1 ----------------------------------------------------------------------------
3 SimGrid (3.27.1) NOT RELEASED YET (v3.28 expected June 21. 2021, 03:32 UTC)
6 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
7 Explicitly cancel the activity to get back to previous behavior.
10 - Lua platform files are deprecated. Their support will be dropped after v3.31.
13 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
15 ----------------------------------------------------------------------------
17 SimGrid (3.27) March 29. 2021
19 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
21 Main user-visible changes:
22 - All functions marked as deprecated for one year were removed.
23 - Removed all that was related to the "Storage" type in all APIs and bindings.
24 "Disks" have been introduced more than a year ago to replace them for the better.
25 It's time to finish this replacement.
26 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
27 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
28 avoiding "initializer element is not constant" errors with some codes. Please be
29 careful to recompile your MPI codes if switching to this new SimGrid version.
30 - Network model ns-3 now requires version ns-3.28 or higher.
31 - Removed some option aliases that were deprecated since v3.23 (June 2019)
32 - 'verbose-exit' is now 'debug/verbose-exit'
33 - 'simix/breakpoint' is now 'debug/breakpoint'
34 - 'clean-atexit' is now 'debug/clean-atexit'
37 - New functions: Comm::sendto_{init,async} to initiate a communication
38 on between two (possibly remote) hosts.
39 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
40 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
41 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
42 or Mailbox::get_async<void>() if you really want to play with void*.
43 - Unify the interface of Activity::on_{start/activity}
44 - New function: Comm::get_dst_data()
45 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
48 - ptask should now work better with multicore hosts (FG#37).
49 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
50 - Internal cleanups and refactoring around models, to prepare for programmatic
52 - Modernize our tests of the performance models
56 - New section: Configuring the logs from the command line.
57 - New section: Programming API of logs.
58 - New section: Getting realistic results
59 - New section: Examples of SMPI Usage
60 - New part: SimGrid's Internals, with a section on Design goals
61 - Move back to the classical Breathe tool chain, and fix many small glitches.
64 - Handle leak checking: SMPI will now display a message if MPI handles are still
65 present at the end of a simulation. Option smpi/list-leaks:n will display the n
66 first leaked pointers and their types.
67 - smpi/display-allocs option : Display Memory allocations analysis, with hints
68 about the largest allocations. This only works for allocations with malloc/calloc
69 and not compiled with SMPI_NO_OVERRIDE_MALLOC
70 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
71 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
72 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
73 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
74 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
75 - Replay now opens and parses the trace file only once for all ranks when possible.
77 - better handling of MPI_PROC_NULL
78 - MPI_Group_incl was broken in some cases
79 - PAPI tracing now fixed
80 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
81 - SMPI will now fail properly when receive size < sent size, instead of
82 just setting MPI_ERR_TRUNCATE and keep going.
83 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
84 - Remove useless build patches and split build/run patches on the page
85 - Build in parallel when possible
86 - Test on yum-based systems
87 - New nightly tested Proxy-apps:
88 - OpenMPI mpi-test-suite
89 - Coral MCB, LSMS, Qbox, CLOMP
90 - ECP/Mantevo Cloverleaf3D, Ember
91 - Codevault bhtree_mpi
94 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
95 - Start to overhaul the implementation to modernize it (still ongoing).
98 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
99 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
100 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
101 - Deprecate xbt_dynar_to_array().
102 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
104 C binding and interface:
105 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
106 - Constness requirement has been lowered for parameter argv of sg_actor_start()
107 and sg_actor_create() It's now (char* const*). The previous version is still
108 available as sg_actor_start_() and sg_actor_create_().
110 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
111 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
112 - FG#37: Parallel tasks are limited to 1 core per host
113 - FG#62: Running "smpirun -replay" on large networks
114 - FG!46: Fix a few potential memory leaks in SMPI colls
115 - GH#343: Wrong consumed energy when an exec timeouts
117 ----------------------------------------------------------------------------
119 SimGrid (3.26) December 15. 2020
121 The Release release (the French lockdown was eased today).
123 Important user-visible changes:
124 - SimGrid now requires a compiler with C++14 support.
125 Sibling projects should upgrade their FindSimgrid.cmake
126 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
127 several users had difficulties to understand issues when using high bandwidth or
128 small latency events. The new value was already the default for SMPI and
129 should not cause too much performance hit. It can change some simulation
130 timings if simulation was skipping a lot of small events (which was actually
131 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
132 - On some network models (L07, CM02, LV08), an implicit loopback link is created
133 for local communication. Default bandwidth was 500MBps and latency was 15µs,
134 which was rather bad. Default values have been changed to 10GBps and 0 latency.
135 Options network/loopback-lat and network/loopback-bw have been added to adapt
136 these values if needed. Network and routing models already providing separate
137 loopbacks (clusters and such) are not impacted by this change.
138 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
139 metadata that are added to each MPI message by MPI libraries (tag,
140 src/dest, and communicator information). This may change simulation times
141 for very small message sizes, but should be more realistic.
144 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
145 - New plugins: wifi_energy and link_load. See the associated examples.
146 - New examples: wifi networks, and communication suspend.
149 - LTO is now enabled for Intel/clang compilers.
150 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
151 Setting it to "auto" will use all available cores, while setting it to n will
152 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
153 - Remove obsolete runtime option 'contexts/parallel-threshold'.
154 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
155 has been kept for compatibility.
156 - Finally remove obsolete snake_case() aliases for runtime options.
157 - Further improve the documentation.
160 - the on_exit() of each actor is also executed when the simulation deadlocks.
161 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
162 An example is provided for s4u::Comm but it should work with Exec and Io.
165 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
166 https://framagit.org/simgrid/SMPI-proxy-apps
167 - MPI/IO: supports names starting with "./"
168 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
170 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
171 building shared libraries, this may cause issues, so environment variable
172 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
173 - most temporary files should now be created in /tmp dir (or equivalent).
174 If this one does not allow execution of code (noexec flag), this may cause issues.
175 Please use another tmp directory (using TMPDIR or equivalent system variable)
179 - A large refactoring is ongoing, but this should be transparent for users so far.
182 - Drop xbt_str_split().
184 C binding and interface:
185 - The return type of the sg_actor_on_exit() callbacks is now 'void'
186 (int value was previously ignored)
187 - Many C functions were renamed to follow a common naming style
188 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
189 names are kept until version 3.30.
190 - Many MSG tests were converted to the new S4U's interface in C, that
191 was extended for that.
193 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
194 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
195 - FG#41: Add sg_actor_create C interface
196 - FG#43: xbt::random needs some care
197 - FG#48: The Impossible Did Happen (yet again)
198 - FG#50: Suspending an actor executed at the current timestamp fails
199 - FG#52: Zero-seconds timeout : "That's in the past already"
200 - FG#53: Crash while using ns-3 network model
201 - FG#54: How to suspend a comm?
202 - FG!22: Stochastic Profiles
203 - FG!24: Documentation and fix for xbt/random
204 - FG!35: Add a modeling hint for parallel links in doc
205 - FG!36: [xbt/random] Read/Write the state of the RNG
206 - FG!42: Link cumulated load plugin
207 - FG!44: Test condition variable wait_until+notify_all
208 - GF#18137: Allow different stack sizes?
209 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
210 - GH#139: Allow pthread creation in SMPI
211 - GH#336: Packet-level simulation using SMPI?
212 - GH#345: Error in the LMM while migrating a VM
213 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
214 - GH#352: pip install / python setup.py install fail to find pybind11
215 - GH!337: Fix link_energy plugin for wifi platforms
216 - GH!339: Add Mailbox set_receiver method to python binding
217 - GH!344: Cast hugepages macros parameters to int64
219 ----------------------------------------------------------------------------
221 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
223 The Palindrom Day Release.
225 Important user-visible changes:
226 - Improve the Python usability (stability and documentation).
227 - A nasty synchronization bug was ironed out, see also below.
228 - Python's doc was organized as a separate tree, now integrated with C++.
229 - C bindings of S4U were not part of the doc.
230 - The C++ doc was also improved as methods are now split by theme.
231 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
232 - This is mandatory to use the Java bindings.
233 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
234 - SimDAG++: Automatic dependencies on S4U activities (experimental)
235 - Some features are already implemented but not all of them
236 - Cannot block an activity until it's scheduled on a resource
237 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
238 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
241 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
242 - Drop support for Viva/Triva (old visualization tools), hence removed
243 TRACE_get_node_types() and TRACE_get_edge_types() functions.
246 - Require pybind11 version 2.4.
247 - Greatly improve locking during context switches with Python.
248 - Declaring Python callbacks to this_actor::on_exit is now working.
249 - No more tests are failing currently, even on win32.
252 - Actor: Merge signals on_migration_start/end into on_host_change
253 - Actor: Rename migrate() into set_host()
254 - Disk: Allow users to get the read and write nominal bandwidth values
255 - Exec: Implement wait_for(timeout)
256 - Io: Implement wait_for(timeout)
259 - Parse errors now raise a simgrid::ParseError that you may want to catch.
262 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
263 been renamed to fix a spelling error. As usual, the previous definitions are
264 kept with a deprecation warning. However, to avoid runtime errors, it is now
265 forbidden to override these deprecated methods in a derived class. Please use
266 the new names immediately if you need to override them.
269 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
270 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
271 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
272 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
273 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
274 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
275 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
276 - New module: random, providing classical random numbers generators.
279 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
280 allocations' size above which they are considered "shared" by default (as if
281 they were performed through SMPI_SHARED_MALLOC macros).
282 Default = 0 = disabled feature.
283 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
284 This can cause some troubles if codes are already overriding these. If this
285 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
286 help, but will make this feature unusable.
287 - Input check is now performed much more consistently and outputs precise
288 warnings for each error found
289 - New supported calls : MPI_File_set_view, MPI_File_get_view
290 - MPI I/O now only supports disks, as storage is discontinued.
291 - MPI I/O allows opening files without specifying absolute path
292 - Note : SMPI configuration options are now initialized by smpi_init_options()
293 call, which should be called if SMPI is not being used through smpirun.
295 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
296 - FG#9: Python bindings crashing
297 - FG#39: Missing s4u::Comm::wait_any_for example
298 - FG#42: Add support for ThreadSanitizer (TSan)
299 - FG!19: Removing RngStream
300 - FG!20: A module for RNG calls
301 - FG!21: Choice between ad-hoc and standard distributions implementations
303 - FG!25: Fix link in Mutex doc
304 - FG!26: Fix links in SMPI interface doc
305 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
306 - GH#305: Unscheduled tasks are still excuted
307 - GH#313: smpirun: manual outdated w.r.t. --help
308 - GH#321: [S4U] Get task remaining work ratio
309 - GH#323: Crash when an actor turn off his physical host
310 - GH#335: Missing links on dragonfly example svg
311 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
312 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
314 ----------------------------------------------------------------------------
316 SimGrid (3.24) October 9. 2019
318 The Clean Disk Release.
321 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
322 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
323 thus becomes the default resource to execute I/O operations. Examples
324 and the file system plugin have been modified to use disks instead of
325 storage elements. Storage can still be used but is doomed to disappear.
326 Users are thus adviced to update their simulators accordingly.
327 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
328 for consistency with pthread_barrier_wait()
329 - Host::get_englobing_zone() returns the englobing netzone
330 - Actor::on_destruction is now called in the destructor
331 Actor::on_termination new signal called when the actor terminates
333 - Global signals are now part of the Engine:
334 - on_platform_creation: after config settings, before the XML parsing
335 - on_platform_created: right after the XML parsing
336 - on_time_advance: each time the clock advances
337 - on_simulation_end: after simulation, before cleanups
338 - on_deadlock: as the name implies.
340 - sg_{actor,host,link}_{data,data_set}() now all exist.
341 Use them to attach user data to the object and retrieve it.
344 - Introduce an experimental Wifi model. It sounds reasonable
345 according to the state of the art, but it still has to be properly
346 validated, at least against ns-3 if not against reality.
347 - Improved the usability of ns-3. Several bugs were ironed out.
348 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
349 It is now expressed as 'idle:epsilon:allCores' to properly model the
350 consumption of non-whole tasks on mono-core hosts. More info in the doc.
353 - convert a new set of functions to the S4U C interface and move the old MSG
354 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
357 - Fortran bindings for DVFS have been removed.
358 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
359 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
360 - SMPI can now be selected by cmake's find_module(MPI) with
361 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
362 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
363 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
364 - trace-call-location can be used with TI traces, and replayed, and is compatible
365 with smpi/comp-adjustment-file.
366 - sleep events are now correctly traced/replayed.
367 - Default for trace-call-location is now to use file names and not full paths.
368 To revert to previous behavior (in case of collision of filenames), option
369 "smpi/trace-call-use-absolute-path" can be set to yes.
372 - Use the included xxHash as an hash implem when C++14 is usable.
373 - Option model-checker/hash was removed. This is always activated now.
374 - New option smpi/buffering controls the MPI buffering in MC mode.
375 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
376 This is useful to check for MPI compliance.
379 - New section on plugins: how to define a new one, and existing ones.
382 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
383 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
386 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
387 and <mount> tags. The rationale is finish to decouple what should be handle
388 by SimGrid kernel (i.e., resources and activities) from plugins built in
389 user space (e.g., the file system plugin). This new tag currently accepts
390 only three attributes (id, read_bw, and write_bw). All the other information
391 that was declared with the storage related tags now has to be expressed as
392 properties. An example of platform using this new tag is available at
393 examples/platforms/hosts_with_disks.xml
396 - 'expect signal' can now accept more than one potential signal.
398 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
399 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
400 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
401 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
402 - FG#31: per-actor data
403 - FG#34: SG_BARRIER_SERIAL_THREAD?
404 - FG#35: model-checker does not like buster-produced binaries
405 - FG!13: MC: complete workaround in the error msg seen on modern systems
406 - FG!15: execute_flops now logs compute
407 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
408 - FG!17: ns-3: unblock the right number of communications + others issues
409 - FG!18: Improving the performance of the ns-3 bindings
410 - GH#189: Energy consumption of parallel tasks
411 - GH#219: Error in the throughput of TCP transfer
412 - GH!330: Minor polishing of English: Configuring_Simgrid
413 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
414 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
415 - GH!334: Rework energy plugin, again
417 ----------------------------------------------------------------------------
419 SimGrid (3.23.2) July 8. 2019
422 - Nicer introduction page.
423 - Migrate the "Deploy your application" page to the new doc.
424 - Move Java as a subtree of MSG.
427 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
430 - Drop xbt_dynar_sort_strings().
433 - Really fix FG#26: Turning off a link should raise NetworkFailureException
434 - FG#27: Wrong exception thrown to wait_any when link is turned off
435 - GH#328: Java: Canceling multiple tasks in a single vm/host
437 ----------------------------------------------------------------------------
439 SimGrid (3.23) June 25. 2019
441 The Exotic Solstice Release.
444 - SunOS and Haiku OS support. Because exotic platforms are fun.
445 - Stop setting random seed with srand48() at initialization.
446 - Use addr2line as a fallback for stacktraces when backtrace is not available.
447 - Build option -Denable_documentation is now OFF by default.
448 - Network model 'NS3' was renamed into 'ns-3'.
451 - Simgrid can now hopefully be installed with pip.
454 - wait_any can now be used for asynchronous executions too.
457 - New log appenders: stdout and stderr. Use stdout for xbt_help.
458 - Drop xbt_dict_dump.
461 - 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
462 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
463 - MPI interface is now const correct for input parameters
464 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
465 - Fortran bindings for async collectives.
466 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
469 - Remove option 'model-check/record': Paths are recorded in any cases now.
470 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
471 always sparse. That's an immense gain in memory, and can even be
472 faster because of cache effects. So there is no need to clutter the
473 code to allow the user to go for the unefficient mode.
476 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
477 models proved to be more accurate than these old experiments.
479 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
480 - FG#1: Broken link in error messages
481 - FG#2: missing installation documentation
482 - FG#3: missing documentation in smpirun
483 - FG#6: Python bindings not available on PyPI
484 - FG#7: simple cmake call requires doxygen
485 - FG#8: make python bindings an optional dependency
486 - FG#10: Can not use MSG_process_set_data from SMPI any more
487 - FG#11: Auto-restart actors forget their on_exit behavior
488 - FG#12: -Denable_lto=OFF doesn't disable LTO
489 - FG#13: Installs unstripped file 'bin/graphicator'
490 - FG#14: Installs the empty directory 'doc/simgrid/html'
491 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
492 - FG#17: Dead link in doc (pls_ns3)
493 - FG#20: 'tesh --help' should return 0
494 - FG#21: Documentation link on http://simgrid.org/ broken
495 - FG#22: Debian installation instruction are broken
496 - FG#26: Turning off a link should raise NetworkFailureException exceptions
497 - GH#133: Java: a process can run on a VM even if its host is off
498 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
499 - GH#326: Valgrind-detected error for join() when energy plugin is activated
501 ----------------------------------------------------------------------------
503 SimGrid (3.22) April 1. 2019
505 The Easter Christmas Release (this one is somewhat late).
508 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
509 - Not complete yet: asynchronous activities (amongst others) are still missing
510 - Still ongoing: the interface may change in the future. We need more
511 testers! Please report any glitches.
512 - No new project using Java should start now. Please switch to Python.
515 - Some of the internal cleanups may lead to speed improvements:
516 - The hard limitation on the amount of simulated actors with
517 Java+Mac was removed. Now, the available memory is the only limit.
518 - Our refcounting was tidyied, leading to 10% speedups in some cases.
519 - We are still working on making our code robust to the actor kills
520 and hosts' churn. Things are improving, but it's not perfect yet.
521 - Replaced our own code to display a backtrace (that was forking addr2line)
522 with the Boost.Stacktrace library.
523 You won't see your backtraces without this optional dependency.
524 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
525 - Stop setting random seed with srand() at initialization.
528 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
529 XML file version remains 4.2 since old files are still compatible.
532 - Process termination which was broken at version 3.21 has been repaired.
533 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
534 - Hide the examples into examples/deprecated. New users should use Python.
537 - Drop MSG_process_create_from_stdfunc() from the API.
538 This C++ function was a pimple in the C API, made necessary at some
539 point by the Java bindings. This is fixed now.
540 - Hide the examples into examples/deprecated. New users should use S4U.
541 - MSG_process_create and MSG_process_attach now crash if the host used
542 to run this process is off.
543 - Fix the protype of MSG_process_on_exit()
544 Now use (int,void*) callbacks instead of (void*,void*) ones.
545 The implementation was ways too messy to actually work, I guess.
548 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
549 too often and provide more accurate timings.
550 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
551 - MPI_Alltoallw support
552 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
553 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
554 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
555 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
557 - Basic implementation of generalized requests (SMPI doesn't
558 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
561 - Drop sg_cmdline. Please use xbt_cmdline instead.
562 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
563 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
564 OsSemaphore is implemented in a portable way with C++11 threads.
565 This should allow much more threads to be created at the same time,
566 allowing Mac and Java users to simulate many more actors.
567 - Implement the 'thread' factory with std::thread instead of xbt ones.
568 It is not possible to set the stack size with threads anymore, but
569 -fsplit-stack is the way to go nowadays when using threads.
570 - Drop the xbt_os_thread_t module (now unused)
571 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
572 - Drop xbt_str_join_array().
573 - Drop cunit, use Catch2 instead.
576 - Many cleanups in the kernel::activity namespace. This was long
577 overdue, and shall open the path to many future endeavors.
580 - #132: Java: a process can not shut down its own host
581 - #220: S4U: detached send causes exception if sender terminates soon after sending
582 - #261: Document the parameters of parallel execution's constructor
583 - #300: [s4u] BarrierPtr is missing
584 - #314: SMPI args internal cleanup
585 - #316: Fix a bug related to the CPU utilization of multi-core VM
586 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
587 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
588 - #325: Turning off a host has different behavior on sleeping actors and computing actors
590 ----------------------------------------------------------------------------
592 SimGrid (3.21) October 3. 2018
594 The Restarting Documentation (TRD) Release.
597 - Convert the user manual to Sphinx for improved usability.
598 Unstable version now hosted on https://simgrid.frama.io/simgrid
599 Many glitches remain; stay tuned.
600 - Restrict the Doxygen reference API to the installed header files.
601 (documenting all internals this way is overkill + doxygen scales badly)
602 - New tutorial on S4U (highly inspired from the old MSG one, but with
603 a git repository to fork as a starting point, and a docker image)
604 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
605 - The Developper manual is still to be converted (not compiled until then)
606 - Some parts are still missing in this conversion (such as the
607 platform and deployment sections) while others were blindly converted
608 and would need more love (such as the configuration flags).
609 Things will certainly further improve in the future releases.
612 - Fully reimplement auto_restart mechanism that was utterly broken
613 (fix #22, #131 and #281 that were all very related)
614 - Implement semaphores (fix #303 and #291)
615 - When creating an actor from a function and its parameters,
616 move-only parameters are not allowed anymore, as it would prevent
617 the actor to be restartable if its parameters are consumed on run.
618 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
619 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
620 IO operations that can be started, waited for, or canceled as a regular
622 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
626 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
627 - New host variable: 'core_count'
630 - Remove xbt_os_thread_specific features
631 - Remove portability wrapper to condition variables
632 - Remove xbt_os_thread_yield()
635 - MPICH collective selector now mimics MPICH 3.3b
636 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
637 - MPI_Init(NULL,NULL) is now allowed
638 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
639 - Command-line arguments should now be handled properly in Fortran simulations
640 - MPI Attributes and related callbacks should work in Fortran
641 - Apps using C + Fortran should now work
642 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
643 - Support MPI_Error_String
646 - Due to an internal bug, Msg.run() must now be your last line.
647 We hope to fix it in a future release, and we are sorry for the inconvenience.
650 - #22: Process autorestart seem to only work with CAS01 cpus
651 - #56: Feature request: dynamic SMPI replay
652 - #93: simgrid should not eat --help
653 - #111: How to change the loopback link characteristics?
654 - #116: Platform cluster radical
655 - #117: Platform cluster/host order
656 - #129: Replay_multiple on steroid
657 - #131: Java : autostart does not work
658 - #143: Setting a breakpoint at a given time
659 - #235: Network energy models should be integrated
660 - #264: Add ptask L07 resource tracing
661 - #271: Dynamic and manual replay of SMPI traces
662 - #279: Breakpoints option uses time deltas instead of absolute time
663 - #281: Daemonized actor no longer daemonized after an auto-restart
664 - #282: TIT convertor erase traces when using absolute path in the trace list file
665 - #285: segfault when a s4u actor kills itself with msg process tracing activated
666 - #286: Pajé traces are not exposing the number of cores
667 - #287: Command-line parsing should stop when encountering '--'
668 - #288: MPI_Init(NULL, NULL)
669 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
670 - #290: Method to check if a mailbox has a message ready to be consumed
671 - #291: [s4u] semaphores do not seem available yet
672 - #294: Alltoall collective implementation of mpich
673 - #295: Starting a Host as OFF is not supported
674 - #296: DTD too permissive
675 - #299: Add a s4u barrier example
677 ----------------------------------------------------------------------------
679 SimGrid (3.20) Released June 24. 2018
681 The proxy snake_case() release.
684 - The whole API is now using snake_case() instead of mixing it with camelCase.
685 Compatibility wrappers in place for one year.
686 - We took the opportunity to increase the consistency of the API.
687 Please report any remaining glitches.
688 - New function simgrid::s4u::Host::get_actor_count:
689 Returns the number of actors running on a specific host.
692 - SMPI is now tested with ~45 proxy apps from various sources, with none or
693 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
694 - Replay: The replay file has been re-written in C++.
695 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
696 supported. They are stored in the trace and used when replayed.
697 - Basic support of MPI_Cancel. Robustness not guaranteed.
698 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
699 (only for MPI_COMM_TYPE_SHARED).
700 - New option: smpi/privatize-libs, to add external shared libs to be privatized
701 by SMPI. They will be copied locally and loaded separately by each process.
702 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
703 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
704 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
705 with internal index optind with multiple processes. Only works if MPI_Init has
707 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
708 syntax without preprocessing tricks.
711 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
712 behavior tracing is no longer limited to MSG
713 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
714 this means that traces now need to be updated or re-obtained; both Irecv
715 and Isend lines in your traces have now in total 5 mandatory fields:
716 <rankid> <command> <to/from rankid> <tag> <size>
717 To update your traces, it suffices to add a 0 for the tag here.
718 - TIT now also supports waiting for a distinct request via MPI_Wait.
719 Wait/Test now wait for a specific request, not just the last one that was
720 issued. This unfortunately means another update, because we need to
721 identify which request you want to wait for. We do this via the
722 triplet (sender, receiver, tag), which needs to be added:
723 <rankid> <command> <sender> <receiver> <tag>
724 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
725 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
728 - The deprecation of MSG is ongoing (but this should not impact you).
729 Many MSG functions are now simple wrappers on the C API of S4U. If
730 you wish to convert your code to S4U, find the S4U counterparts of
731 your MSG calls in src/msg/msg_legacy.cpp.
732 - MSG can still be used, but won't evolve anymore.
735 - Allow to run the Link energy plugin from the command line with
736 --cfg=plugin:link_energy
737 - Rename Energy plugin into host_energy
738 - Rename Load plugin into host_load
741 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
742 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
743 - Improve performance when handling timeouts of simix synchros.
746 - Config: the C API is now deprecated (will be removed in 3.23), and
747 the C++ API has been slightly improved.
750 - Fix several build issues on OSX.
751 - Move simgrid_config.h to simgrid/config.h (old header still working)
754 - #143: Setting a breakpoint at a given time
755 - #150: Inconsistent event names in SMPI replay
756 - #258: daemonized actors hang after all non-daemonized actors have completed
757 - #267: Linker error on unit_tmgr
758 - #269: SMPI: tracing of MPI_Wait/all/any broken
759 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
761 ----------------------------------------------------------------------------
763 SimGrid (3.19.1) Released March 22 2018
765 The Fixed ABI Release.
767 As you may know, we are currently refactoring SimGrid in deep.
768 Upcoming SimGrid4 will be really different from SimGrid3: modular,
769 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
771 Our goal is to smooth this transition, with backward compatibility
772 and automatic update paths, while still progressing toward SimGrid4.
774 SimGrid remains open during works: The last pure SimGrid3 release was
775 v3.12 while all subsequent versions are usable alpha versions of
776 SimGrid4: Existing interfaces remain unchanged, but the new S4U
777 interface is budding and the internals are deeply reorganized.
779 Since 2015, we work hard to reduce the changes to public APIs. When
780 we need to rename a public library symbol in S4U, we let your
781 compiler issue an explicative warning when you use the deprecated
782 function. These messages remain for four releases, i.e. for one full
783 year, before turning into an error. Starting with v3.15, your can
784 also adapt to API changes with the SIMGRID_VERSION macro, that is
785 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
787 Starting with this v3.19.1, our commitment to reduce the changes to
788 the public interfaces is extended from the API to the ABI: a program
789 using only MSG or SimDag and compiled against a given version of
790 simgrid can probably be used with a later version of SimGrid without
791 recompilation. We will do our best... but don't expect too much of
792 it, that's a really difficult goal during such profund refactoring.
794 The difference between v3.19 and v3.19.1 is that the former was
795 accidentally breaking the ABI of MSG, while the later is restoring
798 S4U and kernel APIs will still evolve until SimGrid4, with one-year
799 deprecation warnings as currently. In fact, cleaning up these
800 interfaces and converting them to snake_case() is one release goal of
801 v3.20. But don't worry, we are working to smooth this upgrade path.
803 In summary, new projects should start with S4U to benefit of the
804 future, but old MSG projects should still be usable with no change.
806 ----------------------------------------------------------------------------
808 SimGrid (3.19) Released March 20 2018
810 The Moscovitly-cold Spring release.
812 SMPI user-visible improvements:
813 - Support of MPI_Type_create_subarray
814 - Unimplemented MPI calls now abort (instead of issuing a warning)
815 Some harmless calls around errhandlers still only issue a warning.
816 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
817 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
818 is now possible (mmap privatization was refactored to this end).
819 Adding new SMPI instances should also be technically doable now.
821 SMPI internal cleanups:
822 - Removed index notion from SMPI -- no more "getPid() - 1"!
823 In tracing files, actors are called rank-<PROCESS-ID> but these are
824 NOT the real ranks (tracing several communicators is not supported yet).
825 - Deprecate smpi_process_index() (will be removed in 3.22)
826 - Better testing on classical proxy apps, through an external project:
827 https://github.com/simgrid/SMPI-proxy-apps/
830 - Use a graphical TOC to make it easier to find the documentation you need
831 - Revamp the MSG tutorial
834 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
835 - Remove parameter of MSG_process_killall().
836 Resetting the PID was bogus anyway (several actors could have the same PID).
839 - Execution->setHost() can be called after start(), to migrate it.
840 - Comm::test_any() is now implemented.
841 - s4u::Actor now has onCreation() and onDestruction() signals.
842 - Install some kernel header files for the users' plugins and more.
843 Warning, their API is really not stable yet.
844 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
845 include/simgrid/kernel/routing/*Zone.hpp
848 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
850 Build System and other cleanups:
851 - Remove unused run-time parameter "tracing/onelink-only".
852 - SimGrid now works with intel compilers (MC still buggy)
853 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
854 - tesh kills subprocesses on timeout (but not on windows)
857 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
860 - #194: Feature request: simgrid::s4u::Comm::test_any()
861 - #245: migrating an actor does not migrate its execution
862 - #253: Feature Request: expose clusters as objects
863 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
864 - #255: Tesh broken on Windows
865 - #256: Modernize FindSimGrid.cmake
866 - #257: Fix (ab)use of CMake install
868 ----------------------------------------------------------------------------
870 SimGrid (3.18) Released December 24 2017
872 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
874 MSG IS NOW DEPRECATED. Please use S4U for new projects.
875 - Support for MSG should not be removed from SimGrid before 2020,
876 but future evolutions will be limited to the integration of
877 user-provided patches.
880 - s4u::onDeadlock() signal, run before stopping the simulation.
881 - s4u::Exec: asynchronous executions (abstraction of background
883 - s4u::Host->getLoad() returns the achieved speed in flops/s
886 - Link::name() is deprecated and will be removed in v3.21, use
887 Link::getCname() instead.
888 - Mailbox::getName() changed to return a std::string, use
889 Mailbox::getCname() instead to get a char*.
890 - Storage::getName() changed to return a std::string, use
891 Storage::getCname() instead to get a char*.
892 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
893 to have both allocation and deallocation of the map in user space.
896 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
897 The amount of flops remaining to do is a vector, not a scalar.
898 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
899 name implies on both sequential and parallel tasks.
900 - Both changes fix GitHub's #223 using PR #237 as a basis.
901 Thanks Michael Mercier.
902 - Most examples were converted to S4U and hidden elsewhere at tests.
903 Rationale: we still want MSG to work; we want newcomers to use S4U.
906 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
907 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
910 - Switch to the faster dlopen privatization mechanism by default
911 - Documentation improvments
914 - Live migration is getting moved to a plugin. Dirty page tracking is
915 the first part of this plugin. This imply that VM migration is now
916 only possible if one this function is called:
917 - C/MSG: MSG_vm_live_migration_plugin_init()
918 - C/C++: sg_vm_live_migration_plugin_init()
919 - Java: Msg.liveMigrationInit()
920 For C and C++, "simgrid/plugins/live_migration.h" also has to be
924 - Define class simgrid::xbt::Path to manage file names.
925 - Removed unused functions:
926 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
927 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
928 xbt_graph_export_graphviz()
929 - xbt/str.h: xbt_str_join()
930 - Remove unused datatypes:
931 - xbt/heap.h: use std::priority_queue or boost::heap instead
932 - xbt/swag.h: use boost::intrusive::list instead
935 - New link_energy plugin for the consumption of the links.
936 - All of the operations on files and storage contents have been
937 packaged into a plugin (src/plugins/file_system). The current
938 public interface can be found in
939 include/simgrid/plugins/file_system.h
940 To use these functions you now have to initialize the plugin by
941 calling MSG_storage_file_system_init() just after calling
942 MSG_init() or sg_storage_file_system_init() just after creating
946 - Remove the undocumented/untested tag <include>
949 - Remove viva specific tracing as the tool is no longer maintained
952 - #248: Kill of finished processes leads to segfault
953 - #240: xbt_cond_wait_timeout should gracefully return for C apps
954 - #239: please implement signal s4u::onDeadlock()
955 - #230: segfaults when exit() without run()
956 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
957 - #223: MSG_task_get_flops_amount() not working with parallel tasks
958 - #222: Actor::kill() doesn't really kill and segfaults
959 - #221: odd LMM warning when killing an actor
960 - #120: Memory leak when the processes are forcefully killed
962 ----------------------------------------------------------------------------
964 SimGrid (3.17) Released October 8 2017
966 The Drained Leaks release: (almost) no known leaks despite the tests.
968 Even very long simulations will preserve your memory: our testsuite
969 stresses SimGrid for over 45mn on fast machines for a coverage of
970 over 80%, and there is only one single known leak, of about 4kb.
974 - Comm.detach(): start and forget about asynchronous emission. A cleanup
975 handler may be given to free resources if the comm cannot be completed.
976 - this_actor::send(mailbox) is now mailbox->put()
977 - New: simgrid::s4u::Comm::wait_all()
978 - New: Host.execute() for remote executions.
981 - Unused option network/sender-gap is removed.
984 - Deprecate MSG_task_isend_with_matching(): This unused feature
985 really complicates our internals. Will be removed in v3.20.
988 - Improved context termination. It is now possible to free resources, even
989 when a process is forcibly killed.
990 - ContextBoost: add support for Boost versions above 1.61.
993 - Bring back run-time option --cfg=exception/cutpath to remove exception
995 - Removed unused functions:
996 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
997 xbt_str_rtrim(), xbt_str_trim().
998 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1001 - Removed header files obsolete since SimGrid 3.12:
1002 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1004 - #3: SD_exit should be made optional
1005 - #120: Memory leak when the processes are forcefully killed
1006 - #159: Threading test regression in Actor refcounting
1007 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1008 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1009 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1010 - #191: VM migration and pstate
1011 - #192: Updating the energy consumptions of all hosts crashes with VMs
1012 - #195: All actors have PID=0 in the logs
1013 - #204: Sometimes segfault with thread contexts and mmap privatization
1014 - #222: Actor::kill() doesn't really kill and segfaults
1015 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1017 ----------------------------------------------------------------------------
1019 SimGrid (3.16) Released June 22. 2017.
1021 The Blooming Spring Release: developments are budding.
1024 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1025 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1026 - Java JAR file should be finally fully working on Mac OSX too.
1028 XML platforms: Switch to platform v4.1 format.
1029 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1030 - <zone> can be used as a synonym for the now deprecated <as>
1031 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1032 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1033 - <actor> can be used as a synonym for the now deprecated <process>
1034 - state_file and avail_file periodicity is now easier to express
1035 (check the documentation) the old behavior should still work.
1036 * Storage: not backward compatible, but it seems that nobody ever used it
1037 - Remove attribute 'content_type' of <storage_type>: was never used
1038 - Make attribute 'model' of <storage_type> optional: for future usage
1039 - Remove Bconnection model property: was never (in)validated. Replaced by
1040 the maximum of the read and write bandwidth as a resource constraint.
1043 - New and Backwards Compatibility break:
1044 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1045 When one wants to get the list of tasks whose states have changed during a
1046 simulation round, s/he has to allocate and free a dynar and use it as
1047 argument to this function. The former SD_simulate (double how_long)
1051 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1052 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1055 - New function to update the consumption of all hosts at once.
1056 - Fix the model for multi-core hosts, linear in the amount of busy
1057 cores with an abnormality for pIdle (see paper).
1058 BUG: Still not satisfactory for ptask on mono-cores.
1061 - The netzone are now available from the MSG API.
1062 The old names still work, but are now deprecated.
1065 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1066 - New option: smpi/keep-temps to not cleanup temp files
1067 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1068 mallocs mode is used, allows to change the size of the fake file used
1069 (default 1MB), to potentially limit the number of mappings for large runs.
1070 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1071 - Fortran ifort and flang compilers support
1072 - New RMA calls supported (experimental):
1073 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1074 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1075 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1076 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1077 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1078 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1081 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1082 - Behind the scene, ActivityImpl are now automatically refcounted.
1083 Removing this nasty bottleneck will greatly simplify our internals.
1086 - Replay: New function xbt_replay_action_get():
1087 Retrieve the function previously associated to an event type.
1088 - DROPPED FUNCTION: xbt_str_varsubst()
1089 - DROPPED MODULE: strbuff. We don't need it anymore.
1090 - DROPPED MODULE: matrix. We don't need it anymore.
1091 - DROPPED MODULE: lib. We don't need it anymore.
1093 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1095 SimGrid (3.15) Released March 22. 2017
1097 The Spring Release: continuous integration servers become green
1098 We fixed even the transient bugs on all target architectures:
1099 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1100 Windows, FreeBSD, NetBSD.
1102 - Rename NetCards to NetPoints.
1103 This was intended to help NS3 users, but that's not a netcard.
1104 That's a point in the routing algorithm, let's avoid wrong simplifications.
1107 - New: MSG_process_yield(). Stop and yield to other processes.
1108 - New: MSG_process_daemon(). Daemon processes are automatically killed
1109 when the last non-daemon process terminates
1110 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1111 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1112 to make room for the upcoming network energy plugin.
1113 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1116 - Ensure that an actor can kill itself with Process::exit()
1117 - Kill the obscure NativeException. Nobody want to survive the issues
1118 it denotes, so use JniException that is a RuntimeException (not to
1119 be caught explicitly).
1120 - Partial bug fix in initialization. SimGrid flags on command line were
1121 consumed at C level but stayed in the original Java String[] args.
1122 This could mess users' args[i] if SG flags were not put at the end of
1124 The SimGrid flags are now removed from the Java arguments. However,
1125 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1126 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1127 It might be if you have --log or --cfg flags in the command line.
1128 - Fix numerous memleaks all around the place. In particular, around VMs.
1132 - s4u::onPlatformCreated: right before the simulation starts
1133 - s4u::onSimulationEnd: right after the main simulation loop
1134 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1135 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1136 event from the availability_file changes the avail speed.
1137 - Links are now usable from s4u
1138 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1139 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1140 - New examples: The conversion of MSG examples to S4U has begun
1141 - Actors: create, kill, migration, and suspend
1142 - Applications: master-worker and token-ring
1143 - Action replay: communications and storage
1144 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1147 - Backwards Compatibility breaks
1148 - The SD_route_get_list and SD_route_get_size functions have been removed.
1149 They are replaced by the unique
1150 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1151 The route from src to dst is built in the links dynar whose size can
1152 be obtained with xbt_dynar_length.
1153 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1154 sg_host_route_bandwidth and sg_host_route_latency.
1155 Macros ensure the backwards compatibility, but you should fix your code
1158 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1159 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1160 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1161 We don't support 100% of it, but it should be enough. Contact us if not.
1162 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1163 naive and probably inaccurate)
1164 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1165 mapped onto a unique small file using some system magic.
1166 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1167 - Fortran: cleanups, fixes, support of user-added operations
1168 - RMA: MPI_Accumulate are applied in correct order.
1169 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1170 - Better support for MPI_IN_PLACE
1171 - Support for MPI_Win attrs and keyvals.
1172 - Support MPI_Comm_group_create, MPI_Type_size_x
1173 - MPI ops should be properly applied to all allowed types
1174 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1175 algorithm by a non failing one from mpich.
1178 - Kill the fifo data container: we don't use it anymore.
1181 - A <cluster> can now be created with different speed values to represent pstates.
1182 see examples/platforms/energy_cluster.xml for an example.
1184 SimGrid (3.14.159) Released December 28. 2016
1186 The previous release was only a crude approximation of a Pi release;
1187 this one should be a bit better.
1189 - Revert a last minute change that broke on Mac OSX.
1190 - Fix the changelog and NEWS files.
1191 - Improve the documentation, in particular of the routing module.
1192 - Integrate some patches flying around in Debian.
1194 SimGrid (3.14) Released December 24. 2016
1197 * The whole documentation was reworked and reorganized. There is
1198 still much room for improvement here, but we're on it.
1201 * We now need python3 for our testing infrastructure.
1202 * Model checking needs libevent
1205 * Now works on FreeBSD
1208 * Privatization now works on FreeBSD
1210 * Privatization is now activated by default in smpirun
1211 You can turn it off with -no-privatize if something goes wrong.
1213 * Call-location tracing for SMPI.
1214 You can add the exact location (filename / linenumber) of an MPI call to
1215 your trace files and slow down or speed up the simulation between two
1216 consecutive calls by using an adjustment file (see the documentation).
1218 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1219 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1220 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1221 values were ignored in some cases. The timings of these functions can now
1222 be significantly different.
1224 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1225 smpi/running-power is renamed to smpi/host-speed
1227 * smpi/grow-injected-times option to enable or disable multiplication of the
1228 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1229 make simulation less precise (but also much faster).
1231 * smpirun script should be (much) faster for large deployments.
1233 * SMPI tracing: fixed issue with poor matching of send/receives.
1235 * Replay: Fix broken waitall
1237 New functions and features
1238 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1240 Dropped / renamed functions and features
1241 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1242 - MSG_mailbox_is_empty() -> Mailbox::empty()
1243 - MSG_mailbox_front() -> Mailbox::front()
1244 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1245 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1246 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1247 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1248 * Don't pass the free_f parameter to property related functions:
1249 - MSG_host_set_property_value()
1250 - MSG_as_router_set_property_value()
1251 - MSG_storage_set_property_value()
1252 * VM properties. Since msg_vm_t are msg_host_t, just use
1253 MSG_host_get_property_value() and friends
1254 * VM I/O related things:
1255 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1256 - Unimplemented save/restore methods
1257 * MSG_as_router_get_property_value() was redundant with
1258 MSG_environment_as_get_property_value().
1259 - Removed MSG_as_router_*propert*() functions
1260 - Added MSG_environment_as_set_property_value() for consistency
1261 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1262 Well, they are still there for now, but deprecated with a warning.
1263 Please switch to xbt_dict_new_homogeneous() before this is removed
1265 * Task affinity. Its intended behavior (that was very badly tested
1266 and probably not really working) was deceiving what most users
1267 would have hoped here.
1268 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1269 * Option network/coordinates is now useless and should be dropped.
1272 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1275 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1276 * Added Dragonfly topology support
1279 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1280 EiBps, EBps, Eibps, Ebps for bandwidth.
1281 They may become useful to some lucky ones.
1284 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1285 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1286 * ex/async_waitAll: new example, on asynchronous communications
1289 * Memory usage should be decreased for simulations with a large number
1290 of processes. This also helps for SMPI.
1292 SimGrid (3.13) stable; urgency=low
1294 The Half Release, a.k.a. the Zealous Easter Trim.
1296 - We removed half of the lines, that were mostly experimental cruft.
1297 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1298 (not counting blanks and comments -- according to openhub.net)
1299 - The internals are now compiled in C++ (and will soon be clean C++)
1300 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1301 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1303 * Backwards Compatibility breaks
1304 - Removed Lua simulation bindings (switch to C or Java for that).
1305 Lua can still be used to describe platforms
1306 - Removed Java kernel plug-ins.
1307 Will be reintroduced after the ongoing major internals reorg.
1309 - the following functions were removed.
1310 They were too specific and should be reimplemented in a generic
1311 way, with filter function.
1312 - MSG_task_listen_from_host
1313 - MSG_mailbox_get_count_host_waiting_tasks
1314 - MSG_mailbox_put_with_timeout was removed.
1315 Please use MSG_task_send_with_timeout instead.
1317 - the SD_application_reinit function was removed. It has been a noop for a while.
1318 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1319 replaced by a more flexible mechanism.
1320 - The following functions thus do not exist anymore
1321 - SD_workstation_get_access_mode
1322 - SD_workstation_set_access_mode
1323 - SD_workstation_get_current_task
1324 - Basic estimation functions have been removed but can easily be replaced
1325 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1326 - SD_workstation_get_computation_time => amount / sg_host_speed()
1328 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1329 Use VM.getSpeed()*load/100 for the legacy behavior.
1331 - option enable_tracing was removed. It was not doing anything for a while.
1332 - In the ModelChecker:
1333 - the model-checker now ptraces the model-checked process which means
1334 you cannot use a debugger on the latter anymore (we might make this
1335 optional in the feature);
1336 - removed soft-dirty page tracking;
1337 - remove model-checked side snapshot management,
1338 MC_snapshot() and MC_compare_snapshot();
1339 - keep the MC_cut() function as a stub (it was not really working
1340 in the previous release).
1343 * All options are consistently in kebab-case. Old names are kept as alias.
1346 * Switch to platform v4 format.
1347 - Rename from 'power' to 'speed' the attributes describing the amount of
1348 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1349 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1350 - In <host> and <link>, attributes availability and state are gone.
1351 It was redundant with state and availability traces, and with peak values.
1352 - In <cluster>, attributes availability_file and state_file are gone.
1353 It was too complex and unused.
1354 - Kill <gpu>. Was not doing anything.
1355 - The DOCTYPE points to the right URL:
1356 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1357 (the file at this address now documents the changelog since its v1)
1358 - A warning is emitted for unit-less values (they are still accepted).
1359 - speed. Default: 'f' or 'flops'. Also defined:
1360 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1361 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1362 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1363 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1364 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1365 - latency. Default: 's' second. Also defined:
1366 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1368 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1369 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1372 * s4u::Host is now the preferred public interface to the Host features.
1373 sg_host_* functions are C bindings to the exact same behavior
1374 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1377 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1378 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1379 They were deprecated since a few years, and probably did not even compile anymore.
1382 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1383 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1384 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1385 encouraged to update their codes. A list of the modified functions can be found at the end of
1386 include/simgrid/simdag.h
1389 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1390 the lambda or closure passed as a parameter will run in kernel mode.
1391 All the callback functions should be rewritten to that interface at some point.
1394 * Reorganizing and cleaning the internals all around the place.
1397 * Remove old default barrier/bcast buggy algorithms (see #18407)
1398 * Various bug fixes to handle more codes
1399 * Remove the need for the --foreground option of smpirun (it is still
1400 accepted for backward compatibility).
1403 * Kill the setset data container: MC don't use it anymore.
1404 * Kill the queue data container: it made more sense with GRAS.
1405 * Kill the xbt_peer_t data type: it's useless without GRAS.
1406 * Kill rm_cb feature of config sets: it was never useful.
1407 * Kill graphxml parsing feature. It was not used.
1408 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1410 - xbt_dynar_sort_strings(), when the content is char*
1411 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1412 They throw exceptions on invalid input;
1413 * C++ support for declaring CLI flags (simgrid::config::Flag);
1414 * class for abstracting different signal backends (simgrid::xbt::signal).
1415 with no external dependencies (we need very simple signals).
1418 * refactoring and cleanup of the code;
1419 * ongoing process to cleanly separate the model-checking algorithms
1420 from the code model-checking support.
1422 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1424 SimGrid (3.12) stable; urgency=low
1426 The Facelift Release.
1429 * Require g++ v4.7 at least to not speak prehistorical C++.
1430 * Require Boost 1.48 (for signal2 component).
1431 * Java must be version 7 at least when activated.
1432 * Builds on Windows again (including Java bindings).
1433 * Tracing is now always enabled (no way to turn it out)
1434 * Remove GTNetS. It was not working anyway.
1435 * Various cleanups in the cmake scripts.
1436 * Move headers around to sort them out on installed systems:
1437 - instr/instr.h -> simgrid/instr.h
1438 - instr/jedule/* -> simgrid/jedule
1439 - simdag/datatypes.h was removed
1440 - simdag/simdag.h -> simgrid/simdag.h
1441 - msg/datatypes.h was removed
1442 - msg/msg.h -> simgrid/msg.h
1445 * Interface improvement:
1446 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1447 - Sanitize the interface in MSG_task_ module:
1448 - Merge two functions that were close enough but misleading:
1449 set_compute_duration(t) -> set_flops_amount(t)
1450 get_remaining_computation(t) -> get_flops_amount(t)
1451 - set_data_size(t) -> set_bytes_amount(t)
1452 get_data_size(t) -> get_bytes_amount(t)
1453 - Massive cleanups in the functions related to the energy
1454 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1455 - New: MSG_host_get_pstate()
1457 - msg/energy/onoff: switching hosts on and off
1459 * Interface improvement:
1460 - Rename Host.isAvail() to Host.isOn()
1461 - Rename Process.currentProcess() to Process.getCurrentProcess()
1462 - Rename Task.setDataSize() to Task.setBytesAmount()
1463 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1465 - #18874: Actually allows the GC to reclaim tasks
1469 - SIMIX_process_throw: raises an exception in a remote process
1470 * Refactoring: Separate sub-modules
1471 - libsmx: the public interface, as libc in a real system
1472 - popping: the strange dance that converts a user request into a kernel handling
1473 - smx_context_*: the virtualization mechanisms that embed the user code
1474 - smx_*: the handling of each simcalls
1475 * Interface cleanup:
1476 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1477 * Rename smx_action_t into smx_synchro_t, making explicit that these
1478 things are used to synchronize processes with their environment.
1479 For example, a communication is a sort of synchronization involving
1480 the communicating processes (that may block until the exchange) and
1481 the platform. The same can be said from computations, etc.
1483 - #18888: segfault when a process ends before its kill_time
1487 - 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.
1488 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1489 - MPI_Type_set_name, MPI_Type_get_name
1490 - MPI_*_c2f and MPI_*_f2c functions
1491 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1492 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1493 - Activate a lot of new tests from the mpich 3 testsuite
1495 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1496 - InfiniBand network model added: Based on the works of Jerome Vienne
1497 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1498 - When smpi/display_timing is set, also display global simulation time and application times
1499 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1500 * Collective communications
1501 - 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
1502 - 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/).
1503 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1505 - Replay now uses algorithms from wanted collective selector
1506 - Replay can be used with SMP-aware algorithms
1507 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1508 - 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.
1510 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1511 - Correctly use loopback on fat-tree clusters
1512 - Asynchronous small messages shouldn't trigger deadlocks anymore
1513 * Energy/DVFS cleanup and improvement
1514 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1515 - new: smpi_get_host_pstate()
1518 - "Full" network optimization flag was broken since Surf++
1519 - Better handling of precision flags in maxmin
1520 - Fix bug causing sometimes "Impossible" errors
1521 - Properly pass cluster properties to included hosts
1522 * Improvement of the Energy plugin.
1523 - Always update the consumption before returning that value
1524 - New property: watt_off to denote the dissipation when the host is off
1525 - New functions getWattMinAt and getWattMaxAt to retrieve the
1526 dissipation of pstates that we are not currently at.
1527 * Java: class NetworkLink renamed to Link
1528 * New function: simcall_process_get_kill_time()
1529 * Massive rename s/workstation/host/
1530 - That's intrusive, but that's good for the project consistency. Sorry.
1531 - Change config option "workstation/model" into "host/model"
1534 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1535 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1536 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1537 - xbt_dirname and xbt_basename for non-POSIX systems
1539 * The model checker now runs as a separate process.
1540 * The model checker runs must now be launched with the new simgrid-mc program.
1541 * Record/Replay: the MC can display a textual representation of a path in the
1542 execution graph. It can then be replayed outside of the model checker.
1544 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1546 SimGrid (3.11) stable; urgency=low
1551 * Normalizing pointers addresses tool for better diff between logs
1554 * Add cloud examples using new VMs
1555 - examples/msg/cloud/two_tasks_vm.tesh
1556 - examples/msg/cloud/simple_vm.tesh
1557 - examples/java/cloud/cloud.tesh
1558 - examples/java/cloud/migration/migration.tesh
1559 * Add java surf examples:
1560 - examples/java/surfPlugin/surf_plugin.tesh
1561 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1562 - examples/java/surfCpuModel/surf_cpu_model.tesh
1563 * Add SMPI+MSG example:
1564 - examples/smpi/smpi_msg_masterslave/
1569 - msg task destroy cancel test
1570 - msg_host on/off test
1571 * Move all tests in testsuite to teshsuite (adding tesh files)
1572 * Restructure teshsuites
1573 - one folder for each kind of test
1574 * Restructure AddTests.cmake
1576 - structure the order of tests (with sections)
1579 * Add virtual machine
1580 - creation of a VM on a PM
1581 - migration of a VM from a PM to another PM
1583 - MSG_process_join(msg_process_t process, double timeout)
1584 - msg_bar_t MSG_barrier_init(unsigned int count)
1585 - int MSG_barrier_wait(msg_bar_t barrier)
1586 - void MSG_barrier_destroy(msg_bar_t barrier)
1587 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1588 * New option "msg/debug_multiple_use" to help debugging when a task is used
1591 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1593 - Increase code coverage in test suites
1597 * Protect context stacks against stack overflow. The number of protected memory
1598 pages allocated on the top of each stack (1 by default) can be configured
1599 with the new command line option --cfg=contexts/guard_size:<value>.
1600 * Simcalls are now generated by a python script that
1601 - generates files included by SimGrid
1602 - checks that all the functions exist, and proposes prototypes
1604 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1605 host_set_data, host_get_data
1607 - simcall_process_join(smx_process_t process, double timeout)
1608 * Fix bug where sleeping processing could not be suspended.
1611 * Translate surf models from C to C++
1612 - Generic classes for all models: Model, Resource, Action
1613 - A generic interface for each kind of model (CPU, Network, Storage
1614 Workstation, WorkstationVM)
1616 * Translate surf routings from C to C++
1617 * Add callbacks using sigc++ or boost::signals2
1618 - Add callback functions for resource creation/destruction
1619 - Add callback functions for action state change
1620 - Handle Energy as a plugin
1621 * Replace swag by boost::intrusive
1622 * Add new routing models for clusters. For documentation, see
1623 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1624 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1625 parameters for cluster tag
1626 - Fat trees, with topology="FAT_TREE" and
1627 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1628 - see examples/platforms/torus_cluster.xml and
1629 examples/platforms/fat_tree_cluster.xml
1630 * More documentation
1633 * Hostfiles support host:nb_processes construct to deploy several processes on
1635 * Collective communication algorithms should not crash if used with
1636 improper number of nodes and report the error.
1637 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1638 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1639 MPI_Dims_create, MPI_Cart_sub are supported.
1640 * New interface to use SMPI programmatically (still depends on MSG for
1641 some parts, see examples/smpi/smpi_msg_masterslave):
1642 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1646 * Global variables privatization in MPI executables is now performed at runtime
1647 with the option smpi/privatize_global_variables (default:no).
1648 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1649 dynamic libraries loaded by the application are not privatized (static
1650 linking with these libraries is advised in this case)
1653 * Options defined in XML work correctly now.
1656 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1657 copied into simgrid.jar or not (ON by default). Use this option if you want
1658 to reduce the size of the installed simgrid.jar, *and* the native libraries
1659 are kept installed elsewhere.
1660 * Surf binding with SWIG (code generated in maintainer mode only):
1661 - plugin to handle callbacks
1662 - CPU model only for the moment
1665 * Supernovae build mode is definitively removed. It was used to improve
1666 inlining and inter-module optimizations. It is nowadays superseded by
1667 link-time optimizations commonly available in compilers.
1668 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1669 * Add boost dependency for surf++
1670 * Add new macro for tests
1671 - ADD_TESH(name <tesh_args>)
1672 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1676 - xbt_bar_t XBT_barrier_init(unsigned int count)
1677 - int XBT_barrier_wait(xbt_bar_t barrier)
1678 - void XBT_barrier_destroy(xbt_bar_t barrier)
1679 * Make the xbt_os_time module public
1681 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1683 SimGrid (3.10) stable; urgency=low
1685 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1688 * Reintegrate Java to the main archive as desynchronizing these
1689 package is not acceptable anymore (Java is now considered stable)
1690 * Add explicit synchronization facilities through semaphores
1691 * Bug fix: Task.setDataSize() only changed the C world, not the value
1692 cached in the Java world
1695 * Dramatically change the way files are handled. API and internals changed, but
1696 this part of MSG was not considered as production grade either.
1697 * Add explicit synchronization facilities through semaphores
1698 * Add a new function MSG_host_get_process_list()
1699 * Preliminary DVFS support (see examples/msg/energy/ for details)
1702 * SMPI is now included directly in the libsimgrid as the windows
1703 linker doesn't force us on splitting it anymore.
1704 * Improvements of the SMPI replay tool:
1705 - Most of the collective communications are now rooted in the same process as
1706 in the original application.
1707 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1708 used until now). Multiple data types can now be used in a trace.
1709 - The replay tool now supports traces produce either by TAU or a modified
1711 - Bug Fix: the compute part of the reduce action is now taken into account.
1712 - Gatherv collective is now supported
1713 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1714 of smpirun outputs time independent traces for the current run. One file
1715 is created per process. If too many processes are simulated, this behavior
1716 can be changed to one file for all processes by using the
1717 tracing/smpi/format/ti_one_file flag
1718 * smpirun generates the host file if needed (with given host count and platform)
1719 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1720 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1721 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1722 --cfg=smpi/coll_selector:(mpich/ompi)
1723 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1725 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1726 happening inside a collective SMPI call.
1727 * Fix the behavior of complex data types handling.
1728 * Make MPI_Wtime another synchronization point to take computations into
1730 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1731 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1732 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1733 (no privatization of global variables yet)
1734 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1737 * Verification of liveness properties is now available for SMPI applications
1738 (in addition to MSG applications)
1739 * Bugged examples using SMPI in examples/smpi/mc/
1740 * Add --cfg=model-check/visited option. Allows the verification of infinite
1741 programs. Detection of loops in the execution thanks to the system state
1742 comparison and reduction of the state space to explore. Can be combined with
1743 DPOR for safety properties.
1746 * Allow to change SimGrid configuration (see --help) within the code
1747 thanks to SD_config() as it can be done in MSG.
1748 * Add a new function SD_task_set_amount() upon user request.
1751 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1752 * Remove rule based routing (no more PCRE dependency)
1753 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1754 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1755 * Add a loopback_bw and loopback_lat options to cluster tag.
1756 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1757 routing. Now peers are encapsulated in an AS and have their own private
1758 router but this is transparent.
1761 * Our own implementation of getline is renamed xbt_getline, and gets
1762 used even if the OS provide a getline(). This should reduce the
1763 configuration complexity by using the same code on all platforms.
1764 * New type: xbt_cfg_elm_boolean.
1765 * Allow to use yes/no for boolean configuration options in the command line.
1766 * Allow to disable SimGrid cleanups at exit from command line option.
1767 There are situations where one may want a simulation to end with an exit.
1768 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1769 annoying when scripting around the simulator. Adding a
1770 --cfg=clean_atexit:no allows to circumvent this issue.
1773 * Lots of memory leaks were corrected in this release.
1774 There is no known memory leaks anymore, in all of our 600+ tests.
1775 * New command line option --version, to get SimGrid version information.
1776 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1778 * Supernovae builds are deprecated, and expected to be removed in the next
1781 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1783 SimGrid (3.9) stable; urgency=low
1785 The Grasgory release: GRAS is really dead now.
1787 * Complete overhaul of the internal host structures scheme.
1790 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1791 was considered as experimental and badly maintained since then.
1792 * Keeping it was thus a trap to our potential users, that could take
1793 it instead of MSG or SMPI by mistake despite is pity state.
1794 * GRAS seems to have very few users (if any), and no one volunteered
1795 to maintain it further. It also induces a lot of XBT code (for
1796 portability sake), that must be maintained too.
1797 * For all these reasons, we killed GRAS. If someone wants to revive it
1798 in the future, don't cry, our git history still remembers of GRAS.
1801 * Major overhaul. Merge our documentation again as time proved that
1802 splitting it was really not helping our users.
1803 * Further improve the developer documentation to help newcomers
1804 hacking on SimGrid itself. The user documentation (and in
1805 particular, the beginner documentation) is still in a sorry state.
1808 * Now works on Windows too!
1809 * Much more extensive test suite, from MPICH
1812 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1813 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1814 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1815 * Bug fix: let task be scheduled when the last dependency to be solved is
1816 a control dependency.
1817 * Remove SD_load_environment_script function.
1818 Use the C sg_platf function if you want to declare a platform
1822 * New function: MSG_process_get_number()
1823 * Old function documented: MSG_config()
1824 * Remove MSG_load_platform_script function
1825 Use the C sg_platf function if you want to declare a platform
1829 * Change the default value of the TCP_gamma constant (maximal size of TCP
1830 congestion window) to a more realistic 4MiB value. If you notice changes in
1831 your simulation results, you can fall back to the previous 20k tiny window
1832 by adding --cfg=network/TCP_gamma:20000 on command line.
1833 * (Hopefully) fix a bug wrt periodic availability and state traces
1834 * Bug fix: use default values at start when first event in availability/state
1835 trace is not at time 0.
1838 * remove the "new_" part of function name sg_platf_new_trace_connect
1839 (resulting in sg_platf_trace_connect), since it does not create
1843 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1844 use case where it's really mandatory, and maintaining it was a pain in
1846 * New: xbt_fifo_search(), search an item with a user-provided
1847 comparison function instead of dumb pointer comparison.
1850 * Fix the lua deployment:
1851 Use `simgrid.init_application()` before deployment instead of
1852 `simgrid.msg_register_application()` after.
1855 * Transfer the tracing files into the corresponding modules.
1857 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1859 SimGrid (3.8.1) stable; urgency=low
1861 The "we are told that some people want to also *install* the simgrid
1864 * Add missing file "tesh.1" to the archive.
1866 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1868 SimGrid (3.8) stable; urgency=low
1870 The Psssshiiiit release: SimGrid jumps into the Cloud.
1873 * Add an experimental interface to manipulate VMs. They are mainly
1874 process groups with very few intrinsic semantic, but they should
1875 allow you to build the semantic you want easily.
1876 * New function: MSG_host_set_property_value()
1877 * New function: MSG_process_on_exit(). To clean memory in all cases.
1878 * Bug fixes that made the host (and link) failures unusable.
1879 * Add a way to auto-restart process when the host in which they are
1880 executing comes back (ON_FAILURE="RESTART" on deployment file,
1881 MSG_process_auto_restart_set).
1882 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1883 please stop using the old ones, they are DEPRECATED.
1885 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1886 Please use MSG_init() instead. (reducing the amount of entry
1887 points in the library helps us).
1888 * Make it impossible to link against the wrong version of the lib
1889 * Deprecate MSG_clean(). No need to call it anymore.
1890 * Function MSG_get_host_number() is not deprecated anymore.
1893 * Split the doc into a user guide and a reference guide.
1894 * Start a developper guide to help people hacking on SimGrid.
1897 * Enable tracing by default. This modules rocks you should use it.
1898 * Remove option custom_flags. Now use environment variables CFLAGS
1900 * Use default cmake things to detect lua instead of home grown ones.
1901 * New option "enable_mallocators" to disable mallocators, for debugging
1902 purpose ("on" by default).
1905 * Bug fixes around the resource failures: don't let the processes
1906 survive the host they are running onto.
1907 * Add an interface to auto-restart processes when the host in which they are
1908 executing comes back.
1909 * Ensures that SIMIX_clean is called automatically. It's not part of
1910 the public interface anymore (bindings should be updated).
1913 * Bug fix for when SD_Simulate is called with a positive value: be careful
1914 when comparing doubles. Sometimes they are different for non significant
1916 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1917 parallel task whose initial work is distributed among host according
1918 to the Amdahl's law. Such tasks are created with a parameter alpha
1919 that corresponds to the non-parallelizable part of the computation.
1920 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1921 between two sets of workstations assuming a 1D block distribution (each
1922 workstation owns a similar share of data) on both sides.
1924 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1925 Data redistribution will be automatically scheduled once parent and child
1926 are both scheduled. The filling of computation_amount and
1927 communication_amount structures is now done seamlessly thanks to the chosen
1929 * New function SD_workstation_dump to display various information
1930 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1931 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1932 bandwidth on the route onto which the task is scheduled and the amount of
1934 To divide the nominal bandwidth by 2, the rate then has to be:
1935 rate = bandwidth/(2*amount)
1936 * Compute tasks that have failed can now be rescheduled and executed again
1937 (from their beginning)
1938 * Increasing source code coverage (src/simdag is now covered at 95.8%
1942 * Re-implement time-independent trace replay using SMPI (at the
1943 smpi_smp_* level) instead of MSG. This should replace
1944 examples/msg/actions/actions.c
1945 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1946 hindexed and structs)
1947 * Implement the exchange of non-contiguous data.
1948 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1949 * Correct behavior of smpi/sender_gap and set its default value to 0
1950 * Add option to asynchronously send small messages to allow better
1951 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1952 specifies the size in bytes under which messages will be asynchronously sent.
1953 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1954 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1957 * New C interface to define a platform: XML is now optional.
1958 For more info, please check include/simgrid/platf.h.
1959 * New interface to define random platforms from the C:
1960 For more info, please check include/simgrid/platf_generator.h and
1961 examples/msg/masterslave/masterslave_platfgen.c
1962 * Export a sg_cmdline dynar containing all the arguments we got from
1966 * Two new tracing options for adding comments to trace file so you
1967 can track your experiments (see --help-tracing for details).
1968 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
1969 * Adding the SimGrid version that generated the trace file as a comment.
1970 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1971 * Fix to avoid key clashes on Paje links
1972 * Other minor fixes related to the Paje specification
1975 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1976 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1977 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1978 * Greatly improve the robustness of mmalloc to user errors (such as
1979 using an area after freeing it, or freeing it twice)
1981 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1983 SimGrid-java (3.8.1) stable; urgency=low
1985 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1986 Warning: all methods are visible, but only some of them are
1987 implemented so far. Check the source (src/jtrace.c)
1988 for further information.
1989 * New module: org.simgrid.msg.File (SimGrid File management functions)
1990 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1991 * Change the meaning of Process.restart: now restart the process from
1992 the beginning, like MSG_process_restart in C.
1993 * Add Process.setAutoRestart: handling of process restart when failed
1995 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1996 you to retrieve the properties of the processes/hosts
1997 * Deprecate Msg.clean(): you can just forget about it now.
1998 * New function Process.getCount(), that only works when compiling
1999 with the not yet released version 3.9 of the C library.
2001 * New context factory based on Coroutines. It mandates a modified JVM
2002 but then, the simulations run about five times faster, and there is
2003 no limit to the amount of processes (beside of the available memory).
2005 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2007 SimGrid (3.7.1) stable; urgency=low
2010 * Restore the prototype of MSG_process_create_with_environment() to
2011 the pre-3.7 situation by removing the kill_time argument.
2012 * Add a MSG_process_set_kill_time() function instead.
2015 * Fix weird behaviors when dealing with parallel tasks.
2018 * Simgrid is now built as a dll.
2019 * Simgrid-java now works on Windows.
2020 * Simgrid-Java is now included into Windows package.
2023 * First pre-build package for MacOSX.
2026 * Fix compilation when using MSG_USE_DEPRECATED.
2027 * Fix some compilation issues on Macs and Windows.
2028 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2029 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2031 We discovered that the Lua console is broken, but we are missing the
2032 manpower to fix it right now. The problem existed in 3.7 too, so we
2033 are not blocking the release for that. Sorry if you depended on this
2034 feature, any help would be really welcome.
2036 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2039 SimGrid-java (3.7.1) stable; urgency=low
2041 The "Java aint got to be bloated and slow" release
2044 * Various internal cleanups and performance improvement
2045 Simulations are expected to run up to twice faster or so
2046 * Make Process.kill(process) an instance method, not a static one
2047 * User processes are not java.lang.Thread subclasses.
2048 This breaks the compatibility (sorry), but previous API was
2049 brain-dead, making it impossible to have non-trivial
2050 initializations in the process constructor.
2051 * Require a full constructor per Process sub-class.
2052 Kinda breaks the compatibility (sorry), but this allows a much more
2053 efficient way to launch the processes at simulation startup.
2054 * Do not embeed our version of semaphores, java 1.5 can be considered
2055 as sufficiently prevalent for us to not dupplicate its features.
2059 * Add examples for almost every part of the API
2060 We spotted and fixed a lot of bugs in the process
2062 * New module: asynchronous communication API
2063 * New function: Process.sleep()
2064 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2065 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2066 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2068 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2070 SimGrid (3.7) stable; urgency=low
2072 The "spring cleanups (before the next Big Project kicks in)" release.
2075 * We can specify the SMPI latency/bandwidth factor with command line
2076 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2077 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2078 You can also use the "config tag" from platform file by adding this line
2079 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2080 (see "example/platforms/tag_config.xml" to use "config tag").
2081 Note that the command line supersedes the platform file configuration.
2082 * Change the correction factors used in LMM model, according to
2083 the latest experiments described in INRIA RR-7821.
2084 Accuracy should be improved this way.
2085 * Use the partial invalidation optimization by default for the
2086 network too. Should produce the exact same results, only faster.
2087 * Major cleanup in surf to merge models and split some optimization
2088 mechanisms from the core of the models. As a result you can now
2089 specify which model to use (e.g., --cfg=network/model:LV08
2090 --cfg=cpu/model:Cas01) and which optimization mode to use
2091 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2092 Incompatible combinations should err at initialization. See
2093 --help-models for the list of all models and optimization modes.
2094 * The CLM03 workstation model was dropped for simplicity because it
2095 used the deprecated CM02 network model. Use default instead.
2096 * Rename the TCP_gamma configuration option to network/TCP_gamma
2097 * Rename the coordinates configuration option to
2098 network/coordinates, and document it
2099 * Use now crosstraffic keyword instead of the terribly misleading
2100 fullduplex keyword. It is activated by default now in the current
2101 default model, use --cfg=network/crosstraffic:0 to turn it off.
2102 * Ongoing refactoring the model parsing to make XML files optional
2103 See include/simgrid/platf.h for details (still to be completed)
2106 * Major overhaul of the documentation. Almost instructive now :/
2107 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2108 functions and friends. This interface was considered as
2109 deprecated since over 2 years, it's time to inform our users that it is.
2110 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2111 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2112 use these (crappy) functions in your code.
2113 These functions will be removed soon. Stop using them now.
2114 * Deprecate MSG_get_host_{table,number}
2115 Implement MSG_hosts_as_dynar() instead.
2116 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2117 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2120 * Stabilize the parallel execution mode of user contexts
2121 * Introduce configuration variables to control parallel execution:
2122 - contexts/synchro: Synchronization mode to use when running
2123 contexts in parallel (either futex, posix or busy_wait)
2124 - contexts/parallel_threshold: Minimal number of user contexts
2125 that must be part of a scheduling round to switch to parallel
2126 execution mode (raw contexts only)
2127 * Fix bugs that prevented to use suspend/resume along with
2128 synchronization structures.
2129 * Fix bugs in process termination that lead to invalid memory access
2130 in very specific conditions.
2133 * Introduce a parallel mode for the models (controlled by surf/nthreads
2134 configuration item). In our tests, running the models in parallel
2135 never lead to any speedups because they are so fast that the gain
2136 of computing each model in parallel does not amortizes the
2137 synchronization costs, even when ultra fast futexes are used.
2138 This is released anyway because YMMV.
2141 * Performance boost by using a swag internally to compute the set of
2142 tasks that are finished and should constitute the return value of
2146 * Enable it by default now that it is considered rather stable.
2149 * Documentation of the tracing functions.
2150 * Performance gains when tracing categorized/uncategorized resource
2151 utilization by avoiding calls to get route when updating resource
2152 variables. LMM constraints are being used instead.
2153 * API changed to set task categories. Use MSG_task_set_category instead
2154 of TRACE_msg_set_task_category, and SD_task_set_category instead
2155 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2156 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2157 * Examples for link user variables added (see at examples/msg/tracing/)
2158 * Deprecated function TRACE_msg_set_process_category completely removed
2159 * Trace header updated according to the latest Paje file format
2160 * Tracing network lazy updates, no longer obligate users to use full updates
2161 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2162 * Experimental: let user code declare/set/push/pop application states for hosts
2163 * API changed to allow the manual creation of graph configuration files
2164 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2167 * Improve the API of Lua MSG bindings, using the Lua spirit.
2168 * Each simulated process now lives in its own Lua world (globals are
2169 automatically duplicated). It helps writing simulators. Will allow
2170 to run Splay programs within SimGrid in the future.
2171 * Add a Chord example in Lua, equivalent to the MSG one.
2174 * Start the implementation of a solution to express temporal
2175 properties, not only local assertions. This is still an
2176 experimental work in progress, stay clear from it to be safe.
2180 - Add new runtime parameters --help-logs and --help-log-categories
2181 to display information about supported logging parameters and
2183 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2184 don't exists anymore.
2185 * Mallocators: allow value NULL for the reset function.
2187 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2188 create homogeneous dictionaries, where all the elements share the
2189 same free function. Non homogeneous dictionaries will be
2190 deprecated in the next release.
2191 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2192 - Multi-level dictionaries are deprecated.
2194 - new function xbt_dynar_search_or_negative() that is useful when
2195 you have less than 2 million elements in your dynar and don't
2196 want of the extra complexity of catching exceptions when the
2197 element is not found.
2199 - Make xbt_os_thread module (for thread portability) public.
2200 Documentation is still to come, sorry.
2202 - Cleanups and simplifications to make it maintainable again.
2203 - Exotic features (such as memalign and valloc) were removed.
2204 - The metadata were extended and improved so that the
2205 model-checker becomes able to explore and inspect the heaps.
2206 - This may induce a performance drop when enable_model-checking is
2207 ON in cmake (even if it's not used in the simulation), but it is
2208 necessary at this point to get MC working.
2210 Turn model-checking OFF if simulation performance matters to you.
2211 Not enabling it at runtime is not enough, disable it in cmake.
2213 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2216 SimGrid (3.6.2) stable; urgency=low
2218 The "Not coding new stuff allows to polish old things" release.
2221 * New bindings to the NS3 packet level simulator (experimental)
2222 * Use the raw (efficient) execution contexts instead of the sysv
2223 (portable) ones when possible.
2224 * libpcre is now mandatory in any cases since not using it led to
2225 severe performance loss and possibly other issues
2226 * Update the XML platforms:
2227 - G5K: include the latest machine in Nancy
2228 - GridPP and LCG: new platforms
2229 * Documentation was partially updated, at least (more to come)
2231 Bug fixes, cosmetics and small improvements
2232 * Free terminated processes before the end of the simulation to avoid
2233 exhausting the memory of users having very dynamic amount of
2235 * Bug fix and cosmetics about canceling non-running tasks
2236 * Bug fix about the dot loader's issues when using libcgraph
2239 * Create an installer for windows with nsis (amd64 and win32)
2240 - Add an hello world project to illustrate simgrid project creation.
2241 - Embed libpcre into the Simgrid installer to avoid
2242 its compilation burden
2243 * The raw execution contexts should work on Apple now
2244 * Port to Windows 64 bits
2245 - Sysv contexts now have an implementation for this arch
2246 - GRAS communication features now support this arch
2247 * Drop support for borland compiler on windows
2248 - this code was not maintained, and we kinda depend on gcc nowadays
2249 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2250 * Fix portability issue on unstable ubuntu: linker became picky on
2253 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2255 SimGrid (3.6.1) stable; urgency=low
2257 The "Oops, we broke Macs too" release
2260 * Fixed contexts detection so that raw ones are used when possible
2261 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2262 leads to a strange error, with user code segfaulting sometimes when
2263 the generated code is not perfectly aligned (which is not
2264 controllable from the user side, depends on the amount of code)
2267 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2269 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2271 SimGrid-java (3.6) unstable; urgency=low
2274 * Split of every thing from simgrid v3.5 into a separate package.
2276 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2278 SimGrid (3.6) stable; urgency=medium
2280 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2283 * Bindings now constitute their own package, separated from the main one.
2284 Rationale: reduce our maintenance nightmare by reducing the module coupling
2285 They will soon be released on their own on gforge.
2287 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2288 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2290 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2291 * It's not quite deprecated for now because we have no replacement,
2292 but it may soon become the case.
2295 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2296 * Fortran: New user-level cache variable to store the rank of the running
2297 process. This improves performance by an order of magnitude.
2298 * C: New coccinelle script to automatically locate and modify global and
2299 local static variables.
2300 * Improved SMPI network model with a sender-side gap to account for multiple
2304 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2305 only say if a communication is finished, no matter whether it succeeded or
2306 failed. You can call MSG_comm_get_status() to know the status of a finished
2308 * New function MSG_task_dsend() to send a task and detach it. When a
2309 communication is detached, you are never notified of its success or failure
2310 and the memory is released automatically once it is finished. This function
2311 is useful when you don't care about the end nor the success of a
2313 * Change the prototypes of action replay. Sorry for inconvenience,
2314 but this is really more efficient this way (and to adapt your code,
2315 you just have to fix the initialization, that shouldn't be too long)
2316 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2317 ever managed to do anything useful with it.
2318 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2319 takes a dynar of communications. It returns immediately and gives the
2320 index of a finished communication (if any).
2321 * New example: a basic implementation of the Chord P2P algorithm.
2324 * New model for multi-core CPUs. You can now use the core attribute to
2325 precise the number of cores of a host. This is a basic model. Every
2326 process running on the host receives at most the power provided in
2327 the DTD (throughput<=power). Total throughput of process cannot exceed
2329 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2330 router linked by an up-link and a down-link (possibly asymmetrical).
2331 This kind of pattern allows to easily build last-mile model style platforms.
2332 Aggregating such patterns in a rule-based AS is thus the technique of
2333 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2334 * New model for Vivaldi routing. We transformed the Vivaldi network model
2335 into a Vivaldi routing model (based on the rule-based model). This allows to
2336 combine Vivaldi based latencies with last-mile platforms.
2339 * Added a check for NaN of IEEE754 infinite in the double entries of
2341 * Introduce a new context factory "raw", highly inspirated from the
2342 ucontext factory, but using manually crafted functions in assembly to
2343 do the work in an efficient manner.
2344 * Allow to change the used context factory at run time, not only at
2345 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2346 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2347 contexts at runtime (only with raw contexts or ucontexts).
2348 * Completely rewrote this module to allow parallel execution of user
2349 processes. Use --cfg=contexts/nthreads:N to execute user processes
2350 with N parallel threads (the default is 1, meaning no parallelism).
2351 * Allow to decide dynamically between sequential and parallel modes.
2352 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2353 processes in parallel only when their number is greater than or equal to P
2355 * Added a check for NaN of IEEE754 infinite in the double entries of
2359 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2360 won't output the state of processes when interrupted with Ctrl-C
2361 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2362 NULL-terminated array. This may solve backward compatibility issues
2363 due to the change to return type of SD_simulate. See also:
2364 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2365 * Add new macros with variable number of arguments.
2366 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2367 - in xbt/asserts.h: xbt_assert
2368 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2369 - in xbt/ex.h: THROWF and RETHROWF.
2370 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2372 * Change xbt_die() to accept a format string with arguments, just like printf.
2373 * New data structure: xbt_lib_t, like a dict but more general and with better
2377 * New configuration options
2378 Options triva/categorized and triva/uncategorized can be used to generate
2379 graph configuration files for Triva visualization tool.
2380 * Configuration option tracing/platform is renamed to tracing/categorized
2381 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2382 * New TRACE_user_link_variable interface:
2383 User provides the name of the link and the tracing variable to attach to it
2384 * the declaration of tracing categories must be done after the environment
2386 * simpler tracing interface, just one way to declare categories
2387 TRACE_category or TRACE_category_with_color, it is up to you
2388 * links in the trace file are again identified by their names
2389 * trace contains the full platform hierarchy exactly as declared using the ASes
2390 * Options tracing/msg/[task|process]:1 groups the process by hosts
2391 for both cases, tasks and processes must have names that are unique during the simulation
2392 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2393 * The experimental option tracing/msg/volume is deprecated
2394 its functionality may be reincorporated if needed
2396 The tracing generates a trace file with unordered timestamped events,
2397 because of the way the core simulator (surf) works. A script available
2398 at the tools directory (fix-paje-trace.sh) can be used to put the events
2399 in order. We have changed the tracing so it can generate ordered timestamped
2400 events in the final trace, but depending on the simulator (and how much time
2401 is simulated) that can lead to a huge memory utilization. It is deactivated
2402 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2404 Build Infrastructure
2405 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2406 - We are trying hard to keep the API stable, but it may happen that
2407 some things change (we're a research project after all, not a
2408 nuclear plant operating system). If such things should happen, you
2409 could rely on that macro to adapt.
2410 - current value: 30600 for 3.06.00, aka 3.6
2411 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2412 a number that can be compared with SIMGRID_VERSION.
2413 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2414 assertions and verbose/debug logging events are disabled at compile time.
2416 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2418 SimGrid (3.5) stable; urgency=medium
2421 * New feature to any SimGrid-based simulator: Model-Checking
2422 Check SIN#1 for more details.
2425 * New Model SMPI (three-interval linear regression for correction factors)
2426 See RR-7426, available at http://hal.inria.fr/inria-00527150
2427 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2428 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2429 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2430 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2431 * See also src/smpi/README
2436 - Tracing API changes: TRACE_start and TRACE_end should not be called
2437 by user-code. They are automatically called by simulators created
2438 with SimDAG, MSG and SMPI if the toolkit is compiled with
2439 tracing_enabled=ON. Categories declaration and utilization remain the
2440 same for MSG and SimDag.
2441 - A function was added to the tracing API to declare categories with
2443 - TRACE_category_with_color (char *category, char *color)
2444 where color must be in the following format
2445 "%f %f %f", red, green, blue
2446 and red, green, blue are float values in the interval [0, 1]
2447 - User can specify NULL as color parameter, or continue calling
2448 TRACE_category (cat)
2449 On that case, the tracing system will define random colors
2450 - The following command-line options are supported:
2451 --cfg=tracing/filename:msg.trace
2452 --cfg=tracing:1 (activate tracing, needed to use others)
2453 --cfg=tracing/platform:1 (categorized resource use)
2454 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2455 --cfg=tracing/msg/task:1 (task creation)
2456 --cfg=tracing/msg/process:1 (process creation, migration)
2457 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2458 --cfg=tracing/smpi:1 (SMPI interface tracing)
2459 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2460 - examples of examples/msg/tracing updated
2462 - DAXLoader and DOTLoader functions can generate tasks with categories
2463 - A new function to attribute a category to SD tasks:
2464 TRACE_sd_set_task_category (SD_task_t task, char *category)
2465 * Tracing the MPI interface implemented by SMPI
2466 - Collective operations are traced with states
2467 - Point-to-Point operations are traced with states/links
2468 - Tracing activated by a parameter "-trace filename" passed
2469 to smpirun during execution (considering that simgrid
2470 is compiled with tracing enabled)
2471 - To run the simulation with gdb, the simulator
2472 accepts --cfg=tracing/smpi:1 to trace SMPI
2473 - tesh files to check if smpi tracing is ok
2474 - See examples/smpi/NAS/DT-trace
2475 * GTNetS tracing re-worked
2476 - adaptation to the tracing system of GTNets to cope
2477 with modifications regarding the fullduplex mode
2478 - new tesh files to check if gtnets tracing is ok
2481 * Asynchronous communications through the functions:
2482 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2483 * New function: MSG_load_platform_script()
2484 to make possible using a lua script instead of XML files to set up platforms
2485 * New function: MSG_set_function
2486 to associate functions to processes, used when bypassing the parser
2487 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2489 Platforms: Add some more examples in examples/platforms
2490 * Grid'5000: see www.grid5000.fr
2491 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2494 * Change the XML format. This is a very important modification. SimGrid
2495 3.5 introduces a new hierarchical format based on the notion of
2496 Autonomous Systems. Compatibility with old format is ensured
2497 through the perl script provided in the install bin directory
2498 bin/simgrid_update_xml.
2499 It is now possible to build platforms with specific routing
2500 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2501 connect several platforms together. We will try to provide soon
2502 set of realistic platforms exploiting these properties (have a
2503 look at examples/platforms/ for the moment).
2504 * Take the opportunity of the XML format change to be a good XML citizen:
2505 rename link:ctn to link_ctn and similar changes (also dealt with by
2507 * Add a new routing scheme (rule-based) using regular expressions. It
2508 enables to have an extremely low memory footprint when the
2509 underlying routing is simple and can be compactly described. You
2510 need to have libpcre4-dev (perl regular expressions) installed if
2511 you want to use this routing scheme.
2512 * Revive the cluster TAG and allow to easily and efficiently (both in
2513 term of memory and speed) connect clusters together. Have a look
2514 at teshsuite/simdag/platforms/ to see how this can be done. With
2515 this tag, you can create clusters with thousands of tasks at no
2516 cost (have a look at examples/platforms/).
2517 Note: clusters are implemented as ASes, so there is no need for an
2518 enclosing AS tag if you have only one cluster in your platform.
2519 * Add new generic functions in the public interface that allows the user
2520 to call SURF 'create_resource' methods from your code (same
2521 functionality as the XML bypass mechanism but with a much lighter
2523 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2524 that uses a piecewise linear approximation to produce better
2525 results when exchanging small messages.
2526 * Add a new parameter to handle correctly full duplex link and account
2527 for interferences between uplink and downlink communications
2528 (activate with --cfg=fullduplex:1).
2531 * Rename the SD_READY (all dependencies are satisfied and task is
2532 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2533 dependencies are satisfied) state.
2534 This prevents a confusion between the notion of "ready to schedule"
2535 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2536 simulated" (SD_RUNNABLE) used by the simulation kernel.
2537 * Change the way a task is considered as ready. Instead of removing
2538 dependencies when a task is done, a counter is decreased. This way,
2539 it is always possible to reach ancestors thanks to the
2540 SD_taks_get_parents function (even after the end of the simulation.)
2541 * Change the return type of SD_Simulate from (SD_task_t*) into
2542 xbt_dynar_t. This function was in handling a dynar internally and
2543 converted it into a NULL terminated array for historical reasons.
2544 * New function SD_dotload(char*) to load a DAG described in dot
2545 format. This loader and the corresponding examples require the
2546 installation of the graphviz library.
2547 * Fix a bug in the management of tasks of size 0 in the surf network
2548 models. This problem was only visible with SIMDAG and you should
2549 thus disregard results produced with earlier versions if you
2550 relied on this feature (some tasks were blocked because of this).
2551 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2552 even though your applications were doing only point-to-point
2553 communications and sequential computations. Now you can really use any
2554 model you want (of course, if you create real parallel tasks, which are
2555 not implemented in most models beside ptaskL07, this will abort).
2556 * Add an example that schedules a DAX on an heterogeneous platform
2557 using a Min-Min strategy.
2558 * New function SD_workstation_get_current_task() that returns the kind
2559 of task currently running on a workstation in the sequential access
2561 * Raise some warnings when unexecuted tasks remains at the end of the
2562 simulation. This is usually caused by cycles in the DAG.
2565 * New function: SIMIX_process_set_function() called by MSG_set_function
2566 * Change the underlying waiting queue in semaphores so that a process
2567 can wait on several of them simultaneously (as in waitany).
2568 * Fix the way to handle tokens in semaphores so that all access patterns
2569 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2570 * kill the dirty pimple SIMIX_message_sizes_output()
2571 Please use (proper) visualization instead
2574 * New data container: setset (set of sets of elements)
2575 * New module: mmalloc (mapped malloc, allowing to have several
2576 independent segments of malloc)
2577 * New function: xbt_dict_cursor_set_data()
2578 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2579 * New function: xbt_dynar_is_empty()
2580 * New function: xbt_fifo_get_last_item()
2581 * Fix xbt_dynar_shrink(): use the right element size.
2582 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2583 uninitialized areas during expand.
2584 * Fix semaphores: previous implementation was severely broken.
2585 * Use library init/fini functions for our initialization.
2586 - you can use logs and other feature as soon as you want in your
2587 code (even before the xbt_init / MSG_init)
2588 - xbt_exit is now a no-op and produce a warning when used.
2591 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2592 benefit from the latest improvement to the simulation kernel.
2593 * Kill measurement sockets for now. If you rely on them, sorry. This
2594 release is not for you. This feature will be reintroduced in the
2595 future, but we cannot delay the release any further.
2596 * New function: gras_msgtype_get_name().
2597 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2598 * Fix (at last) the pmm example: it should not randomly fail anymore.
2600 Build chain: bug fixes and overall polishing
2601 * Cmake is now stable enough. Hence, we killed the autotools.
2602 * Port to windows ( TM :)
2603 * Fix the 'make install' target.
2604 No need to use 'make install-simgrid' anymore
2605 * Introduce a 'make dist' target compiling a *source* archive
2606 'make package' compiles a binary archive
2607 * Compile java files only on need
2608 * Add --cd and --setenv command line options to tesh
2609 * Out of source builds are not fully supported yet, but we are close
2610 * Enable supernovae and optimization flags by default for our users
2613 * Add layer to set up environment directly from lua, without XML.
2614 * The effect of gras_stub_generator can be achieved through
2615 lua too (check examples/gras/console/ping_generator.lua)
2617 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2619 SimGrid (3.4.1) stable; urgency=low
2621 The "Polishing easter eggs is probably a good idea" release.
2622 This is a bug fixes release only.
2626 * Fix a bug preventing the tasks from begin garbage collected.
2629 * Fix a bug occurring when a host involved in a communication fails.
2630 This was not detected properly by the other peer involved in the
2631 communication. Now, it's reported as a network error.
2634 * Warn the user about loop dependencies in data flow of DAX files
2635 * Obey the control-flow dependencies of DAX files
2638 * Add option "enable_smpi" allowing to not compile SMPI.
2639 Probably useful for the (Mac) users experiencing a build error here
2640 * Improve the detection of lua5.1 and ruby1.8
2642 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2644 SimGrid (3.4) stable; urgency=low
2646 The "Easter in Cargese" release. Also known as (major changes):
2648 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2649 ~> bindings were greatly improved
2650 ~> new tracing infrastructure for better visualization introduced
2652 * the "Welcome to configury modernity" release.
2653 ~> we switched from autotools to cmake, and improved our cdash
2656 A more detailled list of changes follow (full detail in svn log).
2658 Java Bindings: Various Cleanups
2659 * (install java-gcj-compat-dev on debian-like to use them)
2660 * Remove put/get: no need to export deprecated interface in Java
2661 Use send/receive instead.
2662 * Cleanup the examples and add a README per directory
2663 * Remove example autoDestination (that's the only way to go now)
2664 * Remove example explicitDestination (was a plain copy of basic)
2665 * Make JniException a runtime exception, so that there is no need to
2666 declare the fact that you may encounter such a beast. I guess that
2667 nobody will ever want to survive such error.
2668 * Create specific errors for each MSG case of failure:
2669 host failure, transfer failure, timeout, task cancelled
2670 * Cleanup the exceptions that may get thrown by each function
2671 * Other internal cleanups in Java bindings. Performance still bad :/
2672 Ruby and Lua Bindings: create them
2673 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2674 * That's new and great, you should try them out.
2675 Same functionalities than Java bindings, only even less polished
2677 * Kill the useless "rate" argument of SD_task_get_execution_time()
2678 Everyone used to provide -1 as a value, it was not used, and the
2679 semantic of a possible use wasn't even clear.
2680 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2681 either as comm costs or compute costs to mean that there is no
2682 such thing for that specific task.
2683 * Add a SD_task_set_name() function
2684 * Fix SD_task_unschedule() on typed tasks
2685 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2686 * In DAX loader, accept useless 'level' attributes to <job> since
2687 LIGO DAGs have them (seem to be to ease graphical representation).
2689 * Add an example masterslave_mailbox.c using send/receive and not
2690 the deprecated put/get interface.
2691 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2692 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2694 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2695 (declare MSG_USE_DEPRECATED to still have the old name)
2696 * Add a MSG_task_set_data() function
2697 * About trace replay (see examples/msg/actions):
2699 - Allow to work with splitted trace files for each process
2700 Give the specific trace file as argument of each process,
2701 and call MSG_action_trace_run(NULL)
2702 You can still have one merged file for all processes.
2703 - Fix implementation of collective operations
2704 * Allow task_execute() on 0-sized tasks (closes #10063)
2706 * This is the first release of SimGrid where SMPI is not considered
2707 beta anymore (even if some corners should still be improved)
2708 * Port over the new SIMIX_network submodule (internal refactoring)
2709 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2710 * Implement more missing elements of the standard:
2712 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2714 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2715 MPI_Scan() MPI_Get_processor_name()
2716 - Added implementation of missing case for Alltoall (warning: it's
2717 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2718 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2719 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2720 SMPI_MPI_Allgatherv()
2721 * Bug fixes include:
2722 - MPI_Waitsome() was broken
2723 - Allow relative includes in smpicc
2724 - Command line cfg argument 'reference_speed' was ignored...
2725 - Some functions did not properly lead to auto-benching of user code
2726 - smpicc passes -O2 by default (just like openmpi one)
2728 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2729 * Bug fixes about timeouts during communications
2730 * add SIMIX_message_sizes_output() as a pimple to write to file the
2731 amount of messages per size. Use gnuplot to get histogram.
2732 Pimple because that's the only user-visible function of simix,
2733 defined directly in xbt.h (irk, sorry)
2735 - Add a SIMIX_sem_get_capacity() function
2736 - Fix interactions with processe resume/suspende
2737 - release_forever() was stupidly broken
2738 - Fix SIMIX_display_process_status() for processes in a semaphore
2739 - Make SIMIX_sem_block_onto() user-visible
2740 * Refactoring context stuff:
2741 - Use pseudo-OOP for better modularity
2742 - reimplement SIMIX_process_kill() without process_schedule() so
2743 that the latter can take as invariant that it is called from
2745 - Merge context_start into context_new for sake of simplicity
2747 * Add a Vivaldi network model, coded live during SUD'10 ;)
2748 * Rename configuration variables to start a hierarchy:
2749 o cpu_model -> cpu/model
2750 o network_model -> network/model
2751 o workstation_model -> workstation/model
2752 * New configuration variables:
2753 o network/bandwidth_factor: correction to bandwidth
2754 o network/latency_factor: correction to latency
2755 o netwotk/weight_S: correction to the weight of competing streams
2756 * Add a long description to the models, that users can see with such
2757 argument on the command line: --cfg=cpu/model:help
2758 * --help-models display the long description of all known models
2760 * config: add the ability to set a default value after registration
2761 Does not override any previously set value (e.g. from cmd line)
2762 * dict: allow to have integer key and data.
2763 When so, you need to use the following functions
2764 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2765 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2766 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2767 In contrary to regular dicts, the key is not malloced before copy.
2768 Mixing scalar and regular elements in the same dict is not tested
2770 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2771 Tracing for Visualization:
2772 * SimGrid is now instrumented in order to generate a trace file for
2773 visualization analysis: to use it, need to compile SimGrid with the
2774 "tracing" option enabled, and instrument the program using SimGrid with
2775 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2776 (among other functions).
2777 * The instrumentation only traces the platform utilization for now
2778 * Documentation to use the tracing functions and how to analyze the
2779 traces with the Triva tool is written.
2780 * More information about: SimGrid FAQ (in the section Tracing Simulations
2783 * We moved to cmake as default build system. Autotools support will
2784 be dropped soon. Check the FAQ for more info about how to use it.
2785 * Greatly improved our cdash/ctest interactions
2786 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2787 * Added memory checking tests with valgrind; lot of memleak fixing.
2788 This may be the first release of simgrid with so few memory issues
2789 * Added code coverage tests.
2790 Our coverage is still improvable, but at least we see it on cdash.
2792 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2794 SimGrid (3.3.4) stable; urgency=low
2796 The "Desktop Grid needs love too" release (also called Xmas release).
2798 Models improvements:
2799 * Major speedup in the maxmin system solving by using lazy evaluation
2800 Instead of solving completely the maxmin system at each iteration,
2801 only invalidate (and recompute) the modified parts.
2802 This new feature is enabled in default models but you can try to
2803 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2804 * Cas01 IMproved as default CPU model
2805 This CPU model is the same Cas01 model, but it uses the
2806 maxmin-selective-update flag and a heap structure to manage
2807 actions on SURF kernel.
2808 It reduces the complexity to find the next action to finish and,
2809 consequently, it's faster than the old Cas01.
2810 This is the new default CPU model (Cas01).
2811 * Rename the old Cas01 model to Cas01_fullupdate
2812 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2813 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2814 * CpuTI (CPU Trace Integration)
2815 A new CPU model whose objective is simulate faster when using
2816 availability trace files.
2817 Instead of using a full featured, over engineered maxmin system for
2818 CPU modeling, this model does the pre-integration of traces files
2819 to calculate the amount of CPU power available, and so, executes
2820 faster than the old CPU models.
2821 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2822 * Use LV08 as default network model since it gives better accuracy
2823 for small messages and shouldn't change things for big ones.
2824 Use --cfg=network_model:CM02 to get the previous behavior.
2827 ******************************************
2828 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2829 ******************************************
2830 * The new CPU model may changes simulations!
2831 The point is that events occurring at the exact same timestamp
2832 are not scheduled in the same order with the old and new
2833 version. This may be enough to completely change the execution
2834 of simulations in some cases.
2835 * The new network model will change simulations!
2836 This new model is more realistic than the previous one, so you
2837 should consider redoing your old experiments with this model.
2838 Sorry for the inconvenience.
2841 * Introduce the supernovae compilation mode
2842 When compiled that way, the whole SimGrid (or almost) is put in a
2843 single compilation unit and compiled in one shoot.
2844 This is to help gcc which has difficulties to inline stuff from one
2846 The speedup seem to be above 15%, althrough more tests are needed on
2847 amd64 to confirm that gain.
2850 * Port of MSG's mailbox on top of SIMIX network
2851 The put/get mechanism was greatly simplified on the way.
2854 * New SIMIX network module. Provides:
2855 - Mailbox: rendez-vous mechanism to find with who you want to speak
2856 - Synchronous send/recv: easier and hopefully faster since the
2857 logic is handled in the maestro process directly now
2858 - Asynchronous send/recv: you dreamt of it? It's here now
2859 Too bad that nobody cared enough to propagate the change to MSG.
2860 * Add semaphores as SIMIX synchronization mechanism.
2863 * new function SD_daxload(char*) to load a DAX file
2864 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2865 * Introduce typed tasks. Specify its kind and cost at creation.
2866 At scheduling, just give where it should be placed, and the cost
2867 for each involved resource is automatically computed.
2868 Existing constructors so far (more to come of course):
2869 - SD_task_create_comm_e2e() for end-to-end communication
2870 - SD_task_create_comp_seq() for sequential computation
2871 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2872 * new function SD_task_dump() for debugging display
2873 * new function SD_task_dotty(task,FILE*) writing to file the info
2874 about the task in dotty format
2875 * SD_task_dependency_exists() can now cope with having one of its
2876 arguments NULL. If so, it tests whether the other argument has any
2878 * Add getters on list of preceding/following tasks:
2879 SD_task_get_parents(task) and SD_task_get_children(task)
2880 * Add getters on amount of workstations and list:
2881 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2882 * Add getter on task kind: SD_task_get_kind(task)
2883 * Update the start_time and finish_time of tasks on completion/failure
2884 * Bugfix: Remove task from state swags when destroyed
2887 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2890 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2891 dynar with one separate thread per value of the dynar.
2892 * Change the prototype of xbt_thread_create(), sorry.
2893 Added a boolean parameter indicating whether we want to join this
2894 thread (used in SG only for now)
2895 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2898 * GTNetS wrappers should now be usable again (and betterly tested too)
2899 * Fix a major regression from 3.2 where the timeout provided to
2900 MSG_task_put_with_timeout() was used as absolute time before which
2901 the comm should be done.
2902 * Start to fix the <cluster> tag.
2903 - Internal links should be good now (beside of the loopback, which
2904 use the private link instead)
2905 - paths to the external world is still rather broken
2906 - the <route:multi> tag is just broken. Actually that's brain-dead.
2907 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2908 to make it less stupid
2909 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2910 * Fix a source-level compatibility glitch from 3.2: after defining
2911 MSG_USE_DEPRECATED, you can use the old name
2912 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2913 * Allow to compile from the SVN with automake 1.11
2914 * Fix some problems when using the "start_time" tag in deployment XMLs.
2915 * Fix #8569: XBT/synchro.h has redundant declarations
2916 * Fix #8563: MSG return values and exceptions
2917 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2918 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2919 Thanks to Robson Peixoto.
2920 * Fix a vicious bug in dictionaries inducing that some elements were
2921 not freed on xbt_dict_free()
2923 Portability report of this version:
2924 * Main portability targets:
2925 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2926 on (amd64/i386/ia64)
2927 - mac leopard on i386
2928 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2929 but nothing critical.
2930 * Other platforms: windows, AIX and others were not tested for this release
2932 Timing report of this version:
2933 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2934 scenario parameters). From 8h to a few seconds in desktop grid settings.
2935 * Supernovae brings about 25% speedup on i386.
2937 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2939 SimGrid (3.3.3) stable; urgency=low
2941 The "Need for Speed" release.
2943 The timings done to validate the 3.3.2 were faulty.
2944 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2946 The problem was a conversion from a manually handled vector to
2947 xbt_dynar_t on the critical path.
2948 xbt_dynar_foreach calls functions, inducing stack management crap.
2950 We inlined these functions and xbt_dynar_foreach is now breath taking.
2951 We also inlined xbt_swag_belong on the way.
2953 Here are some approximate speedup measurements (on master/slaves
2954 simulations lasting between 10s and 20s each):
2955 3.3.1 -> 3.3.2: about same performance
2956 3.3.2 -> 3.3.3: 40% speedup
2957 3.3.1 -> 3.3.3: 40% speedup
2958 3.3.1 with inline patch -> 3.3.3: 30% speedup
2960 Our reading is that the refactoring which occurred in 3.3.2 made us
2961 suffer much more from the xbt_dynar_foreach low performance, but
2962 once we solved this, this refactoring proved to be very performance
2963 effective. From the 40% speedup, somehow, 10% are due to the
2964 inlining and 30% to the refactoring.
2966 That's a pitty that gcc cannot inline functions placed in other files
2967 alone. We have to choose between:
2968 - break the encapsulation (by putting private data structures and
2969 accessors in headers files to help gcc)
2970 - live with low performance
2971 - switch to a decent compiler such as icc (not quite possible).
2973 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2975 SimGrid (3.3.2) stable; urgency=low
2977 The "Simplicity does not preceed complexity, but follows it" release.
2979 The main contributors of this release were (lexical order):
2980 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2983 * Extract the routing logic into its own object.
2984 (was dupplicated in network.c and workstation_LV07.c;
2985 Allows to implement other ways of storing that info)
2986 => kill now useless network_card concept
2987 - Use dynar to represent routes (instead of void** + int*)
2988 - kill link_set (use surf_network_model->resource_set instead)
2989 - Add a command-line option to choose the routing schema to use
2990 - Add three new models:
2991 * Floyd (shortest path computed at initialization)
2992 * Dijikstra (shortest path recomputed all the time)
2993 * Cached Dijikstra (shortest path computed on need)
2994 All these models where contributed by Silas De Munck, and are
2995 described in his ICCS09 paper.
2997 * Simplify model declaration
2998 (less redirections, less function to write when defining a model)
2999 - Factorize stuff between models:
3002 surf_model_resource_set(model)
3003 surf_model_resource_by_name(model, name)
3004 - Unify the types of models in s_surf_model_t (using an union)
3005 - Embeed fields of common_public directly into s_surf_model_t
3006 - Rename model methods:
3007 action_free ~> action_unref
3008 action_change_state ~> action_state_set
3009 action_get_state ~> action_state_get
3010 - Change model methods into functions:
3011 (model)->common_public->action_use ~> surf_action_ref
3013 * Implement a generic resource; use it as ancestor to specific ones
3014 (allows to kill duplicated code in models)
3015 Drawback: timer command don't need no name nor properties;
3016 workstation_CLM03 don't need no properties
3017 (but I guess we can live with those few bytes wasted)
3019 * Improve the action object model
3020 - implement a constructor avoiding dupplicated code about field
3021 initialization in generic_action part.
3023 * Kill the SDP model: it has an external dependency, is deprecated
3024 in flavor of modern lmm models, and didn't compile since a while
3027 * Relocation of the context module from XBT to SIMIX.
3028 (the context were decoupled from the simix processes, duplicating a lot of code)
3029 => a lot of code was factorized
3030 - less overhead is introduced during scheduling
3031 - simpler API for the context factory
3032 - the logic for process creation,destruction and manipulation was simplified
3033 * Simplification of the s_smx_process_t data structure.
3034 => accesing the simix level data associated to a process is faster now,
3035 and the code is a lot more readable.
3038 * Implement some more MPI primitives:
3039 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3040 -implementation: Bcast: flat or 2-ary tree (default),
3041 Barrier: 4-ary tree,
3043 Allreduce: Reduce then Bcast
3044 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3045 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3046 Alltoallv: flat tree, like ompi
3048 * Add support for optimized collectives (Bcast is now binomial by default)
3049 * Port smpirun and smpicc to OS X
3052 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3053 Shout out if you used it.
3056 * Display the list of still queued messages in SG mode when existing
3060 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3061 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3062 * Add xbt_str_from_file(FILE*)
3063 * Add xbt_dict_get_key achieving a linear reverse search
3064 * Remove the context module
3066 Portability report of this version:
3067 * Main portability targets:
3068 - Linux(debian)/x86/context
3069 - Linux(debian)/x86/pthreads
3070 - Linux(debian)/amd64/context
3071 - Linux(debian)/amd64/pthreads
3072 On these, we still have the eratic breakages of gras/pmm and
3073 amok/saturate_sg reported in previous version. We still think
3074 that the tests are the cause of the fault, not the tested code.
3076 - Mac OSX Leopard/x86/context
3077 Still false negative in tesh autotesting.
3078 Smpi still fails, but this time because readlink does not accept -f
3079 Everything seems to work properly beside of that.
3082 - AIX version 5.3 (only tested contexts this time)
3083 Smpi still fails there because mktemp is not installed.
3084 Everything seems to work properly beside of that.
3086 I managed to compile it for the first time, but several breakages.
3087 Won't delay the release for this exotic platform.
3089 * Windows: it's still lagging behind. If you want to help, please
3092 Timing report of this version:
3093 This version seem to be more than 5% faster than 3.3.1 (on linux
3094 64bits with contextes). The gain is less than expected, we are
3095 investigating this for next release.
3097 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3099 SimGrid (3.3.1) stable; urgency=low
3102 * Implement a --cfg-help to show existing configuration variables
3103 * Build chain do not require doxygen in maintainer mode
3106 * fix a bug on struct sizeof computation, which prevented the
3107 exchange of arrays of structs in some conditions
3108 - added a regression test about this in datadesc_usage
3109 * Allow the exchange of 0-long dynamic vectors.
3110 - for that, use -1 as indicator of dynamic size instead of 0
3111 - This implied to change any size from unsigned long to long,
3112 reducing a bit communication abilities, but I guess that with
3113 64bits being quite common, this is more than enough.
3114 - This also induce a protocol change, thus bumping network protocol
3115 version from 0 to 1 (if we have external users, we have to get
3116 clean on that point too ;)
3117 - added two regression tests about this in datadesc_usage
3118 * Be more verbose when propagating local exceptions
3119 This helps debugging.
3120 * Display the status of simulated processes when receiving SIGINT in
3124 * Allow to control the simulation from a trace file.
3125 New functions MSG_action_register() and MSG_action_trace_run()
3126 The first one allows to associate a function execution to each
3127 kind of action while the second one parses a trace file and
3128 triggers the corresponding actions within the system.
3129 For now, only a toy example is provided in examples/msg/actions
3130 * Add an example of process migration in examples/msg/migration
3131 * Fix a bug in task exchange which broke MSG_task_get_sender()
3132 Add a teshsuite regression test for that.
3133 [Bug: if MSG_task_get_sender() is called after sender exit,
3135 * Fix a bug which prevented suspend/resume to work properly
3136 * Display the status of simulated processes when receiving SIGINT
3137 This fixes a regression of v3.3. due to the introduction of SIMIX
3138 * Bug fixing in failure management:
3139 - trace could not start by a failure at time 0
3140 - failure during communications were not working
3143 * Add SIMIX_process_set_name() to change the name of the current
3144 process in the log messages.
3145 * Store smx_hosts in a dict since we only retrieve them by name
3146 * Move the configuration infrastructure to surf
3149 * Move the configuration infrastructure to surf
3152 * Massive internal cleanups:
3153 - Store internal structures on processes instead of hosts (allows
3154 to have more than one process per host, in addition of being more
3156 - Cleanup the initialization/finalization process
3157 - Kill a whole bunch of unneeded synchronization:
3158 processes run in exclusive manner within the simulator
3159 - Move queues from global tables to process data fields
3161 - now accept -platform and -hostfile arguments
3162 - Pass the right rank value to processes according to the hostfile
3163 * Compile the examples by default, and use them as regression tests
3164 * Implement MPI_Wtime()
3165 * Change the reference speed to a command line option
3168 * TCP_gamma can now be specified as command line option using
3169 --cfg=TCP_gamma:10000000.0
3170 * Change the --surf-path cmd line option into --cfg=path:
3173 * Also include strbuff from xbt.h public header
3174 * xbt_ex_display(): do not free the exception after displaying
3175 This allows to do more with the given exception afterward.
3176 Users should call xbt_ex_free() themselves.
3180 Portability report of this version:
3181 * Main portability targets:
3182 - Linux(debian)/x86/context
3183 - Linux(debian)/x86/pthreads
3184 - Linux(debian)/amd64/context
3185 - Linux(debian)/amd64/pthreads
3186 These targets fail about 1/10 of times on gras/pmm, but we believe
3187 that this is because of the test, not because of simgrid.
3188 amok/saturate_sg fails even more rarely, and the test may not be
3191 - Mac OSX Leopard/x86/context
3192 The test suite still spits tons of errors because some obscure
3193 force prevents us from removing the temporary directories
3194 arguing that they still contain some metadata I've never heard of.
3195 Smpi fails because seq is not installed.
3196 Everything seems to work properly beside of that.
3199 - AIX version 5.3 (both contexts and pthread)
3200 Smpi still fails there because mktemp is not installed.
3201 XML inclusions seems rosty on AIX.
3203 * Windows: it's still lagging behind. If you want to help, please
3206 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3208 SimGrid (3.3) stable; urgency=high
3212 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3213 [Malek Cherier & Mt]
3215 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3218 Until now, GRAS were using MSG as an interface to SURF. It was
3219 quite difficult because both interface have several differences
3220 (MSG channels vs GRAS sockets were the most notable point).
3222 This also opens the gate to SMPI (which should occur soon) and speed
3223 up simulations by to 40% (even if it were not the main goal).
3225 **************************************
3226 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3227 **************************************
3228 The point is that events occurring at the exact same timestamp are
3229 not scheduled in the same order with the old and new version. This
3230 may be enough to completely change the execution of simulations in
3231 some cases. Sorry for the inconvenience.
3233 * Cleanup and upgrade the XML format to push further scalability
3234 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3236 * Improve the testing infrastructure with tesh. Now a very large part of
3237 the code is tested not only by being run but also by checking that the
3238 output match an expected output [Mt].
3240 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3241 is really less memory-demanding, which should allow you to use
3242 larger files in SimGrid [AL].
3244 * Inform valgrind about our contextes, so that it becomes usable
3245 with the default (and more effecient) version of SimGrid
3246 [contributed by Sékou Diakite, many thanks]
3249 * Introduce a listener thread in charge of receiving incoming
3250 messages from the network. It allows to overlap communication and
3251 computation but most notably, it removes some stupid deadlocks due
3252 to the fact that so far, a process could not send and receive at
3253 the same time. This made most non trivial communication schema
3255 * Convert the PIDs from long int to int to match the MSG ones (and
3256 linux ones too) [Mt]
3257 * New function: gras_agent_spawn() to launch a new process on
3258 current host. Only working in simulation for now. [Mt]
3259 * New function: gras_os_hostport() returning a constant form (ie,
3260 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3263 * Make the backtrace of exceptions more human readable [Mt]
3264 * New module: xbt/str [Mt]
3265 a ton of string utility functions (split, join, printf to a newly
3266 allocated buffer, trim, etc)
3267 * New module: xbt/hash [Mt]
3268 SHA1 hashing algorithm (more to come if needed)
3269 * New module: xbt/synchro [Mt]
3270 synchronization tools (mutex and conditions) working the same way
3271 in simulation and in real life (mainly useful for GRAS, but not
3273 * New module: xbt/queue [Mt]
3274 classical producer/consumer synchronization scheme
3275 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3276 (using the classical functions will get serialized) [Mt]
3277 * Make dictionary internal table dynamic. No need to specify its size
3278 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3280 * Make sure the log channels are organized as a tree under windows
3281 (because of ANSI C compatibility issue, any channel were child of
3282 root directly) [Mt].
3285 * Cleaned many thing in surf and fixed a few bugs [AL].
3286 * Add a nice command line configuration mechanism to compose models [AL].
3287 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3288 the previous one (KCCFLN05). It relies on something that looks like
3289 a max-min sharing mechanism but cannot be written as such. A new solver
3290 was thus designed [AL].
3291 * Add a new solver to lmm. Based on Lagrange optimization and
3292 gradient-based descent, it enables to efficiently maximise systems s.a
3294 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3296 This solver enables to propose two new network models for TCP Reno and
3297 TCP Vegas based on Low's work. These models still need to be fully
3298 tested though [Pedro Velho].
3301 * Bug fix in SD_simulate. Now the time bound given as argument is
3303 * Use the new parallel task model (ptask_L07) as default.
3304 * Use the SURF command line configuration mechanism.
3305 * 0-size tasks (for synchronization) should now work.
3307 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3309 SimGrid (3.2) stable; urgency=high
3313 We still experience issues on this platform, but we believe that at
3314 least MSG is usable.
3316 GRAS API BREAKAGE (for simplification purpose, sorry):
3317 * the gras_msgtype_by_name is not used anymore. Instead of
3318 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3319 you can write (and must)
3320 gras_msg_send(toserver, "request", &request);
3321 - If you still want to pass a gras_msgtype_t to the function (to cache
3322 the type and avoid the lookup time), use the gras_msg_send_() variant.
3323 - Impacted functions:
3324 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3325 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3326 * The callbacks are now expected to return 0 when everything went well
3327 (just like the main() function)
3329 GRAS new features and improvements:
3330 * New module mechanism where user code can use per process globals [Mt]
3331 This is similar to gras_userdata_*() functions, but for libraries. It
3332 factorize some code developped over and over in the examples and AMOK.
3333 It has still to be documented and used (only amok/peermanagement is
3335 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3336 loss when they were small enough to fit into the buffer and sent quickly
3337 enough so that they can all get received in one shoot.
3338 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3339 if not found. Use the *_or_null() variant for the old semantic.
3340 * In gras_msg_handle, do not discard messages without callback.
3341 They are probably messages to be explicitly awaited later (ie, proofs of
3342 mis-synchronization in userland since they are sent before being awaited)
3343 No big deal usually.
3344 * gras_socket_meas_send/recv: semantic changed!
3345 The numerical arguments used to be (1) the total amount of data to send
3346 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3347 messages. This was need for the fool willing to send more than MAXINT
3348 bytes on quite fat pipes.
3351 * Do really rename the hostmanagement module to peermanagement. [Mt]
3352 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3353 but this is rather new and this was documented in the module
3354 documentation (poor excuses, I admit)
3355 * Bandwidth measurement semantic changed! This follows the changes to
3356 gras_socket_meas_send/recv explained above.
3359 * A sequential mode has been added to the workstations. When a workstation
3360 is in sequential mode, it can execute only one task, and the other tasks
3361 are waiting in a FIFO. [Christophe Thiery]
3364 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3365 model for SIMDAG. [Christophe Thiery]
3366 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3367 equal to 0, which caused some bad behaviors in
3368 saturated_constraint_set_update. I now use a threshold mechanism like in
3372 * When running manually src/testall, you select specific units [Mt]
3373 testall is the result of our cunit mechanism, and should replace all
3374 the scripty thingy around since bash don't run easily on billware.
3376 * A mallocator system has been added. [Christophe Thiery]
3377 Mallocators allow you to recycle your unused objects instead of freeing them
3378 and allocating new ones.
3380 Documentation update:
3381 * FAQ reworking + New FAQs:
3382 - "Valgrind spits tons of errors!" [Mt]
3383 - "How to repport bugs" [Mt]
3384 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3385 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3386 - Communication time measurement within MSG [AL]
3387 - I experience weird communication times when I change the latency [AL]
3388 * GRAS tutorial [Mt]
3390 - an introduction to the framework and to the used communication model
3391 - an initiatic tour introducing the most proheminent features:
3393 . Lesson 0: Installing GRAS
3394 . Lesson 1: Setting up your own project
3395 o Part 2: Message passing
3396 . Lesson 2: Exchanging simple messages
3397 . Lesson 3: Passing arguments to the processes (in SG)
3398 . Lesson 4: Attaching callbacks to messages
3399 . Lesson 5: Using globals in processes
3400 . Lesson 6: Logging information properly
3401 . Lesson 7: Using internal timers
3402 . Lesson 8: Handling errors through exceptions
3403 . Lesson 9: Exchanging simple data
3404 . Lesson 10: Remote Procedure Calling (RPC)
3405 . Lesson 11: Explicitly waiting for messages
3406 . Recapping of message passing features in GRAS
3407 - A HOWTO section containing:
3408 o HOWTO design a GRAS application
3409 More are due, of course. They will come latter. In the meanwhile, you can
3410 check the examples which are still here.
3412 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3414 SimGrid (3.1) stable; urgency=high
3418 There was a stack corruption somewhere, visible only when optimizing
3419 with these versions. [Vince]
3422 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3423 module that enabled to play with Directed Acyclic Graphs. It is built
3424 directly on top of SURF and provides an API rather close to the old
3425 SG. Some old codes using SG are currently under rewrite to check that
3426 all needful functions are provided. [Christophe Thiery]
3429 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3430 extension of the classical CLM03 model that gracefully handles
3431 failures. This is now the default model for MSG and GRAS. It doesn't
3432 handle parallel tasks yet though. [AL]
3433 * Bug fix: Weights were not correctly set in the network part.
3434 WARNING: This may have resulted in incorrect results with simulations
3435 where there are more than one flow on a given link. [AL]
3438 * After a (long ?) discussion on simgrid-devel, we have decided that the
3439 convention we had on units was stupid. That is why it has been decided
3440 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3441 WARNING: This means that all previous platform files will not work as
3442 such with this version! A warning is issued to ask users to update
3444 A conversion script can be found in the contrib module of the CVS, under
3445 the name contrib/platform_generation/surfxml_update.pl [MQ]
3448 * Bug fix: Processes were started in reverse order, wrt deployment file.
3449 WARNING: if your code relies on this bug, please fix it. [AL]
3450 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3451 being executed on two different locations. [AL]
3452 * Bug fix: Failures are now better supported thanks to Derrick's tests
3453 (there was many failure situations I hadn't thought of and that weren't
3454 correctly handled). [AL]
3455 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3459 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3460 Exception occurring on server-side are propagated back to client (!).
3462 API CHANGE: the callback changed their prototype. Change:
3463 int my_handler(gras_socket_t expeditor, void *payload_data) {
3465 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3466 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3468 * New! function: gras_msg_handleall to deal with all messages arriving
3469 within a given period.
3470 * New! function: gras_socket_server_range to get a server socket in a
3471 range of port numbers (ease to avoid port number conflicts) [MQ]
3472 * New! gras processes display their backtrace when they get a SIGUSR1
3473 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3474 Sweet to debug RL processes [MQ]
3478 - Do not force experiment sizes to be expressed in kb, or it becomes
3479 impossible to measure the latency this way (needs one byte-long tests)
3480 WARNING: this changes the amok_bw_* function semantic. [MQ]
3481 - Implements the link saturation stuff. [MQ]
3482 * Peer management module:
3483 New! module factorizing code that we wrote over and over [MQ].
3486 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3487 - Test units are placed directly into the library code, they get extracted
3488 automatically and placed into the src/testall binary.
3489 - Convert most of the XBT tests to this system.
3490 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3491 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3492 live in the same C code. This required to change a little bit the API
3493 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3494 * New module: sparse graph structure with basic algorithms (this is work
3495 in progress and the API is not considered to be frozen yet). [AL]
3496 * Display more information on backtraces: source line & function names are
3497 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3498 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3499 * Reworked a little bit some #include statements to load only required
3500 headers. Some user code that relied on SimGrid to include stdlib or
3501 stdio may need to include it by themselves. [AL]
3502 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3504 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3505 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3506 but I guess that almost nobody used those parts. [MQ]
3508 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3510 SimGrid (3.0.1) stable; urgency=low
3513 * Unfortunately, I had missed 5 misnamed functions:
3514 xbt_fifo_item_t xbt_fifo_newitem(void);
3515 void xbt_fifo_freeitem(xbt_fifo_item_t);
3516 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3517 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3518 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3519 They're now deprecated. Please use their new versions:
3520 xbt_fifo_item_t xbt_fifo_new_item(void);
3521 void xbt_fifo_free_item(xbt_fifo_item_t);
3522 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3523 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3524 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3526 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3527 * Documentation: xbt_log module unmercifully reworked [MQ]
3528 * Bugfix: there was a problem with the ending of contexts with
3529 the pthread backend. It caused some weird deadlock or behavior
3530 depending on the pthread implementation. [AL]
3531 * Bugfix: get the exceptions raised in the simulator repport where
3532 and why they come from when they are not caught in time [AL, MQ]
3535 * Bugfix: Do repport the error when two non-connected hosts try to
3536 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3539 * Add additionnal checkings on communications. Assert that two
3540 communicating hosts are connected by a set of links... [AL]
3543 * Add additionnal checkings on channel values in communication [AL]
3544 * New: MSG_task_get_source to see on which host a task was generated [HC]
3545 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3546 the number of tasks waiting to be received on channel and sent
3548 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3549 waits for the first task coming from a given host.. [AL]
3551 GRAS new functionnalities: [MQ]
3552 * Enhance the parsing macro to allow the size of multidimentional objects
3553 to be given thru annotations.
3554 * New example (and documentation): Matrix Multiplication a la RPC
3555 (as when I was young!) and fix a bunch of bugs found on the way.
3557 GRAS performance improvements: [MQ]
3559 * Reduce the amount of cbps creation/destruction by making it static to
3560 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3563 * Change libdata to a set so that we can search for stuff by ID (and thus
3564 reduce the insane amount of dict lookups)
3567 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3568 RL communication on top of SG since it's so uneffective.
3569 It may also be used for inter-thread communication in RL, one day.
3570 * Use gras_datadesc_copy() to exchange messages on top of SG
3572 - improve message exchange performance on top of SG
3573 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3574 * Don't exchange on the network the size of the used part of buffer,
3575 instead, specify the possible buffer size to read().
3577 - reduces the amount of read/write calls (one pair per exchange)
3578 - reduces the amount of exchanged data (the size)
3579 - allows to retrieve all arrived data on receiver side, if we don't need
3580 it right now (subsequent read will peek the buffer)
3581 - allows the receiver to proceed with the beginning of the stream before
3582 everything is arrived
3583 - make it possible to build an iov transport (using readv/writev)
3585 - take care of the data with non-stable storage (like stacked data),
3587 * If possible, TCP send uses vector I/O (when writev() is here)
3588 - Don't use it for receive since we send data sizes and data on the
3589 same stream, so we wouldn't be able to chain large amount of chunks
3590 before having to flush the stuff to read the size.
3591 * Rework the transport plugin mechanism to simplify it and reduce the
3592 amount of pointer dereferencement when searching for the right function
3595 * I guess that now, we do almost as few system calls as possible while
3596 doing as few data copy as possible.
3598 To improve it further, we could try to send all the sizes first and then
3599 all the data (to use iov on receiving size), but it's only a partial
3600 solution: when you have 2 dimensional data, the sizes of the second
3601 dimension is data of the first dimension, so you need 3 streams.
3603 I'm not sure the potential performance gains justify the coding burden.
3605 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3607 SimGrid (3.00) stable; urgency=high
3610 * New! Give the possibility to hijack the surf parser and thus bypass
3611 MSG_create_environment and MSG_launch_application. Have a look at
3612 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3614 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3616 SimGrid (2.96) unstable; urgency=low
3621 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3622 This deprecates the xbt_error_t mechanisms.
3623 It modifies (simplifies) all XBT and GRAS API.
3624 MSG API keeps unchanged (exceptions raised by XBT are caught from
3625 within MSG and masked with existing error handling facilities)
3628 * New! Add a FATPIPE model. [AL]
3629 * New! Add a parallel task model. [AL]
3630 * New! Add automatically a loopback interface (in the default
3631 network model) if none was precised.
3634 * Bugfix: MSG_process_resume now works with the current running process.
3636 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3637 * Modification of MSG_task_get_compute_duration. Once a task has been
3638 processed, the value returned by this function is now equal to 0. [AL]
3639 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3640 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3641 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3642 of a m_task. For now, MSG_task_cancel only works with computation
3644 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3645 of the processor (in Mflop/s) regardless of the current load on the
3647 * API Change: use proper naming convention for MSG_getClock and
3648 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3650 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3651 This function changes the priority of a computation task. This priority
3652 doesn't affect the transfer rate. A priority of 2 will make a task
3653 receive two times more cpu power than the other ones. This function
3654 has been added to suit the needs of Nguyen The Loc and hasn't been that
3655 much tested yet. So if it fails, please report it and send me your code.
3657 * API Change: removed all functions and types that were marked "deprecated"
3658 since many months. Renamed MSG_global_init_args to MSG_global_init.
3660 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3662 SimGrid (2.95) unstable; urgency=low
3665 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3666 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3667 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3670 * Bugfix: Allow absolute paths for platform description files [MQ]
3671 * Bugfix: do free the variables after use. Leads to drastic performance
3673 * Implement max_duration (ie, timeouts) on resources [AL]
3676 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3678 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3679 the needed support (timeouts) [AL]
3682 * Implement measurement sockets. You can now get the bandwidth between two
3683 hosts thanks to AMOK (see below). [MQ]
3684 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3685 dynar over the network (yeah) [MQ]
3686 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3687 * simple/static token-ring example. [Alexandre Colucci and MQ]
3688 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3689 in place before [MQ]
3692 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3693 type parsing. "short" and "long" modifiers were also ignored; other
3694 modifier (such as reference level) are still ignored. [MQ]
3695 * Embeed the buffer size within the buffer itself on SG. [MQ]
3696 That way, send() are atomic and cannot get intermixed anymore (at least
3697 the ones which are less than 100k; bigger messages still have the issue)
3698 * Array size pushed by the field, not by the field type (or each
3699 and every long int will push stuff to the cbps) [MQ]
3700 * use select() to sleep since it allows to portably sleep less than one
3703 GRAS (minor cleanups)
3704 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3707 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3708 * type category 'ignored' killed as it was never used and were difficult
3710 * whether a type can cycle or not is now a flag, leaving room for more
3711 flags (as "ignored", if we feel the need one day ;)
3712 * Rename raw sockets to measurement sockets since "raw" has another
3713 meaning in networking community.
3716 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3717 and offers features not belonging to GRAS but that most applications
3718 need. One day, it may be a set of plugins loadable at runtime.
3719 * New module: bandwidth
3720 bandwidth measurement between arbitrary nodes running this module. [MQ]
3722 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3724 SimGrid (2.94) unstable; urgency=low
3726 The first beta release of SimGrid 3 !
3730 * Update the main page and the FAQ. Adding references to gforge.
3733 * Add a gras_os_getpid function.
3736 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3737 * Extend the logs so that they also print PID, hostname, date, ... if
3739 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3740 and kill the old version which were in testsuite/
3741 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3745 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3746 for users. Instead, each programming environment (GRAS, MSG,...) use it
3747 when needed to provide such a feature to users.
3748 Don't shortcut the mechanism or you will also shortcut the virtualization
3749 you need on the simulator.
3753 * Cleanups in configury with regard to compile optimization/warning flags.
3754 Also add -fno-loop-optimize to any powerpc since it's the optimization
3755 killing gcc (< 3.4.0).
3756 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3757 needed by MSG examples complications
3758 * Borrow configury beautifications from PHP
3761 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3762 warning (thanks loris for stumbling into this one).
3763 * Bugfix: stop loading private headers (gras_config.h) from the public
3767 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3768 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3769 RL would imply a DNS resolver, which is *hard* to do in a portable way
3770 (and therefore delayed).
3771 * Implement a real timer mechanism and use it in timing macros. This allows
3772 to avoid rounding errors and get a 0.000005 sec precision in timing
3773 macros. While I was at it, various cleanups:
3774 - allow to declare more than one timed section per file (fix a stupid bug)
3775 - move some private declaration to the right place
3776 - merge conditional execution and timing macros into emulation module
3777 - document the module
3778 - make sure the module cleanups its mess on gras_exit
3779 * Documentation improvements:
3780 - (new) how to compile applications using GRAS
3781 - (new) emulation support (timing macros)
3783 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3785 SimGrid (2.93) unstable; urgency=low
3787 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3790 - Use Paje properly where used. Still to be sanitized properly.
3791 - Portability fix: Add an implementation of the contexts using pthread
3795 - Add xbt_procname(): returns the name of the current process.
3796 Use it to show the current process's name in all logging.
3798 - fix detection of older flex version and the reaction, since we do
3799 depend on modern ones (we use lex_destroy)
3800 - Better separation of SG and RL in the libs: remove all simulation code
3801 from libgras. As a result, this lib is now only 200k when stripped.
3802 Some of the xbt modules may also be duplicated (two sets and such) and
3803 should be cleaned/killed before SG3.
3804 - Insist on using xlC on AIX because of weird problems involving gcc there.
3805 - Cleanup the make remote stuff. This is now done by scripts
3806 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3807 mainly for our private use, but we're working on changing them to user
3810 - Bugfix: flush the socket on close only if there is some *output*.
3811 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3813 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3816 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3818 SimGrid (2.92) unstable; urgency=low
3820 Alpha 3 on the path to SimGrid 3
3824 - New! First try of benchmarking macros.
3825 - New! First try so that gras_stub_generator generate deployment and
3826 remote compilation helpers.
3828 - Bugfix: Initialization fix in msg_test.
3832 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3834 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3836 - API Change: gras_os_sleep to take the amount of seconds as a double.
3837 Accepting an int was error prone since it was the only location where
3838 seconds were coded as such. It leaded to damn rounding errors.
3839 - Bugfix: Hard to belive that timers ever worked before this.
3841 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3843 SimGrid (2.91) unstable; urgency=low
3845 Alpha 2 on the path to SimGrid 3
3849 - Bug fix in the lmm_solver.
3851 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3852 through the function MSG_paje_output.
3853 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3854 - It is possible to bound the rate of a communication in MSG with
3855 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3856 to write it in the changelog).
3857 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3862 - Various cleanups to the autotools stuff
3863 - Begin to move Gras examples to examples/gras/
3864 - Let make distcheck work again (yeah!)
3866 - documentation overhauled using doxygen.
3867 gtk-doc-tools is dead in SimGrid now.
3868 - Automatically extract all existing logging categories, and add the list
3869 to the documentation (long standing one, to say the less)
3871 - Cleanup the known architecture table. Reorder the entries to group what
3872 should be, and use a more consistent naming scheme.
3873 (some of the test dataset are still to be regenerated)
3874 - New! Allow library to register globals on each process just as userdata
3876 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3877 lookup time (for now).
3878 Use it in msg and trp.
3879 This cleans a lot the internals and helps enforcing privacy of the
3880 headers between the gras components.
3881 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3882 - Bugfix: gras_os_time was delirious in RL.
3883 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3885 - Reenable GRAS now that it works.
3887 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3889 SimGrid (2.90) unstable; urgency=low
3891 Alpha 1 on the path to SimGrid 3
3893 * It is a long time since the last release of SimGrid. I'm sorry about
3894 that but as I had told you, I was rewriting a lot of things. I apologize
3895 to those who had been reporting bugs to me and that I had not answered.
3896 If your bug is still in the new version, please tell me. Here is a
3897 summary of the main changes.
3899 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3900 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3901 portability, speed, and a lot more but you'll figure it out later.
3902 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3903 I wanted to release the new MSG as soon as possible and I have
3904 broken GRAS, which is the reason why, for now, only MSG is fully
3905 functional. A laconic description of these projects is available
3906 in the documentation.
3908 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3909 kernel optimized for our needs (called SURF but only the developers
3910 should use it). Hence, MSG is now roughly 30 times faster and I think
3911 that by rewriting a little bit MSG, I could event speed it up a little
3912 bit more. Beside the gain in speed, it is also much easier to encode a
3913 new platform model with SURF than it was with SG. More to come...
3915 * REVOLUTION 3: I have tried to change a little as possible the API of
3916 MSG but a few things really had to disappear. The main differences
3917 with the previous version are:
3918 1) no more m_links_t and the corresponding functions. Platforms are
3919 directly read from an XML description and cannot be hard-coded
3920 anymore. The same format is used for application deployment
3921 description. The new format is described in the documentation.
3922 Have a look in tools/platform_generation. There is a tiny script
3923 that converts from the old platform format to the new one. Concerning
3924 the application deployment format, parsing the old one is tricky.
3925 I think most of you should however be able to convert your files. If
3926 it is really an issue, I can write a C code that does the conversion.
3928 2) the toolbox tbx does not exist anymore. We now have a library
3929 with much more data-structures but without the hash-tables (we have
3930 dictionaries that are much faster).
3932 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3934 *****************************************************************************
3935 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3936 * don't feel like converting the oldies. (Mt) *
3937 *****************************************************************************
3940 Version 2.90: "the long awaited one"
3941 - Finished rewriting and debugging MSG. Rewrote the documentation.
3942 - disable GRAS for now since it needs to be ported to the newest SG
3945 - Finish the port to windows (using mingw32 for cross-compile)
3948 - Main loop and datastructures of SURF. A cpu resource object is
3949 functional. Surf can thus be used to create cpu's with variable
3950 performance on which you can execute some actions.
3952 2004-11-15 Martin Quinson
3953 - Port to ARM. Simply added the alignment and size descriptions. Should
3954 work, but the ARM machines are so slow that I didn't had the opportunity
3955 to 'make check' over there yet.
3957 2004-11-15 Arnaud Legrand
3958 - Trace manager now written. It uses a heap structure and is therefore
3959 expected to be efficient. It may however be speeded up (particularly
3960 when many events occur at the same date) by using red and black
3961 trees. One day maybe...
3962 - Max-min linear system solver written. It uses a sparse matrix
3963 structure taking advantage of its expected use. Most operations are
3964 O(1) and free/calloc are called as few as possible. The computation of
3965 the minimum could however be improved by using a red and black tree
3968 2004-11-03 Arnaud Legrand
3969 - Rename every gras_* function that was in xbt/ to its xbt_
3971 - Add a heap and a doubly-linked list to xbt
3972 - Added a dichotomy to the dictionaries. make check works as well before
3973 so I assume that the patch is correct. I do not know however if things
3974 run effectively faster than before now. :)
3976 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3977 SimGrid, and the version number is bumped to 2.x
3979 2004-10-29 Martin Quinson
3980 - Introduction of the remote errors.
3981 They are the result of a RMI/RPC on the remote machine.
3982 ErrCodes being scalar values, you can't get the host on which those
3983 errors did happen. Extending the error mechanism as in Gnome is possible.
3984 No idea yet whether it is a good idea.
3986 2004-10-28 Martin Quinson
3987 - Interface revolution: the Starred Structure Eradication.
3988 I used to do typedef struct {} toto_t; and then handle *toto_t.
3989 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3990 - ???_t is a valid type (builded with typedef)
3991 - s_toto_t is a structure (access to fields with .)
3992 - s_toto is a structure needing 'struct' keyword to be used
3993 - e_toto_t is an enum
3994 - toto_t is an 'object' (struct*)
3996 typedef struct s_toto {} s_toto_t, *toto_t;
3997 typedef enum {} e_toto_t;
3998 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3999 s_toto_t) is private.
4001 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4002 so that it can set the variable to NULL. It was so for dicts and sets,
4003 it changed for dynars.
4005 - Fix a bunch of memleaks in dict_remove
4006 - Fix a bug in sg/server_socket opening: it failed all the time.
4008 2004-10-07 Martin Quinson
4009 - Speed up dynar lookup operation a bit.
4011 gras_dynar_get is dead.
4013 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4014 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4015 (which gives you the address of the stored data).
4017 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4018 copy of the data using an affectation to do the job and not a memcpy.
4020 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4021 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4023 It does not really speedup the dynar test because they are
4024 setting elements all the time (and look them seldom). But the dict does
4025 far more lookup than setting.
4027 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4029 2004-10-05 Martin Quinson
4030 - Allow to (en/dis)able the cycle detection at run time.
4032 Whether we should check for cycle or not is now a property of each
4033 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4034 datadesc_cycle_unset allow to remove this property when previously set.
4036 Note that the cycle detection is off by default since it impacts the
4037 performance. Watch the data you feed GRAS with ;)
4039 This property is hereditary. Any element embedded in a structure having it
4040 set have it set for the time of this data exchange.
4042 You should set it both on sender and receiver side. If you don't set it on
4043 sender side, it will enter an endless loop. If you forget on receiver
4044 side, the cycles won't be recreated after communication.
4046 - Header reorganization.
4047 Kill gras_private.h, each submodule must load the headers it needs.
4049 2004-10-04 Martin Quinson
4050 - Interface revolution: do not try to survive to malloc failure.
4052 Now, gras_malloc and friends call gras_abort() on failure.
4053 As a conclusion, malloc_error is not a valid error anymore, and all
4054 functions for which it was the only gras_error_t return value are
4055 changed. They now return void, or there result directly.
4056 This simplify the API a lot.
4058 2004-09-29 Martin Quinson
4059 - Re-enable raw sockets.
4060 Created by gras_socket_{client,server}_ext;
4061 Used with gras_raw_{send,recv}
4064 It should allow to kill the last bits of gras first version soon.
4066 This is not completely satisfactory yet (duplicate code with
4067 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4070 - Simplify transport plugin (internal) interface by not passing any
4071 argument to _server and _client, but embedding them in the socket
4074 2004-09-28 Martin Quinson
4075 - Finish the port to AIX.
4076 autoconf was my problem (segfault within the malloc replacement
4077 function. No idea why)
4079 2004-09-16 Martin Quinson
4080 - Fix some size_t madness on 64bit architectures.
4082 2004-09-08 Martin Quinson
4083 - Reduce the number of system headers loaded, overload some more system
4084 calls (such as malloc to cast the result of the system one, and work
4086 - Fix and reintroduce the config support
4088 2004-09-07 Martin Quinson
4089 - Source code reorganization to allow Arnaud to surf all over there.
4090 - Allow to document the logging categories.
4091 - Remove all uppercase from logging categories and useless cleanup in names.
4093 2004-08-18 Martin Quinson
4094 Version 0.6.2 (protocol not changed; API changed)
4095 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4096 gras_error_t), and NULL when not found. Functions expecting a msgtype
4097 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4098 hopefully usefull message.
4099 - Portability to prehistoric sparcs again
4101 2004-08-17 Martin Quinson
4102 Version 0.6.1 (protocol not changed; ABI not changed)
4103 - prealloc some buffers to speed things up
4105 2004-08-11 Martin Quinson
4106 Version 0.6 (protocol not changed; ABI expended)
4107 - The parsing macro can deal with the references, provided that you add
4108 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4110 2004-08-09 Martin Quinson
4111 Version 0.5 (protocol not changed; ABI changed)
4112 - Allow to off turn the cycle detection code in data exchange at
4113 compilation time. It should be at run time, but I'm short of time (and
4114 the config stuff is still broken). That way, we keep dict out of the
4115 critical path, which is good because the performance is poor:
4116 - search not dichotomial yet
4117 - dynar give no way to access their content and memcpy everytime
4118 - In composed data description (struct, ref and so on), stop foolness of
4119 keeping the subtype's ID, but store the type itself. This keeps sets out
4120 of the critical path, which is good since they rely on dynar and
4121 dictionaries. The only loose of that is that we cannot detect the
4122 redeclaration of a structure/union with another content (but I'm not sure
4123 the code detected well this error before anyway). We still can detect
4124 the redefinition discrepancy for the other types.
4125 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4126 it breaks the code because of type-punning used all over the place).
4127 This breaks on all non-gcc architectures (for now).
4129 All those changes (plus the buffer of last time) allow me to gain 2 order
4130 of magnitude on cruel tests consisting of 800000 array of integers on two
4131 level of a hierarchical structure (200 secondes -> 4 secondes)
4134 - the selector of reference must now return the type it points to, not
4135 the ID of this type.
4137 2004-08-06 Martin Quinson
4138 Version 0.4 (protocol changed; ABI not changed)
4139 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4140 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4141 - Implement a buffer transport, and use it by default (it relies on tcp in
4142 real life and on sg in simulation).
4143 That's a bit hackish since I had a new field to the structure to store
4144 its data without interfering with the subtype ones. Inheritance
4145 is tricky in C. And that's a kind of reverse inheritance with one class
4146 derivating two classes. Or maybe a game with java interfaces. Anyway,
4147 that's damn hard in C (at least).
4148 Moreover, I got tired while trying to ensure plugin separation and
4149 genericity in SG mode. MSG wants me to do weird things, so let's go for
4150 cruel hacks (temporarily of course ;).
4151 See comment in transport_private.h:71
4152 - do not include all the _interface headers in private but in the files
4153 which really need them (to cut the compilation time when they are
4156 2004-07-26 Martin Quinson
4157 Version 0.3 (protocol not changed; ABI changed)
4158 - Major overhault of the datadesc interface to simplify it:
4159 - shorted the function names:
4160 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4161 - add a trivial way to push/pop integers into the cbps without malloc.
4162 This allows to make really generic sub_type description, which simply
4163 pop their size of the stack.
4164 - add a function gras_datadesc_ref_pop_arr() which does what users want
4165 most of the time: Declare a dynamic array (which pops its size of the
4166 stack) and declare a reference to it. Poor name, but anyway.
4167 - kill the post-send callback, add a post-receive one
4169 2004-07-23 Martin Quinson
4170 Version 0.2 (protocol changed; ABI changed)
4171 - add some testing for cpbs in the test cases, and fix some more bugs.
4172 This invalidate again the little64 data file, since I cannot regenerate
4174 - remove an awfull optimization in the logging stuff, allowing me to:
4175 - understand it again
4176 - learn gcc how to check that the argument match the provided format
4177 - fix all errors revealed by gcc after that
4178 - internal keys of dict are not \0 terminated. Deal with it properly in
4179 loggings instead of segfaulting when the user want to see the logs :-/
4181 2004-07-22 Martin Quinson
4182 - Fix some stupid bug preventing cbps (callback postit) from working
4184 2004-07-21 Martin Quinson
4185 - Some documentation cleanups
4186 - remove the useless last argument of msgtype_declare
4187 - rename the Virtu functions to fit into the 'os' namespace
4188 - move headers src/include -> src/include/gras/ and stop fooling with
4189 gras -> . symbolic link
4190 - make distcheck is now successful
4192 2004-07-19 Martin Quinson
4194 - Build shared library also
4195 - Install html doc to the right location
4196 - stop removing maintainer files in make clean
4197 - build tests only on make check
4199 2004-07-13 Martin Quinson
4201 - No major issue in previous version => change versioning schema
4202 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4203 corresponding dataset.
4205 2004-07-11 Martin Quinson
4207 - Get it working with any kind of structure (we can compute the padding
4208 bytes remotely for all the architectures I have access to)
4209 - Implement the structure parsing macro (still not quite robust/complete)
4210 - Improvement to the remote testing toysuite
4212 2004-07-10 Martin Quinson
4213 [autoconf mechanism]
4214 - get ride of a bunch of deprecated macros
4215 - actually run the test for two-compliment, not only compile it :-/
4216 - test whether the structures get packed (and bail out if yes. Damn.
4217 Alignment is a serious matter)
4218 - test whether the structures get compacted (but respecting the alignment
4219 constraints of each types)
4220 - test whether the array fields of structures can straddle alignment boundaries
4222 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4223 (took me 2 hours to find that bug, looking at the wrong place)
4224 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4225 before sending/receiving and are used to compute the offsets of fields
4226 - Given that padding size depend even on compiler options, keep track of
4227 alignment and aligned_size only for the current architecture. Not a big
4228 deal since we send structure fields one after the other (seems
4230 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4231 program, let it work on linux/gcc/little32. portability todo.
4233 2004-07-08 Martin Quinson
4234 - import and improve remote compilation support from FAST
4235 - make sure make check works on half a dozen of machines out there
4237 2004-07-07 Martin Quinson
4238 Let's say it's version 0.0.3 ;)
4239 - Implement conversions (yuhu!)
4240 - Let it work on solaris (beside conversion, of course)
4241 - Stupid me, using rand() to generate the conversion datatests in not wise.
4243 2004-07-06 Martin Quinson
4244 - Let make dist work, since I'm gonna need it to compile on remote hosts
4245 - Let Tests/datadesc_usage write the architecture on which the file was
4246 generated as first byte.
4247 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4248 architecture descriptions.
4249 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4250 on those architectures.
4251 - Optimization: send/recv array of scalar in one shoot
4253 2004-07-05 Martin Quinson
4254 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4256 - Plug a whole bunch of memleaks
4257 - each process now have to call gras_{init,exit}. One day, their log
4258 settings will be separated
4259 - Continue the code factorisation between SG, RL and common in Transport.
4261 2004-07-04 Martin Quinson
4263 - Redistribution between SG and RL.
4264 We wanna have to accept in SG, so move accepted related parts of RL in
4265 the common part. (more precisely, the dynar of all known sockets is no
4266 more a static in transport.c, but part of the process_data)
4268 [gras_stub_generator]
4269 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4271 2004-07-03 Martin Quinson
4272 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4273 them all up in one shot)
4275 - Fix the ugly case of reference to dynamic array.
4276 - New (semi-public) function gras_datadesc_size to allow the messaging
4277 layer to malloc the needed space for the buffer.
4279 - gras_socket_close now expect the socket to close (and not its address to
4280 put NULL in it after it). This is because the socket passed to handlers
4281 is one of their argument (=> not writable).
4283 - propagate the interface cleanup from last week in datadesc, ie remove a
4284 superfluous level of indirection. User pass adress of variable
4285 containing data (both when sending and receiving), and not of a variable
4286 being a pointer to the data. Let's say that I like it better ;)
4287 The price for that is constructs like "int msg=*(int*)payload" in
4288 handlers, but it's a fine price, IMHO.
4290 - Let it work in RL (yuhu)
4292 2004-06-21 Martin Quinson
4294 - porting SG plugin and SG select to new standards (works almost).
4295 - plug memleaks and fix bugs around.
4298 - cleanup the prototype of data recv and force users to specify when they
4299 want to handle references to objects. Test case working even for cycles.
4300 - plug memleaks. Valgrind is perfectly ok with this.
4302 2004-06-12 Martin Quinson
4304 - cleanup the separation between plugin and main code in plugin creation
4306 2004-06-11 Martin Quinson
4308 - Reput hook for raw sockets, needed for BW experiments
4309 - kill a few lines of dead code
4310 [Data description] Interface cleanup
4311 - gras_datadesc_by_name returns the searched type or NULL.
4312 That way, no variable is needed to use a type desc once, which makes
4314 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4315 two parameters were strings (field name, type name), leading to
4317 [Dicos] Interface cleanup
4318 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4319 This is consistant with the dynar API.
4321 2004-04-21 Martin Quinson
4323 - Porting to new standards.
4325 - interface cleanup.
4326 There is no bag anymore, no need to take extra provision to mask the
4327 pointers behind "ID".
4328 Better splitup of functions between files create/exchange/convert.
4329 This is still a bit artificial since convert and receive are so
4330 interleaved, but anyway.
4332 - add a queued message list to procdata (the ones not matching criteria
4334 - factorize some more code between SG and RL wrt procdata
4336 - use gras_exit in example to track memleaks
4337 - get rid of gs_example now that GS is properly integrated into gras
4338 - update run_test to integrate the latest tests (datadesc)
4340 - rename WARNINGn macros to WARNn since it prooved error-prone
4342 2004-04-19 Martin Quinson
4344 - register init/exit functions within gras module mechanism
4345 - send/receive function.
4346 Convertion is not implemented, but short-cutted if not needed.
4347 struct/array elements are sent one by one (instead of block-wise), but
4348 nobody really cares (yet). Get a prototype before optimizing.
4349 - tests (using a file socket) for DD send/receive on:
4350 - base types: int, float
4351 - array: fixed size, string (ie ref to dynamic string)
4352 - structure: homogeneous, heterogeneous
4353 - chained list, graph with cycle
4354 Believe it or not, valgrind is not too unhappy with the results. The
4355 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4359 - Bugfix when using a filename explicitly (instead of '-')
4361 2004-04-09 Martin Quinson
4363 - factorize more code between RL and SG in socket creation
4364 - Complete the implementation and tests of:
4366 o file (only in RL, and mainly for debugging)
4368 I lost 3 days to design a portable address resolver, and then decided
4369 that the prototype mainly have to run on my box.
4370 Addressing portability too early may be like optimizing too early :-/
4372 - use gras_init in the Tests instead of the crappy parse_log_opt
4373 (the latter function is removed)
4374 [Conditional execution]
4375 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4376 [Code reorganisation]
4377 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4378 Build examples against the RL library, since there is no way to disable
4379 its creation for now.
4381 For information, the beginning of coding on GRAS was back in june
4382 2003. I guess that every line has been rewritten at least twice since