a communication is finished, no matter whether it succeeded or
2299 failed. You can call MSG_comm_get_status() to know the status of a finished
2301 * New function MSG_task_dsend() to send a task and detach it. When a
2302 communication is detached, you are never notified of its success or failure
2303 and the memory is released automatically once it is finished. This function
2304 is useful when you don't care about the end nor the success of a
2306 * Change the prototypes of action replay. Sorry for inconvenience,
2307 but this is really more efficient this way (and to adapt your code,
2308 you just have to fix the initialization, that shouldn't be too long)
2309 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2310 ever managed to do anything useful with it.
2311 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2312 takes a dynar of communications. It returns immediately and gives the
2313 index of a finished communication (if any).
2314 * New example: a basic implementation of the Chord P2P algorithm.
2317 * New model for multi-core CPUs. You can now use the core attribute to
2318 precise the number of cores of a host. This is a basic model. Every
2319 process running on the host receives at most the power provided in
2320 the DTD (throughput<=power). Total throughput of process cannot exceed
2322 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2323 router linked by an up-link and a down-link (possibly asymmetrical).
2324 This kind of pattern allows to easily build last-mile model style platforms.
2325 Aggregating such patterns in a rule-based AS is thus the technique of
2326 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2327 * New model for Vivaldi routing. We transformed the Vivaldi network model
2328 into a Vivaldi routing model (based on the rule-based model). This allows to
2329 combine Vivaldi based latencies with last-mile platforms.
2332 * Added a check for NaN of IEEE754 infinite in the double entries of
2334 * Introduce a new context factory "raw", highly inspirated from the
2335 ucontext factory, but using manually crafted functions in assembly to
2336 do the work in an efficient manner.
2337 * Allow to change the used context factory at run time, not only at
2338 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2339 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2340 contexts at runtime (only with raw contexts or ucontexts).
2341 * Completely rewrote this module to allow parallel execution of user
2342 processes. Use --cfg=contexts/nthreads:N to execute user processes
2343 with N parallel threads (the default is 1, meaning no parallelism).
2344 * Allow to decide dynamically between sequential and parallel modes.
2345 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2346 processes in parallel only when their number is greater than or equal to P
2348 * Added a check for NaN of IEEE754 infinite in the double entries of
2352 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2353 won't output the state of processes when interrupted with Ctrl-C
2354 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2355 NULL-terminated array. This may solve backward compatibility issues
2356 due to the change to return type of SD_simulate. See also:
2357 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2358 * Add new macros with variable number of arguments.
2359 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2360 - in xbt/asserts.h: xbt_assert
2361 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2362 - in xbt/ex.h: THROWF and RETHROWF.
2363 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2365 * Change xbt_die() to accept a format string with arguments, just like printf.
2366 * New data structure: xbt_lib_t, like a dict but more general and with better
2370 * New configuration options
2371 Options triva/categorized and triva/uncategorized can be used to generate
2372 graph configuration files for Triva visualization tool.
2373 * Configuration option tracing/platform is renamed to tracing/categorized
2374 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2375 * New TRACE_user_link_variable interface:
2376 User provides the name of the link and the tracing variable to attach to it
2377 * the declaration of tracing categories must be done after the environment
2379 * simpler tracing interface, just one way to declare categories
2380 TRACE_category or TRACE_category_with_color, it is up to you
2381 * links in the trace file are again identified by their names
2382 * trace contains the full platform hierarchy exactly as declared using the ASes
2383 * Options tracing/msg/[task|process]:1 groups the process by hosts
2384 for both cases, tasks and processes must have names that are unique during the simulation
2385 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2386 * The experimental option tracing/msg/volume is deprecated
2387 its functionality may be reincorporated if needed
2389 The tracing generates a trace file with unordered timestamped events,
2390 because of the way the core simulator (surf) works. A script available
2391 at the tools directory (fix-paje-trace.sh) can be used to put the events
2392 in order. We have changed the tracing so it can generate ordered timestamped
2393 events in the final trace, but depending on the simulator (and how much time
2394 is simulated) that can lead to a huge memory utilization. It is deactivated
2395 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2397 Build Infrastructure
2398 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2399 - We are trying hard to keep the API stable, but it may happen that
2400 some things change (we're a research project after all, not a
2401 nuclear plant operating system). If such things should happen, you
2402 could rely on that macro to adapt.
2403 - current value: 30600 for 3.06.00, aka 3.6
2404 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2405 a number that can be compared with SIMGRID_VERSION.
2406 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2407 assertions and verbose/debug logging events are disabled at compile time.
2409 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2411 SimGrid (3.5) stable; urgency=medium
2414 * New feature to any SimGrid-based simulator: Model-Checking
2415 Check SIN#1 for more details.
2418 * New Model SMPI (three-interval linear regression for correction factors)
2419 See RR-7426, available at http://hal.inria.fr/inria-00527150
2420 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2421 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2422 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2423 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2424 * See also src/smpi/README
2429 - Tracing API changes: TRACE_start and TRACE_end should not be called
2430 by user-code. They are automatically called by simulators created
2431 with SimDAG, MSG and SMPI if the toolkit is compiled with
2432 tracing_enabled=ON. Categories declaration and utilization remain the
2433 same for MSG and SimDag.
2434 - A function was added to the tracing API to declare categories with
2436 - TRACE_category_with_color (char *category, char *color)
2437 where color must be in the following format
2438 "%f %f %f", red, green, blue
2439 and red, green, blue are float values in the interval [0, 1]
2440 - User can specify NULL as color parameter, or continue calling
2441 TRACE_category (cat)
2442 On that case, the tracing system will define random colors
2443 - The following command-line options are supported:
2444 --cfg=tracing/filename:msg.trace
2445 --cfg=tracing:1 (activate tracing, needed to use others)
2446 --cfg=tracing/platform:1 (categorized resource use)
2447 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2448 --cfg=tracing/msg/task:1 (task creation)
2449 --cfg=tracing/msg/process:1 (process creation, migration)
2450 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2451 --cfg=tracing/smpi:1 (SMPI interface tracing)
2452 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2453 - examples of examples/msg/tracing updated
2455 - DAXLoader and DOTLoader functions can generate tasks with categories
2456 - A new function to attribute a category to SD tasks:
2457 TRACE_sd_set_task_category (SD_task_t task, char *category)
2458 * Tracing the MPI interface implemented by SMPI
2459 - Collective operations are traced with states
2460 - Point-to-Point operations are traced with states/links
2461 - Tracing activated by a parameter "-trace filename" passed
2462 to smpirun during execution (considering that simgrid
2463 is compiled with tracing enabled)
2464 - To run the simulation with gdb, the simulator
2465 accepts --cfg=tracing/smpi:1 to trace SMPI
2466 - tesh files to check if smpi tracing is ok
2467 - See examples/smpi/NAS/DT-trace
2468 * GTNetS tracing re-worked
2469 - adaptation to the tracing system of GTNets to cope
2470 with modifications regarding the fullduplex mode
2471 - new tesh files to check if gtnets tracing is ok
2474 * Asynchronous communications through the functions:
2475 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2476 * New function: MSG_load_platform_script()
2477 to make possible using a lua script instead of XML files to set up platforms
2478 * New function: MSG_set_function
2479 to associate functions to processes, used when bypassing the parser
2480 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2482 Platforms: Add some more examples in examples/platforms
2483 * Grid'5000: see www.grid5000.fr
2484 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2487 * Change the XML format. This is a very important modification. SimGrid
2488 3.5 introduces a new hierarchical format based on the notion of
2489 Autonomous Systems. Compatibility with old format is ensured
2490 through the perl script provided in the install bin directory
2491 bin/simgrid_update_xml.
2492 It is now possible to build platforms with specific routing
2493 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2494 connect several platforms together. We will try to provide soon
2495 set of realistic platforms exploiting these properties (have a
2496 look at examples/platforms/ for the moment).
2497 * Take the opportunity of the XML format change to be a good XML citizen:
2498 rename link:ctn to link_ctn and similar changes (also dealt with by
2500 * Add a new routing scheme (rule-based) using regular expressions. It
2501 enables to have an extremely low memory footprint when the
2502 underlying routing is simple and can be compactly described. You
2503 need to have libpcre4-dev (perl regular expressions) installed if
2504 you want to use this routing scheme.
2505 * Revive the cluster TAG and allow to easily and efficiently (both in
2506 term of memory and speed) connect clusters together. Have a look
2507 at teshsuite/simdag/platforms/ to see how this can be done. With
2508 this tag, you can create clusters with thousands of tasks at no
2509 cost (have a look at examples/platforms/).
2510 Note: clusters are implemented as ASes, so there is no need for an
2511 enclosing AS tag if you have only one cluster in your platform.
2512 * Add new generic functions in the public interface that allows the user
2513 to call SURF 'create_resource' methods from your code (same
2514 functionality as the XML bypass mechanism but with a much lighter
2516 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2517 that uses a piecewise linear approximation to produce better
2518 results when exchanging small messages.
2519 * Add a new parameter to handle correctly full duplex link and account
2520 for interferences between uplink and downlink communications
2521 (activate with --cfg=fullduplex:1).
2524 * Rename the SD_READY (all dependencies are satisfied and task is
2525 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2526 dependencies are satisfied) state.
2527 This prevents a confusion between the notion of "ready to schedule"
2528 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2529 simulated" (SD_RUNNABLE) used by the simulation kernel.
2530 * Change the way a task is considered as ready. Instead of removing
2531 dependencies when a task is done, a counter is decreased. This way,
2532 it is always possible to reach ancestors thanks to the
2533 SD_taks_get_parents function (even after the end of the simulation.)
2534 * Change the return type of SD_Simulate from (SD_task_t*) into
2535 xbt_dynar_t. This function was in handling a dynar internally and
2536 converted it into a NULL terminated array for historical reasons.
2537 * New function SD_dotload(char*) to load a DAG described in dot
2538 format. This loader and the corresponding examples require the
2539 installation of the graphviz library.
2540 * Fix a bug in the management of tasks of size 0 in the surf network
2541 models. This problem was only visible with SIMDAG and you should
2542 thus disregard results produced with earlier versions if you
2543 relied on this feature (some tasks were blocked because of this).
2544 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2545 even though your applications were doing only point-to-point
2546 communications and sequential computations. Now you can really use any
2547 model you want (of course, if you create real parallel tasks, which are
2548 not implemented in most models beside ptaskL07, this will abort).
2549 * Add an example that schedules a DAX on an heterogeneous platform
2550 using a Min-Min strategy.
2551 * New function SD_workstation_get_current_task() that returns the kind
2552 of task currently running on a workstation in the sequential access
2554 * Raise some warnings when unexecuted tasks remains at the end of the
2555 simulation. This is usually caused by cycles in the DAG.
2558 * New function: SIMIX_process_set_function() called by MSG_set_function
2559 * Change the underlying waiting queue in semaphores so that a process
2560 can wait on several of them simultaneously (as in waitany).
2561 * Fix the way to handle tokens in semaphores so that all access patterns
2562 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2563 * kill the dirty pimple SIMIX_message_sizes_output()
2564 Please use (proper) visualization instead
2567 * New data container: setset (set of sets of elements)
2568 * New module: mmalloc (mapped malloc, allowing to have several
2569 independent segments of malloc)
2570 * New function: xbt_dict_cursor_set_data()
2571 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2572 * New function: xbt_dynar_is_empty()
2573 * New function: xbt_fifo_get_last_item()
2574 * Fix xbt_dynar_shrink(): use the right element size.
2575 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2576 uninitialized areas during expand.
2577 * Fix semaphores: previous implementation was severely broken.
2578 * Use library init/fini functions for our initialization.
2579 - you can use logs and other feature as soon as you want in your
2580 code (even before the xbt_init / MSG_init)
2581 - xbt_exit is now a no-op and produce a warning when used.
2584 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2585 benefit from the latest improvement to the simulation kernel.
2586 * Kill measurement sockets for now. If you rely on them, sorry. This
2587 release is not for you. This feature will be reintroduced in the
2588 future, but we cannot delay the release any further.
2589 * New function: gras_msgtype_get_name().
2590 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2591 * Fix (at last) the pmm example: it should not randomly fail anymore.
2593 Build chain: bug fixes and overall polishing
2594 * Cmake is now stable enough. Hence, we killed the autotools.
2595 * Port to windows ( TM :)
2596 * Fix the 'make install' target.
2597 No need to use 'make install-simgrid' anymore
2598 * Introduce a 'make dist' target compiling a *source* archive
2599 'make package' compiles a binary archive
2600 * Compile java files only on need
2601 * Add --cd and --setenv command line options to tesh
2602 * Out of source builds are not fully supported yet, but we are close
2603 * Enable supernovae and optimization flags by default for our users
2606 * Add layer to set up environment directly from lua, without XML.
2607 * The effect of gras_stub_generator can be achieved through
2608 lua too (check examples/gras/console/ping_generator.lua)
2610 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2612 SimGrid (3.4.1) stable; urgency=low
2614 The "Polishing easter eggs is probably a good idea" release.
2615 This is a bug fixes release only.
2619 * Fix a bug preventing the tasks from begin garbage collected.
2622 * Fix a bug occurring when a host involved in a communication fails.
2623 This was not detected properly by the other peer involved in the
2624 communication. Now, it's reported as a network error.
2627 * Warn the user about loop dependencies in data flow of DAX files
2628 * Obey the control-flow dependencies of DAX files
2631 * Add option "enable_smpi" allowing to not compile SMPI.
2632 Probably useful for the (Mac) users experiencing a build error here
2633 * Improve the detection of lua5.1 and ruby1.8
2635 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2637 SimGrid (3.4) stable; urgency=low
2639 The "Easter in Cargese" release. Also known as (major changes):
2641 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2642 ~> bindings were greatly improved
2643 ~> new tracing infrastructure for better visualization introduced
2645 * the "Welcome to configury modernity" release.
2646 ~> we switched from autotools to cmake, and improved our cdash
2649 A more detailled list of changes follow (full detail in svn log).
2651 Java Bindings: Various Cleanups
2652 * (install java-gcj-compat-dev on debian-like to use them)
2653 * Remove put/get: no need to export deprecated interface in Java
2654 Use send/receive instead.
2655 * Cleanup the examples and add a README per directory
2656 * Remove example autoDestination (that's the only way to go now)
2657 * Remove example explicitDestination (was a plain copy of basic)
2658 * Make JniException a runtime exception, so that there is no need to
2659 declare the fact that you may encounter such a beast. I guess that
2660 nobody will ever want to survive such error.
2661 * Create specific errors for each MSG case of failure:
2662 host failure, transfer failure, timeout, task cancelled
2663 * Cleanup the exceptions that may get thrown by each function
2664 * Other internal cleanups in Java bindings. Performance still bad :/
2665 Ruby and Lua Bindings: create them
2666 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2667 * That's new and great, you should try them out.
2668 Same functionalities than Java bindings, only even less polished
2670 * Kill the useless "rate" argument of SD_task_get_execution_time()
2671 Everyone used to provide -1 as a value, it was not used, and the
2672 semantic of a possible use wasn't even clear.
2673 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2674 either as comm costs or compute costs to mean that there is no
2675 such thing for that specific task.
2676 * Add a SD_task_set_name() function
2677 * Fix SD_task_unschedule() on typed tasks
2678 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2679 * In DAX loader, accept useless 'level' attributes to <job> since
2680 LIGO DAGs have them (seem to be to ease graphical representation).
2682 * Add an example masterslave_mailbox.c using send/receive and not
2683 the deprecated put/get interface.
2684 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2685 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2687 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2688 (declare MSG_USE_DEPRECATED to still have the old name)
2689 * Add a MSG_task_set_data() function
2690 * About trace replay (see examples/msg/actions):
2692 - Allow to work with splitted trace files for each process
2693 Give the specific trace file as argument of each process,
2694 and call MSG_action_trace_run(NULL)
2695 You can still have one merged file for all processes.
2696 - Fix implementation of collective operations
2697 * Allow task_execute() on 0-sized tasks (closes #10063)
2699 * This is the first release of SimGrid where SMPI is not considered
2700 beta anymore (even if some corners should still be improved)
2701 * Port over the new SIMIX_network submodule (internal refactoring)
2702 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2703 * Implement more missing elements of the standard:
2705 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2707 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2708 MPI_Scan() MPI_Get_processor_name()
2709 - Added implementation of missing case for Alltoall (warning: it's
2710 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2711 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2712 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2713 SMPI_MPI_Allgatherv()
2714 * Bug fixes include:
2715 - MPI_Waitsome() was broken
2716 - Allow relative includes in smpicc
2717 - Command line cfg argument 'reference_speed' was ignored...
2718 - Some functions did not properly lead to auto-benching of user code
2719 - smpicc passes -O2 by default (just like openmpi one)
2721 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2722 * Bug fixes about timeouts during communications
2723 * add SIMIX_message_sizes_output() as a pimple to write to file the
2724 amount of messages per size. Use gnuplot to get histogram.
2725 Pimple because that's the only user-visible function of simix,
2726 defined directly in xbt.h (irk, sorry)
2728 - Add a SIMIX_sem_get_capacity() function
2729 - Fix interactions with processe resume/suspende
2730 - release_forever() was stupidly broken
2731 - Fix SIMIX_display_process_status() for processes in a semaphore
2732 - Make SIMIX_sem_block_onto() user-visible
2733 * Refactoring context stuff:
2734 - Use pseudo-OOP for better modularity
2735 - reimplement SIMIX_process_kill() without process_schedule() so
2736 that the latter can take as invariant that it is called from
2738 - Merge context_start into context_new for sake of simplicity
2740 * Add a Vivaldi network model, coded live during SUD'10 ;)
2741 * Rename configuration variables to start a hierarchy:
2742 o cpu_model -> cpu/model
2743 o network_model -> network/model
2744 o workstation_model -> workstation/model
2745 * New configuration variables:
2746 o network/bandwidth_factor: correction to bandwidth
2747 o network/latency_factor: correction to latency
2748 o netwotk/weight_S: correction to the weight of competing streams
2749 * Add a long description to the models, that users can see with such
2750 argument on the command line: --cfg=cpu/model:help
2751 * --help-models display the long description of all known models
2753 * config: add the ability to set a default value after registration
2754 Does not override any previously set value (e.g. from cmd line)
2755 * dict: allow to have integer key and data.
2756 When so, you need to use the following functions
2757 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2758 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2759 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2760 In contrary to regular dicts, the key is not malloced before copy.
2761 Mixing scalar and regular elements in the same dict is not tested
2763 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2764 Tracing for Visualization:
2765 * SimGrid is now instrumented in order to generate a trace file for
2766 visualization analysis: to use it, need to compile SimGrid with the
2767 "tracing" option enabled, and instrument the program using SimGrid with
2768 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2769 (among other functions).
2770 * The instrumentation only traces the platform utilization for now
2771 * Documentation to use the tracing functions and how to analyze the
2772 traces with the Triva tool is written.
2773 * More information about: SimGrid FAQ (in the section Tracing Simulations
2776 * We moved to cmake as default build system. Autotools support will
2777 be dropped soon. Check the FAQ for more info about how to use it.
2778 * Greatly improved our cdash/ctest interactions
2779 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2780 * Added memory checking tests with valgrind; lot of memleak fixing.
2781 This may be the first release of simgrid with so few memory issues
2782 * Added code coverage tests.
2783 Our coverage is still improvable, but at least we see it on cdash.
2785 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2787 SimGrid (3.3.4) stable; urgency=low
2789 The "Desktop Grid needs love too" release (also called Xmas release).
2791 Models improvements:
2792 * Major speedup in the maxmin system solving by using lazy evaluation
2793 Instead of solving completely the maxmin system at each iteration,
2794 only invalidate (and recompute) the modified parts.
2795 This new feature is enabled in default models but you can try to
2796 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2797 * Cas01 IMproved as default CPU model
2798 This CPU model is the same Cas01 model, but it uses the
2799 maxmin-selective-update flag and a heap structure to manage
2800 actions on SURF kernel.
2801 It reduces the complexity to find the next action to finish and,
2802 consequently, it's faster than the old Cas01.
2803 This is the new default CPU model (Cas01).
2804 * Rename the old Cas01 model to Cas01_fullupdate
2805 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2806 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2807 * CpuTI (CPU Trace Integration)
2808 A new CPU model whose objective is simulate faster when using
2809 availability trace files.
2810 Instead of using a full featured, over engineered maxmin system for
2811 CPU modeling, this model does the pre-integration of traces files
2812 to calculate the amount of CPU power available, and so, executes
2813 faster than the old CPU models.
2814 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2815 * Use LV08 as default network model since it gives better accuracy
2816 for small messages and shouldn't change things for big ones.
2817 Use --cfg=network_model:CM02 to get the previous behavior.
2820 ******************************************
2821 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2822 ******************************************
2823 * The new CPU model may changes simulations!
2824 The point is that events occurring at the exact same timestamp
2825 are not scheduled in the same order with the old and new
2826 version. This may be enough to completely change the execution
2827 of simulations in some cases.
2828 * The new network model will change simulations!
2829 This new model is more realistic than the previous one, so you
2830 should consider redoing your old experiments with this model.
2831 Sorry for the inconvenience.
2834 * Introduce the supernovae compilation mode
2835 When compiled that way, the whole SimGrid (or almost) is put in a
2836 single compilation unit and compiled in one shoot.
2837 This is to help gcc which has difficulties to inline stuff from one
2839 The speedup seem to be above 15%, althrough more tests are needed on
2840 amd64 to confirm that gain.
2843 * Port of MSG's mailbox on top of SIMIX network
2844 The put/get mechanism was greatly simplified on the way.
2847 * New SIMIX network module. Provides:
2848 - Mailbox: rendez-vous mechanism to find with who you want to speak
2849 - Synchronous send/recv: easier and hopefully faster since the
2850 logic is handled in the maestro process directly now
2851 - Asynchronous send/recv: you dreamt of it? It's here now
2852 Too bad that nobody cared enough to propagate the change to MSG.
2853 * Add semaphores as SIMIX synchronization mechanism.
2856 * new function SD_daxload(char*) to load a DAX file
2857 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2858 * Introduce typed tasks. Specify its kind and cost at creation.
2859 At scheduling, just give where it should be placed, and the cost
2860 for each involved resource is automatically computed.
2861 Existing constructors so far (more to come of course):
2862 - SD_task_create_comm_e2e() for end-to-end communication
2863 - SD_task_create_comp_seq() for sequential computation
2864 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2865 * new function SD_task_dump() for debugging display
2866 * new function SD_task_dotty(task,FILE*) writing to file the info
2867 about the task in dotty format
2868 * SD_task_dependency_exists() can now cope with having one of its
2869 arguments NULL. If so, it tests whether the other argument has any
2871 * Add getters on list of preceding/following tasks:
2872 SD_task_get_parents(task) and SD_task_get_children(task)
2873 * Add getters on amount of workstations and list:
2874 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2875 * Add getter on task kind: SD_task_get_kind(task)
2876 * Update the start_time and finish_time of tasks on completion/failure
2877 * Bugfix: Remove task from state swags when destroyed
2880 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2883 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2884 dynar with one separate thread per value of the dynar.
2885 * Change the prototype of xbt_thread_create(), sorry.
2886 Added a boolean parameter indicating whether we want to join this
2887 thread (used in SG only for now)
2888 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2891 * GTNetS wrappers should now be usable again (and betterly tested too)
2892 * Fix a major regression from 3.2 where the timeout provided to
2893 MSG_task_put_with_timeout() was used as absolute time before which
2894 the comm should be done.
2895 * Start to fix the <cluster> tag.
2896 - Internal links should be good now (beside of the loopback, which
2897 use the private link instead)
2898 - paths to the external world is still rather broken
2899 - the <route:multi> tag is just broken. Actually that's brain-dead.
2900 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2901 to make it less stupid
2902 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2903 * Fix a source-level compatibility glitch from 3.2: after defining
2904 MSG_USE_DEPRECATED, you can use the old name
2905 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2906 * Allow to compile from the SVN with automake 1.11
2907 * Fix some problems when using the "start_time" tag in deployment XMLs.
2908 * Fix #8569: XBT/synchro.h has redundant declarations
2909 * Fix #8563: MSG return values and exceptions
2910 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2911 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2912 Thanks to Robson Peixoto.
2913 * Fix a vicious bug in dictionaries inducing that some elements were
2914 not freed on xbt_dict_free()
2916 Portability report of this version:
2917 * Main portability targets:
2918 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2919 on (amd64/i386/ia64)
2920 - mac leopard on i386
2921 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2922 but nothing critical.
2923 * Other platforms: windows, AIX and others were not tested for this release
2925 Timing report of this version:
2926 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2927 scenario parameters). From 8h to a few seconds in desktop grid settings.
2928 * Supernovae brings about 25% speedup on i386.
2930 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2932 SimGrid (3.3.3) stable; urgency=low
2934 The "Need for Speed" release.
2936 The timings done to validate the 3.3.2 were faulty.
2937 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2939 The problem was a conversion from a manually handled vector to
2940 xbt_dynar_t on the critical path.
2941 xbt_dynar_foreach calls functions, inducing stack management crap.
2943 We inlined these functions and xbt_dynar_foreach is now breath taking.
2944 We also inlined xbt_swag_belong on the way.
2946 Here are some approximate speedup measurements (on master/slaves
2947 simulations lasting between 10s and 20s each):
2948 3.3.1 -> 3.3.2: about same performance
2949 3.3.2 -> 3.3.3: 40% speedup
2950 3.3.1 -> 3.3.3: 40% speedup
2951 3.3.1 with inline patch -> 3.3.3: 30% speedup
2953 Our reading is that the refactoring which occurred in 3.3.2 made us
2954 suffer much more from the xbt_dynar_foreach low performance, but
2955 once we solved this, this refactoring proved to be very performance
2956 effective. From the 40% speedup, somehow, 10% are due to the
2957 inlining and 30% to the refactoring.
2959 That's a pitty that gcc cannot inline functions placed in other files
2960 alone. We have to choose between:
2961 - break the encapsulation (by putting private data structures and
2962 accessors in headers files to help gcc)
2963 - live with low performance
2964 - switch to a decent compiler such as icc (not quite possible).
2966 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2968 SimGrid (3.3.2) stable; urgency=low
2970 The "Simplicity does not preceed complexity, but follows it" release.
2972 The main contributors of this release were (lexical order):
2973 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2976 * Extract the routing logic into its own object.
2977 (was dupplicated in network.c and workstation_LV07.c;
2978 Allows to implement other ways of storing that info)
2979 => kill now useless network_card concept
2980 - Use dynar to represent routes (instead of void** + int*)
2981 - kill link_set (use surf_network_model->resource_set instead)
2982 - Add a command-line option to choose the routing schema to use
2983 - Add three new models:
2984 * Floyd (shortest path computed at initialization)
2985 * Dijikstra (shortest path recomputed all the time)
2986 * Cached Dijikstra (shortest path computed on need)
2987 All these models where contributed by Silas De Munck, and are
2988 described in his ICCS09 paper.
2990 * Simplify model declaration
2991 (less redirections, less function to write when defining a model)
2992 - Factorize stuff between models:
2995 surf_model_resource_set(model)
2996 surf_model_resource_by_name(model, name)
2997 - Unify the types of models in s_surf_model_t (using an union)
2998 - Embeed fields of common_public directly into s_surf_model_t
2999 - Rename model methods:
3000 action_free ~> action_unref
3001 action_change_state ~> action_state_set
3002 action_get_state ~> action_state_get
3003 - Change model methods into functions:
3004 (model)->common_public->action_use ~> surf_action_ref
3006 * Implement a generic resource; use it as ancestor to specific ones
3007 (allows to kill duplicated code in models)
3008 Drawback: timer command don't need no name nor properties;
3009 workstation_CLM03 don't need no properties
3010 (but I guess we can live with those few bytes wasted)
3012 * Improve the action object model
3013 - implement a constructor avoiding dupplicated code about field
3014 initialization in generic_action part.
3016 * Kill the SDP model: it has an external dependency, is deprecated
3017 in flavor of modern lmm models, and didn't compile since a while
3020 * Relocation of the context module from XBT to SIMIX.
3021 (the context were decoupled from the simix processes, duplicating a lot of code)
3022 => a lot of code was factorized
3023 - less overhead is introduced during scheduling
3024 - simpler API for the context factory
3025 - the logic for process creation,destruction and manipulation was simplified
3026 * Simplification of the s_smx_process_t data structure.
3027 => accesing the simix level data associated to a process is faster now,
3028 and the code is a lot more readable.
3031 * Implement some more MPI primitives:
3032 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3033 -implementation: Bcast: flat or 2-ary tree (default),
3034 Barrier: 4-ary tree,
3036 Allreduce: Reduce then Bcast
3037 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3038 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3039 Alltoallv: flat tree, like ompi
3041 * Add support for optimized collectives (Bcast is now binomial by default)
3042 * Port smpirun and smpicc to OS X
3045 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3046 Shout out if you used it.
3049 * Display the list of still queued messages in SG mode when existing
3053 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3054 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3055 * Add xbt_str_from_file(FILE*)
3056 * Add xbt_dict_get_key achieving a linear reverse search
3057 * Remove the context module
3059 Portability report of this version:
3060 * Main portability targets:
3061 - Linux(debian)/x86/context
3062 - Linux(debian)/x86/pthreads
3063 - Linux(debian)/amd64/context
3064 - Linux(debian)/amd64/pthreads
3065 On these, we still have the eratic breakages of gras/pmm and
3066 amok/saturate_sg reported in previous version. We still think
3067 that the tests are the cause of the fault, not the tested code.
3069 - Mac OSX Leopard/x86/context
3070 Still false negative in tesh autotesting.
3071 Smpi still fails, but this time because readlink does not accept -f
3072 Everything seems to work properly beside of that.
3075 - AIX version 5.3 (only tested contexts this time)
3076 Smpi still fails there because mktemp is not installed.
3077 Everything seems to work properly beside of that.
3079 I managed to compile it for the first time, but several breakages.
3080 Won't delay the release for this exotic platform.
3082 * Windows: it's still lagging behind. If you want to help, please
3085 Timing report of this version:
3086 This version seem to be more than 5% faster than 3.3.1 (on linux
3087 64bits with contextes). The gain is less than expected, we are
3088 investigating this for next release.
3090 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3092 SimGrid (3.3.1) stable; urgency=low
3095 * Implement a --cfg-help to show existing configuration variables
3096 * Build chain do not require doxygen in maintainer mode
3099 * fix a bug on struct sizeof computation, which prevented the
3100 exchange of arrays of structs in some conditions
3101 - added a regression test about this in datadesc_usage
3102 * Allow the exchange of 0-long dynamic vectors.
3103 - for that, use -1 as indicator of dynamic size instead of 0
3104 - This implied to change any size from unsigned long to long,
3105 reducing a bit communication abilities, but I guess that with
3106 64bits being quite common, this is more than enough.
3107 - This also induce a protocol change, thus bumping network protocol
3108 version from 0 to 1 (if we have external users, we have to get
3109 clean on that point too ;)
3110 - added two regression tests about this in datadesc_usage
3111 * Be more verbose when propagating local exceptions
3112 This helps debugging.
3113 * Display the status of simulated processes when receiving SIGINT in
3117 * Allow to control the simulation from a trace file.
3118 New functions MSG_action_register() and MSG_action_trace_run()
3119 The first one allows to associate a function execution to each
3120 kind of action while the second one parses a trace file and
3121 triggers the corresponding actions within the system.
3122 For now, only a toy example is provided in examples/msg/actions
3123 * Add an example of process migration in examples/msg/migration
3124 * Fix a bug in task exchange which broke MSG_task_get_sender()
3125 Add a teshsuite regression test for that.
3126 [Bug: if MSG_task_get_sender() is called after sender exit,
3128 * Fix a bug which prevented suspend/resume to work properly
3129 * Display the status of simulated processes when receiving SIGINT
3130 This fixes a regression of v3.3. due to the introduction of SIMIX
3131 * Bug fixing in failure management:
3132 - trace could not start by a failure at time 0
3133 - failure during communications were not working
3136 * Add SIMIX_process_set_name() to change the name of the current
3137 process in the log messages.
3138 * Store smx_hosts in a dict since we only retrieve them by name
3139 * Move the configuration infrastructure to surf
3142 * Move the configuration infrastructure to surf
3145 * Massive internal cleanups:
3146 - Store internal structures on processes instead of hosts (allows
3147 to have more than one process per host, in addition of being more
3149 - Cleanup the initialization/finalization process
3150 - Kill a whole bunch of unneeded synchronization:
3151 processes run in exclusive manner within the simulator
3152 - Move queues from global tables to process data fields
3154 - now accept -platform and -hostfile arguments
3155 - Pass the right rank value to processes according to the hostfile
3156 * Compile the examples by default, and use them as regression tests
3157 * Implement MPI_Wtime()
3158 * Change the reference speed to a command line option
3161 * TCP_gamma can now be specified as command line option using
3162 --cfg=TCP_gamma:10000000.0
3163 * Change the --surf-path cmd line option into --cfg=path:
3166 * Also include strbuff from xbt.h public header
3167 * xbt_ex_display(): do not free the exception after displaying
3168 This allows to do more with the given exception afterward.
3169 Users should call xbt_ex_free() themselves.
3173 Portability report of this version:
3174 * Main portability targets:
3175 - Linux(debian)/x86/context
3176 - Linux(debian)/x86/pthreads
3177 - Linux(debian)/amd64/context
3178 - Linux(debian)/amd64/pthreads
3179 These targets fail about 1/10 of times on gras/pmm, but we believe
3180 that this is because of the test, not because of simgrid.
3181 amok/saturate_sg fails even more rarely, and the test may not be
3184 - Mac OSX Leopard/x86/context
3185 The test suite still spits tons of errors because some obscure
3186 force prevents us from removing the temporary directories
3187 arguing that they still contain some metadata I've never heard of.
3188 Smpi fails because seq is not installed.
3189 Everything seems to work properly beside of that.
3192 - AIX version 5.3 (both contexts and pthread)
3193 Smpi still fails there because mktemp is not installed.
3194 XML inclusions seems rosty on AIX.
3196 * Windows: it's still lagging behind. If you want to help, please
3199 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3201 SimGrid (3.3) stable; urgency=high
3205 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3206 [Malek Cherier & Mt]
3208 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3211 Until now, GRAS were using MSG as an interface to SURF. It was
3212 quite difficult because both interface have several differences
3213 (MSG channels vs GRAS sockets were the most notable point).
3215 This also opens the gate to SMPI (which should occur soon) and speed
3216 up simulations by to 40% (even if it were not the main goal).
3218 **************************************
3219 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3220 **************************************
3221 The point is that events occurring at the exact same timestamp are
3222 not scheduled in the same order with the old and new version. This
3223 may be enough to completely change the execution of simulations in
3224 some cases. Sorry for the inconvenience.
3226 * Cleanup and upgrade the XML format to push further scalability
3227 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3229 * Improve the testing infrastructure with tesh. Now a very large part of
3230 the code is tested not only by being run but also by checking that the
3231 output match an expected output [Mt].
3233 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3234 is really less memory-demanding, which should allow you to use
3235 larger files in SimGrid [AL].
3237 * Inform valgrind about our contextes, so that it becomes usable
3238 with the default (and more effecient) version of SimGrid
3239 [contributed by Sékou Diakite, many thanks]
3242 * Introduce a listener thread in charge of receiving incoming
3243 messages from the network. It allows to overlap communication and
3244 computation but most notably, it removes some stupid deadlocks due
3245 to the fact that so far, a process could not send and receive at
3246 the same time. This made most non trivial communication schema
3248 * Convert the PIDs from long int to int to match the MSG ones (and
3249 linux ones too) [Mt]
3250 * New function: gras_agent_spawn() to launch a new process on
3251 current host. Only working in simulation for now. [Mt]
3252 * New function: gras_os_hostport() returning a constant form (ie,
3253 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3256 * Make the backtrace of exceptions more human readable [Mt]
3257 * New module: xbt/str [Mt]
3258 a ton of string utility functions (split, join, printf to a newly
3259 allocated buffer, trim, etc)
3260 * New module: xbt/hash [Mt]
3261 SHA1 hashing algorithm (more to come if needed)
3262 * New module: xbt/synchro [Mt]
3263 synchronization tools (mutex and conditions) working the same way
3264 in simulation and in real life (mainly useful for GRAS, but not
3266 * New module: xbt/queue [Mt]
3267 classical producer/consumer synchronization scheme
3268 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3269 (using the classical functions will get serialized) [Mt]
3270 * Make dictionary internal table dynamic. No need to specify its size
3271 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3273 * Make sure the log channels are organized as a tree under windows
3274 (because of ANSI C compatibility issue, any channel were child of
3275 root directly) [Mt].
3278 * Cleaned many thing in surf and fixed a few bugs [AL].
3279 * Add a nice command line configuration mechanism to compose models [AL].
3280 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3281 the previous one (KCCFLN05). It relies on something that looks like
3282 a max-min sharing mechanism but cannot be written as such. A new solver
3283 was thus designed [AL].
3284 * Add a new solver to lmm. Based on Lagrange optimization and
3285 gradient-based descent, it enables to efficiently maximise systems s.a
3287 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3289 This solver enables to propose two new network models for TCP Reno and
3290 TCP Vegas based on Low's work. These models still need to be fully
3291 tested though [Pedro Velho].
3294 * Bug fix in SD_simulate. Now the time bound given as argument is
3296 * Use the new parallel task model (ptask_L07) as default.
3297 * Use the SURF command line configuration mechanism.
3298 * 0-size tasks (for synchronization) should now work.
3300 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3302 SimGrid (3.2) stable; urgency=high
3306 We still experience issues on this platform, but we believe that at
3307 least MSG is usable.
3309 GRAS API BREAKAGE (for simplification purpose, sorry):
3310 * the gras_msgtype_by_name is not used anymore. Instead of
3311 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3312 you can write (and must)
3313 gras_msg_send(toserver, "request", &request);
3314 - If you still want to pass a gras_msgtype_t to the function (to cache
3315 the type and avoid the lookup time), use the gras_msg_send_() variant.
3316 - Impacted functions:
3317 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3318 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3319 * The callbacks are now expected to return 0 when everything went well
3320 (just like the main() function)
3322 GRAS new features and improvements:
3323 * New module mechanism where user code can use per process globals [Mt]
3324 This is similar to gras_userdata_*() functions, but for libraries. It
3325 factorize some code developped over and over in the examples and AMOK.
3326 It has still to be documented and used (only amok/peermanagement is
3328 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3329 loss when they were small enough to fit into the buffer and sent quickly
3330 enough so that they can all get received in one shoot.
3331 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3332 if not found. Use the *_or_null() variant for the old semantic.
3333 * In gras_msg_handle, do not discard messages without callback.
3334 They are probably messages to be explicitly awaited later (ie, proofs of
3335 mis-synchronization in userland since they are sent before being awaited)
3336 No big deal usually.
3337 * gras_socket_meas_send/recv: semantic changed!
3338 The numerical arguments used to be (1) the total amount of data to send
3339 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3340 messages. This was need for the fool willing to send more than MAXINT
3341 bytes on quite fat pipes.
3344 * Do really rename the hostmanagement module to peermanagement. [Mt]
3345 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3346 but this is rather new and this was documented in the module
3347 documentation (poor excuses, I admit)
3348 * Bandwidth measurement semantic changed! This follows the changes to
3349 gras_socket_meas_send/recv explained above.
3352 * A sequential mode has been added to the workstations. When a workstation
3353 is in sequential mode, it can execute only one task, and the other tasks
3354 are waiting in a FIFO. [Christophe Thiery]
3357 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3358 model for SIMDAG. [Christophe Thiery]
3359 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3360 equal to 0, which caused some bad behaviors in
3361 saturated_constraint_set_update. I now use a threshold mechanism like in
3365 * When running manually src/testall, you select specific units [Mt]
3366 testall is the result of our cunit mechanism, and should replace all
3367 the scripty thingy around since bash don't run easily on billware.
3369 * A mallocator system has been added. [Christophe Thiery]
3370 Mallocators allow you to recycle your unused objects instead of freeing them
3371 and allocating new ones.
3373 Documentation update:
3374 * FAQ reworking + New FAQs:
3375 - "Valgrind spits tons of errors!" [Mt]
3376 - "How to repport bugs" [Mt]
3377 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3378 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3379 - Communication time measurement within MSG [AL]
3380 - I experience weird communication times when I change the latency [AL]
3381 * GRAS tutorial [Mt]
3383 - an introduction to the framework and to the used communication model
3384 - an initiatic tour introducing the most proheminent features:
3386 . Lesson 0: Installing GRAS
3387 . Lesson 1: Setting up your own project
3388 o Part 2: Message passing
3389 . Lesson 2: Exchanging simple messages
3390 . Lesson 3: Passing arguments to the processes (in SG)
3391 . Lesson 4: Attaching callbacks to messages
3392 . Lesson 5: Using globals in processes
3393 . Lesson 6: Logging information properly
3394 . Lesson 7: Using internal timers
3395 . Lesson 8: Handling errors through exceptions
3396 . Lesson 9: Exchanging simple data
3397 . Lesson 10: Remote Procedure Calling (RPC)
3398 . Lesson 11: Explicitly waiting for messages
3399 . Recapping of message passing features in GRAS
3400 - A HOWTO section containing:
3401 o HOWTO design a GRAS application
3402 More are due, of course. They will come latter. In the meanwhile, you can
3403 check the examples which are still here.
3405 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3407 SimGrid (3.1) stable; urgency=high
3411 There was a stack corruption somewhere, visible only when optimizing
3412 with these versions. [Vince]
3415 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3416 module that enabled to play with Directed Acyclic Graphs. It is built
3417 directly on top of SURF and provides an API rather close to the old
3418 SG. Some old codes using SG are currently under rewrite to check that
3419 all needful functions are provided. [Christophe Thiery]
3422 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3423 extension of the classical CLM03 model that gracefully handles
3424 failures. This is now the default model for MSG and GRAS. It doesn't
3425 handle parallel tasks yet though. [AL]
3426 * Bug fix: Weights were not correctly set in the network part.
3427 WARNING: This may have resulted in incorrect results with simulations
3428 where there are more than one flow on a given link. [AL]
3431 * After a (long ?) discussion on simgrid-devel, we have decided that the
3432 convention we had on units was stupid. That is why it has been decided
3433 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3434 WARNING: This means that all previous platform files will not work as
3435 such with this version! A warning is issued to ask users to update
3437 A conversion script can be found in the contrib module of the CVS, under
3438 the name contrib/platform_generation/surfxml_update.pl [MQ]
3441 * Bug fix: Processes were started in reverse order, wrt deployment file.
3442 WARNING: if your code relies on this bug, please fix it. [AL]
3443 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3444 being executed on two different locations. [AL]
3445 * Bug fix: Failures are now better supported thanks to Derrick's tests
3446 (there was many failure situations I hadn't thought of and that weren't
3447 correctly handled). [AL]
3448 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3452 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3453 Exception occurring on server-side are propagated back to client (!).
3455 API CHANGE: the callback changed their prototype. Change:
3456 int my_handler(gras_socket_t expeditor, void *payload_data) {
3458 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3459 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3461 * New! function: gras_msg_handleall to deal with all messages arriving
3462 within a given period.
3463 * New! function: gras_socket_server_range to get a server socket in a
3464 range of port numbers (ease to avoid port number conflicts) [MQ]
3465 * New! gras processes display their backtrace when they get a SIGUSR1
3466 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3467 Sweet to debug RL processes [MQ]
3471 - Do not force experiment sizes to be expressed in kb, or it becomes
3472 impossible to measure the latency this way (needs one byte-long tests)
3473 WARNING: this changes the amok_bw_* function semantic. [MQ]
3474 - Implements the link saturation stuff. [MQ]
3475 * Peer management module:
3476 New! module factorizing code that we wrote over and over [MQ].
3479 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3480 - Test units are placed directly into the library code, they get extracted
3481 automatically and placed into the src/testall binary.
3482 - Convert most of the XBT tests to this system.
3483 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3484 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3485 live in the same C code. This required to change a little bit the API
3486 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3487 * New module: sparse graph structure with basic algorithms (this is work
3488 in progress and the API is not considered to be frozen yet). [AL]
3489 * Display more information on backtraces: source line & function names are
3490 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3491 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3492 * Reworked a little bit some #include statements to load only required
3493 headers. Some user code that relied on SimGrid to include stdlib or
3494 stdio may need to include it by themselves. [AL]
3495 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3497 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3498 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3499 but I guess that almost nobody used those parts. [MQ]
3501 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3503 SimGrid (3.0.1) stable; urgency=low
3506 * Unfortunately, I had missed 5 misnamed functions:
3507 xbt_fifo_item_t xbt_fifo_newitem(void);
3508 void xbt_fifo_freeitem(xbt_fifo_item_t);
3509 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3510 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3511 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3512 They're now deprecated. Please use their new versions:
3513 xbt_fifo_item_t xbt_fifo_new_item(void);
3514 void xbt_fifo_free_item(xbt_fifo_item_t);
3515 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3516 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3517 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3519 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3520 * Documentation: xbt_log module unmercifully reworked [MQ]
3521 * Bugfix: there was a problem with the ending of contexts with
3522 the pthread backend. It caused some weird deadlock or behavior
3523 depending on the pthread implementation. [AL]
3524 * Bugfix: get the exceptions raised in the simulator repport where
3525 and why they come from when they are not caught in time [AL, MQ]
3528 * Bugfix: Do repport the error when two non-connected hosts try to
3529 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3532 * Add additionnal checkings on communications. Assert that two
3533 communicating hosts are connected by a set of links... [AL]
3536 * Add additionnal checkings on channel values in communication [AL]
3537 * New: MSG_task_get_source to see on which host a task was generated [HC]
3538 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3539 the number of tasks waiting to be received on channel and sent
3541 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3542 waits for the first task coming from a given host.. [AL]
3544 GRAS new functionnalities: [MQ]
3545 * Enhance the parsing macro to allow the size of multidimentional objects
3546 to be given thru annotations.
3547 * New example (and documentation): Matrix Multiplication a la RPC
3548 (as when I was young!) and fix a bunch of bugs found on the way.
3550 GRAS performance improvements: [MQ]
3552 * Reduce the amount of cbps creation/destruction by making it static to
3553 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3556 * Change libdata to a set so that we can search for stuff by ID (and thus
3557 reduce the insane amount of dict lookups)
3560 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3561 RL communication on top of SG since it's so uneffective.
3562 It may also be used for inter-thread communication in RL, one day.
3563 * Use gras_datadesc_copy() to exchange messages on top of SG
3565 - improve message exchange performance on top of SG
3566 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3567 * Don't exchange on the network the size of the used part of buffer,
3568 instead, specify the possible buffer size to read().
3570 - reduces the amount of read/write calls (one pair per exchange)
3571 - reduces the amount of exchanged data (the size)
3572 - allows to retrieve all arrived data on receiver side, if we don't need
3573 it right now (subsequent read will peek the buffer)
3574 - allows the receiver to proceed with the beginning of the stream before
3575 everything is arrived
3576 - make it possible to build an iov transport (using readv/writev)
3578 - take care of the data with non-stable storage (like stacked data),
3580 * If possible, TCP send uses vector I/O (when writev() is here)
3581 - Don't use it for receive since we send data sizes and data on the
3582 same stream, so we wouldn't be able to chain large amount of chunks
3583 before having to flush the stuff to read the size.
3584 * Rework the transport plugin mechanism to simplify it and reduce the
3585 amount of pointer dereferencement when searching for the right function
3588 * I guess that now, we do almost as few system calls as possible while
3589 doing as few data copy as possible.
3591 To improve it further, we could try to send all the sizes first and then
3592 all the data (to use iov on receiving size), but it's only a partial
3593 solution: when you have 2 dimensional data, the sizes of the second
3594 dimension is data of the first dimension, so you need 3 streams.
3596 I'm not sure the potential performance gains justify the coding burden.
3598 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3600 SimGrid (3.00) stable; urgency=high
3603 * New! Give the possibility to hijack the surf parser and thus bypass
3604 MSG_create_environment and MSG_launch_application. Have a look at
3605 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3607 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3609 SimGrid (2.96) unstable; urgency=low
3614 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3615 This deprecates the xbt_error_t mechanisms.
3616 It modifies (simplifies) all XBT and GRAS API.
3617 MSG API keeps unchanged (exceptions raised by XBT are caught from
3618 within MSG and masked with existing error handling facilities)
3621 * New! Add a FATPIPE model. [AL]
3622 * New! Add a parallel task model. [AL]
3623 * New! Add automatically a loopback interface (in the default
3624 network model) if none was precised.
3627 * Bugfix: MSG_process_resume now works with the current running process.
3629 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3630 * Modification of MSG_task_get_compute_duration. Once a task has been
3631 processed, the value returned by this function is now equal to 0. [AL]
3632 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3633 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3634 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3635 of a m_task. For now, MSG_task_cancel only works with computation
3637 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3638 of the processor (in Mflop/s) regardless of the current load on the
3640 * API Change: use proper naming convention for MSG_getClock and
3641 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3643 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3644 This function changes the priority of a computation task. This priority
3645 doesn't affect the transfer rate. A priority of 2 will make a task
3646 receive two times more cpu power than the other ones. This function
3647 has been added to suit the needs of Nguyen The Loc and hasn't been that
3648 much tested yet. So if it fails, please report it and send me your code.
3650 * API Change: removed all functions and types that were marked "deprecated"
3651 since many months. Renamed MSG_global_init_args to MSG_global_init.
3653 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3655 SimGrid (2.95) unstable; urgency=low
3658 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3659 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3660 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3663 * Bugfix: Allow absolute paths for platform description files [MQ]
3664 * Bugfix: do free the variables after use. Leads to drastic performance
3666 * Implement max_duration (ie, timeouts) on resources [AL]
3669 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3671 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3672 the needed support (timeouts) [AL]
3675 * Implement measurement sockets. You can now get the bandwidth between two
3676 hosts thanks to AMOK (see below). [MQ]
3677 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3678 dynar over the network (yeah) [MQ]
3679 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3680 * simple/static token-ring example. [Alexandre Colucci and MQ]
3681 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3682 in place before [MQ]
3685 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3686 type parsing. "short" and "long" modifiers were also ignored; other
3687 modifier (such as reference level) are still ignored. [MQ]
3688 * Embeed the buffer size within the buffer itself on SG. [MQ]
3689 That way, send() are atomic and cannot get intermixed anymore (at least
3690 the ones which are less than 100k; bigger messages still have the issue)
3691 * Array size pushed by the field, not by the field type (or each
3692 and every long int will push stuff to the cbps) [MQ]
3693 * use select() to sleep since it allows to portably sleep less than one
3696 GRAS (minor cleanups)
3697 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3700 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3701 * type category 'ignored' killed as it was never used and were difficult
3703 * whether a type can cycle or not is now a flag, leaving room for more
3704 flags (as "ignored", if we feel the need one day ;)
3705 * Rename raw sockets to measurement sockets since "raw" has another
3706 meaning in networking community.
3709 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3710 and offers features not belonging to GRAS but that most applications
3711 need. One day, it may be a set of plugins loadable at runtime.
3712 * New module: bandwidth
3713 bandwidth measurement between arbitrary nodes running this module. [MQ]
3715 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3717 SimGrid (2.94) unstable; urgency=low
3719 The first beta release of SimGrid 3 !
3723 * Update the main page and the FAQ. Adding references to gforge.
3726 * Add a gras_os_getpid function.
3729 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3730 * Extend the logs so that they also print PID, hostname, date, ... if
3732 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3733 and kill the old version which were in testsuite/
3734 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3738 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3739 for users. Instead, each programming environment (GRAS, MSG,...) use it
3740 when needed to provide such a feature to users.
3741 Don't shortcut the mechanism or you will also shortcut the virtualization
3742 you need on the simulator.
3746 * Cleanups in configury with regard to compile optimization/warning flags.
3747 Also add -fno-loop-optimize to any powerpc since it's the optimization
3748 killing gcc (< 3.4.0).
3749 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3750 needed by MSG examples complications
3751 * Borrow configury beautifications from PHP
3754 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3755 warning (thanks loris for stumbling into this one).
3756 * Bugfix: stop loading private headers (gras_config.h) from the public
3760 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3761 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3762 RL would imply a DNS resolver, which is *hard* to do in a portable way
3763 (and therefore delayed).
3764 * Implement a real timer mechanism and use it in timing macros. This allows
3765 to avoid rounding errors and get a 0.000005 sec precision in timing
3766 macros. While I was at it, various cleanups:
3767 - allow to declare more than one timed section per file (fix a stupid bug)
3768 - move some private declaration to the right place
3769 - merge conditional execution and timing macros into emulation module
3770 - document the module
3771 - make sure the module cleanups its mess on gras_exit
3772 * Documentation improvements:
3773 - (new) how to compile applications using GRAS
3774 - (new) emulation support (timing macros)
3776 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3778 SimGrid (2.93) unstable; urgency=low
3780 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3783 - Use Paje properly where used. Still to be sanitized properly.
3784 - Portability fix: Add an implementation of the contexts using pthread
3788 - Add xbt_procname(): returns the name of the current process.
3789 Use it to show the current process's name in all logging.
3791 - fix detection of older flex version and the reaction, since we do
3792 depend on modern ones (we use lex_destroy)
3793 - Better separation of SG and RL in the libs: remove all simulation code
3794 from libgras. As a result, this lib is now only 200k when stripped.
3795 Some of the xbt modules may also be duplicated (two sets and such) and
3796 should be cleaned/killed before SG3.
3797 - Insist on using xlC on AIX because of weird problems involving gcc there.
3798 - Cleanup the make remote stuff. This is now done by scripts
3799 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3800 mainly for our private use, but we're working on changing them to user
3803 - Bugfix: flush the socket on close only if there is some *output*.
3804 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3806 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3809 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3811 SimGrid (2.92) unstable; urgency=low
3813 Alpha 3 on the path to SimGrid 3
3817 - New! First try of benchmarking macros.
3818 - New! First try so that gras_stub_generator generate deployment and
3819 remote compilation helpers.
3821 - Bugfix: Initialization fix in msg_test.
3825 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3827 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3829 - API Change: gras_os_sleep to take the amount of seconds as a double.
3830 Accepting an int was error prone since it was the only location where
3831 seconds were coded as such. It leaded to damn rounding errors.
3832 - Bugfix: Hard to belive that timers ever worked before this.
3834 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3836 SimGrid (2.91) unstable; urgency=low
3838 Alpha 2 on the path to SimGrid 3
3842 - Bug fix in the lmm_solver.
3844 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3845 through the function MSG_paje_output.
3846 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3847 - It is possible to bound the rate of a communication in MSG with
3848 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3849 to write it in the changelog).
3850 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3855 - Various cleanups to the autotools stuff
3856 - Begin to move Gras examples to examples/gras/
3857 - Let make distcheck work again (yeah!)
3859 - documentation overhauled using doxygen.
3860 gtk-doc-tools is dead in SimGrid now.
3861 - Automatically extract all existing logging categories, and add the list
3862 to the documentation (long standing one, to say the less)
3864 - Cleanup the known architecture table. Reorder the entries to group what
3865 should be, and use a more consistent naming scheme.
3866 (some of the test dataset are still to be regenerated)
3867 - New! Allow library to register globals on each process just as userdata
3869 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3870 lookup time (for now).
3871 Use it in msg and trp.
3872 This cleans a lot the internals and helps enforcing privacy of the
3873 headers between the gras components.
3874 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3875 - Bugfix: gras_os_time was delirious in RL.
3876 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3878 - Reenable GRAS now that it works.
3880 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3882 SimGrid (2.90) unstable; urgency=low
3884 Alpha 1 on the path to SimGrid 3
3886 * It is a long time since the last release of SimGrid. I'm sorry about
3887 that but as I had told you, I was rewriting a lot of things. I apologize
3888 to those who had been reporting bugs to me and that I had not answered.
3889 If your bug is still in the new version, please tell me. Here is a
3890 summary of the main changes.
3892 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3893 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3894 portability, speed, and a lot more but you'll figure it out later.
3895 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3896 I wanted to release the new MSG as soon as possible and I have
3897 broken GRAS, which is the reason why, for now, only MSG is fully
3898 functional. A laconic description of these projects is available
3899 in the documentation.
3901 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3902 kernel optimized for our needs (called SURF but only the developers
3903 should use it). Hence, MSG is now roughly 30 times faster and I think
3904 that by rewriting a little bit MSG, I could event speed it up a little
3905 bit more. Beside the gain in speed, it is also much easier to encode a
3906 new platform model with SURF than it was with SG. More to come...
3908 * REVOLUTION 3: I have tried to change a little as possible the API of
3909 MSG but a few things really had to disappear. The main differences
3910 with the previous version are:
3911 1) no more m_links_t and the corresponding functions. Platforms are
3912 directly read from an XML description and cannot be hard-coded
3913 anymore. The same format is used for application deployment
3914 description. The new format is described in the documentation.
3915 Have a look in tools/platform_generation. There is a tiny script
3916 that converts from the old platform format to the new one. Concerning
3917 the application deployment format, parsing the old one is tricky.
3918 I think most of you should however be able to convert your files. If
3919 it is really an issue, I can write a C code that does the conversion.
3921 2) the toolbox tbx does not exist anymore. We now have a library
3922 with much more data-structures but without the hash-tables (we have
3923 dictionaries that are much faster).
3925 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3927 *****************************************************************************
3928 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3929 * don't feel like converting the oldies. (Mt) *
3930 *****************************************************************************
3933 Version 2.90: "the long awaited one"
3934 - Finished rewriting and debugging MSG. Rewrote the documentation.
3935 - disable GRAS for now since it needs to be ported to the newest SG
3938 - Finish the port to windows (using mingw32 for cross-compile)
3941 - Main loop and datastructures of SURF. A cpu resource object is
3942 functional. Surf can thus be used to create cpu's with variable
3943 performance on which you can execute some actions.
3945 2004-11-15 Martin Quinson
3946 - Port to ARM. Simply added the alignment and size descriptions. Should
3947 work, but the ARM machines are so slow that I didn't had the opportunity
3948 to 'make check' over there yet.
3950 2004-11-15 Arnaud Legrand
3951 - Trace manager now written. It uses a heap structure and is therefore
3952 expected to be efficient. It may however be speeded up (particularly
3953 when many events occur at the same date) by using red and black
3954 trees. One day maybe...
3955 - Max-min linear system solver written. It uses a sparse matrix
3956 structure taking advantage of its expected use. Most operations are
3957 O(1) and free/calloc are called as few as possible. The computation of
3958 the minimum could however be improved by using a red and black tree
3961 2004-11-03 Arnaud Legrand
3962 - Rename every gras_* function that was in xbt/ to its xbt_
3964 - Add a heap and a doubly-linked list to xbt
3965 - Added a dichotomy to the dictionaries. make check works as well before
3966 so I assume that the patch is correct. I do not know however if things
3967 run effectively faster than before now. :)
3969 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3970 SimGrid, and the version number is bumped to 2.x
3972 2004-10-29 Martin Quinson
3973 - Introduction of the remote errors.
3974 They are the result of a RMI/RPC on the remote machine.
3975 ErrCodes being scalar values, you can't get the host on which those
3976 errors did happen. Extending the error mechanism as in Gnome is possible.
3977 No idea yet whether it is a good idea.
3979 2004-10-28 Martin Quinson
3980 - Interface revolution: the Starred Structure Eradication.
3981 I used to do typedef struct {} toto_t; and then handle *toto_t.
3982 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3983 - ???_t is a valid type (builded with typedef)
3984 - s_toto_t is a structure (access to fields with .)
3985 - s_toto is a structure needing 'struct' keyword to be used
3986 - e_toto_t is an enum
3987 - toto_t is an 'object' (struct*)
3989 typedef struct s_toto {} s_toto_t, *toto_t;
3990 typedef enum {} e_toto_t;
3991 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3992 s_toto_t) is private.
3994 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3995 so that it can set the variable to NULL. It was so for dicts and sets,
3996 it changed for dynars.
3998 - Fix a bunch of memleaks in dict_remove
3999 - Fix a bug in sg/server_socket opening: it failed all the time.
4001 2004-10-07 Martin Quinson
4002 - Speed up dynar lookup operation a bit.
4004 gras_dynar_get is dead.
4006 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4007 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4008 (which gives you the address of the stored data).
4010 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4011 copy of the data using an affectation to do the job and not a memcpy.
4013 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4014 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4016 It does not really speedup the dynar test because they are
4017 setting elements all the time (and look them seldom). But the dict does
4018 far more lookup than setting.
4020 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4022 2004-10-05 Martin Quinson
4023 - Allow to (en/dis)able the cycle detection at run time.
4025 Whether we should check for cycle or not is now a property of each
4026 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4027 datadesc_cycle_unset allow to remove this property when previously set.
4029 Note that the cycle detection is off by default since it impacts the
4030 performance. Watch the data you feed GRAS with ;)
4032 This property is hereditary. Any element embedded in a structure having it
4033 set have it set for the time of this data exchange.
4035 You should set it both on sender and receiver side. If you don't set it on
4036 sender side, it will enter an endless loop. If you forget on receiver
4037 side, the cycles won't be recreated after communication.
4039 - Header reorganization.
4040 Kill gras_private.h, each submodule must load the headers it needs.
4042 2004-10-04 Martin Quinson
4043 - Interface revolution: do not try to survive to malloc failure.
4045 Now, gras_malloc and friends call gras_abort() on failure.
4046 As a conclusion, malloc_error is not a valid error anymore, and all
4047 functions for which it was the only gras_error_t return value are
4048 changed. They now return void, or there result directly.
4049 This simplify the API a lot.
4051 2004-09-29 Martin Quinson
4052 - Re-enable raw sockets.
4053 Created by gras_socket_{client,server}_ext;
4054 Used with gras_raw_{send,recv}
4057 It should allow to kill the last bits of gras first version soon.
4059 This is not completely satisfactory yet (duplicate code with
4060 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4063 - Simplify transport plugin (internal) interface by not passing any
4064 argument to _server and _client, but embedding them in the socket
4067 2004-09-28 Martin Quinson
4068 - Finish the port to AIX.
4069 autoconf was my problem (segfault within the malloc replacement
4070 function. No idea why)
4072 2004-09-16 Martin Quinson
4073 - Fix some size_t madness on 64bit architectures.
4075 2004-09-08 Martin Quinson
4076 - Reduce the number of system headers loaded, overload some more system
4077 calls (such as malloc to cast the result of the system one, and work
4079 - Fix and reintroduce the config support
4081 2004-09-07 Martin Quinson
4082 - Source code reorganization to allow Arnaud to surf all over there.
4083 - Allow to document the logging categories.
4084 - Remove all uppercase from logging categories and useless cleanup in names.
4086 2004-08-18 Martin Quinson
4087 Version 0.6.2 (protocol not changed; API changed)
4088 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4089 gras_error_t), and NULL when not found. Functions expecting a msgtype
4090 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4091 hopefully usefull message.
4092 - Portability to prehistoric sparcs again
4094 2004-08-17 Martin Quinson
4095 Version 0.6.1 (protocol not changed; ABI not changed)
4096 - prealloc some buffers to speed things up
4098 2004-08-11 Martin Quinson
4099 Version 0.6 (protocol not changed; ABI expended)
4100 - The parsing macro can deal with the references, provided that you add
4101 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4103 2004-08-09 Martin Quinson
4104 Version 0.5 (protocol not changed; ABI changed)
4105 - Allow to off turn the cycle detection code in data exchange at
4106 compilation time. It should be at run time, but I'm short of time (and
4107 the config stuff is still broken). That way, we keep dict out of the
4108 critical path, which is good because the performance is poor:
4109 - search not dichotomial yet
4110 - dynar give no way to access their content and memcpy everytime
4111 - In composed data description (struct, ref and so on), stop foolness of
4112 keeping the subtype's ID, but store the type itself. This keeps sets out
4113 of the critical path, which is good since they rely on dynar and
4114 dictionaries. The only loose of that is that we cannot detect the
4115 redeclaration of a structure/union with another content (but I'm not sure
4116 the code detected well this error before anyway). We still can detect
4117 the redefinition discrepancy for the other types.
4118 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4119 it breaks the code because of type-punning used all over the place).
4120 This breaks on all non-gcc architectures (for now).
4122 All those changes (plus the buffer of last time) allow me to gain 2 order
4123 of magnitude on cruel tests consisting of 800000 array of integers on two
4124 level of a hierarchical structure (200 secondes -> 4 secondes)
4127 - the selector of reference must now return the type it points to, not
4128 the ID of this type.
4130 2004-08-06 Martin Quinson
4131 Version 0.4 (protocol changed; ABI not changed)
4132 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4133 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4134 - Implement a buffer transport, and use it by default (it relies on tcp in
4135 real life and on sg in simulation).
4136 That's a bit hackish since I had a new field to the structure to store
4137 its data without interfering with the subtype ones. Inheritance
4138 is tricky in C. And that's a kind of reverse inheritance with one class
4139 derivating two classes. Or maybe a game with java interfaces. Anyway,
4140 that's damn hard in C (at least).
4141 Moreover, I got tired while trying to ensure plugin separation and
4142 genericity in SG mode. MSG wants me to do weird things, so let's go for
4143 cruel hacks (temporarily of course ;).
4144 See comment in transport_private.h:71
4145 - do not include all the _interface headers in private but in the files
4146 which really need them (to cut the compilation time when they are
4149 2004-07-26 Martin Quinson
4150 Version 0.3 (protocol not changed; ABI changed)
4151 - Major overhault of the datadesc interface to simplify it:
4152 - shorted the function names:
4153 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4154 - add a trivial way to push/pop integers into the cbps without malloc.
4155 This allows to make really generic sub_type description, which simply
4156 pop their size of the stack.
4157 - add a function gras_datadesc_ref_pop_arr() which does what users want
4158 most of the time: Declare a dynamic array (which pops its size of the
4159 stack) and declare a reference to it. Poor name, but anyway.
4160 - kill the post-send callback, add a post-receive one
4162 2004-07-23 Martin Quinson
4163 Version 0.2 (protocol changed; ABI changed)
4164 - add some testing for cpbs in the test cases, and fix some more bugs.
4165 This invalidate again the little64 data file, since I cannot regenerate
4167 - remove an awfull optimization in the logging stuff, allowing me to:
4168 - understand it again
4169 - learn gcc how to check that the argument match the provided format
4170 - fix all errors revealed by gcc after that
4171 - internal keys of dict are not \0 terminated. Deal with it properly in
4172 loggings instead of segfaulting when the user want to see the logs :-/
4174 2004-07-22 Martin Quinson
4175 - Fix some stupid bug preventing cbps (callback postit) from working
4177 2004-07-21 Martin Quinson
4178 - Some documentation cleanups
4179 - remove the useless last argument of msgtype_declare
4180 - rename the Virtu functions to fit into the 'os' namespace
4181 - move headers src/include -> src/include/gras/ and stop fooling with
4182 gras -> . symbolic link
4183 - make distcheck is now successful
4185 2004-07-19 Martin Quinson
4187 - Build shared library also
4188 - Install html doc to the right location
4189 - stop removing maintainer files in make clean
4190 - build tests only on make check
4192 2004-07-13 Martin Quinson
4194 - No major issue in previous version => change versioning schema
4195 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4196 corresponding dataset.
4198 2004-07-11 Martin Quinson
4200 - Get it working with any kind of structure (we can compute the padding
4201 bytes remotely for all the architectures I have access to)
4202 - Implement the structure parsing macro (still not quite robust/complete)
4203 - Improvement to the remote testing toysuite
4205 2004-07-10 Martin Quinson
4206 [autoconf mechanism]
4207 - get ride of a bunch of deprecated macros
4208 - actually run the test for two-compliment, not only compile it :-/
4209 - test whether the structures get packed (and bail out if yes. Damn.
4210 Alignment is a serious matter)
4211 - test whether the structures get compacted (but respecting the alignment
4212 constraints of each types)
4213 - test whether the array fields of structures can straddle alignment boundaries
4215 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4216 (took me 2 hours to find that bug, looking at the wrong place)
4217 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4218 before sending/receiving and are used to compute the offsets of fields
4219 - Given that padding size depend even on compiler options, keep track of
4220 alignment and aligned_size only for the current architecture. Not a big
4221 deal since we send structure fields one after the other (seems
4223 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4224 program, let it work on linux/gcc/little32. portability todo.
4226 2004-07-08 Martin Quinson
4227 - import and improve remote compilation support from FAST
4228 - make sure make check works on half a dozen of machines out there
4230 2004-07-07 Martin Quinson
4231 Let's say it's version 0.0.3 ;)
4232 - Implement conversions (yuhu!)
4233 - Let it work on solaris (beside conversion, of course)
4234 - Stupid me, using rand() to generate the conversion datatests in not wise.
4236 2004-07-06 Martin Quinson
4237 - Let make dist work, since I'm gonna need it to compile on remote hosts
4238 - Let Tests/datadesc_usage write the architecture on which the file was
4239 generated as first byte.
4240 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4241 architecture descriptions.
4242 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4243 on those architectures.
4244 - Optimization: send/recv array of scalar in one shoot
4246 2004-07-05 Martin Quinson
4247 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4249 - Plug a whole bunch of memleaks
4250 - each process now have to call gras_{init,exit}. One day, their log
4251 settings will be separated
4252 - Continue the code factorisation between SG, RL and common in Transport.
4254 2004-07-04 Martin Quinson
4256 - Redistribution between SG and RL.
4257 We wanna have to accept in SG, so move accepted related parts of RL in
4258 the common part. (more precisely, the dynar of all known sockets is no
4259 more a static in transport.c, but part of the process_data)
4261 [gras_stub_generator]
4262 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4264 2004-07-03 Martin Quinson
4265 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4266 them all up in one shot)
4268 - Fix the ugly case of reference to dynamic array.
4269 - New (semi-public) function gras_datadesc_size to allow the messaging
4270 layer to malloc the needed space for the buffer.
4272 - gras_socket_close now expect the socket to close (and not its address to
4273 put NULL in it after it). This is because the socket passed to handlers
4274 is one of their argument (=> not writable).
4276 - propagate the interface cleanup from last week in datadesc, ie remove a
4277 superfluous level of indirection. User pass adress of variable
4278 containing data (both when sending and receiving), and not of a variable
4279 being a pointer to the data. Let's say that I like it better ;)
4280 The price for that is constructs like "int msg=*(int*)payload" in
4281 handlers, but it's a fine price, IMHO.
4283 - Let it work in RL (yuhu)
4285 2004-06-21 Martin Quinson
4287 - porting SG plugin and SG select to new standards (works almost).
4288 - plug memleaks and fix bugs around.
4291 - cleanup the prototype of data recv and force users to specify when they
4292 want to handle references to objects. Test case working even for cycles.
4293 - plug memleaks. Valgrind is perfectly ok with this.
4295 2004-06-12 Martin Quinson
4297 - cleanup the separation between plugin and main code in plugin creation
4299 2004-06-11 Martin Quinson
4301 - Reput hook for raw sockets, needed for BW experiments
4302 - kill a few lines of dead code
4303 [Data description] Interface cleanup
4304 - gras_datadesc_by_name returns the searched type or NULL.
4305 That way, no variable is needed to use a type desc once, which makes
4307 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4308 two parameters were strings (field name, type name), leading to
4310 [Dicos] Interface cleanup
4311 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4312 This is consistant with the dynar API.
4314 2004-04-21 Martin Quinson
4316 - Porting to new standards.
4318 - interface cleanup.
4319 There is no bag anymore, no need to take extra provision to mask the
4320 pointers behind "ID".
4321 Better splitup of functions between files create/exchange/convert.
4322 This is still a bit artificial since convert and receive are so
4323 interleaved, but anyway.
4325 - add a queued message list to procdata (the ones not matching criteria
4327 - factorize some more code between SG and RL wrt procdata
4329 - use gras_exit in example to track memleaks
4330 - get rid of gs_example now that GS is properly integrated into gras
4331 - update run_test to integrate the latest tests (datadesc)
4333 - rename WARNINGn macros to WARNn since it prooved error-prone
4335 2004-04-19 Martin Quinson
4337 - register init/exit functions within gras module mechanism
4338 - send/receive function.
4339 Convertion is not implemented, but short-cutted if not needed.
4340 struct/array elements are sent one by one (instead of block-wise), but
4341 nobody really cares (yet). Get a prototype before optimizing.
4342 - tests (using a file socket) for DD send/receive on:
4343 - base types: int, float
4344 - array: fixed size, string (ie ref to dynamic string)
4345 - structure: homogeneous, heterogeneous
4346 - chained list, graph with cycle
4347 Believe it or not, valgrind is not too unhappy with the results. The
4348 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4352 - Bugfix when using a filename explicitly (instead of '-')
4354 2004-04-09 Martin Quinson
4356 - factorize more code between RL and SG in socket creation
4357 - Complete the implementation and tests of:
4359 o file (only in RL, and mainly for debugging)
4361 I lost 3 days to design a portable address resolver, and then decided
4362 that the prototype mainly have to run on my box.
4363 Addressing portability too early may be like optimizing too early :-/
4365 - use gras_init in the Tests instead of the crappy parse_log_opt
4366 (the latter function is removed)
4367 [Conditional execution]
4368 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4369 [Code reorganisation]
4370 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4371 Build examples against the RL library, since there is no way to disable
4372 its creation for now.
4374 For information, the beginning of coding on GRAS was back in june
4375 2003. I guess that every line has been rewritten at least twice since