1 SimGrid (3.30.1) NOT RELEASED YET (v3.31 expected March 20. 2022, 15:33 UTC)
3 ----------------------------------------------------------------------------
5 SimGrid (3.30) January 30. 2022.
7 The Sunday Bloody Sunday release.
9 Main user-visible changes:
10 - The SimDag API for the simulation of the scheduling of Directed Acyclic
11 Graphs has been dropped. It was marked as deprecated for a couple of years.
12 We finally complete the implementation of what has been called SimDag++
13 internally, i.e., porting the different features of SimDag on top of S4U.
14 The new way to simulate the execution of dependent activities directly by
15 maestro (without any other actor) is details in the examples/cpp/dag-* series
17 - The removal of SimDag led us to also remove the export to Jedule files that
18 was tightly coupled to SimDag. The instrumentation of DAG simulation is still
19 possible through the regular instrumentation API based on the Paje format.
20 - We also dropped the old and clumsy Lua bindings to create platforms in a
21 programmatic way. It can be done in C++ in a much cleaner way now, which
22 motivates this suppression.
25 - Introduce on_X_cb() functions for all signals, to attach a new
26 callback to the signal X. The signal variables are now hidden and
27 only these functions should be used.
28 Rationale: this enables the usual deprecation scheme where functions
29 remain for 4 releases if we need to modify the signals, while the
30 current code with the signal variables directly visible prevents any
32 - New function: Engine::run_until(date), to split the simulation.
33 - New signal: Activity::on_veto, to detect when an activity fails to start.
34 - Signal change: Comm::on_start(Comm&, bool) has been replaced by
35 Comm::on_send and Comm::recv. These two signals respectively correspond to
36 when the sending or receiving side of a Comm is ready. They are raised at
37 the same locations as the former Comm::on_start signal.
38 - New function: Engine::track_vetoed_activities() to interrupt run()
39 when an activity fails to start, and to keep track of such activities.
40 Please see the corresponding example for more info.
41 - New functions: s4u::Comm::{sendto_init, set_source, set_destination} to enable
42 the use of vetoers with direct host-to-host communications. Both source and
43 destination have to set for a comm to start. Each call to these setters check
44 if all vetoes are satisfied. When it is the case, the comm starts. A use case of
45 these functions is given in examples/cpp/dag-scheduling.
46 - New functions: {Exec, Io}::update_priority allow you to modify the priority of
47 these kinds of activities during their execution. Behavior is detailed in
48 examples/cpp/io-priority/.
51 - Dynamic costs for MPI operations: New API to allow users to dynamically
52 change injected costs for MPI_Recv, MPI_Send and MPI_Isend operations.
53 Alternative for smpi/or, smpi/os and smpi/ois configuration options.
54 - Fix some issues with the replay mechanism.
57 - Function xbt::Extendable::get_data() is now templated with the type of the
58 pointee. Untyped function is deprecated. Use get_data<void>() if you still
59 want to retrieve void*.
62 - New section: "SimGrid MPI calibration of a Grid5000 cluster"
63 presenting how to properly calibrate MPI communications in SimGrid.
64 - Complete and reword the platform section, which is now completed.
67 - Thread contexts are used by default with Python bindings. Other kinds of
68 contexts revealed unstable, specially starting with pybind11 v2.8.0.
70 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
71 (FG: issues on Framagit; GH: issues on GitHub)
72 - FG#95: Wrong computation time for multicore execution after pstate change
73 - FG#97: Wrong computation time for ptask+multicore+pstates
74 - FG#98: SMPI offline simulation is inconsistent with the online simulation
75 (deadlocks / message truncation)
76 - FG#99: Weird segfault when not sealing an host
78 ----------------------------------------------------------------------------
80 SimGrid (3.29) October 7. 2021
82 The "Ask a stupid question" release.
84 We wish that every user ask one question about SimGrid to celebrate.
85 On Mattermost, Stack Overflow or using the issues tracker.
88 New modeling features:
89 - Non-linear resource sharing, modeling resources whose performance heavily degrades with contention:
90 - The total capacity may be updated dynamically through a callback
91 and depends mainly on the number of concurrent flows.
92 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
94 - Dynamic factors: model variability in the speed of activities
95 - Each action can now have a factor that affects its progression.
96 This multiplicative factor is applied when updating the amount of work
97 remaining, thereby an activity with factor=0.5 only uses half of the
98 instantaneous power/bandwidth it is allocated and will appear twice
99 slower than what it actually consumes.
100 - This can be used to model a overhead (e.g., there is a 20 bytes
101 header in a 480 bytes TCP packet so the factor 0.9583) but the novelty
102 is this factor can now easily be adjusted depending on activity's and
103 resources characteristics.
104 - This existed for network (e.g., the effective bandwidth depends
105 on the message in SMPI piecewise-linear network model) but it is now
106 more general (the factor may depend on the source and destination and
107 thus account to different behaviors for intra-node communications and
108 extra-node communications) and is available for CPUs (e.g., if you
109 want to model an affinity as in the "Unrelated Machines" problem in
110 scheduling) and disks (e.g., if you want to model a stochastic
112 - For that, resources can be provided with a callback that computes
113 the activity factor when creating the action.
114 - Example: examples/cpp/exec-cpu-factors
115 - The same mechanism is also available for the latency, which
116 allows to easily introduce complex variability patterns.
119 - Added support to programmatic platform creation in Python.
120 Example: examples/python/clusters-multicpu
123 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
124 This can only be set through the API, not through XML files.
128 - Multiple fixes to ensure reproducibility of tracing
129 - scan/excan can now be replayed
130 - wait action now uses ranks and not pid, as the other ones.
131 - smpi/init and smpi/finalization-barrier are now valid for replays.
132 - exit() is now intercepted by SMPI to avoid premature shutdown of
133 simulation. First non 0 return codes is returned as simulation return
137 * New section "Release Notes" documenting recent and current developments.
138 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
139 * Improvements in API Reference for C++ and Python interfaces.
142 - Make wifi creation compatible with ns-3 version 3.34 too.
144 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
145 (FG: issues on Framagit; GH: issues on GitHub)
146 - FG#77: Search feature of doc is broken (update sphinx theme version)
147 - FG#78: Multiple fixes for SMPI replay:
148 - TI tracing of allotallv/w was outputting wrong values
149 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
150 ----------------------------------------------------------------------------
152 SimGrid (3.28) July 14. 2021
154 The Victoriadagarna Release.
157 - C++ platform interface: Users can now describe their platform directly in C++.
158 This provides greatly flexibility and performance improvement for complex
159 platforms. Main features:
160 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
161 "zones", instead of single hosts. This feature enables the description
162 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
163 - StarZone: new zone with a Star-like topology. The routes are defined
164 as a set of links used to communicate from node to everybody (node<->ALL).
165 - Split-Duplex links: auxiliary method to create split-duplex links in
166 the platform, easing its utilisation. It automatically creates both UP
167 and DOWN links (similarly as done in XML).
168 - Please refer to the documentation and the examples included:
169 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
170 - New plugin: Producer-Consumer with monitor. Just requires to include the
171 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
172 associated example (examples/cpp/plugin-prodcons).
175 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
176 s4u::Io::wait_any(), s4u::Io::wait_any_for().
177 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
178 parameter by reference, instead of a pointer.
179 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
180 Explicitly cancel the activity to get back to previous behavior.
181 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
184 - The default SMPI compiler flags are no more taken from the environment.
185 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
186 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
188 --cfg=smpi/finalization-barrier: which can be used to add
189 a barrier inside MPI_Finalize. This can help for some codes which cleanup
190 data attached to a process, but still used in other SMPI processes.
191 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
192 is active when set to false, to keep going after a small error
193 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
194 which may or may not be problematic in the end.
196 - fix behaviour, as maximum iteration count could be ignored
197 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
198 use sampling when the same kernel is called with a different set of parameters
199 which have an impact on the timing.
200 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
201 It should now work with smpi/auto-shared-malloc-thresh.
202 - Improve error handling and reporting in multiple places
203 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
204 validity checks, truncated messages are now an error, return errors
205 when explicitely deleted handles are reused, ...)
206 - RMA: multiple fixes and stability improvements.
207 - analysis (-analyze flag in smpirun):
208 - SMPI can now report buffer leaks as well as MPI handles leaks,
209 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
210 - if -trace-call-location is used when compiling, SMPI can report
211 origin of leaked handles/buffers
212 - group leaks by type/origin in output message if possible
213 - New implemented MPI calls: MPI_Comm_test_inter
216 - Changed internal implementation of bandwidth factors in network models.
217 Models affected: CM02, LV08 (default), SMPI, IB.
218 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
219 Bandwidth factors are applied to communications to describe that users
220 cannot use 100% of the available bandwidth. For example, the default network model,
221 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
222 behavior was implemented by limiting the bandwidth available in the LMM
223 system for this flow. This may give the false impression that there is
224 bandwidth available for other flows due to its underutilization, especially
225 for the dynamic bandwidth factors used in SMPI models.
226 To avoid this, we have modified the implementation so that each flow uses the
227 maximum physical bandwidth according to the LMM system.
228 However, the actual throughput of the flow seen by the user is defined by
229 the physical bandwidth multiplied by the bandwidth factor.
230 This change impacts on the simulation results for all network models on
231 which we have bandwidth factors configured.
232 *****************************************
233 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
234 *****************************************
235 This change may impact on the timing of your simulation results.
236 Take care when comparing simulations from different SimGrid's
237 versions. Sorry for the inconvenience.
238 - Dynamic network factors: users can configure a callback to define
239 the network factors dynamically. This API is available at
240 simgrid::kernel::resource::NetworkModelIntf.
241 - Users have access to complete information about the current communication
242 to decide which factor to apply. This includes: message size, source and
243 destination hosts, links and zones traversed.
244 - Dynamic factors for both latency and bandwidth.
245 - For more details, see the example in (examples/cpp/network-factors).
246 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
247 deprecated since version 3.24, are no longer supported. Instead, use
248 "wattage_off" and "wattage_per_state".
251 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
254 - New tutorial: Model-checking and formal assessment
255 - New sections: "Demystifying the routing" and "C++ platforms"
256 - Update and improve the part on visualization in MPI and Algo tutorials.
257 - Improve the section on routing: how to define it, how it's used internally
258 - Fix many issues, broken links and missing references in doxygen and Sphinx
261 - Lua platform files are deprecated. Their support will be dropped after v3.31.
264 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
266 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
267 (FG: issues on Framagit; GH: issues on GitHub)
268 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
269 - FG#64: Configuring smpi/IB-penalty-factors
270 - FG#67: Running computation concurrently with MPI_Iallreduce
271 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
272 - FG#71: Segmentation fault on invalid gw_src/gw_dst
273 - GH#322: Issue when an actor kills his host vm
275 ----------------------------------------------------------------------------
277 SimGrid (3.27) March 29. 2021
279 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
281 Main user-visible changes:
282 - All functions marked as deprecated for one year were removed.
283 - Removed all that was related to the "Storage" type in all APIs and bindings.
284 "Disks" have been introduced more than a year ago to replace them for the better.
285 It's time to finish this replacement.
286 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
287 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
288 avoiding "initializer element is not constant" errors with some codes. Please be
289 careful to recompile your MPI codes if switching to this new SimGrid version.
290 - Network model ns-3 now requires version ns-3.28 or higher.
291 - Removed some option aliases that were deprecated since v3.23 (June 2019)
292 - 'verbose-exit' is now 'debug/verbose-exit'
293 - 'simix/breakpoint' is now 'debug/breakpoint'
294 - 'clean-atexit' is now 'debug/clean-atexit'
297 - New functions: Comm::sendto_{init,async} to initiate a communication
298 on between two (possibly remote) hosts.
299 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
300 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
301 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
302 or Mailbox::get_async<void>() if you really want to play with void*.
303 - Unify the interface of Activity::on_{start/activity}
304 - New function: Comm::get_dst_data()
305 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
308 - ptask should now work better with multicore hosts (FG#37).
309 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
310 - Internal cleanups and refactoring around models, to prepare for programmatic
312 - Modernize our tests of the performance models
316 - New section: Configuring the logs from the command line.
317 - New section: Programming API of logs.
318 - New section: Getting realistic results
319 - New section: Examples of SMPI Usage
320 - New part: SimGrid's Internals, with a section on Design goals
321 - Move back to the classical Breathe tool chain, and fix many small glitches.
324 - Handle leak checking: SMPI will now display a message if MPI handles are still
325 present at the end of a simulation. Option smpi/list-leaks:n will display the n
326 first leaked pointers and their types.
327 - smpi/display-allocs option : Display Memory allocations analysis, with hints
328 about the largest allocations. This only works for allocations with malloc/calloc
329 and not compiled with SMPI_NO_OVERRIDE_MALLOC
330 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
331 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
332 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
333 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
334 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
335 - Replay now opens and parses the trace file only once for all ranks when possible.
337 - better handling of MPI_PROC_NULL
338 - MPI_Group_incl was broken in some cases
339 - PAPI tracing now fixed
340 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
341 - SMPI will now fail properly when receive size < sent size, instead of
342 just setting MPI_ERR_TRUNCATE and keep going.
343 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
344 - Remove useless build patches and split build/run patches on the page
345 - Build in parallel when possible
346 - Test on yum-based systems
347 - New nightly tested Proxy-apps:
348 - OpenMPI mpi-test-suite
349 - Coral MCB, LSMS, Qbox, CLOMP
350 - ECP/Mantevo Cloverleaf3D, Ember
351 - Codevault bhtree_mpi
354 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
355 - Start to overhaul the implementation to modernize it (still ongoing).
358 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
359 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
360 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
361 - Deprecate xbt_dynar_to_array().
362 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
364 C binding and interface:
365 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
366 - Constness requirement has been lowered for parameter argv of sg_actor_start()
367 and sg_actor_create() It's now (char* const*). The previous version is still
368 available as sg_actor_start_() and sg_actor_create_().
370 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
371 (FG: issues on Framagit; GH: issues on GitHub)
372 - FG#37: Parallel tasks are limited to 1 core per host
373 - FG#62: Running "smpirun -replay" on large networks
374 - FG!46: Fix a few potential memory leaks in SMPI colls
375 - GH#343: Wrong consumed energy when an exec timeouts
377 ----------------------------------------------------------------------------
379 SimGrid (3.26) December 15. 2020
381 The Release release (the French lockdown was eased today).
383 Important user-visible changes:
384 - SimGrid now requires a compiler with C++14 support.
385 Sibling projects should upgrade their FindSimgrid.cmake
386 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
387 several users had difficulties to understand issues when using high bandwidth or
388 small latency events. The new value was already the default for SMPI and
389 should not cause too much performance hit. It can change some simulation
390 timings if simulation was skipping a lot of small events (which was actually
391 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
392 - On some network models (L07, CM02, LV08), an implicit loopback link is created
393 for local communication. Default bandwidth was 500MBps and latency was 15µs,
394 which was rather bad. Default values have been changed to 10GBps and 0 latency.
395 Options network/loopback-lat and network/loopback-bw have been added to adapt
396 these values if needed. Network and routing models already providing separate
397 loopbacks (clusters and such) are not impacted by this change.
398 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
399 metadata that are added to each MPI message by MPI libraries (tag,
400 src/dest, and communicator information). This may change simulation times
401 for very small message sizes, but should be more realistic.
404 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
405 - New plugins: wifi_energy and link_load. See the associated examples.
406 - New examples: wifi networks, and communication suspend.
409 - LTO is now enabled for Intel/clang compilers.
410 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
411 Setting it to "auto" will use all available cores, while setting it to n will
412 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
413 - Remove obsolete runtime option 'contexts/parallel-threshold'.
414 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
415 has been kept for compatibility.
416 - Finally remove obsolete snake_case() aliases for runtime options.
417 - Further improve the documentation.
420 - the on_exit() of each actor is also executed when the simulation deadlocks.
421 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
422 An example is provided for s4u::Comm but it should work with Exec and Io.
425 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
426 https://framagit.org/simgrid/SMPI-proxy-apps
427 - MPI/IO: supports names starting with "./"
428 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
430 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
431 building shared libraries, this may cause issues, so environment variable
432 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
433 - most temporary files should now be created in /tmp dir (or equivalent).
434 If this one does not allow execution of code (noexec flag), this may cause issues.
435 Please use another tmp directory (using TMPDIR or equivalent system variable)
439 - A large refactoring is ongoing, but this should be transparent for users so far.
442 - Drop xbt_str_split().
444 C binding and interface:
445 - The return type of the sg_actor_on_exit() callbacks is now 'void'
446 (int value was previously ignored)
447 - Many C functions were renamed to follow a common naming style
448 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
449 names are kept until version 3.30.
450 - Many MSG tests were converted to the new S4U's interface in C, that
451 was extended for that.
453 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
454 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
455 - FG#41: Add sg_actor_create C interface
456 - FG#43: xbt::random needs some care
457 - FG#48: The Impossible Did Happen (yet again)
458 - FG#50: Suspending an actor executed at the current timestamp fails
459 - FG#52: Zero-seconds timeout : "That's in the past already"
460 - FG#53: Crash while using ns-3 network model
461 - FG#54: How to suspend a comm?
462 - FG!22: Stochastic Profiles
463 - FG!24: Documentation and fix for xbt/random
464 - FG!35: Add a modeling hint for parallel links in doc
465 - FG!36: [xbt/random] Read/Write the state of the RNG
466 - FG!42: Link cumulated load plugin
467 - FG!44: Test condition variable wait_until+notify_all
468 - GF#18137: Allow different stack sizes?
469 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
470 - GH#139: Allow pthread creation in SMPI
471 - GH#336: Packet-level simulation using SMPI?
472 - GH#345: Error in the LMM while migrating a VM
473 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
474 - GH#352: pip install / python setup.py install fail to find pybind11
475 - GH!337: Fix link_energy plugin for wifi platforms
476 - GH!339: Add Mailbox set_receiver method to python binding
477 - GH!344: Cast hugepages macros parameters to int64
479 ----------------------------------------------------------------------------
481 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
483 The Palindrom Day Release.
485 Important user-visible changes:
486 - Improve the Python usability (stability and documentation).
487 - A nasty synchronization bug was ironed out, see also below.
488 - Python's doc was organized as a separate tree, now integrated with C++.
489 - C bindings of S4U were not part of the doc.
490 - The C++ doc was also improved as methods are now split by theme.
491 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
492 - This is mandatory to use the Java bindings.
493 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
494 - SimDAG++: Automatic dependencies on S4U activities (experimental)
495 - Some features are already implemented but not all of them
496 - Cannot block an activity until it's scheduled on a resource
497 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
498 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
501 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
502 - Drop support for Viva/Triva (old visualization tools), hence removed
503 TRACE_get_node_types() and TRACE_get_edge_types() functions.
506 - Require pybind11 version 2.4.
507 - Greatly improve locking during context switches with Python.
508 - Declaring Python callbacks to this_actor::on_exit is now working.
509 - No more tests are failing currently, even on win32.
512 - Actor: Merge signals on_migration_start/end into on_host_change
513 - Actor: Rename migrate() into set_host()
514 - Disk: Allow users to get the read and write nominal bandwidth values
515 - Exec: Implement wait_for(timeout)
516 - Io: Implement wait_for(timeout)
519 - Parse errors now raise a simgrid::ParseError that you may want to catch.
522 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
523 been renamed to fix a spelling error. As usual, the previous definitions are
524 kept with a deprecation warning. However, to avoid runtime errors, it is now
525 forbidden to override these deprecated methods in a derived class. Please use
526 the new names immediately if you need to override them.
529 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
530 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
531 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
532 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
533 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
534 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
535 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
536 - New module: random, providing classical random numbers generators.
539 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
540 allocations' size above which they are considered "shared" by default (as if
541 they were performed through SMPI_SHARED_MALLOC macros).
542 Default = 0 = disabled feature.
543 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
544 This can cause some troubles if codes are already overriding these. If this
545 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
546 help, but will make this feature unusable.
547 - Input check is now performed much more consistently and outputs precise
548 warnings for each error found
549 - New supported calls : MPI_File_set_view, MPI_File_get_view
550 - MPI I/O now only supports disks, as storage is discontinued.
551 - MPI I/O allows opening files without specifying absolute path
552 - Note : SMPI configuration options are now initialized by smpi_init_options()
553 call, which should be called if SMPI is not being used through smpirun.
555 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
556 - FG#9: Python bindings crashing
557 - FG#39: Missing s4u::Comm::wait_any_for example
558 - FG#42: Add support for ThreadSanitizer (TSan)
559 - FG!19: Removing RngStream
560 - FG!20: A module for RNG calls
561 - FG!21: Choice between ad-hoc and standard distributions implementations
563 - FG!25: Fix link in Mutex doc
564 - FG!26: Fix links in SMPI interface doc
565 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
566 - GH#305: Unscheduled tasks are still excuted
567 - GH#313: smpirun: manual outdated w.r.t. --help
568 - GH#321: [S4U] Get task remaining work ratio
569 - GH#323: Crash when an actor turn off his physical host
570 - GH#335: Missing links on dragonfly example svg
571 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
572 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
574 ----------------------------------------------------------------------------
576 SimGrid (3.24) October 9. 2019
578 The Clean Disk Release.
581 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
582 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
583 thus becomes the default resource to execute I/O operations. Examples
584 and the file system plugin have been modified to use disks instead of
585 storage elements. Storage can still be used but is doomed to disappear.
586 Users are thus adviced to update their simulators accordingly.
587 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
588 for consistency with pthread_barrier_wait()
589 - Host::get_englobing_zone() returns the englobing netzone
590 - Actor::on_destruction is now called in the destructor
591 Actor::on_termination new signal called when the actor terminates
593 - Global signals are now part of the Engine:
594 - on_platform_creation: after config settings, before the XML parsing
595 - on_platform_created: right after the XML parsing
596 - on_time_advance: each time the clock advances
597 - on_simulation_end: after simulation, before cleanups
598 - on_deadlock: as the name implies.
600 - sg_{actor,host,link}_{data,data_set}() now all exist.
601 Use them to attach user data to the object and retrieve it.
604 - Introduce an experimental Wifi model. It sounds reasonable
605 according to the state of the art, but it still has to be properly
606 validated, at least against ns-3 if not against reality.
607 - Improved the usability of ns-3. Several bugs were ironed out.
608 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
609 It is now expressed as 'idle:epsilon:allCores' to properly model the
610 consumption of non-whole tasks on mono-core hosts. More info in the doc.
613 - convert a new set of functions to the S4U C interface and move the old MSG
614 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
617 - Fortran bindings for DVFS have been removed.
618 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
619 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
620 - SMPI can now be selected by cmake's find_module(MPI) with
621 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
622 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
623 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
624 - trace-call-location can be used with TI traces, and replayed, and is compatible
625 with smpi/comp-adjustment-file.
626 - sleep events are now correctly traced/replayed.
627 - Default for trace-call-location is now to use file names and not full paths.
628 To revert to previous behavior (in case of collision of filenames), option
629 "smpi/trace-call-use-absolute-path" can be set to yes.
632 - Use the included xxHash as an hash implem when C++14 is usable.
633 - Option model-checker/hash was removed. This is always activated now.
634 - New option smpi/buffering controls the MPI buffering in MC mode.
635 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
636 This is useful to check for MPI compliance.
639 - New section on plugins: how to define a new one, and existing ones.
642 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
643 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
646 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
647 and <mount> tags. The rationale is finish to decouple what should be handle
648 by SimGrid kernel (i.e., resources and activities) from plugins built in
649 user space (e.g., the file system plugin). This new tag currently accepts
650 only three attributes (id, read_bw, and write_bw). All the other information
651 that was declared with the storage related tags now has to be expressed as
652 properties. An example of platform using this new tag is available at
653 examples/platforms/hosts_with_disks.xml
656 - 'expect signal' can now accept more than one potential signal.
658 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
659 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
660 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
661 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
662 - FG#31: per-actor data
663 - FG#34: SG_BARRIER_SERIAL_THREAD?
664 - FG#35: model-checker does not like buster-produced binaries
665 - FG!13: MC: complete workaround in the error msg seen on modern systems
666 - FG!15: execute_flops now logs compute
667 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
668 - FG!17: ns-3: unblock the right number of communications + others issues
669 - FG!18: Improving the performance of the ns-3 bindings
670 - GH#189: Energy consumption of parallel tasks
671 - GH#219: Error in the throughput of TCP transfer
672 - GH!330: Minor polishing of English: Configuring_Simgrid
673 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
674 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
675 - GH!334: Rework energy plugin, again
677 ----------------------------------------------------------------------------
679 SimGrid (3.23.2) July 8. 2019
682 - Nicer introduction page.
683 - Migrate the "Deploy your application" page to the new doc.
684 - Move Java as a subtree of MSG.
687 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
690 - Drop xbt_dynar_sort_strings().
693 - Really fix FG#26: Turning off a link should raise NetworkFailureException
694 - FG#27: Wrong exception thrown to wait_any when link is turned off
695 - GH#328: Java: Canceling multiple tasks in a single vm/host
697 ----------------------------------------------------------------------------
699 SimGrid (3.23) June 25. 2019
701 The Exotic Solstice Release.
704 - SunOS and Haiku OS support. Because exotic platforms are fun.
705 - Stop setting random seed with srand48() at initialization.
706 - Use addr2line as a fallback for stacktraces when backtrace is not available.
707 - Build option -Denable_documentation is now OFF by default.
708 - Network model 'NS3' was renamed into 'ns-3'.
711 - Simgrid can now hopefully be installed with pip.
714 - wait_any can now be used for asynchronous executions too.
717 - New log appenders: stdout and stderr. Use stdout for xbt_help.
718 - Drop xbt_dict_dump.
721 - 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
722 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
723 - MPI interface is now const correct for input parameters
724 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
725 - Fortran bindings for async collectives.
726 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
729 - Remove option 'model-check/record': Paths are recorded in any cases now.
730 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
731 always sparse. That's an immense gain in memory, and can even be
732 faster because of cache effects. So there is no need to clutter the
733 code to allow the user to go for the unefficient mode.
736 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
737 models proved to be more accurate than these old experiments.
739 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
740 - FG#1: Broken link in error messages
741 - FG#2: missing installation documentation
742 - FG#3: missing documentation in smpirun
743 - FG#6: Python bindings not available on PyPI
744 - FG#7: simple cmake call requires doxygen
745 - FG#8: make python bindings an optional dependency
746 - FG#10: Can not use MSG_process_set_data from SMPI any more
747 - FG#11: Auto-restart actors forget their on_exit behavior
748 - FG#12: -Denable_lto=OFF doesn't disable LTO
749 - FG#13: Installs unstripped file 'bin/graphicator'
750 - FG#14: Installs the empty directory 'doc/simgrid/html'
751 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
752 - FG#17: Dead link in doc (pls_ns3)
753 - FG#20: 'tesh --help' should return 0
754 - FG#21: Documentation link on http://simgrid.org/ broken
755 - FG#22: Debian installation instruction are broken
756 - FG#26: Turning off a link should raise NetworkFailureException exceptions
757 - GH#133: Java: a process can run on a VM even if its host is off
758 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
759 - GH#326: Valgrind-detected error for join() when energy plugin is activated
761 ----------------------------------------------------------------------------
763 SimGrid (3.22) April 1. 2019
765 The Easter Christmas Release (this one is somewhat late).
768 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
769 - Not complete yet: asynchronous activities (amongst others) are still missing
770 - Still ongoing: the interface may change in the future. We need more
771 testers! Please report any glitches.
772 - No new project using Java should start now. Please switch to Python.
775 - Some of the internal cleanups may lead to speed improvements:
776 - The hard limitation on the amount of simulated actors with
777 Java+Mac was removed. Now, the available memory is the only limit.
778 - Our refcounting was tidyied, leading to 10% speedups in some cases.
779 - We are still working on making our code robust to the actor kills
780 and hosts' churn. Things are improving, but it's not perfect yet.
781 - Replaced our own code to display a backtrace (that was forking addr2line)
782 with the Boost.Stacktrace library.
783 You won't see your backtraces without this optional dependency.
784 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
785 - Stop setting random seed with srand() at initialization.
788 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
789 XML file version remains 4.2 since old files are still compatible.
792 - Process termination which was broken at version 3.21 has been repaired.
793 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
794 - Hide the examples into examples/deprecated. New users should use Python.
797 - Drop MSG_process_create_from_stdfunc() from the API.
798 This C++ function was a pimple in the C API, made necessary at some
799 point by the Java bindings. This is fixed now.
800 - Hide the examples into examples/deprecated. New users should use S4U.
801 - MSG_process_create and MSG_process_attach now crash if the host used
802 to run this process is off.
803 - Fix the protype of MSG_process_on_exit()
804 Now use (int,void*) callbacks instead of (void*,void*) ones.
805 The implementation was ways too messy to actually work, I guess.
808 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
809 too often and provide more accurate timings.
810 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
811 - MPI_Alltoallw support
812 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
813 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
814 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
815 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
817 - Basic implementation of generalized requests (SMPI doesn't
818 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
821 - Drop sg_cmdline. Please use xbt_cmdline instead.
822 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
823 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
824 OsSemaphore is implemented in a portable way with C++11 threads.
825 This should allow much more threads to be created at the same time,
826 allowing Mac and Java users to simulate many more actors.
827 - Implement the 'thread' factory with std::thread instead of xbt ones.
828 It is not possible to set the stack size with threads anymore, but
829 -fsplit-stack is the way to go nowadays when using threads.
830 - Drop the xbt_os_thread_t module (now unused)
831 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
832 - Drop xbt_str_join_array().
833 - Drop cunit, use Catch2 instead.
836 - Many cleanups in the kernel::activity namespace. This was long
837 overdue, and shall open the path to many future endeavors.
840 - #132: Java: a process can not shut down its own host
841 - #220: S4U: detached send causes exception if sender terminates soon after sending
842 - #261: Document the parameters of parallel execution's constructor
843 - #300: [s4u] BarrierPtr is missing
844 - #314: SMPI args internal cleanup
845 - #316: Fix a bug related to the CPU utilization of multi-core VM
846 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
847 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
848 - #325: Turning off a host has different behavior on sleeping actors and computing actors
850 ----------------------------------------------------------------------------
852 SimGrid (3.21) October 3. 2018
854 The Restarting Documentation (TRD) Release.
857 - Convert the user manual to Sphinx for improved usability.
858 Unstable version now hosted on https://simgrid.frama.io/simgrid
859 Many glitches remain; stay tuned.
860 - Restrict the Doxygen reference API to the installed header files.
861 (documenting all internals this way is overkill + doxygen scales badly)
862 - New tutorial on S4U (highly inspired from the old MSG one, but with
863 a git repository to fork as a starting point, and a docker image)
864 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
865 - The Developper manual is still to be converted (not compiled until then)
866 - Some parts are still missing in this conversion (such as the
867 platform and deployment sections) while others were blindly converted
868 and would need more love (such as the configuration flags).
869 Things will certainly further improve in the future releases.
872 - Fully reimplement auto_restart mechanism that was utterly broken
873 (fix #22, #131 and #281 that were all very related)
874 - Implement semaphores (fix #303 and #291)
875 - When creating an actor from a function and its parameters,
876 move-only parameters are not allowed anymore, as it would prevent
877 the actor to be restartable if its parameters are consumed on run.
878 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
879 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
880 IO operations that can be started, waited for, or canceled as a regular
882 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
886 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
887 - New host variable: 'core_count'
890 - Remove xbt_os_thread_specific features
891 - Remove portability wrapper to condition variables
892 - Remove xbt_os_thread_yield()
895 - MPICH collective selector now mimics MPICH 3.3b
896 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
897 - MPI_Init(NULL,NULL) is now allowed
898 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
899 - Command-line arguments should now be handled properly in Fortran simulations
900 - MPI Attributes and related callbacks should work in Fortran
901 - Apps using C + Fortran should now work
902 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
903 - Support MPI_Error_String
906 - Due to an internal bug, Msg.run() must now be your last line.
907 We hope to fix it in a future release, and we are sorry for the inconvenience.
910 - #22: Process autorestart seem to only work with CAS01 cpus
911 - #56: Feature request: dynamic SMPI replay
912 - #93: simgrid should not eat --help
913 - #111: How to change the loopback link characteristics?
914 - #116: Platform cluster radical
915 - #117: Platform cluster/host order
916 - #129: Replay_multiple on steroid
917 - #131: Java : autostart does not work
918 - #143: Setting a breakpoint at a given time
919 - #235: Network energy models should be integrated
920 - #264: Add ptask L07 resource tracing
921 - #271: Dynamic and manual replay of SMPI traces
922 - #279: Breakpoints option uses time deltas instead of absolute time
923 - #281: Daemonized actor no longer daemonized after an auto-restart
924 - #282: TIT convertor erase traces when using absolute path in the trace list file
925 - #285: segfault when a s4u actor kills itself with msg process tracing activated
926 - #286: Pajé traces are not exposing the number of cores
927 - #287: Command-line parsing should stop when encountering '--'
928 - #288: MPI_Init(NULL, NULL)
929 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
930 - #290: Method to check if a mailbox has a message ready to be consumed
931 - #291: [s4u] semaphores do not seem available yet
932 - #294: Alltoall collective implementation of mpich
933 - #295: Starting a Host as OFF is not supported
934 - #296: DTD too permissive
935 - #299: Add a s4u barrier example
937 ----------------------------------------------------------------------------
939 SimGrid (3.20) Released June 24. 2018
941 The proxy snake_case() release.
944 - The whole API is now using snake_case() instead of mixing it with camelCase.
945 Compatibility wrappers in place for one year.
946 - We took the opportunity to increase the consistency of the API.
947 Please report any remaining glitches.
948 - New function simgrid::s4u::Host::get_actor_count:
949 Returns the number of actors running on a specific host.
952 - SMPI is now tested with ~45 proxy apps from various sources, with none or
953 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
954 - Replay: The replay file has been re-written in C++.
955 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
956 supported. They are stored in the trace and used when replayed.
957 - Basic support of MPI_Cancel. Robustness not guaranteed.
958 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
959 (only for MPI_COMM_TYPE_SHARED).
960 - New option: smpi/privatize-libs, to add external shared libs to be privatized
961 by SMPI. They will be copied locally and loaded separately by each process.
962 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
963 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
964 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
965 with internal index optind with multiple processes. Only works if MPI_Init has
967 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
968 syntax without preprocessing tricks.
971 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
972 behavior tracing is no longer limited to MSG
973 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
974 this means that traces now need to be updated or re-obtained; both Irecv
975 and Isend lines in your traces have now in total 5 mandatory fields:
976 <rankid> <command> <to/from rankid> <tag> <size>
977 To update your traces, it suffices to add a 0 for the tag here.
978 - TIT now also supports waiting for a distinct request via MPI_Wait.
979 Wait/Test now wait for a specific request, not just the last one that was
980 issued. This unfortunately means another update, because we need to
981 identify which request you want to wait for. We do this via the
982 triplet (sender, receiver, tag), which needs to be added:
983 <rankid> <command> <sender> <receiver> <tag>
984 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
985 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
988 - The deprecation of MSG is ongoing (but this should not impact you).
989 Many MSG functions are now simple wrappers on the C API of S4U. If
990 you wish to convert your code to S4U, find the S4U counterparts of
991 your MSG calls in src/msg/msg_legacy.cpp.
992 - MSG can still be used, but won't evolve anymore.
995 - Allow to run the Link energy plugin from the command line with
996 --cfg=plugin:link_energy
997 - Rename Energy plugin into host_energy
998 - Rename Load plugin into host_load
1001 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
1002 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
1003 - Improve performance when handling timeouts of simix synchros.
1006 - Config: the C API is now deprecated (will be removed in 3.23), and
1007 the C++ API has been slightly improved.
1010 - Fix several build issues on OSX.
1011 - Move simgrid_config.h to simgrid/config.h (old header still working)
1014 - #143: Setting a breakpoint at a given time
1015 - #150: Inconsistent event names in SMPI replay
1016 - #258: daemonized actors hang after all non-daemonized actors have completed
1017 - #267: Linker error on unit_tmgr
1018 - #269: SMPI: tracing of MPI_Wait/all/any broken
1019 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
1021 ----------------------------------------------------------------------------
1023 SimGrid (3.19.1) Released March 22 2018
1025 The Fixed ABI Release.
1027 As you may know, we are currently refactoring SimGrid in deep.
1028 Upcoming SimGrid4 will be really different from SimGrid3: modular,
1029 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
1031 Our goal is to smooth this transition, with backward compatibility
1032 and automatic update paths, while still progressing toward SimGrid4.
1034 SimGrid remains open during works: The last pure SimGrid3 release was
1035 v3.12 while all subsequent versions are usable alpha versions of
1036 SimGrid4: Existing interfaces remain unchanged, but the new S4U
1037 interface is budding and the internals are deeply reorganized.
1039 Since 2015, we work hard to reduce the changes to public APIs. When
1040 we need to rename a public library symbol in S4U, we let your
1041 compiler issue an explicative warning when you use the deprecated
1042 function. These messages remain for four releases, i.e. for one full
1043 year, before turning into an error. Starting with v3.15, your can
1044 also adapt to API changes with the SIMGRID_VERSION macro, that is
1045 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
1047 Starting with this v3.19.1, our commitment to reduce the changes to
1048 the public interfaces is extended from the API to the ABI: a program
1049 using only MSG or SimDag and compiled against a given version of
1050 simgrid can probably be used with a later version of SimGrid without
1051 recompilation. We will do our best... but don't expect too much of
1052 it, that's a really difficult goal during such profund refactoring.
1054 The difference between v3.19 and v3.19.1 is that the former was
1055 accidentally breaking the ABI of MSG, while the later is restoring
1058 S4U and kernel APIs will still evolve until SimGrid4, with one-year
1059 deprecation warnings as currently. In fact, cleaning up these
1060 interfaces and converting them to snake_case() is one release goal of
1061 v3.20. But don't worry, we are working to smooth this upgrade path.
1063 In summary, new projects should start with S4U to benefit of the
1064 future, but old MSG projects should still be usable with no change.
1066 ----------------------------------------------------------------------------
1068 SimGrid (3.19) Released March 20 2018
1070 The Moscovitly-cold Spring release.
1072 SMPI user-visible improvements:
1073 - Support of MPI_Type_create_subarray
1074 - Unimplemented MPI calls now abort (instead of issuing a warning)
1075 Some harmless calls around errhandlers still only issue a warning.
1076 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
1077 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
1078 is now possible (mmap privatization was refactored to this end).
1079 Adding new SMPI instances should also be technically doable now.
1081 SMPI internal cleanups:
1082 - Removed index notion from SMPI -- no more "getPid() - 1"!
1083 In tracing files, actors are called rank-<PROCESS-ID> but these are
1084 NOT the real ranks (tracing several communicators is not supported yet).
1085 - Deprecate smpi_process_index() (will be removed in 3.22)
1086 - Better testing on classical proxy apps, through an external project:
1087 https://github.com/simgrid/SMPI-proxy-apps/
1090 - Use a graphical TOC to make it easier to find the documentation you need
1091 - Revamp the MSG tutorial
1094 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
1095 - Remove parameter of MSG_process_killall().
1096 Resetting the PID was bogus anyway (several actors could have the same PID).
1099 - Execution->setHost() can be called after start(), to migrate it.
1100 - Comm::test_any() is now implemented.
1101 - s4u::Actor now has onCreation() and onDestruction() signals.
1102 - Install some kernel header files for the users' plugins and more.
1103 Warning, their API is really not stable yet.
1104 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1105 include/simgrid/kernel/routing/*Zone.hpp
1108 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1110 Build System and other cleanups:
1111 - Remove unused run-time parameter "tracing/onelink-only".
1112 - SimGrid now works with intel compilers (MC still buggy)
1113 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1114 - tesh kills subprocesses on timeout (but not on windows)
1117 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1120 - #194: Feature request: simgrid::s4u::Comm::test_any()
1121 - #245: migrating an actor does not migrate its execution
1122 - #253: Feature Request: expose clusters as objects
1123 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1124 - #255: Tesh broken on Windows
1125 - #256: Modernize FindSimGrid.cmake
1126 - #257: Fix (ab)use of CMake install
1128 ----------------------------------------------------------------------------
1130 SimGrid (3.18) Released December 24 2017
1132 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1134 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1135 - Support for MSG should not be removed from SimGrid before 2020,
1136 but future evolutions will be limited to the integration of
1137 user-provided patches.
1140 - s4u::onDeadlock() signal, run before stopping the simulation.
1141 - s4u::Exec: asynchronous executions (abstraction of background
1143 - s4u::Host->getLoad() returns the achieved speed in flops/s
1146 - Link::name() is deprecated and will be removed in v3.21, use
1147 Link::getCname() instead.
1148 - Mailbox::getName() changed to return a std::string, use
1149 Mailbox::getCname() instead to get a char*.
1150 - Storage::getName() changed to return a std::string, use
1151 Storage::getCname() instead to get a char*.
1152 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1153 to have both allocation and deallocation of the map in user space.
1156 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1157 The amount of flops remaining to do is a vector, not a scalar.
1158 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1159 name implies on both sequential and parallel tasks.
1160 - Both changes fix GitHub's #223 using PR #237 as a basis.
1161 Thanks Michael Mercier.
1162 - Most examples were converted to S4U and hidden elsewhere at tests.
1163 Rationale: we still want MSG to work; we want newcomers to use S4U.
1166 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1167 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1170 - Switch to the faster dlopen privatization mechanism by default
1171 - Documentation improvments
1174 - Live migration is getting moved to a plugin. Dirty page tracking is
1175 the first part of this plugin. This imply that VM migration is now
1176 only possible if one this function is called:
1177 - C/MSG: MSG_vm_live_migration_plugin_init()
1178 - C/C++: sg_vm_live_migration_plugin_init()
1179 - Java: Msg.liveMigrationInit()
1180 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1184 - Define class simgrid::xbt::Path to manage file names.
1185 - Removed unused functions:
1186 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1187 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1188 xbt_graph_export_graphviz()
1189 - xbt/str.h: xbt_str_join()
1190 - Remove unused datatypes:
1191 - xbt/heap.h: use std::priority_queue or boost::heap instead
1192 - xbt/swag.h: use boost::intrusive::list instead
1195 - New link_energy plugin for the consumption of the links.
1196 - All of the operations on files and storage contents have been
1197 packaged into a plugin (src/plugins/file_system). The current
1198 public interface can be found in
1199 include/simgrid/plugins/file_system.h
1200 To use these functions you now have to initialize the plugin by
1201 calling MSG_storage_file_system_init() just after calling
1202 MSG_init() or sg_storage_file_system_init() just after creating
1206 - Remove the undocumented/untested tag <include>
1209 - Remove viva specific tracing as the tool is no longer maintained
1212 - #248: Kill of finished processes leads to segfault
1213 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1214 - #239: please implement signal s4u::onDeadlock()
1215 - #230: segfaults when exit() without run()
1216 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1217 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1218 - #222: Actor::kill() doesn't really kill and segfaults
1219 - #221: odd LMM warning when killing an actor
1220 - #120: Memory leak when the processes are forcefully killed
1222 ----------------------------------------------------------------------------
1224 SimGrid (3.17) Released October 8 2017
1226 The Drained Leaks release: (almost) no known leaks despite the tests.
1228 Even very long simulations will preserve your memory: our testsuite
1229 stresses SimGrid for over 45mn on fast machines for a coverage of
1230 over 80%, and there is only one single known leak, of about 4kb.
1234 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1235 handler may be given to free resources if the comm cannot be completed.
1236 - this_actor::send(mailbox) is now mailbox->put()
1237 - New: simgrid::s4u::Comm::wait_all()
1238 - New: Host.execute() for remote executions.
1241 - Unused option network/sender-gap is removed.
1244 - Deprecate MSG_task_isend_with_matching(): This unused feature
1245 really complicates our internals. Will be removed in v3.20.
1248 - Improved context termination. It is now possible to free resources, even
1249 when a process is forcibly killed.
1250 - ContextBoost: add support for Boost versions above 1.61.
1253 - Bring back run-time option --cfg=exception/cutpath to remove exception
1255 - Removed unused functions:
1256 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1257 xbt_str_rtrim(), xbt_str_trim().
1258 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1261 - Removed header files obsolete since SimGrid 3.12:
1262 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1264 - #3: SD_exit should be made optional
1265 - #120: Memory leak when the processes are forcefully killed
1266 - #159: Threading test regression in Actor refcounting
1267 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1268 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1269 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1270 - #191: VM migration and pstate
1271 - #192: Updating the energy consumptions of all hosts crashes with VMs
1272 - #195: All actors have PID=0 in the logs
1273 - #204: Sometimes segfault with thread contexts and mmap privatization
1274 - #222: Actor::kill() doesn't really kill and segfaults
1275 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1277 ----------------------------------------------------------------------------
1279 SimGrid (3.16) Released June 22. 2017.
1281 The Blooming Spring Release: developments are budding.
1284 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1285 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1286 - Java JAR file should be finally fully working on Mac OSX too.
1288 XML platforms: Switch to platform v4.1 format.
1289 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1290 - <zone> can be used as a synonym for the now deprecated <as>
1291 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1292 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1293 - <actor> can be used as a synonym for the now deprecated <process>
1294 - state_file and avail_file periodicity is now easier to express
1295 (check the documentation) the old behavior should still work.
1296 * Storage: not backward compatible, but it seems that nobody ever used it
1297 - Remove attribute 'content_type' of <storage_type>: was never used
1298 - Make attribute 'model' of <storage_type> optional: for future usage
1299 - Remove Bconnection model property: was never (in)validated. Replaced by
1300 the maximum of the read and write bandwidth as a resource constraint.
1303 - New and Backwards Compatibility break:
1304 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1305 When one wants to get the list of tasks whose states have changed during a
1306 simulation round, s/he has to allocate and free a dynar and use it as
1307 argument to this function. The former SD_simulate (double how_long)
1311 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1312 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1315 - New function to update the consumption of all hosts at once.
1316 - Fix the model for multi-core hosts, linear in the amount of busy
1317 cores with an abnormality for pIdle (see paper).
1318 BUG: Still not satisfactory for ptask on mono-cores.
1321 - The netzone are now available from the MSG API.
1322 The old names still work, but are now deprecated.
1325 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1326 - New option: smpi/keep-temps to not cleanup temp files
1327 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1328 mallocs mode is used, allows to change the size of the fake file used
1329 (default 1MB), to potentially limit the number of mappings for large runs.
1330 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1331 - Fortran ifort and flang compilers support
1332 - New RMA calls supported (experimental):
1333 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1334 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1335 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1336 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1337 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1338 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1341 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1342 - Behind the scene, ActivityImpl are now automatically refcounted.
1343 Removing this nasty bottleneck will greatly simplify our internals.
1346 - Replay: New function xbt_replay_action_get():
1347 Retrieve the function previously associated to an event type.
1348 - DROPPED FUNCTION: xbt_str_varsubst()
1349 - DROPPED MODULE: strbuff. We don't need it anymore.
1350 - DROPPED MODULE: matrix. We don't need it anymore.
1351 - DROPPED MODULE: lib. We don't need it anymore.
1353 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1355 SimGrid (3.15) Released March 22. 2017
1357 The Spring Release: continuous integration servers become green
1358 We fixed even the transient bugs on all target architectures:
1359 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1360 Windows, FreeBSD, NetBSD.
1362 - Rename NetCards to NetPoints.
1363 This was intended to help NS3 users, but that's not a netcard.
1364 That's a point in the routing algorithm, let's avoid wrong simplifications.
1367 - New: MSG_process_yield(). Stop and yield to other processes.
1368 - New: MSG_process_daemon(). Daemon processes are automatically killed
1369 when the last non-daemon process terminates
1370 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1371 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1372 to make room for the upcoming network energy plugin.
1373 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1376 - Ensure that an actor can kill itself with Process::exit()
1377 - Kill the obscure NativeException. Nobody want to survive the issues
1378 it denotes, so use JniException that is a RuntimeException (not to
1379 be caught explicitly).
1380 - Partial bug fix in initialization. SimGrid flags on command line were
1381 consumed at C level but stayed in the original Java String[] args.
1382 This could mess users' args[i] if SG flags were not put at the end of
1384 The SimGrid flags are now removed from the Java arguments. However,
1385 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1386 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1387 It might be if you have --log or --cfg flags in the command line.
1388 - Fix numerous memleaks all around the place. In particular, around VMs.
1392 - s4u::onPlatformCreated: right before the simulation starts
1393 - s4u::onSimulationEnd: right after the main simulation loop
1394 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1395 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1396 event from the availability_file changes the avail speed.
1397 - Links are now usable from s4u
1398 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1399 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1400 - New examples: The conversion of MSG examples to S4U has begun
1401 - Actors: create, kill, migration, and suspend
1402 - Applications: master-worker and token-ring
1403 - Action replay: communications and storage
1404 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1407 - Backwards Compatibility breaks
1408 - The SD_route_get_list and SD_route_get_size functions have been removed.
1409 They are replaced by the unique
1410 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1411 The route from src to dst is built in the links dynar whose size can
1412 be obtained with xbt_dynar_length.
1413 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1414 sg_host_route_bandwidth and sg_host_route_latency.
1415 Macros ensure the backwards compatibility, but you should fix your code
1418 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1419 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1420 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1421 We don't support 100% of it, but it should be enough. Contact us if not.
1422 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1423 naive and probably inaccurate)
1424 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1425 mapped onto a unique small file using some system magic.
1426 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1427 - Fortran: cleanups, fixes, support of user-added operations
1428 - RMA: MPI_Accumulate are applied in correct order.
1429 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1430 - Better support for MPI_IN_PLACE
1431 - Support for MPI_Win attrs and keyvals.
1432 - Support MPI_Comm_group_create, MPI_Type_size_x
1433 - MPI ops should be properly applied to all allowed types
1434 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1435 algorithm by a non failing one from mpich.
1438 - Kill the fifo data container: we don't use it anymore.
1441 - A <cluster> can now be created with different speed values to represent pstates.
1442 see examples/platforms/energy_cluster.xml for an example.
1444 SimGrid (3.14.159) Released December 28. 2016
1446 The previous release was only a crude approximation of a Pi release;
1447 this one should be a bit better.
1449 - Revert a last minute change that broke on Mac OSX.
1450 - Fix the changelog and NEWS files.
1451 - Improve the documentation, in particular of the routing module.
1452 - Integrate some patches flying around in Debian.
1454 SimGrid (3.14) Released December 24. 2016
1457 * The whole documentation was reworked and reorganized. There is
1458 still much room for improvement here, but we're on it.
1461 * We now need python3 for our testing infrastructure.
1462 * Model checking needs libevent
1465 * Now works on FreeBSD
1468 * Privatization now works on FreeBSD
1470 * Privatization is now activated by default in smpirun
1471 You can turn it off with -no-privatize if something goes wrong.
1473 * Call-location tracing for SMPI.
1474 You can add the exact location (filename / linenumber) of an MPI call to
1475 your trace files and slow down or speed up the simulation between two
1476 consecutive calls by using an adjustment file (see the documentation).
1478 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1479 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1480 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1481 values were ignored in some cases. The timings of these functions can now
1482 be significantly different.
1484 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1485 smpi/running-power is renamed to smpi/host-speed
1487 * smpi/grow-injected-times option to enable or disable multiplication of the
1488 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1489 make simulation less precise (but also much faster).
1491 * smpirun script should be (much) faster for large deployments.
1493 * SMPI tracing: fixed issue with poor matching of send/receives.
1495 * Replay: Fix broken waitall
1497 New functions and features
1498 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1500 Dropped / renamed functions and features
1501 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1502 - MSG_mailbox_is_empty() -> Mailbox::empty()
1503 - MSG_mailbox_front() -> Mailbox::front()
1504 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1505 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1506 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1507 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1508 * Don't pass the free_f parameter to property related functions:
1509 - MSG_host_set_property_value()
1510 - MSG_as_router_set_property_value()
1511 - MSG_storage_set_property_value()
1512 * VM properties. Since msg_vm_t are msg_host_t, just use
1513 MSG_host_get_property_value() and friends
1514 * VM I/O related things:
1515 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1516 - Unimplemented save/restore methods
1517 * MSG_as_router_get_property_value() was redundant with
1518 MSG_environment_as_get_property_value().
1519 - Removed MSG_as_router_*propert*() functions
1520 - Added MSG_environment_as_set_property_value() for consistency
1521 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1522 Well, they are still there for now, but deprecated with a warning.
1523 Please switch to xbt_dict_new_homogeneous() before this is removed
1525 * Task affinity. Its intended behavior (that was very badly tested
1526 and probably not really working) was deceiving what most users
1527 would have hoped here.
1528 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1529 * Option network/coordinates is now useless and should be dropped.
1532 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1535 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1536 * Added Dragonfly topology support
1539 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1540 EiBps, EBps, Eibps, Ebps for bandwidth.
1541 They may become useful to some lucky ones.
1544 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1545 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1546 * ex/async_waitAll: new example, on asynchronous communications
1549 * Memory usage should be decreased for simulations with a large number
1550 of processes. This also helps for SMPI.
1552 SimGrid (3.13) stable; urgency=low
1554 The Half Release, a.k.a. the Zealous Easter Trim.
1556 - We removed half of the lines, that were mostly experimental cruft.
1557 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1558 (not counting blanks and comments -- according to openhub.net)
1559 - The internals are now compiled in C++ (and will soon be clean C++)
1560 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1561 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1563 * Backwards Compatibility breaks
1564 - Removed Lua simulation bindings (switch to C or Java for that).
1565 Lua can still be used to describe platforms
1566 - Removed Java kernel plug-ins.
1567 Will be reintroduced after the ongoing major internals reorg.
1569 - the following functions were removed.
1570 They were too specific and should be reimplemented in a generic
1571 way, with filter function.
1572 - MSG_task_listen_from_host
1573 - MSG_mailbox_get_count_host_waiting_tasks
1574 - MSG_mailbox_put_with_timeout was removed.
1575 Please use MSG_task_send_with_timeout instead.
1577 - the SD_application_reinit function was removed. It has been a noop for a while.
1578 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1579 replaced by a more flexible mechanism.
1580 - The following functions thus do not exist anymore
1581 - SD_workstation_get_access_mode
1582 - SD_workstation_set_access_mode
1583 - SD_workstation_get_current_task
1584 - Basic estimation functions have been removed but can easily be replaced
1585 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1586 - SD_workstation_get_computation_time => amount / sg_host_speed()
1588 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1589 Use VM.getSpeed()*load/100 for the legacy behavior.
1591 - option enable_tracing was removed. It was not doing anything for a while.
1592 - In the ModelChecker:
1593 - the model-checker now ptraces the model-checked process which means
1594 you cannot use a debugger on the latter anymore (we might make this
1595 optional in the feature);
1596 - removed soft-dirty page tracking;
1597 - remove model-checked side snapshot management,
1598 MC_snapshot() and MC_compare_snapshot();
1599 - keep the MC_cut() function as a stub (it was not really working
1600 in the previous release).
1603 * All options are consistently in kebab-case. Old names are kept as alias.
1606 * Switch to platform v4 format.
1607 - Rename from 'power' to 'speed' the attributes describing the amount of
1608 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1609 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1610 - In <host> and <link>, attributes availability and state are gone.
1611 It was redundant with state and availability traces, and with peak values.
1612 - In <cluster>, attributes availability_file and state_file are gone.
1613 It was too complex and unused.
1614 - Kill <gpu>. Was not doing anything.
1615 - The DOCTYPE points to the right URL:
1616 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1617 (the file at this address now documents the changelog since its v1)
1618 - A warning is emitted for unit-less values (they are still accepted).
1619 - speed. Default: 'f' or 'flops'. Also defined:
1620 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1621 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1622 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1623 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1624 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1625 - latency. Default: 's' second. Also defined:
1626 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1628 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1629 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1632 * s4u::Host is now the preferred public interface to the Host features.
1633 sg_host_* functions are C bindings to the exact same behavior
1634 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1637 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1638 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1639 They were deprecated since a few years, and probably did not even compile anymore.
1642 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1643 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1644 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1645 encouraged to update their codes. A list of the modified functions can be found at the end of
1646 include/simgrid/simdag.h
1649 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1650 the lambda or closure passed as a parameter will run in kernel mode.
1651 All the callback functions should be rewritten to that interface at some point.
1654 * Reorganizing and cleaning the internals all around the place.
1657 * Remove old default barrier/bcast buggy algorithms (see #18407)
1658 * Various bug fixes to handle more codes
1659 * Remove the need for the --foreground option of smpirun (it is still
1660 accepted for backward compatibility).
1663 * Kill the setset data container: MC don't use it anymore.
1664 * Kill the queue data container: it made more sense with GRAS.
1665 * Kill the xbt_peer_t data type: it's useless without GRAS.
1666 * Kill rm_cb feature of config sets: it was never useful.
1667 * Kill graphxml parsing feature. It was not used.
1668 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1670 - xbt_dynar_sort_strings(), when the content is char*
1671 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1672 They throw exceptions on invalid input;
1673 * C++ support for declaring CLI flags (simgrid::config::Flag);
1674 * class for abstracting different signal backends (simgrid::xbt::signal).
1675 with no external dependencies (we need very simple signals).
1678 * refactoring and cleanup of the code;
1679 * ongoing process to cleanly separate the model-checking algorithms
1680 from the code model-checking support.
1682 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1684 SimGrid (3.12) stable; urgency=low
1686 The Facelift Release.
1689 * Require g++ v4.7 at least to not speak prehistorical C++.
1690 * Require Boost 1.48 (for signal2 component).
1691 * Java must be version 7 at least when activated.
1692 * Builds on Windows again (including Java bindings).
1693 * Tracing is now always enabled (no way to turn it out)
1694 * Remove GTNetS. It was not working anyway.
1695 * Various cleanups in the cmake scripts.
1696 * Move headers around to sort them out on installed systems:
1697 - instr/instr.h -> simgrid/instr.h
1698 - instr/jedule/* -> simgrid/jedule
1699 - simdag/datatypes.h was removed
1700 - simdag/simdag.h -> simgrid/simdag.h
1701 - msg/datatypes.h was removed
1702 - msg/msg.h -> simgrid/msg.h
1705 * Interface improvement:
1706 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1707 - Sanitize the interface in MSG_task_ module:
1708 - Merge two functions that were close enough but misleading:
1709 set_compute_duration(t) -> set_flops_amount(t)
1710 get_remaining_computation(t) -> get_flops_amount(t)
1711 - set_data_size(t) -> set_bytes_amount(t)
1712 get_data_size(t) -> get_bytes_amount(t)
1713 - Massive cleanups in the functions related to the energy
1714 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1715 - New: MSG_host_get_pstate()
1717 - msg/energy/onoff: switching hosts on and off
1719 * Interface improvement:
1720 - Rename Host.isAvail() to Host.isOn()
1721 - Rename Process.currentProcess() to Process.getCurrentProcess()
1722 - Rename Task.setDataSize() to Task.setBytesAmount()
1723 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1725 - #18874: Actually allows the GC to reclaim tasks
1729 - SIMIX_process_throw: raises an exception in a remote process
1730 * Refactoring: Separate sub-modules
1731 - libsmx: the public interface, as libc in a real system
1732 - popping: the strange dance that converts a user request into a kernel handling
1733 - smx_context_*: the virtualization mechanisms that embed the user code
1734 - smx_*: the handling of each simcalls
1735 * Interface cleanup:
1736 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1737 * Rename smx_action_t into smx_synchro_t, making explicit that these
1738 things are used to synchronize processes with their environment.
1739 For example, a communication is a sort of synchronization involving
1740 the communicating processes (that may block until the exchange) and
1741 the platform. The same can be said from computations, etc.
1743 - #18888: segfault when a process ends before its kill_time
1747 - 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.
1748 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1749 - MPI_Type_set_name, MPI_Type_get_name
1750 - MPI_*_c2f and MPI_*_f2c functions
1751 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1752 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1753 - Activate a lot of new tests from the mpich 3 testsuite
1755 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1756 - InfiniBand network model added: Based on the works of Jerome Vienne
1757 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1758 - When smpi/display_timing is set, also display global simulation time and application times
1759 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1760 * Collective communications
1761 - 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
1762 - 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/).
1763 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1765 - Replay now uses algorithms from wanted collective selector
1766 - Replay can be used with SMP-aware algorithms
1767 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1768 - 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.
1770 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1771 - Correctly use loopback on fat-tree clusters
1772 - Asynchronous small messages shouldn't trigger deadlocks anymore
1773 * Energy/DVFS cleanup and improvement
1774 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1775 - new: smpi_get_host_pstate()
1778 - "Full" network optimization flag was broken since Surf++
1779 - Better handling of precision flags in maxmin
1780 - Fix bug causing sometimes "Impossible" errors
1781 - Properly pass cluster properties to included hosts
1782 * Improvement of the Energy plugin.
1783 - Always update the consumption before returning that value
1784 - New property: watt_off to denote the dissipation when the host is off
1785 - New functions getWattMinAt and getWattMaxAt to retrieve the
1786 dissipation of pstates that we are not currently at.
1787 * Java: class NetworkLink renamed to Link
1788 * New function: simcall_process_get_kill_time()
1789 * Massive rename s/workstation/host/
1790 - That's intrusive, but that's good for the project consistency. Sorry.
1791 - Change config option "workstation/model" into "host/model"
1794 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1795 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1796 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1797 - xbt_dirname and xbt_basename for non-POSIX systems
1799 * The model checker now runs as a separate process.
1800 * The model checker runs must now be launched with the new simgrid-mc program.
1801 * Record/Replay: the MC can display a textual representation of a path in the
1802 execution graph. It can then be replayed outside of the model checker.
1804 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1806 SimGrid (3.11) stable; urgency=low
1811 * Normalizing pointers addresses tool for better diff between logs
1814 * Add cloud examples using new VMs
1815 - examples/msg/cloud/two_tasks_vm.tesh
1816 - examples/msg/cloud/simple_vm.tesh
1817 - examples/java/cloud/cloud.tesh
1818 - examples/java/cloud/migration/migration.tesh
1819 * Add java surf examples:
1820 - examples/java/surfPlugin/surf_plugin.tesh
1821 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1822 - examples/java/surfCpuModel/surf_cpu_model.tesh
1823 * Add SMPI+MSG example:
1824 - examples/smpi/smpi_msg_masterslave/
1829 - msg task destroy cancel test
1830 - msg_host on/off test
1831 * Move all tests in testsuite to teshsuite (adding tesh files)
1832 * Restructure teshsuites
1833 - one folder for each kind of test
1834 * Restructure AddTests.cmake
1836 - structure the order of tests (with sections)
1839 * Add virtual machine
1840 - creation of a VM on a PM
1841 - migration of a VM from a PM to another PM
1843 - MSG_process_join(msg_process_t process, double timeout)
1844 - msg_bar_t MSG_barrier_init(unsigned int count)
1845 - int MSG_barrier_wait(msg_bar_t barrier)
1846 - void MSG_barrier_destroy(msg_bar_t barrier)
1847 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1848 * New option "msg/debug_multiple_use" to help debugging when a task is used
1851 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1853 - Increase code coverage in test suites
1857 * Protect context stacks against stack overflow. The number of protected memory
1858 pages allocated on the top of each stack (1 by default) can be configured
1859 with the new command line option --cfg=contexts/guard_size:<value>.
1860 * Simcalls are now generated by a python script that
1861 - generates files included by SimGrid
1862 - checks that all the functions exist, and proposes prototypes
1864 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1865 host_set_data, host_get_data
1867 - simcall_process_join(smx_process_t process, double timeout)
1868 * Fix bug where sleeping processing could not be suspended.
1871 * Translate surf models from C to C++
1872 - Generic classes for all models: Model, Resource, Action
1873 - A generic interface for each kind of model (CPU, Network, Storage
1874 Workstation, WorkstationVM)
1876 * Translate surf routings from C to C++
1877 * Add callbacks using sigc++ or boost::signals2
1878 - Add callback functions for resource creation/destruction
1879 - Add callback functions for action state change
1880 - Handle Energy as a plugin
1881 * Replace swag by boost::intrusive
1882 * Add new routing models for clusters. For documentation, see
1883 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1884 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1885 parameters for cluster tag
1886 - Fat trees, with topology="FAT_TREE" and
1887 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1888 - see examples/platforms/torus_cluster.xml and
1889 examples/platforms/fat_tree_cluster.xml
1890 * More documentation
1893 * Hostfiles support host:nb_processes construct to deploy several processes on
1895 * Collective communication algorithms should not crash if used with
1896 improper number of nodes and report the error.
1897 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1898 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1899 MPI_Dims_create, MPI_Cart_sub are supported.
1900 * New interface to use SMPI programmatically (still depends on MSG for
1901 some parts, see examples/smpi/smpi_msg_masterslave):
1902 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1906 * Global variables privatization in MPI executables is now performed at runtime
1907 with the option smpi/privatize_global_variables (default:no).
1908 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1909 dynamic libraries loaded by the application are not privatized (static
1910 linking with these libraries is advised in this case)
1913 * Options defined in XML work correctly now.
1916 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1917 copied into simgrid.jar or not (ON by default). Use this option if you want
1918 to reduce the size of the installed simgrid.jar, *and* the native libraries
1919 are kept installed elsewhere.
1920 * Surf binding with SWIG (code generated in maintainer mode only):
1921 - plugin to handle callbacks
1922 - CPU model only for the moment
1925 * Supernovae build mode is definitively removed. It was used to improve
1926 inlining and inter-module optimizations. It is nowadays superseded by
1927 link-time optimizations commonly available in compilers.
1928 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1929 * Add boost dependency for surf++
1930 * Add new macro for tests
1931 - ADD_TESH(name <tesh_args>)
1932 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1936 - xbt_bar_t XBT_barrier_init(unsigned int count)
1937 - int XBT_barrier_wait(xbt_bar_t barrier)
1938 - void XBT_barrier_destroy(xbt_bar_t barrier)
1939 * Make the xbt_os_time module public
1941 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1943 SimGrid (3.10) stable; urgency=low
1945 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1948 * Reintegrate Java to the main archive as desynchronizing these
1949 package is not acceptable anymore (Java is now considered stable)
1950 * Add explicit synchronization facilities through semaphores
1951 * Bug fix: Task.setDataSize() only changed the C world, not the value
1952 cached in the Java world
1955 * Dramatically change the way files are handled. API and internals changed, but
1956 this part of MSG was not considered as production grade either.
1957 * Add explicit synchronization facilities through semaphores
1958 * Add a new function MSG_host_get_process_list()
1959 * Preliminary DVFS support (see examples/msg/energy/ for details)
1962 * SMPI is now included directly in the libsimgrid as the windows
1963 linker doesn't force us on splitting it anymore.
1964 * Improvements of the SMPI replay tool:
1965 - Most of the collective communications are now rooted in the same process as
1966 in the original application.
1967 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1968 used until now). Multiple data types can now be used in a trace.
1969 - The replay tool now supports traces produce either by TAU or a modified
1971 - Bug Fix: the compute part of the reduce action is now taken into account.
1972 - Gatherv collective is now supported
1973 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1974 of smpirun outputs time independent traces for the current run. One file
1975 is created per process. If too many processes are simulated, this behavior
1976 can be changed to one file for all processes by using the
1977 tracing/smpi/format/ti_one_file flag
1978 * smpirun generates the host file if needed (with given host count and platform)
1979 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1980 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1981 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1982 --cfg=smpi/coll_selector:(mpich/ompi)
1983 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1985 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1986 happening inside a collective SMPI call.
1987 * Fix the behavior of complex data types handling.
1988 * Make MPI_Wtime another synchronization point to take computations into
1990 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1991 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1992 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1993 (no privatization of global variables yet)
1994 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1997 * Verification of liveness properties is now available for SMPI applications
1998 (in addition to MSG applications)
1999 * Bugged examples using SMPI in examples/smpi/mc/
2000 * Add --cfg=model-check/visited option. Allows the verification of infinite
2001 programs. Detection of loops in the execution thanks to the system state
2002 comparison and reduction of the state space to explore. Can be combined with
2003 DPOR for safety properties.
2006 * Allow to change SimGrid configuration (see --help) within the code
2007 thanks to SD_config() as it can be done in MSG.
2008 * Add a new function SD_task_set_amount() upon user request.
2011 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
2012 * Remove rule based routing (no more PCRE dependency)
2013 * Add a limiter_link option to cluster tag, to specify a maximum reachable
2014 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
2015 * Add a loopback_bw and loopback_lat options to cluster tag.
2016 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
2017 routing. Now peers are encapsulated in an AS and have their own private
2018 router but this is transparent.
2021 * Our own implementation of getline is renamed xbt_getline, and gets
2022 used even if the OS provide a getline(). This should reduce the
2023 configuration complexity by using the same code on all platforms.
2024 * New type: xbt_cfg_elm_boolean.
2025 * Allow to use yes/no for boolean configuration options in the command line.
2026 * Allow to disable SimGrid cleanups at exit from command line option.
2027 There are situations where one may want a simulation to end with an exit.
2028 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
2029 annoying when scripting around the simulator. Adding a
2030 --cfg=clean_atexit:no allows to circumvent this issue.
2033 * Lots of memory leaks were corrected in this release.
2034 There is no known memory leaks anymore, in all of our 600+ tests.
2035 * New command line option --version, to get SimGrid version information.
2036 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
2038 * Supernovae builds are deprecated, and expected to be removed in the next
2041 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2043 SimGrid (3.9) stable; urgency=low
2045 The Grasgory release: GRAS is really dead now.
2047 * Complete overhaul of the internal host structures scheme.
2050 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
2051 was considered as experimental and badly maintained since then.
2052 * Keeping it was thus a trap to our potential users, that could take
2053 it instead of MSG or SMPI by mistake despite is pity state.
2054 * GRAS seems to have very few users (if any), and no one volunteered
2055 to maintain it further. It also induces a lot of XBT code (for
2056 portability sake), that must be maintained too.
2057 * For all these reasons, we killed GRAS. If someone wants to revive it
2058 in the future, don't cry, our git history still remembers of GRAS.
2061 * Major overhaul. Merge our documentation again as time proved that
2062 splitting it was really not helping our users.
2063 * Further improve the developer documentation to help newcomers
2064 hacking on SimGrid itself. The user documentation (and in
2065 particular, the beginner documentation) is still in a sorry state.
2068 * Now works on Windows too!
2069 * Much more extensive test suite, from MPICH
2072 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
2073 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
2074 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
2075 * Bug fix: let task be scheduled when the last dependency to be solved is
2076 a control dependency.
2077 * Remove SD_load_environment_script function.
2078 Use the C sg_platf function if you want to declare a platform
2082 * New function: MSG_process_get_number()
2083 * Old function documented: MSG_config()
2084 * Remove MSG_load_platform_script function
2085 Use the C sg_platf function if you want to declare a platform
2089 * Change the default value of the TCP_gamma constant (maximal size of TCP
2090 congestion window) to a more realistic 4MiB value. If you notice changes in
2091 your simulation results, you can fall back to the previous 20k tiny window
2092 by adding --cfg=network/TCP_gamma:20000 on command line.
2093 * (Hopefully) fix a bug wrt periodic availability and state traces
2094 * Bug fix: use default values at start when first event in availability/state
2095 trace is not at time 0.
2098 * remove the "new_" part of function name sg_platf_new_trace_connect
2099 (resulting in sg_platf_trace_connect), since it does not create
2103 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2104 use case where it's really mandatory, and maintaining it was a pain in
2106 * New: xbt_fifo_search(), search an item with a user-provided
2107 comparison function instead of dumb pointer comparison.
2110 * Fix the lua deployment:
2111 Use `simgrid.init_application()` before deployment instead of
2112 `simgrid.msg_register_application()` after.
2115 * Transfer the tracing files into the corresponding modules.
2117 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2119 SimGrid (3.8.1) stable; urgency=low
2121 The "we are told that some people want to also *install* the simgrid
2124 * Add missing file "tesh.1" to the archive.
2126 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2128 SimGrid (3.8) stable; urgency=low
2130 The Psssshiiiit release: SimGrid jumps into the Cloud.
2133 * Add an experimental interface to manipulate VMs. They are mainly
2134 process groups with very few intrinsic semantic, but they should
2135 allow you to build the semantic you want easily.
2136 * New function: MSG_host_set_property_value()
2137 * New function: MSG_process_on_exit(). To clean memory in all cases.
2138 * Bug fixes that made the host (and link) failures unusable.
2139 * Add a way to auto-restart process when the host in which they are
2140 executing comes back (ON_FAILURE="RESTART" on deployment file,
2141 MSG_process_auto_restart_set).
2142 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2143 please stop using the old ones, they are DEPRECATED.
2145 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2146 Please use MSG_init() instead. (reducing the amount of entry
2147 points in the library helps us).
2148 * Make it impossible to link against the wrong version of the lib
2149 * Deprecate MSG_clean(). No need to call it anymore.
2150 * Function MSG_get_host_number() is not deprecated anymore.
2153 * Split the doc into a user guide and a reference guide.
2154 * Start a developper guide to help people hacking on SimGrid.
2157 * Enable tracing by default. This modules rocks you should use it.
2158 * Remove option custom_flags. Now use environment variables CFLAGS
2160 * Use default cmake things to detect lua instead of home grown ones.
2161 * New option "enable_mallocators" to disable mallocators, for debugging
2162 purpose ("on" by default).
2165 * Bug fixes around the resource failures: don't let the processes
2166 survive the host they are running onto.
2167 * Add an interface to auto-restart processes when the host in which they are
2168 executing comes back.
2169 * Ensures that SIMIX_clean is called automatically. It's not part of
2170 the public interface anymore (bindings should be updated).
2173 * Bug fix for when SD_Simulate is called with a positive value: be careful
2174 when comparing doubles. Sometimes they are different for non significant
2176 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2177 parallel task whose initial work is distributed among host according
2178 to the Amdahl's law. Such tasks are created with a parameter alpha
2179 that corresponds to the non-parallelizable part of the computation.
2180 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2181 between two sets of workstations assuming a 1D block distribution (each
2182 workstation owns a similar share of data) on both sides.
2184 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2185 Data redistribution will be automatically scheduled once parent and child
2186 are both scheduled. The filling of computation_amount and
2187 communication_amount structures is now done seamlessly thanks to the chosen
2189 * New function SD_workstation_dump to display various information
2190 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2191 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2192 bandwidth on the route onto which the task is scheduled and the amount of
2194 To divide the nominal bandwidth by 2, the rate then has to be:
2195 rate = bandwidth/(2*amount)
2196 * Compute tasks that have failed can now be rescheduled and executed again
2197 (from their beginning)
2198 * Increasing source code coverage (src/simdag is now covered at 95.8%
2202 * Re-implement time-independent trace replay using SMPI (at the
2203 smpi_smp_* level) instead of MSG. This should replace
2204 examples/msg/actions/actions.c
2205 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2206 hindexed and structs)
2207 * Implement the exchange of non-contiguous data.
2208 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2209 * Correct behavior of smpi/sender_gap and set its default value to 0
2210 * Add option to asynchronously send small messages to allow better
2211 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2212 specifies the size in bytes under which messages will be asynchronously sent.
2213 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2214 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2217 * New C interface to define a platform: XML is now optional.
2218 For more info, please check include/simgrid/platf.h.
2219 * New interface to define random platforms from the C:
2220 For more info, please check include/simgrid/platf_generator.h and
2221 examples/msg/masterslave/masterslave_platfgen.c
2222 * Export a sg_cmdline dynar containing all the arguments we got from
2226 * Two new tracing options for adding comments to trace file so you
2227 can track your experiments (see --help-tracing for details).
2228 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2229 * Adding the SimGrid version that generated the trace file as a comment.
2230 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2231 * Fix to avoid key clashes on Paje links
2232 * Other minor fixes related to the Paje specification
2235 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2236 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2237 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2238 * Greatly improve the robustness of mmalloc to user errors (such as
2239 using an area after freeing it, or freeing it twice)
2241 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2243 SimGrid-java (3.8.1) stable; urgency=low
2245 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2246 Warning: all methods are visible, but only some of them are
2247 implemented so far. Check the source (src/jtrace.c)
2248 for further information.
2249 * New module: org.simgrid.msg.File (SimGrid File management functions)
2250 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2251 * Change the meaning of Process.restart: now restart the process from
2252 the beginning, like MSG_process_restart in C.
2253 * Add Process.setAutoRestart: handling of process restart when failed
2255 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2256 you to retrieve the properties of the processes/hosts
2257 * Deprecate Msg.clean(): you can just forget about it now.
2258 * New function Process.getCount(), that only works when compiling
2259 with the not yet released version 3.9 of the C library.
2261 * New context factory based on Coroutines. It mandates a modified JVM
2262 but then, the simulations run about five times faster, and there is
2263 no limit to the amount of processes (beside of the available memory).
2265 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2267 SimGrid (3.7.1) stable; urgency=low
2270 * Restore the prototype of MSG_process_create_with_environment() to
2271 the pre-3.7 situation by removing the kill_time argument.
2272 * Add a MSG_process_set_kill_time() function instead.
2275 * Fix weird behaviors when dealing with parallel tasks.
2278 * Simgrid is now built as a dll.
2279 * Simgrid-java now works on Windows.
2280 * Simgrid-Java is now included into Windows package.
2283 * First pre-build package for MacOSX.
2286 * Fix compilation when using MSG_USE_DEPRECATED.
2287 * Fix some compilation issues on Macs and Windows.
2288 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2289 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2291 We discovered that the Lua console is broken, but we are missing the
2292 manpower to fix it right now. The problem existed in 3.7 too, so we
2293 are not blocking the release for that. Sorry if you depended on this
2294 feature, any help would be really welcome.
2296 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2299 SimGrid-java (3.7.1) stable; urgency=low
2301 The "Java aint got to be bloated and slow" release
2304 * Various internal cleanups and performance improvement
2305 Simulations are expected to run up to twice faster or so
2306 * Make Process.kill(process) an instance method, not a static one
2307 * User processes are not java.lang.Thread subclasses.
2308 This breaks the compatibility (sorry), but previous API was
2309 brain-dead, making it impossible to have non-trivial
2310 initializations in the process constructor.
2311 * Require a full constructor per Process sub-class.
2312 Kinda breaks the compatibility (sorry), but this allows a much more
2313 efficient way to launch the processes at simulation startup.
2314 * Do not embeed our version of semaphores, java 1.5 can be considered
2315 as sufficiently prevalent for us to not dupplicate its features.
2319 * Add examples for almost every part of the API
2320 We spotted and fixed a lot of bugs in the process
2322 * New module: asynchronous communication API
2323 * New function: Process.sleep()
2324 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2325 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2326 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2328 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2330 SimGrid (3.7) stable; urgency=low
2332 The "spring cleanups (before the next Big Project kicks in)" release.
2335 * We can specify the SMPI latency/bandwidth factor with command line
2336 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2337 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2338 You can also use the "config tag" from platform file by adding this line
2339 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2340 (see "example/platforms/tag_config.xml" to use "config tag").
2341 Note that the command line supersedes the platform file configuration.
2342 * Change the correction factors used in LMM model, according to
2343 the latest experiments described in INRIA RR-7821.
2344 Accuracy should be improved this way.
2345 * Use the partial invalidation optimization by default for the
2346 network too. Should produce the exact same results, only faster.
2347 * Major cleanup in surf to merge models and split some optimization
2348 mechanisms from the core of the models. As a result you can now
2349 specify which model to use (e.g., --cfg=network/model:LV08
2350 --cfg=cpu/model:Cas01) and which optimization mode to use
2351 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2352 Incompatible combinations should err at initialization. See
2353 --help-models for the list of all models and optimization modes.
2354 * The CLM03 workstation model was dropped for simplicity because it
2355 used the deprecated CM02 network model. Use default instead.
2356 * Rename the TCP_gamma configuration option to network/TCP_gamma
2357 * Rename the coordinates configuration option to
2358 network/coordinates, and document it
2359 * Use now crosstraffic keyword instead of the terribly misleading
2360 fullduplex keyword. It is activated by default now in the current
2361 default model, use --cfg=network/crosstraffic:0 to turn it off.
2362 * Ongoing refactoring the model parsing to make XML files optional
2363 See include/simgrid/platf.h for details (still to be completed)
2366 * Major overhaul of the documentation. Almost instructive now :/
2367 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2368 functions and friends. This interface was considered as
2369 deprecated since over 2 years, it's time to inform our users that it is.
2370 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2371 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2372 use these (crappy) functions in your code.
2373 These functions will be removed soon. Stop using them now.
2374 * Deprecate MSG_get_host_{table,number}
2375 Implement MSG_hosts_as_dynar() instead.
2376 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2377 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2380 * Stabilize the parallel execution mode of user contexts
2381 * Introduce configuration variables to control parallel execution:
2382 - contexts/synchro: Synchronization mode to use when running
2383 contexts in parallel (either futex, posix or busy_wait)
2384 - contexts/parallel_threshold: Minimal number of user contexts
2385 that must be part of a scheduling round to switch to parallel
2386 execution mode (raw contexts only)
2387 * Fix bugs that prevented to use suspend/resume along with
2388 synchronization structures.
2389 * Fix bugs in process termination that lead to invalid memory access
2390 in very specific conditions.
2393 * Introduce a parallel mode for the models (controlled by surf/nthreads
2394 configuration item). In our tests, running the models in parallel
2395 never lead to any speedups because they are so fast that the gain
2396 of computing each model in parallel does not amortizes the
2397 synchronization costs, even when ultra fast futexes are used.
2398 This is released anyway because YMMV.
2401 * Performance boost by using a swag internally to compute the set of
2402 tasks that are finished and should constitute the return value of
2406 * Enable it by default now that it is considered rather stable.
2409 * Documentation of the tracing functions.
2410 * Performance gains when tracing categorized/uncategorized resource
2411 utilization by avoiding calls to get route when updating resource
2412 variables. LMM constraints are being used instead.
2413 * API changed to set task categories. Use MSG_task_set_category instead
2414 of TRACE_msg_set_task_category, and SD_task_set_category instead
2415 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2416 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2417 * Examples for link user variables added (see at examples/msg/tracing/)
2418 * Deprecated function TRACE_msg_set_process_category completely removed
2419 * Trace header updated according to the latest Paje file format
2420 * Tracing network lazy updates, no longer obligate users to use full updates
2421 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2422 * Experimental: let user code declare/set/push/pop application states for hosts
2423 * API changed to allow the manual creation of graph configuration files
2424 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2427 * Improve the API of Lua MSG bindings, using the Lua spirit.
2428 * Each simulated process now lives in its own Lua world (globals are
2429 automatically duplicated). It helps writing simulators. Will allow
2430 to run Splay programs within SimGrid in the future.
2431 * Add a Chord example in Lua, equivalent to the MSG one.
2434 * Start the implementation of a solution to express temporal
2435 properties, not only local assertions. This is still an
2436 experimental work in progress, stay clear from it to be safe.
2440 - Add new runtime parameters --help-logs and --help-log-categories
2441 to display information about supported logging parameters and
2443 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2444 don't exists anymore.
2445 * Mallocators: allow value NULL for the reset function.
2447 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2448 create homogeneous dictionaries, where all the elements share the
2449 same free function. Non homogeneous dictionaries will be
2450 deprecated in the next release.
2451 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2452 - Multi-level dictionaries are deprecated.
2454 - new function xbt_dynar_search_or_negative() that is useful when
2455 you have less than 2 million elements in your dynar and don't
2456 want of the extra complexity of catching exceptions when the
2457 element is not found.
2459 - Make xbt_os_thread module (for thread portability) public.
2460 Documentation is still to come, sorry.
2462 - Cleanups and simplifications to make it maintainable again.
2463 - Exotic features (such as memalign and valloc) were removed.
2464 - The metadata were extended and improved so that the
2465 model-checker becomes able to explore and inspect the heaps.
2466 - This may induce a performance drop when enable_model-checking is
2467 ON in cmake (even if it's not used in the simulation), but it is
2468 necessary at this point to get MC working.
2470 Turn model-checking OFF if simulation performance matters to you.
2471 Not enabling it at runtime is not enough, disable it in cmake.
2473 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2476 SimGrid (3.6.2) stable; urgency=low
2478 The "Not coding new stuff allows to polish old things" release.
2481 * New bindings to the NS3 packet level simulator (experimental)
2482 * Use the raw (efficient) execution contexts instead of the sysv
2483 (portable) ones when possible.
2484 * libpcre is now mandatory in any cases since not using it led to
2485 severe performance loss and possibly other issues
2486 * Update the XML platforms:
2487 - G5K: include the latest machine in Nancy
2488 - GridPP and LCG: new platforms
2489 * Documentation was partially updated, at least (more to come)
2491 Bug fixes, cosmetics and small improvements
2492 * Free terminated processes before the end of the simulation to avoid
2493 exhausting the memory of users having very dynamic amount of
2495 * Bug fix and cosmetics about canceling non-running tasks
2496 * Bug fix about the dot loader's issues when using libcgraph
2499 * Create an installer for windows with nsis (amd64 and win32)
2500 - Add an hello world project to illustrate simgrid project creation.
2501 - Embed libpcre into the Simgrid installer to avoid
2502 its compilation burden
2503 * The raw execution contexts should work on Apple now
2504 * Port to Windows 64 bits
2505 - Sysv contexts now have an implementation for this arch
2506 - GRAS communication features now support this arch
2507 * Drop support for borland compiler on windows
2508 - this code was not maintained, and we kinda depend on gcc nowadays
2509 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2510 * Fix portability issue on unstable ubuntu: linker became picky on
2513 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2515 SimGrid (3.6.1) stable; urgency=low
2517 The "Oops, we broke Macs too" release
2520 * Fixed contexts detection so that raw ones are used when possible
2521 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2522 leads to a strange error, with user code segfaulting sometimes when
2523 the generated code is not perfectly aligned (which is not
2524 controllable from the user side, depends on the amount of code)
2527 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2529 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2531 SimGrid-java (3.6) unstable; urgency=low
2534 * Split of every thing from simgrid v3.5 into a separate package.
2536 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2538 SimGrid (3.6) stable; urgency=medium
2540 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2543 * Bindings now constitute their own package, separated from the main one.
2544 Rationale: reduce our maintenance nightmare by reducing the module coupling
2545 They will soon be released on their own on gforge.
2547 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2548 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2550 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2551 * It's not quite deprecated for now because we have no replacement,
2552 but it may soon become the case.
2555 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2556 * Fortran: New user-level cache variable to store the rank of the running
2557 process. This improves performance by an order of magnitude.
2558 * C: New coccinelle script to automatically locate and modify global and
2559 local static variables.
2560 * Improved SMPI network model with a sender-side gap to account for multiple
2564 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2565 only say if a communication is finished, no matter whether it succeeded or
2566 failed. You can call MSG_comm_get_status() to know the status of a finished
2568 * New function MSG_task_dsend() to send a task and detach it. When a
2569 communication is detached, you are never notified of its success or failure
2570 and the memory is released automatically once it is finished. This function
2571 is useful when you don't care about the end nor the success of a
2573 * Change the prototypes of action replay. Sorry for inconvenience,
2574 but this is really more efficient this way (and to adapt your code,
2575 you just have to fix the initialization, that shouldn't be too long)
2576 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2577 ever managed to do anything useful with it.
2578 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2579 takes a dynar of communications. It returns immediately and gives the
2580 index of a finished communication (if any).
2581 * New example: a basic implementation of the Chord P2P algorithm.
2584 * New model for multi-core CPUs. You can now use the core attribute to
2585 precise the number of cores of a host. This is a basic model. Every
2586 process running on the host receives at most the power provided in
2587 the DTD (throughput<=power). Total throughput of process cannot exceed
2589 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2590 router linked by an up-link and a down-link (possibly asymmetrical).
2591 This kind of pattern allows to easily build last-mile model style platforms.
2592 Aggregating such patterns in a rule-based AS is thus the technique of
2593 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2594 * New model for Vivaldi routing. We transformed the Vivaldi network model
2595 into a Vivaldi routing model (based on the rule-based model). This allows to
2596 combine Vivaldi based latencies with last-mile platforms.
2599 * Added a check for NaN of IEEE754 infinite in the double entries of
2601 * Introduce a new context factory "raw", highly inspirated from the
2602 ucontext factory, but using manually crafted functions in assembly to
2603 do the work in an efficient manner.
2604 * Allow to change the used context factory at run time, not only at
2605 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2606 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2607 contexts at runtime (only with raw contexts or ucontexts).
2608 * Completely rewrote this module to allow parallel execution of user
2609 processes. Use --cfg=contexts/nthreads:N to execute user processes
2610 with N parallel threads (the default is 1, meaning no parallelism).
2611 * Allow to decide dynamically between sequential and parallel modes.
2612 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2613 processes in parallel only when their number is greater than or equal to P
2615 * Added a check for NaN of IEEE754 infinite in the double entries of
2619 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2620 won't output the state of processes when interrupted with Ctrl-C
2621 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2622 NULL-terminated array. This may solve backward compatibility issues
2623 due to the change to return type of SD_simulate. See also:
2624 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2625 * Add new macros with variable number of arguments.
2626 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2627 - in xbt/asserts.h: xbt_assert
2628 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2629 - in xbt/ex.h: THROWF and RETHROWF.
2630 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2632 * Change xbt_die() to accept a format string with arguments, just like printf.
2633 * New data structure: xbt_lib_t, like a dict but more general and with better
2637 * New configuration options
2638 Options triva/categorized and triva/uncategorized can be used to generate
2639 graph configuration files for Triva visualization tool.
2640 * Configuration option tracing/platform is renamed to tracing/categorized
2641 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2642 * New TRACE_user_link_variable interface:
2643 User provides the name of the link and the tracing variable to attach to it
2644 * the declaration of tracing categories must be done after the environment
2646 * simpler tracing interface, just one way to declare categories
2647 TRACE_category or TRACE_category_with_color, it is up to you
2648 * links in the trace file are again identified by their names
2649 * trace contains the full platform hierarchy exactly as declared using the ASes
2650 * Options tracing/msg/[task|process]:1 groups the process by hosts
2651 for both cases, tasks and processes must have names that are unique during the simulation
2652 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2653 * The experimental option tracing/msg/volume is deprecated
2654 its functionality may be reincorporated if needed
2656 The tracing generates a trace file with unordered timestamped events,
2657 because of the way the core simulator (surf) works. A script available
2658 at the tools directory (fix-paje-trace.sh) can be used to put the events
2659 in order. We have changed the tracing so it can generate ordered timestamped
2660 events in the final trace, but depending on the simulator (and how much time
2661 is simulated) that can lead to a huge memory utilization. It is deactivated
2662 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2664 Build Infrastructure
2665 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2666 - We are trying hard to keep the API stable, but it may happen that
2667 some things change (we're a research project after all, not a
2668 nuclear plant operating system). If such things should happen, you
2669 could rely on that macro to adapt.
2670 - current value: 30600 for 3.06.00, aka 3.6
2671 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2672 a number that can be compared with SIMGRID_VERSION.
2673 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2674 assertions and verbose/debug logging events are disabled at compile time.
2676 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2678 SimGrid (3.5) stable; urgency=medium
2681 * New feature to any SimGrid-based simulator: Model-Checking
2682 Check SIN#1 for more details.
2685 * New Model SMPI (three-interval linear regression for correction factors)
2686 See RR-7426, available at http://hal.inria.fr/inria-00527150
2687 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2688 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2689 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2690 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2691 * See also src/smpi/README
2696 - Tracing API changes: TRACE_start and TRACE_end should not be called
2697 by user-code. They are automatically called by simulators created
2698 with SimDAG, MSG and SMPI if the toolkit is compiled with
2699 tracing_enabled=ON. Categories declaration and utilization remain the
2700 same for MSG and SimDag.
2701 - A function was added to the tracing API to declare categories with
2703 - TRACE_category_with_color (char *category, char *color)
2704 where color must be in the following format
2705 "%f %f %f", red, green, blue
2706 and red, green, blue are float values in the interval [0, 1]
2707 - User can specify NULL as color parameter, or continue calling
2708 TRACE_category (cat)
2709 On that case, the tracing system will define random colors
2710 - The following command-line options are supported:
2711 --cfg=tracing/filename:msg.trace
2712 --cfg=tracing:1 (activate tracing, needed to use others)
2713 --cfg=tracing/platform:1 (categorized resource use)
2714 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2715 --cfg=tracing/msg/task:1 (task creation)
2716 --cfg=tracing/msg/process:1 (process creation, migration)
2717 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2718 --cfg=tracing/smpi:1 (SMPI interface tracing)
2719 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2720 - examples of examples/msg/tracing updated
2722 - DAXLoader and DOTLoader functions can generate tasks with categories
2723 - A new function to attribute a category to SD tasks:
2724 TRACE_sd_set_task_category (SD_task_t task, char *category)
2725 * Tracing the MPI interface implemented by SMPI
2726 - Collective operations are traced with states
2727 - Point-to-Point operations are traced with states/links
2728 - Tracing activated by a parameter "-trace filename" passed
2729 to smpirun during execution (considering that simgrid
2730 is compiled with tracing enabled)
2731 - To run the simulation with gdb, the simulator
2732 accepts --cfg=tracing/smpi:1 to trace SMPI
2733 - tesh files to check if smpi tracing is ok
2734 - See examples/smpi/NAS/DT-trace
2735 * GTNetS tracing re-worked
2736 - adaptation to the tracing system of GTNets to cope
2737 with modifications regarding the fullduplex mode
2738 - new tesh files to check if gtnets tracing is ok
2741 * Asynchronous communications through the functions:
2742 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2743 * New function: MSG_load_platform_script()
2744 to make possible using a lua script instead of XML files to set up platforms
2745 * New function: MSG_set_function
2746 to associate functions to processes, used when bypassing the parser
2747 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2749 Platforms: Add some more examples in examples/platforms
2750 * Grid'5000: see www.grid5000.fr
2751 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2754 * Change the XML format. This is a very important modification. SimGrid
2755 3.5 introduces a new hierarchical format based on the notion of
2756 Autonomous Systems. Compatibility with old format is ensured
2757 through the perl script provided in the install bin directory
2758 bin/simgrid_update_xml.
2759 It is now possible to build platforms with specific routing
2760 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2761 connect several platforms together. We will try to provide soon
2762 set of realistic platforms exploiting these properties (have a
2763 look at examples/platforms/ for the moment).
2764 * Take the opportunity of the XML format change to be a good XML citizen:
2765 rename link:ctn to link_ctn and similar changes (also dealt with by
2767 * Add a new routing scheme (rule-based) using regular expressions. It
2768 enables to have an extremely low memory footprint when the
2769 underlying routing is simple and can be compactly described. You
2770 need to have libpcre4-dev (perl regular expressions) installed if
2771 you want to use this routing scheme.
2772 * Revive the cluster TAG and allow to easily and efficiently (both in
2773 term of memory and speed) connect clusters together. Have a look
2774 at teshsuite/simdag/platforms/ to see how this can be done. With
2775 this tag, you can create clusters with thousands of tasks at no
2776 cost (have a look at examples/platforms/).
2777 Note: clusters are implemented as ASes, so there is no need for an
2778 enclosing AS tag if you have only one cluster in your platform.
2779 * Add new generic functions in the public interface that allows the user
2780 to call SURF 'create_resource' methods from your code (same
2781 functionality as the XML bypass mechanism but with a much lighter
2783 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2784 that uses a piecewise linear approximation to produce better
2785 results when exchanging small messages.
2786 * Add a new parameter to handle correctly full duplex link and account
2787 for interferences between uplink and downlink communications
2788 (activate with --cfg=fullduplex:1).
2791 * Rename the SD_READY (all dependencies are satisfied and task is
2792 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2793 dependencies are satisfied) state.
2794 This prevents a confusion between the notion of "ready to schedule"
2795 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2796 simulated" (SD_RUNNABLE) used by the simulation kernel.
2797 * Change the way a task is considered as ready. Instead of removing
2798 dependencies when a task is done, a counter is decreased. This way,
2799 it is always possible to reach ancestors thanks to the
2800 SD_taks_get_parents function (even after the end of the simulation.)
2801 * Change the return type of SD_Simulate from (SD_task_t*) into
2802 xbt_dynar_t. This function was in handling a dynar internally and
2803 converted it into a NULL terminated array for historical reasons.
2804 * New function SD_dotload(char*) to load a DAG described in dot
2805 format. This loader and the corresponding examples require the
2806 installation of the graphviz library.
2807 * Fix a bug in the management of tasks of size 0 in the surf network
2808 models. This problem was only visible with SIMDAG and you should
2809 thus disregard results produced with earlier versions if you
2810 relied on this feature (some tasks were blocked because of this).
2811 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2812 even though your applications were doing only point-to-point
2813 communications and sequential computations. Now you can really use any
2814 model you want (of course, if you create real parallel tasks, which are
2815 not implemented in most models beside ptaskL07, this will abort).
2816 * Add an example that schedules a DAX on an heterogeneous platform
2817 using a Min-Min strategy.
2818 * New function SD_workstation_get_current_task() that returns the kind
2819 of task currently running on a workstation in the sequential access
2821 * Raise some warnings when unexecuted tasks remains at the end of the
2822 simulation. This is usually caused by cycles in the DAG.
2825 * New function: SIMIX_process_set_function() called by MSG_set_function
2826 * Change the underlying waiting queue in semaphores so that a process
2827 can wait on several of them simultaneously (as in waitany).
2828 * Fix the way to handle tokens in semaphores so that all access patterns
2829 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2830 * kill the dirty pimple SIMIX_message_sizes_output()
2831 Please use (proper) visualization instead
2834 * New data container: setset (set of sets of elements)
2835 * New module: mmalloc (mapped malloc, allowing to have several
2836 independent segments of malloc)
2837 * New function: xbt_dict_cursor_set_data()
2838 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2839 * New function: xbt_dynar_is_empty()
2840 * New function: xbt_fifo_get_last_item()
2841 * Fix xbt_dynar_shrink(): use the right element size.
2842 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2843 uninitialized areas during expand.
2844 * Fix semaphores: previous implementation was severely broken.
2845 * Use library init/fini functions for our initialization.
2846 - you can use logs and other feature as soon as you want in your
2847 code (even before the xbt_init / MSG_init)
2848 - xbt_exit is now a no-op and produce a warning when used.
2851 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2852 benefit from the latest improvement to the simulation kernel.
2853 * Kill measurement sockets for now. If you rely on them, sorry. This
2854 release is not for you. This feature will be reintroduced in the
2855 future, but we cannot delay the release any further.
2856 * New function: gras_msgtype_get_name().
2857 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2858 * Fix (at last) the pmm example: it should not randomly fail anymore.
2860 Build chain: bug fixes and overall polishing
2861 * Cmake is now stable enough. Hence, we killed the autotools.
2862 * Port to windows ( TM :)
2863 * Fix the 'make install' target.
2864 No need to use 'make install-simgrid' anymore
2865 * Introduce a 'make dist' target compiling a *source* archive
2866 'make package' compiles a binary archive
2867 * Compile java files only on need
2868 * Add --cd and --setenv command line options to tesh
2869 * Out of source builds are not fully supported yet, but we are close
2870 * Enable supernovae and optimization flags by default for our users
2873 * Add layer to set up environment directly from lua, without XML.
2874 * The effect of gras_stub_generator can be achieved through
2875 lua too (check examples/gras/console/ping_generator.lua)
2877 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2879 SimGrid (3.4.1) stable; urgency=low
2881 The "Polishing easter eggs is probably a good idea" release.
2882 This is a bug fixes release only.
2886 * Fix a bug preventing the tasks from begin garbage collected.
2889 * Fix a bug occurring when a host involved in a communication fails.
2890 This was not detected properly by the other peer involved in the
2891 communication. Now, it's reported as a network error.
2894 * Warn the user about loop dependencies in data flow of DAX files
2895 * Obey the control-flow dependencies of DAX files
2898 * Add option "enable_smpi" allowing to not compile SMPI.
2899 Probably useful for the (Mac) users experiencing a build error here
2900 * Improve the detection of lua5.1 and ruby1.8
2902 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2904 SimGrid (3.4) stable; urgency=low
2906 The "Easter in Cargese" release. Also known as (major changes):
2908 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2909 ~> bindings were greatly improved
2910 ~> new tracing infrastructure for better visualization introduced
2912 * the "Welcome to configury modernity" release.
2913 ~> we switched from autotools to cmake, and improved our cdash
2916 A more detailled list of changes follow (full detail in svn log).
2918 Java Bindings: Various Cleanups
2919 * (install java-gcj-compat-dev on debian-like to use them)
2920 * Remove put/get: no need to export deprecated interface in Java
2921 Use send/receive instead.
2922 * Cleanup the examples and add a README per directory
2923 * Remove example autoDestination (that's the only way to go now)
2924 * Remove example explicitDestination (was a plain copy of basic)
2925 * Make JniException a runtime exception, so that there is no need to
2926 declare the fact that you may encounter such a beast. I guess that
2927 nobody will ever want to survive such error.
2928 * Create specific errors for each MSG case of failure:
2929 host failure, transfer failure, timeout, task cancelled
2930 * Cleanup the exceptions that may get thrown by each function
2931 * Other internal cleanups in Java bindings. Performance still bad :/
2932 Ruby and Lua Bindings: create them
2933 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2934 * That's new and great, you should try them out.
2935 Same functionalities than Java bindings, only even less polished
2937 * Kill the useless "rate" argument of SD_task_get_execution_time()
2938 Everyone used to provide -1 as a value, it was not used, and the
2939 semantic of a possible use wasn't even clear.
2940 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2941 either as comm costs or compute costs to mean that there is no
2942 such thing for that specific task.
2943 * Add a SD_task_set_name() function
2944 * Fix SD_task_unschedule() on typed tasks
2945 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2946 * In DAX loader, accept useless 'level' attributes to <job> since
2947 LIGO DAGs have them (seem to be to ease graphical representation).
2949 * Add an example masterslave_mailbox.c using send/receive and not
2950 the deprecated put/get interface.
2951 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2952 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2954 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2955 (declare MSG_USE_DEPRECATED to still have the old name)
2956 * Add a MSG_task_set_data() function
2957 * About trace replay (see examples/msg/actions):
2959 - Allow to work with splitted trace files for each process
2960 Give the specific trace file as argument of each process,
2961 and call MSG_action_trace_run(NULL)
2962 You can still have one merged file for all processes.
2963 - Fix implementation of collective operations
2964 * Allow task_execute() on 0-sized tasks (closes #10063)
2966 * This is the first release of SimGrid where SMPI is not considered
2967 beta anymore (even if some corners should still be improved)
2968 * Port over the new SIMIX_network submodule (internal refactoring)
2969 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2970 * Implement more missing elements of the standard:
2972 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2974 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2975 MPI_Scan() MPI_Get_processor_name()
2976 - Added implementation of missing case for Alltoall (warning: it's
2977 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2978 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2979 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2980 SMPI_MPI_Allgatherv()
2981 * Bug fixes include:
2982 - MPI_Waitsome() was broken
2983 - Allow relative includes in smpicc
2984 - Command line cfg argument 'reference_speed' was ignored...
2985 - Some functions did not properly lead to auto-benching of user code
2986 - smpicc passes -O2 by default (just like openmpi one)
2988 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2989 * Bug fixes about timeouts during communications
2990 * add SIMIX_message_sizes_output() as a pimple to write to file the
2991 amount of messages per size. Use gnuplot to get histogram.
2992 Pimple because that's the only user-visible function of simix,
2993 defined directly in xbt.h (irk, sorry)
2995 - Add a SIMIX_sem_get_capacity() function
2996 - Fix interactions with processe resume/suspende
2997 - release_forever() was stupidly broken
2998 - Fix SIMIX_display_process_status() for processes in a semaphore
2999 - Make SIMIX_sem_block_onto() user-visible
3000 * Refactoring context stuff:
3001 - Use pseudo-OOP for better modularity
3002 - reimplement SIMIX_process_kill() without process_schedule() so
3003 that the latter can take as invariant that it is called from
3005 - Merge context_start into context_new for sake of simplicity
3007 * Add a Vivaldi network model, coded live during SUD'10 ;)
3008 * Rename configuration variables to start a hierarchy:
3009 o cpu_model -> cpu/model
3010 o network_model -> network/model
3011 o workstation_model -> workstation/model
3012 * New configuration variables:
3013 o network/bandwidth_factor: correction to bandwidth
3014 o network/latency_factor: correction to latency
3015 o netwotk/weight_S: correction to the weight of competing streams
3016 * Add a long description to the models, that users can see with such
3017 argument on the command line: --cfg=cpu/model:help
3018 * --help-models display the long description of all known models
3020 * config: add the ability to set a default value after registration
3021 Does not override any previously set value (e.g. from cmd line)
3022 * dict: allow to have integer key and data.
3023 When so, you need to use the following functions
3024 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
3025 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
3026 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
3027 In contrary to regular dicts, the key is not malloced before copy.
3028 Mixing scalar and regular elements in the same dict is not tested
3030 * Allow to use xbt_dynar_shrink() to expend the dynar instead
3031 Tracing for Visualization:
3032 * SimGrid is now instrumented in order to generate a trace file for
3033 visualization analysis: to use it, need to compile SimGrid with the
3034 "tracing" option enabled, and instrument the program using SimGrid with
3035 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
3036 (among other functions).
3037 * The instrumentation only traces the platform utilization for now
3038 * Documentation to use the tracing functions and how to analyze the
3039 traces with the Triva tool is written.
3040 * More information about: SimGrid FAQ (in the section Tracing Simulations
3043 * We moved to cmake as default build system. Autotools support will
3044 be dropped soon. Check the FAQ for more info about how to use it.
3045 * Greatly improved our cdash/ctest interactions
3046 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
3047 * Added memory checking tests with valgrind; lot of memleak fixing.
3048 This may be the first release of simgrid with so few memory issues
3049 * Added code coverage tests.
3050 Our coverage is still improvable, but at least we see it on cdash.
3052 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
3054 SimGrid (3.3.4) stable; urgency=low
3056 The "Desktop Grid needs love too" release (also called Xmas release).
3058 Models improvements:
3059 * Major speedup in the maxmin system solving by using lazy evaluation
3060 Instead of solving completely the maxmin system at each iteration,
3061 only invalidate (and recompute) the modified parts.
3062 This new feature is enabled in default models but you can try to
3063 turn it on with "--cfg:maxmin-selective-update=1" for other models.
3064 * Cas01 IMproved as default CPU model
3065 This CPU model is the same Cas01 model, but it uses the
3066 maxmin-selective-update flag and a heap structure to manage
3067 actions on SURF kernel.
3068 It reduces the complexity to find the next action to finish and,
3069 consequently, it's faster than the old Cas01.
3070 This is the new default CPU model (Cas01).
3071 * Rename the old Cas01 model to Cas01_fullupdate
3072 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
3073 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
3074 * CpuTI (CPU Trace Integration)
3075 A new CPU model whose objective is simulate faster when using
3076 availability trace files.
3077 Instead of using a full featured, over engineered maxmin system for
3078 CPU modeling, this model does the pre-integration of traces files
3079 to calculate the amount of CPU power available, and so, executes
3080 faster than the old CPU models.
3081 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
3082 * Use LV08 as default network model since it gives better accuracy
3083 for small messages and shouldn't change things for big ones.
3084 Use --cfg=network_model:CM02 to get the previous behavior.
3087 ******************************************
3088 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
3089 ******************************************
3090 * The new CPU model may changes simulations!
3091 The point is that events occurring at the exact same timestamp
3092 are not scheduled in the same order with the old and new
3093 version. This may be enough to completely change the execution
3094 of simulations in some cases.
3095 * The new network model will change simulations!
3096 This new model is more realistic than the previous one, so you
3097 should consider redoing your old experiments with this model.
3098 Sorry for the inconvenience.
3101 * Introduce the supernovae compilation mode
3102 When compiled that way, the whole SimGrid (or almost) is put in a
3103 single compilation unit and compiled in one shoot.
3104 This is to help gcc which has difficulties to inline stuff from one
3106 The speedup seem to be above 15%, althrough more tests are needed on
3107 amd64 to confirm that gain.
3110 * Port of MSG's mailbox on top of SIMIX network
3111 The put/get mechanism was greatly simplified on the way.
3114 * New SIMIX network module. Provides:
3115 - Mailbox: rendez-vous mechanism to find with who you want to speak
3116 - Synchronous send/recv: easier and hopefully faster since the
3117 logic is handled in the maestro process directly now
3118 - Asynchronous send/recv: you dreamt of it? It's here now
3119 Too bad that nobody cared enough to propagate the change to MSG.
3120 * Add semaphores as SIMIX synchronization mechanism.
3123 * new function SD_daxload(char*) to load a DAX file
3124 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3125 * Introduce typed tasks. Specify its kind and cost at creation.
3126 At scheduling, just give where it should be placed, and the cost
3127 for each involved resource is automatically computed.
3128 Existing constructors so far (more to come of course):
3129 - SD_task_create_comm_e2e() for end-to-end communication
3130 - SD_task_create_comp_seq() for sequential computation
3131 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3132 * new function SD_task_dump() for debugging display
3133 * new function SD_task_dotty(task,FILE*) writing to file the info
3134 about the task in dotty format
3135 * SD_task_dependency_exists() can now cope with having one of its
3136 arguments NULL. If so, it tests whether the other argument has any
3138 * Add getters on list of preceding/following tasks:
3139 SD_task_get_parents(task) and SD_task_get_children(task)
3140 * Add getters on amount of workstations and list:
3141 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3142 * Add getter on task kind: SD_task_get_kind(task)
3143 * Update the start_time and finish_time of tasks on completion/failure
3144 * Bugfix: Remove task from state swags when destroyed
3147 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3150 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3151 dynar with one separate thread per value of the dynar.
3152 * Change the prototype of xbt_thread_create(), sorry.
3153 Added a boolean parameter indicating whether we want to join this
3154 thread (used in SG only for now)
3155 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3158 * GTNetS wrappers should now be usable again (and betterly tested too)
3159 * Fix a major regression from 3.2 where the timeout provided to
3160 MSG_task_put_with_timeout() was used as absolute time before which
3161 the comm should be done.
3162 * Start to fix the <cluster> tag.
3163 - Internal links should be good now (beside of the loopback, which
3164 use the private link instead)
3165 - paths to the external world is still rather broken
3166 - the <route:multi> tag is just broken. Actually that's brain-dead.
3167 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3168 to make it less stupid
3169 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3170 * Fix a source-level compatibility glitch from 3.2: after defining
3171 MSG_USE_DEPRECATED, you can use the old name
3172 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3173 * Allow to compile from the SVN with automake 1.11
3174 * Fix some problems when using the "start_time" tag in deployment XMLs.
3175 * Fix #8569: XBT/synchro.h has redundant declarations
3176 * Fix #8563: MSG return values and exceptions
3177 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3178 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3179 Thanks to Robson Peixoto.
3180 * Fix a vicious bug in dictionaries inducing that some elements were
3181 not freed on xbt_dict_free()
3183 Portability report of this version:
3184 * Main portability targets:
3185 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3186 on (amd64/i386/ia64)
3187 - mac leopard on i386
3188 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3189 but nothing critical.
3190 * Other platforms: windows, AIX and others were not tested for this release
3192 Timing report of this version:
3193 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3194 scenario parameters). From 8h to a few seconds in desktop grid settings.
3195 * Supernovae brings about 25% speedup on i386.
3197 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3199 SimGrid (3.3.3) stable; urgency=low
3201 The "Need for Speed" release.
3203 The timings done to validate the 3.3.2 were faulty.
3204 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3206 The problem was a conversion from a manually handled vector to
3207 xbt_dynar_t on the critical path.
3208 xbt_dynar_foreach calls functions, inducing stack management crap.
3210 We inlined these functions and xbt_dynar_foreach is now breath taking.
3211 We also inlined xbt_swag_belong on the way.
3213 Here are some approximate speedup measurements (on master/slaves
3214 simulations lasting between 10s and 20s each):
3215 3.3.1 -> 3.3.2: about same performance
3216 3.3.2 -> 3.3.3: 40% speedup
3217 3.3.1 -> 3.3.3: 40% speedup
3218 3.3.1 with inline patch -> 3.3.3: 30% speedup
3220 Our reading is that the refactoring which occurred in 3.3.2 made us
3221 suffer much more from the xbt_dynar_foreach low performance, but
3222 once we solved this, this refactoring proved to be very performance
3223 effective. From the 40% speedup, somehow, 10% are due to the
3224 inlining and 30% to the refactoring.
3226 That's a pitty that gcc cannot inline functions placed in other files
3227 alone. We have to choose between:
3228 - break the encapsulation (by putting private data structures and
3229 accessors in headers files to help gcc)
3230 - live with low performance
3231 - switch to a decent compiler such as icc (not quite possible).
3233 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3235 SimGrid (3.3.2) stable; urgency=low
3237 The "Simplicity does not preceed complexity, but follows it" release.
3239 The main contributors of this release were (lexical order):
3240 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3243 * Extract the routing logic into its own object.
3244 (was dupplicated in network.c and workstation_LV07.c;
3245 Allows to implement other ways of storing that info)
3246 => kill now useless network_card concept
3247 - Use dynar to represent routes (instead of void** + int*)
3248 - kill link_set (use surf_network_model->resource_set instead)
3249 - Add a command-line option to choose the routing schema to use
3250 - Add three new models:
3251 * Floyd (shortest path computed at initialization)
3252 * Dijikstra (shortest path recomputed all the time)
3253 * Cached Dijikstra (shortest path computed on need)
3254 All these models where contributed by Silas De Munck, and are
3255 described in his ICCS09 paper.
3257 * Simplify model declaration
3258 (less redirections, less function to write when defining a model)
3259 - Factorize stuff between models:
3262 surf_model_resource_set(model)
3263 surf_model_resource_by_name(model, name)
3264 - Unify the types of models in s_surf_model_t (using an union)
3265 - Embeed fields of common_public directly into s_surf_model_t
3266 - Rename model methods:
3267 action_free ~> action_unref
3268 action_change_state ~> action_state_set
3269 action_get_state ~> action_state_get
3270 - Change model methods into functions:
3271 (model)->common_public->action_use ~> surf_action_ref
3273 * Implement a generic resource; use it as ancestor to specific ones
3274 (allows to kill duplicated code in models)
3275 Drawback: timer command don't need no name nor properties;
3276 workstation_CLM03 don't need no properties
3277 (but I guess we can live with those few bytes wasted)
3279 * Improve the action object model
3280 - implement a constructor avoiding dupplicated code about field
3281 initialization in generic_action part.
3283 * Kill the SDP model: it has an external dependency, is deprecated
3284 in flavor of modern lmm models, and didn't compile since a while
3287 * Relocation of the context module from XBT to SIMIX.
3288 (the context were decoupled from the simix processes, duplicating a lot of code)
3289 => a lot of code was factorized
3290 - less overhead is introduced during scheduling
3291 - simpler API for the context factory
3292 - the logic for process creation,destruction and manipulation was simplified
3293 * Simplification of the s_smx_process_t data structure.
3294 => accesing the simix level data associated to a process is faster now,
3295 and the code is a lot more readable.
3298 * Implement some more MPI primitives:
3299 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3300 -implementation: Bcast: flat or 2-ary tree (default),
3301 Barrier: 4-ary tree,
3303 Allreduce: Reduce then Bcast
3304 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3305 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3306 Alltoallv: flat tree, like ompi
3308 * Add support for optimized collectives (Bcast is now binomial by default)
3309 * Port smpirun and smpicc to OS X
3312 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3313 Shout out if you used it.
3316 * Display the list of still queued messages in SG mode when existing
3320 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3321 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3322 * Add xbt_str_from_file(FILE*)
3323 * Add xbt_dict_get_key achieving a linear reverse search
3324 * Remove the context module
3326 Portability report of this version:
3327 * Main portability targets:
3328 - Linux(debian)/x86/context
3329 - Linux(debian)/x86/pthreads
3330 - Linux(debian)/amd64/context
3331 - Linux(debian)/amd64/pthreads
3332 On these, we still have the eratic breakages of gras/pmm and
3333 amok/saturate_sg reported in previous version. We still think
3334 that the tests are the cause of the fault, not the tested code.
3336 - Mac OSX Leopard/x86/context
3337 Still false negative in tesh autotesting.
3338 Smpi still fails, but this time because readlink does not accept -f
3339 Everything seems to work properly beside of that.
3342 - AIX version 5.3 (only tested contexts this time)
3343 Smpi still fails there because mktemp is not installed.
3344 Everything seems to work properly beside of that.
3346 I managed to compile it for the first time, but several breakages.
3347 Won't delay the release for this exotic platform.
3349 * Windows: it's still lagging behind. If you want to help, please
3352 Timing report of this version:
3353 This version seem to be more than 5% faster than 3.3.1 (on linux
3354 64bits with contextes). The gain is less than expected, we are
3355 investigating this for next release.
3357 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3359 SimGrid (3.3.1) stable; urgency=low
3362 * Implement a --cfg-help to show existing configuration variables
3363 * Build chain do not require doxygen in maintainer mode
3366 * fix a bug on struct sizeof computation, which prevented the
3367 exchange of arrays of structs in some conditions
3368 - added a regression test about this in datadesc_usage
3369 * Allow the exchange of 0-long dynamic vectors.
3370 - for that, use -1 as indicator of dynamic size instead of 0
3371 - This implied to change any size from unsigned long to long,
3372 reducing a bit communication abilities, but I guess that with
3373 64bits being quite common, this is more than enough.
3374 - This also induce a protocol change, thus bumping network protocol
3375 version from 0 to 1 (if we have external users, we have to get
3376 clean on that point too ;)
3377 - added two regression tests about this in datadesc_usage
3378 * Be more verbose when propagating local exceptions
3379 This helps debugging.
3380 * Display the status of simulated processes when receiving SIGINT in
3384 * Allow to control the simulation from a trace file.
3385 New functions MSG_action_register() and MSG_action_trace_run()
3386 The first one allows to associate a function execution to each
3387 kind of action while the second one parses a trace file and
3388 triggers the corresponding actions within the system.
3389 For now, only a toy example is provided in examples/msg/actions
3390 * Add an example of process migration in examples/msg/migration
3391 * Fix a bug in task exchange which broke MSG_task_get_sender()
3392 Add a teshsuite regression test for that.
3393 [Bug: if MSG_task_get_sender() is called after sender exit,
3395 * Fix a bug which prevented suspend/resume to work properly
3396 * Display the status of simulated processes when receiving SIGINT
3397 This fixes a regression of v3.3. due to the introduction of SIMIX
3398 * Bug fixing in failure management:
3399 - trace could not start by a failure at time 0
3400 - failure during communications were not working
3403 * Add SIMIX_process_set_name() to change the name of the current
3404 process in the log messages.
3405 * Store smx_hosts in a dict since we only retrieve them by name
3406 * Move the configuration infrastructure to surf
3409 * Move the configuration infrastructure to surf
3412 * Massive internal cleanups:
3413 - Store internal structures on processes instead of hosts (allows
3414 to have more than one process per host, in addition of being more
3416 - Cleanup the initialization/finalization process
3417 - Kill a whole bunch of unneeded synchronization:
3418 processes run in exclusive manner within the simulator
3419 - Move queues from global tables to process data fields
3421 - now accept -platform and -hostfile arguments
3422 - Pass the right rank value to processes according to the hostfile
3423 * Compile the examples by default, and use them as regression tests
3424 * Implement MPI_Wtime()
3425 * Change the reference speed to a command line option
3428 * TCP_gamma can now be specified as command line option using
3429 --cfg=TCP_gamma:10000000.0
3430 * Change the --surf-path cmd line option into --cfg=path:
3433 * Also include strbuff from xbt.h public header
3434 * xbt_ex_display(): do not free the exception after displaying
3435 This allows to do more with the given exception afterward.
3436 Users should call xbt_ex_free() themselves.
3440 Portability report of this version:
3441 * Main portability targets:
3442 - Linux(debian)/x86/context
3443 - Linux(debian)/x86/pthreads
3444 - Linux(debian)/amd64/context
3445 - Linux(debian)/amd64/pthreads
3446 These targets fail about 1/10 of times on gras/pmm, but we believe
3447 that this is because of the test, not because of simgrid.
3448 amok/saturate_sg fails even more rarely, and the test may not be
3451 - Mac OSX Leopard/x86/context
3452 The test suite still spits tons of errors because some obscure
3453 force prevents us from removing the temporary directories
3454 arguing that they still contain some metadata I've never heard of.
3455 Smpi fails because seq is not installed.
3456 Everything seems to work properly beside of that.
3459 - AIX version 5.3 (both contexts and pthread)
3460 Smpi still fails there because mktemp is not installed.
3461 XML inclusions seems rosty on AIX.
3463 * Windows: it's still lagging behind. If you want to help, please
3466 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3468 SimGrid (3.3) stable; urgency=high
3472 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3473 [Malek Cherier & Mt]
3475 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3478 Until now, GRAS were using MSG as an interface to SURF. It was
3479 quite difficult because both interface have several differences
3480 (MSG channels vs GRAS sockets were the most notable point).
3482 This also opens the gate to SMPI (which should occur soon) and speed
3483 up simulations by to 40% (even if it were not the main goal).
3485 **************************************
3486 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3487 **************************************
3488 The point is that events occurring at the exact same timestamp are
3489 not scheduled in the same order with the old and new version. This
3490 may be enough to completely change the execution of simulations in
3491 some cases. Sorry for the inconvenience.
3493 * Cleanup and upgrade the XML format to push further scalability
3494 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3496 * Improve the testing infrastructure with tesh. Now a very large part of
3497 the code is tested not only by being run but also by checking that the
3498 output match an expected output [Mt].
3500 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3501 is really less memory-demanding, which should allow you to use
3502 larger files in SimGrid [AL].
3504 * Inform valgrind about our contextes, so that it becomes usable
3505 with the default (and more effecient) version of SimGrid
3506 [contributed by Sékou Diakite, many thanks]
3509 * Introduce a listener thread in charge of receiving incoming
3510 messages from the network. It allows to overlap communication and
3511 computation but most notably, it removes some stupid deadlocks due
3512 to the fact that so far, a process could not send and receive at
3513 the same time. This made most non trivial communication schema
3515 * Convert the PIDs from long int to int to match the MSG ones (and
3516 linux ones too) [Mt]
3517 * New function: gras_agent_spawn() to launch a new process on
3518 current host. Only working in simulation for now. [Mt]
3519 * New function: gras_os_hostport() returning a constant form (ie,
3520 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3523 * Make the backtrace of exceptions more human readable [Mt]
3524 * New module: xbt/str [Mt]
3525 a ton of string utility functions (split, join, printf to a newly
3526 allocated buffer, trim, etc)
3527 * New module: xbt/hash [Mt]
3528 SHA1 hashing algorithm (more to come if needed)
3529 * New module: xbt/synchro [Mt]
3530 synchronization tools (mutex and conditions) working the same way
3531 in simulation and in real life (mainly useful for GRAS, but not
3533 * New module: xbt/queue [Mt]
3534 classical producer/consumer synchronization scheme
3535 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3536 (using the classical functions will get serialized) [Mt]
3537 * Make dictionary internal table dynamic. No need to specify its size
3538 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3540 * Make sure the log channels are organized as a tree under windows
3541 (because of ANSI C compatibility issue, any channel were child of
3542 root directly) [Mt].
3545 * Cleaned many thing in surf and fixed a few bugs [AL].
3546 * Add a nice command line configuration mechanism to compose models [AL].
3547 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3548 the previous one (KCCFLN05). It relies on something that looks like
3549 a max-min sharing mechanism but cannot be written as such. A new solver
3550 was thus designed [AL].
3551 * Add a new solver to lmm. Based on Lagrange optimization and
3552 gradient-based descent, it enables to efficiently maximise systems s.a
3554 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3556 This solver enables to propose two new network models for TCP Reno and
3557 TCP Vegas based on Low's work. These models still need to be fully
3558 tested though [Pedro Velho].
3561 * Bug fix in SD_simulate. Now the time bound given as argument is
3563 * Use the new parallel task model (ptask_L07) as default.
3564 * Use the SURF command line configuration mechanism.
3565 * 0-size tasks (for synchronization) should now work.
3567 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3569 SimGrid (3.2) stable; urgency=high
3573 We still experience issues on this platform, but we believe that at
3574 least MSG is usable.
3576 GRAS API BREAKAGE (for simplification purpose, sorry):
3577 * the gras_msgtype_by_name is not used anymore. Instead of
3578 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3579 you can write (and must)
3580 gras_msg_send(toserver, "request", &request);
3581 - If you still want to pass a gras_msgtype_t to the function (to cache
3582 the type and avoid the lookup time), use the gras_msg_send_() variant.
3583 - Impacted functions:
3584 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3585 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3586 * The callbacks are now expected to return 0 when everything went well
3587 (just like the main() function)
3589 GRAS new features and improvements:
3590 * New module mechanism where user code can use per process globals [Mt]
3591 This is similar to gras_userdata_*() functions, but for libraries. It
3592 factorize some code developped over and over in the examples and AMOK.
3593 It has still to be documented and used (only amok/peermanagement is
3595 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3596 loss when they were small enough to fit into the buffer and sent quickly
3597 enough so that they can all get received in one shoot.
3598 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3599 if not found. Use the *_or_null() variant for the old semantic.
3600 * In gras_msg_handle, do not discard messages without callback.
3601 They are probably messages to be explicitly awaited later (ie, proofs of
3602 mis-synchronization in userland since they are sent before being awaited)
3603 No big deal usually.
3604 * gras_socket_meas_send/recv: semantic changed!
3605 The numerical arguments used to be (1) the total amount of data to send
3606 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3607 messages. This was need for the fool willing to send more than MAXINT
3608 bytes on quite fat pipes.
3611 * Do really rename the hostmanagement module to peermanagement. [Mt]
3612 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3613 but this is rather new and this was documented in the module
3614 documentation (poor excuses, I admit)
3615 * Bandwidth measurement semantic changed! This follows the changes to
3616 gras_socket_meas_send/recv explained above.
3619 * A sequential mode has been added to the workstations. When a workstation
3620 is in sequential mode, it can execute only one task, and the other tasks
3621 are waiting in a FIFO. [Christophe Thiery]
3624 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3625 model for SIMDAG. [Christophe Thiery]
3626 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3627 equal to 0, which caused some bad behaviors in
3628 saturated_constraint_set_update. I now use a threshold mechanism like in
3632 * When running manually src/testall, you select specific units [Mt]
3633 testall is the result of our cunit mechanism, and should replace all
3634 the scripty thingy around since bash don't run easily on billware.
3636 * A mallocator system has been added. [Christophe Thiery]
3637 Mallocators allow you to recycle your unused objects instead of freeing them
3638 and allocating new ones.
3640 Documentation update:
3641 * FAQ reworking + New FAQs:
3642 - "Valgrind spits tons of errors!" [Mt]
3643 - "How to repport bugs" [Mt]
3644 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3645 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3646 - Communication time measurement within MSG [AL]
3647 - I experience weird communication times when I change the latency [AL]
3648 * GRAS tutorial [Mt]
3650 - an introduction to the framework and to the used communication model
3651 - an initiatic tour introducing the most proheminent features:
3653 . Lesson 0: Installing GRAS
3654 . Lesson 1: Setting up your own project
3655 o Part 2: Message passing
3656 . Lesson 2: Exchanging simple messages
3657 . Lesson 3: Passing arguments to the processes (in SG)
3658 . Lesson 4: Attaching callbacks to messages
3659 . Lesson 5: Using globals in processes
3660 . Lesson 6: Logging information properly
3661 . Lesson 7: Using internal timers
3662 . Lesson 8: Handling errors through exceptions
3663 . Lesson 9: Exchanging simple data
3664 . Lesson 10: Remote Procedure Calling (RPC)
3665 . Lesson 11: Explicitly waiting for messages
3666 . Recapping of message passing features in GRAS
3667 - A HOWTO section containing:
3668 o HOWTO design a GRAS application
3669 More are due, of course. They will come latter. In the meanwhile, you can
3670 check the examples which are still here.
3672 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3674 SimGrid (3.1) stable; urgency=high
3678 There was a stack corruption somewhere, visible only when optimizing
3679 with these versions. [Vince]
3682 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3683 module that enabled to play with Directed Acyclic Graphs. It is built
3684 directly on top of SURF and provides an API rather close to the old
3685 SG. Some old codes using SG are currently under rewrite to check that
3686 all needful functions are provided. [Christophe Thiery]
3689 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3690 extension of the classical CLM03 model that gracefully handles
3691 failures. This is now the default model for MSG and GRAS. It doesn't
3692 handle parallel tasks yet though. [AL]
3693 * Bug fix: Weights were not correctly set in the network part.
3694 WARNING: This may have resulted in incorrect results with simulations
3695 where there are more than one flow on a given link. [AL]
3698 * After a (long ?) discussion on simgrid-devel, we have decided that the
3699 convention we had on units was stupid. That is why it has been decided
3700 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3701 WARNING: This means that all previous platform files will not work as
3702 such with this version! A warning is issued to ask users to update
3704 A conversion script can be found in the contrib module of the CVS, under
3705 the name contrib/platform_generation/surfxml_update.pl [MQ]
3708 * Bug fix: Processes were started in reverse order, wrt deployment file.
3709 WARNING: if your code relies on this bug, please fix it. [AL]
3710 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3711 being executed on two different locations. [AL]
3712 * Bug fix: Failures are now better supported thanks to Derrick's tests
3713 (there was many failure situations I hadn't thought of and that weren't
3714 correctly handled). [AL]
3715 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3719 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3720 Exception occurring on server-side are propagated back to client (!).
3722 API CHANGE: the callback changed their prototype. Change:
3723 int my_handler(gras_socket_t expeditor, void *payload_data) {
3725 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3726 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3728 * New! function: gras_msg_handleall to deal with all messages arriving
3729 within a given period.
3730 * New! function: gras_socket_server_range to get a server socket in a
3731 range of port numbers (ease to avoid port number conflicts) [MQ]
3732 * New! gras processes display their backtrace when they get a SIGUSR1
3733 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3734 Sweet to debug RL processes [MQ]
3738 - Do not force experiment sizes to be expressed in kb, or it becomes
3739 impossible to measure the latency this way (needs one byte-long tests)
3740 WARNING: this changes the amok_bw_* function semantic. [MQ]
3741 - Implements the link saturation stuff. [MQ]
3742 * Peer management module:
3743 New! module factorizing code that we wrote over and over [MQ].
3746 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3747 - Test units are placed directly into the library code, they get extracted
3748 automatically and placed into the src/testall binary.
3749 - Convert most of the XBT tests to this system.
3750 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3751 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3752 live in the same C code. This required to change a little bit the API
3753 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3754 * New module: sparse graph structure with basic algorithms (this is work
3755 in progress and the API is not considered to be frozen yet). [AL]
3756 * Display more information on backtraces: source line & function names are
3757 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3758 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3759 * Reworked a little bit some #include statements to load only required
3760 headers. Some user code that relied on SimGrid to include stdlib or
3761 stdio may need to include it by themselves. [AL]
3762 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3764 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3765 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3766 but I guess that almost nobody used those parts. [MQ]
3768 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3770 SimGrid (3.0.1) stable; urgency=low
3773 * Unfortunately, I had missed 5 misnamed functions:
3774 xbt_fifo_item_t xbt_fifo_newitem(void);
3775 void xbt_fifo_freeitem(xbt_fifo_item_t);
3776 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3777 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3778 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3779 They're now deprecated. Please use their new versions:
3780 xbt_fifo_item_t xbt_fifo_new_item(void);
3781 void xbt_fifo_free_item(xbt_fifo_item_t);
3782 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3783 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3784 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3786 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3787 * Documentation: xbt_log module unmercifully reworked [MQ]
3788 * Bugfix: there was a problem with the ending of contexts with
3789 the pthread backend. It caused some weird deadlock or behavior
3790 depending on the pthread implementation. [AL]
3791 * Bugfix: get the exceptions raised in the simulator repport where
3792 and why they come from when they are not caught in time [AL, MQ]
3795 * Bugfix: Do repport the error when two non-connected hosts try to
3796 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3799 * Add additionnal checkings on communications. Assert that two
3800 communicating hosts are connected by a set of links... [AL]
3803 * Add additionnal checkings on channel values in communication [AL]
3804 * New: MSG_task_get_source to see on which host a task was generated [HC]
3805 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3806 the number of tasks waiting to be received on channel and sent
3808 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3809 waits for the first task coming from a given host.. [AL]
3811 GRAS new functionnalities: [MQ]
3812 * Enhance the parsing macro to allow the size of multidimentional objects
3813 to be given thru annotations.
3814 * New example (and documentation): Matrix Multiplication a la RPC
3815 (as when I was young!) and fix a bunch of bugs found on the way.
3817 GRAS performance improvements: [MQ]
3819 * Reduce the amount of cbps creation/destruction by making it static to
3820 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3823 * Change libdata to a set so that we can search for stuff by ID (and thus
3824 reduce the insane amount of dict lookups)
3827 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3828 RL communication on top of SG since it's so uneffective.
3829 It may also be used for inter-thread communication in RL, one day.
3830 * Use gras_datadesc_copy() to exchange messages on top of SG
3832 - improve message exchange performance on top of SG
3833 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3834 * Don't exchange on the network the size of the used part of buffer,
3835 instead, specify the possible buffer size to read().
3837 - reduces the amount of read/write calls (one pair per exchange)
3838 - reduces the amount of exchanged data (the size)
3839 - allows to retrieve all arrived data on receiver side, if we don't need
3840 it right now (subsequent read will peek the buffer)
3841 - allows the receiver to proceed with the beginning of the stream before
3842 everything is arrived
3843 - make it possible to build an iov transport (using readv/writev)
3845 - take care of the data with non-stable storage (like stacked data),
3847 * If possible, TCP send uses vector I/O (when writev() is here)
3848 - Don't use it for receive since we send data sizes and data on the
3849 same stream, so we wouldn't be able to chain large amount of chunks
3850 before having to flush the stuff to read the size.
3851 * Rework the transport plugin mechanism to simplify it and reduce the
3852 amount of pointer dereferencement when searching for the right function
3855 * I guess that now, we do almost as few system calls as possible while
3856 doing as few data copy as possible.
3858 To improve it further, we could try to send all the sizes first and then
3859 all the data (to use iov on receiving size), but it's only a partial
3860 solution: when you have 2 dimensional data, the sizes of the second
3861 dimension is data of the first dimension, so you need 3 streams.
3863 I'm not sure the potential performance gains justify the coding burden.
3865 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3867 SimGrid (3.00) stable; urgency=high
3870 * New! Give the possibility to hijack the surf parser and thus bypass
3871 MSG_create_environment and MSG_launch_application. Have a look at
3872 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3874 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3876 SimGrid (2.96) unstable; urgency=low
3881 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3882 This deprecates the xbt_error_t mechanisms.
3883 It modifies (simplifies) all XBT and GRAS API.
3884 MSG API keeps unchanged (exceptions raised by XBT are caught from
3885 within MSG and masked with existing error handling facilities)
3888 * New! Add a FATPIPE model. [AL]
3889 * New! Add a parallel task model. [AL]
3890 * New! Add automatically a loopback interface (in the default
3891 network model) if none was precised.
3894 * Bugfix: MSG_process_resume now works with the current running process.
3896 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3897 * Modification of MSG_task_get_compute_duration. Once a task has been
3898 processed, the value returned by this function is now equal to 0. [AL]
3899 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3900 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3901 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3902 of a m_task. For now, MSG_task_cancel only works with computation
3904 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3905 of the processor (in Mflop/s) regardless of the current load on the
3907 * API Change: use proper naming convention for MSG_getClock and
3908 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3910 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3911 This function changes the priority of a computation task. This priority
3912 doesn't affect the transfer rate. A priority of 2 will make a task
3913 receive two times more cpu power than the other ones. This function
3914 has been added to suit the needs of Nguyen The Loc and hasn't been that
3915 much tested yet. So if it fails, please report it and send me your code.
3917 * API Change: removed all functions and types that were marked "deprecated"
3918 since many months. Renamed MSG_global_init_args to MSG_global_init.
3920 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3922 SimGrid (2.95) unstable; urgency=low
3925 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3926 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3927 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3930 * Bugfix: Allow absolute paths for platform description files [MQ]
3931 * Bugfix: do free the variables after use. Leads to drastic performance
3933 * Implement max_duration (ie, timeouts) on resources [AL]
3936 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3938 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3939 the needed support (timeouts) [AL]
3942 * Implement measurement sockets. You can now get the bandwidth between two
3943 hosts thanks to AMOK (see below). [MQ]
3944 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3945 dynar over the network (yeah) [MQ]
3946 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3947 * simple/static token-ring example. [Alexandre Colucci and MQ]
3948 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3949 in place before [MQ]
3952 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3953 type parsing. "short" and "long" modifiers were also ignored; other
3954 modifier (such as reference level) are still ignored. [MQ]
3955 * Embeed the buffer size within the buffer itself on SG. [MQ]
3956 That way, send() are atomic and cannot get intermixed anymore (at least
3957 the ones which are less than 100k; bigger messages still have the issue)
3958 * Array size pushed by the field, not by the field type (or each
3959 and every long int will push stuff to the cbps) [MQ]
3960 * use select() to sleep since it allows to portably sleep less than one
3963 GRAS (minor cleanups)
3964 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3967 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3968 * type category 'ignored' killed as it was never used and were difficult
3970 * whether a type can cycle or not is now a flag, leaving room for more
3971 flags (as "ignored", if we feel the need one day ;)
3972 * Rename raw sockets to measurement sockets since "raw" has another
3973 meaning in networking community.
3976 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3977 and offers features not belonging to GRAS but that most applications
3978 need. One day, it may be a set of plugins loadable at runtime.
3979 * New module: bandwidth
3980 bandwidth measurement between arbitrary nodes running this module. [MQ]
3982 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3984 SimGrid (2.94) unstable; urgency=low
3986 The first beta release of SimGrid 3 !
3990 * Update the main page and the FAQ. Adding references to gforge.
3993 * Add a gras_os_getpid function.
3996 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3997 * Extend the logs so that they also print PID, hostname, date, ... if
3999 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
4000 and kill the old version which were in testsuite/
4001 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
4005 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
4006 for users. Instead, each programming environment (GRAS, MSG,...) use it
4007 when needed to provide such a feature to users.
4008 Don't shortcut the mechanism or you will also shortcut the virtualization
4009 you need on the simulator.
4013 * Cleanups in configury with regard to compile optimization/warning flags.
4014 Also add -fno-loop-optimize to any powerpc since it's the optimization
4015 killing gcc (< 3.4.0).
4016 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
4017 needed by MSG examples complications
4018 * Borrow configury beautifications from PHP
4021 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
4022 warning (thanks loris for stumbling into this one).
4023 * Bugfix: stop loading private headers (gras_config.h) from the public
4027 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
4028 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
4029 RL would imply a DNS resolver, which is *hard* to do in a portable way
4030 (and therefore delayed).
4031 * Implement a real timer mechanism and use it in timing macros. This allows
4032 to avoid rounding errors and get a 0.000005 sec precision in timing
4033 macros. While I was at it, various cleanups:
4034 - allow to declare more than one timed section per file (fix a stupid bug)
4035 - move some private declaration to the right place
4036 - merge conditional execution and timing macros into emulation module
4037 - document the module
4038 - make sure the module cleanups its mess on gras_exit
4039 * Documentation improvements:
4040 - (new) how to compile applications using GRAS
4041 - (new) emulation support (timing macros)
4043 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
4045 SimGrid (2.93) unstable; urgency=low
4047 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
4050 - Use Paje properly where used. Still to be sanitized properly.
4051 - Portability fix: Add an implementation of the contexts using pthread
4055 - Add xbt_procname(): returns the name of the current process.
4056 Use it to show the current process's name in all logging.
4058 - fix detection of older flex version and the reaction, since we do
4059 depend on modern ones (we use lex_destroy)
4060 - Better separation of SG and RL in the libs: remove all simulation code
4061 from libgras. As a result, this lib is now only 200k when stripped.
4062 Some of the xbt modules may also be duplicated (two sets and such) and
4063 should be cleaned/killed before SG3.
4064 - Insist on using xlC on AIX because of weird problems involving gcc there.
4065 - Cleanup the make remote stuff. This is now done by scripts
4066 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
4067 mainly for our private use, but we're working on changing them to user
4070 - Bugfix: flush the socket on close only if there is some *output*.
4071 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
4073 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
4076 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
4078 SimGrid (2.92) unstable; urgency=low
4080 Alpha 3 on the path to SimGrid 3
4084 - New! First try of benchmarking macros.
4085 - New! First try so that gras_stub_generator generate deployment and
4086 remote compilation helpers.
4088 - Bugfix: Initialization fix in msg_test.
4092 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
4094 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
4096 - API Change: gras_os_sleep to take the amount of seconds as a double.
4097 Accepting an int was error prone since it was the only location where
4098 seconds were coded as such. It leaded to damn rounding errors.
4099 - Bugfix: Hard to belive that timers ever worked before this.
4101 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4103 SimGrid (2.91) unstable; urgency=low
4105 Alpha 2 on the path to SimGrid 3
4109 - Bug fix in the lmm_solver.
4111 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4112 through the function MSG_paje_output.
4113 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4114 - It is possible to bound the rate of a communication in MSG with
4115 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4116 to write it in the changelog).
4117 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4122 - Various cleanups to the autotools stuff
4123 - Begin to move Gras examples to examples/gras/
4124 - Let make distcheck work again (yeah!)
4126 - documentation overhauled using doxygen.
4127 gtk-doc-tools is dead in SimGrid now.
4128 - Automatically extract all existing logging categories, and add the list
4129 to the documentation (long standing one, to say the less)
4131 - Cleanup the known architecture table. Reorder the entries to group what
4132 should be, and use a more consistent naming scheme.
4133 (some of the test dataset are still to be regenerated)
4134 - New! Allow library to register globals on each process just as userdata
4136 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4137 lookup time (for now).
4138 Use it in msg and trp.
4139 This cleans a lot the internals and helps enforcing privacy of the
4140 headers between the gras components.
4141 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4142 - Bugfix: gras_os_time was delirious in RL.
4143 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4145 - Reenable GRAS now that it works.
4147 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4149 SimGrid (2.90) unstable; urgency=low
4151 Alpha 1 on the path to SimGrid 3
4153 * It is a long time since the last release of SimGrid. I'm sorry about
4154 that but as I had told you, I was rewriting a lot of things. I apologize
4155 to those who had been reporting bugs to me and that I had not answered.
4156 If your bug is still in the new version, please tell me. Here is a
4157 summary of the main changes.
4159 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4160 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4161 portability, speed, and a lot more but you'll figure it out later.
4162 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4163 I wanted to release the new MSG as soon as possible and I have
4164 broken GRAS, which is the reason why, for now, only MSG is fully
4165 functional. A laconic description of these projects is available
4166 in the documentation.
4168 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4169 kernel optimized for our needs (called SURF but only the developers
4170 should use it). Hence, MSG is now roughly 30 times faster and I think
4171 that by rewriting a little bit MSG, I could event speed it up a little
4172 bit more. Beside the gain in speed, it is also much easier to encode a
4173 new platform model with SURF than it was with SG. More to come...
4175 * REVOLUTION 3: I have tried to change a little as possible the API of
4176 MSG but a few things really had to disappear. The main differences
4177 with the previous version are:
4178 1) no more m_links_t and the corresponding functions. Platforms are
4179 directly read from an XML description and cannot be hard-coded
4180 anymore. The same format is used for application deployment
4181 description. The new format is described in the documentation.
4182 Have a look in tools/platform_generation. There is a tiny script
4183 that converts from the old platform format to the new one. Concerning
4184 the application deployment format, parsing the old one is tricky.
4185 I think most of you should however be able to convert your files. If
4186 it is really an issue, I can write a C code that does the conversion.
4188 2) the toolbox tbx does not exist anymore. We now have a library
4189 with much more data-structures but without the hash-tables (we have
4190 dictionaries that are much faster).
4192 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4194 *****************************************************************************
4195 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4196 * don't feel like converting the oldies. (Mt) *
4197 *****************************************************************************
4200 Version 2.90: "the long awaited one"
4201 - Finished rewriting and debugging MSG. Rewrote the documentation.
4202 - disable GRAS for now since it needs to be ported to the newest SG
4205 - Finish the port to windows (using mingw32 for cross-compile)
4208 - Main loop and datastructures of SURF. A cpu resource object is
4209 functional. Surf can thus be used to create cpu's with variable
4210 performance on which you can execute some actions.
4212 2004-11-15 Martin Quinson
4213 - Port to ARM. Simply added the alignment and size descriptions. Should
4214 work, but the ARM machines are so slow that I didn't had the opportunity
4215 to 'make check' over there yet.
4217 2004-11-15 Arnaud Legrand
4218 - Trace manager now written. It uses a heap structure and is therefore
4219 expected to be efficient. It may however be speeded up (particularly
4220 when many events occur at the same date) by using red and black
4221 trees. One day maybe...
4222 - Max-min linear system solver written. It uses a sparse matrix
4223 structure taking advantage of its expected use. Most operations are
4224 O(1) and free/calloc are called as few as possible. The computation of
4225 the minimum could however be improved by using a red and black tree
4228 2004-11-03 Arnaud Legrand
4229 - Rename every gras_* function that was in xbt/ to its xbt_
4231 - Add a heap and a doubly-linked list to xbt
4232 - Added a dichotomy to the dictionaries. make check works as well before
4233 so I assume that the patch is correct. I do not know however if things
4234 run effectively faster than before now. :)
4236 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4237 SimGrid, and the version number is bumped to 2.x
4239 2004-10-29 Martin Quinson
4240 - Introduction of the remote errors.
4241 They are the result of a RMI/RPC on the remote machine.
4242 ErrCodes being scalar values, you can't get the host on which those
4243 errors did happen. Extending the error mechanism as in Gnome is possible.
4244 No idea yet whether it is a good idea.
4246 2004-10-28 Martin Quinson
4247 - Interface revolution: the Starred Structure Eradication.
4248 I used to do typedef struct {} toto_t; and then handle *toto_t.
4249 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4250 - ???_t is a valid type (builded with typedef)
4251 - s_toto_t is a structure (access to fields with .)
4252 - s_toto is a structure needing 'struct' keyword to be used
4253 - e_toto_t is an enum
4254 - toto_t is an 'object' (struct*)
4256 typedef struct s_toto {} s_toto_t, *toto_t;
4257 typedef enum {} e_toto_t;
4258 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4259 s_toto_t) is private.
4261 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4262 so that it can set the variable to NULL. It was so for dicts and sets,
4263 it changed for dynars.
4265 - Fix a bunch of memleaks in dict_remove
4266 - Fix a bug in sg/server_socket opening: it failed all the time.
4268 2004-10-07 Martin Quinson
4269 - Speed up dynar lookup operation a bit.
4271 gras_dynar_get is dead.
4273 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4274 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4275 (which gives you the address of the stored data).
4277 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4278 copy of the data using an affectation to do the job and not a memcpy.
4280 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4281 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4283 It does not really speedup the dynar test because they are
4284 setting elements all the time (and look them seldom). But the dict does
4285 far more lookup than setting.
4287 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4289 2004-10-05 Martin Quinson
4290 - Allow to (en/dis)able the cycle detection at run time.
4292 Whether we should check for cycle or not is now a property of each
4293 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4294 datadesc_cycle_unset allow to remove this property when previously set.
4296 Note that the cycle detection is off by default since it impacts the
4297 performance. Watch the data you feed GRAS with ;)
4299 This property is hereditary. Any element embedded in a structure having it
4300 set have it set for the time of this data exchange.
4302 You should set it both on sender and receiver side. If you don't set it on
4303 sender side, it will enter an endless loop. If you forget on receiver
4304 side, the cycles won't be recreated after communication.
4306 - Header reorganization.
4307 Kill gras_private.h, each submodule must load the headers it needs.
4309 2004-10-04 Martin Quinson
4310 - Interface revolution: do not try to survive to malloc failure.
4312 Now, gras_malloc and friends call gras_abort() on failure.
4313 As a conclusion, malloc_error is not a valid error anymore, and all
4314 functions for which it was the only gras_error_t return value are
4315 changed. They now return void, or there result directly.
4316 This simplify the API a lot.
4318 2004-09-29 Martin Quinson
4319 - Re-enable raw sockets.
4320 Created by gras_socket_{client,server}_ext;
4321 Used with gras_raw_{send,recv}
4324 It should allow to kill the last bits of gras first version soon.
4326 This is not completely satisfactory yet (duplicate code with
4327 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4330 - Simplify transport plugin (internal) interface by not passing any
4331 argument to _server and _client, but embedding them in the socket
4334 2004-09-28 Martin Quinson
4335 - Finish the port to AIX.
4336 autoconf was my problem (segfault within the malloc replacement
4337 function. No idea why)
4339 2004-09-16 Martin Quinson
4340 - Fix some size_t madness on 64bit architectures.
4342 2004-09-08 Martin Quinson
4343 - Reduce the number of system headers loaded, overload some more system
4344 calls (such as malloc to cast the result of the system one, and work
4346 - Fix and reintroduce the config support
4348 2004-09-07 Martin Quinson
4349 - Source code reorganization to allow Arnaud to surf all over there.
4350 - Allow to document the logging categories.
4351 - Remove all uppercase from logging categories and useless cleanup in names.
4353 2004-08-18 Martin Quinson
4354 Version 0.6.2 (protocol not changed; API changed)
4355 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4356 gras_error_t), and NULL when not found. Functions expecting a msgtype
4357 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4358 hopefully usefull message.
4359 - Portability to prehistoric sparcs again
4361 2004-08-17 Martin Quinson
4362 Version 0.6.1 (protocol not changed; ABI not changed)
4363 - prealloc some buffers to speed things up
4365 2004-08-11 Martin Quinson
4366 Version 0.6 (protocol not changed; ABI expended)
4367 - The parsing macro can deal with the references, provided that you add
4368 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4370 2004-08-09 Martin Quinson
4371 Version 0.5 (protocol not changed; ABI changed)
4372 - Allow to off turn the cycle detection code in data exchange at
4373 compilation time. It should be at run time, but I'm short of time (and
4374 the config stuff is still broken). That way, we keep dict out of the
4375 critical path, which is good because the performance is poor:
4376 - search not dichotomial yet
4377 - dynar give no way to access their content and memcpy everytime
4378 - In composed data description (struct, ref and so on), stop foolness of
4379 keeping the subtype's ID, but store the type itself. This keeps sets out
4380 of the critical path, which is good since they rely on dynar and
4381 dictionaries. The only loose of that is that we cannot detect the
4382 redeclaration of a structure/union with another content (but I'm not sure
4383 the code detected well this error before anyway). We still can detect
4384 the redefinition discrepancy for the other types.
4385 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4386 it breaks the code because of type-punning used all over the place).
4387 This breaks on all non-gcc architectures (for now).
4389 All those changes (plus the buffer of last time) allow me to gain 2 order
4390 of magnitude on cruel tests consisting of 800000 array of integers on two
4391 level of a hierarchical structure (200 secondes -> 4 secondes)
4394 - the selector of reference must now return the type it points to, not
4395 the ID of this type.
4397 2004-08-06 Martin Quinson
4398 Version 0.4 (protocol changed; ABI not changed)
4399 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4400 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4401 - Implement a buffer transport, and use it by default (it relies on tcp in
4402 real life and on sg in simulation).
4403 That's a bit hackish since I had a new field to the structure to store
4404 its data without interfering with the subtype ones. Inheritance
4405 is tricky in C. And that's a kind of reverse inheritance with one class
4406 derivating two classes. Or maybe a game with java interfaces. Anyway,
4407 that's damn hard in C (at least).
4408 Moreover, I got tired while trying to ensure plugin separation and
4409 genericity in SG mode. MSG wants me to do weird things, so let's go for
4410 cruel hacks (temporarily of course ;).
4411 See comment in transport_private.h:71
4412 - do not include all the _interface headers in private but in the files
4413 which really need them (to cut the compilation time when they are
4416 2004-07-26 Martin Quinson
4417 Version 0.3 (protocol not changed; ABI changed)
4418 - Major overhault of the datadesc interface to simplify it:
4419 - shorted the function names:
4420 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4421 - add a trivial way to push/pop integers into the cbps without malloc.
4422 This allows to make really generic sub_type description, which simply
4423 pop their size of the stack.
4424 - add a function gras_datadesc_ref_pop_arr() which does what users want
4425 most of the time: Declare a dynamic array (which pops its size of the
4426 stack) and declare a reference to it. Poor name, but anyway.
4427 - kill the post-send callback, add a post-receive one
4429 2004-07-23 Martin Quinson
4430 Version 0.2 (protocol changed; ABI changed)
4431 - add some testing for cpbs in the test cases, and fix some more bugs.
4432 This invalidate again the little64 data file, since I cannot regenerate
4434 - remove an awfull optimization in the logging stuff, allowing me to:
4435 - understand it again
4436 - learn gcc how to check that the argument match the provided format
4437 - fix all errors revealed by gcc after that
4438 - internal keys of dict are not \0 terminated. Deal with it properly in
4439 loggings instead of segfaulting when the user want to see the logs :-/
4441 2004-07-22 Martin Quinson
4442 - Fix some stupid bug preventing cbps (callback postit) from working
4444 2004-07-21 Martin Quinson
4445 - Some documentation cleanups
4446 - remove the useless last argument of msgtype_declare
4447 - rename the Virtu functions to fit into the 'os' namespace
4448 - move headers src/include -> src/include/gras/ and stop fooling with
4449 gras -> . symbolic link
4450 - make distcheck is now successful
4452 2004-07-19 Martin Quinson
4454 - Build shared library also
4455 - Install html doc to the right location
4456 - stop removing maintainer files in make clean
4457 - build tests only on make check
4459 2004-07-13 Martin Quinson
4461 - No major issue in previous version => change versioning schema
4462 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4463 corresponding dataset.
4465 2004-07-11 Martin Quinson
4467 - Get it working with any kind of structure (we can compute the padding
4468 bytes remotely for all the architectures I have access to)
4469 - Implement the structure parsing macro (still not quite robust/complete)
4470 - Improvement to the remote testing toysuite
4472 2004-07-10 Martin Quinson
4473 [autoconf mechanism]
4474 - get ride of a bunch of deprecated macros
4475 - actually run the test for two-compliment, not only compile it :-/
4476 - test whether the structures get packed (and bail out if yes. Damn.
4477 Alignment is a serious matter)
4478 - test whether the structures get compacted (but respecting the alignment
4479 constraints of each types)
4480 - test whether the array fields of structures can straddle alignment boundaries
4482 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4483 (took me 2 hours to find that bug, looking at the wrong place)
4484 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4485 before sending/receiving and are used to compute the offsets of fields
4486 - Given that padding size depend even on compiler options, keep track of
4487 alignment and aligned_size only for the current architecture. Not a big
4488 deal since we send structure fields one after the other (seems
4490 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4491 program, let it work on linux/gcc/little32. portability todo.
4493 2004-07-08 Martin Quinson
4494 - import and improve remote compilation support from FAST
4495 - make sure make check works on half a dozen of machines out there
4497 2004-07-07 Martin Quinson
4498 Let's say it's version 0.0.3 ;)
4499 - Implement conversions (yuhu!)
4500 - Let it work on solaris (beside conversion, of course)
4501 - Stupid me, using rand() to generate the conversion datatests in not wise.
4503 2004-07-06 Martin Quinson
4504 - Let make dist work, since I'm gonna need it to compile on remote hosts
4505 - Let Tests/datadesc_usage write the architecture on which the file was
4506 generated as first byte.
4507 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4508 architecture descriptions.
4509 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4510 on those architectures.
4511 - Optimization: send/recv array of scalar in one shoot
4513 2004-07-05 Martin Quinson
4514 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4516 - Plug a whole bunch of memleaks
4517 - each process now have to call gras_{init,exit}. One day, their log
4518 settings will be separated
4519 - Continue the code factorisation between SG, RL and common in Transport.
4521 2004-07-04 Martin Quinson
4523 - Redistribution between SG and RL.
4524 We wanna have to accept in SG, so move accepted related parts of RL in
4525 the common part. (more precisely, the dynar of all known sockets is no
4526 more a static in transport.c, but part of the process_data)
4528 [gras_stub_generator]
4529 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4531 2004-07-03 Martin Quinson
4532 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4533 them all up in one shot)
4535 - Fix the ugly case of reference to dynamic array.
4536 - New (semi-public) function gras_datadesc_size to allow the messaging
4537 layer to malloc the needed space for the buffer.
4539 - gras_socket_close now expect the socket to close (and not its address to
4540 put NULL in it after it). This is because the socket passed to handlers
4541 is one of their argument (=> not writable).
4543 - propagate the interface cleanup from last week in datadesc, ie remove a
4544 superfluous level of indirection. User pass adress of variable
4545 containing data (both when sending and receiving), and not of a variable
4546 being a pointer to the data. Let's say that I like it better ;)
4547 The price for that is constructs like "int msg=*(int*)payload" in
4548 handlers, but it's a fine price, IMHO.
4550 - Let it work in RL (yuhu)
4552 2004-06-21 Martin Quinson
4554 - porting SG plugin and SG select to new standards (works almost).
4555 - plug memleaks and fix bugs around.
4558 - cleanup the prototype of data recv and force users to specify when they
4559 want to handle references to objects. Test case working even for cycles.
4560 - plug memleaks. Valgrind is perfectly ok with this.
4562 2004-06-12 Martin Quinson
4564 - cleanup the separation between plugin and main code in plugin creation
4566 2004-06-11 Martin Quinson
4568 - Reput hook for raw sockets, needed for BW experiments
4569 - kill a few lines of dead code
4570 [Data description] Interface cleanup
4571 - gras_datadesc_by_name returns the searched type or NULL.
4572 That way, no variable is needed to use a type desc once, which makes
4574 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4575 two parameters were strings (field name, type name), leading to
4577 [Dicos] Interface cleanup
4578 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4579 This is consistant with the dynar API.
4581 2004-04-21 Martin Quinson
4583 - Porting to new standards.
4585 - interface cleanup.
4586 There is no bag anymore, no need to take extra provision to mask the
4587 pointers behind "ID".
4588 Better splitup of functions between files create/exchange/convert.
4589 This is still a bit artificial since convert and receive are so
4590 interleaved, but anyway.
4592 - add a queued message list to procdata (the ones not matching criteria
4594 - factorize some more code between SG and RL wrt procdata
4596 - use gras_exit in example to track memleaks
4597 - get rid of gs_example now that GS is properly integrated into gras
4598 - update run_test to integrate the latest tests (datadesc)
4600 - rename WARNINGn macros to WARNn since it prooved error-prone
4602 2004-04-19 Martin Quinson
4604 - register init/exit functions within gras module mechanism
4605 - send/receive function.
4606 Convertion is not implemented, but short-cutted if not needed.
4607 struct/array elements are sent one by one (instead of block-wise), but
4608 nobody really cares (yet). Get a prototype before optimizing.
4609 - tests (using a file socket) for DD send/receive on:
4610 - base types: int, float
4611 - array: fixed size, string (ie ref to dynamic string)
4612 - structure: homogeneous, heterogeneous
4613 - chained list, graph with cycle
4614 Believe it or not, valgrind is not too unhappy with the results. The
4615 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4619 - Bugfix when using a filename explicitly (instead of '-')
4621 2004-04-09 Martin Quinson
4623 - factorize more code between RL and SG in socket creation
4624 - Complete the implementation and tests of:
4626 o file (only in RL, and mainly for debugging)
4628 I lost 3 days to design a portable address resolver, and then decided
4629 that the prototype mainly have to run on my box.
4630 Addressing portability too early may be like optimizing too early :-/
4632 - use gras_init in the Tests instead of the crappy parse_log_opt
4633 (the latter function is removed)
4634 [Conditional execution]
4635 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4636 [Code reorganisation]
4637 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4638 Build examples against the RL library, since there is no way to disable
4639 its creation for now.
4641 For information, the beginning of coding on GRAS was back in june
4642 2003. I guess that every line has been rewritten at least twice since