1 ----------------------------------------------------------------------------
3 SimGrid (3.29.1) NOT RELEASED YET (v3.30 expected December 21. 2021, 15:59 UTC)
5 Main user-visible changes:
6 - The SimDag API for the simulation of the scheduling of Directed Acyclic
7 Graphs has been dropped. It was marked as deprecated for a couple of years.
8 We finally complete the implementation of what has been called SimDag++
9 internally, i.e., porting the different features of SimDag on top of S4U.
10 The new way to simulate the execution of dependent activities directly by
11 maestro (without any other actor) is details in the examples/cpp/dag-* series
13 - The removal of SimDag led us to also remove the export to Jedule files that
14 was tightly coupled to SimDag. The instrumentation of DAG simulation is still
15 possible through the regular instrumentation API based on the Paje format.
18 - New function: Engine::run_until(date), to split the simulation.
19 - New signal: Activity::on_veto, to detect when an activity fails to start.
20 - New function: Engine::track_vetoed_activities() to interrupt run()
21 when an activity fails to start, and to keep track of such activities.
22 Please see the corresponding example for more info.
25 - Dynamic costs for MPI operations: New API to allow users to dynamically
26 change injected costs for MPI_Recv, MPI_Send and MPI_Isend operations.
27 Alternative for smpi/or, smpi/os and smpi/ois configuration options.
30 - New section: "SimGrid MPI calibration of a Grid5000 cluster"
31 presenting how to properly calibrate MPI communications in SimGrid.
32 - Complete and reword the platform section, which is now completed.
35 - Thread contexts are used by default with Python bindings. Other kinds of
36 contexts revealed unstable, specially starting with pybind11 v2.8.0.
38 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
39 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
40 - FG#99: Weird segfault when not sealing an host
42 ----------------------------------------------------------------------------
44 SimGrid (3.29) October 7. 2021
46 The "Ask a stupid question" release.
48 We wish that every user ask one question about SimGrid to celebrate.
49 On Mattermost, Stack Overflow or using the issues tracker.
52 New modeling features:
53 - Non-linear resource sharing, modeling resources whose performance heavily degrades with contention:
54 - The total capacity may be updated dynamically through a callback
55 and depends mainly on the number of concurrent flows.
56 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
58 - Dynamic factors: model variability in the speed of activities
59 - Each action can now have a factor that affects its progression.
60 This multiplicative factor is applied when updating the amount of work
61 remaining, thereby an activity with factor=0.5 only uses half of the
62 instantaneous power/bandwidth it is allocated and will appear twice
63 slower than what it actually consumes.
64 - This can be used to model a overhead (e.g., there is a 20 bytes
65 header in a 480 bytes TCP packet so the factor 0.9583) but the novelty
66 is this factor can now easily be adjusted depending on activity's and
67 resources characteristics.
68 - This existed for network (e.g., the effective bandwidth depends
69 on the message in SMPI piecewise-linear network model) but it is now
70 more general (the factor may depend on the source and destination and
71 thus account to different behaviors for intra-node communications and
72 extra-node communications) and is available for CPUs (e.g., if you
73 want to model an affinity as in the "Unrelated Machines" problem in
74 scheduling) and disks (e.g., if you want to model a stochastic
76 - For that, resources can be provided with a callback that computes
77 the activity factor when creating the action.
78 - Example: examples/cpp/exec-cpu-factors
79 - The same mechanism is also available for the latency, which
80 allows to easily introduce complex variability patterns.
83 - Added support to programmatic platform creation in Python.
84 Example: examples/python/clusters-multicpu
87 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
88 This can only be set through the API, not through XML files.
92 - Multiple fixes to ensure reproducibility of tracing
93 - scan/excan can now be replayed
94 - wait action now uses ranks and not pid, as the other ones.
95 - smpi/init and smpi/finalization-barrier are now valid for replays.
96 - exit() is now intercepted by SMPI to avoid premature shutdown of
97 simulation. First non 0 return codes is returned as simulation return
101 * New section "Release Notes" documenting recent and current developments.
102 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
103 * Improvements in API Reference for C++ and Python interfaces.
106 - Make wifi creation compatible with ns-3 version 3.34 too.
108 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
109 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
110 - FG#78: Multiple fixes for SMPI replay:
111 - TI tracing of allotallv/w was outputting wrong values
112 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
113 - FG#77: Search feature of doc is broken (update sphinx theme version)
114 ----------------------------------------------------------------------------
116 SimGrid (3.28) July 14. 2021
118 The Victoriadagarna Release.
121 - C++ platform interface: Users can now describe their platform directly in C++.
122 This provides greatly flexibility and performance improvement for complex
123 platforms. Main features:
124 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
125 "zones", instead of single hosts. This feature enables the description
126 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
127 - StarZone: new zone with a Star-like topology. The routes are defined
128 as a set of links used to communicate from node to everybody (node<->ALL).
129 - Split-Duplex links: auxiliary method to create split-duplex links in
130 the platform, easing its utilisation. It automatically creates both UP
131 and DOWN links (similarly as done in XML).
132 - Please refer to the documentation and the examples included:
133 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
134 - New plugin: Producer-Consumer with monitor. Just requires to include the
135 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
136 associated example (examples/cpp/plugin-prodcons).
139 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
140 s4u::Io::wait_any(), s4u::Io::wait_any_for().
141 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
142 parameter by reference, instead of a pointer.
143 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
144 Explicitly cancel the activity to get back to previous behavior.
145 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
148 - The default SMPI compiler flags are no more taken from the environment.
149 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
150 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
152 --cfg=smpi/finalization-barrier: which can be used to add
153 a barrier inside MPI_Finalize. This can help for some codes which cleanup
154 data attached to a process, but still used in other SMPI processes.
155 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
156 is active when set to false, to keep going after a small error
157 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
158 which may or may not be problematic in the end.
160 - fix behaviour, as maximum iteration count could be ignored
161 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
162 use sampling when the same kernel is called with a different set of parameters
163 which have an impact on the timing.
164 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
165 It should now work with smpi/auto-shared-malloc-thresh.
166 - Improve error handling and reporting in multiple places
167 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
168 validity checks, truncated messages are now an error, return errors
169 when explicitely deleted handles are reused, ...)
170 - RMA: multiple fixes and stability improvements.
171 - analysis (-analyze flag in smpirun):
172 - SMPI can now report buffer leaks as well as MPI handles leaks,
173 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
174 - if -trace-call-location is used when compiling, SMPI can report
175 origin of leaked handles/buffers
176 - group leaks by type/origin in output message if possible
177 - New implemented MPI calls: MPI_Comm_test_inter
180 - Changed internal implementation of bandwidth factors in network models.
181 Models affected: CM02, LV08 (default), SMPI, IB.
182 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
183 Bandwidth factors are applied to communications to describe that users
184 cannot use 100% of the available bandwidth. For example, the default network model,
185 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
186 behavior was implemented by limiting the bandwidth available in the LMM
187 system for this flow. This may give the false impression that there is
188 bandwidth available for other flows due to its underutilization, especially
189 for the dynamic bandwidth factors used in SMPI models.
190 To avoid this, we have modified the implementation so that each flow uses the
191 maximum physical bandwidth according to the LMM system.
192 However, the actual throughput of the flow seen by the user is defined by
193 the physical bandwidth multiplied by the bandwidth factor.
194 This change impacts on the simulation results for all network models on
195 which we have bandwidth factors configured.
196 *****************************************
197 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
198 *****************************************
199 This change may impact on the timing of your simulation results.
200 Take care when comparing simulations from different SimGrid's
201 versions. Sorry for the inconvenience.
202 - Dynamic network factors: users can configure a callback to define
203 the network factors dynamically. This API is available at
204 simgrid::kernel::resource::NetworkModelIntf.
205 - Users have access to complete information about the current communication
206 to decide which factor to apply. This includes: message size, source and
207 destination hosts, links and zones traversed.
208 - Dynamic factors for both latency and bandwidth.
209 - For more details, see the example in (examples/cpp/network-factors).
210 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
211 deprecated since version 3.24, are no longer supported. Instead, use
212 "wattage_off" and "wattage_per_state".
215 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
218 - New tutorial: Model-checking and formal assessment
219 - New sections: "Demystifying the routing" and "C++ platforms"
220 - Update and improve the part on visualization in MPI and Algo tutorials.
221 - Improve the section on routing: how to define it, how it's used internally
222 - Fix many issues, broken links and missing references in doxygen and Sphinx
225 - Lua platform files are deprecated. Their support will be dropped after v3.31.
228 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
230 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
231 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
232 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
233 - FG#64: Configuring smpi/IB-penalty-factors
234 - FG#67: Running computation concurrently with MPI_Iallreduce
235 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
236 - FG#71: Segmentation fault on invalid gw_src/gw_dst
237 - GH#322: Issue when an actor kills his host vm
239 ----------------------------------------------------------------------------
241 SimGrid (3.27) March 29. 2021
243 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
245 Main user-visible changes:
246 - All functions marked as deprecated for one year were removed.
247 - Removed all that was related to the "Storage" type in all APIs and bindings.
248 "Disks" have been introduced more than a year ago to replace them for the better.
249 It's time to finish this replacement.
250 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
251 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
252 avoiding "initializer element is not constant" errors with some codes. Please be
253 careful to recompile your MPI codes if switching to this new SimGrid version.
254 - Network model ns-3 now requires version ns-3.28 or higher.
255 - Removed some option aliases that were deprecated since v3.23 (June 2019)
256 - 'verbose-exit' is now 'debug/verbose-exit'
257 - 'simix/breakpoint' is now 'debug/breakpoint'
258 - 'clean-atexit' is now 'debug/clean-atexit'
261 - New functions: Comm::sendto_{init,async} to initiate a communication
262 on between two (possibly remote) hosts.
263 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
264 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
265 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
266 or Mailbox::get_async<void>() if you really want to play with void*.
267 - Unify the interface of Activity::on_{start/activity}
268 - New function: Comm::get_dst_data()
269 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
272 - ptask should now work better with multicore hosts (FG#37).
273 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
274 - Internal cleanups and refactoring around models, to prepare for programmatic
276 - Modernize our tests of the performance models
280 - New section: Configuring the logs from the command line.
281 - New section: Programming API of logs.
282 - New section: Getting realistic results
283 - New section: Examples of SMPI Usage
284 - New part: SimGrid's Internals, with a section on Design goals
285 - Move back to the classical Breathe tool chain, and fix many small glitches.
288 - Handle leak checking: SMPI will now display a message if MPI handles are still
289 present at the end of a simulation. Option smpi/list-leaks:n will display the n
290 first leaked pointers and their types.
291 - smpi/display-allocs option : Display Memory allocations analysis, with hints
292 about the largest allocations. This only works for allocations with malloc/calloc
293 and not compiled with SMPI_NO_OVERRIDE_MALLOC
294 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
295 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
296 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
297 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
298 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
299 - Replay now opens and parses the trace file only once for all ranks when possible.
301 - better handling of MPI_PROC_NULL
302 - MPI_Group_incl was broken in some cases
303 - PAPI tracing now fixed
304 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
305 - SMPI will now fail properly when receive size < sent size, instead of
306 just setting MPI_ERR_TRUNCATE and keep going.
307 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
308 - Remove useless build patches and split build/run patches on the page
309 - Build in parallel when possible
310 - Test on yum-based systems
311 - New nightly tested Proxy-apps:
312 - OpenMPI mpi-test-suite
313 - Coral MCB, LSMS, Qbox, CLOMP
314 - ECP/Mantevo Cloverleaf3D, Ember
315 - Codevault bhtree_mpi
318 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
319 - Start to overhaul the implementation to modernize it (still ongoing).
322 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
323 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
324 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
325 - Deprecate xbt_dynar_to_array().
326 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
328 C binding and interface:
329 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
330 - Constness requirement has been lowered for parameter argv of sg_actor_start()
331 and sg_actor_create() It's now (char* const*). The previous version is still
332 available as sg_actor_start_() and sg_actor_create_().
334 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
335 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
336 - FG#37: Parallel tasks are limited to 1 core per host
337 - FG#62: Running "smpirun -replay" on large networks
338 - FG!46: Fix a few potential memory leaks in SMPI colls
339 - GH#343: Wrong consumed energy when an exec timeouts
341 ----------------------------------------------------------------------------
343 SimGrid (3.26) December 15. 2020
345 The Release release (the French lockdown was eased today).
347 Important user-visible changes:
348 - SimGrid now requires a compiler with C++14 support.
349 Sibling projects should upgrade their FindSimgrid.cmake
350 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
351 several users had difficulties to understand issues when using high bandwidth or
352 small latency events. The new value was already the default for SMPI and
353 should not cause too much performance hit. It can change some simulation
354 timings if simulation was skipping a lot of small events (which was actually
355 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
356 - On some network models (L07, CM02, LV08), an implicit loopback link is created
357 for local communication. Default bandwidth was 500MBps and latency was 15µs,
358 which was rather bad. Default values have been changed to 10GBps and 0 latency.
359 Options network/loopback-lat and network/loopback-bw have been added to adapt
360 these values if needed. Network and routing models already providing separate
361 loopbacks (clusters and such) are not impacted by this change.
362 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
363 metadata that are added to each MPI message by MPI libraries (tag,
364 src/dest, and communicator information). This may change simulation times
365 for very small message sizes, but should be more realistic.
368 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
369 - New plugins: wifi_energy and link_load. See the associated examples.
370 - New examples: wifi networks, and communication suspend.
373 - LTO is now enabled for Intel/clang compilers.
374 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
375 Setting it to "auto" will use all available cores, while setting it to n will
376 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
377 - Remove obsolete runtime option 'contexts/parallel-threshold'.
378 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
379 has been kept for compatibility.
380 - Finally remove obsolete snake_case() aliases for runtime options.
381 - Further improve the documentation.
384 - the on_exit() of each actor is also executed when the simulation deadlocks.
385 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
386 An example is provided for s4u::Comm but it should work with Exec and Io.
389 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
390 https://framagit.org/simgrid/SMPI-proxy-apps
391 - MPI/IO: supports names starting with "./"
392 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
394 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
395 building shared libraries, this may cause issues, so environment variable
396 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
397 - most temporary files should now be created in /tmp dir (or equivalent).
398 If this one does not allow execution of code (noexec flag), this may cause issues.
399 Please use another tmp directory (using TMPDIR or equivalent system variable)
403 - A large refactoring is ongoing, but this should be transparent for users so far.
406 - Drop xbt_str_split().
408 C binding and interface:
409 - The return type of the sg_actor_on_exit() callbacks is now 'void'
410 (int value was previously ignored)
411 - Many C functions were renamed to follow a common naming style
412 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
413 names are kept until version 3.30.
414 - Many MSG tests were converted to the new S4U's interface in C, that
415 was extended for that.
417 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
418 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
419 - FG#41: Add sg_actor_create C interface
420 - FG#43: xbt::random needs some care
421 - FG#48: The Impossible Did Happen (yet again)
422 - FG#50: Suspending an actor executed at the current timestamp fails
423 - FG#52: Zero-seconds timeout : "That's in the past already"
424 - FG#53: Crash while using ns-3 network model
425 - FG#54: How to suspend a comm?
426 - FG!22: Stochastic Profiles
427 - FG!24: Documentation and fix for xbt/random
428 - FG!35: Add a modeling hint for parallel links in doc
429 - FG!36: [xbt/random] Read/Write the state of the RNG
430 - FG!42: Link cumulated load plugin
431 - FG!44: Test condition variable wait_until+notify_all
432 - GF#18137: Allow different stack sizes?
433 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
434 - GH#139: Allow pthread creation in SMPI
435 - GH#336: Packet-level simulation using SMPI?
436 - GH#345: Error in the LMM while migrating a VM
437 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
438 - GH#352: pip install / python setup.py install fail to find pybind11
439 - GH!337: Fix link_energy plugin for wifi platforms
440 - GH!339: Add Mailbox set_receiver method to python binding
441 - GH!344: Cast hugepages macros parameters to int64
443 ----------------------------------------------------------------------------
445 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
447 The Palindrom Day Release.
449 Important user-visible changes:
450 - Improve the Python usability (stability and documentation).
451 - A nasty synchronization bug was ironed out, see also below.
452 - Python's doc was organized as a separate tree, now integrated with C++.
453 - C bindings of S4U were not part of the doc.
454 - The C++ doc was also improved as methods are now split by theme.
455 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
456 - This is mandatory to use the Java bindings.
457 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
458 - SimDAG++: Automatic dependencies on S4U activities (experimental)
459 - Some features are already implemented but not all of them
460 - Cannot block an activity until it's scheduled on a resource
461 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
462 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
465 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
466 - Drop support for Viva/Triva (old visualization tools), hence removed
467 TRACE_get_node_types() and TRACE_get_edge_types() functions.
470 - Require pybind11 version 2.4.
471 - Greatly improve locking during context switches with Python.
472 - Declaring Python callbacks to this_actor::on_exit is now working.
473 - No more tests are failing currently, even on win32.
476 - Actor: Merge signals on_migration_start/end into on_host_change
477 - Actor: Rename migrate() into set_host()
478 - Disk: Allow users to get the read and write nominal bandwidth values
479 - Exec: Implement wait_for(timeout)
480 - Io: Implement wait_for(timeout)
483 - Parse errors now raise a simgrid::ParseError that you may want to catch.
486 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
487 been renamed to fix a spelling error. As usual, the previous definitions are
488 kept with a deprecation warning. However, to avoid runtime errors, it is now
489 forbidden to override these deprecated methods in a derived class. Please use
490 the new names immediately if you need to override them.
493 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
494 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
495 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
496 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
497 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
498 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
499 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
500 - New module: random, providing classical random numbers generators.
503 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
504 allocations' size above which they are considered "shared" by default (as if
505 they were performed through SMPI_SHARED_MALLOC macros).
506 Default = 0 = disabled feature.
507 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
508 This can cause some troubles if codes are already overriding these. If this
509 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
510 help, but will make this feature unusable.
511 - Input check is now performed much more consistently and outputs precise
512 warnings for each error found
513 - New supported calls : MPI_File_set_view, MPI_File_get_view
514 - MPI I/O now only supports disks, as storage is discontinued.
515 - MPI I/O allows opening files without specifying absolute path
516 - Note : SMPI configuration options are now initialized by smpi_init_options()
517 call, which should be called if SMPI is not being used through smpirun.
519 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
520 - FG#9: Python bindings crashing
521 - FG#39: Missing s4u::Comm::wait_any_for example
522 - FG#42: Add support for ThreadSanitizer (TSan)
523 - FG!19: Removing RngStream
524 - FG!20: A module for RNG calls
525 - FG!21: Choice between ad-hoc and standard distributions implementations
527 - FG!25: Fix link in Mutex doc
528 - FG!26: Fix links in SMPI interface doc
529 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
530 - GH#305: Unscheduled tasks are still excuted
531 - GH#313: smpirun: manual outdated w.r.t. --help
532 - GH#321: [S4U] Get task remaining work ratio
533 - GH#323: Crash when an actor turn off his physical host
534 - GH#335: Missing links on dragonfly example svg
535 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
536 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
538 ----------------------------------------------------------------------------
540 SimGrid (3.24) October 9. 2019
542 The Clean Disk Release.
545 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
546 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
547 thus becomes the default resource to execute I/O operations. Examples
548 and the file system plugin have been modified to use disks instead of
549 storage elements. Storage can still be used but is doomed to disappear.
550 Users are thus adviced to update their simulators accordingly.
551 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
552 for consistency with pthread_barrier_wait()
553 - Host::get_englobing_zone() returns the englobing netzone
554 - Actor::on_destruction is now called in the destructor
555 Actor::on_termination new signal called when the actor terminates
557 - Global signals are now part of the Engine:
558 - on_platform_creation: after config settings, before the XML parsing
559 - on_platform_created: right after the XML parsing
560 - on_time_advance: each time the clock advances
561 - on_simulation_end: after simulation, before cleanups
562 - on_deadlock: as the name implies.
564 - sg_{actor,host,link}_{data,data_set}() now all exist.
565 Use them to attach user data to the object and retrieve it.
568 - Introduce an experimental Wifi model. It sounds reasonable
569 according to the state of the art, but it still has to be properly
570 validated, at least against ns-3 if not against reality.
571 - Improved the usability of ns-3. Several bugs were ironed out.
572 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
573 It is now expressed as 'idle:epsilon:allCores' to properly model the
574 consumption of non-whole tasks on mono-core hosts. More info in the doc.
577 - convert a new set of functions to the S4U C interface and move the old MSG
578 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
581 - Fortran bindings for DVFS have been removed.
582 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
583 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
584 - SMPI can now be selected by cmake's find_module(MPI) with
585 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
586 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
587 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
588 - trace-call-location can be used with TI traces, and replayed, and is compatible
589 with smpi/comp-adjustment-file.
590 - sleep events are now correctly traced/replayed.
591 - Default for trace-call-location is now to use file names and not full paths.
592 To revert to previous behavior (in case of collision of filenames), option
593 "smpi/trace-call-use-absolute-path" can be set to yes.
596 - Use the included xxHash as an hash implem when C++14 is usable.
597 - Option model-checker/hash was removed. This is always activated now.
598 - New option smpi/buffering controls the MPI buffering in MC mode.
599 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
600 This is useful to check for MPI compliance.
603 - New section on plugins: how to define a new one, and existing ones.
606 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
607 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
610 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
611 and <mount> tags. The rationale is finish to decouple what should be handle
612 by SimGrid kernel (i.e., resources and activities) from plugins built in
613 user space (e.g., the file system plugin). This new tag currently accepts
614 only three attributes (id, read_bw, and write_bw). All the other information
615 that was declared with the storage related tags now has to be expressed as
616 properties. An example of platform using this new tag is available at
617 examples/platforms/hosts_with_disks.xml
620 - 'expect signal' can now accept more than one potential signal.
622 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
623 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
624 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
625 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
626 - FG#31: per-actor data
627 - FG#34: SG_BARRIER_SERIAL_THREAD?
628 - FG#35: model-checker does not like buster-produced binaries
629 - FG!13: MC: complete workaround in the error msg seen on modern systems
630 - FG!15: execute_flops now logs compute
631 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
632 - FG!17: ns-3: unblock the right number of communications + others issues
633 - FG!18: Improving the performance of the ns-3 bindings
634 - GH#189: Energy consumption of parallel tasks
635 - GH#219: Error in the throughput of TCP transfer
636 - GH!330: Minor polishing of English: Configuring_Simgrid
637 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
638 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
639 - GH!334: Rework energy plugin, again
641 ----------------------------------------------------------------------------
643 SimGrid (3.23.2) July 8. 2019
646 - Nicer introduction page.
647 - Migrate the "Deploy your application" page to the new doc.
648 - Move Java as a subtree of MSG.
651 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
654 - Drop xbt_dynar_sort_strings().
657 - Really fix FG#26: Turning off a link should raise NetworkFailureException
658 - FG#27: Wrong exception thrown to wait_any when link is turned off
659 - GH#328: Java: Canceling multiple tasks in a single vm/host
661 ----------------------------------------------------------------------------
663 SimGrid (3.23) June 25. 2019
665 The Exotic Solstice Release.
668 - SunOS and Haiku OS support. Because exotic platforms are fun.
669 - Stop setting random seed with srand48() at initialization.
670 - Use addr2line as a fallback for stacktraces when backtrace is not available.
671 - Build option -Denable_documentation is now OFF by default.
672 - Network model 'NS3' was renamed into 'ns-3'.
675 - Simgrid can now hopefully be installed with pip.
678 - wait_any can now be used for asynchronous executions too.
681 - New log appenders: stdout and stderr. Use stdout for xbt_help.
682 - Drop xbt_dict_dump.
685 - 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
686 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
687 - MPI interface is now const correct for input parameters
688 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
689 - Fortran bindings for async collectives.
690 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
693 - Remove option 'model-check/record': Paths are recorded in any cases now.
694 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
695 always sparse. That's an immense gain in memory, and can even be
696 faster because of cache effects. So there is no need to clutter the
697 code to allow the user to go for the unefficient mode.
700 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
701 models proved to be more accurate than these old experiments.
703 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
704 - FG#1: Broken link in error messages
705 - FG#2: missing installation documentation
706 - FG#3: missing documentation in smpirun
707 - FG#6: Python bindings not available on PyPI
708 - FG#7: simple cmake call requires doxygen
709 - FG#8: make python bindings an optional dependency
710 - FG#10: Can not use MSG_process_set_data from SMPI any more
711 - FG#11: Auto-restart actors forget their on_exit behavior
712 - FG#12: -Denable_lto=OFF doesn't disable LTO
713 - FG#13: Installs unstripped file 'bin/graphicator'
714 - FG#14: Installs the empty directory 'doc/simgrid/html'
715 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
716 - FG#17: Dead link in doc (pls_ns3)
717 - FG#20: 'tesh --help' should return 0
718 - FG#21: Documentation link on http://simgrid.org/ broken
719 - FG#22: Debian installation instruction are broken
720 - FG#26: Turning off a link should raise NetworkFailureException exceptions
721 - GH#133: Java: a process can run on a VM even if its host is off
722 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
723 - GH#326: Valgrind-detected error for join() when energy plugin is activated
725 ----------------------------------------------------------------------------
727 SimGrid (3.22) April 1. 2019
729 The Easter Christmas Release (this one is somewhat late).
732 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
733 - Not complete yet: asynchronous activities (amongst others) are still missing
734 - Still ongoing: the interface may change in the future. We need more
735 testers! Please report any glitches.
736 - No new project using Java should start now. Please switch to Python.
739 - Some of the internal cleanups may lead to speed improvements:
740 - The hard limitation on the amount of simulated actors with
741 Java+Mac was removed. Now, the available memory is the only limit.
742 - Our refcounting was tidyied, leading to 10% speedups in some cases.
743 - We are still working on making our code robust to the actor kills
744 and hosts' churn. Things are improving, but it's not perfect yet.
745 - Replaced our own code to display a backtrace (that was forking addr2line)
746 with the Boost.Stacktrace library.
747 You won't see your backtraces without this optional dependency.
748 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
749 - Stop setting random seed with srand() at initialization.
752 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
753 XML file version remains 4.2 since old files are still compatible.
756 - Process termination which was broken at version 3.21 has been repaired.
757 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
758 - Hide the examples into examples/deprecated. New users should use Python.
761 - Drop MSG_process_create_from_stdfunc() from the API.
762 This C++ function was a pimple in the C API, made necessary at some
763 point by the Java bindings. This is fixed now.
764 - Hide the examples into examples/deprecated. New users should use S4U.
765 - MSG_process_create and MSG_process_attach now crash if the host used
766 to run this process is off.
767 - Fix the protype of MSG_process_on_exit()
768 Now use (int,void*) callbacks instead of (void*,void*) ones.
769 The implementation was ways too messy to actually work, I guess.
772 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
773 too often and provide more accurate timings.
774 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
775 - MPI_Alltoallw support
776 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
777 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
778 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
779 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
781 - Basic implementation of generalized requests (SMPI doesn't
782 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
785 - Drop sg_cmdline. Please use xbt_cmdline instead.
786 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
787 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
788 OsSemaphore is implemented in a portable way with C++11 threads.
789 This should allow much more threads to be created at the same time,
790 allowing Mac and Java users to simulate many more actors.
791 - Implement the 'thread' factory with std::thread instead of xbt ones.
792 It is not possible to set the stack size with threads anymore, but
793 -fsplit-stack is the way to go nowadays when using threads.
794 - Drop the xbt_os_thread_t module (now unused)
795 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
796 - Drop xbt_str_join_array().
797 - Drop cunit, use Catch2 instead.
800 - Many cleanups in the kernel::activity namespace. This was long
801 overdue, and shall open the path to many future endeavors.
804 - #132: Java: a process can not shut down its own host
805 - #220: S4U: detached send causes exception if sender terminates soon after sending
806 - #261: Document the parameters of parallel execution's constructor
807 - #300: [s4u] BarrierPtr is missing
808 - #314: SMPI args internal cleanup
809 - #316: Fix a bug related to the CPU utilization of multi-core VM
810 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
811 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
812 - #325: Turning off a host has different behavior on sleeping actors and computing actors
814 ----------------------------------------------------------------------------
816 SimGrid (3.21) October 3. 2018
818 The Restarting Documentation (TRD) Release.
821 - Convert the user manual to Sphinx for improved usability.
822 Unstable version now hosted on https://simgrid.frama.io/simgrid
823 Many glitches remain; stay tuned.
824 - Restrict the Doxygen reference API to the installed header files.
825 (documenting all internals this way is overkill + doxygen scales badly)
826 - New tutorial on S4U (highly inspired from the old MSG one, but with
827 a git repository to fork as a starting point, and a docker image)
828 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
829 - The Developper manual is still to be converted (not compiled until then)
830 - Some parts are still missing in this conversion (such as the
831 platform and deployment sections) while others were blindly converted
832 and would need more love (such as the configuration flags).
833 Things will certainly further improve in the future releases.
836 - Fully reimplement auto_restart mechanism that was utterly broken
837 (fix #22, #131 and #281 that were all very related)
838 - Implement semaphores (fix #303 and #291)
839 - When creating an actor from a function and its parameters,
840 move-only parameters are not allowed anymore, as it would prevent
841 the actor to be restartable if its parameters are consumed on run.
842 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
843 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
844 IO operations that can be started, waited for, or canceled as a regular
846 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
850 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
851 - New host variable: 'core_count'
854 - Remove xbt_os_thread_specific features
855 - Remove portability wrapper to condition variables
856 - Remove xbt_os_thread_yield()
859 - MPICH collective selector now mimics MPICH 3.3b
860 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
861 - MPI_Init(NULL,NULL) is now allowed
862 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
863 - Command-line arguments should now be handled properly in Fortran simulations
864 - MPI Attributes and related callbacks should work in Fortran
865 - Apps using C + Fortran should now work
866 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
867 - Support MPI_Error_String
870 - Due to an internal bug, Msg.run() must now be your last line.
871 We hope to fix it in a future release, and we are sorry for the inconvenience.
874 - #22: Process autorestart seem to only work with CAS01 cpus
875 - #56: Feature request: dynamic SMPI replay
876 - #93: simgrid should not eat --help
877 - #111: How to change the loopback link characteristics?
878 - #116: Platform cluster radical
879 - #117: Platform cluster/host order
880 - #129: Replay_multiple on steroid
881 - #131: Java : autostart does not work
882 - #143: Setting a breakpoint at a given time
883 - #235: Network energy models should be integrated
884 - #264: Add ptask L07 resource tracing
885 - #271: Dynamic and manual replay of SMPI traces
886 - #279: Breakpoints option uses time deltas instead of absolute time
887 - #281: Daemonized actor no longer daemonized after an auto-restart
888 - #282: TIT convertor erase traces when using absolute path in the trace list file
889 - #285: segfault when a s4u actor kills itself with msg process tracing activated
890 - #286: Pajé traces are not exposing the number of cores
891 - #287: Command-line parsing should stop when encountering '--'
892 - #288: MPI_Init(NULL, NULL)
893 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
894 - #290: Method to check if a mailbox has a message ready to be consumed
895 - #291: [s4u] semaphores do not seem available yet
896 - #294: Alltoall collective implementation of mpich
897 - #295: Starting a Host as OFF is not supported
898 - #296: DTD too permissive
899 - #299: Add a s4u barrier example
901 ----------------------------------------------------------------------------
903 SimGrid (3.20) Released June 24. 2018
905 The proxy snake_case() release.
908 - The whole API is now using snake_case() instead of mixing it with camelCase.
909 Compatibility wrappers in place for one year.
910 - We took the opportunity to increase the consistency of the API.
911 Please report any remaining glitches.
912 - New function simgrid::s4u::Host::get_actor_count:
913 Returns the number of actors running on a specific host.
916 - SMPI is now tested with ~45 proxy apps from various sources, with none or
917 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
918 - Replay: The replay file has been re-written in C++.
919 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
920 supported. They are stored in the trace and used when replayed.
921 - Basic support of MPI_Cancel. Robustness not guaranteed.
922 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
923 (only for MPI_COMM_TYPE_SHARED).
924 - New option: smpi/privatize-libs, to add external shared libs to be privatized
925 by SMPI. They will be copied locally and loaded separately by each process.
926 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
927 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
928 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
929 with internal index optind with multiple processes. Only works if MPI_Init has
931 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
932 syntax without preprocessing tricks.
935 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
936 behavior tracing is no longer limited to MSG
937 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
938 this means that traces now need to be updated or re-obtained; both Irecv
939 and Isend lines in your traces have now in total 5 mandatory fields:
940 <rankid> <command> <to/from rankid> <tag> <size>
941 To update your traces, it suffices to add a 0 for the tag here.
942 - TIT now also supports waiting for a distinct request via MPI_Wait.
943 Wait/Test now wait for a specific request, not just the last one that was
944 issued. This unfortunately means another update, because we need to
945 identify which request you want to wait for. We do this via the
946 triplet (sender, receiver, tag), which needs to be added:
947 <rankid> <command> <sender> <receiver> <tag>
948 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
949 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
952 - The deprecation of MSG is ongoing (but this should not impact you).
953 Many MSG functions are now simple wrappers on the C API of S4U. If
954 you wish to convert your code to S4U, find the S4U counterparts of
955 your MSG calls in src/msg/msg_legacy.cpp.
956 - MSG can still be used, but won't evolve anymore.
959 - Allow to run the Link energy plugin from the command line with
960 --cfg=plugin:link_energy
961 - Rename Energy plugin into host_energy
962 - Rename Load plugin into host_load
965 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
966 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
967 - Improve performance when handling timeouts of simix synchros.
970 - Config: the C API is now deprecated (will be removed in 3.23), and
971 the C++ API has been slightly improved.
974 - Fix several build issues on OSX.
975 - Move simgrid_config.h to simgrid/config.h (old header still working)
978 - #143: Setting a breakpoint at a given time
979 - #150: Inconsistent event names in SMPI replay
980 - #258: daemonized actors hang after all non-daemonized actors have completed
981 - #267: Linker error on unit_tmgr
982 - #269: SMPI: tracing of MPI_Wait/all/any broken
983 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
985 ----------------------------------------------------------------------------
987 SimGrid (3.19.1) Released March 22 2018
989 The Fixed ABI Release.
991 As you may know, we are currently refactoring SimGrid in deep.
992 Upcoming SimGrid4 will be really different from SimGrid3: modular,
993 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
995 Our goal is to smooth this transition, with backward compatibility
996 and automatic update paths, while still progressing toward SimGrid4.
998 SimGrid remains open during works: The last pure SimGrid3 release was
999 v3.12 while all subsequent versions are usable alpha versions of
1000 SimGrid4: Existing interfaces remain unchanged, but the new S4U
1001 interface is budding and the internals are deeply reorganized.
1003 Since 2015, we work hard to reduce the changes to public APIs. When
1004 we need to rename a public library symbol in S4U, we let your
1005 compiler issue an explicative warning when you use the deprecated
1006 function. These messages remain for four releases, i.e. for one full
1007 year, before turning into an error. Starting with v3.15, your can
1008 also adapt to API changes with the SIMGRID_VERSION macro, that is
1009 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
1011 Starting with this v3.19.1, our commitment to reduce the changes to
1012 the public interfaces is extended from the API to the ABI: a program
1013 using only MSG or SimDag and compiled against a given version of
1014 simgrid can probably be used with a later version of SimGrid without
1015 recompilation. We will do our best... but don't expect too much of
1016 it, that's a really difficult goal during such profund refactoring.
1018 The difference between v3.19 and v3.19.1 is that the former was
1019 accidentally breaking the ABI of MSG, while the later is restoring
1022 S4U and kernel APIs will still evolve until SimGrid4, with one-year
1023 deprecation warnings as currently. In fact, cleaning up these
1024 interfaces and converting them to snake_case() is one release goal of
1025 v3.20. But don't worry, we are working to smooth this upgrade path.
1027 In summary, new projects should start with S4U to benefit of the
1028 future, but old MSG projects should still be usable with no change.
1030 ----------------------------------------------------------------------------
1032 SimGrid (3.19) Released March 20 2018
1034 The Moscovitly-cold Spring release.
1036 SMPI user-visible improvements:
1037 - Support of MPI_Type_create_subarray
1038 - Unimplemented MPI calls now abort (instead of issuing a warning)
1039 Some harmless calls around errhandlers still only issue a warning.
1040 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
1041 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
1042 is now possible (mmap privatization was refactored to this end).
1043 Adding new SMPI instances should also be technically doable now.
1045 SMPI internal cleanups:
1046 - Removed index notion from SMPI -- no more "getPid() - 1"!
1047 In tracing files, actors are called rank-<PROCESS-ID> but these are
1048 NOT the real ranks (tracing several communicators is not supported yet).
1049 - Deprecate smpi_process_index() (will be removed in 3.22)
1050 - Better testing on classical proxy apps, through an external project:
1051 https://github.com/simgrid/SMPI-proxy-apps/
1054 - Use a graphical TOC to make it easier to find the documentation you need
1055 - Revamp the MSG tutorial
1058 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
1059 - Remove parameter of MSG_process_killall().
1060 Resetting the PID was bogus anyway (several actors could have the same PID).
1063 - Execution->setHost() can be called after start(), to migrate it.
1064 - Comm::test_any() is now implemented.
1065 - s4u::Actor now has onCreation() and onDestruction() signals.
1066 - Install some kernel header files for the users' plugins and more.
1067 Warning, their API is really not stable yet.
1068 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1069 include/simgrid/kernel/routing/*Zone.hpp
1072 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1074 Build System and other cleanups:
1075 - Remove unused run-time parameter "tracing/onelink-only".
1076 - SimGrid now works with intel compilers (MC still buggy)
1077 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1078 - tesh kills subprocesses on timeout (but not on windows)
1081 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1084 - #194: Feature request: simgrid::s4u::Comm::test_any()
1085 - #245: migrating an actor does not migrate its execution
1086 - #253: Feature Request: expose clusters as objects
1087 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1088 - #255: Tesh broken on Windows
1089 - #256: Modernize FindSimGrid.cmake
1090 - #257: Fix (ab)use of CMake install
1092 ----------------------------------------------------------------------------
1094 SimGrid (3.18) Released December 24 2017
1096 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1098 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1099 - Support for MSG should not be removed from SimGrid before 2020,
1100 but future evolutions will be limited to the integration of
1101 user-provided patches.
1104 - s4u::onDeadlock() signal, run before stopping the simulation.
1105 - s4u::Exec: asynchronous executions (abstraction of background
1107 - s4u::Host->getLoad() returns the achieved speed in flops/s
1110 - Link::name() is deprecated and will be removed in v3.21, use
1111 Link::getCname() instead.
1112 - Mailbox::getName() changed to return a std::string, use
1113 Mailbox::getCname() instead to get a char*.
1114 - Storage::getName() changed to return a std::string, use
1115 Storage::getCname() instead to get a char*.
1116 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1117 to have both allocation and deallocation of the map in user space.
1120 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1121 The amount of flops remaining to do is a vector, not a scalar.
1122 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1123 name implies on both sequential and parallel tasks.
1124 - Both changes fix GitHub's #223 using PR #237 as a basis.
1125 Thanks Michael Mercier.
1126 - Most examples were converted to S4U and hidden elsewhere at tests.
1127 Rationale: we still want MSG to work; we want newcomers to use S4U.
1130 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1131 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1134 - Switch to the faster dlopen privatization mechanism by default
1135 - Documentation improvments
1138 - Live migration is getting moved to a plugin. Dirty page tracking is
1139 the first part of this plugin. This imply that VM migration is now
1140 only possible if one this function is called:
1141 - C/MSG: MSG_vm_live_migration_plugin_init()
1142 - C/C++: sg_vm_live_migration_plugin_init()
1143 - Java: Msg.liveMigrationInit()
1144 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1148 - Define class simgrid::xbt::Path to manage file names.
1149 - Removed unused functions:
1150 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1151 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1152 xbt_graph_export_graphviz()
1153 - xbt/str.h: xbt_str_join()
1154 - Remove unused datatypes:
1155 - xbt/heap.h: use std::priority_queue or boost::heap instead
1156 - xbt/swag.h: use boost::intrusive::list instead
1159 - New link_energy plugin for the consumption of the links.
1160 - All of the operations on files and storage contents have been
1161 packaged into a plugin (src/plugins/file_system). The current
1162 public interface can be found in
1163 include/simgrid/plugins/file_system.h
1164 To use these functions you now have to initialize the plugin by
1165 calling MSG_storage_file_system_init() just after calling
1166 MSG_init() or sg_storage_file_system_init() just after creating
1170 - Remove the undocumented/untested tag <include>
1173 - Remove viva specific tracing as the tool is no longer maintained
1176 - #248: Kill of finished processes leads to segfault
1177 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1178 - #239: please implement signal s4u::onDeadlock()
1179 - #230: segfaults when exit() without run()
1180 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1181 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1182 - #222: Actor::kill() doesn't really kill and segfaults
1183 - #221: odd LMM warning when killing an actor
1184 - #120: Memory leak when the processes are forcefully killed
1186 ----------------------------------------------------------------------------
1188 SimGrid (3.17) Released October 8 2017
1190 The Drained Leaks release: (almost) no known leaks despite the tests.
1192 Even very long simulations will preserve your memory: our testsuite
1193 stresses SimGrid for over 45mn on fast machines for a coverage of
1194 over 80%, and there is only one single known leak, of about 4kb.
1198 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1199 handler may be given to free resources if the comm cannot be completed.
1200 - this_actor::send(mailbox) is now mailbox->put()
1201 - New: simgrid::s4u::Comm::wait_all()
1202 - New: Host.execute() for remote executions.
1205 - Unused option network/sender-gap is removed.
1208 - Deprecate MSG_task_isend_with_matching(): This unused feature
1209 really complicates our internals. Will be removed in v3.20.
1212 - Improved context termination. It is now possible to free resources, even
1213 when a process is forcibly killed.
1214 - ContextBoost: add support for Boost versions above 1.61.
1217 - Bring back run-time option --cfg=exception/cutpath to remove exception
1219 - Removed unused functions:
1220 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1221 xbt_str_rtrim(), xbt_str_trim().
1222 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1225 - Removed header files obsolete since SimGrid 3.12:
1226 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1228 - #3: SD_exit should be made optional
1229 - #120: Memory leak when the processes are forcefully killed
1230 - #159: Threading test regression in Actor refcounting
1231 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1232 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1233 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1234 - #191: VM migration and pstate
1235 - #192: Updating the energy consumptions of all hosts crashes with VMs
1236 - #195: All actors have PID=0 in the logs
1237 - #204: Sometimes segfault with thread contexts and mmap privatization
1238 - #222: Actor::kill() doesn't really kill and segfaults
1239 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1241 ----------------------------------------------------------------------------
1243 SimGrid (3.16) Released June 22. 2017.
1245 The Blooming Spring Release: developments are budding.
1248 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1249 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1250 - Java JAR file should be finally fully working on Mac OSX too.
1252 XML platforms: Switch to platform v4.1 format.
1253 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1254 - <zone> can be used as a synonym for the now deprecated <as>
1255 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1256 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1257 - <actor> can be used as a synonym for the now deprecated <process>
1258 - state_file and avail_file periodicity is now easier to express
1259 (check the documentation) the old behavior should still work.
1260 * Storage: not backward compatible, but it seems that nobody ever used it
1261 - Remove attribute 'content_type' of <storage_type>: was never used
1262 - Make attribute 'model' of <storage_type> optional: for future usage
1263 - Remove Bconnection model property: was never (in)validated. Replaced by
1264 the maximum of the read and write bandwidth as a resource constraint.
1267 - New and Backwards Compatibility break:
1268 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1269 When one wants to get the list of tasks whose states have changed during a
1270 simulation round, s/he has to allocate and free a dynar and use it as
1271 argument to this function. The former SD_simulate (double how_long)
1275 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1276 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1279 - New function to update the consumption of all hosts at once.
1280 - Fix the model for multi-core hosts, linear in the amount of busy
1281 cores with an abnormality for pIdle (see paper).
1282 BUG: Still not satisfactory for ptask on mono-cores.
1285 - The netzone are now available from the MSG API.
1286 The old names still work, but are now deprecated.
1289 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1290 - New option: smpi/keep-temps to not cleanup temp files
1291 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1292 mallocs mode is used, allows to change the size of the fake file used
1293 (default 1MB), to potentially limit the number of mappings for large runs.
1294 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1295 - Fortran ifort and flang compilers support
1296 - New RMA calls supported (experimental):
1297 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1298 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1299 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1300 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1301 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1302 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1305 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1306 - Behind the scene, ActivityImpl are now automatically refcounted.
1307 Removing this nasty bottleneck will greatly simplify our internals.
1310 - Replay: New function xbt_replay_action_get():
1311 Retrieve the function previously associated to an event type.
1312 - DROPPED FUNCTION: xbt_str_varsubst()
1313 - DROPPED MODULE: strbuff. We don't need it anymore.
1314 - DROPPED MODULE: matrix. We don't need it anymore.
1315 - DROPPED MODULE: lib. We don't need it anymore.
1317 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1319 SimGrid (3.15) Released March 22. 2017
1321 The Spring Release: continuous integration servers become green
1322 We fixed even the transient bugs on all target architectures:
1323 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1324 Windows, FreeBSD, NetBSD.
1326 - Rename NetCards to NetPoints.
1327 This was intended to help NS3 users, but that's not a netcard.
1328 That's a point in the routing algorithm, let's avoid wrong simplifications.
1331 - New: MSG_process_yield(). Stop and yield to other processes.
1332 - New: MSG_process_daemon(). Daemon processes are automatically killed
1333 when the last non-daemon process terminates
1334 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1335 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1336 to make room for the upcoming network energy plugin.
1337 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1340 - Ensure that an actor can kill itself with Process::exit()
1341 - Kill the obscure NativeException. Nobody want to survive the issues
1342 it denotes, so use JniException that is a RuntimeException (not to
1343 be caught explicitly).
1344 - Partial bug fix in initialization. SimGrid flags on command line were
1345 consumed at C level but stayed in the original Java String[] args.
1346 This could mess users' args[i] if SG flags were not put at the end of
1348 The SimGrid flags are now removed from the Java arguments. However,
1349 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1350 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1351 It might be if you have --log or --cfg flags in the command line.
1352 - Fix numerous memleaks all around the place. In particular, around VMs.
1356 - s4u::onPlatformCreated: right before the simulation starts
1357 - s4u::onSimulationEnd: right after the main simulation loop
1358 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1359 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1360 event from the availability_file changes the avail speed.
1361 - Links are now usable from s4u
1362 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1363 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1364 - New examples: The conversion of MSG examples to S4U has begun
1365 - Actors: create, kill, migration, and suspend
1366 - Applications: master-worker and token-ring
1367 - Action replay: communications and storage
1368 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1371 - Backwards Compatibility breaks
1372 - The SD_route_get_list and SD_route_get_size functions have been removed.
1373 They are replaced by the unique
1374 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1375 The route from src to dst is built in the links dynar whose size can
1376 be obtained with xbt_dynar_length.
1377 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1378 sg_host_route_bandwidth and sg_host_route_latency.
1379 Macros ensure the backwards compatibility, but you should fix your code
1382 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1383 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1384 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1385 We don't support 100% of it, but it should be enough. Contact us if not.
1386 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1387 naive and probably inaccurate)
1388 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1389 mapped onto a unique small file using some system magic.
1390 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1391 - Fortran: cleanups, fixes, support of user-added operations
1392 - RMA: MPI_Accumulate are applied in correct order.
1393 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1394 - Better support for MPI_IN_PLACE
1395 - Support for MPI_Win attrs and keyvals.
1396 - Support MPI_Comm_group_create, MPI_Type_size_x
1397 - MPI ops should be properly applied to all allowed types
1398 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1399 algorithm by a non failing one from mpich.
1402 - Kill the fifo data container: we don't use it anymore.
1405 - A <cluster> can now be created with different speed values to represent pstates.
1406 see examples/platforms/energy_cluster.xml for an example.
1408 SimGrid (3.14.159) Released December 28. 2016
1410 The previous release was only a crude approximation of a Pi release;
1411 this one should be a bit better.
1413 - Revert a last minute change that broke on Mac OSX.
1414 - Fix the changelog and NEWS files.
1415 - Improve the documentation, in particular of the routing module.
1416 - Integrate some patches flying around in Debian.
1418 SimGrid (3.14) Released December 24. 2016
1421 * The whole documentation was reworked and reorganized. There is
1422 still much room for improvement here, but we're on it.
1425 * We now need python3 for our testing infrastructure.
1426 * Model checking needs libevent
1429 * Now works on FreeBSD
1432 * Privatization now works on FreeBSD
1434 * Privatization is now activated by default in smpirun
1435 You can turn it off with -no-privatize if something goes wrong.
1437 * Call-location tracing for SMPI.
1438 You can add the exact location (filename / linenumber) of an MPI call to
1439 your trace files and slow down or speed up the simulation between two
1440 consecutive calls by using an adjustment file (see the documentation).
1442 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1443 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1444 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1445 values were ignored in some cases. The timings of these functions can now
1446 be significantly different.
1448 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1449 smpi/running-power is renamed to smpi/host-speed
1451 * smpi/grow-injected-times option to enable or disable multiplication of the
1452 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1453 make simulation less precise (but also much faster).
1455 * smpirun script should be (much) faster for large deployments.
1457 * SMPI tracing: fixed issue with poor matching of send/receives.
1459 * Replay: Fix broken waitall
1461 New functions and features
1462 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1464 Dropped / renamed functions and features
1465 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1466 - MSG_mailbox_is_empty() -> Mailbox::empty()
1467 - MSG_mailbox_front() -> Mailbox::front()
1468 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1469 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1470 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1471 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1472 * Don't pass the free_f parameter to property related functions:
1473 - MSG_host_set_property_value()
1474 - MSG_as_router_set_property_value()
1475 - MSG_storage_set_property_value()
1476 * VM properties. Since msg_vm_t are msg_host_t, just use
1477 MSG_host_get_property_value() and friends
1478 * VM I/O related things:
1479 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1480 - Unimplemented save/restore methods
1481 * MSG_as_router_get_property_value() was redundant with
1482 MSG_environment_as_get_property_value().
1483 - Removed MSG_as_router_*propert*() functions
1484 - Added MSG_environment_as_set_property_value() for consistency
1485 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1486 Well, they are still there for now, but deprecated with a warning.
1487 Please switch to xbt_dict_new_homogeneous() before this is removed
1489 * Task affinity. Its intended behavior (that was very badly tested
1490 and probably not really working) was deceiving what most users
1491 would have hoped here.
1492 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1493 * Option network/coordinates is now useless and should be dropped.
1496 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1499 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1500 * Added Dragonfly topology support
1503 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1504 EiBps, EBps, Eibps, Ebps for bandwidth.
1505 They may become useful to some lucky ones.
1508 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1509 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1510 * ex/async_waitAll: new example, on asynchronous communications
1513 * Memory usage should be decreased for simulations with a large number
1514 of processes. This also helps for SMPI.
1516 SimGrid (3.13) stable; urgency=low
1518 The Half Release, a.k.a. the Zealous Easter Trim.
1520 - We removed half of the lines, that were mostly experimental cruft.
1521 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1522 (not counting blanks and comments -- according to openhub.net)
1523 - The internals are now compiled in C++ (and will soon be clean C++)
1524 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1525 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1527 * Backwards Compatibility breaks
1528 - Removed Lua simulation bindings (switch to C or Java for that).
1529 Lua can still be used to describe platforms
1530 - Removed Java kernel plug-ins.
1531 Will be reintroduced after the ongoing major internals reorg.
1533 - the following functions were removed.
1534 They were too specific and should be reimplemented in a generic
1535 way, with filter function.
1536 - MSG_task_listen_from_host
1537 - MSG_mailbox_get_count_host_waiting_tasks
1538 - MSG_mailbox_put_with_timeout was removed.
1539 Please use MSG_task_send_with_timeout instead.
1541 - the SD_application_reinit function was removed. It has been a noop for a while.
1542 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1543 replaced by a more flexible mechanism.
1544 - The following functions thus do not exist anymore
1545 - SD_workstation_get_access_mode
1546 - SD_workstation_set_access_mode
1547 - SD_workstation_get_current_task
1548 - Basic estimation functions have been removed but can easily be replaced
1549 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1550 - SD_workstation_get_computation_time => amount / sg_host_speed()
1552 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1553 Use VM.getSpeed()*load/100 for the legacy behavior.
1555 - option enable_tracing was removed. It was not doing anything for a while.
1556 - In the ModelChecker:
1557 - the model-checker now ptraces the model-checked process which means
1558 you cannot use a debugger on the latter anymore (we might make this
1559 optional in the feature);
1560 - removed soft-dirty page tracking;
1561 - remove model-checked side snapshot management,
1562 MC_snapshot() and MC_compare_snapshot();
1563 - keep the MC_cut() function as a stub (it was not really working
1564 in the previous release).
1567 * All options are consistently in kebab-case. Old names are kept as alias.
1570 * Switch to platform v4 format.
1571 - Rename from 'power' to 'speed' the attributes describing the amount of
1572 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1573 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1574 - In <host> and <link>, attributes availability and state are gone.
1575 It was redundant with state and availability traces, and with peak values.
1576 - In <cluster>, attributes availability_file and state_file are gone.
1577 It was too complex and unused.
1578 - Kill <gpu>. Was not doing anything.
1579 - The DOCTYPE points to the right URL:
1580 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1581 (the file at this address now documents the changelog since its v1)
1582 - A warning is emitted for unit-less values (they are still accepted).
1583 - speed. Default: 'f' or 'flops'. Also defined:
1584 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1585 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1586 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1587 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1588 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1589 - latency. Default: 's' second. Also defined:
1590 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1592 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1593 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1596 * s4u::Host is now the preferred public interface to the Host features.
1597 sg_host_* functions are C bindings to the exact same behavior
1598 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1601 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1602 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1603 They were deprecated since a few years, and probably did not even compile anymore.
1606 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1607 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1608 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1609 encouraged to update their codes. A list of the modified functions can be found at the end of
1610 include/simgrid/simdag.h
1613 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1614 the lambda or closure passed as a parameter will run in kernel mode.
1615 All the callback functions should be rewritten to that interface at some point.
1618 * Reorganizing and cleaning the internals all around the place.
1621 * Remove old default barrier/bcast buggy algorithms (see #18407)
1622 * Various bug fixes to handle more codes
1623 * Remove the need for the --foreground option of smpirun (it is still
1624 accepted for backward compatibility).
1627 * Kill the setset data container: MC don't use it anymore.
1628 * Kill the queue data container: it made more sense with GRAS.
1629 * Kill the xbt_peer_t data type: it's useless without GRAS.
1630 * Kill rm_cb feature of config sets: it was never useful.
1631 * Kill graphxml parsing feature. It was not used.
1632 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1634 - xbt_dynar_sort_strings(), when the content is char*
1635 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1636 They throw exceptions on invalid input;
1637 * C++ support for declaring CLI flags (simgrid::config::Flag);
1638 * class for abstracting different signal backends (simgrid::xbt::signal).
1639 with no external dependencies (we need very simple signals).
1642 * refactoring and cleanup of the code;
1643 * ongoing process to cleanly separate the model-checking algorithms
1644 from the code model-checking support.
1646 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1648 SimGrid (3.12) stable; urgency=low
1650 The Facelift Release.
1653 * Require g++ v4.7 at least to not speak prehistorical C++.
1654 * Require Boost 1.48 (for signal2 component).
1655 * Java must be version 7 at least when activated.
1656 * Builds on Windows again (including Java bindings).
1657 * Tracing is now always enabled (no way to turn it out)
1658 * Remove GTNetS. It was not working anyway.
1659 * Various cleanups in the cmake scripts.
1660 * Move headers around to sort them out on installed systems:
1661 - instr/instr.h -> simgrid/instr.h
1662 - instr/jedule/* -> simgrid/jedule
1663 - simdag/datatypes.h was removed
1664 - simdag/simdag.h -> simgrid/simdag.h
1665 - msg/datatypes.h was removed
1666 - msg/msg.h -> simgrid/msg.h
1669 * Interface improvement:
1670 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1671 - Sanitize the interface in MSG_task_ module:
1672 - Merge two functions that were close enough but misleading:
1673 set_compute_duration(t) -> set_flops_amount(t)
1674 get_remaining_computation(t) -> get_flops_amount(t)
1675 - set_data_size(t) -> set_bytes_amount(t)
1676 get_data_size(t) -> get_bytes_amount(t)
1677 - Massive cleanups in the functions related to the energy
1678 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1679 - New: MSG_host_get_pstate()
1681 - msg/energy/onoff: switching hosts on and off
1683 * Interface improvement:
1684 - Rename Host.isAvail() to Host.isOn()
1685 - Rename Process.currentProcess() to Process.getCurrentProcess()
1686 - Rename Task.setDataSize() to Task.setBytesAmount()
1687 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1689 - #18874: Actually allows the GC to reclaim tasks
1693 - SIMIX_process_throw: raises an exception in a remote process
1694 * Refactoring: Separate sub-modules
1695 - libsmx: the public interface, as libc in a real system
1696 - popping: the strange dance that converts a user request into a kernel handling
1697 - smx_context_*: the virtualization mechanisms that embed the user code
1698 - smx_*: the handling of each simcalls
1699 * Interface cleanup:
1700 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1701 * Rename smx_action_t into smx_synchro_t, making explicit that these
1702 things are used to synchronize processes with their environment.
1703 For example, a communication is a sort of synchronization involving
1704 the communicating processes (that may block until the exchange) and
1705 the platform. The same can be said from computations, etc.
1707 - #18888: segfault when a process ends before its kill_time
1711 - 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.
1712 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1713 - MPI_Type_set_name, MPI_Type_get_name
1714 - MPI_*_c2f and MPI_*_f2c functions
1715 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1716 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1717 - Activate a lot of new tests from the mpich 3 testsuite
1719 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1720 - InfiniBand network model added: Based on the works of Jerome Vienne
1721 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1722 - When smpi/display_timing is set, also display global simulation time and application times
1723 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1724 * Collective communications
1725 - 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
1726 - 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/).
1727 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1729 - Replay now uses algorithms from wanted collective selector
1730 - Replay can be used with SMP-aware algorithms
1731 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1732 - 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.
1734 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1735 - Correctly use loopback on fat-tree clusters
1736 - Asynchronous small messages shouldn't trigger deadlocks anymore
1737 * Energy/DVFS cleanup and improvement
1738 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1739 - new: smpi_get_host_pstate()
1742 - "Full" network optimization flag was broken since Surf++
1743 - Better handling of precision flags in maxmin
1744 - Fix bug causing sometimes "Impossible" errors
1745 - Properly pass cluster properties to included hosts
1746 * Improvement of the Energy plugin.
1747 - Always update the consumption before returning that value
1748 - New property: watt_off to denote the dissipation when the host is off
1749 - New functions getWattMinAt and getWattMaxAt to retrieve the
1750 dissipation of pstates that we are not currently at.
1751 * Java: class NetworkLink renamed to Link
1752 * New function: simcall_process_get_kill_time()
1753 * Massive rename s/workstation/host/
1754 - That's intrusive, but that's good for the project consistency. Sorry.
1755 - Change config option "workstation/model" into "host/model"
1758 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1759 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1760 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1761 - xbt_dirname and xbt_basename for non-POSIX systems
1763 * The model checker now runs as a separate process.
1764 * The model checker runs must now be launched with the new simgrid-mc program.
1765 * Record/Replay: the MC can display a textual representation of a path in the
1766 execution graph. It can then be replayed outside of the model checker.
1768 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1770 SimGrid (3.11) stable; urgency=low
1775 * Normalizing pointers addresses tool for better diff between logs
1778 * Add cloud examples using new VMs
1779 - examples/msg/cloud/two_tasks_vm.tesh
1780 - examples/msg/cloud/simple_vm.tesh
1781 - examples/java/cloud/cloud.tesh
1782 - examples/java/cloud/migration/migration.tesh
1783 * Add java surf examples:
1784 - examples/java/surfPlugin/surf_plugin.tesh
1785 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1786 - examples/java/surfCpuModel/surf_cpu_model.tesh
1787 * Add SMPI+MSG example:
1788 - examples/smpi/smpi_msg_masterslave/
1793 - msg task destroy cancel test
1794 - msg_host on/off test
1795 * Move all tests in testsuite to teshsuite (adding tesh files)
1796 * Restructure teshsuites
1797 - one folder for each kind of test
1798 * Restructure AddTests.cmake
1800 - structure the order of tests (with sections)
1803 * Add virtual machine
1804 - creation of a VM on a PM
1805 - migration of a VM from a PM to another PM
1807 - MSG_process_join(msg_process_t process, double timeout)
1808 - msg_bar_t MSG_barrier_init(unsigned int count)
1809 - int MSG_barrier_wait(msg_bar_t barrier)
1810 - void MSG_barrier_destroy(msg_bar_t barrier)
1811 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1812 * New option "msg/debug_multiple_use" to help debugging when a task is used
1815 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1817 - Increase code coverage in test suites
1821 * Protect context stacks against stack overflow. The number of protected memory
1822 pages allocated on the top of each stack (1 by default) can be configured
1823 with the new command line option --cfg=contexts/guard_size:<value>.
1824 * Simcalls are now generated by a python script that
1825 - generates files included by SimGrid
1826 - checks that all the functions exist, and proposes prototypes
1828 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1829 host_set_data, host_get_data
1831 - simcall_process_join(smx_process_t process, double timeout)
1832 * Fix bug where sleeping processing could not be suspended.
1835 * Translate surf models from C to C++
1836 - Generic classes for all models: Model, Resource, Action
1837 - A generic interface for each kind of model (CPU, Network, Storage
1838 Workstation, WorkstationVM)
1840 * Translate surf routings from C to C++
1841 * Add callbacks using sigc++ or boost::signals2
1842 - Add callback functions for resource creation/destruction
1843 - Add callback functions for action state change
1844 - Handle Energy as a plugin
1845 * Replace swag by boost::intrusive
1846 * Add new routing models for clusters. For documentation, see
1847 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1848 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1849 parameters for cluster tag
1850 - Fat trees, with topology="FAT_TREE" and
1851 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1852 - see examples/platforms/torus_cluster.xml and
1853 examples/platforms/fat_tree_cluster.xml
1854 * More documentation
1857 * Hostfiles support host:nb_processes construct to deploy several processes on
1859 * Collective communication algorithms should not crash if used with
1860 improper number of nodes and report the error.
1861 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1862 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1863 MPI_Dims_create, MPI_Cart_sub are supported.
1864 * New interface to use SMPI programmatically (still depends on MSG for
1865 some parts, see examples/smpi/smpi_msg_masterslave):
1866 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1870 * Global variables privatization in MPI executables is now performed at runtime
1871 with the option smpi/privatize_global_variables (default:no).
1872 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1873 dynamic libraries loaded by the application are not privatized (static
1874 linking with these libraries is advised in this case)
1877 * Options defined in XML work correctly now.
1880 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1881 copied into simgrid.jar or not (ON by default). Use this option if you want
1882 to reduce the size of the installed simgrid.jar, *and* the native libraries
1883 are kept installed elsewhere.
1884 * Surf binding with SWIG (code generated in maintainer mode only):
1885 - plugin to handle callbacks
1886 - CPU model only for the moment
1889 * Supernovae build mode is definitively removed. It was used to improve
1890 inlining and inter-module optimizations. It is nowadays superseded by
1891 link-time optimizations commonly available in compilers.
1892 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1893 * Add boost dependency for surf++
1894 * Add new macro for tests
1895 - ADD_TESH(name <tesh_args>)
1896 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1900 - xbt_bar_t XBT_barrier_init(unsigned int count)
1901 - int XBT_barrier_wait(xbt_bar_t barrier)
1902 - void XBT_barrier_destroy(xbt_bar_t barrier)
1903 * Make the xbt_os_time module public
1905 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1907 SimGrid (3.10) stable; urgency=low
1909 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1912 * Reintegrate Java to the main archive as desynchronizing these
1913 package is not acceptable anymore (Java is now considered stable)
1914 * Add explicit synchronization facilities through semaphores
1915 * Bug fix: Task.setDataSize() only changed the C world, not the value
1916 cached in the Java world
1919 * Dramatically change the way files are handled. API and internals changed, but
1920 this part of MSG was not considered as production grade either.
1921 * Add explicit synchronization facilities through semaphores
1922 * Add a new function MSG_host_get_process_list()
1923 * Preliminary DVFS support (see examples/msg/energy/ for details)
1926 * SMPI is now included directly in the libsimgrid as the windows
1927 linker doesn't force us on splitting it anymore.
1928 * Improvements of the SMPI replay tool:
1929 - Most of the collective communications are now rooted in the same process as
1930 in the original application.
1931 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1932 used until now). Multiple data types can now be used in a trace.
1933 - The replay tool now supports traces produce either by TAU or a modified
1935 - Bug Fix: the compute part of the reduce action is now taken into account.
1936 - Gatherv collective is now supported
1937 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1938 of smpirun outputs time independent traces for the current run. One file
1939 is created per process. If too many processes are simulated, this behavior
1940 can be changed to one file for all processes by using the
1941 tracing/smpi/format/ti_one_file flag
1942 * smpirun generates the host file if needed (with given host count and platform)
1943 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1944 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1945 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1946 --cfg=smpi/coll_selector:(mpich/ompi)
1947 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1949 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1950 happening inside a collective SMPI call.
1951 * Fix the behavior of complex data types handling.
1952 * Make MPI_Wtime another synchronization point to take computations into
1954 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1955 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1956 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1957 (no privatization of global variables yet)
1958 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1961 * Verification of liveness properties is now available for SMPI applications
1962 (in addition to MSG applications)
1963 * Bugged examples using SMPI in examples/smpi/mc/
1964 * Add --cfg=model-check/visited option. Allows the verification of infinite
1965 programs. Detection of loops in the execution thanks to the system state
1966 comparison and reduction of the state space to explore. Can be combined with
1967 DPOR for safety properties.
1970 * Allow to change SimGrid configuration (see --help) within the code
1971 thanks to SD_config() as it can be done in MSG.
1972 * Add a new function SD_task_set_amount() upon user request.
1975 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1976 * Remove rule based routing (no more PCRE dependency)
1977 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1978 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1979 * Add a loopback_bw and loopback_lat options to cluster tag.
1980 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1981 routing. Now peers are encapsulated in an AS and have their own private
1982 router but this is transparent.
1985 * Our own implementation of getline is renamed xbt_getline, and gets
1986 used even if the OS provide a getline(). This should reduce the
1987 configuration complexity by using the same code on all platforms.
1988 * New type: xbt_cfg_elm_boolean.
1989 * Allow to use yes/no for boolean configuration options in the command line.
1990 * Allow to disable SimGrid cleanups at exit from command line option.
1991 There are situations where one may want a simulation to end with an exit.
1992 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1993 annoying when scripting around the simulator. Adding a
1994 --cfg=clean_atexit:no allows to circumvent this issue.
1997 * Lots of memory leaks were corrected in this release.
1998 There is no known memory leaks anymore, in all of our 600+ tests.
1999 * New command line option --version, to get SimGrid version information.
2000 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
2002 * Supernovae builds are deprecated, and expected to be removed in the next
2005 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2007 SimGrid (3.9) stable; urgency=low
2009 The Grasgory release: GRAS is really dead now.
2011 * Complete overhaul of the internal host structures scheme.
2014 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
2015 was considered as experimental and badly maintained since then.
2016 * Keeping it was thus a trap to our potential users, that could take
2017 it instead of MSG or SMPI by mistake despite is pity state.
2018 * GRAS seems to have very few users (if any), and no one volunteered
2019 to maintain it further. It also induces a lot of XBT code (for
2020 portability sake), that must be maintained too.
2021 * For all these reasons, we killed GRAS. If someone wants to revive it
2022 in the future, don't cry, our git history still remembers of GRAS.
2025 * Major overhaul. Merge our documentation again as time proved that
2026 splitting it was really not helping our users.
2027 * Further improve the developer documentation to help newcomers
2028 hacking on SimGrid itself. The user documentation (and in
2029 particular, the beginner documentation) is still in a sorry state.
2032 * Now works on Windows too!
2033 * Much more extensive test suite, from MPICH
2036 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
2037 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
2038 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
2039 * Bug fix: let task be scheduled when the last dependency to be solved is
2040 a control dependency.
2041 * Remove SD_load_environment_script function.
2042 Use the C sg_platf function if you want to declare a platform
2046 * New function: MSG_process_get_number()
2047 * Old function documented: MSG_config()
2048 * Remove MSG_load_platform_script function
2049 Use the C sg_platf function if you want to declare a platform
2053 * Change the default value of the TCP_gamma constant (maximal size of TCP
2054 congestion window) to a more realistic 4MiB value. If you notice changes in
2055 your simulation results, you can fall back to the previous 20k tiny window
2056 by adding --cfg=network/TCP_gamma:20000 on command line.
2057 * (Hopefully) fix a bug wrt periodic availability and state traces
2058 * Bug fix: use default values at start when first event in availability/state
2059 trace is not at time 0.
2062 * remove the "new_" part of function name sg_platf_new_trace_connect
2063 (resulting in sg_platf_trace_connect), since it does not create
2067 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2068 use case where it's really mandatory, and maintaining it was a pain in
2070 * New: xbt_fifo_search(), search an item with a user-provided
2071 comparison function instead of dumb pointer comparison.
2074 * Fix the lua deployment:
2075 Use `simgrid.init_application()` before deployment instead of
2076 `simgrid.msg_register_application()` after.
2079 * Transfer the tracing files into the corresponding modules.
2081 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2083 SimGrid (3.8.1) stable; urgency=low
2085 The "we are told that some people want to also *install* the simgrid
2088 * Add missing file "tesh.1" to the archive.
2090 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2092 SimGrid (3.8) stable; urgency=low
2094 The Psssshiiiit release: SimGrid jumps into the Cloud.
2097 * Add an experimental interface to manipulate VMs. They are mainly
2098 process groups with very few intrinsic semantic, but they should
2099 allow you to build the semantic you want easily.
2100 * New function: MSG_host_set_property_value()
2101 * New function: MSG_process_on_exit(). To clean memory in all cases.
2102 * Bug fixes that made the host (and link) failures unusable.
2103 * Add a way to auto-restart process when the host in which they are
2104 executing comes back (ON_FAILURE="RESTART" on deployment file,
2105 MSG_process_auto_restart_set).
2106 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2107 please stop using the old ones, they are DEPRECATED.
2109 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2110 Please use MSG_init() instead. (reducing the amount of entry
2111 points in the library helps us).
2112 * Make it impossible to link against the wrong version of the lib
2113 * Deprecate MSG_clean(). No need to call it anymore.
2114 * Function MSG_get_host_number() is not deprecated anymore.
2117 * Split the doc into a user guide and a reference guide.
2118 * Start a developper guide to help people hacking on SimGrid.
2121 * Enable tracing by default. This modules rocks you should use it.
2122 * Remove option custom_flags. Now use environment variables CFLAGS
2124 * Use default cmake things to detect lua instead of home grown ones.
2125 * New option "enable_mallocators" to disable mallocators, for debugging
2126 purpose ("on" by default).
2129 * Bug fixes around the resource failures: don't let the processes
2130 survive the host they are running onto.
2131 * Add an interface to auto-restart processes when the host in which they are
2132 executing comes back.
2133 * Ensures that SIMIX_clean is called automatically. It's not part of
2134 the public interface anymore (bindings should be updated).
2137 * Bug fix for when SD_Simulate is called with a positive value: be careful
2138 when comparing doubles. Sometimes they are different for non significant
2140 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2141 parallel task whose initial work is distributed among host according
2142 to the Amdahl's law. Such tasks are created with a parameter alpha
2143 that corresponds to the non-parallelizable part of the computation.
2144 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2145 between two sets of workstations assuming a 1D block distribution (each
2146 workstation owns a similar share of data) on both sides.
2148 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2149 Data redistribution will be automatically scheduled once parent and child
2150 are both scheduled. The filling of computation_amount and
2151 communication_amount structures is now done seamlessly thanks to the chosen
2153 * New function SD_workstation_dump to display various information
2154 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2155 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2156 bandwidth on the route onto which the task is scheduled and the amount of
2158 To divide the nominal bandwidth by 2, the rate then has to be:
2159 rate = bandwidth/(2*amount)
2160 * Compute tasks that have failed can now be rescheduled and executed again
2161 (from their beginning)
2162 * Increasing source code coverage (src/simdag is now covered at 95.8%
2166 * Re-implement time-independent trace replay using SMPI (at the
2167 smpi_smp_* level) instead of MSG. This should replace
2168 examples/msg/actions/actions.c
2169 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2170 hindexed and structs)
2171 * Implement the exchange of non-contiguous data.
2172 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2173 * Correct behavior of smpi/sender_gap and set its default value to 0
2174 * Add option to asynchronously send small messages to allow better
2175 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2176 specifies the size in bytes under which messages will be asynchronously sent.
2177 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2178 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2181 * New C interface to define a platform: XML is now optional.
2182 For more info, please check include/simgrid/platf.h.
2183 * New interface to define random platforms from the C:
2184 For more info, please check include/simgrid/platf_generator.h and
2185 examples/msg/masterslave/masterslave_platfgen.c
2186 * Export a sg_cmdline dynar containing all the arguments we got from
2190 * Two new tracing options for adding comments to trace file so you
2191 can track your experiments (see --help-tracing for details).
2192 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2193 * Adding the SimGrid version that generated the trace file as a comment.
2194 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2195 * Fix to avoid key clashes on Paje links
2196 * Other minor fixes related to the Paje specification
2199 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2200 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2201 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2202 * Greatly improve the robustness of mmalloc to user errors (such as
2203 using an area after freeing it, or freeing it twice)
2205 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2207 SimGrid-java (3.8.1) stable; urgency=low
2209 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2210 Warning: all methods are visible, but only some of them are
2211 implemented so far. Check the source (src/jtrace.c)
2212 for further information.
2213 * New module: org.simgrid.msg.File (SimGrid File management functions)
2214 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2215 * Change the meaning of Process.restart: now restart the process from
2216 the beginning, like MSG_process_restart in C.
2217 * Add Process.setAutoRestart: handling of process restart when failed
2219 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2220 you to retrieve the properties of the processes/hosts
2221 * Deprecate Msg.clean(): you can just forget about it now.
2222 * New function Process.getCount(), that only works when compiling
2223 with the not yet released version 3.9 of the C library.
2225 * New context factory based on Coroutines. It mandates a modified JVM
2226 but then, the simulations run about five times faster, and there is
2227 no limit to the amount of processes (beside of the available memory).
2229 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2231 SimGrid (3.7.1) stable; urgency=low
2234 * Restore the prototype of MSG_process_create_with_environment() to
2235 the pre-3.7 situation by removing the kill_time argument.
2236 * Add a MSG_process_set_kill_time() function instead.
2239 * Fix weird behaviors when dealing with parallel tasks.
2242 * Simgrid is now built as a dll.
2243 * Simgrid-java now works on Windows.
2244 * Simgrid-Java is now included into Windows package.
2247 * First pre-build package for MacOSX.
2250 * Fix compilation when using MSG_USE_DEPRECATED.
2251 * Fix some compilation issues on Macs and Windows.
2252 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2253 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2255 We discovered that the Lua console is broken, but we are missing the
2256 manpower to fix it right now. The problem existed in 3.7 too, so we
2257 are not blocking the release for that. Sorry if you depended on this
2258 feature, any help would be really welcome.
2260 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2263 SimGrid-java (3.7.1) stable; urgency=low
2265 The "Java aint got to be bloated and slow" release
2268 * Various internal cleanups and performance improvement
2269 Simulations are expected to run up to twice faster or so
2270 * Make Process.kill(process) an instance method, not a static one
2271 * User processes are not java.lang.Thread subclasses.
2272 This breaks the compatibility (sorry), but previous API was
2273 brain-dead, making it impossible to have non-trivial
2274 initializations in the process constructor.
2275 * Require a full constructor per Process sub-class.
2276 Kinda breaks the compatibility (sorry), but this allows a much more
2277 efficient way to launch the processes at simulation startup.
2278 * Do not embeed our version of semaphores, java 1.5 can be considered
2279 as sufficiently prevalent for us to not dupplicate its features.
2283 * Add examples for almost every part of the API
2284 We spotted and fixed a lot of bugs in the process
2286 * New module: asynchronous communication API
2287 * New function: Process.sleep()
2288 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2289 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2290 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2292 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2294 SimGrid (3.7) stable; urgency=low
2296 The "spring cleanups (before the next Big Project kicks in)" release.
2299 * We can specify the SMPI latency/bandwidth factor with command line
2300 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2301 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2302 You can also use the "config tag" from platform file by adding this line
2303 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2304 (see "example/platforms/tag_config.xml" to use "config tag").
2305 Note that the command line supersedes the platform file configuration.
2306 * Change the correction factors used in LMM model, according to
2307 the latest experiments described in INRIA RR-7821.
2308 Accuracy should be improved this way.
2309 * Use the partial invalidation optimization by default for the
2310 network too. Should produce the exact same results, only faster.
2311 * Major cleanup in surf to merge models and split some optimization
2312 mechanisms from the core of the models. As a result you can now
2313 specify which model to use (e.g., --cfg=network/model:LV08
2314 --cfg=cpu/model:Cas01) and which optimization mode to use
2315 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2316 Incompatible combinations should err at initialization. See
2317 --help-models for the list of all models and optimization modes.
2318 * The CLM03 workstation model was dropped for simplicity because it
2319 used the deprecated CM02 network model. Use default instead.
2320 * Rename the TCP_gamma configuration option to network/TCP_gamma
2321 * Rename the coordinates configuration option to
2322 network/coordinates, and document it
2323 * Use now crosstraffic keyword instead of the terribly misleading
2324 fullduplex keyword. It is activated by default now in the current
2325 default model, use --cfg=network/crosstraffic:0 to turn it off.
2326 * Ongoing refactoring the model parsing to make XML files optional
2327 See include/simgrid/platf.h for details (still to be completed)
2330 * Major overhaul of the documentation. Almost instructive now :/
2331 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2332 functions and friends. This interface was considered as
2333 deprecated since over 2 years, it's time to inform our users that it is.
2334 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2335 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2336 use these (crappy) functions in your code.
2337 These functions will be removed soon. Stop using them now.
2338 * Deprecate MSG_get_host_{table,number}
2339 Implement MSG_hosts_as_dynar() instead.
2340 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2341 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2344 * Stabilize the parallel execution mode of user contexts
2345 * Introduce configuration variables to control parallel execution:
2346 - contexts/synchro: Synchronization mode to use when running
2347 contexts in parallel (either futex, posix or busy_wait)
2348 - contexts/parallel_threshold: Minimal number of user contexts
2349 that must be part of a scheduling round to switch to parallel
2350 execution mode (raw contexts only)
2351 * Fix bugs that prevented to use suspend/resume along with
2352 synchronization structures.
2353 * Fix bugs in process termination that lead to invalid memory access
2354 in very specific conditions.
2357 * Introduce a parallel mode for the models (controlled by surf/nthreads
2358 configuration item). In our tests, running the models in parallel
2359 never lead to any speedups because they are so fast that the gain
2360 of computing each model in parallel does not amortizes the
2361 synchronization costs, even when ultra fast futexes are used.
2362 This is released anyway because YMMV.
2365 * Performance boost by using a swag internally to compute the set of
2366 tasks that are finished and should constitute the return value of
2370 * Enable it by default now that it is considered rather stable.
2373 * Documentation of the tracing functions.
2374 * Performance gains when tracing categorized/uncategorized resource
2375 utilization by avoiding calls to get route when updating resource
2376 variables. LMM constraints are being used instead.
2377 * API changed to set task categories. Use MSG_task_set_category instead
2378 of TRACE_msg_set_task_category, and SD_task_set_category instead
2379 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2380 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2381 * Examples for link user variables added (see at examples/msg/tracing/)
2382 * Deprecated function TRACE_msg_set_process_category completely removed
2383 * Trace header updated according to the latest Paje file format
2384 * Tracing network lazy updates, no longer obligate users to use full updates
2385 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2386 * Experimental: let user code declare/set/push/pop application states for hosts
2387 * API changed to allow the manual creation of graph configuration files
2388 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2391 * Improve the API of Lua MSG bindings, using the Lua spirit.
2392 * Each simulated process now lives in its own Lua world (globals are
2393 automatically duplicated). It helps writing simulators. Will allow
2394 to run Splay programs within SimGrid in the future.
2395 * Add a Chord example in Lua, equivalent to the MSG one.
2398 * Start the implementation of a solution to express temporal
2399 properties, not only local assertions. This is still an
2400 experimental work in progress, stay clear from it to be safe.
2404 - Add new runtime parameters --help-logs and --help-log-categories
2405 to display information about supported logging parameters and
2407 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2408 don't exists anymore.
2409 * Mallocators: allow value NULL for the reset function.
2411 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2412 create homogeneous dictionaries, where all the elements share the
2413 same free function. Non homogeneous dictionaries will be
2414 deprecated in the next release.
2415 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2416 - Multi-level dictionaries are deprecated.
2418 - new function xbt_dynar_search_or_negative() that is useful when
2419 you have less than 2 million elements in your dynar and don't
2420 want of the extra complexity of catching exceptions when the
2421 element is not found.
2423 - Make xbt_os_thread module (for thread portability) public.
2424 Documentation is still to come, sorry.
2426 - Cleanups and simplifications to make it maintainable again.
2427 - Exotic features (such as memalign and valloc) were removed.
2428 - The metadata were extended and improved so that the
2429 model-checker becomes able to explore and inspect the heaps.
2430 - This may induce a performance drop when enable_model-checking is
2431 ON in cmake (even if it's not used in the simulation), but it is
2432 necessary at this point to get MC working.
2434 Turn model-checking OFF if simulation performance matters to you.
2435 Not enabling it at runtime is not enough, disable it in cmake.
2437 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2440 SimGrid (3.6.2) stable; urgency=low
2442 The "Not coding new stuff allows to polish old things" release.
2445 * New bindings to the NS3 packet level simulator (experimental)
2446 * Use the raw (efficient) execution contexts instead of the sysv
2447 (portable) ones when possible.
2448 * libpcre is now mandatory in any cases since not using it led to
2449 severe performance loss and possibly other issues
2450 * Update the XML platforms:
2451 - G5K: include the latest machine in Nancy
2452 - GridPP and LCG: new platforms
2453 * Documentation was partially updated, at least (more to come)
2455 Bug fixes, cosmetics and small improvements
2456 * Free terminated processes before the end of the simulation to avoid
2457 exhausting the memory of users having very dynamic amount of
2459 * Bug fix and cosmetics about canceling non-running tasks
2460 * Bug fix about the dot loader's issues when using libcgraph
2463 * Create an installer for windows with nsis (amd64 and win32)
2464 - Add an hello world project to illustrate simgrid project creation.
2465 - Embed libpcre into the Simgrid installer to avoid
2466 its compilation burden
2467 * The raw execution contexts should work on Apple now
2468 * Port to Windows 64 bits
2469 - Sysv contexts now have an implementation for this arch
2470 - GRAS communication features now support this arch
2471 * Drop support for borland compiler on windows
2472 - this code was not maintained, and we kinda depend on gcc nowadays
2473 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2474 * Fix portability issue on unstable ubuntu: linker became picky on
2477 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2479 SimGrid (3.6.1) stable; urgency=low
2481 The "Oops, we broke Macs too" release
2484 * Fixed contexts detection so that raw ones are used when possible
2485 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2486 leads to a strange error, with user code segfaulting sometimes when
2487 the generated code is not perfectly aligned (which is not
2488 controllable from the user side, depends on the amount of code)
2491 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2493 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2495 SimGrid-java (3.6) unstable; urgency=low
2498 * Split of every thing from simgrid v3.5 into a separate package.
2500 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2502 SimGrid (3.6) stable; urgency=medium
2504 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2507 * Bindings now constitute their own package, separated from the main one.
2508 Rationale: reduce our maintenance nightmare by reducing the module coupling
2509 They will soon be released on their own on gforge.
2511 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2512 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2514 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2515 * It's not quite deprecated for now because we have no replacement,
2516 but it may soon become the case.
2519 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2520 * Fortran: New user-level cache variable to store the rank of the running
2521 process. This improves performance by an order of magnitude.
2522 * C: New coccinelle script to automatically locate and modify global and
2523 local static variables.
2524 * Improved SMPI network model with a sender-side gap to account for multiple
2528 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2529 only say if a communication is finished, no matter whether it succeeded or
2530 failed. You can call MSG_comm_get_status() to know the status of a finished
2532 * New function MSG_task_dsend() to send a task and detach it. When a
2533 communication is detached, you are never notified of its success or failure
2534 and the memory is released automatically once it is finished. This function
2535 is useful when you don't care about the end nor the success of a
2537 * Change the prototypes of action replay. Sorry for inconvenience,
2538 but this is really more efficient this way (and to adapt your code,
2539 you just have to fix the initialization, that shouldn't be too long)
2540 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2541 ever managed to do anything useful with it.
2542 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2543 takes a dynar of communications. It returns immediately and gives the
2544 index of a finished communication (if any).
2545 * New example: a basic implementation of the Chord P2P algorithm.
2548 * New model for multi-core CPUs. You can now use the core attribute to
2549 precise the number of cores of a host. This is a basic model. Every
2550 process running on the host receives at most the power provided in
2551 the DTD (throughput<=power). Total throughput of process cannot exceed
2553 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2554 router linked by an up-link and a down-link (possibly asymmetrical).
2555 This kind of pattern allows to easily build last-mile model style platforms.
2556 Aggregating such patterns in a rule-based AS is thus the technique of
2557 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2558 * New model for Vivaldi routing. We transformed the Vivaldi network model
2559 into a Vivaldi routing model (based on the rule-based model). This allows to
2560 combine Vivaldi based latencies with last-mile platforms.
2563 * Added a check for NaN of IEEE754 infinite in the double entries of
2565 * Introduce a new context factory "raw", highly inspirated from the
2566 ucontext factory, but using manually crafted functions in assembly to
2567 do the work in an efficient manner.
2568 * Allow to change the used context factory at run time, not only at
2569 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2570 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2571 contexts at runtime (only with raw contexts or ucontexts).
2572 * Completely rewrote this module to allow parallel execution of user
2573 processes. Use --cfg=contexts/nthreads:N to execute user processes
2574 with N parallel threads (the default is 1, meaning no parallelism).
2575 * Allow to decide dynamically between sequential and parallel modes.
2576 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2577 processes in parallel only when their number is greater than or equal to P
2579 * Added a check for NaN of IEEE754 infinite in the double entries of
2583 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2584 won't output the state of processes when interrupted with Ctrl-C
2585 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2586 NULL-terminated array. This may solve backward compatibility issues
2587 due to the change to return type of SD_simulate. See also:
2588 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2589 * Add new macros with variable number of arguments.
2590 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2591 - in xbt/asserts.h: xbt_assert
2592 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2593 - in xbt/ex.h: THROWF and RETHROWF.
2594 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2596 * Change xbt_die() to accept a format string with arguments, just like printf.
2597 * New data structure: xbt_lib_t, like a dict but more general and with better
2601 * New configuration options
2602 Options triva/categorized and triva/uncategorized can be used to generate
2603 graph configuration files for Triva visualization tool.
2604 * Configuration option tracing/platform is renamed to tracing/categorized
2605 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2606 * New TRACE_user_link_variable interface:
2607 User provides the name of the link and the tracing variable to attach to it
2608 * the declaration of tracing categories must be done after the environment
2610 * simpler tracing interface, just one way to declare categories
2611 TRACE_category or TRACE_category_with_color, it is up to you
2612 * links in the trace file are again identified by their names
2613 * trace contains the full platform hierarchy exactly as declared using the ASes
2614 * Options tracing/msg/[task|process]:1 groups the process by hosts
2615 for both cases, tasks and processes must have names that are unique during the simulation
2616 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2617 * The experimental option tracing/msg/volume is deprecated
2618 its functionality may be reincorporated if needed
2620 The tracing generates a trace file with unordered timestamped events,
2621 because of the way the core simulator (surf) works. A script available
2622 at the tools directory (fix-paje-trace.sh) can be used to put the events
2623 in order. We have changed the tracing so it can generate ordered timestamped
2624 events in the final trace, but depending on the simulator (and how much time
2625 is simulated) that can lead to a huge memory utilization. It is deactivated
2626 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2628 Build Infrastructure
2629 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2630 - We are trying hard to keep the API stable, but it may happen that
2631 some things change (we're a research project after all, not a
2632 nuclear plant operating system). If such things should happen, you
2633 could rely on that macro to adapt.
2634 - current value: 30600 for 3.06.00, aka 3.6
2635 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2636 a number that can be compared with SIMGRID_VERSION.
2637 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2638 assertions and verbose/debug logging events are disabled at compile time.
2640 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2642 SimGrid (3.5) stable; urgency=medium
2645 * New feature to any SimGrid-based simulator: Model-Checking
2646 Check SIN#1 for more details.
2649 * New Model SMPI (three-interval linear regression for correction factors)
2650 See RR-7426, available at http://hal.inria.fr/inria-00527150
2651 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2652 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2653 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2654 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2655 * See also src/smpi/README
2660 - Tracing API changes: TRACE_start and TRACE_end should not be called
2661 by user-code. They are automatically called by simulators created
2662 with SimDAG, MSG and SMPI if the toolkit is compiled with
2663 tracing_enabled=ON. Categories declaration and utilization remain the
2664 same for MSG and SimDag.
2665 - A function was added to the tracing API to declare categories with
2667 - TRACE_category_with_color (char *category, char *color)
2668 where color must be in the following format
2669 "%f %f %f", red, green, blue
2670 and red, green, blue are float values in the interval [0, 1]
2671 - User can specify NULL as color parameter, or continue calling
2672 TRACE_category (cat)
2673 On that case, the tracing system will define random colors
2674 - The following command-line options are supported:
2675 --cfg=tracing/filename:msg.trace
2676 --cfg=tracing:1 (activate tracing, needed to use others)
2677 --cfg=tracing/platform:1 (categorized resource use)
2678 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2679 --cfg=tracing/msg/task:1 (task creation)
2680 --cfg=tracing/msg/process:1 (process creation, migration)
2681 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2682 --cfg=tracing/smpi:1 (SMPI interface tracing)
2683 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2684 - examples of examples/msg/tracing updated
2686 - DAXLoader and DOTLoader functions can generate tasks with categories
2687 - A new function to attribute a category to SD tasks:
2688 TRACE_sd_set_task_category (SD_task_t task, char *category)
2689 * Tracing the MPI interface implemented by SMPI
2690 - Collective operations are traced with states
2691 - Point-to-Point operations are traced with states/links
2692 - Tracing activated by a parameter "-trace filename" passed
2693 to smpirun during execution (considering that simgrid
2694 is compiled with tracing enabled)
2695 - To run the simulation with gdb, the simulator
2696 accepts --cfg=tracing/smpi:1 to trace SMPI
2697 - tesh files to check if smpi tracing is ok
2698 - See examples/smpi/NAS/DT-trace
2699 * GTNetS tracing re-worked
2700 - adaptation to the tracing system of GTNets to cope
2701 with modifications regarding the fullduplex mode
2702 - new tesh files to check if gtnets tracing is ok
2705 * Asynchronous communications through the functions:
2706 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2707 * New function: MSG_load_platform_script()
2708 to make possible using a lua script instead of XML files to set up platforms
2709 * New function: MSG_set_function
2710 to associate functions to processes, used when bypassing the parser
2711 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2713 Platforms: Add some more examples in examples/platforms
2714 * Grid'5000: see www.grid5000.fr
2715 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2718 * Change the XML format. This is a very important modification. SimGrid
2719 3.5 introduces a new hierarchical format based on the notion of
2720 Autonomous Systems. Compatibility with old format is ensured
2721 through the perl script provided in the install bin directory
2722 bin/simgrid_update_xml.
2723 It is now possible to build platforms with specific routing
2724 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2725 connect several platforms together. We will try to provide soon
2726 set of realistic platforms exploiting these properties (have a
2727 look at examples/platforms/ for the moment).
2728 * Take the opportunity of the XML format change to be a good XML citizen:
2729 rename link:ctn to link_ctn and similar changes (also dealt with by
2731 * Add a new routing scheme (rule-based) using regular expressions. It
2732 enables to have an extremely low memory footprint when the
2733 underlying routing is simple and can be compactly described. You
2734 need to have libpcre4-dev (perl regular expressions) installed if
2735 you want to use this routing scheme.
2736 * Revive the cluster TAG and allow to easily and efficiently (both in
2737 term of memory and speed) connect clusters together. Have a look
2738 at teshsuite/simdag/platforms/ to see how this can be done. With
2739 this tag, you can create clusters with thousands of tasks at no
2740 cost (have a look at examples/platforms/).
2741 Note: clusters are implemented as ASes, so there is no need for an
2742 enclosing AS tag if you have only one cluster in your platform.
2743 * Add new generic functions in the public interface that allows the user
2744 to call SURF 'create_resource' methods from your code (same
2745 functionality as the XML bypass mechanism but with a much lighter
2747 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2748 that uses a piecewise linear approximation to produce better
2749 results when exchanging small messages.
2750 * Add a new parameter to handle correctly full duplex link and account
2751 for interferences between uplink and downlink communications
2752 (activate with --cfg=fullduplex:1).
2755 * Rename the SD_READY (all dependencies are satisfied and task is
2756 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2757 dependencies are satisfied) state.
2758 This prevents a confusion between the notion of "ready to schedule"
2759 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2760 simulated" (SD_RUNNABLE) used by the simulation kernel.
2761 * Change the way a task is considered as ready. Instead of removing
2762 dependencies when a task is done, a counter is decreased. This way,
2763 it is always possible to reach ancestors thanks to the
2764 SD_taks_get_parents function (even after the end of the simulation.)
2765 * Change the return type of SD_Simulate from (SD_task_t*) into
2766 xbt_dynar_t. This function was in handling a dynar internally and
2767 converted it into a NULL terminated array for historical reasons.
2768 * New function SD_dotload(char*) to load a DAG described in dot
2769 format. This loader and the corresponding examples require the
2770 installation of the graphviz library.
2771 * Fix a bug in the management of tasks of size 0 in the surf network
2772 models. This problem was only visible with SIMDAG and you should
2773 thus disregard results produced with earlier versions if you
2774 relied on this feature (some tasks were blocked because of this).
2775 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2776 even though your applications were doing only point-to-point
2777 communications and sequential computations. Now you can really use any
2778 model you want (of course, if you create real parallel tasks, which are
2779 not implemented in most models beside ptaskL07, this will abort).
2780 * Add an example that schedules a DAX on an heterogeneous platform
2781 using a Min-Min strategy.
2782 * New function SD_workstation_get_current_task() that returns the kind
2783 of task currently running on a workstation in the sequential access
2785 * Raise some warnings when unexecuted tasks remains at the end of the
2786 simulation. This is usually caused by cycles in the DAG.
2789 * New function: SIMIX_process_set_function() called by MSG_set_function
2790 * Change the underlying waiting queue in semaphores so that a process
2791 can wait on several of them simultaneously (as in waitany).
2792 * Fix the way to handle tokens in semaphores so that all access patterns
2793 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2794 * kill the dirty pimple SIMIX_message_sizes_output()
2795 Please use (proper) visualization instead
2798 * New data container: setset (set of sets of elements)
2799 * New module: mmalloc (mapped malloc, allowing to have several
2800 independent segments of malloc)
2801 * New function: xbt_dict_cursor_set_data()
2802 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2803 * New function: xbt_dynar_is_empty()
2804 * New function: xbt_fifo_get_last_item()
2805 * Fix xbt_dynar_shrink(): use the right element size.
2806 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2807 uninitialized areas during expand.
2808 * Fix semaphores: previous implementation was severely broken.
2809 * Use library init/fini functions for our initialization.
2810 - you can use logs and other feature as soon as you want in your
2811 code (even before the xbt_init / MSG_init)
2812 - xbt_exit is now a no-op and produce a warning when used.
2815 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2816 benefit from the latest improvement to the simulation kernel.
2817 * Kill measurement sockets for now. If you rely on them, sorry. This
2818 release is not for you. This feature will be reintroduced in the
2819 future, but we cannot delay the release any further.
2820 * New function: gras_msgtype_get_name().
2821 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2822 * Fix (at last) the pmm example: it should not randomly fail anymore.
2824 Build chain: bug fixes and overall polishing
2825 * Cmake is now stable enough. Hence, we killed the autotools.
2826 * Port to windows ( TM :)
2827 * Fix the 'make install' target.
2828 No need to use 'make install-simgrid' anymore
2829 * Introduce a 'make dist' target compiling a *source* archive
2830 'make package' compiles a binary archive
2831 * Compile java files only on need
2832 * Add --cd and --setenv command line options to tesh
2833 * Out of source builds are not fully supported yet, but we are close
2834 * Enable supernovae and optimization flags by default for our users
2837 * Add layer to set up environment directly from lua, without XML.
2838 * The effect of gras_stub_generator can be achieved through
2839 lua too (check examples/gras/console/ping_generator.lua)
2841 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2843 SimGrid (3.4.1) stable; urgency=low
2845 The "Polishing easter eggs is probably a good idea" release.
2846 This is a bug fixes release only.
2850 * Fix a bug preventing the tasks from begin garbage collected.
2853 * Fix a bug occurring when a host involved in a communication fails.
2854 This was not detected properly by the other peer involved in the
2855 communication. Now, it's reported as a network error.
2858 * Warn the user about loop dependencies in data flow of DAX files
2859 * Obey the control-flow dependencies of DAX files
2862 * Add option "enable_smpi" allowing to not compile SMPI.
2863 Probably useful for the (Mac) users experiencing a build error here
2864 * Improve the detection of lua5.1 and ruby1.8
2866 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2868 SimGrid (3.4) stable; urgency=low
2870 The "Easter in Cargese" release. Also known as (major changes):
2872 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2873 ~> bindings were greatly improved
2874 ~> new tracing infrastructure for better visualization introduced
2876 * the "Welcome to configury modernity" release.
2877 ~> we switched from autotools to cmake, and improved our cdash
2880 A more detailled list of changes follow (full detail in svn log).
2882 Java Bindings: Various Cleanups
2883 * (install java-gcj-compat-dev on debian-like to use them)
2884 * Remove put/get: no need to export deprecated interface in Java
2885 Use send/receive instead.
2886 * Cleanup the examples and add a README per directory
2887 * Remove example autoDestination (that's the only way to go now)
2888 * Remove example explicitDestination (was a plain copy of basic)
2889 * Make JniException a runtime exception, so that there is no need to
2890 declare the fact that you may encounter such a beast. I guess that
2891 nobody will ever want to survive such error.
2892 * Create specific errors for each MSG case of failure:
2893 host failure, transfer failure, timeout, task cancelled
2894 * Cleanup the exceptions that may get thrown by each function
2895 * Other internal cleanups in Java bindings. Performance still bad :/
2896 Ruby and Lua Bindings: create them
2897 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2898 * That's new and great, you should try them out.
2899 Same functionalities than Java bindings, only even less polished
2901 * Kill the useless "rate" argument of SD_task_get_execution_time()
2902 Everyone used to provide -1 as a value, it was not used, and the
2903 semantic of a possible use wasn't even clear.
2904 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2905 either as comm costs or compute costs to mean that there is no
2906 such thing for that specific task.
2907 * Add a SD_task_set_name() function
2908 * Fix SD_task_unschedule() on typed tasks
2909 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2910 * In DAX loader, accept useless 'level' attributes to <job> since
2911 LIGO DAGs have them (seem to be to ease graphical representation).
2913 * Add an example masterslave_mailbox.c using send/receive and not
2914 the deprecated put/get interface.
2915 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2916 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2918 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2919 (declare MSG_USE_DEPRECATED to still have the old name)
2920 * Add a MSG_task_set_data() function
2921 * About trace replay (see examples/msg/actions):
2923 - Allow to work with splitted trace files for each process
2924 Give the specific trace file as argument of each process,
2925 and call MSG_action_trace_run(NULL)
2926 You can still have one merged file for all processes.
2927 - Fix implementation of collective operations
2928 * Allow task_execute() on 0-sized tasks (closes #10063)
2930 * This is the first release of SimGrid where SMPI is not considered
2931 beta anymore (even if some corners should still be improved)
2932 * Port over the new SIMIX_network submodule (internal refactoring)
2933 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2934 * Implement more missing elements of the standard:
2936 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2938 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2939 MPI_Scan() MPI_Get_processor_name()
2940 - Added implementation of missing case for Alltoall (warning: it's
2941 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2942 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2943 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2944 SMPI_MPI_Allgatherv()
2945 * Bug fixes include:
2946 - MPI_Waitsome() was broken
2947 - Allow relative includes in smpicc
2948 - Command line cfg argument 'reference_speed' was ignored...
2949 - Some functions did not properly lead to auto-benching of user code
2950 - smpicc passes -O2 by default (just like openmpi one)
2952 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2953 * Bug fixes about timeouts during communications
2954 * add SIMIX_message_sizes_output() as a pimple to write to file the
2955 amount of messages per size. Use gnuplot to get histogram.
2956 Pimple because that's the only user-visible function of simix,
2957 defined directly in xbt.h (irk, sorry)
2959 - Add a SIMIX_sem_get_capacity() function
2960 - Fix interactions with processe resume/suspende
2961 - release_forever() was stupidly broken
2962 - Fix SIMIX_display_process_status() for processes in a semaphore
2963 - Make SIMIX_sem_block_onto() user-visible
2964 * Refactoring context stuff:
2965 - Use pseudo-OOP for better modularity
2966 - reimplement SIMIX_process_kill() without process_schedule() so
2967 that the latter can take as invariant that it is called from
2969 - Merge context_start into context_new for sake of simplicity
2971 * Add a Vivaldi network model, coded live during SUD'10 ;)
2972 * Rename configuration variables to start a hierarchy:
2973 o cpu_model -> cpu/model
2974 o network_model -> network/model
2975 o workstation_model -> workstation/model
2976 * New configuration variables:
2977 o network/bandwidth_factor: correction to bandwidth
2978 o network/latency_factor: correction to latency
2979 o netwotk/weight_S: correction to the weight of competing streams
2980 * Add a long description to the models, that users can see with such
2981 argument on the command line: --cfg=cpu/model:help
2982 * --help-models display the long description of all known models
2984 * config: add the ability to set a default value after registration
2985 Does not override any previously set value (e.g. from cmd line)
2986 * dict: allow to have integer key and data.
2987 When so, you need to use the following functions
2988 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2989 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2990 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2991 In contrary to regular dicts, the key is not malloced before copy.
2992 Mixing scalar and regular elements in the same dict is not tested
2994 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2995 Tracing for Visualization:
2996 * SimGrid is now instrumented in order to generate a trace file for
2997 visualization analysis: to use it, need to compile SimGrid with the
2998 "tracing" option enabled, and instrument the program using SimGrid with
2999 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
3000 (among other functions).
3001 * The instrumentation only traces the platform utilization for now
3002 * Documentation to use the tracing functions and how to analyze the
3003 traces with the Triva tool is written.
3004 * More information about: SimGrid FAQ (in the section Tracing Simulations
3007 * We moved to cmake as default build system. Autotools support will
3008 be dropped soon. Check the FAQ for more info about how to use it.
3009 * Greatly improved our cdash/ctest interactions
3010 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
3011 * Added memory checking tests with valgrind; lot of memleak fixing.
3012 This may be the first release of simgrid with so few memory issues
3013 * Added code coverage tests.
3014 Our coverage is still improvable, but at least we see it on cdash.
3016 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
3018 SimGrid (3.3.4) stable; urgency=low
3020 The "Desktop Grid needs love too" release (also called Xmas release).
3022 Models improvements:
3023 * Major speedup in the maxmin system solving by using lazy evaluation
3024 Instead of solving completely the maxmin system at each iteration,
3025 only invalidate (and recompute) the modified parts.
3026 This new feature is enabled in default models but you can try to
3027 turn it on with "--cfg:maxmin-selective-update=1" for other models.
3028 * Cas01 IMproved as default CPU model
3029 This CPU model is the same Cas01 model, but it uses the
3030 maxmin-selective-update flag and a heap structure to manage
3031 actions on SURF kernel.
3032 It reduces the complexity to find the next action to finish and,
3033 consequently, it's faster than the old Cas01.
3034 This is the new default CPU model (Cas01).
3035 * Rename the old Cas01 model to Cas01_fullupdate
3036 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
3037 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
3038 * CpuTI (CPU Trace Integration)
3039 A new CPU model whose objective is simulate faster when using
3040 availability trace files.
3041 Instead of using a full featured, over engineered maxmin system for
3042 CPU modeling, this model does the pre-integration of traces files
3043 to calculate the amount of CPU power available, and so, executes
3044 faster than the old CPU models.
3045 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
3046 * Use LV08 as default network model since it gives better accuracy
3047 for small messages and shouldn't change things for big ones.
3048 Use --cfg=network_model:CM02 to get the previous behavior.
3051 ******************************************
3052 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
3053 ******************************************
3054 * The new CPU model may changes simulations!
3055 The point is that events occurring at the exact same timestamp
3056 are not scheduled in the same order with the old and new
3057 version. This may be enough to completely change the execution
3058 of simulations in some cases.
3059 * The new network model will change simulations!
3060 This new model is more realistic than the previous one, so you
3061 should consider redoing your old experiments with this model.
3062 Sorry for the inconvenience.
3065 * Introduce the supernovae compilation mode
3066 When compiled that way, the whole SimGrid (or almost) is put in a
3067 single compilation unit and compiled in one shoot.
3068 This is to help gcc which has difficulties to inline stuff from one
3070 The speedup seem to be above 15%, althrough more tests are needed on
3071 amd64 to confirm that gain.
3074 * Port of MSG's mailbox on top of SIMIX network
3075 The put/get mechanism was greatly simplified on the way.
3078 * New SIMIX network module. Provides:
3079 - Mailbox: rendez-vous mechanism to find with who you want to speak
3080 - Synchronous send/recv: easier and hopefully faster since the
3081 logic is handled in the maestro process directly now
3082 - Asynchronous send/recv: you dreamt of it? It's here now
3083 Too bad that nobody cared enough to propagate the change to MSG.
3084 * Add semaphores as SIMIX synchronization mechanism.
3087 * new function SD_daxload(char*) to load a DAX file
3088 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3089 * Introduce typed tasks. Specify its kind and cost at creation.
3090 At scheduling, just give where it should be placed, and the cost
3091 for each involved resource is automatically computed.
3092 Existing constructors so far (more to come of course):
3093 - SD_task_create_comm_e2e() for end-to-end communication
3094 - SD_task_create_comp_seq() for sequential computation
3095 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3096 * new function SD_task_dump() for debugging display
3097 * new function SD_task_dotty(task,FILE*) writing to file the info
3098 about the task in dotty format
3099 * SD_task_dependency_exists() can now cope with having one of its
3100 arguments NULL. If so, it tests whether the other argument has any
3102 * Add getters on list of preceding/following tasks:
3103 SD_task_get_parents(task) and SD_task_get_children(task)
3104 * Add getters on amount of workstations and list:
3105 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3106 * Add getter on task kind: SD_task_get_kind(task)
3107 * Update the start_time and finish_time of tasks on completion/failure
3108 * Bugfix: Remove task from state swags when destroyed
3111 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3114 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3115 dynar with one separate thread per value of the dynar.
3116 * Change the prototype of xbt_thread_create(), sorry.
3117 Added a boolean parameter indicating whether we want to join this
3118 thread (used in SG only for now)
3119 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3122 * GTNetS wrappers should now be usable again (and betterly tested too)
3123 * Fix a major regression from 3.2 where the timeout provided to
3124 MSG_task_put_with_timeout() was used as absolute time before which
3125 the comm should be done.
3126 * Start to fix the <cluster> tag.
3127 - Internal links should be good now (beside of the loopback, which
3128 use the private link instead)
3129 - paths to the external world is still rather broken
3130 - the <route:multi> tag is just broken. Actually that's brain-dead.
3131 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3132 to make it less stupid
3133 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3134 * Fix a source-level compatibility glitch from 3.2: after defining
3135 MSG_USE_DEPRECATED, you can use the old name
3136 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3137 * Allow to compile from the SVN with automake 1.11
3138 * Fix some problems when using the "start_time" tag in deployment XMLs.
3139 * Fix #8569: XBT/synchro.h has redundant declarations
3140 * Fix #8563: MSG return values and exceptions
3141 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3142 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3143 Thanks to Robson Peixoto.
3144 * Fix a vicious bug in dictionaries inducing that some elements were
3145 not freed on xbt_dict_free()
3147 Portability report of this version:
3148 * Main portability targets:
3149 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3150 on (amd64/i386/ia64)
3151 - mac leopard on i386
3152 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3153 but nothing critical.
3154 * Other platforms: windows, AIX and others were not tested for this release
3156 Timing report of this version:
3157 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3158 scenario parameters). From 8h to a few seconds in desktop grid settings.
3159 * Supernovae brings about 25% speedup on i386.
3161 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3163 SimGrid (3.3.3) stable; urgency=low
3165 The "Need for Speed" release.
3167 The timings done to validate the 3.3.2 were faulty.
3168 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3170 The problem was a conversion from a manually handled vector to
3171 xbt_dynar_t on the critical path.
3172 xbt_dynar_foreach calls functions, inducing stack management crap.
3174 We inlined these functions and xbt_dynar_foreach is now breath taking.
3175 We also inlined xbt_swag_belong on the way.
3177 Here are some approximate speedup measurements (on master/slaves
3178 simulations lasting between 10s and 20s each):
3179 3.3.1 -> 3.3.2: about same performance
3180 3.3.2 -> 3.3.3: 40% speedup
3181 3.3.1 -> 3.3.3: 40% speedup
3182 3.3.1 with inline patch -> 3.3.3: 30% speedup
3184 Our reading is that the refactoring which occurred in 3.3.2 made us
3185 suffer much more from the xbt_dynar_foreach low performance, but
3186 once we solved this, this refactoring proved to be very performance
3187 effective. From the 40% speedup, somehow, 10% are due to the
3188 inlining and 30% to the refactoring.
3190 That's a pitty that gcc cannot inline functions placed in other files
3191 alone. We have to choose between:
3192 - break the encapsulation (by putting private data structures and
3193 accessors in headers files to help gcc)
3194 - live with low performance
3195 - switch to a decent compiler such as icc (not quite possible).
3197 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3199 SimGrid (3.3.2) stable; urgency=low
3201 The "Simplicity does not preceed complexity, but follows it" release.
3203 The main contributors of this release were (lexical order):
3204 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3207 * Extract the routing logic into its own object.
3208 (was dupplicated in network.c and workstation_LV07.c;
3209 Allows to implement other ways of storing that info)
3210 => kill now useless network_card concept
3211 - Use dynar to represent routes (instead of void** + int*)
3212 - kill link_set (use surf_network_model->resource_set instead)
3213 - Add a command-line option to choose the routing schema to use
3214 - Add three new models:
3215 * Floyd (shortest path computed at initialization)
3216 * Dijikstra (shortest path recomputed all the time)
3217 * Cached Dijikstra (shortest path computed on need)
3218 All these models where contributed by Silas De Munck, and are
3219 described in his ICCS09 paper.
3221 * Simplify model declaration
3222 (less redirections, less function to write when defining a model)
3223 - Factorize stuff between models:
3226 surf_model_resource_set(model)
3227 surf_model_resource_by_name(model, name)
3228 - Unify the types of models in s_surf_model_t (using an union)
3229 - Embeed fields of common_public directly into s_surf_model_t
3230 - Rename model methods:
3231 action_free ~> action_unref
3232 action_change_state ~> action_state_set
3233 action_get_state ~> action_state_get
3234 - Change model methods into functions:
3235 (model)->common_public->action_use ~> surf_action_ref
3237 * Implement a generic resource; use it as ancestor to specific ones
3238 (allows to kill duplicated code in models)
3239 Drawback: timer command don't need no name nor properties;
3240 workstation_CLM03 don't need no properties
3241 (but I guess we can live with those few bytes wasted)
3243 * Improve the action object model
3244 - implement a constructor avoiding dupplicated code about field
3245 initialization in generic_action part.
3247 * Kill the SDP model: it has an external dependency, is deprecated
3248 in flavor of modern lmm models, and didn't compile since a while
3251 * Relocation of the context module from XBT to SIMIX.
3252 (the context were decoupled from the simix processes, duplicating a lot of code)
3253 => a lot of code was factorized
3254 - less overhead is introduced during scheduling
3255 - simpler API for the context factory
3256 - the logic for process creation,destruction and manipulation was simplified
3257 * Simplification of the s_smx_process_t data structure.
3258 => accesing the simix level data associated to a process is faster now,
3259 and the code is a lot more readable.
3262 * Implement some more MPI primitives:
3263 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3264 -implementation: Bcast: flat or 2-ary tree (default),
3265 Barrier: 4-ary tree,
3267 Allreduce: Reduce then Bcast
3268 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3269 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3270 Alltoallv: flat tree, like ompi
3272 * Add support for optimized collectives (Bcast is now binomial by default)
3273 * Port smpirun and smpicc to OS X
3276 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3277 Shout out if you used it.
3280 * Display the list of still queued messages in SG mode when existing
3284 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3285 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3286 * Add xbt_str_from_file(FILE*)
3287 * Add xbt_dict_get_key achieving a linear reverse search
3288 * Remove the context module
3290 Portability report of this version:
3291 * Main portability targets:
3292 - Linux(debian)/x86/context
3293 - Linux(debian)/x86/pthreads
3294 - Linux(debian)/amd64/context
3295 - Linux(debian)/amd64/pthreads
3296 On these, we still have the eratic breakages of gras/pmm and
3297 amok/saturate_sg reported in previous version. We still think
3298 that the tests are the cause of the fault, not the tested code.
3300 - Mac OSX Leopard/x86/context
3301 Still false negative in tesh autotesting.
3302 Smpi still fails, but this time because readlink does not accept -f
3303 Everything seems to work properly beside of that.
3306 - AIX version 5.3 (only tested contexts this time)
3307 Smpi still fails there because mktemp is not installed.
3308 Everything seems to work properly beside of that.
3310 I managed to compile it for the first time, but several breakages.
3311 Won't delay the release for this exotic platform.
3313 * Windows: it's still lagging behind. If you want to help, please
3316 Timing report of this version:
3317 This version seem to be more than 5% faster than 3.3.1 (on linux
3318 64bits with contextes). The gain is less than expected, we are
3319 investigating this for next release.
3321 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3323 SimGrid (3.3.1) stable; urgency=low
3326 * Implement a --cfg-help to show existing configuration variables
3327 * Build chain do not require doxygen in maintainer mode
3330 * fix a bug on struct sizeof computation, which prevented the
3331 exchange of arrays of structs in some conditions
3332 - added a regression test about this in datadesc_usage
3333 * Allow the exchange of 0-long dynamic vectors.
3334 - for that, use -1 as indicator of dynamic size instead of 0
3335 - This implied to change any size from unsigned long to long,
3336 reducing a bit communication abilities, but I guess that with
3337 64bits being quite common, this is more than enough.
3338 - This also induce a protocol change, thus bumping network protocol
3339 version from 0 to 1 (if we have external users, we have to get
3340 clean on that point too ;)
3341 - added two regression tests about this in datadesc_usage
3342 * Be more verbose when propagating local exceptions
3343 This helps debugging.
3344 * Display the status of simulated processes when receiving SIGINT in
3348 * Allow to control the simulation from a trace file.
3349 New functions MSG_action_register() and MSG_action_trace_run()
3350 The first one allows to associate a function execution to each
3351 kind of action while the second one parses a trace file and
3352 triggers the corresponding actions within the system.
3353 For now, only a toy example is provided in examples/msg/actions
3354 * Add an example of process migration in examples/msg/migration
3355 * Fix a bug in task exchange which broke MSG_task_get_sender()
3356 Add a teshsuite regression test for that.
3357 [Bug: if MSG_task_get_sender() is called after sender exit,
3359 * Fix a bug which prevented suspend/resume to work properly
3360 * Display the status of simulated processes when receiving SIGINT
3361 This fixes a regression of v3.3. due to the introduction of SIMIX
3362 * Bug fixing in failure management:
3363 - trace could not start by a failure at time 0
3364 - failure during communications were not working
3367 * Add SIMIX_process_set_name() to change the name of the current
3368 process in the log messages.
3369 * Store smx_hosts in a dict since we only retrieve them by name
3370 * Move the configuration infrastructure to surf
3373 * Move the configuration infrastructure to surf
3376 * Massive internal cleanups:
3377 - Store internal structures on processes instead of hosts (allows
3378 to have more than one process per host, in addition of being more
3380 - Cleanup the initialization/finalization process
3381 - Kill a whole bunch of unneeded synchronization:
3382 processes run in exclusive manner within the simulator
3383 - Move queues from global tables to process data fields
3385 - now accept -platform and -hostfile arguments
3386 - Pass the right rank value to processes according to the hostfile
3387 * Compile the examples by default, and use them as regression tests
3388 * Implement MPI_Wtime()
3389 * Change the reference speed to a command line option
3392 * TCP_gamma can now be specified as command line option using
3393 --cfg=TCP_gamma:10000000.0
3394 * Change the --surf-path cmd line option into --cfg=path:
3397 * Also include strbuff from xbt.h public header
3398 * xbt_ex_display(): do not free the exception after displaying
3399 This allows to do more with the given exception afterward.
3400 Users should call xbt_ex_free() themselves.
3404 Portability report of this version:
3405 * Main portability targets:
3406 - Linux(debian)/x86/context
3407 - Linux(debian)/x86/pthreads
3408 - Linux(debian)/amd64/context
3409 - Linux(debian)/amd64/pthreads
3410 These targets fail about 1/10 of times on gras/pmm, but we believe
3411 that this is because of the test, not because of simgrid.
3412 amok/saturate_sg fails even more rarely, and the test may not be
3415 - Mac OSX Leopard/x86/context
3416 The test suite still spits tons of errors because some obscure
3417 force prevents us from removing the temporary directories
3418 arguing that they still contain some metadata I've never heard of.
3419 Smpi fails because seq is not installed.
3420 Everything seems to work properly beside of that.
3423 - AIX version 5.3 (both contexts and pthread)
3424 Smpi still fails there because mktemp is not installed.
3425 XML inclusions seems rosty on AIX.
3427 * Windows: it's still lagging behind. If you want to help, please
3430 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3432 SimGrid (3.3) stable; urgency=high
3436 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3437 [Malek Cherier & Mt]
3439 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3442 Until now, GRAS were using MSG as an interface to SURF. It was
3443 quite difficult because both interface have several differences
3444 (MSG channels vs GRAS sockets were the most notable point).
3446 This also opens the gate to SMPI (which should occur soon) and speed
3447 up simulations by to 40% (even if it were not the main goal).
3449 **************************************
3450 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3451 **************************************
3452 The point is that events occurring at the exact same timestamp are
3453 not scheduled in the same order with the old and new version. This
3454 may be enough to completely change the execution of simulations in
3455 some cases. Sorry for the inconvenience.
3457 * Cleanup and upgrade the XML format to push further scalability
3458 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3460 * Improve the testing infrastructure with tesh. Now a very large part of
3461 the code is tested not only by being run but also by checking that the
3462 output match an expected output [Mt].
3464 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3465 is really less memory-demanding, which should allow you to use
3466 larger files in SimGrid [AL].
3468 * Inform valgrind about our contextes, so that it becomes usable
3469 with the default (and more effecient) version of SimGrid
3470 [contributed by Sékou Diakite, many thanks]
3473 * Introduce a listener thread in charge of receiving incoming
3474 messages from the network. It allows to overlap communication and
3475 computation but most notably, it removes some stupid deadlocks due
3476 to the fact that so far, a process could not send and receive at
3477 the same time. This made most non trivial communication schema
3479 * Convert the PIDs from long int to int to match the MSG ones (and
3480 linux ones too) [Mt]
3481 * New function: gras_agent_spawn() to launch a new process on
3482 current host. Only working in simulation for now. [Mt]
3483 * New function: gras_os_hostport() returning a constant form (ie,
3484 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3487 * Make the backtrace of exceptions more human readable [Mt]
3488 * New module: xbt/str [Mt]
3489 a ton of string utility functions (split, join, printf to a newly
3490 allocated buffer, trim, etc)
3491 * New module: xbt/hash [Mt]
3492 SHA1 hashing algorithm (more to come if needed)
3493 * New module: xbt/synchro [Mt]
3494 synchronization tools (mutex and conditions) working the same way
3495 in simulation and in real life (mainly useful for GRAS, but not
3497 * New module: xbt/queue [Mt]
3498 classical producer/consumer synchronization scheme
3499 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3500 (using the classical functions will get serialized) [Mt]
3501 * Make dictionary internal table dynamic. No need to specify its size
3502 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3504 * Make sure the log channels are organized as a tree under windows
3505 (because of ANSI C compatibility issue, any channel were child of
3506 root directly) [Mt].
3509 * Cleaned many thing in surf and fixed a few bugs [AL].
3510 * Add a nice command line configuration mechanism to compose models [AL].
3511 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3512 the previous one (KCCFLN05). It relies on something that looks like
3513 a max-min sharing mechanism but cannot be written as such. A new solver
3514 was thus designed [AL].
3515 * Add a new solver to lmm. Based on Lagrange optimization and
3516 gradient-based descent, it enables to efficiently maximise systems s.a
3518 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3520 This solver enables to propose two new network models for TCP Reno and
3521 TCP Vegas based on Low's work. These models still need to be fully
3522 tested though [Pedro Velho].
3525 * Bug fix in SD_simulate. Now the time bound given as argument is
3527 * Use the new parallel task model (ptask_L07) as default.
3528 * Use the SURF command line configuration mechanism.
3529 * 0-size tasks (for synchronization) should now work.
3531 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3533 SimGrid (3.2) stable; urgency=high
3537 We still experience issues on this platform, but we believe that at
3538 least MSG is usable.
3540 GRAS API BREAKAGE (for simplification purpose, sorry):
3541 * the gras_msgtype_by_name is not used anymore. Instead of
3542 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3543 you can write (and must)
3544 gras_msg_send(toserver, "request", &request);
3545 - If you still want to pass a gras_msgtype_t to the function (to cache
3546 the type and avoid the lookup time), use the gras_msg_send_() variant.
3547 - Impacted functions:
3548 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3549 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3550 * The callbacks are now expected to return 0 when everything went well
3551 (just like the main() function)
3553 GRAS new features and improvements:
3554 * New module mechanism where user code can use per process globals [Mt]
3555 This is similar to gras_userdata_*() functions, but for libraries. It
3556 factorize some code developped over and over in the examples and AMOK.
3557 It has still to be documented and used (only amok/peermanagement is
3559 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3560 loss when they were small enough to fit into the buffer and sent quickly
3561 enough so that they can all get received in one shoot.
3562 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3563 if not found. Use the *_or_null() variant for the old semantic.
3564 * In gras_msg_handle, do not discard messages without callback.
3565 They are probably messages to be explicitly awaited later (ie, proofs of
3566 mis-synchronization in userland since they are sent before being awaited)
3567 No big deal usually.
3568 * gras_socket_meas_send/recv: semantic changed!
3569 The numerical arguments used to be (1) the total amount of data to send
3570 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3571 messages. This was need for the fool willing to send more than MAXINT
3572 bytes on quite fat pipes.
3575 * Do really rename the hostmanagement module to peermanagement. [Mt]
3576 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3577 but this is rather new and this was documented in the module
3578 documentation (poor excuses, I admit)
3579 * Bandwidth measurement semantic changed! This follows the changes to
3580 gras_socket_meas_send/recv explained above.
3583 * A sequential mode has been added to the workstations. When a workstation
3584 is in sequential mode, it can execute only one task, and the other tasks
3585 are waiting in a FIFO. [Christophe Thiery]
3588 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3589 model for SIMDAG. [Christophe Thiery]
3590 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3591 equal to 0, which caused some bad behaviors in
3592 saturated_constraint_set_update. I now use a threshold mechanism like in
3596 * When running manually src/testall, you select specific units [Mt]
3597 testall is the result of our cunit mechanism, and should replace all
3598 the scripty thingy around since bash don't run easily on billware.
3600 * A mallocator system has been added. [Christophe Thiery]
3601 Mallocators allow you to recycle your unused objects instead of freeing them
3602 and allocating new ones.
3604 Documentation update:
3605 * FAQ reworking + New FAQs:
3606 - "Valgrind spits tons of errors!" [Mt]
3607 - "How to repport bugs" [Mt]
3608 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3609 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3610 - Communication time measurement within MSG [AL]
3611 - I experience weird communication times when I change the latency [AL]
3612 * GRAS tutorial [Mt]
3614 - an introduction to the framework and to the used communication model
3615 - an initiatic tour introducing the most proheminent features:
3617 . Lesson 0: Installing GRAS
3618 . Lesson 1: Setting up your own project
3619 o Part 2: Message passing
3620 . Lesson 2: Exchanging simple messages
3621 . Lesson 3: Passing arguments to the processes (in SG)
3622 . Lesson 4: Attaching callbacks to messages
3623 . Lesson 5: Using globals in processes
3624 . Lesson 6: Logging information properly
3625 . Lesson 7: Using internal timers
3626 . Lesson 8: Handling errors through exceptions
3627 . Lesson 9: Exchanging simple data
3628 . Lesson 10: Remote Procedure Calling (RPC)
3629 . Lesson 11: Explicitly waiting for messages
3630 . Recapping of message passing features in GRAS
3631 - A HOWTO section containing:
3632 o HOWTO design a GRAS application
3633 More are due, of course. They will come latter. In the meanwhile, you can
3634 check the examples which are still here.
3636 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3638 SimGrid (3.1) stable; urgency=high
3642 There was a stack corruption somewhere, visible only when optimizing
3643 with these versions. [Vince]
3646 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3647 module that enabled to play with Directed Acyclic Graphs. It is built
3648 directly on top of SURF and provides an API rather close to the old
3649 SG. Some old codes using SG are currently under rewrite to check that
3650 all needful functions are provided. [Christophe Thiery]
3653 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3654 extension of the classical CLM03 model that gracefully handles
3655 failures. This is now the default model for MSG and GRAS. It doesn't
3656 handle parallel tasks yet though. [AL]
3657 * Bug fix: Weights were not correctly set in the network part.
3658 WARNING: This may have resulted in incorrect results with simulations
3659 where there are more than one flow on a given link. [AL]
3662 * After a (long ?) discussion on simgrid-devel, we have decided that the
3663 convention we had on units was stupid. That is why it has been decided
3664 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3665 WARNING: This means that all previous platform files will not work as
3666 such with this version! A warning is issued to ask users to update
3668 A conversion script can be found in the contrib module of the CVS, under
3669 the name contrib/platform_generation/surfxml_update.pl [MQ]
3672 * Bug fix: Processes were started in reverse order, wrt deployment file.
3673 WARNING: if your code relies on this bug, please fix it. [AL]
3674 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3675 being executed on two different locations. [AL]
3676 * Bug fix: Failures are now better supported thanks to Derrick's tests
3677 (there was many failure situations I hadn't thought of and that weren't
3678 correctly handled). [AL]
3679 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3683 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3684 Exception occurring on server-side are propagated back to client (!).
3686 API CHANGE: the callback changed their prototype. Change:
3687 int my_handler(gras_socket_t expeditor, void *payload_data) {
3689 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3690 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3692 * New! function: gras_msg_handleall to deal with all messages arriving
3693 within a given period.
3694 * New! function: gras_socket_server_range to get a server socket in a
3695 range of port numbers (ease to avoid port number conflicts) [MQ]
3696 * New! gras processes display their backtrace when they get a SIGUSR1
3697 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3698 Sweet to debug RL processes [MQ]
3702 - Do not force experiment sizes to be expressed in kb, or it becomes
3703 impossible to measure the latency this way (needs one byte-long tests)
3704 WARNING: this changes the amok_bw_* function semantic. [MQ]
3705 - Implements the link saturation stuff. [MQ]
3706 * Peer management module:
3707 New! module factorizing code that we wrote over and over [MQ].
3710 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3711 - Test units are placed directly into the library code, they get extracted
3712 automatically and placed into the src/testall binary.
3713 - Convert most of the XBT tests to this system.
3714 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3715 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3716 live in the same C code. This required to change a little bit the API
3717 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3718 * New module: sparse graph structure with basic algorithms (this is work
3719 in progress and the API is not considered to be frozen yet). [AL]
3720 * Display more information on backtraces: source line & function names are
3721 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3722 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3723 * Reworked a little bit some #include statements to load only required
3724 headers. Some user code that relied on SimGrid to include stdlib or
3725 stdio may need to include it by themselves. [AL]
3726 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3728 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3729 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3730 but I guess that almost nobody used those parts. [MQ]
3732 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3734 SimGrid (3.0.1) stable; urgency=low
3737 * Unfortunately, I had missed 5 misnamed functions:
3738 xbt_fifo_item_t xbt_fifo_newitem(void);
3739 void xbt_fifo_freeitem(xbt_fifo_item_t);
3740 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3741 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3742 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3743 They're now deprecated. Please use their new versions:
3744 xbt_fifo_item_t xbt_fifo_new_item(void);
3745 void xbt_fifo_free_item(xbt_fifo_item_t);
3746 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3747 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3748 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3750 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3751 * Documentation: xbt_log module unmercifully reworked [MQ]
3752 * Bugfix: there was a problem with the ending of contexts with
3753 the pthread backend. It caused some weird deadlock or behavior
3754 depending on the pthread implementation. [AL]
3755 * Bugfix: get the exceptions raised in the simulator repport where
3756 and why they come from when they are not caught in time [AL, MQ]
3759 * Bugfix: Do repport the error when two non-connected hosts try to
3760 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3763 * Add additionnal checkings on communications. Assert that two
3764 communicating hosts are connected by a set of links... [AL]
3767 * Add additionnal checkings on channel values in communication [AL]
3768 * New: MSG_task_get_source to see on which host a task was generated [HC]
3769 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3770 the number of tasks waiting to be received on channel and sent
3772 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3773 waits for the first task coming from a given host.. [AL]
3775 GRAS new functionnalities: [MQ]
3776 * Enhance the parsing macro to allow the size of multidimentional objects
3777 to be given thru annotations.
3778 * New example (and documentation): Matrix Multiplication a la RPC
3779 (as when I was young!) and fix a bunch of bugs found on the way.
3781 GRAS performance improvements: [MQ]
3783 * Reduce the amount of cbps creation/destruction by making it static to
3784 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3787 * Change libdata to a set so that we can search for stuff by ID (and thus
3788 reduce the insane amount of dict lookups)
3791 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3792 RL communication on top of SG since it's so uneffective.
3793 It may also be used for inter-thread communication in RL, one day.
3794 * Use gras_datadesc_copy() to exchange messages on top of SG
3796 - improve message exchange performance on top of SG
3797 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3798 * Don't exchange on the network the size of the used part of buffer,
3799 instead, specify the possible buffer size to read().
3801 - reduces the amount of read/write calls (one pair per exchange)
3802 - reduces the amount of exchanged data (the size)
3803 - allows to retrieve all arrived data on receiver side, if we don't need
3804 it right now (subsequent read will peek the buffer)
3805 - allows the receiver to proceed with the beginning of the stream before
3806 everything is arrived
3807 - make it possible to build an iov transport (using readv/writev)
3809 - take care of the data with non-stable storage (like stacked data),
3811 * If possible, TCP send uses vector I/O (when writev() is here)
3812 - Don't use it for receive since we send data sizes and data on the
3813 same stream, so we wouldn't be able to chain large amount of chunks
3814 before having to flush the stuff to read the size.
3815 * Rework the transport plugin mechanism to simplify it and reduce the
3816 amount of pointer dereferencement when searching for the right function
3819 * I guess that now, we do almost as few system calls as possible while
3820 doing as few data copy as possible.
3822 To improve it further, we could try to send all the sizes first and then
3823 all the data (to use iov on receiving size), but it's only a partial
3824 solution: when you have 2 dimensional data, the sizes of the second
3825 dimension is data of the first dimension, so you need 3 streams.
3827 I'm not sure the potential performance gains justify the coding burden.
3829 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3831 SimGrid (3.00) stable; urgency=high
3834 * New! Give the possibility to hijack the surf parser and thus bypass
3835 MSG_create_environment and MSG_launch_application. Have a look at
3836 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3838 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3840 SimGrid (2.96) unstable; urgency=low
3845 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3846 This deprecates the xbt_error_t mechanisms.
3847 It modifies (simplifies) all XBT and GRAS API.
3848 MSG API keeps unchanged (exceptions raised by XBT are caught from
3849 within MSG and masked with existing error handling facilities)
3852 * New! Add a FATPIPE model. [AL]
3853 * New! Add a parallel task model. [AL]
3854 * New! Add automatically a loopback interface (in the default
3855 network model) if none was precised.
3858 * Bugfix: MSG_process_resume now works with the current running process.
3860 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3861 * Modification of MSG_task_get_compute_duration. Once a task has been
3862 processed, the value returned by this function is now equal to 0. [AL]
3863 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3864 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3865 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3866 of a m_task. For now, MSG_task_cancel only works with computation
3868 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3869 of the processor (in Mflop/s) regardless of the current load on the
3871 * API Change: use proper naming convention for MSG_getClock and
3872 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3874 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3875 This function changes the priority of a computation task. This priority
3876 doesn't affect the transfer rate. A priority of 2 will make a task
3877 receive two times more cpu power than the other ones. This function
3878 has been added to suit the needs of Nguyen The Loc and hasn't been that
3879 much tested yet. So if it fails, please report it and send me your code.
3881 * API Change: removed all functions and types that were marked "deprecated"
3882 since many months. Renamed MSG_global_init_args to MSG_global_init.
3884 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3886 SimGrid (2.95) unstable; urgency=low
3889 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3890 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3891 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3894 * Bugfix: Allow absolute paths for platform description files [MQ]
3895 * Bugfix: do free the variables after use. Leads to drastic performance
3897 * Implement max_duration (ie, timeouts) on resources [AL]
3900 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3902 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3903 the needed support (timeouts) [AL]
3906 * Implement measurement sockets. You can now get the bandwidth between two
3907 hosts thanks to AMOK (see below). [MQ]
3908 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3909 dynar over the network (yeah) [MQ]
3910 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3911 * simple/static token-ring example. [Alexandre Colucci and MQ]
3912 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3913 in place before [MQ]
3916 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3917 type parsing. "short" and "long" modifiers were also ignored; other
3918 modifier (such as reference level) are still ignored. [MQ]
3919 * Embeed the buffer size within the buffer itself on SG. [MQ]
3920 That way, send() are atomic and cannot get intermixed anymore (at least
3921 the ones which are less than 100k; bigger messages still have the issue)
3922 * Array size pushed by the field, not by the field type (or each
3923 and every long int will push stuff to the cbps) [MQ]
3924 * use select() to sleep since it allows to portably sleep less than one
3927 GRAS (minor cleanups)
3928 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3931 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3932 * type category 'ignored' killed as it was never used and were difficult
3934 * whether a type can cycle or not is now a flag, leaving room for more
3935 flags (as "ignored", if we feel the need one day ;)
3936 * Rename raw sockets to measurement sockets since "raw" has another
3937 meaning in networking community.
3940 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3941 and offers features not belonging to GRAS but that most applications
3942 need. One day, it may be a set of plugins loadable at runtime.
3943 * New module: bandwidth
3944 bandwidth measurement between arbitrary nodes running this module. [MQ]
3946 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3948 SimGrid (2.94) unstable; urgency=low
3950 The first beta release of SimGrid 3 !
3954 * Update the main page and the FAQ. Adding references to gforge.
3957 * Add a gras_os_getpid function.
3960 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3961 * Extend the logs so that they also print PID, hostname, date, ... if
3963 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3964 and kill the old version which were in testsuite/
3965 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3969 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3970 for users. Instead, each programming environment (GRAS, MSG,...) use it
3971 when needed to provide such a feature to users.
3972 Don't shortcut the mechanism or you will also shortcut the virtualization
3973 you need on the simulator.
3977 * Cleanups in configury with regard to compile optimization/warning flags.
3978 Also add -fno-loop-optimize to any powerpc since it's the optimization
3979 killing gcc (< 3.4.0).
3980 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3981 needed by MSG examples complications
3982 * Borrow configury beautifications from PHP
3985 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3986 warning (thanks loris for stumbling into this one).
3987 * Bugfix: stop loading private headers (gras_config.h) from the public
3991 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3992 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3993 RL would imply a DNS resolver, which is *hard* to do in a portable way
3994 (and therefore delayed).
3995 * Implement a real timer mechanism and use it in timing macros. This allows
3996 to avoid rounding errors and get a 0.000005 sec precision in timing
3997 macros. While I was at it, various cleanups:
3998 - allow to declare more than one timed section per file (fix a stupid bug)
3999 - move some private declaration to the right place
4000 - merge conditional execution and timing macros into emulation module
4001 - document the module
4002 - make sure the module cleanups its mess on gras_exit
4003 * Documentation improvements:
4004 - (new) how to compile applications using GRAS
4005 - (new) emulation support (timing macros)
4007 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
4009 SimGrid (2.93) unstable; urgency=low
4011 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
4014 - Use Paje properly where used. Still to be sanitized properly.
4015 - Portability fix: Add an implementation of the contexts using pthread
4019 - Add xbt_procname(): returns the name of the current process.
4020 Use it to show the current process's name in all logging.
4022 - fix detection of older flex version and the reaction, since we do
4023 depend on modern ones (we use lex_destroy)
4024 - Better separation of SG and RL in the libs: remove all simulation code
4025 from libgras. As a result, this lib is now only 200k when stripped.
4026 Some of the xbt modules may also be duplicated (two sets and such) and
4027 should be cleaned/killed before SG3.
4028 - Insist on using xlC on AIX because of weird problems involving gcc there.
4029 - Cleanup the make remote stuff. This is now done by scripts
4030 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
4031 mainly for our private use, but we're working on changing them to user
4034 - Bugfix: flush the socket on close only if there is some *output*.
4035 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
4037 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
4040 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
4042 SimGrid (2.92) unstable; urgency=low
4044 Alpha 3 on the path to SimGrid 3
4048 - New! First try of benchmarking macros.
4049 - New! First try so that gras_stub_generator generate deployment and
4050 remote compilation helpers.
4052 - Bugfix: Initialization fix in msg_test.
4056 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
4058 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
4060 - API Change: gras_os_sleep to take the amount of seconds as a double.
4061 Accepting an int was error prone since it was the only location where
4062 seconds were coded as such. It leaded to damn rounding errors.
4063 - Bugfix: Hard to belive that timers ever worked before this.
4065 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4067 SimGrid (2.91) unstable; urgency=low
4069 Alpha 2 on the path to SimGrid 3
4073 - Bug fix in the lmm_solver.
4075 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4076 through the function MSG_paje_output.
4077 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4078 - It is possible to bound the rate of a communication in MSG with
4079 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4080 to write it in the changelog).
4081 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4086 - Various cleanups to the autotools stuff
4087 - Begin to move Gras examples to examples/gras/
4088 - Let make distcheck work again (yeah!)
4090 - documentation overhauled using doxygen.
4091 gtk-doc-tools is dead in SimGrid now.
4092 - Automatically extract all existing logging categories, and add the list
4093 to the documentation (long standing one, to say the less)
4095 - Cleanup the known architecture table. Reorder the entries to group what
4096 should be, and use a more consistent naming scheme.
4097 (some of the test dataset are still to be regenerated)
4098 - New! Allow library to register globals on each process just as userdata
4100 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4101 lookup time (for now).
4102 Use it in msg and trp.
4103 This cleans a lot the internals and helps enforcing privacy of the
4104 headers between the gras components.
4105 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4106 - Bugfix: gras_os_time was delirious in RL.
4107 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4109 - Reenable GRAS now that it works.
4111 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4113 SimGrid (2.90) unstable; urgency=low
4115 Alpha 1 on the path to SimGrid 3
4117 * It is a long time since the last release of SimGrid. I'm sorry about
4118 that but as I had told you, I was rewriting a lot of things. I apologize
4119 to those who had been reporting bugs to me and that I had not answered.
4120 If your bug is still in the new version, please tell me. Here is a
4121 summary of the main changes.
4123 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4124 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4125 portability, speed, and a lot more but you'll figure it out later.
4126 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4127 I wanted to release the new MSG as soon as possible and I have
4128 broken GRAS, which is the reason why, for now, only MSG is fully
4129 functional. A laconic description of these projects is available
4130 in the documentation.
4132 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4133 kernel optimized for our needs (called SURF but only the developers
4134 should use it). Hence, MSG is now roughly 30 times faster and I think
4135 that by rewriting a little bit MSG, I could event speed it up a little
4136 bit more. Beside the gain in speed, it is also much easier to encode a
4137 new platform model with SURF than it was with SG. More to come...
4139 * REVOLUTION 3: I have tried to change a little as possible the API of
4140 MSG but a few things really had to disappear. The main differences
4141 with the previous version are:
4142 1) no more m_links_t and the corresponding functions. Platforms are
4143 directly read from an XML description and cannot be hard-coded
4144 anymore. The same format is used for application deployment
4145 description. The new format is described in the documentation.
4146 Have a look in tools/platform_generation. There is a tiny script
4147 that converts from the old platform format to the new one. Concerning
4148 the application deployment format, parsing the old one is tricky.
4149 I think most of you should however be able to convert your files. If
4150 it is really an issue, I can write a C code that does the conversion.
4152 2) the toolbox tbx does not exist anymore. We now have a library
4153 with much more data-structures but without the hash-tables (we have
4154 dictionaries that are much faster).
4156 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4158 *****************************************************************************
4159 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4160 * don't feel like converting the oldies. (Mt) *
4161 *****************************************************************************
4164 Version 2.90: "the long awaited one"
4165 - Finished rewriting and debugging MSG. Rewrote the documentation.
4166 - disable GRAS for now since it needs to be ported to the newest SG
4169 - Finish the port to windows (using mingw32 for cross-compile)
4172 - Main loop and datastructures of SURF. A cpu resource object is
4173 functional. Surf can thus be used to create cpu's with variable
4174 performance on which you can execute some actions.
4176 2004-11-15 Martin Quinson
4177 - Port to ARM. Simply added the alignment and size descriptions. Should
4178 work, but the ARM machines are so slow that I didn't had the opportunity
4179 to 'make check' over there yet.
4181 2004-11-15 Arnaud Legrand
4182 - Trace manager now written. It uses a heap structure and is therefore
4183 expected to be efficient. It may however be speeded up (particularly
4184 when many events occur at the same date) by using red and black
4185 trees. One day maybe...
4186 - Max-min linear system solver written. It uses a sparse matrix
4187 structure taking advantage of its expected use. Most operations are
4188 O(1) and free/calloc are called as few as possible. The computation of
4189 the minimum could however be improved by using a red and black tree
4192 2004-11-03 Arnaud Legrand
4193 - Rename every gras_* function that was in xbt/ to its xbt_
4195 - Add a heap and a doubly-linked list to xbt
4196 - Added a dichotomy to the dictionaries. make check works as well before
4197 so I assume that the patch is correct. I do not know however if things
4198 run effectively faster than before now. :)
4200 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4201 SimGrid, and the version number is bumped to 2.x
4203 2004-10-29 Martin Quinson
4204 - Introduction of the remote errors.
4205 They are the result of a RMI/RPC on the remote machine.
4206 ErrCodes being scalar values, you can't get the host on which those
4207 errors did happen. Extending the error mechanism as in Gnome is possible.
4208 No idea yet whether it is a good idea.
4210 2004-10-28 Martin Quinson
4211 - Interface revolution: the Starred Structure Eradication.
4212 I used to do typedef struct {} toto_t; and then handle *toto_t.
4213 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4214 - ???_t is a valid type (builded with typedef)
4215 - s_toto_t is a structure (access to fields with .)
4216 - s_toto is a structure needing 'struct' keyword to be used
4217 - e_toto_t is an enum
4218 - toto_t is an 'object' (struct*)
4220 typedef struct s_toto {} s_toto_t, *toto_t;
4221 typedef enum {} e_toto_t;
4222 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4223 s_toto_t) is private.
4225 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4226 so that it can set the variable to NULL. It was so for dicts and sets,
4227 it changed for dynars.
4229 - Fix a bunch of memleaks in dict_remove
4230 - Fix a bug in sg/server_socket opening: it failed all the time.
4232 2004-10-07 Martin Quinson
4233 - Speed up dynar lookup operation a bit.
4235 gras_dynar_get is dead.
4237 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4238 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4239 (which gives you the address of the stored data).
4241 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4242 copy of the data using an affectation to do the job and not a memcpy.
4244 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4245 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4247 It does not really speedup the dynar test because they are
4248 setting elements all the time (and look them seldom). But the dict does
4249 far more lookup than setting.
4251 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4253 2004-10-05 Martin Quinson
4254 - Allow to (en/dis)able the cycle detection at run time.
4256 Whether we should check for cycle or not is now a property of each
4257 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4258 datadesc_cycle_unset allow to remove this property when previously set.
4260 Note that the cycle detection is off by default since it impacts the
4261 performance. Watch the data you feed GRAS with ;)
4263 This property is hereditary. Any element embedded in a structure having it
4264 set have it set for the time of this data exchange.
4266 You should set it both on sender and receiver side. If you don't set it on
4267 sender side, it will enter an endless loop. If you forget on receiver
4268 side, the cycles won't be recreated after communication.
4270 - Header reorganization.
4271 Kill gras_private.h, each submodule must load the headers it needs.
4273 2004-10-04 Martin Quinson
4274 - Interface revolution: do not try to survive to malloc failure.
4276 Now, gras_malloc and friends call gras_abort() on failure.
4277 As a conclusion, malloc_error is not a valid error anymore, and all
4278 functions for which it was the only gras_error_t return value are
4279 changed. They now return void, or there result directly.
4280 This simplify the API a lot.
4282 2004-09-29 Martin Quinson
4283 - Re-enable raw sockets.
4284 Created by gras_socket_{client,server}_ext;
4285 Used with gras_raw_{send,recv}
4288 It should allow to kill the last bits of gras first version soon.
4290 This is not completely satisfactory yet (duplicate code with
4291 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4294 - Simplify transport plugin (internal) interface by not passing any
4295 argument to _server and _client, but embedding them in the socket
4298 2004-09-28 Martin Quinson
4299 - Finish the port to AIX.
4300 autoconf was my problem (segfault within the malloc replacement
4301 function. No idea why)
4303 2004-09-16 Martin Quinson
4304 - Fix some size_t madness on 64bit architectures.
4306 2004-09-08 Martin Quinson
4307 - Reduce the number of system headers loaded, overload some more system
4308 calls (such as malloc to cast the result of the system one, and work
4310 - Fix and reintroduce the config support
4312 2004-09-07 Martin Quinson
4313 - Source code reorganization to allow Arnaud to surf all over there.
4314 - Allow to document the logging categories.
4315 - Remove all uppercase from logging categories and useless cleanup in names.
4317 2004-08-18 Martin Quinson
4318 Version 0.6.2 (protocol not changed; API changed)
4319 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4320 gras_error_t), and NULL when not found. Functions expecting a msgtype
4321 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4322 hopefully usefull message.
4323 - Portability to prehistoric sparcs again
4325 2004-08-17 Martin Quinson
4326 Version 0.6.1 (protocol not changed; ABI not changed)
4327 - prealloc some buffers to speed things up
4329 2004-08-11 Martin Quinson
4330 Version 0.6 (protocol not changed; ABI expended)
4331 - The parsing macro can deal with the references, provided that you add
4332 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4334 2004-08-09 Martin Quinson
4335 Version 0.5 (protocol not changed; ABI changed)
4336 - Allow to off turn the cycle detection code in data exchange at
4337 compilation time. It should be at run time, but I'm short of time (and
4338 the config stuff is still broken). That way, we keep dict out of the
4339 critical path, which is good because the performance is poor:
4340 - search not dichotomial yet
4341 - dynar give no way to access their content and memcpy everytime
4342 - In composed data description (struct, ref and so on), stop foolness of
4343 keeping the subtype's ID, but store the type itself. This keeps sets out
4344 of the critical path, which is good since they rely on dynar and
4345 dictionaries. The only loose of that is that we cannot detect the
4346 redeclaration of a structure/union with another content (but I'm not sure
4347 the code detected well this error before anyway). We still can detect
4348 the redefinition discrepancy for the other types.
4349 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4350 it breaks the code because of type-punning used all over the place).
4351 This breaks on all non-gcc architectures (for now).
4353 All those changes (plus the buffer of last time) allow me to gain 2 order
4354 of magnitude on cruel tests consisting of 800000 array of integers on two
4355 level of a hierarchical structure (200 secondes -> 4 secondes)
4358 - the selector of reference must now return the type it points to, not
4359 the ID of this type.
4361 2004-08-06 Martin Quinson
4362 Version 0.4 (protocol changed; ABI not changed)
4363 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4364 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4365 - Implement a buffer transport, and use it by default (it relies on tcp in
4366 real life and on sg in simulation).
4367 That's a bit hackish since I had a new field to the structure to store
4368 its data without interfering with the subtype ones. Inheritance
4369 is tricky in C. And that's a kind of reverse inheritance with one class
4370 derivating two classes. Or maybe a game with java interfaces. Anyway,
4371 that's damn hard in C (at least).
4372 Moreover, I got tired while trying to ensure plugin separation and
4373 genericity in SG mode. MSG wants me to do weird things, so let's go for
4374 cruel hacks (temporarily of course ;).
4375 See comment in transport_private.h:71
4376 - do not include all the _interface headers in private but in the files
4377 which really need them (to cut the compilation time when they are
4380 2004-07-26 Martin Quinson
4381 Version 0.3 (protocol not changed; ABI changed)
4382 - Major overhault of the datadesc interface to simplify it:
4383 - shorted the function names:
4384 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4385 - add a trivial way to push/pop integers into the cbps without malloc.
4386 This allows to make really generic sub_type description, which simply
4387 pop their size of the stack.
4388 - add a function gras_datadesc_ref_pop_arr() which does what users want
4389 most of the time: Declare a dynamic array (which pops its size of the
4390 stack) and declare a reference to it. Poor name, but anyway.
4391 - kill the post-send callback, add a post-receive one
4393 2004-07-23 Martin Quinson
4394 Version 0.2 (protocol changed; ABI changed)
4395 - add some testing for cpbs in the test cases, and fix some more bugs.
4396 This invalidate again the little64 data file, since I cannot regenerate
4398 - remove an awfull optimization in the logging stuff, allowing me to:
4399 - understand it again
4400 - learn gcc how to check that the argument match the provided format
4401 - fix all errors revealed by gcc after that
4402 - internal keys of dict are not \0 terminated. Deal with it properly in
4403 loggings instead of segfaulting when the user want to see the logs :-/
4405 2004-07-22 Martin Quinson
4406 - Fix some stupid bug preventing cbps (callback postit) from working
4408 2004-07-21 Martin Quinson
4409 - Some documentation cleanups
4410 - remove the useless last argument of msgtype_declare
4411 - rename the Virtu functions to fit into the 'os' namespace
4412 - move headers src/include -> src/include/gras/ and stop fooling with
4413 gras -> . symbolic link
4414 - make distcheck is now successful
4416 2004-07-19 Martin Quinson
4418 - Build shared library also
4419 - Install html doc to the right location
4420 - stop removing maintainer files in make clean
4421 - build tests only on make check
4423 2004-07-13 Martin Quinson
4425 - No major issue in previous version => change versioning schema
4426 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4427 corresponding dataset.
4429 2004-07-11 Martin Quinson
4431 - Get it working with any kind of structure (we can compute the padding
4432 bytes remotely for all the architectures I have access to)
4433 - Implement the structure parsing macro (still not quite robust/complete)
4434 - Improvement to the remote testing toysuite
4436 2004-07-10 Martin Quinson
4437 [autoconf mechanism]
4438 - get ride of a bunch of deprecated macros
4439 - actually run the test for two-compliment, not only compile it :-/
4440 - test whether the structures get packed (and bail out if yes. Damn.
4441 Alignment is a serious matter)
4442 - test whether the structures get compacted (but respecting the alignment
4443 constraints of each types)
4444 - test whether the array fields of structures can straddle alignment boundaries
4446 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4447 (took me 2 hours to find that bug, looking at the wrong place)
4448 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4449 before sending/receiving and are used to compute the offsets of fields
4450 - Given that padding size depend even on compiler options, keep track of
4451 alignment and aligned_size only for the current architecture. Not a big
4452 deal since we send structure fields one after the other (seems
4454 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4455 program, let it work on linux/gcc/little32. portability todo.
4457 2004-07-08 Martin Quinson
4458 - import and improve remote compilation support from FAST
4459 - make sure make check works on half a dozen of machines out there
4461 2004-07-07 Martin Quinson
4462 Let's say it's version 0.0.3 ;)
4463 - Implement conversions (yuhu!)
4464 - Let it work on solaris (beside conversion, of course)
4465 - Stupid me, using rand() to generate the conversion datatests in not wise.
4467 2004-07-06 Martin Quinson
4468 - Let make dist work, since I'm gonna need it to compile on remote hosts
4469 - Let Tests/datadesc_usage write the architecture on which the file was
4470 generated as first byte.
4471 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4472 architecture descriptions.
4473 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4474 on those architectures.
4475 - Optimization: send/recv array of scalar in one shoot
4477 2004-07-05 Martin Quinson
4478 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4480 - Plug a whole bunch of memleaks
4481 - each process now have to call gras_{init,exit}. One day, their log
4482 settings will be separated
4483 - Continue the code factorisation between SG, RL and common in Transport.
4485 2004-07-04 Martin Quinson
4487 - Redistribution between SG and RL.
4488 We wanna have to accept in SG, so move accepted related parts of RL in
4489 the common part. (more precisely, the dynar of all known sockets is no
4490 more a static in transport.c, but part of the process_data)
4492 [gras_stub_generator]
4493 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4495 2004-07-03 Martin Quinson
4496 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4497 them all up in one shot)
4499 - Fix the ugly case of reference to dynamic array.
4500 - New (semi-public) function gras_datadesc_size to allow the messaging
4501 layer to malloc the needed space for the buffer.
4503 - gras_socket_close now expect the socket to close (and not its address to
4504 put NULL in it after it). This is because the socket passed to handlers
4505 is one of their argument (=> not writable).
4507 - propagate the interface cleanup from last week in datadesc, ie remove a
4508 superfluous level of indirection. User pass adress of variable
4509 containing data (both when sending and receiving), and not of a variable
4510 being a pointer to the data. Let's say that I like it better ;)
4511 The price for that is constructs like "int msg=*(int*)payload" in
4512 handlers, but it's a fine price, IMHO.
4514 - Let it work in RL (yuhu)
4516 2004-06-21 Martin Quinson
4518 - porting SG plugin and SG select to new standards (works almost).
4519 - plug memleaks and fix bugs around.
4522 - cleanup the prototype of data recv and force users to specify when they
4523 want to handle references to objects. Test case working even for cycles.
4524 - plug memleaks. Valgrind is perfectly ok with this.
4526 2004-06-12 Martin Quinson
4528 - cleanup the separation between plugin and main code in plugin creation
4530 2004-06-11 Martin Quinson
4532 - Reput hook for raw sockets, needed for BW experiments
4533 - kill a few lines of dead code
4534 [Data description] Interface cleanup
4535 - gras_datadesc_by_name returns the searched type or NULL.
4536 That way, no variable is needed to use a type desc once, which makes
4538 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4539 two parameters were strings (field name, type name), leading to
4541 [Dicos] Interface cleanup
4542 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4543 This is consistant with the dynar API.
4545 2004-04-21 Martin Quinson
4547 - Porting to new standards.
4549 - interface cleanup.
4550 There is no bag anymore, no need to take extra provision to mask the
4551 pointers behind "ID".
4552 Better splitup of functions between files create/exchange/convert.
4553 This is still a bit artificial since convert and receive are so
4554 interleaved, but anyway.
4556 - add a queued message list to procdata (the ones not matching criteria
4558 - factorize some more code between SG and RL wrt procdata
4560 - use gras_exit in example to track memleaks
4561 - get rid of gs_example now that GS is properly integrated into gras
4562 - update run_test to integrate the latest tests (datadesc)
4564 - rename WARNINGn macros to WARNn since it prooved error-prone
4566 2004-04-19 Martin Quinson
4568 - register init/exit functions within gras module mechanism
4569 - send/receive function.
4570 Convertion is not implemented, but short-cutted if not needed.
4571 struct/array elements are sent one by one (instead of block-wise), but
4572 nobody really cares (yet). Get a prototype before optimizing.
4573 - tests (using a file socket) for DD send/receive on:
4574 - base types: int, float
4575 - array: fixed size, string (ie ref to dynamic string)
4576 - structure: homogeneous, heterogeneous
4577 - chained list, graph with cycle
4578 Believe it or not, valgrind is not too unhappy with the results. The
4579 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4583 - Bugfix when using a filename explicitly (instead of '-')
4585 2004-04-09 Martin Quinson
4587 - factorize more code between RL and SG in socket creation
4588 - Complete the implementation and tests of:
4590 o file (only in RL, and mainly for debugging)
4592 I lost 3 days to design a portable address resolver, and then decided
4593 that the prototype mainly have to run on my box.
4594 Addressing portability too early may be like optimizing too early :-/
4596 - use gras_init in the Tests instead of the crappy parse_log_opt
4597 (the latter function is removed)
4598 [Conditional execution]
4599 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4600 [Code reorganisation]
4601 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4602 Build examples against the RL library, since there is no way to disable
4603 its creation for now.
4605 For information, the beginning of coding on GRAS was back in june
4606 2003. I guess that every line has been rewritten at least twice since