1 ----------------------------------------------------------------------------
3 SimGrid (3.27) March 29. 2021
5 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
7 Main user-visible changes:
8 - All functions marked as deprecated for one year were removed.
9 - Removed all that was related to the "Storage" type in all APIs and bindings.
10 "Disks" have been introduced more than a year ago to replace them for the better.
11 It's time to finish this replacement.
12 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
13 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
14 avoiding "initializer element is not constant" errors with some codes. Please be
15 careful to recompile your MPI codes if switching to this new SimGrid version.
16 - Network model ns-3 now requires version ns-3.28 or higher.
17 - Removed some option aliases that were deprecated since v3.23 (June 2019)
18 - 'verbose-exit' is now 'debug/verbose-exit'
19 - 'simix/breakpoint' is now 'debug/breakpoint'
20 - 'clean-atexit' is now 'debug/clean-atexit'
23 - New functions: Comm::sendto_{init,async} to initiate a communication
24 on between two (possibly remote) hosts.
25 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
26 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
27 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
28 or Mailbox::get_async<void>() if you really want to play with void*.
29 - Unify the interface of Activity::on_{start/activity}
30 - New function: Comm::get_dst_data()
31 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
34 - ptask should now work better with multicore hosts (FG#37).
35 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
36 - Internal cleanups and refactoring around models, to prepare for programmatic
38 - Modernize our tests of the performance models
42 - New section: Configuring the logs from the command line.
43 - New section: Programming API of logs.
44 - New section: Getting realistic results
45 - New section: Examples of SMPI Usage
46 - New part: SimGrid's Internals, with a section on Design goals
47 - Move back to the classical Breathe tool chain, and fix many small glitches.
50 - Handle leak checking: SMPI will now display a message if MPI handles are still
51 present at the end of a simulation. Option smpi/list-leaks:n will display the n
52 first leaked pointers and their types.
53 - smpi/display-allocs option : Display Memory allocations analysis, with hints
54 about the largest allocations. This only works for allocations with malloc/calloc
55 and not compiled with SMPI_NO_OVERRIDE_MALLOC
56 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
57 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
58 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
59 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
60 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
61 - Replay now opens and parses the trace file only once for all ranks when possible.
63 - better handling of MPI_PROC_NULL
64 - MPI_Group_incl was broken in some cases
65 - PAPI tracing now fixed
66 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
67 - SMPI will now fail properly when receive size < sent size, instead of
68 just setting MPI_ERR_TRUNCATE and keep going.
69 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
70 - Remove useless build patches and split build/run patches on the page
71 - Build in parallel when possible
72 - Test on yum-based systems
73 - New nightly tested Proxy-apps:
74 - OpenMPI mpi-test-suite
75 - Coral MCB, LSMS, Qbox, CLOMP
76 - ECP/Mantevo Cloverleaf3D, Ember
77 - Codevault bhtree_mpi
80 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
81 - Start to overhaul the implementation to modernize it (still ongoing).
84 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
85 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
86 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
87 - Deprecate xbt_dynar_to_array().
88 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
90 C binding and interface:
91 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
92 - Constness requirement has been lowered for parameter argv of sg_actor_start()
93 and sg_actor_create() It's now (char* const*). The previous version is still
94 available as sg_actor_start_() and sg_actor_create_().
96 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
97 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
98 - FG#37: Parallel tasks are limited to 1 core per host
99 - FG#62: Running "smpirun -replay" on large networks
100 - FG!46: Fix a few potential memory leaks in SMPI colls
101 - GH#343: Wrong consumed energy when an exec timeouts
103 ----------------------------------------------------------------------------
105 SimGrid (3.26) December 15. 2020
107 The Release release (the French lockdown was eased today).
109 Important user-visible changes:
110 - SimGrid now requires a compiler with C++14 support.
111 Sibling projects should upgrade their FindSimgrid.cmake
112 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
113 several users had difficulties to understand issues when using high bandwidth or
114 small latency events. The new value was already the default for SMPI and
115 should not cause too much performance hit. It can change some simulation
116 timings if simulation was skipping a lot of small events (which was actually
117 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
118 - On some network models (L07, CM02, LV08), an implicit loopback link is created
119 for local communication. Default bandwidth was 500MBps and latency was 15µs,
120 which was rather bad. Default values have been changed to 10GBps and 0 latency.
121 Options network/loopback-lat and network/loopback-bw have been added to adapt
122 these values if needed. Network and routing models already providing separate
123 loopbacks (clusters and such) are not impacted by this change.
124 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
125 metadata that are added to each MPI message by MPI libraries (tag,
126 src/dest, and communicator information). This may change simulation times
127 for very small message sizes, but should be more realistic.
130 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
131 - New plugins: wifi_energy and link_load. See the associated examples.
132 - New examples: wifi networks, and communication suspend.
135 - LTO is now enabled for Intel/clang compilers.
136 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
137 Setting it to "auto" will use all available cores, while setting it to n will
138 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
139 - Remove obsolete runtime option 'contexts/parallel-threshold'.
140 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
141 has been kept for compatibility.
142 - Finally remove obsolete snake_case() aliases for runtime options.
143 - Further improve the documentation.
146 - the on_exit() of each actor is also executed when the simulation deadlocks.
147 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
148 An example is provided for s4u::Comm but it should work with Exec and Io.
151 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
152 https://framagit.org/simgrid/SMPI-proxy-apps
153 - MPI/IO: supports names starting with "./"
154 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
156 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
157 building shared libraries, this may cause issues, so environment variable
158 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
159 - most temporary files should now be created in /tmp dir (or equivalent).
160 If this one does not allow execution of code (noexec flag), this may cause issues.
161 Please use another tmp directory (using TMPDIR or equivalent system variable)
165 - A large refactoring is ongoing, but this should be transparent for users so far.
168 - Drop xbt_str_split().
170 C binding and interface:
171 - The return type of the sg_actor_on_exit() callbacks is now 'void'
172 (int value was previously ignored)
173 - Many C functions were renamed to follow a common naming style
174 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
175 names are kept until version 3.30.
176 - Many MSG tests were converted to the new S4U's interface in C, that
177 was extended for that.
179 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
180 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
181 - FG#41: Add sg_actor_create C interface
182 - FG#43: xbt::random needs some care
183 - FG#48: The Impossible Did Happen (yet again)
184 - FG#50: Suspending an actor executed at the current timestamp fails
185 - FG#52: Zero-seconds timeout : "That's in the past already"
186 - FG#53: Crash while using ns-3 network model
187 - FG#54: How to suspend a comm?
188 - FG!22: Stochastic Profiles
189 - FG!24: Documentation and fix for xbt/random
190 - FG!35: Add a modeling hint for parallel links in doc
191 - FG!36: [xbt/random] Read/Write the state of the RNG
192 - FG!42: Link cumulated load plugin
193 - FG!44: Test condition variable wait_until+notify_all
194 - GF#18137: Allow different stack sizes?
195 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
196 - GH#139: Allow pthread creation in SMPI
197 - GH#336: Packet-level simulation using SMPI?
198 - GH#345: Error in the LMM while migrating a VM
199 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
200 - GH#352: pip install / python setup.py install fail to find pybind11
201 - GH!337: Fix link_energy plugin for wifi platforms
202 - GH!339: Add Mailbox set_receiver method to python binding
203 - GH!344: Cast hugepages macros parameters to int64
205 ----------------------------------------------------------------------------
207 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
209 The Palindrom Day Release.
211 Important user-visible changes:
212 - Improve the Python usability (stability and documentation).
213 - A nasty synchronization bug was ironed out, see also below.
214 - Python's doc was organized as a separate tree, now integrated with C++.
215 - C bindings of S4U were not part of the doc.
216 - The C++ doc was also improved as methods are now split by theme.
217 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
218 - This is mandatory to use the Java bindings.
219 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
220 - SimDAG++: Automatic dependencies on S4U activities (experimental)
221 - Some features are already implemented but not all of them
222 - Cannot block an activity until it's scheduled on a resource
223 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
224 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
227 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
228 - Drop support for Viva/Triva (old visualization tools), hence removed
229 TRACE_get_node_types() and TRACE_get_edge_types() functions.
232 - Require pybind11 version 2.4.
233 - Greatly improve locking during context switches with Python.
234 - Declaring Python callbacks to this_actor::on_exit is now working.
235 - No more tests are failing currently, even on win32.
238 - Actor: Merge signals on_migration_start/end into on_host_change
239 - Actor: Rename migrate() into set_host()
240 - Disk: Allow users to get the read and write nominal bandwidth values
241 - Exec: Implement wait_for(timeout)
242 - Io: Implement wait_for(timeout)
245 - Parse errors now raise a simgrid::ParseError that you may want to catch.
248 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
249 been renamed to fix a spelling error. As usual, the previous definitions are
250 kept with a deprecation warning. However, to avoid runtime errors, it is now
251 forbidden to override these deprecated methods in a derived class. Please use
252 the new names immediately if you need to override them.
255 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
256 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
257 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
258 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
259 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
260 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
261 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
262 - New module: random, providing classical random numbers generators.
265 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
266 allocations' size above which they are considered "shared" by default (as if
267 they were performed through SMPI_SHARED_MALLOC macros).
268 Default = 0 = disabled feature.
269 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
270 This can cause some troubles if codes are already overriding these. If this
271 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
272 help, but will make this feature unusable.
273 - Input check is now performed much more consistently and outputs precise
274 warnings for each error found
275 - New supported calls : MPI_File_set_view, MPI_File_get_view
276 - MPI I/O now only supports disks, as storage is discontinued.
277 - MPI I/O allows opening files without specifying absolute path
278 - Note : SMPI configuration options are now initialized by smpi_init_options()
279 call, which should be called if SMPI is not being used through smpirun.
281 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
282 - FG#9: Python bindings crashing
283 - FG#39: Missing s4u::Comm::wait_any_for example
284 - FG#42: Add support for ThreadSanitizer (TSan)
285 - FG!19: Removing RngStream
286 - FG!20: A module for RNG calls
287 - FG!21: Choice between ad-hoc and standard distributions implementations
289 - FG!25: Fix link in Mutex doc
290 - FG!26: Fix links in SMPI interface doc
291 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
292 - GH#305: Unscheduled tasks are still excuted
293 - GH#313: smpirun: manual outdated w.r.t. --help
294 - GH#321: [S4U] Get task remaining work ratio
295 - GH#323: Crash when an actor turn off his physical host
296 - GH#335: Missing links on dragonfly example svg
297 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
298 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
300 ----------------------------------------------------------------------------
302 SimGrid (3.24) October 9. 2019
304 The Clean Disk Release.
307 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
308 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
309 thus becomes the default resource to execute I/O operations. Examples
310 and the file system plugin have been modified to use disks instead of
311 storage elements. Storage can still be used but is doomed to disappear.
312 Users are thus adviced to update their simulators accordingly.
313 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
314 for consistency with pthread_barrier_wait()
315 - Host::get_englobing_zone() returns the englobing netzone
316 - Actor::on_destruction is now called in the destructor
317 Actor::on_termination new signal called when the actor terminates
319 - Global signals are now part of the Engine:
320 - on_platform_creation: after config settings, before the XML parsing
321 - on_platform_created: right after the XML parsing
322 - on_time_advance: each time the clock advances
323 - on_simulation_end: after simulation, before cleanups
324 - on_deadlock: as the name implies.
326 - sg_{actor,host,link}_{data,data_set}() now all exist.
327 Use them to attach user data to the object and retrieve it.
330 - Introduce an experimental Wifi model. It sounds reasonable
331 according to the state of the art, but it still has to be properly
332 validated, at least against ns-3 if not against reality.
333 - Improved the usability of ns-3. Several bugs were ironed out.
334 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
335 It is now expressed as 'idle:epsilon:allCores' to properly model the
336 consumption of non-whole tasks on mono-core hosts. More info in the doc.
339 - convert a new set of functions to the S4U C interface and move the old MSG
340 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
343 - Fortran bindings for DVFS have been removed.
344 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
345 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
346 - SMPI can now be selected by cmake's find_module(MPI) with
347 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
348 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
349 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
350 - trace-call-location can be used with TI traces, and replayed, and is compatible
351 with smpi/comp-adjustment-file.
352 - sleep events are now correctly traced/replayed.
353 - Default for trace-call-location is now to use file names and not full paths.
354 To revert to previous behavior (in case of collision of filenames), option
355 "smpi/trace-call-use-absolute-path" can be set to yes.
358 - Use the included xxHash as an hash implem when C++14 is usable.
359 - Option model-checker/hash was removed. This is always activated now.
360 - New option smpi/buffering controls the MPI buffering in MC mode.
361 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
362 This is useful to check for MPI compliance.
365 - New section on plugins: how to define a new one, and existing ones.
368 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
369 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
372 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
373 and <mount> tags. The rationale is finish to decouple what should be handle
374 by SimGrid kernel (i.e., resources and activities) from plugins built in
375 user space (e.g., the file system plugin). This new tag currently accepts
376 only three attributes (id, read_bw, and write_bw). All the other information
377 that was declared with the storage related tags now has to be expressed as
378 properties. An example of platform using this new tag is available at
379 examples/platforms/hosts_with_disks.xml
382 - 'expect signal' can now accept more than one potential signal.
384 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
385 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
386 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
387 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
388 - FG#31: per-actor data
389 - FG#34: SG_BARRIER_SERIAL_THREAD?
390 - FG#35: model-checker does not like buster-produced binaries
391 - FG!13: MC: complete workaround in the error msg seen on modern systems
392 - FG!15: execute_flops now logs compute
393 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
394 - FG!17: ns-3: unblock the right number of communications + others issues
395 - FG!18: Improving the performance of the ns-3 bindings
396 - GH#189: Energy consumption of parallel tasks
397 - GH#219: Error in the throughput of TCP transfer
398 - GH!330: Minor polishing of English: Configuring_Simgrid
399 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
400 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
401 - GH!334: Rework energy plugin, again
403 ----------------------------------------------------------------------------
405 SimGrid (3.23.2) July 8. 2019
408 - Nicer introduction page.
409 - Migrate the "Deploy your application" page to the new doc.
410 - Move Java as a subtree of MSG.
413 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
416 - Drop xbt_dynar_sort_strings().
419 - Really fix FG#26: Turning off a link should raise NetworkFailureException
420 - FG#27: Wrong exception thrown to wait_any when link is turned off
421 - GH#328: Java: Canceling multiple tasks in a single vm/host
423 ----------------------------------------------------------------------------
425 SimGrid (3.23) June 25. 2019
427 The Exotic Solstice Release.
430 - SunOS and Haiku OS support. Because exotic platforms are fun.
431 - Stop setting random seed with srand48() at initialization.
432 - Use addr2line as a fallback for stacktraces when backtrace is not available.
433 - Build option -Denable_documentation is now OFF by default.
434 - Network model 'NS3' was renamed into 'ns-3'.
437 - Simgrid can now hopefully be installed with pip.
440 - wait_any can now be used for asynchronous executions too.
443 - New log appenders: stdout and stderr. Use stdout for xbt_help.
444 - Drop xbt_dict_dump.
447 - 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
448 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
449 - MPI interface is now const correct for input parameters
450 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
451 - Fortran bindings for async collectives.
452 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
455 - Remove option 'model-check/record': Paths are recorded in any cases now.
456 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
457 always sparse. That's an immense gain in memory, and can even be
458 faster because of cache effects. So there is no need to clutter the
459 code to allow the user to go for the unefficient mode.
462 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
463 models proved to be more accurate than these old experiments.
465 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
466 - FG#1: Broken link in error messages
467 - FG#2: missing installation documentation
468 - FG#3: missing documentation in smpirun
469 - FG#6: Python bindings not available on PyPI
470 - FG#7: simple cmake call requires doxygen
471 - FG#8: make python bindings an optional dependency
472 - FG#10: Can not use MSG_process_set_data from SMPI any more
473 - FG#11: Auto-restart actors forget their on_exit behavior
474 - FG#12: -Denable_lto=OFF doesn't disable LTO
475 - FG#13: Installs unstripped file 'bin/graphicator'
476 - FG#14: Installs the empty directory 'doc/simgrid/html'
477 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
478 - FG#17: Dead link in doc (pls_ns3)
479 - FG#20: 'tesh --help' should return 0
480 - FG#21: Documentation link on http://simgrid.org/ broken
481 - FG#22: Debian installation instruction are broken
482 - FG#26: Turning off a link should raise NetworkFailureException exceptions
483 - GH#133: Java: a process can run on a VM even if its host is off
484 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
485 - GH#326: Valgrind-detected error for join() when energy plugin is activated
487 ----------------------------------------------------------------------------
489 SimGrid (3.22) April 1. 2019
491 The Easter Christmas Release (this one is somewhat late).
494 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
495 - Not complete yet: asynchronous activities (amongst others) are still missing
496 - Still ongoing: the interface may change in the future. We need more
497 testers! Please report any glitches.
498 - No new project using Java should start now. Please switch to Python.
501 - Some of the internal cleanups may lead to speed improvements:
502 - The hard limitation on the amount of simulated actors with
503 Java+Mac was removed. Now, the available memory is the only limit.
504 - Our refcounting was tidyied, leading to 10% speedups in some cases.
505 - We are still working on making our code robust to the actor kills
506 and hosts' churn. Things are improving, but it's not perfect yet.
507 - Replaced our own code to display a backtrace (that was forking addr2line)
508 with the Boost.Stacktrace library.
509 You won't see your backtraces without this optional dependency.
510 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
511 - Stop setting random seed with srand() at initialization.
514 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
515 XML file version remains 4.2 since old files are still compatible.
518 - Process termination which was broken at version 3.21 has been repaired.
519 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
520 - Hide the examples into examples/deprecated. New users should use Python.
523 - Drop MSG_process_create_from_stdfunc() from the API.
524 This C++ function was a pimple in the C API, made necessary at some
525 point by the Java bindings. This is fixed now.
526 - Hide the examples into examples/deprecated. New users should use S4U.
527 - MSG_process_create and MSG_process_attach now crash if the host used
528 to run this process is off.
529 - Fix the protype of MSG_process_on_exit()
530 Now use (int,void*) callbacks instead of (void*,void*) ones.
531 The implementation was ways too messy to actually work, I guess.
534 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
535 too often and provide more accurate timings.
536 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
537 - MPI_Alltoallw support
538 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
539 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
540 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
541 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
543 - Basic implementation of generalized requests (SMPI doesn't
544 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
547 - Drop sg_cmdline. Please use xbt_cmdline instead.
548 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
549 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
550 OsSemaphore is implemented in a portable way with C++11 threads.
551 This should allow much more threads to be created at the same time,
552 allowing Mac and Java users to simulate many more actors.
553 - Implement the 'thread' factory with std::thread instead of xbt ones.
554 It is not possible to set the stack size with threads anymore, but
555 -fsplit-stack is the way to go nowadays when using threads.
556 - Drop the xbt_os_thread_t module (now unused)
557 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
558 - Drop xbt_str_join_array().
559 - Drop cunit, use Catch2 instead.
562 - Many cleanups in the kernel::activity namespace. This was long
563 overdue, and shall open the path to many future endeavors.
566 - #132: Java: a process can not shut down its own host
567 - #220: S4U: detached send causes exception if sender terminates soon after sending
568 - #261: Document the parameters of parallel execution's constructor
569 - #300: [s4u] BarrierPtr is missing
570 - #314: SMPI args internal cleanup
571 - #316: Fix a bug related to the CPU utilization of multi-core VM
572 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
573 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
574 - #325: Turning off a host has different behavior on sleeping actors and computing actors
576 ----------------------------------------------------------------------------
578 SimGrid (3.21) October 3. 2018
580 The Restarting Documentation (TRD) Release.
583 - Convert the user manual to Sphinx for improved usability.
584 Unstable version now hosted on https://simgrid.frama.io/simgrid
585 Many glitches remain; stay tuned.
586 - Restrict the Doxygen reference API to the installed header files.
587 (documenting all internals this way is overkill + doxygen scales badly)
588 - New tutorial on S4U (highly inspired from the old MSG one, but with
589 a git repository to fork as a starting point, and a docker image)
590 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
591 - The Developper manual is still to be converted (not compiled until then)
592 - Some parts are still missing in this conversion (such as the
593 platform and deployment sections) while others were blindly converted
594 and would need more love (such as the configuration flags).
595 Things will certainly further improve in the future releases.
598 - Fully reimplement auto_restart mechanism that was utterly broken
599 (fix #22, #131 and #281 that were all very related)
600 - Implement semaphores (fix #303 and #291)
601 - When creating an actor from a function and its parameters,
602 move-only parameters are not allowed anymore, as it would prevent
603 the actor to be restartable if its parameters are consumed on run.
604 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
605 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
606 IO operations that can be started, waited for, or canceled as a regular
608 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
612 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
613 - New host variable: 'core_count'
616 - Remove xbt_os_thread_specific features
617 - Remove portability wrapper to condition variables
618 - Remove xbt_os_thread_yield()
621 - MPICH collective selector now mimics MPICH 3.3b
622 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
623 - MPI_Init(NULL,NULL) is now allowed
624 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
625 - Command-line arguments should now be handled properly in Fortran simulations
626 - MPI Attributes and related callbacks should work in Fortran
627 - Apps using C + Fortran should now work
628 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
629 - Support MPI_Error_String
632 - Due to an internal bug, Msg.run() must now be your last line.
633 We hope to fix it in a future release, and we are sorry for the inconvenience.
636 - #22: Process autorestart seem to only work with CAS01 cpus
637 - #56: Feature request: dynamic SMPI replay
638 - #93: simgrid should not eat --help
639 - #111: How to change the loopback link characteristics?
640 - #116: Platform cluster radical
641 - #117: Platform cluster/host order
642 - #129: Replay_multiple on steroid
643 - #131: Java : autostart does not work
644 - #143: Setting a breakpoint at a given time
645 - #235: Network energy models should be integrated
646 - #264: Add ptask L07 resource tracing
647 - #271: Dynamic and manual replay of SMPI traces
648 - #279: Breakpoints option uses time deltas instead of absolute time
649 - #281: Daemonized actor no longer daemonized after an auto-restart
650 - #282: TIT convertor erase traces when using absolute path in the trace list file
651 - #285: segfault when a s4u actor kills itself with msg process tracing activated
652 - #286: Pajé traces are not exposing the number of cores
653 - #287: Command-line parsing should stop when encountering '--'
654 - #288: MPI_Init(NULL, NULL)
655 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
656 - #290: Method to check if a mailbox has a message ready to be consumed
657 - #291: [s4u] semaphores do not seem available yet
658 - #294: Alltoall collective implementation of mpich
659 - #295: Starting a Host as OFF is not supported
660 - #296: DTD too permissive
661 - #299: Add a s4u barrier example
663 ----------------------------------------------------------------------------
665 SimGrid (3.20) Released June 24. 2018
667 The proxy snake_case() release.
670 - The whole API is now using snake_case() instead of mixing it with camelCase.
671 Compatibility wrappers in place for one year.
672 - We took the opportunity to increase the consistency of the API.
673 Please report any remaining glitches.
674 - New function simgrid::s4u::Host::get_actor_count:
675 Returns the number of actors running on a specific host.
678 - SMPI is now tested with ~45 proxy apps from various sources, with none or
679 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
680 - Replay: The replay file has been re-written in C++.
681 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
682 supported. They are stored in the trace and used when replayed.
683 - Basic support of MPI_Cancel. Robustness not guaranteed.
684 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
685 (only for MPI_COMM_TYPE_SHARED).
686 - New option: smpi/privatize-libs, to add external shared libs to be privatized
687 by SMPI. They will be copied locally and loaded separately by each process.
688 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
689 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
690 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
691 with internal index optind with multiple processes. Only works if MPI_Init has
693 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
694 syntax without preprocessing tricks.
697 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
698 behavior tracing is no longer limited to MSG
699 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
700 this means that traces now need to be updated or re-obtained; both Irecv
701 and Isend lines in your traces have now in total 5 mandatory fields:
702 <rankid> <command> <to/from rankid> <tag> <size>
703 To update your traces, it suffices to add a 0 for the tag here.
704 - TIT now also supports waiting for a distinct request via MPI_Wait.
705 Wait/Test now wait for a specific request, not just the last one that was
706 issued. This unfortunately means another update, because we need to
707 identify which request you want to wait for. We do this via the
708 triplet (sender, receiver, tag), which needs to be added:
709 <rankid> <command> <sender> <receiver> <tag>
710 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
711 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
714 - The deprecation of MSG is ongoing (but this should not impact you).
715 Many MSG functions are now simple wrappers on the C API of S4U. If
716 you wish to convert your code to S4U, find the S4U counterparts of
717 your MSG calls in src/msg/msg_legacy.cpp.
718 - MSG can still be used, but won't evolve anymore.
721 - Allow to run the Link energy plugin from the command line with
722 --cfg=plugin:link_energy
723 - Rename Energy plugin into host_energy
724 - Rename Load plugin into host_load
727 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
728 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
729 - Improve performance when handling timeouts of simix synchros.
732 - Config: the C API is now deprecated (will be removed in 3.23), and
733 the C++ API has been slightly improved.
736 - Fix several build issues on OSX.
737 - Move simgrid_config.h to simgrid/config.h (old header still working)
740 - #143: Setting a breakpoint at a given time
741 - #150: Inconsistent event names in SMPI replay
742 - #258: daemonized actors hang after all non-daemonized actors have completed
743 - #267: Linker error on unit_tmgr
744 - #269: SMPI: tracing of MPI_Wait/all/any broken
745 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
747 ----------------------------------------------------------------------------
749 SimGrid (3.19.1) Released March 22 2018
751 The Fixed ABI Release.
753 As you may know, we are currently refactoring SimGrid in deep.
754 Upcoming SimGrid4 will be really different from SimGrid3: modular,
755 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
757 Our goal is to smooth this transition, with backward compatibility
758 and automatic update paths, while still progressing toward SimGrid4.
760 SimGrid remains open during works: The last pure SimGrid3 release was
761 v3.12 while all subsequent versions are usable alpha versions of
762 SimGrid4: Existing interfaces remain unchanged, but the new S4U
763 interface is budding and the internals are deeply reorganized.
765 Since 2015, we work hard to reduce the changes to public APIs. When
766 we need to rename a public library symbol in S4U, we let your
767 compiler issue an explicative warning when you use the deprecated
768 function. These messages remain for four releases, i.e. for one full
769 year, before turning into an error. Starting with v3.15, your can
770 also adapt to API changes with the SIMGRID_VERSION macro, that is
771 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
773 Starting with this v3.19.1, our commitment to reduce the changes to
774 the public interfaces is extended from the API to the ABI: a program
775 using only MSG or SimDag and compiled against a given version of
776 simgrid can probably be used with a later version of SimGrid without
777 recompilation. We will do our best... but don't expect too much of
778 it, that's a really difficult goal during such profund refactoring.
780 The difference between v3.19 and v3.19.1 is that the former was
781 accidentally breaking the ABI of MSG, while the later is restoring
784 S4U and kernel APIs will still evolve until SimGrid4, with one-year
785 deprecation warnings as currently. In fact, cleaning up these
786 interfaces and converting them to snake_case() is one release goal of
787 v3.20. But don't worry, we are working to smooth this upgrade path.
789 In summary, new projects should start with S4U to benefit of the
790 future, but old MSG projects should still be usable with no change.
792 ----------------------------------------------------------------------------
794 SimGrid (3.19) Released March 20 2018
796 The Moscovitly-cold Spring release.
798 SMPI user-visible improvements:
799 - Support of MPI_Type_create_subarray
800 - Unimplemented MPI calls now abort (instead of issuing a warning)
801 Some harmless calls around errhandlers still only issue a warning.
802 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
803 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
804 is now possible (mmap privatization was refactored to this end).
805 Adding new SMPI instances should also be technically doable now.
807 SMPI internal cleanups:
808 - Removed index notion from SMPI -- no more "getPid() - 1"!
809 In tracing files, actors are called rank-<PROCESS-ID> but these are
810 NOT the real ranks (tracing several communicators is not supported yet).
811 - Deprecate smpi_process_index() (will be removed in 3.22)
812 - Better testing on classical proxy apps, through an external project:
813 https://github.com/simgrid/SMPI-proxy-apps/
816 - Use a graphical TOC to make it easier to find the documentation you need
817 - Revamp the MSG tutorial
820 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
821 - Remove parameter of MSG_process_killall().
822 Resetting the PID was bogus anyway (several actors could have the same PID).
825 - Execution->setHost() can be called after start(), to migrate it.
826 - Comm::test_any() is now implemented.
827 - s4u::Actor now has onCreation() and onDestruction() signals.
828 - Install some kernel header files for the users' plugins and more.
829 Warning, their API is really not stable yet.
830 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
831 include/simgrid/kernel/routing/*Zone.hpp
834 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
836 Build System and other cleanups:
837 - Remove unused run-time parameter "tracing/onelink-only".
838 - SimGrid now works with intel compilers (MC still buggy)
839 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
840 - tesh kills subprocesses on timeout (but not on windows)
843 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
846 - #194: Feature request: simgrid::s4u::Comm::test_any()
847 - #245: migrating an actor does not migrate its execution
848 - #253: Feature Request: expose clusters as objects
849 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
850 - #255: Tesh broken on Windows
851 - #256: Modernize FindSimGrid.cmake
852 - #257: Fix (ab)use of CMake install
854 ----------------------------------------------------------------------------
856 SimGrid (3.18) Released December 24 2017
858 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
860 MSG IS NOW DEPRECATED. Please use S4U for new projects.
861 - Support for MSG should not be removed from SimGrid before 2020,
862 but future evolutions will be limited to the integration of
863 user-provided patches.
866 - s4u::onDeadlock() signal, run before stopping the simulation.
867 - s4u::Exec: asynchronous executions (abstraction of background
869 - s4u::Host->getLoad() returns the achieved speed in flops/s
872 - Link::name() is deprecated and will be removed in v3.21, use
873 Link::getCname() instead.
874 - Mailbox::getName() changed to return a std::string, use
875 Mailbox::getCname() instead to get a char*.
876 - Storage::getName() changed to return a std::string, use
877 Storage::getCname() instead to get a char*.
878 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
879 to have both allocation and deallocation of the map in user space.
882 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
883 The amount of flops remaining to do is a vector, not a scalar.
884 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
885 name implies on both sequential and parallel tasks.
886 - Both changes fix GitHub's #223 using PR #237 as a basis.
887 Thanks Michael Mercier.
888 - Most examples were converted to S4U and hidden elsewhere at tests.
889 Rationale: we still want MSG to work; we want newcomers to use S4U.
892 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
893 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
896 - Switch to the faster dlopen privatization mechanism by default
897 - Documentation improvments
900 - Live migration is getting moved to a plugin. Dirty page tracking is
901 the first part of this plugin. This imply that VM migration is now
902 only possible if one this function is called:
903 - C/MSG: MSG_vm_live_migration_plugin_init()
904 - C/C++: sg_vm_live_migration_plugin_init()
905 - Java: Msg.liveMigrationInit()
906 For C and C++, "simgrid/plugins/live_migration.h" also has to be
910 - Define class simgrid::xbt::Path to manage file names.
911 - Removed unused functions:
912 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
913 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
914 xbt_graph_export_graphviz()
915 - xbt/str.h: xbt_str_join()
916 - Remove unused datatypes:
917 - xbt/heap.h: use std::priority_queue or boost::heap instead
918 - xbt/swag.h: use boost::intrusive::list instead
921 - New link_energy plugin for the consumption of the links.
922 - All of the operations on files and storage contents have been
923 packaged into a plugin (src/plugins/file_system). The current
924 public interface can be found in
925 include/simgrid/plugins/file_system.h
926 To use these functions you now have to initialize the plugin by
927 calling MSG_storage_file_system_init() just after calling
928 MSG_init() or sg_storage_file_system_init() just after creating
932 - Remove the undocumented/untested tag <include>
935 - Remove viva specific tracing as the tool is no longer maintained
938 - #248: Kill of finished processes leads to segfault
939 - #240: xbt_cond_wait_timeout should gracefully return for C apps
940 - #239: please implement signal s4u::onDeadlock()
941 - #230: segfaults when exit() without run()
942 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
943 - #223: MSG_task_get_flops_amount() not working with parallel tasks
944 - #222: Actor::kill() doesn't really kill and segfaults
945 - #221: odd LMM warning when killing an actor
946 - #120: Memory leak when the processes are forcefully killed
948 ----------------------------------------------------------------------------
950 SimGrid (3.17) Released October 8 2017
952 The Drained Leaks release: (almost) no known leaks despite the tests.
954 Even very long simulations will preserve your memory: our testsuite
955 stresses SimGrid for over 45mn on fast machines for a coverage of
956 over 80%, and there is only one single known leak, of about 4kb.
960 - Comm.detach(): start and forget about asynchronous emission. A cleanup
961 handler may be given to free resources if the comm cannot be completed.
962 - this_actor::send(mailbox) is now mailbox->put()
963 - New: simgrid::s4u::Comm::wait_all()
964 - New: Host.execute() for remote executions.
967 - Unused option network/sender-gap is removed.
970 - Deprecate MSG_task_isend_with_matching(): This unused feature
971 really complicates our internals. Will be removed in v3.20.
974 - Improved context termination. It is now possible to free resources, even
975 when a process is forcibly killed.
976 - ContextBoost: add support for Boost versions above 1.61.
979 - Bring back run-time option --cfg=exception/cutpath to remove exception
981 - Removed unused functions:
982 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
983 xbt_str_rtrim(), xbt_str_trim().
984 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
987 - Removed header files obsolete since SimGrid 3.12:
988 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
990 - #3: SD_exit should be made optional
991 - #120: Memory leak when the processes are forcefully killed
992 - #159: Threading test regression in Actor refcounting
993 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
994 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
995 - #186: Actor::killAll() segfaults if some process is blocked on wait()
996 - #191: VM migration and pstate
997 - #192: Updating the energy consumptions of all hosts crashes with VMs
998 - #195: All actors have PID=0 in the logs
999 - #204: Sometimes segfault with thread contexts and mmap privatization
1000 - #222: Actor::kill() doesn't really kill and segfaults
1001 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1003 ----------------------------------------------------------------------------
1005 SimGrid (3.16) Released June 22. 2017.
1007 The Blooming Spring Release: developments are budding.
1010 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1011 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1012 - Java JAR file should be finally fully working on Mac OSX too.
1014 XML platforms: Switch to platform v4.1 format.
1015 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1016 - <zone> can be used as a synonym for the now deprecated <as>
1017 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1018 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1019 - <actor> can be used as a synonym for the now deprecated <process>
1020 - state_file and avail_file periodicity is now easier to express
1021 (check the documentation) the old behavior should still work.
1022 * Storage: not backward compatible, but it seems that nobody ever used it
1023 - Remove attribute 'content_type' of <storage_type>: was never used
1024 - Make attribute 'model' of <storage_type> optional: for future usage
1025 - Remove Bconnection model property: was never (in)validated. Replaced by
1026 the maximum of the read and write bandwidth as a resource constraint.
1029 - New and Backwards Compatibility break:
1030 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1031 When one wants to get the list of tasks whose states have changed during a
1032 simulation round, s/he has to allocate and free a dynar and use it as
1033 argument to this function. The former SD_simulate (double how_long)
1037 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1038 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1041 - New function to update the consumption of all hosts at once.
1042 - Fix the model for multi-core hosts, linear in the amount of busy
1043 cores with an abnormality for pIdle (see paper).
1044 BUG: Still not satisfactory for ptask on mono-cores.
1047 - The netzone are now available from the MSG API.
1048 The old names still work, but are now deprecated.
1051 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1052 - New option: smpi/keep-temps to not cleanup temp files
1053 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1054 mallocs mode is used, allows to change the size of the fake file used
1055 (default 1MB), to potentially limit the number of mappings for large runs.
1056 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1057 - Fortran ifort and flang compilers support
1058 - New RMA calls supported (experimental):
1059 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1060 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1061 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1062 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1063 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1064 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1067 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1068 - Behind the scene, ActivityImpl are now automatically refcounted.
1069 Removing this nasty bottleneck will greatly simplify our internals.
1072 - Replay: New function xbt_replay_action_get():
1073 Retrieve the function previously associated to an event type.
1074 - DROPPED FUNCTION: xbt_str_varsubst()
1075 - DROPPED MODULE: strbuff. We don't need it anymore.
1076 - DROPPED MODULE: matrix. We don't need it anymore.
1077 - DROPPED MODULE: lib. We don't need it anymore.
1079 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1081 SimGrid (3.15) Released March 22. 2017
1083 The Spring Release: continuous integration servers become green
1084 We fixed even the transient bugs on all target architectures:
1085 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1086 Windows, FreeBSD, NetBSD.
1088 - Rename NetCards to NetPoints.
1089 This was intended to help NS3 users, but that's not a netcard.
1090 That's a point in the routing algorithm, let's avoid wrong simplifications.
1093 - New: MSG_process_yield(). Stop and yield to other processes.
1094 - New: MSG_process_daemon(). Daemon processes are automatically killed
1095 when the last non-daemon process terminates
1096 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1097 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1098 to make room for the upcoming network energy plugin.
1099 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1102 - Ensure that an actor can kill itself with Process::exit()
1103 - Kill the obscure NativeException. Nobody want to survive the issues
1104 it denotes, so use JniException that is a RuntimeException (not to
1105 be caught explicitly).
1106 - Partial bug fix in initialization. SimGrid flags on command line were
1107 consumed at C level but stayed in the original Java String[] args.
1108 This could mess users' args[i] if SG flags were not put at the end of
1110 The SimGrid flags are now removed from the Java arguments. However,
1111 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1112 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1113 It might be if you have --log or --cfg flags in the command line.
1114 - Fix numerous memleaks all around the place. In particular, around VMs.
1118 - s4u::onPlatformCreated: right before the simulation starts
1119 - s4u::onSimulationEnd: right after the main simulation loop
1120 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1121 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1122 event from the availability_file changes the avail speed.
1123 - Links are now usable from s4u
1124 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1125 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1126 - New examples: The conversion of MSG examples to S4U has begun
1127 - Actors: create, kill, migration, and suspend
1128 - Applications: master-worker and token-ring
1129 - Action replay: communications and storage
1130 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1133 - Backwards Compatibility breaks
1134 - The SD_route_get_list and SD_route_get_size functions have been removed.
1135 They are replaced by the unique
1136 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1137 The route from src to dst is built in the links dynar whose size can
1138 be obtained with xbt_dynar_length.
1139 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1140 sg_host_route_bandwidth and sg_host_route_latency.
1141 Macros ensure the backwards compatibility, but you should fix your code
1144 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1145 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1146 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1147 We don't support 100% of it, but it should be enough. Contact us if not.
1148 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1149 naive and probably inaccurate)
1150 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1151 mapped onto a unique small file using some system magic.
1152 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1153 - Fortran: cleanups, fixes, support of user-added operations
1154 - RMA: MPI_Accumulate are applied in correct order.
1155 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1156 - Better support for MPI_IN_PLACE
1157 - Support for MPI_Win attrs and keyvals.
1158 - Support MPI_Comm_group_create, MPI_Type_size_x
1159 - MPI ops should be properly applied to all allowed types
1160 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1161 algorithm by a non failing one from mpich.
1164 - Kill the fifo data container: we don't use it anymore.
1167 - A <cluster> can now be created with different speed values to represent pstates.
1168 see examples/platforms/energy_cluster.xml for an example.
1170 SimGrid (3.14.159) Released December 28. 2016
1172 The previous release was only a crude approximation of a Pi release;
1173 this one should be a bit better.
1175 - Revert a last minute change that broke on Mac OSX.
1176 - Fix the changelog and NEWS files.
1177 - Improve the documentation, in particular of the routing module.
1178 - Integrate some patches flying around in Debian.
1180 SimGrid (3.14) Released December 24. 2016
1183 * The whole documentation was reworked and reorganized. There is
1184 still much room for improvement here, but we're on it.
1187 * We now need python3 for our testing infrastructure.
1188 * Model checking needs libevent
1191 * Now works on FreeBSD
1194 * Privatization now works on FreeBSD
1196 * Privatization is now activated by default in smpirun
1197 You can turn it off with -no-privatize if something goes wrong.
1199 * Call-location tracing for SMPI.
1200 You can add the exact location (filename / linenumber) of an MPI call to
1201 your trace files and slow down or speed up the simulation between two
1202 consecutive calls by using an adjustment file (see the documentation).
1204 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1205 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1206 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1207 values were ignored in some cases. The timings of these functions can now
1208 be significantly different.
1210 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1211 smpi/running-power is renamed to smpi/host-speed
1213 * smpi/grow-injected-times option to enable or disable multiplication of the
1214 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1215 make simulation less precise (but also much faster).
1217 * smpirun script should be (much) faster for large deployments.
1219 * SMPI tracing: fixed issue with poor matching of send/receives.
1221 * Replay: Fix broken waitall
1223 New functions and features
1224 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1226 Dropped / renamed functions and features
1227 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1228 - MSG_mailbox_is_empty() -> Mailbox::empty()
1229 - MSG_mailbox_front() -> Mailbox::front()
1230 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1231 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1232 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1233 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1234 * Don't pass the free_f parameter to property related functions:
1235 - MSG_host_set_property_value()
1236 - MSG_as_router_set_property_value()
1237 - MSG_storage_set_property_value()
1238 * VM properties. Since msg_vm_t are msg_host_t, just use
1239 MSG_host_get_property_value() and friends
1240 * VM I/O related things:
1241 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1242 - Unimplemented save/restore methods
1243 * MSG_as_router_get_property_value() was redundant with
1244 MSG_environment_as_get_property_value().
1245 - Removed MSG_as_router_*propert*() functions
1246 - Added MSG_environment_as_set_property_value() for consistency
1247 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1248 Well, they are still there for now, but deprecated with a warning.
1249 Please switch to xbt_dict_new_homogeneous() before this is removed
1251 * Task affinity. Its intended behavior (that was very badly tested
1252 and probably not really working) was deceiving what most users
1253 would have hoped here.
1254 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1255 * Option network/coordinates is now useless and should be dropped.
1258 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1261 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1262 * Added Dragonfly topology support
1265 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1266 EiBps, EBps, Eibps, Ebps for bandwidth.
1267 They may become useful to some lucky ones.
1270 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1271 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1272 * ex/async_waitAll: new example, on asynchronous communications
1275 * Memory usage should be decreased for simulations with a large number
1276 of processes. This also helps for SMPI.
1278 SimGrid (3.13) stable; urgency=low
1280 The Half Release, a.k.a. the Zealous Easter Trim.
1282 - We removed half of the lines, that were mostly experimental cruft.
1283 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1284 (not counting blanks and comments -- according to openhub.net)
1285 - The internals are now compiled in C++ (and will soon be clean C++)
1286 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1287 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1289 * Backwards Compatibility breaks
1290 - Removed Lua simulation bindings (switch to C or Java for that).
1291 Lua can still be used to describe platforms
1292 - Removed Java kernel plug-ins.
1293 Will be reintroduced after the ongoing major internals reorg.
1295 - the following functions were removed.
1296 They were too specific and should be reimplemented in a generic
1297 way, with filter function.
1298 - MSG_task_listen_from_host
1299 - MSG_mailbox_get_count_host_waiting_tasks
1300 - MSG_mailbox_put_with_timeout was removed.
1301 Please use MSG_task_send_with_timeout instead.
1303 - the SD_application_reinit function was removed. It has been a noop for a while.
1304 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1305 replaced by a more flexible mechanism.
1306 - The following functions thus do not exist anymore
1307 - SD_workstation_get_access_mode
1308 - SD_workstation_set_access_mode
1309 - SD_workstation_get_current_task
1310 - Basic estimation functions have been removed but can easily be replaced
1311 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1312 - SD_workstation_get_computation_time => amount / sg_host_speed()
1314 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1315 Use VM.getSpeed()*load/100 for the legacy behavior.
1317 - option enable_tracing was removed. It was not doing anything for a while.
1318 - In the ModelChecker:
1319 - the model-checker now ptraces the model-checked process which means
1320 you cannot use a debugger on the latter anymore (we might make this
1321 optional in the feature);
1322 - removed soft-dirty page tracking;
1323 - remove model-checked side snapshot management,
1324 MC_snapshot() and MC_compare_snapshot();
1325 - keep the MC_cut() function as a stub (it was not really working
1326 in the previous release).
1329 * All options are consistently in kebab-case. Old names are kept as alias.
1332 * Switch to platform v4 format.
1333 - Rename from 'power' to 'speed' the attributes describing the amount of
1334 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1335 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1336 - In <host> and <link>, attributes availability and state are gone.
1337 It was redundant with state and availability traces, and with peak values.
1338 - In <cluster>, attributes availability_file and state_file are gone.
1339 It was too complex and unused.
1340 - Kill <gpu>. Was not doing anything.
1341 - The DOCTYPE points to the right URL:
1342 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1343 (the file at this address now documents the changelog since its v1)
1344 - A warning is emitted for unit-less values (they are still accepted).
1345 - speed. Default: 'f' or 'flops'. Also defined:
1346 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1347 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1348 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1349 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1350 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1351 - latency. Default: 's' second. Also defined:
1352 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1354 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1355 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1358 * s4u::Host is now the preferred public interface to the Host features.
1359 sg_host_* functions are C bindings to the exact same behavior
1360 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1363 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1364 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1365 They were deprecated since a few years, and probably did not even compile anymore.
1368 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1369 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1370 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1371 encouraged to update their codes. A list of the modified functions can be found at the end of
1372 include/simgrid/simdag.h
1375 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1376 the lambda or closure passed as a parameter will run in kernel mode.
1377 All the callback functions should be rewritten to that interface at some point.
1380 * Reorganizing and cleaning the internals all around the place.
1383 * Remove old default barrier/bcast buggy algorithms (see #18407)
1384 * Various bug fixes to handle more codes
1385 * Remove the need for the --foreground option of smpirun (it is still
1386 accepted for backward compatibility).
1389 * Kill the setset data container: MC don't use it anymore.
1390 * Kill the queue data container: it made more sense with GRAS.
1391 * Kill the xbt_peer_t data type: it's useless without GRAS.
1392 * Kill rm_cb feature of config sets: it was never useful.
1393 * Kill graphxml parsing feature. It was not used.
1394 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1396 - xbt_dynar_sort_strings(), when the content is char*
1397 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1398 They throw exceptions on invalid input;
1399 * C++ support for declaring CLI flags (simgrid::config::Flag);
1400 * class for abstracting different signal backends (simgrid::xbt::signal).
1401 with no external dependencies (we need very simple signals).
1404 * refactoring and cleanup of the code;
1405 * ongoing process to cleanly separate the model-checking algorithms
1406 from the code model-checking support.
1408 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1410 SimGrid (3.12) stable; urgency=low
1412 The Facelift Release.
1415 * Require g++ v4.7 at least to not speak prehistorical C++.
1416 * Require Boost 1.48 (for signal2 component).
1417 * Java must be version 7 at least when activated.
1418 * Builds on Windows again (including Java bindings).
1419 * Tracing is now always enabled (no way to turn it out)
1420 * Remove GTNetS. It was not working anyway.
1421 * Various cleanups in the cmake scripts.
1422 * Move headers around to sort them out on installed systems:
1423 - instr/instr.h -> simgrid/instr.h
1424 - instr/jedule/* -> simgrid/jedule
1425 - simdag/datatypes.h was removed
1426 - simdag/simdag.h -> simgrid/simdag.h
1427 - msg/datatypes.h was removed
1428 - msg/msg.h -> simgrid/msg.h
1431 * Interface improvement:
1432 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1433 - Sanitize the interface in MSG_task_ module:
1434 - Merge two functions that were close enough but misleading:
1435 set_compute_duration(t) -> set_flops_amount(t)
1436 get_remaining_computation(t) -> get_flops_amount(t)
1437 - set_data_size(t) -> set_bytes_amount(t)
1438 get_data_size(t) -> get_bytes_amount(t)
1439 - Massive cleanups in the functions related to the energy
1440 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1441 - New: MSG_host_get_pstate()
1443 - msg/energy/onoff: switching hosts on and off
1445 * Interface improvement:
1446 - Rename Host.isAvail() to Host.isOn()
1447 - Rename Process.currentProcess() to Process.getCurrentProcess()
1448 - Rename Task.setDataSize() to Task.setBytesAmount()
1449 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1451 - #18874: Actually allows the GC to reclaim tasks
1455 - SIMIX_process_throw: raises an exception in a remote process
1456 * Refactoring: Separate sub-modules
1457 - libsmx: the public interface, as libc in a real system
1458 - popping: the strange dance that converts a user request into a kernel handling
1459 - smx_context_*: the virtualization mechanisms that embed the user code
1460 - smx_*: the handling of each simcalls
1461 * Interface cleanup:
1462 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1463 * Rename smx_action_t into smx_synchro_t, making explicit that these
1464 things are used to synchronize processes with their environment.
1465 For example, a communication is a sort of synchronization involving
1466 the communicating processes (that may block until the exchange) and
1467 the platform. The same can be said from computations, etc.
1469 - #18888: segfault when a process ends before its kill_time
1473 - 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.
1474 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1475 - MPI_Type_set_name, MPI_Type_get_name
1476 - MPI_*_c2f and MPI_*_f2c functions
1477 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1478 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1479 - Activate a lot of new tests from the mpich 3 testsuite
1481 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1482 - InfiniBand network model added: Based on the works of Jerome Vienne
1483 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1484 - When smpi/display_timing is set, also display global simulation time and application times
1485 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1486 * Collective communications
1487 - 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
1488 - 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/).
1489 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1491 - Replay now uses algorithms from wanted collective selector
1492 - Replay can be used with SMP-aware algorithms
1493 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1494 - 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.
1496 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1497 - Correctly use loopback on fat-tree clusters
1498 - Asynchronous small messages shouldn't trigger deadlocks anymore
1499 * Energy/DVFS cleanup and improvement
1500 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1501 - new: smpi_get_host_pstate()
1504 - "Full" network optimization flag was broken since Surf++
1505 - Better handling of precision flags in maxmin
1506 - Fix bug causing sometimes "Impossible" errors
1507 - Properly pass cluster properties to included hosts
1508 * Improvement of the Energy plugin.
1509 - Always update the consumption before returning that value
1510 - New property: watt_off to denote the dissipation when the host is off
1511 - New functions getWattMinAt and getWattMaxAt to retrieve the
1512 dissipation of pstates that we are not currently at.
1513 * Java: class NetworkLink renamed to Link
1514 * New function: simcall_process_get_kill_time()
1515 * Massive rename s/workstation/host/
1516 - That's intrusive, but that's good for the project consistency. Sorry.
1517 - Change config option "workstation/model" into "host/model"
1520 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1521 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1522 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1523 - xbt_dirname and xbt_basename for non-POSIX systems
1525 * The model checker now runs as a separate process.
1526 * The model checker runs must now be launched with the new simgrid-mc program.
1527 * Record/Replay: the MC can display a textual representation of a path in the
1528 execution graph. It can then be replayed outside of the model checker.
1530 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1532 SimGrid (3.11) stable; urgency=low
1537 * Normalizing pointers addresses tool for better diff between logs
1540 * Add cloud examples using new VMs
1541 - examples/msg/cloud/two_tasks_vm.tesh
1542 - examples/msg/cloud/simple_vm.tesh
1543 - examples/java/cloud/cloud.tesh
1544 - examples/java/cloud/migration/migration.tesh
1545 * Add java surf examples:
1546 - examples/java/surfPlugin/surf_plugin.tesh
1547 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1548 - examples/java/surfCpuModel/surf_cpu_model.tesh
1549 * Add SMPI+MSG example:
1550 - examples/smpi/smpi_msg_masterslave/
1555 - msg task destroy cancel test
1556 - msg_host on/off test
1557 * Move all tests in testsuite to teshsuite (adding tesh files)
1558 * Restructure teshsuites
1559 - one folder for each kind of test
1560 * Restructure AddTests.cmake
1562 - structure the order of tests (with sections)
1565 * Add virtual machine
1566 - creation of a VM on a PM
1567 - migration of a VM from a PM to another PM
1569 - MSG_process_join(msg_process_t process, double timeout)
1570 - msg_bar_t MSG_barrier_init(unsigned int count)
1571 - int MSG_barrier_wait(msg_bar_t barrier)
1572 - void MSG_barrier_destroy(msg_bar_t barrier)
1573 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1574 * New option "msg/debug_multiple_use" to help debugging when a task is used
1577 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1579 - Increase code coverage in test suites
1583 * Protect context stacks against stack overflow. The number of protected memory
1584 pages allocated on the top of each stack (1 by default) can be configured
1585 with the new command line option --cfg=contexts/guard_size:<value>.
1586 * Simcalls are now generated by a python script that
1587 - generates files included by SimGrid
1588 - checks that all the functions exist, and proposes prototypes
1590 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1591 host_set_data, host_get_data
1593 - simcall_process_join(smx_process_t process, double timeout)
1594 * Fix bug where sleeping processing could not be suspended.
1597 * Translate surf models from C to C++
1598 - Generic classes for all models: Model, Resource, Action
1599 - A generic interface for each kind of model (CPU, Network, Storage
1600 Workstation, WorkstationVM)
1602 * Translate surf routings from C to C++
1603 * Add callbacks using sigc++ or boost::signals2
1604 - Add callback functions for resource creation/destruction
1605 - Add callback functions for action state change
1606 - Handle Energy as a plugin
1607 * Replace swag by boost::intrusive
1608 * Add new routing models for clusters. For documentation, see
1609 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1610 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1611 parameters for cluster tag
1612 - Fat trees, with topology="FAT_TREE" and
1613 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1614 - see examples/platforms/torus_cluster.xml and
1615 examples/platforms/fat_tree_cluster.xml
1616 * More documentation
1619 * Hostfiles support host:nb_processes construct to deploy several processes on
1621 * Collective communication algorithms should not crash if used with
1622 improper number of nodes and report the error.
1623 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1624 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1625 MPI_Dims_create, MPI_Cart_sub are supported.
1626 * New interface to use SMPI programmatically (still depends on MSG for
1627 some parts, see examples/smpi/smpi_msg_masterslave):
1628 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1632 * Global variables privatization in MPI executables is now performed at runtime
1633 with the option smpi/privatize_global_variables (default:no).
1634 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1635 dynamic libraries loaded by the application are not privatized (static
1636 linking with these libraries is advised in this case)
1639 * Options defined in XML work correctly now.
1642 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1643 copied into simgrid.jar or not (ON by default). Use this option if you want
1644 to reduce the size of the installed simgrid.jar, *and* the native libraries
1645 are kept installed elsewhere.
1646 * Surf binding with SWIG (code generated in maintainer mode only):
1647 - plugin to handle callbacks
1648 - CPU model only for the moment
1651 * Supernovae build mode is definitively removed. It was used to improve
1652 inlining and inter-module optimizations. It is nowadays superseded by
1653 link-time optimizations commonly available in compilers.
1654 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1655 * Add boost dependency for surf++
1656 * Add new macro for tests
1657 - ADD_TESH(name <tesh_args>)
1658 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1662 - xbt_bar_t XBT_barrier_init(unsigned int count)
1663 - int XBT_barrier_wait(xbt_bar_t barrier)
1664 - void XBT_barrier_destroy(xbt_bar_t barrier)
1665 * Make the xbt_os_time module public
1667 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1669 SimGrid (3.10) stable; urgency=low
1671 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1674 * Reintegrate Java to the main archive as desynchronizing these
1675 package is not acceptable anymore (Java is now considered stable)
1676 * Add explicit synchronization facilities through semaphores
1677 * Bug fix: Task.setDataSize() only changed the C world, not the value
1678 cached in the Java world
1681 * Dramatically change the way files are handled. API and internals changed, but
1682 this part of MSG was not considered as production grade either.
1683 * Add explicit synchronization facilities through semaphores
1684 * Add a new function MSG_host_get_process_list()
1685 * Preliminary DVFS support (see examples/msg/energy/ for details)
1688 * SMPI is now included directly in the libsimgrid as the windows
1689 linker doesn't force us on splitting it anymore.
1690 * Improvements of the SMPI replay tool:
1691 - Most of the collective communications are now rooted in the same process as
1692 in the original application.
1693 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1694 used until now). Multiple data types can now be used in a trace.
1695 - The replay tool now supports traces produce either by TAU or a modified
1697 - Bug Fix: the compute part of the reduce action is now taken into account.
1698 - Gatherv collective is now supported
1699 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1700 of smpirun outputs time independent traces for the current run. One file
1701 is created per process. If too many processes are simulated, this behavior
1702 can be changed to one file for all processes by using the
1703 tracing/smpi/format/ti_one_file flag
1704 * smpirun generates the host file if needed (with given host count and platform)
1705 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1706 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1707 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1708 --cfg=smpi/coll_selector:(mpich/ompi)
1709 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1711 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1712 happening inside a collective SMPI call.
1713 * Fix the behavior of complex data types handling.
1714 * Make MPI_Wtime another synchronization point to take computations into
1716 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1717 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1718 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1719 (no privatization of global variables yet)
1720 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1723 * Verification of liveness properties is now available for SMPI applications
1724 (in addition to MSG applications)
1725 * Bugged examples using SMPI in examples/smpi/mc/
1726 * Add --cfg=model-check/visited option. Allows the verification of infinite
1727 programs. Detection of loops in the execution thanks to the system state
1728 comparison and reduction of the state space to explore. Can be combined with
1729 DPOR for safety properties.
1732 * Allow to change SimGrid configuration (see --help) within the code
1733 thanks to SD_config() as it can be done in MSG.
1734 * Add a new function SD_task_set_amount() upon user request.
1737 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1738 * Remove rule based routing (no more PCRE dependency)
1739 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1740 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1741 * Add a loopback_bw and loopback_lat options to cluster tag.
1742 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1743 routing. Now peers are encapsulated in an AS and have their own private
1744 router but this is transparent.
1747 * Our own implementation of getline is renamed xbt_getline, and gets
1748 used even if the OS provide a getline(). This should reduce the
1749 configuration complexity by using the same code on all platforms.
1750 * New type: xbt_cfg_elm_boolean.
1751 * Allow to use yes/no for boolean configuration options in the command line.
1752 * Allow to disable SimGrid cleanups at exit from command line option.
1753 There are situations where one may want a simulation to end with an exit.
1754 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1755 annoying when scripting around the simulator. Adding a
1756 --cfg=clean_atexit:no allows to circumvent this issue.
1759 * Lots of memory leaks were corrected in this release.
1760 There is no known memory leaks anymore, in all of our 600+ tests.
1761 * New command line option --version, to get SimGrid version information.
1762 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1764 * Supernovae builds are deprecated, and expected to be removed in the next
1767 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1769 SimGrid (3.9) stable; urgency=low
1771 The Grasgory release: GRAS is really dead now.
1773 * Complete overhaul of the internal host structures scheme.
1776 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1777 was considered as experimental and badly maintained since then.
1778 * Keeping it was thus a trap to our potential users, that could take
1779 it instead of MSG or SMPI by mistake despite is pity state.
1780 * GRAS seems to have very few users (if any), and no one volunteered
1781 to maintain it further. It also induces a lot of XBT code (for
1782 portability sake), that must be maintained too.
1783 * For all these reasons, we killed GRAS. If someone wants to revive it
1784 in the future, don't cry, our git history still remembers of GRAS.
1787 * Major overhaul. Merge our documentation again as time proved that
1788 splitting it was really not helping our users.
1789 * Further improve the developer documentation to help newcomers
1790 hacking on SimGrid itself. The user documentation (and in
1791 particular, the beginner documentation) is still in a sorry state.
1794 * Now works on Windows too!
1795 * Much more extensive test suite, from MPICH
1798 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1799 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1800 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1801 * Bug fix: let task be scheduled when the last dependency to be solved is
1802 a control dependency.
1803 * Remove SD_load_environment_script function.
1804 Use the C sg_platf function if you want to declare a platform
1808 * New function: MSG_process_get_number()
1809 * Old function documented: MSG_config()
1810 * Remove MSG_load_platform_script function
1811 Use the C sg_platf function if you want to declare a platform
1815 * Change the default value of the TCP_gamma constant (maximal size of TCP
1816 congestion window) to a more realistic 4MiB value. If you notice changes in
1817 your simulation results, you can fall back to the previous 20k tiny window
1818 by adding --cfg=network/TCP_gamma:20000 on command line.
1819 * (Hopefully) fix a bug wrt periodic availability and state traces
1820 * Bug fix: use default values at start when first event in availability/state
1821 trace is not at time 0.
1824 * remove the "new_" part of function name sg_platf_new_trace_connect
1825 (resulting in sg_platf_trace_connect), since it does not create
1829 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1830 use case where it's really mandatory, and maintaining it was a pain in
1832 * New: xbt_fifo_search(), search an item with a user-provided
1833 comparison function instead of dumb pointer comparison.
1836 * Fix the lua deployment:
1837 Use `simgrid.init_application()` before deployment instead of
1838 `simgrid.msg_register_application()` after.
1841 * Transfer the tracing files into the corresponding modules.
1843 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1845 SimGrid (3.8.1) stable; urgency=low
1847 The "we are told that some people want to also *install* the simgrid
1850 * Add missing file "tesh.1" to the archive.
1852 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1854 SimGrid (3.8) stable; urgency=low
1856 The Psssshiiiit release: SimGrid jumps into the Cloud.
1859 * Add an experimental interface to manipulate VMs. They are mainly
1860 process groups with very few intrinsic semantic, but they should
1861 allow you to build the semantic you want easily.
1862 * New function: MSG_host_set_property_value()
1863 * New function: MSG_process_on_exit(). To clean memory in all cases.
1864 * Bug fixes that made the host (and link) failures unusable.
1865 * Add a way to auto-restart process when the host in which they are
1866 executing comes back (ON_FAILURE="RESTART" on deployment file,
1867 MSG_process_auto_restart_set).
1868 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1869 please stop using the old ones, they are DEPRECATED.
1871 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1872 Please use MSG_init() instead. (reducing the amount of entry
1873 points in the library helps us).
1874 * Make it impossible to link against the wrong version of the lib
1875 * Deprecate MSG_clean(). No need to call it anymore.
1876 * Function MSG_get_host_number() is not deprecated anymore.
1879 * Split the doc into a user guide and a reference guide.
1880 * Start a developper guide to help people hacking on SimGrid.
1883 * Enable tracing by default. This modules rocks you should use it.
1884 * Remove option custom_flags. Now use environment variables CFLAGS
1886 * Use default cmake things to detect lua instead of home grown ones.
1887 * New option "enable_mallocators" to disable mallocators, for debugging
1888 purpose ("on" by default).
1891 * Bug fixes around the resource failures: don't let the processes
1892 survive the host they are running onto.
1893 * Add an interface to auto-restart processes when the host in which they are
1894 executing comes back.
1895 * Ensures that SIMIX_clean is called automatically. It's not part of
1896 the public interface anymore (bindings should be updated).
1899 * Bug fix for when SD_Simulate is called with a positive value: be careful
1900 when comparing doubles. Sometimes they are different for non significant
1902 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1903 parallel task whose initial work is distributed among host according
1904 to the Amdahl's law. Such tasks are created with a parameter alpha
1905 that corresponds to the non-parallelizable part of the computation.
1906 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1907 between two sets of workstations assuming a 1D block distribution (each
1908 workstation owns a similar share of data) on both sides.
1910 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1911 Data redistribution will be automatically scheduled once parent and child
1912 are both scheduled. The filling of computation_amount and
1913 communication_amount structures is now done seamlessly thanks to the chosen
1915 * New function SD_workstation_dump to display various information
1916 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1917 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1918 bandwidth on the route onto which the task is scheduled and the amount of
1920 To divide the nominal bandwidth by 2, the rate then has to be:
1921 rate = bandwidth/(2*amount)
1922 * Compute tasks that have failed can now be rescheduled and executed again
1923 (from their beginning)
1924 * Increasing source code coverage (src/simdag is now covered at 95.8%
1928 * Re-implement time-independent trace replay using SMPI (at the
1929 smpi_smp_* level) instead of MSG. This should replace
1930 examples/msg/actions/actions.c
1931 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1932 hindexed and structs)
1933 * Implement the exchange of non-contiguous data.
1934 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1935 * Correct behavior of smpi/sender_gap and set its default value to 0
1936 * Add option to asynchronously send small messages to allow better
1937 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1938 specifies the size in bytes under which messages will be asynchronously sent.
1939 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1940 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1943 * New C interface to define a platform: XML is now optional.
1944 For more info, please check include/simgrid/platf.h.
1945 * New interface to define random platforms from the C:
1946 For more info, please check include/simgrid/platf_generator.h and
1947 examples/msg/masterslave/masterslave_platfgen.c
1948 * Export a sg_cmdline dynar containing all the arguments we got from
1952 * Two new tracing options for adding comments to trace file so you
1953 can track your experiments (see --help-tracing for details).
1954 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
1955 * Adding the SimGrid version that generated the trace file as a comment.
1956 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1957 * Fix to avoid key clashes on Paje links
1958 * Other minor fixes related to the Paje specification
1961 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1962 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1963 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1964 * Greatly improve the robustness of mmalloc to user errors (such as
1965 using an area after freeing it, or freeing it twice)
1967 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1969 SimGrid-java (3.8.1) stable; urgency=low
1971 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1972 Warning: all methods are visible, but only some of them are
1973 implemented so far. Check the source (src/jtrace.c)
1974 for further information.
1975 * New module: org.simgrid.msg.File (SimGrid File management functions)
1976 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1977 * Change the meaning of Process.restart: now restart the process from
1978 the beginning, like MSG_process_restart in C.
1979 * Add Process.setAutoRestart: handling of process restart when failed
1981 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1982 you to retrieve the properties of the processes/hosts
1983 * Deprecate Msg.clean(): you can just forget about it now.
1984 * New function Process.getCount(), that only works when compiling
1985 with the not yet released version 3.9 of the C library.
1987 * New context factory based on Coroutines. It mandates a modified JVM
1988 but then, the simulations run about five times faster, and there is
1989 no limit to the amount of processes (beside of the available memory).
1991 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1993 SimGrid (3.7.1) stable; urgency=low
1996 * Restore the prototype of MSG_process_create_with_environment() to
1997 the pre-3.7 situation by removing the kill_time argument.
1998 * Add a MSG_process_set_kill_time() function instead.
2001 * Fix weird behaviors when dealing with parallel tasks.
2004 * Simgrid is now built as a dll.
2005 * Simgrid-java now works on Windows.
2006 * Simgrid-Java is now included into Windows package.
2009 * First pre-build package for MacOSX.
2012 * Fix compilation when using MSG_USE_DEPRECATED.
2013 * Fix some compilation issues on Macs and Windows.
2014 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2015 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2017 We discovered that the Lua console is broken, but we are missing the
2018 manpower to fix it right now. The problem existed in 3.7 too, so we
2019 are not blocking the release for that. Sorry if you depended on this
2020 feature, any help would be really welcome.
2022 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2025 SimGrid-java (3.7.1) stable; urgency=low
2027 The "Java aint got to be bloated and slow" release
2030 * Various internal cleanups and performance improvement
2031 Simulations are expected to run up to twice faster or so
2032 * Make Process.kill(process) an instance method, not a static one
2033 * User processes are not java.lang.Thread subclasses.
2034 This breaks the compatibility (sorry), but previous API was
2035 brain-dead, making it impossible to have non-trivial
2036 initializations in the process constructor.
2037 * Require a full constructor per Process sub-class.
2038 Kinda breaks the compatibility (sorry), but this allows a much more
2039 efficient way to launch the processes at simulation startup.
2040 * Do not embeed our version of semaphores, java 1.5 can be considered
2041 as sufficiently prevalent for us to not dupplicate its features.
2045 * Add examples for almost every part of the API
2046 We spotted and fixed a lot of bugs in the process
2048 * New module: asynchronous communication API
2049 * New function: Process.sleep()
2050 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2051 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2052 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2054 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2056 SimGrid (3.7) stable; urgency=low
2058 The "spring cleanups (before the next Big Project kicks in)" release.
2061 * We can specify the SMPI latency/bandwidth factor with command line
2062 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2063 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2064 You can also use the "config tag" from platform file by adding this line
2065 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2066 (see "example/platforms/tag_config.xml" to use "config tag").
2067 Note that the command line supersedes the platform file configuration.
2068 * Change the correction factors used in LMM model, according to
2069 the latest experiments described in INRIA RR-7821.
2070 Accuracy should be improved this way.
2071 * Use the partial invalidation optimization by default for the
2072 network too. Should produce the exact same results, only faster.
2073 * Major cleanup in surf to merge models and split some optimization
2074 mechanisms from the core of the models. As a result you can now
2075 specify which model to use (e.g., --cfg=network/model:LV08
2076 --cfg=cpu/model:Cas01) and which optimization mode to use
2077 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2078 Incompatible combinations should err at initialization. See
2079 --help-models for the list of all models and optimization modes.
2080 * The CLM03 workstation model was dropped for simplicity because it
2081 used the deprecated CM02 network model. Use default instead.
2082 * Rename the TCP_gamma configuration option to network/TCP_gamma
2083 * Rename the coordinates configuration option to
2084 network/coordinates, and document it
2085 * Use now crosstraffic keyword instead of the terribly misleading
2086 fullduplex keyword. It is activated by default now in the current
2087 default model, use --cfg=network/crosstraffic:0 to turn it off.
2088 * Ongoing refactoring the model parsing to make XML files optional
2089 See include/simgrid/platf.h for details (still to be completed)
2092 * Major overhaul of the documentation. Almost instructive now :/
2093 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2094 functions and friends. This interface was considered as
2095 deprecated since over 2 years, it's time to inform our users that it is.
2096 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2097 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2098 use these (crappy) functions in your code.
2099 These functions will be removed soon. Stop using them now.
2100 * Deprecate MSG_get_host_{table,number}
2101 Implement MSG_hosts_as_dynar() instead.
2102 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2103 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2106 * Stabilize the parallel execution mode of user contexts
2107 * Introduce configuration variables to control parallel execution:
2108 - contexts/synchro: Synchronization mode to use when running
2109 contexts in parallel (either futex, posix or busy_wait)
2110 - contexts/parallel_threshold: Minimal number of user contexts
2111 that must be part of a scheduling round to switch to parallel
2112 execution mode (raw contexts only)
2113 * Fix bugs that prevented to use suspend/resume along with
2114 synchronization structures.
2115 * Fix bugs in process termination that lead to invalid memory access
2116 in very specific conditions.
2119 * Introduce a parallel mode for the models (controlled by surf/nthreads
2120 configuration item). In our tests, running the models in parallel
2121 never lead to any speedups because they are so fast that the gain
2122 of computing each model in parallel does not amortizes the
2123 synchronization costs, even when ultra fast futexes are used.
2124 This is released anyway because YMMV.
2127 * Performance boost by using a swag internally to compute the set of
2128 tasks that are finished and should constitute the return value of
2132 * Enable it by default now that it is considered rather stable.
2135 * Documentation of the tracing functions.
2136 * Performance gains when tracing categorized/uncategorized resource
2137 utilization by avoiding calls to get route when updating resource
2138 variables. LMM constraints are being used instead.
2139 * API changed to set task categories. Use MSG_task_set_category instead
2140 of TRACE_msg_set_task_category, and SD_task_set_category instead
2141 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2142 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2143 * Examples for link user variables added (see at examples/msg/tracing/)
2144 * Deprecated function TRACE_msg_set_process_category completely removed
2145 * Trace header updated according to the latest Paje file format
2146 * Tracing network lazy updates, no longer obligate users to use full updates
2147 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2148 * Experimental: let user code declare/set/push/pop application states for hosts
2149 * API changed to allow the manual creation of graph configuration files
2150 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2153 * Improve the API of Lua MSG bindings, using the Lua spirit.
2154 * Each simulated process now lives in its own Lua world (globals are
2155 automatically duplicated). It helps writing simulators. Will allow
2156 to run Splay programs within SimGrid in the future.
2157 * Add a Chord example in Lua, equivalent to the MSG one.
2160 * Start the implementation of a solution to express temporal
2161 properties, not only local assertions. This is still an
2162 experimental work in progress, stay clear from it to be safe.
2166 - Add new runtime parameters --help-logs and --help-log-categories
2167 to display information about supported logging parameters and
2169 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2170 don't exists anymore.
2171 * Mallocators: allow value NULL for the reset function.
2173 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2174 create homogeneous dictionaries, where all the elements share the
2175 same free function. Non homogeneous dictionaries will be
2176 deprecated in the next release.
2177 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2178 - Multi-level dictionaries are deprecated.
2180 - new function xbt_dynar_search_or_negative() that is useful when
2181 you have less than 2 million elements in your dynar and don't
2182 want of the extra complexity of catching exceptions when the
2183 element is not found.
2185 - Make xbt_os_thread module (for thread portability) public.
2186 Documentation is still to come, sorry.
2188 - Cleanups and simplifications to make it maintainable again.
2189 - Exotic features (such as memalign and valloc) were removed.
2190 - The metadata were extended and improved so that the
2191 model-checker becomes able to explore and inspect the heaps.
2192 - This may induce a performance drop when enable_model-checking is
2193 ON in cmake (even if it's not used in the simulation), but it is
2194 necessary at this point to get MC working.
2196 Turn model-checking OFF if simulation performance matters to you.
2197 Not enabling it at runtime is not enough, disable it in cmake.
2199 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2202 SimGrid (3.6.2) stable; urgency=low
2204 The "Not coding new stuff allows to polish old things" release.
2207 * New bindings to the NS3 packet level simulator (experimental)
2208 * Use the raw (efficient) execution contexts instead of the sysv
2209 (portable) ones when possible.
2210 * libpcre is now mandatory in any cases since not using it led to
2211 severe performance loss and possibly other issues
2212 * Update the XML platforms:
2213 - G5K: include the latest machine in Nancy
2214 - GridPP and LCG: new platforms
2215 * Documentation was partially updated, at least (more to come)
2217 Bug fixes, cosmetics and small improvements
2218 * Free terminated processes before the end of the simulation to avoid
2219 exhausting the memory of users having very dynamic amount of
2221 * Bug fix and cosmetics about canceling non-running tasks
2222 * Bug fix about the dot loader's issues when using libcgraph
2225 * Create an installer for windows with nsis (amd64 and win32)
2226 - Add an hello world project to illustrate simgrid project creation.
2227 - Embed libpcre into the Simgrid installer to avoid
2228 its compilation burden
2229 * The raw execution contexts should work on Apple now
2230 * Port to Windows 64 bits
2231 - Sysv contexts now have an implementation for this arch
2232 - GRAS communication features now support this arch
2233 * Drop support for borland compiler on windows
2234 - this code was not maintained, and we kinda depend on gcc nowadays
2235 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2236 * Fix portability issue on unstable ubuntu: linker became picky on
2239 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2241 SimGrid (3.6.1) stable; urgency=low
2243 The "Oops, we broke Macs too" release
2246 * Fixed contexts detection so that raw ones are used when possible
2247 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2248 leads to a strange error, with user code segfaulting sometimes when
2249 the generated code is not perfectly aligned (which is not
2250 controllable from the user side, depends on the amount of code)
2253 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2255 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2257 SimGrid-java (3.6) unstable; urgency=low
2260 * Split of every thing from simgrid v3.5 into a separate package.
2262 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2264 SimGrid (3.6) stable; urgency=medium
2266 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2269 * Bindings now constitute their own package, separated from the main one.
2270 Rationale: reduce our maintenance nightmare by reducing the module coupling
2271 They will soon be released on their own on gforge.
2273 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2274 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2276 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2277 * It's not quite deprecated for now because we have no replacement,
2278 but it may soon become the case.
2281 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2282 * Fortran: New user-level cache variable to store the rank of the running
2283 process. This improves performance by an order of magnitude.
2284 * C: New coccinelle script to automatically locate and modify global and
2285 local static variables.
2286 * Improved SMPI network model with a sender-side gap to account for multiple
2290 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2291 only say if a communication is finished, no matter whether it succeeded or
2292 failed. You can call MSG_comm_get_status() to know the status of a finished
2294 * New function MSG_task_dsend() to send a task and detach it. When a
2295 communication is detached, you are never notified of its success or failure
2296 and the memory is released automatically once it is finished. This function
2297 is useful when you don't care about the end nor the success of a
2299 * Change the prototypes of action replay. Sorry for inconvenience,
2300 but this is really more efficient this way (and to adapt your code,
2301 you just have to fix the initialization, that shouldn't be too long)
2302 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2303 ever managed to do anything useful with it.
2304 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2305 takes a dynar of communications. It returns immediately and gives the
2306 index of a finished communication (if any).
2307 * New example: a basic implementation of the Chord P2P algorithm.
2310 * New model for multi-core CPUs. You can now use the core attribute to
2311 precise the number of cores of a host. This is a basic model. Every
2312 process running on the host receives at most the power provided in
2313 the DTD (throughput<=power). Total throughput of process cannot exceed
2315 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2316 router linked by an up-link and a down-link (possibly asymmetrical).
2317 This kind of pattern allows to easily build last-mile model style platforms.
2318 Aggregating such patterns in a rule-based AS is thus the technique of
2319 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2320 * New model for Vivaldi routing. We transformed the Vivaldi network model
2321 into a Vivaldi routing model (based on the rule-based model). This allows to
2322 combine Vivaldi based latencies with last-mile platforms.
2325 * Added a check for NaN of IEEE754 infinite in the double entries of
2327 * Introduce a new context factory "raw", highly inspirated from the
2328 ucontext factory, but using manually crafted functions in assembly to
2329 do the work in an efficient manner.
2330 * Allow to change the used context factory at run time, not only at
2331 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2332 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2333 contexts at runtime (only with raw contexts or ucontexts).
2334 * Completely rewrote this module to allow parallel execution of user
2335 processes. Use --cfg=contexts/nthreads:N to execute user processes
2336 with N parallel threads (the default is 1, meaning no parallelism).
2337 * Allow to decide dynamically between sequential and parallel modes.
2338 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2339 processes in parallel only when their number is greater than or equal to P
2341 * Added a check for NaN of IEEE754 infinite in the double entries of
2345 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2346 won't output the state of processes when interrupted with Ctrl-C
2347 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2348 NULL-terminated array. This may solve backward compatibility issues
2349 due to the change to return type of SD_simulate. See also:
2350 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2351 * Add new macros with variable number of arguments.
2352 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2353 - in xbt/asserts.h: xbt_assert
2354 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2355 - in xbt/ex.h: THROWF and RETHROWF.
2356 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2358 * Change xbt_die() to accept a format string with arguments, just like printf.
2359 * New data structure: xbt_lib_t, like a dict but more general and with better
2363 * New configuration options
2364 Options triva/categorized and triva/uncategorized can be used to generate
2365 graph configuration files for Triva visualization tool.
2366 * Configuration option tracing/platform is renamed to tracing/categorized
2367 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2368 * New TRACE_user_link_variable interface:
2369 User provides the name of the link and the tracing variable to attach to it
2370 * the declaration of tracing categories must be done after the environment
2372 * simpler tracing interface, just one way to declare categories
2373 TRACE_category or TRACE_category_with_color, it is up to you
2374 * links in the trace file are again identified by their names
2375 * trace contains the full platform hierarchy exactly as declared using the ASes
2376 * Options tracing/msg/[task|process]:1 groups the process by hosts
2377 for both cases, tasks and processes must have names that are unique during the simulation
2378 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2379 * The experimental option tracing/msg/volume is deprecated
2380 its functionality may be reincorporated if needed
2382 The tracing generates a trace file with unordered timestamped events,
2383 because of the way the core simulator (surf) works. A script available
2384 at the tools directory (fix-paje-trace.sh) can be used to put the events
2385 in order. We have changed the tracing so it can generate ordered timestamped
2386 events in the final trace, but depending on the simulator (and how much time
2387 is simulated) that can lead to a huge memory utilization. It is deactivated
2388 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2390 Build Infrastructure
2391 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2392 - We are trying hard to keep the API stable, but it may happen that
2393 some things change (we're a research project after all, not a
2394 nuclear plant operating system). If such things should happen, you
2395 could rely on that macro to adapt.
2396 - current value: 30600 for 3.06.00, aka 3.6
2397 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2398 a number that can be compared with SIMGRID_VERSION.
2399 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2400 assertions and verbose/debug logging events are disabled at compile time.
2402 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2404 SimGrid (3.5) stable; urgency=medium
2407 * New feature to any SimGrid-based simulator: Model-Checking
2408 Check SIN#1 for more details.
2411 * New Model SMPI (three-interval linear regression for correction factors)
2412 See RR-7426, available at http://hal.inria.fr/inria-00527150
2413 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2414 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2415 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2416 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2417 * See also src/smpi/README
2422 - Tracing API changes: TRACE_start and TRACE_end should not be called
2423 by user-code. They are automatically called by simulators created
2424 with SimDAG, MSG and SMPI if the toolkit is compiled with
2425 tracing_enabled=ON. Categories declaration and utilization remain the
2426 same for MSG and SimDag.
2427 - A function was added to the tracing API to declare categories with
2429 - TRACE_category_with_color (char *category, char *color)
2430 where color must be in the following format
2431 "%f %f %f", red, green, blue
2432 and red, green, blue are float values in the interval [0, 1]
2433 - User can specify NULL as color parameter, or continue calling
2434 TRACE_category (cat)
2435 On that case, the tracing system will define random colors
2436 - The following command-line options are supported:
2437 --cfg=tracing/filename:msg.trace
2438 --cfg=tracing:1 (activate tracing, needed to use others)
2439 --cfg=tracing/platform:1 (categorized resource use)
2440 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2441 --cfg=tracing/msg/task:1 (task creation)
2442 --cfg=tracing/msg/process:1 (process creation, migration)
2443 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2444 --cfg=tracing/smpi:1 (SMPI interface tracing)
2445 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2446 - examples of examples/msg/tracing updated
2448 - DAXLoader and DOTLoader functions can generate tasks with categories
2449 - A new function to attribute a category to SD tasks:
2450 TRACE_sd_set_task_category (SD_task_t task, char *category)
2451 * Tracing the MPI interface implemented by SMPI
2452 - Collective operations are traced with states
2453 - Point-to-Point operations are traced with states/links
2454 - Tracing activated by a parameter "-trace filename" passed
2455 to smpirun during execution (considering that simgrid
2456 is compiled with tracing enabled)
2457 - To run the simulation with gdb, the simulator
2458 accepts --cfg=tracing/smpi:1 to trace SMPI
2459 - tesh files to check if smpi tracing is ok
2460 - See examples/smpi/NAS/DT-trace
2461 * GTNetS tracing re-worked
2462 - adaptation to the tracing system of GTNets to cope
2463 with modifications regarding the fullduplex mode
2464 - new tesh files to check if gtnets tracing is ok
2467 * Asynchronous communications through the functions:
2468 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2469 * New function: MSG_load_platform_script()
2470 to make possible using a lua script instead of XML files to set up platforms
2471 * New function: MSG_set_function
2472 to associate functions to processes, used when bypassing the parser
2473 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2475 Platforms: Add some more examples in examples/platforms
2476 * Grid'5000: see www.grid5000.fr
2477 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2480 * Change the XML format. This is a very important modification. SimGrid
2481 3.5 introduces a new hierarchical format based on the notion of
2482 Autonomous Systems. Compatibility with old format is ensured
2483 through the perl script provided in the install bin directory
2484 bin/simgrid_update_xml.
2485 It is now possible to build platforms with specific routing
2486 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2487 connect several platforms together. We will try to provide soon
2488 set of realistic platforms exploiting these properties (have a
2489 look at examples/platforms/ for the moment).
2490 * Take the opportunity of the XML format change to be a good XML citizen:
2491 rename link:ctn to link_ctn and similar changes (also dealt with by
2493 * Add a new routing scheme (rule-based) using regular expressions. It
2494 enables to have an extremely low memory footprint when the
2495 underlying routing is simple and can be compactly described. You
2496 need to have libpcre4-dev (perl regular expressions) installed if
2497 you want to use this routing scheme.
2498 * Revive the cluster TAG and allow to easily and efficiently (both in
2499 term of memory and speed) connect clusters together. Have a look
2500 at teshsuite/simdag/platforms/ to see how this can be done. With
2501 this tag, you can create clusters with thousands of tasks at no
2502 cost (have a look at examples/platforms/).
2503 Note: clusters are implemented as ASes, so there is no need for an
2504 enclosing AS tag if you have only one cluster in your platform.
2505 * Add new generic functions in the public interface that allows the user
2506 to call SURF 'create_resource' methods from your code (same
2507 functionality as the XML bypass mechanism but with a much lighter
2509 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2510 that uses a piecewise linear approximation to produce better
2511 results when exchanging small messages.
2512 * Add a new parameter to handle correctly full duplex link and account
2513 for interferences between uplink and downlink communications
2514 (activate with --cfg=fullduplex:1).
2517 * Rename the SD_READY (all dependencies are satisfied and task is
2518 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2519 dependencies are satisfied) state.
2520 This prevents a confusion between the notion of "ready to schedule"
2521 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2522 simulated" (SD_RUNNABLE) used by the simulation kernel.
2523 * Change the way a task is considered as ready. Instead of removing
2524 dependencies when a task is done, a counter is decreased. This way,
2525 it is always possible to reach ancestors thanks to the
2526 SD_taks_get_parents function (even after the end of the simulation.)
2527 * Change the return type of SD_Simulate from (SD_task_t*) into
2528 xbt_dynar_t. This function was in handling a dynar internally and
2529 converted it into a NULL terminated array for historical reasons.
2530 * New function SD_dotload(char*) to load a DAG described in dot
2531 format. This loader and the corresponding examples require the
2532 installation of the graphviz library.
2533 * Fix a bug in the management of tasks of size 0 in the surf network
2534 models. This problem was only visible with SIMDAG and you should
2535 thus disregard results produced with earlier versions if you
2536 relied on this feature (some tasks were blocked because of this).
2537 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2538 even though your applications were doing only point-to-point
2539 communications and sequential computations. Now you can really use any
2540 model you want (of course, if you create real parallel tasks, which are
2541 not implemented in most models beside ptaskL07, this will abort).
2542 * Add an example that schedules a DAX on an heterogeneous platform
2543 using a Min-Min strategy.
2544 * New function SD_workstation_get_current_task() that returns the kind
2545 of task currently running on a workstation in the sequential access
2547 * Raise some warnings when unexecuted tasks remains at the end of the
2548 simulation. This is usually caused by cycles in the DAG.
2551 * New function: SIMIX_process_set_function() called by MSG_set_function
2552 * Change the underlying waiting queue in semaphores so that a process
2553 can wait on several of them simultaneously (as in waitany).
2554 * Fix the way to handle tokens in semaphores so that all access patterns
2555 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2556 * kill the dirty pimple SIMIX_message_sizes_output()
2557 Please use (proper) visualization instead
2560 * New data container: setset (set of sets of elements)
2561 * New module: mmalloc (mapped malloc, allowing to have several
2562 independent segments of malloc)
2563 * New function: xbt_dict_cursor_set_data()
2564 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2565 * New function: xbt_dynar_is_empty()
2566 * New function: xbt_fifo_get_last_item()
2567 * Fix xbt_dynar_shrink(): use the right element size.
2568 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2569 uninitialized areas during expand.
2570 * Fix semaphores: previous implementation was severely broken.
2571 * Use library init/fini functions for our initialization.
2572 - you can use logs and other feature as soon as you want in your
2573 code (even before the xbt_init / MSG_init)
2574 - xbt_exit is now a no-op and produce a warning when used.
2577 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2578 benefit from the latest improvement to the simulation kernel.
2579 * Kill measurement sockets for now. If you rely on them, sorry. This
2580 release is not for you. This feature will be reintroduced in the
2581 future, but we cannot delay the release any further.
2582 * New function: gras_msgtype_get_name().
2583 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2584 * Fix (at last) the pmm example: it should not randomly fail anymore.
2586 Build chain: bug fixes and overall polishing
2587 * Cmake is now stable enough. Hence, we killed the autotools.
2588 * Port to windows ( TM :)
2589 * Fix the 'make install' target.
2590 No need to use 'make install-simgrid' anymore
2591 * Introduce a 'make dist' target compiling a *source* archive
2592 'make package' compiles a binary archive
2593 * Compile java files only on need
2594 * Add --cd and --setenv command line options to tesh
2595 * Out of source builds are not fully supported yet, but we are close
2596 * Enable supernovae and optimization flags by default for our users
2599 * Add layer to set up environment directly from lua, without XML.
2600 * The effect of gras_stub_generator can be achieved through
2601 lua too (check examples/gras/console/ping_generator.lua)
2603 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2605 SimGrid (3.4.1) stable; urgency=low
2607 The "Polishing easter eggs is probably a good idea" release.
2608 This is a bug fixes release only.
2612 * Fix a bug preventing the tasks from begin garbage collected.
2615 * Fix a bug occurring when a host involved in a communication fails.
2616 This was not detected properly by the other peer involved in the
2617 communication. Now, it's reported as a network error.
2620 * Warn the user about loop dependencies in data flow of DAX files
2621 * Obey the control-flow dependencies of DAX files
2624 * Add option "enable_smpi" allowing to not compile SMPI.
2625 Probably useful for the (Mac) users experiencing a build error here
2626 * Improve the detection of lua5.1 and ruby1.8
2628 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2630 SimGrid (3.4) stable; urgency=low
2632 The "Easter in Cargese" release. Also known as (major changes):
2634 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2635 ~> bindings were greatly improved
2636 ~> new tracing infrastructure for better visualization introduced
2638 * the "Welcome to configury modernity" release.
2639 ~> we switched from autotools to cmake, and improved our cdash
2642 A more detailled list of changes follow (full detail in svn log).
2644 Java Bindings: Various Cleanups
2645 * (install java-gcj-compat-dev on debian-like to use them)
2646 * Remove put/get: no need to export deprecated interface in Java
2647 Use send/receive instead.
2648 * Cleanup the examples and add a README per directory
2649 * Remove example autoDestination (that's the only way to go now)
2650 * Remove example explicitDestination (was a plain copy of basic)
2651 * Make JniException a runtime exception, so that there is no need to
2652 declare the fact that you may encounter such a beast. I guess that
2653 nobody will ever want to survive such error.
2654 * Create specific errors for each MSG case of failure:
2655 host failure, transfer failure, timeout, task cancelled
2656 * Cleanup the exceptions that may get thrown by each function
2657 * Other internal cleanups in Java bindings. Performance still bad :/
2658 Ruby and Lua Bindings: create them
2659 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2660 * That's new and great, you should try them out.
2661 Same functionalities than Java bindings, only even less polished
2663 * Kill the useless "rate" argument of SD_task_get_execution_time()
2664 Everyone used to provide -1 as a value, it was not used, and the
2665 semantic of a possible use wasn't even clear.
2666 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2667 either as comm costs or compute costs to mean that there is no
2668 such thing for that specific task.
2669 * Add a SD_task_set_name() function
2670 * Fix SD_task_unschedule() on typed tasks
2671 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2672 * In DAX loader, accept useless 'level' attributes to <job> since
2673 LIGO DAGs have them (seem to be to ease graphical representation).
2675 * Add an example masterslave_mailbox.c using send/receive and not
2676 the deprecated put/get interface.
2677 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2678 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2680 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2681 (declare MSG_USE_DEPRECATED to still have the old name)
2682 * Add a MSG_task_set_data() function
2683 * About trace replay (see examples/msg/actions):
2685 - Allow to work with splitted trace files for each process
2686 Give the specific trace file as argument of each process,
2687 and call MSG_action_trace_run(NULL)
2688 You can still have one merged file for all processes.
2689 - Fix implementation of collective operations
2690 * Allow task_execute() on 0-sized tasks (closes #10063)
2692 * This is the first release of SimGrid where SMPI is not considered
2693 beta anymore (even if some corners should still be improved)
2694 * Port over the new SIMIX_network submodule (internal refactoring)
2695 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2696 * Implement more missing elements of the standard:
2698 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2700 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2701 MPI_Scan() MPI_Get_processor_name()
2702 - Added implementation of missing case for Alltoall (warning: it's
2703 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2704 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2705 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2706 SMPI_MPI_Allgatherv()
2707 * Bug fixes include:
2708 - MPI_Waitsome() was broken
2709 - Allow relative includes in smpicc
2710 - Command line cfg argument 'reference_speed' was ignored...
2711 - Some functions did not properly lead to auto-benching of user code
2712 - smpicc passes -O2 by default (just like openmpi one)
2714 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2715 * Bug fixes about timeouts during communications
2716 * add SIMIX_message_sizes_output() as a pimple to write to file the
2717 amount of messages per size. Use gnuplot to get histogram.
2718 Pimple because that's the only user-visible function of simix,
2719 defined directly in xbt.h (irk, sorry)
2721 - Add a SIMIX_sem_get_capacity() function
2722 - Fix interactions with processe resume/suspende
2723 - release_forever() was stupidly broken
2724 - Fix SIMIX_display_process_status() for processes in a semaphore
2725 - Make SIMIX_sem_block_onto() user-visible
2726 * Refactoring context stuff:
2727 - Use pseudo-OOP for better modularity
2728 - reimplement SIMIX_process_kill() without process_schedule() so
2729 that the latter can take as invariant that it is called from
2731 - Merge context_start into context_new for sake of simplicity
2733 * Add a Vivaldi network model, coded live during SUD'10 ;)
2734 * Rename configuration variables to start a hierarchy:
2735 o cpu_model -> cpu/model
2736 o network_model -> network/model
2737 o workstation_model -> workstation/model
2738 * New configuration variables:
2739 o network/bandwidth_factor: correction to bandwidth
2740 o network/latency_factor: correction to latency
2741 o netwotk/weight_S: correction to the weight of competing streams
2742 * Add a long description to the models, that users can see with such
2743 argument on the command line: --cfg=cpu/model:help
2744 * --help-models display the long description of all known models
2746 * config: add the ability to set a default value after registration
2747 Does not override any previously set value (e.g. from cmd line)
2748 * dict: allow to have integer key and data.
2749 When so, you need to use the following functions
2750 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2751 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2752 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2753 In contrary to regular dicts, the key is not malloced before copy.
2754 Mixing scalar and regular elements in the same dict is not tested
2756 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2757 Tracing for Visualization:
2758 * SimGrid is now instrumented in order to generate a trace file for
2759 visualization analysis: to use it, need to compile SimGrid with the
2760 "tracing" option enabled, and instrument the program using SimGrid with
2761 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2762 (among other functions).
2763 * The instrumentation only traces the platform utilization for now
2764 * Documentation to use the tracing functions and how to analyze the
2765 traces with the Triva tool is written.
2766 * More information about: SimGrid FAQ (in the section Tracing Simulations
2769 * We moved to cmake as default build system. Autotools support will
2770 be dropped soon. Check the FAQ for more info about how to use it.
2771 * Greatly improved our cdash/ctest interactions
2772 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2773 * Added memory checking tests with valgrind; lot of memleak fixing.
2774 This may be the first release of simgrid with so few memory issues
2775 * Added code coverage tests.
2776 Our coverage is still improvable, but at least we see it on cdash.
2778 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2780 SimGrid (3.3.4) stable; urgency=low
2782 The "Desktop Grid needs love too" release (also called Xmas release).
2784 Models improvements:
2785 * Major speedup in the maxmin system solving by using lazy evaluation
2786 Instead of solving completely the maxmin system at each iteration,
2787 only invalidate (and recompute) the modified parts.
2788 This new feature is enabled in default models but you can try to
2789 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2790 * Cas01 IMproved as default CPU model
2791 This CPU model is the same Cas01 model, but it uses the
2792 maxmin-selective-update flag and a heap structure to manage
2793 actions on SURF kernel.
2794 It reduces the complexity to find the next action to finish and,
2795 consequently, it's faster than the old Cas01.
2796 This is the new default CPU model (Cas01).
2797 * Rename the old Cas01 model to Cas01_fullupdate
2798 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2799 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2800 * CpuTI (CPU Trace Integration)
2801 A new CPU model whose objective is simulate faster when using
2802 availability trace files.
2803 Instead of using a full featured, over engineered maxmin system for
2804 CPU modeling, this model does the pre-integration of traces files
2805 to calculate the amount of CPU power available, and so, executes
2806 faster than the old CPU models.
2807 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2808 * Use LV08 as default network model since it gives better accuracy
2809 for small messages and shouldn't change things for big ones.
2810 Use --cfg=network_model:CM02 to get the previous behavior.
2813 ******************************************
2814 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2815 ******************************************
2816 * The new CPU model may changes simulations!
2817 The point is that events occurring at the exact same timestamp
2818 are not scheduled in the same order with the old and new
2819 version. This may be enough to completely change the execution
2820 of simulations in some cases.
2821 * The new network model will change simulations!
2822 This new model is more realistic than the previous one, so you
2823 should consider redoing your old experiments with this model.
2824 Sorry for the inconvenience.
2827 * Introduce the supernovae compilation mode
2828 When compiled that way, the whole SimGrid (or almost) is put in a
2829 single compilation unit and compiled in one shoot.
2830 This is to help gcc which has difficulties to inline stuff from one
2832 The speedup seem to be above 15%, althrough more tests are needed on
2833 amd64 to confirm that gain.
2836 * Port of MSG's mailbox on top of SIMIX network
2837 The put/get mechanism was greatly simplified on the way.
2840 * New SIMIX network module. Provides:
2841 - Mailbox: rendez-vous mechanism to find with who you want to speak
2842 - Synchronous send/recv: easier and hopefully faster since the
2843 logic is handled in the maestro process directly now
2844 - Asynchronous send/recv: you dreamt of it? It's here now
2845 Too bad that nobody cared enough to propagate the change to MSG.
2846 * Add semaphores as SIMIX synchronization mechanism.
2849 * new function SD_daxload(char*) to load a DAX file
2850 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2851 * Introduce typed tasks. Specify its kind and cost at creation.
2852 At scheduling, just give where it should be placed, and the cost
2853 for each involved resource is automatically computed.
2854 Existing constructors so far (more to come of course):
2855 - SD_task_create_comm_e2e() for end-to-end communication
2856 - SD_task_create_comp_seq() for sequential computation
2857 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2858 * new function SD_task_dump() for debugging display
2859 * new function SD_task_dotty(task,FILE*) writing to file the info
2860 about the task in dotty format
2861 * SD_task_dependency_exists() can now cope with having one of its
2862 arguments NULL. If so, it tests whether the other argument has any
2864 * Add getters on list of preceding/following tasks:
2865 SD_task_get_parents(task) and SD_task_get_children(task)
2866 * Add getters on amount of workstations and list:
2867 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2868 * Add getter on task kind: SD_task_get_kind(task)
2869 * Update the start_time and finish_time of tasks on completion/failure
2870 * Bugfix: Remove task from state swags when destroyed
2873 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2876 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2877 dynar with one separate thread per value of the dynar.
2878 * Change the prototype of xbt_thread_create(), sorry.
2879 Added a boolean parameter indicating whether we want to join this
2880 thread (used in SG only for now)
2881 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2884 * GTNetS wrappers should now be usable again (and betterly tested too)
2885 * Fix a major regression from 3.2 where the timeout provided to
2886 MSG_task_put_with_timeout() was used as absolute time before which
2887 the comm should be done.
2888 * Start to fix the <cluster> tag.
2889 - Internal links should be good now (beside of the loopback, which
2890 use the private link instead)
2891 - paths to the external world is still rather broken
2892 - the <route:multi> tag is just broken. Actually that's brain-dead.
2893 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2894 to make it less stupid
2895 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2896 * Fix a source-level compatibility glitch from 3.2: after defining
2897 MSG_USE_DEPRECATED, you can use the old name
2898 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2899 * Allow to compile from the SVN with automake 1.11
2900 * Fix some problems when using the "start_time" tag in deployment XMLs.
2901 * Fix #8569: XBT/synchro.h has redundant declarations
2902 * Fix #8563: MSG return values and exceptions
2903 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2904 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2905 Thanks to Robson Peixoto.
2906 * Fix a vicious bug in dictionaries inducing that some elements were
2907 not freed on xbt_dict_free()
2909 Portability report of this version:
2910 * Main portability targets:
2911 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2912 on (amd64/i386/ia64)
2913 - mac leopard on i386
2914 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2915 but nothing critical.
2916 * Other platforms: windows, AIX and others were not tested for this release
2918 Timing report of this version:
2919 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2920 scenario parameters). From 8h to a few seconds in desktop grid settings.
2921 * Supernovae brings about 25% speedup on i386.
2923 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2925 SimGrid (3.3.3) stable; urgency=low
2927 The "Need for Speed" release.
2929 The timings done to validate the 3.3.2 were faulty.
2930 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2932 The problem was a conversion from a manually handled vector to
2933 xbt_dynar_t on the critical path.
2934 xbt_dynar_foreach calls functions, inducing stack management crap.
2936 We inlined these functions and xbt_dynar_foreach is now breath taking.
2937 We also inlined xbt_swag_belong on the way.
2939 Here are some approximate speedup measurements (on master/slaves
2940 simulations lasting between 10s and 20s each):
2941 3.3.1 -> 3.3.2: about same performance
2942 3.3.2 -> 3.3.3: 40% speedup
2943 3.3.1 -> 3.3.3: 40% speedup
2944 3.3.1 with inline patch -> 3.3.3: 30% speedup
2946 Our reading is that the refactoring which occurred in 3.3.2 made us
2947 suffer much more from the xbt_dynar_foreach low performance, but
2948 once we solved this, this refactoring proved to be very performance
2949 effective. From the 40% speedup, somehow, 10% are due to the
2950 inlining and 30% to the refactoring.
2952 That's a pitty that gcc cannot inline functions placed in other files
2953 alone. We have to choose between:
2954 - break the encapsulation (by putting private data structures and
2955 accessors in headers files to help gcc)
2956 - live with low performance
2957 - switch to a decent compiler such as icc (not quite possible).
2959 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2961 SimGrid (3.3.2) stable; urgency=low
2963 The "Simplicity does not preceed complexity, but follows it" release.
2965 The main contributors of this release were (lexical order):
2966 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2969 * Extract the routing logic into its own object.
2970 (was dupplicated in network.c and workstation_LV07.c;
2971 Allows to implement other ways of storing that info)
2972 => kill now useless network_card concept
2973 - Use dynar to represent routes (instead of void** + int*)
2974 - kill link_set (use surf_network_model->resource_set instead)
2975 - Add a command-line option to choose the routing schema to use
2976 - Add three new models:
2977 * Floyd (shortest path computed at initialization)
2978 * Dijikstra (shortest path recomputed all the time)
2979 * Cached Dijikstra (shortest path computed on need)
2980 All these models where contributed by Silas De Munck, and are
2981 described in his ICCS09 paper.
2983 * Simplify model declaration
2984 (less redirections, less function to write when defining a model)
2985 - Factorize stuff between models:
2988 surf_model_resource_set(model)
2989 surf_model_resource_by_name(model, name)
2990 - Unify the types of models in s_surf_model_t (using an union)
2991 - Embeed fields of common_public directly into s_surf_model_t
2992 - Rename model methods:
2993 action_free ~> action_unref
2994 action_change_state ~> action_state_set
2995 action_get_state ~> action_state_get
2996 - Change model methods into functions:
2997 (model)->common_public->action_use ~> surf_action_ref
2999 * Implement a generic resource; use it as ancestor to specific ones
3000 (allows to kill duplicated code in models)
3001 Drawback: timer command don't need no name nor properties;
3002 workstation_CLM03 don't need no properties
3003 (but I guess we can live with those few bytes wasted)
3005 * Improve the action object model
3006 - implement a constructor avoiding dupplicated code about field
3007 initialization in generic_action part.
3009 * Kill the SDP model: it has an external dependency, is deprecated
3010 in flavor of modern lmm models, and didn't compile since a while
3013 * Relocation of the context module from XBT to SIMIX.
3014 (the context were decoupled from the simix processes, duplicating a lot of code)
3015 => a lot of code was factorized
3016 - less overhead is introduced during scheduling
3017 - simpler API for the context factory
3018 - the logic for process creation,destruction and manipulation was simplified
3019 * Simplification of the s_smx_process_t data structure.
3020 => accesing the simix level data associated to a process is faster now,
3021 and the code is a lot more readable.
3024 * Implement some more MPI primitives:
3025 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3026 -implementation: Bcast: flat or 2-ary tree (default),
3027 Barrier: 4-ary tree,
3029 Allreduce: Reduce then Bcast
3030 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3031 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3032 Alltoallv: flat tree, like ompi
3034 * Add support for optimized collectives (Bcast is now binomial by default)
3035 * Port smpirun and smpicc to OS X
3038 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3039 Shout out if you used it.
3042 * Display the list of still queued messages in SG mode when existing
3046 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3047 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3048 * Add xbt_str_from_file(FILE*)
3049 * Add xbt_dict_get_key achieving a linear reverse search
3050 * Remove the context module
3052 Portability report of this version:
3053 * Main portability targets:
3054 - Linux(debian)/x86/context
3055 - Linux(debian)/x86/pthreads
3056 - Linux(debian)/amd64/context
3057 - Linux(debian)/amd64/pthreads
3058 On these, we still have the eratic breakages of gras/pmm and
3059 amok/saturate_sg reported in previous version. We still think
3060 that the tests are the cause of the fault, not the tested code.
3062 - Mac OSX Leopard/x86/context
3063 Still false negative in tesh autotesting.
3064 Smpi still fails, but this time because readlink does not accept -f
3065 Everything seems to work properly beside of that.
3068 - AIX version 5.3 (only tested contexts this time)
3069 Smpi still fails there because mktemp is not installed.
3070 Everything seems to work properly beside of that.
3072 I managed to compile it for the first time, but several breakages.
3073 Won't delay the release for this exotic platform.
3075 * Windows: it's still lagging behind. If you want to help, please
3078 Timing report of this version:
3079 This version seem to be more than 5% faster than 3.3.1 (on linux
3080 64bits with contextes). The gain is less than expected, we are
3081 investigating this for next release.
3083 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3085 SimGrid (3.3.1) stable; urgency=low
3088 * Implement a --cfg-help to show existing configuration variables
3089 * Build chain do not require doxygen in maintainer mode
3092 * fix a bug on struct sizeof computation, which prevented the
3093 exchange of arrays of structs in some conditions
3094 - added a regression test about this in datadesc_usage
3095 * Allow the exchange of 0-long dynamic vectors.
3096 - for that, use -1 as indicator of dynamic size instead of 0
3097 - This implied to change any size from unsigned long to long,
3098 reducing a bit communication abilities, but I guess that with
3099 64bits being quite common, this is more than enough.
3100 - This also induce a protocol change, thus bumping network protocol
3101 version from 0 to 1 (if we have external users, we have to get
3102 clean on that point too ;)
3103 - added two regression tests about this in datadesc_usage
3104 * Be more verbose when propagating local exceptions
3105 This helps debugging.
3106 * Display the status of simulated processes when receiving SIGINT in
3110 * Allow to control the simulation from a trace file.
3111 New functions MSG_action_register() and MSG_action_trace_run()
3112 The first one allows to associate a function execution to each
3113 kind of action while the second one parses a trace file and
3114 triggers the corresponding actions within the system.
3115 For now, only a toy example is provided in examples/msg/actions
3116 * Add an example of process migration in examples/msg/migration
3117 * Fix a bug in task exchange which broke MSG_task_get_sender()
3118 Add a teshsuite regression test for that.
3119 [Bug: if MSG_task_get_sender() is called after sender exit,
3121 * Fix a bug which prevented suspend/resume to work properly
3122 * Display the status of simulated processes when receiving SIGINT
3123 This fixes a regression of v3.3. due to the introduction of SIMIX
3124 * Bug fixing in failure management:
3125 - trace could not start by a failure at time 0
3126 - failure during communications were not working
3129 * Add SIMIX_process_set_name() to change the name of the current
3130 process in the log messages.
3131 * Store smx_hosts in a dict since we only retrieve them by name
3132 * Move the configuration infrastructure to surf
3135 * Move the configuration infrastructure to surf
3138 * Massive internal cleanups:
3139 - Store internal structures on processes instead of hosts (allows
3140 to have more than one process per host, in addition of being more
3142 - Cleanup the initialization/finalization process
3143 - Kill a whole bunch of unneeded synchronization:
3144 processes run in exclusive manner within the simulator
3145 - Move queues from global tables to process data fields
3147 - now accept -platform and -hostfile arguments
3148 - Pass the right rank value to processes according to the hostfile
3149 * Compile the examples by default, and use them as regression tests
3150 * Implement MPI_Wtime()
3151 * Change the reference speed to a command line option
3154 * TCP_gamma can now be specified as command line option using
3155 --cfg=TCP_gamma:10000000.0
3156 * Change the --surf-path cmd line option into --cfg=path:
3159 * Also include strbuff from xbt.h public header
3160 * xbt_ex_display(): do not free the exception after displaying
3161 This allows to do more with the given exception afterward.
3162 Users should call xbt_ex_free() themselves.
3166 Portability report of this version:
3167 * Main portability targets:
3168 - Linux(debian)/x86/context
3169 - Linux(debian)/x86/pthreads
3170 - Linux(debian)/amd64/context
3171 - Linux(debian)/amd64/pthreads
3172 These targets fail about 1/10 of times on gras/pmm, but we believe
3173 that this is because of the test, not because of simgrid.
3174 amok/saturate_sg fails even more rarely, and the test may not be
3177 - Mac OSX Leopard/x86/context
3178 The test suite still spits tons of errors because some obscure
3179 force prevents us from removing the temporary directories
3180 arguing that they still contain some metadata I've never heard of.
3181 Smpi fails because seq is not installed.
3182 Everything seems to work properly beside of that.
3185 - AIX version 5.3 (both contexts and pthread)
3186 Smpi still fails there because mktemp is not installed.
3187 XML inclusions seems rosty on AIX.
3189 * Windows: it's still lagging behind. If you want to help, please
3192 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3194 SimGrid (3.3) stable; urgency=high
3198 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3199 [Malek Cherier & Mt]
3201 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3204 Until now, GRAS were using MSG as an interface to SURF. It was
3205 quite difficult because both interface have several differences
3206 (MSG channels vs GRAS sockets were the most notable point).
3208 This also opens the gate to SMPI (which should occur soon) and speed
3209 up simulations by to 40% (even if it were not the main goal).
3211 **************************************
3212 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3213 **************************************
3214 The point is that events occurring at the exact same timestamp are
3215 not scheduled in the same order with the old and new version. This
3216 may be enough to completely change the execution of simulations in
3217 some cases. Sorry for the inconvenience.
3219 * Cleanup and upgrade the XML format to push further scalability
3220 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3222 * Improve the testing infrastructure with tesh. Now a very large part of
3223 the code is tested not only by being run but also by checking that the
3224 output match an expected output [Mt].
3226 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3227 is really less memory-demanding, which should allow you to use
3228 larger files in SimGrid [AL].
3230 * Inform valgrind about our contextes, so that it becomes usable
3231 with the default (and more effecient) version of SimGrid
3232 [contributed by Sékou Diakite, many thanks]
3235 * Introduce a listener thread in charge of receiving incoming
3236 messages from the network. It allows to overlap communication and
3237 computation but most notably, it removes some stupid deadlocks due
3238 to the fact that so far, a process could not send and receive at
3239 the same time. This made most non trivial communication schema
3241 * Convert the PIDs from long int to int to match the MSG ones (and
3242 linux ones too) [Mt]
3243 * New function: gras_agent_spawn() to launch a new process on
3244 current host. Only working in simulation for now. [Mt]
3245 * New function: gras_os_hostport() returning a constant form (ie,
3246 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3249 * Make the backtrace of exceptions more human readable [Mt]
3250 * New module: xbt/str [Mt]
3251 a ton of string utility functions (split, join, printf to a newly
3252 allocated buffer, trim, etc)
3253 * New module: xbt/hash [Mt]
3254 SHA1 hashing algorithm (more to come if needed)
3255 * New module: xbt/synchro [Mt]
3256 synchronization tools (mutex and conditions) working the same way
3257 in simulation and in real life (mainly useful for GRAS, but not
3259 * New module: xbt/queue [Mt]
3260 classical producer/consumer synchronization scheme
3261 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3262 (using the classical functions will get serialized) [Mt]
3263 * Make dictionary internal table dynamic. No need to specify its size
3264 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3266 * Make sure the log channels are organized as a tree under windows
3267 (because of ANSI C compatibility issue, any channel were child of
3268 root directly) [Mt].
3271 * Cleaned many thing in surf and fixed a few bugs [AL].
3272 * Add a nice command line configuration mechanism to compose models [AL].
3273 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3274 the previous one (KCCFLN05). It relies on something that looks like
3275 a max-min sharing mechanism but cannot be written as such. A new solver
3276 was thus designed [AL].
3277 * Add a new solver to lmm. Based on Lagrange optimization and
3278 gradient-based descent, it enables to efficiently maximise systems s.a
3280 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3282 This solver enables to propose two new network models for TCP Reno and
3283 TCP Vegas based on Low's work. These models still need to be fully
3284 tested though [Pedro Velho].
3287 * Bug fix in SD_simulate. Now the time bound given as argument is
3289 * Use the new parallel task model (ptask_L07) as default.
3290 * Use the SURF command line configuration mechanism.
3291 * 0-size tasks (for synchronization) should now work.
3293 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3295 SimGrid (3.2) stable; urgency=high
3299 We still experience issues on this platform, but we believe that at
3300 least MSG is usable.
3302 GRAS API BREAKAGE (for simplification purpose, sorry):
3303 * the gras_msgtype_by_name is not used anymore. Instead of
3304 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3305 you can write (and must)
3306 gras_msg_send(toserver, "request", &request);
3307 - If you still want to pass a gras_msgtype_t to the function (to cache
3308 the type and avoid the lookup time), use the gras_msg_send_() variant.
3309 - Impacted functions:
3310 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3311 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3312 * The callbacks are now expected to return 0 when everything went well
3313 (just like the main() function)
3315 GRAS new features and improvements:
3316 * New module mechanism where user code can use per process globals [Mt]
3317 This is similar to gras_userdata_*() functions, but for libraries. It
3318 factorize some code developped over and over in the examples and AMOK.
3319 It has still to be documented and used (only amok/peermanagement is
3321 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3322 loss when they were small enough to fit into the buffer and sent quickly
3323 enough so that they can all get received in one shoot.
3324 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3325 if not found. Use the *_or_null() variant for the old semantic.
3326 * In gras_msg_handle, do not discard messages without callback.
3327 They are probably messages to be explicitly awaited later (ie, proofs of
3328 mis-synchronization in userland since they are sent before being awaited)
3329 No big deal usually.
3330 * gras_socket_meas_send/recv: semantic changed!
3331 The numerical arguments used to be (1) the total amount of data to send
3332 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3333 messages. This was need for the fool willing to send more than MAXINT
3334 bytes on quite fat pipes.
3337 * Do really rename the hostmanagement module to peermanagement. [Mt]
3338 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3339 but this is rather new and this was documented in the module
3340 documentation (poor excuses, I admit)
3341 * Bandwidth measurement semantic changed! This follows the changes to
3342 gras_socket_meas_send/recv explained above.
3345 * A sequential mode has been added to the workstations. When a workstation
3346 is in sequential mode, it can execute only one task, and the other tasks
3347 are waiting in a FIFO. [Christophe Thiery]
3350 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3351 model for SIMDAG. [Christophe Thiery]
3352 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3353 equal to 0, which caused some bad behaviors in
3354 saturated_constraint_set_update. I now use a threshold mechanism like in
3358 * When running manually src/testall, you select specific units [Mt]
3359 testall is the result of our cunit mechanism, and should replace all
3360 the scripty thingy around since bash don't run easily on billware.
3362 * A mallocator system has been added. [Christophe Thiery]
3363 Mallocators allow you to recycle your unused objects instead of freeing them
3364 and allocating new ones.
3366 Documentation update:
3367 * FAQ reworking + New FAQs:
3368 - "Valgrind spits tons of errors!" [Mt]
3369 - "How to repport bugs" [Mt]
3370 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3371 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3372 - Communication time measurement within MSG [AL]
3373 - I experience weird communication times when I change the latency [AL]
3374 * GRAS tutorial [Mt]
3376 - an introduction to the framework and to the used communication model
3377 - an initiatic tour introducing the most proheminent features:
3379 . Lesson 0: Installing GRAS
3380 . Lesson 1: Setting up your own project
3381 o Part 2: Message passing
3382 . Lesson 2: Exchanging simple messages
3383 . Lesson 3: Passing arguments to the processes (in SG)
3384 . Lesson 4: Attaching callbacks to messages
3385 . Lesson 5: Using globals in processes
3386 . Lesson 6: Logging information properly
3387 . Lesson 7: Using internal timers
3388 . Lesson 8: Handling errors through exceptions
3389 . Lesson 9: Exchanging simple data
3390 . Lesson 10: Remote Procedure Calling (RPC)
3391 . Lesson 11: Explicitly waiting for messages
3392 . Recapping of message passing features in GRAS
3393 - A HOWTO section containing:
3394 o HOWTO design a GRAS application
3395 More are due, of course. They will come latter. In the meanwhile, you can
3396 check the examples which are still here.
3398 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3400 SimGrid (3.1) stable; urgency=high
3404 There was a stack corruption somewhere, visible only when optimizing
3405 with these versions. [Vince]
3408 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3409 module that enabled to play with Directed Acyclic Graphs. It is built
3410 directly on top of SURF and provides an API rather close to the old
3411 SG. Some old codes using SG are currently under rewrite to check that
3412 all needful functions are provided. [Christophe Thiery]
3415 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3416 extension of the classical CLM03 model that gracefully handles
3417 failures. This is now the default model for MSG and GRAS. It doesn't
3418 handle parallel tasks yet though. [AL]
3419 * Bug fix: Weights were not correctly set in the network part.
3420 WARNING: This may have resulted in incorrect results with simulations
3421 where there are more than one flow on a given link. [AL]
3424 * After a (long ?) discussion on simgrid-devel, we have decided that the
3425 convention we had on units was stupid. That is why it has been decided
3426 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3427 WARNING: This means that all previous platform files will not work as
3428 such with this version! A warning is issued to ask users to update
3430 A conversion script can be found in the contrib module of the CVS, under
3431 the name contrib/platform_generation/surfxml_update.pl [MQ]
3434 * Bug fix: Processes were started in reverse order, wrt deployment file.
3435 WARNING: if your code relies on this bug, please fix it. [AL]
3436 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3437 being executed on two different locations. [AL]
3438 * Bug fix: Failures are now better supported thanks to Derrick's tests
3439 (there was many failure situations I hadn't thought of and that weren't
3440 correctly handled). [AL]
3441 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3445 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3446 Exception occurring on server-side are propagated back to client (!).
3448 API CHANGE: the callback changed their prototype. Change:
3449 int my_handler(gras_socket_t expeditor, void *payload_data) {
3451 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3452 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3454 * New! function: gras_msg_handleall to deal with all messages arriving
3455 within a given period.
3456 * New! function: gras_socket_server_range to get a server socket in a
3457 range of port numbers (ease to avoid port number conflicts) [MQ]
3458 * New! gras processes display their backtrace when they get a SIGUSR1
3459 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3460 Sweet to debug RL processes [MQ]
3464 - Do not force experiment sizes to be expressed in kb, or it becomes
3465 impossible to measure the latency this way (needs one byte-long tests)
3466 WARNING: this changes the amok_bw_* function semantic. [MQ]
3467 - Implements the link saturation stuff. [MQ]
3468 * Peer management module:
3469 New! module factorizing code that we wrote over and over [MQ].
3472 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3473 - Test units are placed directly into the library code, they get extracted
3474 automatically and placed into the src/testall binary.
3475 - Convert most of the XBT tests to this system.
3476 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3477 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3478 live in the same C code. This required to change a little bit the API
3479 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3480 * New module: sparse graph structure with basic algorithms (this is work
3481 in progress and the API is not considered to be frozen yet). [AL]
3482 * Display more information on backtraces: source line & function names are
3483 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3484 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3485 * Reworked a little bit some #include statements to load only required
3486 headers. Some user code that relied on SimGrid to include stdlib or
3487 stdio may need to include it by themselves. [AL]
3488 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3490 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3491 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3492 but I guess that almost nobody used those parts. [MQ]
3494 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3496 SimGrid (3.0.1) stable; urgency=low
3499 * Unfortunately, I had missed 5 misnamed functions:
3500 xbt_fifo_item_t xbt_fifo_newitem(void);
3501 void xbt_fifo_freeitem(xbt_fifo_item_t);
3502 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3503 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3504 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3505 They're now deprecated. Please use their new versions:
3506 xbt_fifo_item_t xbt_fifo_new_item(void);
3507 void xbt_fifo_free_item(xbt_fifo_item_t);
3508 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3509 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3510 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3512 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3513 * Documentation: xbt_log module unmercifully reworked [MQ]
3514 * Bugfix: there was a problem with the ending of contexts with
3515 the pthread backend. It caused some weird deadlock or behavior
3516 depending on the pthread implementation. [AL]
3517 * Bugfix: get the exceptions raised in the simulator repport where
3518 and why they come from when they are not caught in time [AL, MQ]
3521 * Bugfix: Do repport the error when two non-connected hosts try to
3522 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3525 * Add additionnal checkings on communications. Assert that two
3526 communicating hosts are connected by a set of links... [AL]
3529 * Add additionnal checkings on channel values in communication [AL]
3530 * New: MSG_task_get_source to see on which host a task was generated [HC]
3531 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3532 the number of tasks waiting to be received on channel and sent
3534 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3535 waits for the first task coming from a given host.. [AL]
3537 GRAS new functionnalities: [MQ]
3538 * Enhance the parsing macro to allow the size of multidimentional objects
3539 to be given thru annotations.
3540 * New example (and documentation): Matrix Multiplication a la RPC
3541 (as when I was young!) and fix a bunch of bugs found on the way.
3543 GRAS performance improvements: [MQ]
3545 * Reduce the amount of cbps creation/destruction by making it static to
3546 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3549 * Change libdata to a set so that we can search for stuff by ID (and thus
3550 reduce the insane amount of dict lookups)
3553 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3554 RL communication on top of SG since it's so uneffective.
3555 It may also be used for inter-thread communication in RL, one day.
3556 * Use gras_datadesc_copy() to exchange messages on top of SG
3558 - improve message exchange performance on top of SG
3559 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3560 * Don't exchange on the network the size of the used part of buffer,
3561 instead, specify the possible buffer size to read().
3563 - reduces the amount of read/write calls (one pair per exchange)
3564 - reduces the amount of exchanged data (the size)
3565 - allows to retrieve all arrived data on receiver side, if we don't need
3566 it right now (subsequent read will peek the buffer)
3567 - allows the receiver to proceed with the beginning of the stream before
3568 everything is arrived
3569 - make it possible to build an iov transport (using readv/writev)
3571 - take care of the data with non-stable storage (like stacked data),
3573 * If possible, TCP send uses vector I/O (when writev() is here)
3574 - Don't use it for receive since we send data sizes and data on the
3575 same stream, so we wouldn't be able to chain large amount of chunks
3576 before having to flush the stuff to read the size.
3577 * Rework the transport plugin mechanism to simplify it and reduce the
3578 amount of pointer dereferencement when searching for the right function
3581 * I guess that now, we do almost as few system calls as possible while
3582 doing as few data copy as possible.
3584 To improve it further, we could try to send all the sizes first and then
3585 all the data (to use iov on receiving size), but it's only a partial
3586 solution: when you have 2 dimensional data, the sizes of the second
3587 dimension is data of the first dimension, so you need 3 streams.
3589 I'm not sure the potential performance gains justify the coding burden.
3591 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3593 SimGrid (3.00) stable; urgency=high
3596 * New! Give the possibility to hijack the surf parser and thus bypass
3597 MSG_create_environment and MSG_launch_application. Have a look at
3598 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3600 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3602 SimGrid (2.96) unstable; urgency=low
3607 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3608 This deprecates the xbt_error_t mechanisms.
3609 It modifies (simplifies) all XBT and GRAS API.
3610 MSG API keeps unchanged (exceptions raised by XBT are caught from
3611 within MSG and masked with existing error handling facilities)
3614 * New! Add a FATPIPE model. [AL]
3615 * New! Add a parallel task model. [AL]
3616 * New! Add automatically a loopback interface (in the default
3617 network model) if none was precised.
3620 * Bugfix: MSG_process_resume now works with the current running process.
3622 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3623 * Modification of MSG_task_get_compute_duration. Once a task has been
3624 processed, the value returned by this function is now equal to 0. [AL]
3625 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3626 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3627 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3628 of a m_task. For now, MSG_task_cancel only works with computation
3630 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3631 of the processor (in Mflop/s) regardless of the current load on the
3633 * API Change: use proper naming convention for MSG_getClock and
3634 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3636 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3637 This function changes the priority of a computation task. This priority
3638 doesn't affect the transfer rate. A priority of 2 will make a task
3639 receive two times more cpu power than the other ones. This function
3640 has been added to suit the needs of Nguyen The Loc and hasn't been that
3641 much tested yet. So if it fails, please report it and send me your code.
3643 * API Change: removed all functions and types that were marked "deprecated"
3644 since many months. Renamed MSG_global_init_args to MSG_global_init.
3646 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3648 SimGrid (2.95) unstable; urgency=low
3651 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3652 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3653 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3656 * Bugfix: Allow absolute paths for platform description files [MQ]
3657 * Bugfix: do free the variables after use. Leads to drastic performance
3659 * Implement max_duration (ie, timeouts) on resources [AL]
3662 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3664 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3665 the needed support (timeouts) [AL]
3668 * Implement measurement sockets. You can now get the bandwidth between two
3669 hosts thanks to AMOK (see below). [MQ]
3670 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3671 dynar over the network (yeah) [MQ]
3672 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3673 * simple/static token-ring example. [Alexandre Colucci and MQ]
3674 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3675 in place before [MQ]
3678 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3679 type parsing. "short" and "long" modifiers were also ignored; other
3680 modifier (such as reference level) are still ignored. [MQ]
3681 * Embeed the buffer size within the buffer itself on SG. [MQ]
3682 That way, send() are atomic and cannot get intermixed anymore (at least
3683 the ones which are less than 100k; bigger messages still have the issue)
3684 * Array size pushed by the field, not by the field type (or each
3685 and every long int will push stuff to the cbps) [MQ]
3686 * use select() to sleep since it allows to portably sleep less than one
3689 GRAS (minor cleanups)
3690 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3693 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3694 * type category 'ignored' killed as it was never used and were difficult
3696 * whether a type can cycle or not is now a flag, leaving room for more
3697 flags (as "ignored", if we feel the need one day ;)
3698 * Rename raw sockets to measurement sockets since "raw" has another
3699 meaning in networking community.
3702 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3703 and offers features not belonging to GRAS but that most applications
3704 need. One day, it may be a set of plugins loadable at runtime.
3705 * New module: bandwidth
3706 bandwidth measurement between arbitrary nodes running this module. [MQ]
3708 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3710 SimGrid (2.94) unstable; urgency=low
3712 The first beta release of SimGrid 3 !
3716 * Update the main page and the FAQ. Adding references to gforge.
3719 * Add a gras_os_getpid function.
3722 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3723 * Extend the logs so that they also print PID, hostname, date, ... if
3725 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3726 and kill the old version which were in testsuite/
3727 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3731 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3732 for users. Instead, each programming environment (GRAS, MSG,...) use it
3733 when needed to provide such a feature to users.
3734 Don't shortcut the mechanism or you will also shortcut the virtualization
3735 you need on the simulator.
3739 * Cleanups in configury with regard to compile optimization/warning flags.
3740 Also add -fno-loop-optimize to any powerpc since it's the optimization
3741 killing gcc (< 3.4.0).
3742 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3743 needed by MSG examples complications
3744 * Borrow configury beautifications from PHP
3747 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3748 warning (thanks loris for stumbling into this one).
3749 * Bugfix: stop loading private headers (gras_config.h) from the public
3753 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3754 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3755 RL would imply a DNS resolver, which is *hard* to do in a portable way
3756 (and therefore delayed).
3757 * Implement a real timer mechanism and use it in timing macros. This allows
3758 to avoid rounding errors and get a 0.000005 sec precision in timing
3759 macros. While I was at it, various cleanups:
3760 - allow to declare more than one timed section per file (fix a stupid bug)
3761 - move some private declaration to the right place
3762 - merge conditional execution and timing macros into emulation module
3763 - document the module
3764 - make sure the module cleanups its mess on gras_exit
3765 * Documentation improvements:
3766 - (new) how to compile applications using GRAS
3767 - (new) emulation support (timing macros)
3769 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3771 SimGrid (2.93) unstable; urgency=low
3773 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3776 - Use Paje properly where used. Still to be sanitized properly.
3777 - Portability fix: Add an implementation of the contexts using pthread
3781 - Add xbt_procname(): returns the name of the current process.
3782 Use it to show the current process's name in all logging.
3784 - fix detection of older flex version and the reaction, since we do
3785 depend on modern ones (we use lex_destroy)
3786 - Better separation of SG and RL in the libs: remove all simulation code
3787 from libgras. As a result, this lib is now only 200k when stripped.
3788 Some of the xbt modules may also be duplicated (two sets and such) and
3789 should be cleaned/killed before SG3.
3790 - Insist on using xlC on AIX because of weird problems involving gcc there.
3791 - Cleanup the make remote stuff. This is now done by scripts
3792 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3793 mainly for our private use, but we're working on changing them to user
3796 - Bugfix: flush the socket on close only if there is some *output*.
3797 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3799 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3802 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3804 SimGrid (2.92) unstable; urgency=low
3806 Alpha 3 on the path to SimGrid 3
3810 - New! First try of benchmarking macros.
3811 - New! First try so that gras_stub_generator generate deployment and
3812 remote compilation helpers.
3814 - Bugfix: Initialization fix in msg_test.
3818 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3820 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3822 - API Change: gras_os_sleep to take the amount of seconds as a double.
3823 Accepting an int was error prone since it was the only location where
3824 seconds were coded as such. It leaded to damn rounding errors.
3825 - Bugfix: Hard to belive that timers ever worked before this.
3827 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3829 SimGrid (2.91) unstable; urgency=low
3831 Alpha 2 on the path to SimGrid 3
3835 - Bug fix in the lmm_solver.
3837 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3838 through the function MSG_paje_output.
3839 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3840 - It is possible to bound the rate of a communication in MSG with
3841 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3842 to write it in the changelog).
3843 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3848 - Various cleanups to the autotools stuff
3849 - Begin to move Gras examples to examples/gras/
3850 - Let make distcheck work again (yeah!)
3852 - documentation overhauled using doxygen.
3853 gtk-doc-tools is dead in SimGrid now.
3854 - Automatically extract all existing logging categories, and add the list
3855 to the documentation (long standing one, to say the less)
3857 - Cleanup the known architecture table. Reorder the entries to group what
3858 should be, and use a more consistent naming scheme.
3859 (some of the test dataset are still to be regenerated)
3860 - New! Allow library to register globals on each process just as userdata
3862 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3863 lookup time (for now).
3864 Use it in msg and trp.
3865 This cleans a lot the internals and helps enforcing privacy of the
3866 headers between the gras components.
3867 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3868 - Bugfix: gras_os_time was delirious in RL.
3869 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3871 - Reenable GRAS now that it works.
3873 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3875 SimGrid (2.90) unstable; urgency=low
3877 Alpha 1 on the path to SimGrid 3
3879 * It is a long time since the last release of SimGrid. I'm sorry about
3880 that but as I had told you, I was rewriting a lot of things. I apologize
3881 to those who had been reporting bugs to me and that I had not answered.
3882 If your bug is still in the new version, please tell me. Here is a
3883 summary of the main changes.
3885 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3886 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3887 portability, speed, and a lot more but you'll figure it out later.
3888 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3889 I wanted to release the new MSG as soon as possible and I have
3890 broken GRAS, which is the reason why, for now, only MSG is fully
3891 functional. A laconic description of these projects is available
3892 in the documentation.
3894 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3895 kernel optimized for our needs (called SURF but only the developers
3896 should use it). Hence, MSG is now roughly 30 times faster and I think
3897 that by rewriting a little bit MSG, I could event speed it up a little
3898 bit more. Beside the gain in speed, it is also much easier to encode a
3899 new platform model with SURF than it was with SG. More to come...
3901 * REVOLUTION 3: I have tried to change a little as possible the API of
3902 MSG but a few things really had to disappear. The main differences
3903 with the previous version are:
3904 1) no more m_links_t and the corresponding functions. Platforms are
3905 directly read from an XML description and cannot be hard-coded
3906 anymore. The same format is used for application deployment
3907 description. The new format is described in the documentation.
3908 Have a look in tools/platform_generation. There is a tiny script
3909 that converts from the old platform format to the new one. Concerning
3910 the application deployment format, parsing the old one is tricky.
3911 I think most of you should however be able to convert your files. If
3912 it is really an issue, I can write a C code that does the conversion.
3914 2) the toolbox tbx does not exist anymore. We now have a library
3915 with much more data-structures but without the hash-tables (we have
3916 dictionaries that are much faster).
3918 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3920 *****************************************************************************
3921 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3922 * don't feel like converting the oldies. (Mt) *
3923 *****************************************************************************
3926 Version 2.90: "the long awaited one"
3927 - Finished rewriting and debugging MSG. Rewrote the documentation.
3928 - disable GRAS for now since it needs to be ported to the newest SG
3931 - Finish the port to windows (using mingw32 for cross-compile)
3934 - Main loop and datastructures of SURF. A cpu resource object is
3935 functional. Surf can thus be used to create cpu's with variable
3936 performance on which you can execute some actions.
3938 2004-11-15 Martin Quinson
3939 - Port to ARM. Simply added the alignment and size descriptions. Should
3940 work, but the ARM machines are so slow that I didn't had the opportunity
3941 to 'make check' over there yet.
3943 2004-11-15 Arnaud Legrand
3944 - Trace manager now written. It uses a heap structure and is therefore
3945 expected to be efficient. It may however be speeded up (particularly
3946 when many events occur at the same date) by using red and black
3947 trees. One day maybe...
3948 - Max-min linear system solver written. It uses a sparse matrix
3949 structure taking advantage of its expected use. Most operations are
3950 O(1) and free/calloc are called as few as possible. The computation of
3951 the minimum could however be improved by using a red and black tree
3954 2004-11-03 Arnaud Legrand
3955 - Rename every gras_* function that was in xbt/ to its xbt_
3957 - Add a heap and a doubly-linked list to xbt
3958 - Added a dichotomy to the dictionaries. make check works as well before
3959 so I assume that the patch is correct. I do not know however if things
3960 run effectively faster than before now. :)
3962 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3963 SimGrid, and the version number is bumped to 2.x
3965 2004-10-29 Martin Quinson
3966 - Introduction of the remote errors.
3967 They are the result of a RMI/RPC on the remote machine.
3968 ErrCodes being scalar values, you can't get the host on which those
3969 errors did happen. Extending the error mechanism as in Gnome is possible.
3970 No idea yet whether it is a good idea.
3972 2004-10-28 Martin Quinson
3973 - Interface revolution: the Starred Structure Eradication.
3974 I used to do typedef struct {} toto_t; and then handle *toto_t.
3975 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3976 - ???_t is a valid type (builded with typedef)
3977 - s_toto_t is a structure (access to fields with .)
3978 - s_toto is a structure needing 'struct' keyword to be used
3979 - e_toto_t is an enum
3980 - toto_t is an 'object' (struct*)
3982 typedef struct s_toto {} s_toto_t, *toto_t;
3983 typedef enum {} e_toto_t;
3984 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3985 s_toto_t) is private.
3987 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3988 so that it can set the variable to NULL. It was so for dicts and sets,
3989 it changed for dynars.
3991 - Fix a bunch of memleaks in dict_remove
3992 - Fix a bug in sg/server_socket opening: it failed all the time.
3994 2004-10-07 Martin Quinson
3995 - Speed up dynar lookup operation a bit.
3997 gras_dynar_get is dead.
3999 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4000 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4001 (which gives you the address of the stored data).
4003 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4004 copy of the data using an affectation to do the job and not a memcpy.
4006 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4007 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4009 It does not really speedup the dynar test because they are
4010 setting elements all the time (and look them seldom). But the dict does
4011 far more lookup than setting.
4013 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4015 2004-10-05 Martin Quinson
4016 - Allow to (en/dis)able the cycle detection at run time.
4018 Whether we should check for cycle or not is now a property of each
4019 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4020 datadesc_cycle_unset allow to remove this property when previously set.
4022 Note that the cycle detection is off by default since it impacts the
4023 performance. Watch the data you feed GRAS with ;)
4025 This property is hereditary. Any element embedded in a structure having it
4026 set have it set for the time of this data exchange.
4028 You should set it both on sender and receiver side. If you don't set it on
4029 sender side, it will enter an endless loop. If you forget on receiver
4030 side, the cycles won't be recreated after communication.
4032 - Header reorganization.
4033 Kill gras_private.h, each submodule must load the headers it needs.
4035 2004-10-04 Martin Quinson
4036 - Interface revolution: do not try to survive to malloc failure.
4038 Now, gras_malloc and friends call gras_abort() on failure.
4039 As a conclusion, malloc_error is not a valid error anymore, and all
4040 functions for which it was the only gras_error_t return value are
4041 changed. They now return void, or there result directly.
4042 This simplify the API a lot.
4044 2004-09-29 Martin Quinson
4045 - Re-enable raw sockets.
4046 Created by gras_socket_{client,server}_ext;
4047 Used with gras_raw_{send,recv}
4050 It should allow to kill the last bits of gras first version soon.
4052 This is not completely satisfactory yet (duplicate code with
4053 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4056 - Simplify transport plugin (internal) interface by not passing any
4057 argument to _server and _client, but embedding them in the socket
4060 2004-09-28 Martin Quinson
4061 - Finish the port to AIX.
4062 autoconf was my problem (segfault within the malloc replacement
4063 function. No idea why)
4065 2004-09-16 Martin Quinson
4066 - Fix some size_t madness on 64bit architectures.
4068 2004-09-08 Martin Quinson
4069 - Reduce the number of system headers loaded, overload some more system
4070 calls (such as malloc to cast the result of the system one, and work
4072 - Fix and reintroduce the config support
4074 2004-09-07 Martin Quinson
4075 - Source code reorganization to allow Arnaud to surf all over there.
4076 - Allow to document the logging categories.
4077 - Remove all uppercase from logging categories and useless cleanup in names.
4079 2004-08-18 Martin Quinson
4080 Version 0.6.2 (protocol not changed; API changed)
4081 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4082 gras_error_t), and NULL when not found. Functions expecting a msgtype
4083 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4084 hopefully usefull message.
4085 - Portability to prehistoric sparcs again
4087 2004-08-17 Martin Quinson
4088 Version 0.6.1 (protocol not changed; ABI not changed)
4089 - prealloc some buffers to speed things up
4091 2004-08-11 Martin Quinson
4092 Version 0.6 (protocol not changed; ABI expended)
4093 - The parsing macro can deal with the references, provided that you add
4094 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4096 2004-08-09 Martin Quinson
4097 Version 0.5 (protocol not changed; ABI changed)
4098 - Allow to off turn the cycle detection code in data exchange at
4099 compilation time. It should be at run time, but I'm short of time (and
4100 the config stuff is still broken). That way, we keep dict out of the
4101 critical path, which is good because the performance is poor:
4102 - search not dichotomial yet
4103 - dynar give no way to access their content and memcpy everytime
4104 - In composed data description (struct, ref and so on), stop foolness of
4105 keeping the subtype's ID, but store the type itself. This keeps sets out
4106 of the critical path, which is good since they rely on dynar and
4107 dictionaries. The only loose of that is that we cannot detect the
4108 redeclaration of a structure/union with another content (but I'm not sure
4109 the code detected well this error before anyway). We still can detect
4110 the redefinition discrepancy for the other types.
4111 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4112 it breaks the code because of type-punning used all over the place).
4113 This breaks on all non-gcc architectures (for now).
4115 All those changes (plus the buffer of last time) allow me to gain 2 order
4116 of magnitude on cruel tests consisting of 800000 array of integers on two
4117 level of a hierarchical structure (200 secondes -> 4 secondes)
4120 - the selector of reference must now return the type it points to, not
4121 the ID of this type.
4123 2004-08-06 Martin Quinson
4124 Version 0.4 (protocol changed; ABI not changed)
4125 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4126 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4127 - Implement a buffer transport, and use it by default (it relies on tcp in
4128 real life and on sg in simulation).
4129 That's a bit hackish since I had a new field to the structure to store
4130 its data without interfering with the subtype ones. Inheritance
4131 is tricky in C. And that's a kind of reverse inheritance with one class
4132 derivating two classes. Or maybe a game with java interfaces. Anyway,
4133 that's damn hard in C (at least).
4134 Moreover, I got tired while trying to ensure plugin separation and
4135 genericity in SG mode. MSG wants me to do weird things, so let's go for
4136 cruel hacks (temporarily of course ;).
4137 See comment in transport_private.h:71
4138 - do not include all the _interface headers in private but in the files
4139 which really need them (to cut the compilation time when they are
4142 2004-07-26 Martin Quinson
4143 Version 0.3 (protocol not changed; ABI changed)
4144 - Major overhault of the datadesc interface to simplify it:
4145 - shorted the function names:
4146 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4147 - add a trivial way to push/pop integers into the cbps without malloc.
4148 This allows to make really generic sub_type description, which simply
4149 pop their size of the stack.
4150 - add a function gras_datadesc_ref_pop_arr() which does what users want
4151 most of the time: Declare a dynamic array (which pops its size of the
4152 stack) and declare a reference to it. Poor name, but anyway.
4153 - kill the post-send callback, add a post-receive one
4155 2004-07-23 Martin Quinson
4156 Version 0.2 (protocol changed; ABI changed)
4157 - add some testing for cpbs in the test cases, and fix some more bugs.
4158 This invalidate again the little64 data file, since I cannot regenerate
4160 - remove an awfull optimization in the logging stuff, allowing me to:
4161 - understand it again
4162 - learn gcc how to check that the argument match the provided format
4163 - fix all errors revealed by gcc after that
4164 - internal keys of dict are not \0 terminated. Deal with it properly in
4165 loggings instead of segfaulting when the user want to see the logs :-/
4167 2004-07-22 Martin Quinson
4168 - Fix some stupid bug preventing cbps (callback postit) from working
4170 2004-07-21 Martin Quinson
4171 - Some documentation cleanups
4172 - remove the useless last argument of msgtype_declare
4173 - rename the Virtu functions to fit into the 'os' namespace
4174 - move headers src/include -> src/include/gras/ and stop fooling with
4175 gras -> . symbolic link
4176 - make distcheck is now successful
4178 2004-07-19 Martin Quinson
4180 - Build shared library also
4181 - Install html doc to the right location
4182 - stop removing maintainer files in make clean
4183 - build tests only on make check
4185 2004-07-13 Martin Quinson
4187 - No major issue in previous version => change versioning schema
4188 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4189 corresponding dataset.
4191 2004-07-11 Martin Quinson
4193 - Get it working with any kind of structure (we can compute the padding
4194 bytes remotely for all the architectures I have access to)
4195 - Implement the structure parsing macro (still not quite robust/complete)
4196 - Improvement to the remote testing toysuite
4198 2004-07-10 Martin Quinson
4199 [autoconf mechanism]
4200 - get ride of a bunch of deprecated macros
4201 - actually run the test for two-compliment, not only compile it :-/
4202 - test whether the structures get packed (and bail out if yes. Damn.
4203 Alignment is a serious matter)
4204 - test whether the structures get compacted (but respecting the alignment
4205 constraints of each types)
4206 - test whether the array fields of structures can straddle alignment boundaries
4208 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4209 (took me 2 hours to find that bug, looking at the wrong place)
4210 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4211 before sending/receiving and are used to compute the offsets of fields
4212 - Given that padding size depend even on compiler options, keep track of
4213 alignment and aligned_size only for the current architecture. Not a big
4214 deal since we send structure fields one after the other (seems
4216 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4217 program, let it work on linux/gcc/little32. portability todo.
4219 2004-07-08 Martin Quinson
4220 - import and improve remote compilation support from FAST
4221 - make sure make check works on half a dozen of machines out there
4223 2004-07-07 Martin Quinson
4224 Let's say it's version 0.0.3 ;)
4225 - Implement conversions (yuhu!)
4226 - Let it work on solaris (beside conversion, of course)
4227 - Stupid me, using rand() to generate the conversion datatests in not wise.
4229 2004-07-06 Martin Quinson
4230 - Let make dist work, since I'm gonna need it to compile on remote hosts
4231 - Let Tests/datadesc_usage write the architecture on which the file was
4232 generated as first byte.
4233 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4234 architecture descriptions.
4235 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4236 on those architectures.
4237 - Optimization: send/recv array of scalar in one shoot
4239 2004-07-05 Martin Quinson
4240 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4242 - Plug a whole bunch of memleaks
4243 - each process now have to call gras_{init,exit}. One day, their log
4244 settings will be separated
4245 - Continue the code factorisation between SG, RL and common in Transport.
4247 2004-07-04 Martin Quinson
4249 - Redistribution between SG and RL.
4250 We wanna have to accept in SG, so move accepted related parts of RL in
4251 the common part. (more precisely, the dynar of all known sockets is no
4252 more a static in transport.c, but part of the process_data)
4254 [gras_stub_generator]
4255 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4257 2004-07-03 Martin Quinson
4258 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4259 them all up in one shot)
4261 - Fix the ugly case of reference to dynamic array.
4262 - New (semi-public) function gras_datadesc_size to allow the messaging
4263 layer to malloc the needed space for the buffer.
4265 - gras_socket_close now expect the socket to close (and not its address to
4266 put NULL in it after it). This is because the socket passed to handlers
4267 is one of their argument (=> not writable).
4269 - propagate the interface cleanup from last week in datadesc, ie remove a
4270 superfluous level of indirection. User pass adress of variable
4271 containing data (both when sending and receiving), and not of a variable
4272 being a pointer to the data. Let's say that I like it better ;)
4273 The price for that is constructs like "int msg=*(int*)payload" in
4274 handlers, but it's a fine price, IMHO.
4276 - Let it work in RL (yuhu)
4278 2004-06-21 Martin Quinson
4280 - porting SG plugin and SG select to new standards (works almost).
4281 - plug memleaks and fix bugs around.
4284 - cleanup the prototype of data recv and force users to specify when they
4285 want to handle references to objects. Test case working even for cycles.
4286 - plug memleaks. Valgrind is perfectly ok with this.
4288 2004-06-12 Martin Quinson
4290 - cleanup the separation between plugin and main code in plugin creation
4292 2004-06-11 Martin Quinson
4294 - Reput hook for raw sockets, needed for BW experiments
4295 - kill a few lines of dead code
4296 [Data description] Interface cleanup
4297 - gras_datadesc_by_name returns the searched type or NULL.
4298 That way, no variable is needed to use a type desc once, which makes
4300 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4301 two parameters were strings (field name, type name), leading to
4303 [Dicos] Interface cleanup
4304 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4305 This is consistant with the dynar API.
4307 2004-04-21 Martin Quinson
4309 - Porting to new standards.
4311 - interface cleanup.
4312 There is no bag anymore, no need to take extra provision to mask the
4313 pointers behind "ID".
4314 Better splitup of functions between files create/exchange/convert.
4315 This is still a bit artificial since convert and receive are so
4316 interleaved, but anyway.
4318 - add a queued message list to procdata (the ones not matching criteria
4320 - factorize some more code between SG and RL wrt procdata
4322 - use gras_exit in example to track memleaks
4323 - get rid of gs_example now that GS is properly integrated into gras
4324 - update run_test to integrate the latest tests (datadesc)
4326 - rename WARNINGn macros to WARNn since it prooved error-prone
4328 2004-04-19 Martin Quinson
4330 - register init/exit functions within gras module mechanism
4331 - send/receive function.
4332 Convertion is not implemented, but short-cutted if not needed.
4333 struct/array elements are sent one by one (instead of block-wise), but
4334 nobody really cares (yet). Get a prototype before optimizing.
4335 - tests (using a file socket) for DD send/receive on:
4336 - base types: int, float
4337 - array: fixed size, string (ie ref to dynamic string)
4338 - structure: homogeneous, heterogeneous
4339 - chained list, graph with cycle
4340 Believe it or not, valgrind is not too unhappy with the results. The
4341 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4345 - Bugfix when using a filename explicitly (instead of '-')
4347 2004-04-09 Martin Quinson
4349 - factorize more code between RL and SG in socket creation
4350 - Complete the implementation and tests of:
4352 o file (only in RL, and mainly for debugging)
4354 I lost 3 days to design a portable address resolver, and then decided
4355 that the prototype mainly have to run on my box.
4356 Addressing portability too early may be like optimizing too early :-/
4358 - use gras_init in the Tests instead of the crappy parse_log_opt
4359 (the latter function is removed)
4360 [Conditional execution]
4361 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4362 [Code reorganisation]
4363 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4364 Build examples against the RL library, since there is no way to disable
4365 its creation for now.
4367 For information, the beginning of coding on GRAS was back in june
4368 2003. I guess that every line has been rewritten at least twice since