1 ----------------------------------------------------------------------------
3 SimGrid (3.29.1) NOT RELEASED YET (v3.30 expected December 21. 2021, 15:59 UTC)
5 Main user-visible changes:
6 - The SimDag API for the simulation of the scheduling of Directed Acyclic
7 Graphs has been dropped. It was marked as deprecated for a couple of years.
8 We finally complete the implementation of what has been called SimDag++
9 internally, i.e., porting the different features of SimDag on top of S4U.
10 The new way to simulate the execution of dependent activities directly by
11 maestro (without any other actor) is details in the examples/cpp/dag-* series
13 - The removal of SimDag led us to also remove the export to Jedule files that
14 was tightly coupled to SimDag. The instrumentation of DAG simulation is still
15 possible through the regular instrumentation API based on the Paje format.
16 - We also dropped the old and clumsy Lua bindings to create platforms in a
17 programmatic way. It can be done in C++ in a much cleaner way now, which
18 motivates this suppression.
21 - Introduce on_X_cb() functions for all signals, to attach a new
22 callback to the signal X. The signal variables are now hidden and
23 only these functions should be used.
24 Rationale: this enables the usual deprecation schema where functions
25 remain for 4 releases if we need to modify the signals, while the
26 current code with the signal variables directly visible prevents any
28 - New function: Engine::run_until(date), to split the simulation.
29 - New signal: Activity::on_veto, to detect when an activity fails to start.
30 - Signal change: Comm::on_start(Comm&, bool) has been replaced by
31 Comm::on_send and Comm::recv. These two signals respectively correspond to
32 when the sending or receiving side of a Comm is ready. They are raised at
33 the same locations as the former Comm::on_start signal.
34 - New function: Engine::track_vetoed_activities() to interrupt run()
35 when an activity fails to start, and to keep track of such activities.
36 Please see the corresponding example for more info.
37 - New functions: s4u::Comm::{sendto_init, set_source, set_destination} to enable
38 the use of vetoers with direct host-to-host communications. Both source and
39 destination have to set for a comm to start. Each call to these setters check
40 if all vetos are satistied. When it is the case, the comm starts. A use case of
41 these functions is given in examples/cpp/dag-scheduling.
42 - New functions: {Exec, Io}::update_priority allow you to modify the priority of
43 these kinds of activities during their execution. Behavior is detailed in
44 examples/cpp/io-priority/
46 - Dynamic costs for MPI operations: New API to allow users to dynamically
47 change injected costs for MPI_Recv, MPI_Send and MPI_Isend operations.
48 Alternative for smpi/or, smpi/os and smpi/ois configuration options.
51 - Function xbt::Extendable::get_data() is now templated with the type of the
52 pointee. Untyped function is deprecated. Use get_data<void>() if you still
53 want to retrieve void*.
56 - New section: "SimGrid MPI calibration of a Grid5000 cluster"
57 presenting how to properly calibrate MPI communications in SimGrid.
58 - Complete and reword the platform section, which is now completed.
61 - Thread contexts are used by default with Python bindings. Other kinds of
62 contexts revealed unstable, specially starting with pybind11 v2.8.0.
64 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
65 (FG: issues on Framagit; GH: issues on GitHub)
66 - FG#95: Wrong computation time for multicore execution after pstate change
67 - FG#97: Wrong computation time for ptask+multicore+pstates
68 - FG#98: SMPI offline simulation is inconsistent with the online simulation
69 (deadlocks / message truncation)
70 - FG#99: Weird segfault when not sealing an host
72 ----------------------------------------------------------------------------
74 SimGrid (3.29) October 7. 2021
76 The "Ask a stupid question" release.
78 We wish that every user ask one question about SimGrid to celebrate.
79 On Mattermost, Stack Overflow or using the issues tracker.
82 New modeling features:
83 - Non-linear resource sharing, modeling resources whose performance heavily degrades with contention:
84 - The total capacity may be updated dynamically through a callback
85 and depends mainly on the number of concurrent flows.
86 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
88 - Dynamic factors: model variability in the speed of activities
89 - Each action can now have a factor that affects its progression.
90 This multiplicative factor is applied when updating the amount of work
91 remaining, thereby an activity with factor=0.5 only uses half of the
92 instantaneous power/bandwidth it is allocated and will appear twice
93 slower than what it actually consumes.
94 - This can be used to model a overhead (e.g., there is a 20 bytes
95 header in a 480 bytes TCP packet so the factor 0.9583) but the novelty
96 is this factor can now easily be adjusted depending on activity's and
97 resources characteristics.
98 - This existed for network (e.g., the effective bandwidth depends
99 on the message in SMPI piecewise-linear network model) but it is now
100 more general (the factor may depend on the source and destination and
101 thus account to different behaviors for intra-node communications and
102 extra-node communications) and is available for CPUs (e.g., if you
103 want to model an affinity as in the "Unrelated Machines" problem in
104 scheduling) and disks (e.g., if you want to model a stochastic
106 - For that, resources can be provided with a callback that computes
107 the activity factor when creating the action.
108 - Example: examples/cpp/exec-cpu-factors
109 - The same mechanism is also available for the latency, which
110 allows to easily introduce complex variability patterns.
113 - Added support to programmatic platform creation in Python.
114 Example: examples/python/clusters-multicpu
117 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
118 This can only be set through the API, not through XML files.
122 - Multiple fixes to ensure reproducibility of tracing
123 - scan/excan can now be replayed
124 - wait action now uses ranks and not pid, as the other ones.
125 - smpi/init and smpi/finalization-barrier are now valid for replays.
126 - exit() is now intercepted by SMPI to avoid premature shutdown of
127 simulation. First non 0 return codes is returned as simulation return
131 * New section "Release Notes" documenting recent and current developments.
132 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
133 * Improvements in API Reference for C++ and Python interfaces.
136 - Make wifi creation compatible with ns-3 version 3.34 too.
138 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
139 (FG: issues on Framagit; GH: issues on GitHub)
140 - FG#77: Search feature of doc is broken (update sphinx theme version)
141 - FG#78: Multiple fixes for SMPI replay:
142 - TI tracing of allotallv/w was outputting wrong values
143 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
144 ----------------------------------------------------------------------------
146 SimGrid (3.28) July 14. 2021
148 The Victoriadagarna Release.
151 - C++ platform interface: Users can now describe their platform directly in C++.
152 This provides greatly flexibility and performance improvement for complex
153 platforms. Main features:
154 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
155 "zones", instead of single hosts. This feature enables the description
156 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
157 - StarZone: new zone with a Star-like topology. The routes are defined
158 as a set of links used to communicate from node to everybody (node<->ALL).
159 - Split-Duplex links: auxiliary method to create split-duplex links in
160 the platform, easing its utilisation. It automatically creates both UP
161 and DOWN links (similarly as done in XML).
162 - Please refer to the documentation and the examples included:
163 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
164 - New plugin: Producer-Consumer with monitor. Just requires to include the
165 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
166 associated example (examples/cpp/plugin-prodcons).
169 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
170 s4u::Io::wait_any(), s4u::Io::wait_any_for().
171 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
172 parameter by reference, instead of a pointer.
173 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
174 Explicitly cancel the activity to get back to previous behavior.
175 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
178 - The default SMPI compiler flags are no more taken from the environment.
179 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
180 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
182 --cfg=smpi/finalization-barrier: which can be used to add
183 a barrier inside MPI_Finalize. This can help for some codes which cleanup
184 data attached to a process, but still used in other SMPI processes.
185 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
186 is active when set to false, to keep going after a small error
187 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
188 which may or may not be problematic in the end.
190 - fix behaviour, as maximum iteration count could be ignored
191 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
192 use sampling when the same kernel is called with a different set of parameters
193 which have an impact on the timing.
194 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
195 It should now work with smpi/auto-shared-malloc-thresh.
196 - Improve error handling and reporting in multiple places
197 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
198 validity checks, truncated messages are now an error, return errors
199 when explicitely deleted handles are reused, ...)
200 - RMA: multiple fixes and stability improvements.
201 - analysis (-analyze flag in smpirun):
202 - SMPI can now report buffer leaks as well as MPI handles leaks,
203 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
204 - if -trace-call-location is used when compiling, SMPI can report
205 origin of leaked handles/buffers
206 - group leaks by type/origin in output message if possible
207 - New implemented MPI calls: MPI_Comm_test_inter
210 - Changed internal implementation of bandwidth factors in network models.
211 Models affected: CM02, LV08 (default), SMPI, IB.
212 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
213 Bandwidth factors are applied to communications to describe that users
214 cannot use 100% of the available bandwidth. For example, the default network model,
215 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
216 behavior was implemented by limiting the bandwidth available in the LMM
217 system for this flow. This may give the false impression that there is
218 bandwidth available for other flows due to its underutilization, especially
219 for the dynamic bandwidth factors used in SMPI models.
220 To avoid this, we have modified the implementation so that each flow uses the
221 maximum physical bandwidth according to the LMM system.
222 However, the actual throughput of the flow seen by the user is defined by
223 the physical bandwidth multiplied by the bandwidth factor.
224 This change impacts on the simulation results for all network models on
225 which we have bandwidth factors configured.
226 *****************************************
227 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
228 *****************************************
229 This change may impact on the timing of your simulation results.
230 Take care when comparing simulations from different SimGrid's
231 versions. Sorry for the inconvenience.
232 - Dynamic network factors: users can configure a callback to define
233 the network factors dynamically. This API is available at
234 simgrid::kernel::resource::NetworkModelIntf.
235 - Users have access to complete information about the current communication
236 to decide which factor to apply. This includes: message size, source and
237 destination hosts, links and zones traversed.
238 - Dynamic factors for both latency and bandwidth.
239 - For more details, see the example in (examples/cpp/network-factors).
240 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
241 deprecated since version 3.24, are no longer supported. Instead, use
242 "wattage_off" and "wattage_per_state".
245 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
248 - New tutorial: Model-checking and formal assessment
249 - New sections: "Demystifying the routing" and "C++ platforms"
250 - Update and improve the part on visualization in MPI and Algo tutorials.
251 - Improve the section on routing: how to define it, how it's used internally
252 - Fix many issues, broken links and missing references in doxygen and Sphinx
255 - Lua platform files are deprecated. Their support will be dropped after v3.31.
258 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
260 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
261 (FG: issues on Framagit; GH: issues on GitHub)
262 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
263 - FG#64: Configuring smpi/IB-penalty-factors
264 - FG#67: Running computation concurrently with MPI_Iallreduce
265 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
266 - FG#71: Segmentation fault on invalid gw_src/gw_dst
267 - GH#322: Issue when an actor kills his host vm
269 ----------------------------------------------------------------------------
271 SimGrid (3.27) March 29. 2021
273 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
275 Main user-visible changes:
276 - All functions marked as deprecated for one year were removed.
277 - Removed all that was related to the "Storage" type in all APIs and bindings.
278 "Disks" have been introduced more than a year ago to replace them for the better.
279 It's time to finish this replacement.
280 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
281 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
282 avoiding "initializer element is not constant" errors with some codes. Please be
283 careful to recompile your MPI codes if switching to this new SimGrid version.
284 - Network model ns-3 now requires version ns-3.28 or higher.
285 - Removed some option aliases that were deprecated since v3.23 (June 2019)
286 - 'verbose-exit' is now 'debug/verbose-exit'
287 - 'simix/breakpoint' is now 'debug/breakpoint'
288 - 'clean-atexit' is now 'debug/clean-atexit'
291 - New functions: Comm::sendto_{init,async} to initiate a communication
292 on between two (possibly remote) hosts.
293 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
294 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
295 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
296 or Mailbox::get_async<void>() if you really want to play with void*.
297 - Unify the interface of Activity::on_{start/activity}
298 - New function: Comm::get_dst_data()
299 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
302 - ptask should now work better with multicore hosts (FG#37).
303 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
304 - Internal cleanups and refactoring around models, to prepare for programmatic
306 - Modernize our tests of the performance models
310 - New section: Configuring the logs from the command line.
311 - New section: Programming API of logs.
312 - New section: Getting realistic results
313 - New section: Examples of SMPI Usage
314 - New part: SimGrid's Internals, with a section on Design goals
315 - Move back to the classical Breathe tool chain, and fix many small glitches.
318 - Handle leak checking: SMPI will now display a message if MPI handles are still
319 present at the end of a simulation. Option smpi/list-leaks:n will display the n
320 first leaked pointers and their types.
321 - smpi/display-allocs option : Display Memory allocations analysis, with hints
322 about the largest allocations. This only works for allocations with malloc/calloc
323 and not compiled with SMPI_NO_OVERRIDE_MALLOC
324 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
325 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
326 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
327 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
328 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
329 - Replay now opens and parses the trace file only once for all ranks when possible.
331 - better handling of MPI_PROC_NULL
332 - MPI_Group_incl was broken in some cases
333 - PAPI tracing now fixed
334 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
335 - SMPI will now fail properly when receive size < sent size, instead of
336 just setting MPI_ERR_TRUNCATE and keep going.
337 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
338 - Remove useless build patches and split build/run patches on the page
339 - Build in parallel when possible
340 - Test on yum-based systems
341 - New nightly tested Proxy-apps:
342 - OpenMPI mpi-test-suite
343 - Coral MCB, LSMS, Qbox, CLOMP
344 - ECP/Mantevo Cloverleaf3D, Ember
345 - Codevault bhtree_mpi
348 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
349 - Start to overhaul the implementation to modernize it (still ongoing).
352 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
353 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
354 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
355 - Deprecate xbt_dynar_to_array().
356 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
358 C binding and interface:
359 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
360 - Constness requirement has been lowered for parameter argv of sg_actor_start()
361 and sg_actor_create() It's now (char* const*). The previous version is still
362 available as sg_actor_start_() and sg_actor_create_().
364 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
365 (FG: issues on Framagit; GH: issues on GitHub)
366 - FG#37: Parallel tasks are limited to 1 core per host
367 - FG#62: Running "smpirun -replay" on large networks
368 - FG!46: Fix a few potential memory leaks in SMPI colls
369 - GH#343: Wrong consumed energy when an exec timeouts
371 ----------------------------------------------------------------------------
373 SimGrid (3.26) December 15. 2020
375 The Release release (the French lockdown was eased today).
377 Important user-visible changes:
378 - SimGrid now requires a compiler with C++14 support.
379 Sibling projects should upgrade their FindSimgrid.cmake
380 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
381 several users had difficulties to understand issues when using high bandwidth or
382 small latency events. The new value was already the default for SMPI and
383 should not cause too much performance hit. It can change some simulation
384 timings if simulation was skipping a lot of small events (which was actually
385 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
386 - On some network models (L07, CM02, LV08), an implicit loopback link is created
387 for local communication. Default bandwidth was 500MBps and latency was 15µs,
388 which was rather bad. Default values have been changed to 10GBps and 0 latency.
389 Options network/loopback-lat and network/loopback-bw have been added to adapt
390 these values if needed. Network and routing models already providing separate
391 loopbacks (clusters and such) are not impacted by this change.
392 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
393 metadata that are added to each MPI message by MPI libraries (tag,
394 src/dest, and communicator information). This may change simulation times
395 for very small message sizes, but should be more realistic.
398 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
399 - New plugins: wifi_energy and link_load. See the associated examples.
400 - New examples: wifi networks, and communication suspend.
403 - LTO is now enabled for Intel/clang compilers.
404 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
405 Setting it to "auto" will use all available cores, while setting it to n will
406 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
407 - Remove obsolete runtime option 'contexts/parallel-threshold'.
408 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
409 has been kept for compatibility.
410 - Finally remove obsolete snake_case() aliases for runtime options.
411 - Further improve the documentation.
414 - the on_exit() of each actor is also executed when the simulation deadlocks.
415 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
416 An example is provided for s4u::Comm but it should work with Exec and Io.
419 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
420 https://framagit.org/simgrid/SMPI-proxy-apps
421 - MPI/IO: supports names starting with "./"
422 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
424 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
425 building shared libraries, this may cause issues, so environment variable
426 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
427 - most temporary files should now be created in /tmp dir (or equivalent).
428 If this one does not allow execution of code (noexec flag), this may cause issues.
429 Please use another tmp directory (using TMPDIR or equivalent system variable)
433 - A large refactoring is ongoing, but this should be transparent for users so far.
436 - Drop xbt_str_split().
438 C binding and interface:
439 - The return type of the sg_actor_on_exit() callbacks is now 'void'
440 (int value was previously ignored)
441 - Many C functions were renamed to follow a common naming style
442 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
443 names are kept until version 3.30.
444 - Many MSG tests were converted to the new S4U's interface in C, that
445 was extended for that.
447 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
448 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
449 - FG#41: Add sg_actor_create C interface
450 - FG#43: xbt::random needs some care
451 - FG#48: The Impossible Did Happen (yet again)
452 - FG#50: Suspending an actor executed at the current timestamp fails
453 - FG#52: Zero-seconds timeout : "That's in the past already"
454 - FG#53: Crash while using ns-3 network model
455 - FG#54: How to suspend a comm?
456 - FG!22: Stochastic Profiles
457 - FG!24: Documentation and fix for xbt/random
458 - FG!35: Add a modeling hint for parallel links in doc
459 - FG!36: [xbt/random] Read/Write the state of the RNG
460 - FG!42: Link cumulated load plugin
461 - FG!44: Test condition variable wait_until+notify_all
462 - GF#18137: Allow different stack sizes?
463 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
464 - GH#139: Allow pthread creation in SMPI
465 - GH#336: Packet-level simulation using SMPI?
466 - GH#345: Error in the LMM while migrating a VM
467 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
468 - GH#352: pip install / python setup.py install fail to find pybind11
469 - GH!337: Fix link_energy plugin for wifi platforms
470 - GH!339: Add Mailbox set_receiver method to python binding
471 - GH!344: Cast hugepages macros parameters to int64
473 ----------------------------------------------------------------------------
475 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
477 The Palindrom Day Release.
479 Important user-visible changes:
480 - Improve the Python usability (stability and documentation).
481 - A nasty synchronization bug was ironed out, see also below.
482 - Python's doc was organized as a separate tree, now integrated with C++.
483 - C bindings of S4U were not part of the doc.
484 - The C++ doc was also improved as methods are now split by theme.
485 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
486 - This is mandatory to use the Java bindings.
487 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
488 - SimDAG++: Automatic dependencies on S4U activities (experimental)
489 - Some features are already implemented but not all of them
490 - Cannot block an activity until it's scheduled on a resource
491 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
492 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
495 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
496 - Drop support for Viva/Triva (old visualization tools), hence removed
497 TRACE_get_node_types() and TRACE_get_edge_types() functions.
500 - Require pybind11 version 2.4.
501 - Greatly improve locking during context switches with Python.
502 - Declaring Python callbacks to this_actor::on_exit is now working.
503 - No more tests are failing currently, even on win32.
506 - Actor: Merge signals on_migration_start/end into on_host_change
507 - Actor: Rename migrate() into set_host()
508 - Disk: Allow users to get the read and write nominal bandwidth values
509 - Exec: Implement wait_for(timeout)
510 - Io: Implement wait_for(timeout)
513 - Parse errors now raise a simgrid::ParseError that you may want to catch.
516 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
517 been renamed to fix a spelling error. As usual, the previous definitions are
518 kept with a deprecation warning. However, to avoid runtime errors, it is now
519 forbidden to override these deprecated methods in a derived class. Please use
520 the new names immediately if you need to override them.
523 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
524 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
525 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
526 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
527 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
528 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
529 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
530 - New module: random, providing classical random numbers generators.
533 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
534 allocations' size above which they are considered "shared" by default (as if
535 they were performed through SMPI_SHARED_MALLOC macros).
536 Default = 0 = disabled feature.
537 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
538 This can cause some troubles if codes are already overriding these. If this
539 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
540 help, but will make this feature unusable.
541 - Input check is now performed much more consistently and outputs precise
542 warnings for each error found
543 - New supported calls : MPI_File_set_view, MPI_File_get_view
544 - MPI I/O now only supports disks, as storage is discontinued.
545 - MPI I/O allows opening files without specifying absolute path
546 - Note : SMPI configuration options are now initialized by smpi_init_options()
547 call, which should be called if SMPI is not being used through smpirun.
549 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
550 - FG#9: Python bindings crashing
551 - FG#39: Missing s4u::Comm::wait_any_for example
552 - FG#42: Add support for ThreadSanitizer (TSan)
553 - FG!19: Removing RngStream
554 - FG!20: A module for RNG calls
555 - FG!21: Choice between ad-hoc and standard distributions implementations
557 - FG!25: Fix link in Mutex doc
558 - FG!26: Fix links in SMPI interface doc
559 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
560 - GH#305: Unscheduled tasks are still excuted
561 - GH#313: smpirun: manual outdated w.r.t. --help
562 - GH#321: [S4U] Get task remaining work ratio
563 - GH#323: Crash when an actor turn off his physical host
564 - GH#335: Missing links on dragonfly example svg
565 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
566 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
568 ----------------------------------------------------------------------------
570 SimGrid (3.24) October 9. 2019
572 The Clean Disk Release.
575 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
576 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
577 thus becomes the default resource to execute I/O operations. Examples
578 and the file system plugin have been modified to use disks instead of
579 storage elements. Storage can still be used but is doomed to disappear.
580 Users are thus adviced to update their simulators accordingly.
581 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
582 for consistency with pthread_barrier_wait()
583 - Host::get_englobing_zone() returns the englobing netzone
584 - Actor::on_destruction is now called in the destructor
585 Actor::on_termination new signal called when the actor terminates
587 - Global signals are now part of the Engine:
588 - on_platform_creation: after config settings, before the XML parsing
589 - on_platform_created: right after the XML parsing
590 - on_time_advance: each time the clock advances
591 - on_simulation_end: after simulation, before cleanups
592 - on_deadlock: as the name implies.
594 - sg_{actor,host,link}_{data,data_set}() now all exist.
595 Use them to attach user data to the object and retrieve it.
598 - Introduce an experimental Wifi model. It sounds reasonable
599 according to the state of the art, but it still has to be properly
600 validated, at least against ns-3 if not against reality.
601 - Improved the usability of ns-3. Several bugs were ironed out.
602 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
603 It is now expressed as 'idle:epsilon:allCores' to properly model the
604 consumption of non-whole tasks on mono-core hosts. More info in the doc.
607 - convert a new set of functions to the S4U C interface and move the old MSG
608 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
611 - Fortran bindings for DVFS have been removed.
612 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
613 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
614 - SMPI can now be selected by cmake's find_module(MPI) with
615 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
616 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
617 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
618 - trace-call-location can be used with TI traces, and replayed, and is compatible
619 with smpi/comp-adjustment-file.
620 - sleep events are now correctly traced/replayed.
621 - Default for trace-call-location is now to use file names and not full paths.
622 To revert to previous behavior (in case of collision of filenames), option
623 "smpi/trace-call-use-absolute-path" can be set to yes.
626 - Use the included xxHash as an hash implem when C++14 is usable.
627 - Option model-checker/hash was removed. This is always activated now.
628 - New option smpi/buffering controls the MPI buffering in MC mode.
629 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
630 This is useful to check for MPI compliance.
633 - New section on plugins: how to define a new one, and existing ones.
636 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
637 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
640 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
641 and <mount> tags. The rationale is finish to decouple what should be handle
642 by SimGrid kernel (i.e., resources and activities) from plugins built in
643 user space (e.g., the file system plugin). This new tag currently accepts
644 only three attributes (id, read_bw, and write_bw). All the other information
645 that was declared with the storage related tags now has to be expressed as
646 properties. An example of platform using this new tag is available at
647 examples/platforms/hosts_with_disks.xml
650 - 'expect signal' can now accept more than one potential signal.
652 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
653 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
654 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
655 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
656 - FG#31: per-actor data
657 - FG#34: SG_BARRIER_SERIAL_THREAD?
658 - FG#35: model-checker does not like buster-produced binaries
659 - FG!13: MC: complete workaround in the error msg seen on modern systems
660 - FG!15: execute_flops now logs compute
661 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
662 - FG!17: ns-3: unblock the right number of communications + others issues
663 - FG!18: Improving the performance of the ns-3 bindings
664 - GH#189: Energy consumption of parallel tasks
665 - GH#219: Error in the throughput of TCP transfer
666 - GH!330: Minor polishing of English: Configuring_Simgrid
667 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
668 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
669 - GH!334: Rework energy plugin, again
671 ----------------------------------------------------------------------------
673 SimGrid (3.23.2) July 8. 2019
676 - Nicer introduction page.
677 - Migrate the "Deploy your application" page to the new doc.
678 - Move Java as a subtree of MSG.
681 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
684 - Drop xbt_dynar_sort_strings().
687 - Really fix FG#26: Turning off a link should raise NetworkFailureException
688 - FG#27: Wrong exception thrown to wait_any when link is turned off
689 - GH#328: Java: Canceling multiple tasks in a single vm/host
691 ----------------------------------------------------------------------------
693 SimGrid (3.23) June 25. 2019
695 The Exotic Solstice Release.
698 - SunOS and Haiku OS support. Because exotic platforms are fun.
699 - Stop setting random seed with srand48() at initialization.
700 - Use addr2line as a fallback for stacktraces when backtrace is not available.
701 - Build option -Denable_documentation is now OFF by default.
702 - Network model 'NS3' was renamed into 'ns-3'.
705 - Simgrid can now hopefully be installed with pip.
708 - wait_any can now be used for asynchronous executions too.
711 - New log appenders: stdout and stderr. Use stdout for xbt_help.
712 - Drop xbt_dict_dump.
715 - 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
716 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
717 - MPI interface is now const correct for input parameters
718 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
719 - Fortran bindings for async collectives.
720 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
723 - Remove option 'model-check/record': Paths are recorded in any cases now.
724 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
725 always sparse. That's an immense gain in memory, and can even be
726 faster because of cache effects. So there is no need to clutter the
727 code to allow the user to go for the unefficient mode.
730 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
731 models proved to be more accurate than these old experiments.
733 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
734 - FG#1: Broken link in error messages
735 - FG#2: missing installation documentation
736 - FG#3: missing documentation in smpirun
737 - FG#6: Python bindings not available on PyPI
738 - FG#7: simple cmake call requires doxygen
739 - FG#8: make python bindings an optional dependency
740 - FG#10: Can not use MSG_process_set_data from SMPI any more
741 - FG#11: Auto-restart actors forget their on_exit behavior
742 - FG#12: -Denable_lto=OFF doesn't disable LTO
743 - FG#13: Installs unstripped file 'bin/graphicator'
744 - FG#14: Installs the empty directory 'doc/simgrid/html'
745 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
746 - FG#17: Dead link in doc (pls_ns3)
747 - FG#20: 'tesh --help' should return 0
748 - FG#21: Documentation link on http://simgrid.org/ broken
749 - FG#22: Debian installation instruction are broken
750 - FG#26: Turning off a link should raise NetworkFailureException exceptions
751 - GH#133: Java: a process can run on a VM even if its host is off
752 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
753 - GH#326: Valgrind-detected error for join() when energy plugin is activated
755 ----------------------------------------------------------------------------
757 SimGrid (3.22) April 1. 2019
759 The Easter Christmas Release (this one is somewhat late).
762 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
763 - Not complete yet: asynchronous activities (amongst others) are still missing
764 - Still ongoing: the interface may change in the future. We need more
765 testers! Please report any glitches.
766 - No new project using Java should start now. Please switch to Python.
769 - Some of the internal cleanups may lead to speed improvements:
770 - The hard limitation on the amount of simulated actors with
771 Java+Mac was removed. Now, the available memory is the only limit.
772 - Our refcounting was tidyied, leading to 10% speedups in some cases.
773 - We are still working on making our code robust to the actor kills
774 and hosts' churn. Things are improving, but it's not perfect yet.
775 - Replaced our own code to display a backtrace (that was forking addr2line)
776 with the Boost.Stacktrace library.
777 You won't see your backtraces without this optional dependency.
778 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
779 - Stop setting random seed with srand() at initialization.
782 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
783 XML file version remains 4.2 since old files are still compatible.
786 - Process termination which was broken at version 3.21 has been repaired.
787 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
788 - Hide the examples into examples/deprecated. New users should use Python.
791 - Drop MSG_process_create_from_stdfunc() from the API.
792 This C++ function was a pimple in the C API, made necessary at some
793 point by the Java bindings. This is fixed now.
794 - Hide the examples into examples/deprecated. New users should use S4U.
795 - MSG_process_create and MSG_process_attach now crash if the host used
796 to run this process is off.
797 - Fix the protype of MSG_process_on_exit()
798 Now use (int,void*) callbacks instead of (void*,void*) ones.
799 The implementation was ways too messy to actually work, I guess.
802 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
803 too often and provide more accurate timings.
804 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
805 - MPI_Alltoallw support
806 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
807 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
808 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
809 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
811 - Basic implementation of generalized requests (SMPI doesn't
812 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
815 - Drop sg_cmdline. Please use xbt_cmdline instead.
816 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
817 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
818 OsSemaphore is implemented in a portable way with C++11 threads.
819 This should allow much more threads to be created at the same time,
820 allowing Mac and Java users to simulate many more actors.
821 - Implement the 'thread' factory with std::thread instead of xbt ones.
822 It is not possible to set the stack size with threads anymore, but
823 -fsplit-stack is the way to go nowadays when using threads.
824 - Drop the xbt_os_thread_t module (now unused)
825 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
826 - Drop xbt_str_join_array().
827 - Drop cunit, use Catch2 instead.
830 - Many cleanups in the kernel::activity namespace. This was long
831 overdue, and shall open the path to many future endeavors.
834 - #132: Java: a process can not shut down its own host
835 - #220: S4U: detached send causes exception if sender terminates soon after sending
836 - #261: Document the parameters of parallel execution's constructor
837 - #300: [s4u] BarrierPtr is missing
838 - #314: SMPI args internal cleanup
839 - #316: Fix a bug related to the CPU utilization of multi-core VM
840 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
841 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
842 - #325: Turning off a host has different behavior on sleeping actors and computing actors
844 ----------------------------------------------------------------------------
846 SimGrid (3.21) October 3. 2018
848 The Restarting Documentation (TRD) Release.
851 - Convert the user manual to Sphinx for improved usability.
852 Unstable version now hosted on https://simgrid.frama.io/simgrid
853 Many glitches remain; stay tuned.
854 - Restrict the Doxygen reference API to the installed header files.
855 (documenting all internals this way is overkill + doxygen scales badly)
856 - New tutorial on S4U (highly inspired from the old MSG one, but with
857 a git repository to fork as a starting point, and a docker image)
858 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
859 - The Developper manual is still to be converted (not compiled until then)
860 - Some parts are still missing in this conversion (such as the
861 platform and deployment sections) while others were blindly converted
862 and would need more love (such as the configuration flags).
863 Things will certainly further improve in the future releases.
866 - Fully reimplement auto_restart mechanism that was utterly broken
867 (fix #22, #131 and #281 that were all very related)
868 - Implement semaphores (fix #303 and #291)
869 - When creating an actor from a function and its parameters,
870 move-only parameters are not allowed anymore, as it would prevent
871 the actor to be restartable if its parameters are consumed on run.
872 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
873 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
874 IO operations that can be started, waited for, or canceled as a regular
876 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
880 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
881 - New host variable: 'core_count'
884 - Remove xbt_os_thread_specific features
885 - Remove portability wrapper to condition variables
886 - Remove xbt_os_thread_yield()
889 - MPICH collective selector now mimics MPICH 3.3b
890 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
891 - MPI_Init(NULL,NULL) is now allowed
892 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
893 - Command-line arguments should now be handled properly in Fortran simulations
894 - MPI Attributes and related callbacks should work in Fortran
895 - Apps using C + Fortran should now work
896 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
897 - Support MPI_Error_String
900 - Due to an internal bug, Msg.run() must now be your last line.
901 We hope to fix it in a future release, and we are sorry for the inconvenience.
904 - #22: Process autorestart seem to only work with CAS01 cpus
905 - #56: Feature request: dynamic SMPI replay
906 - #93: simgrid should not eat --help
907 - #111: How to change the loopback link characteristics?
908 - #116: Platform cluster radical
909 - #117: Platform cluster/host order
910 - #129: Replay_multiple on steroid
911 - #131: Java : autostart does not work
912 - #143: Setting a breakpoint at a given time
913 - #235: Network energy models should be integrated
914 - #264: Add ptask L07 resource tracing
915 - #271: Dynamic and manual replay of SMPI traces
916 - #279: Breakpoints option uses time deltas instead of absolute time
917 - #281: Daemonized actor no longer daemonized after an auto-restart
918 - #282: TIT convertor erase traces when using absolute path in the trace list file
919 - #285: segfault when a s4u actor kills itself with msg process tracing activated
920 - #286: Pajé traces are not exposing the number of cores
921 - #287: Command-line parsing should stop when encountering '--'
922 - #288: MPI_Init(NULL, NULL)
923 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
924 - #290: Method to check if a mailbox has a message ready to be consumed
925 - #291: [s4u] semaphores do not seem available yet
926 - #294: Alltoall collective implementation of mpich
927 - #295: Starting a Host as OFF is not supported
928 - #296: DTD too permissive
929 - #299: Add a s4u barrier example
931 ----------------------------------------------------------------------------
933 SimGrid (3.20) Released June 24. 2018
935 The proxy snake_case() release.
938 - The whole API is now using snake_case() instead of mixing it with camelCase.
939 Compatibility wrappers in place for one year.
940 - We took the opportunity to increase the consistency of the API.
941 Please report any remaining glitches.
942 - New function simgrid::s4u::Host::get_actor_count:
943 Returns the number of actors running on a specific host.
946 - SMPI is now tested with ~45 proxy apps from various sources, with none or
947 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
948 - Replay: The replay file has been re-written in C++.
949 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
950 supported. They are stored in the trace and used when replayed.
951 - Basic support of MPI_Cancel. Robustness not guaranteed.
952 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
953 (only for MPI_COMM_TYPE_SHARED).
954 - New option: smpi/privatize-libs, to add external shared libs to be privatized
955 by SMPI. They will be copied locally and loaded separately by each process.
956 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
957 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
958 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
959 with internal index optind with multiple processes. Only works if MPI_Init has
961 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
962 syntax without preprocessing tricks.
965 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
966 behavior tracing is no longer limited to MSG
967 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
968 this means that traces now need to be updated or re-obtained; both Irecv
969 and Isend lines in your traces have now in total 5 mandatory fields:
970 <rankid> <command> <to/from rankid> <tag> <size>
971 To update your traces, it suffices to add a 0 for the tag here.
972 - TIT now also supports waiting for a distinct request via MPI_Wait.
973 Wait/Test now wait for a specific request, not just the last one that was
974 issued. This unfortunately means another update, because we need to
975 identify which request you want to wait for. We do this via the
976 triplet (sender, receiver, tag), which needs to be added:
977 <rankid> <command> <sender> <receiver> <tag>
978 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
979 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
982 - The deprecation of MSG is ongoing (but this should not impact you).
983 Many MSG functions are now simple wrappers on the C API of S4U. If
984 you wish to convert your code to S4U, find the S4U counterparts of
985 your MSG calls in src/msg/msg_legacy.cpp.
986 - MSG can still be used, but won't evolve anymore.
989 - Allow to run the Link energy plugin from the command line with
990 --cfg=plugin:link_energy
991 - Rename Energy plugin into host_energy
992 - Rename Load plugin into host_load
995 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
996 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
997 - Improve performance when handling timeouts of simix synchros.
1000 - Config: the C API is now deprecated (will be removed in 3.23), and
1001 the C++ API has been slightly improved.
1004 - Fix several build issues on OSX.
1005 - Move simgrid_config.h to simgrid/config.h (old header still working)
1008 - #143: Setting a breakpoint at a given time
1009 - #150: Inconsistent event names in SMPI replay
1010 - #258: daemonized actors hang after all non-daemonized actors have completed
1011 - #267: Linker error on unit_tmgr
1012 - #269: SMPI: tracing of MPI_Wait/all/any broken
1013 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
1015 ----------------------------------------------------------------------------
1017 SimGrid (3.19.1) Released March 22 2018
1019 The Fixed ABI Release.
1021 As you may know, we are currently refactoring SimGrid in deep.
1022 Upcoming SimGrid4 will be really different from SimGrid3: modular,
1023 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
1025 Our goal is to smooth this transition, with backward compatibility
1026 and automatic update paths, while still progressing toward SimGrid4.
1028 SimGrid remains open during works: The last pure SimGrid3 release was
1029 v3.12 while all subsequent versions are usable alpha versions of
1030 SimGrid4: Existing interfaces remain unchanged, but the new S4U
1031 interface is budding and the internals are deeply reorganized.
1033 Since 2015, we work hard to reduce the changes to public APIs. When
1034 we need to rename a public library symbol in S4U, we let your
1035 compiler issue an explicative warning when you use the deprecated
1036 function. These messages remain for four releases, i.e. for one full
1037 year, before turning into an error. Starting with v3.15, your can
1038 also adapt to API changes with the SIMGRID_VERSION macro, that is
1039 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
1041 Starting with this v3.19.1, our commitment to reduce the changes to
1042 the public interfaces is extended from the API to the ABI: a program
1043 using only MSG or SimDag and compiled against a given version of
1044 simgrid can probably be used with a later version of SimGrid without
1045 recompilation. We will do our best... but don't expect too much of
1046 it, that's a really difficult goal during such profund refactoring.
1048 The difference between v3.19 and v3.19.1 is that the former was
1049 accidentally breaking the ABI of MSG, while the later is restoring
1052 S4U and kernel APIs will still evolve until SimGrid4, with one-year
1053 deprecation warnings as currently. In fact, cleaning up these
1054 interfaces and converting them to snake_case() is one release goal of
1055 v3.20. But don't worry, we are working to smooth this upgrade path.
1057 In summary, new projects should start with S4U to benefit of the
1058 future, but old MSG projects should still be usable with no change.
1060 ----------------------------------------------------------------------------
1062 SimGrid (3.19) Released March 20 2018
1064 The Moscovitly-cold Spring release.
1066 SMPI user-visible improvements:
1067 - Support of MPI_Type_create_subarray
1068 - Unimplemented MPI calls now abort (instead of issuing a warning)
1069 Some harmless calls around errhandlers still only issue a warning.
1070 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
1071 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
1072 is now possible (mmap privatization was refactored to this end).
1073 Adding new SMPI instances should also be technically doable now.
1075 SMPI internal cleanups:
1076 - Removed index notion from SMPI -- no more "getPid() - 1"!
1077 In tracing files, actors are called rank-<PROCESS-ID> but these are
1078 NOT the real ranks (tracing several communicators is not supported yet).
1079 - Deprecate smpi_process_index() (will be removed in 3.22)
1080 - Better testing on classical proxy apps, through an external project:
1081 https://github.com/simgrid/SMPI-proxy-apps/
1084 - Use a graphical TOC to make it easier to find the documentation you need
1085 - Revamp the MSG tutorial
1088 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
1089 - Remove parameter of MSG_process_killall().
1090 Resetting the PID was bogus anyway (several actors could have the same PID).
1093 - Execution->setHost() can be called after start(), to migrate it.
1094 - Comm::test_any() is now implemented.
1095 - s4u::Actor now has onCreation() and onDestruction() signals.
1096 - Install some kernel header files for the users' plugins and more.
1097 Warning, their API is really not stable yet.
1098 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1099 include/simgrid/kernel/routing/*Zone.hpp
1102 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1104 Build System and other cleanups:
1105 - Remove unused run-time parameter "tracing/onelink-only".
1106 - SimGrid now works with intel compilers (MC still buggy)
1107 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1108 - tesh kills subprocesses on timeout (but not on windows)
1111 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1114 - #194: Feature request: simgrid::s4u::Comm::test_any()
1115 - #245: migrating an actor does not migrate its execution
1116 - #253: Feature Request: expose clusters as objects
1117 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1118 - #255: Tesh broken on Windows
1119 - #256: Modernize FindSimGrid.cmake
1120 - #257: Fix (ab)use of CMake install
1122 ----------------------------------------------------------------------------
1124 SimGrid (3.18) Released December 24 2017
1126 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1128 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1129 - Support for MSG should not be removed from SimGrid before 2020,
1130 but future evolutions will be limited to the integration of
1131 user-provided patches.
1134 - s4u::onDeadlock() signal, run before stopping the simulation.
1135 - s4u::Exec: asynchronous executions (abstraction of background
1137 - s4u::Host->getLoad() returns the achieved speed in flops/s
1140 - Link::name() is deprecated and will be removed in v3.21, use
1141 Link::getCname() instead.
1142 - Mailbox::getName() changed to return a std::string, use
1143 Mailbox::getCname() instead to get a char*.
1144 - Storage::getName() changed to return a std::string, use
1145 Storage::getCname() instead to get a char*.
1146 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1147 to have both allocation and deallocation of the map in user space.
1150 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1151 The amount of flops remaining to do is a vector, not a scalar.
1152 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1153 name implies on both sequential and parallel tasks.
1154 - Both changes fix GitHub's #223 using PR #237 as a basis.
1155 Thanks Michael Mercier.
1156 - Most examples were converted to S4U and hidden elsewhere at tests.
1157 Rationale: we still want MSG to work; we want newcomers to use S4U.
1160 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1161 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1164 - Switch to the faster dlopen privatization mechanism by default
1165 - Documentation improvments
1168 - Live migration is getting moved to a plugin. Dirty page tracking is
1169 the first part of this plugin. This imply that VM migration is now
1170 only possible if one this function is called:
1171 - C/MSG: MSG_vm_live_migration_plugin_init()
1172 - C/C++: sg_vm_live_migration_plugin_init()
1173 - Java: Msg.liveMigrationInit()
1174 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1178 - Define class simgrid::xbt::Path to manage file names.
1179 - Removed unused functions:
1180 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1181 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1182 xbt_graph_export_graphviz()
1183 - xbt/str.h: xbt_str_join()
1184 - Remove unused datatypes:
1185 - xbt/heap.h: use std::priority_queue or boost::heap instead
1186 - xbt/swag.h: use boost::intrusive::list instead
1189 - New link_energy plugin for the consumption of the links.
1190 - All of the operations on files and storage contents have been
1191 packaged into a plugin (src/plugins/file_system). The current
1192 public interface can be found in
1193 include/simgrid/plugins/file_system.h
1194 To use these functions you now have to initialize the plugin by
1195 calling MSG_storage_file_system_init() just after calling
1196 MSG_init() or sg_storage_file_system_init() just after creating
1200 - Remove the undocumented/untested tag <include>
1203 - Remove viva specific tracing as the tool is no longer maintained
1206 - #248: Kill of finished processes leads to segfault
1207 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1208 - #239: please implement signal s4u::onDeadlock()
1209 - #230: segfaults when exit() without run()
1210 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1211 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1212 - #222: Actor::kill() doesn't really kill and segfaults
1213 - #221: odd LMM warning when killing an actor
1214 - #120: Memory leak when the processes are forcefully killed
1216 ----------------------------------------------------------------------------
1218 SimGrid (3.17) Released October 8 2017
1220 The Drained Leaks release: (almost) no known leaks despite the tests.
1222 Even very long simulations will preserve your memory: our testsuite
1223 stresses SimGrid for over 45mn on fast machines for a coverage of
1224 over 80%, and there is only one single known leak, of about 4kb.
1228 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1229 handler may be given to free resources if the comm cannot be completed.
1230 - this_actor::send(mailbox) is now mailbox->put()
1231 - New: simgrid::s4u::Comm::wait_all()
1232 - New: Host.execute() for remote executions.
1235 - Unused option network/sender-gap is removed.
1238 - Deprecate MSG_task_isend_with_matching(): This unused feature
1239 really complicates our internals. Will be removed in v3.20.
1242 - Improved context termination. It is now possible to free resources, even
1243 when a process is forcibly killed.
1244 - ContextBoost: add support for Boost versions above 1.61.
1247 - Bring back run-time option --cfg=exception/cutpath to remove exception
1249 - Removed unused functions:
1250 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1251 xbt_str_rtrim(), xbt_str_trim().
1252 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1255 - Removed header files obsolete since SimGrid 3.12:
1256 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1258 - #3: SD_exit should be made optional
1259 - #120: Memory leak when the processes are forcefully killed
1260 - #159: Threading test regression in Actor refcounting
1261 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1262 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1263 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1264 - #191: VM migration and pstate
1265 - #192: Updating the energy consumptions of all hosts crashes with VMs
1266 - #195: All actors have PID=0 in the logs
1267 - #204: Sometimes segfault with thread contexts and mmap privatization
1268 - #222: Actor::kill() doesn't really kill and segfaults
1269 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1271 ----------------------------------------------------------------------------
1273 SimGrid (3.16) Released June 22. 2017.
1275 The Blooming Spring Release: developments are budding.
1278 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1279 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1280 - Java JAR file should be finally fully working on Mac OSX too.
1282 XML platforms: Switch to platform v4.1 format.
1283 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1284 - <zone> can be used as a synonym for the now deprecated <as>
1285 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1286 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1287 - <actor> can be used as a synonym for the now deprecated <process>
1288 - state_file and avail_file periodicity is now easier to express
1289 (check the documentation) the old behavior should still work.
1290 * Storage: not backward compatible, but it seems that nobody ever used it
1291 - Remove attribute 'content_type' of <storage_type>: was never used
1292 - Make attribute 'model' of <storage_type> optional: for future usage
1293 - Remove Bconnection model property: was never (in)validated. Replaced by
1294 the maximum of the read and write bandwidth as a resource constraint.
1297 - New and Backwards Compatibility break:
1298 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1299 When one wants to get the list of tasks whose states have changed during a
1300 simulation round, s/he has to allocate and free a dynar and use it as
1301 argument to this function. The former SD_simulate (double how_long)
1305 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1306 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1309 - New function to update the consumption of all hosts at once.
1310 - Fix the model for multi-core hosts, linear in the amount of busy
1311 cores with an abnormality for pIdle (see paper).
1312 BUG: Still not satisfactory for ptask on mono-cores.
1315 - The netzone are now available from the MSG API.
1316 The old names still work, but are now deprecated.
1319 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1320 - New option: smpi/keep-temps to not cleanup temp files
1321 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1322 mallocs mode is used, allows to change the size of the fake file used
1323 (default 1MB), to potentially limit the number of mappings for large runs.
1324 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1325 - Fortran ifort and flang compilers support
1326 - New RMA calls supported (experimental):
1327 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1328 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1329 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1330 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1331 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1332 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1335 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1336 - Behind the scene, ActivityImpl are now automatically refcounted.
1337 Removing this nasty bottleneck will greatly simplify our internals.
1340 - Replay: New function xbt_replay_action_get():
1341 Retrieve the function previously associated to an event type.
1342 - DROPPED FUNCTION: xbt_str_varsubst()
1343 - DROPPED MODULE: strbuff. We don't need it anymore.
1344 - DROPPED MODULE: matrix. We don't need it anymore.
1345 - DROPPED MODULE: lib. We don't need it anymore.
1347 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1349 SimGrid (3.15) Released March 22. 2017
1351 The Spring Release: continuous integration servers become green
1352 We fixed even the transient bugs on all target architectures:
1353 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1354 Windows, FreeBSD, NetBSD.
1356 - Rename NetCards to NetPoints.
1357 This was intended to help NS3 users, but that's not a netcard.
1358 That's a point in the routing algorithm, let's avoid wrong simplifications.
1361 - New: MSG_process_yield(). Stop and yield to other processes.
1362 - New: MSG_process_daemon(). Daemon processes are automatically killed
1363 when the last non-daemon process terminates
1364 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1365 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1366 to make room for the upcoming network energy plugin.
1367 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1370 - Ensure that an actor can kill itself with Process::exit()
1371 - Kill the obscure NativeException. Nobody want to survive the issues
1372 it denotes, so use JniException that is a RuntimeException (not to
1373 be caught explicitly).
1374 - Partial bug fix in initialization. SimGrid flags on command line were
1375 consumed at C level but stayed in the original Java String[] args.
1376 This could mess users' args[i] if SG flags were not put at the end of
1378 The SimGrid flags are now removed from the Java arguments. However,
1379 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1380 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1381 It might be if you have --log or --cfg flags in the command line.
1382 - Fix numerous memleaks all around the place. In particular, around VMs.
1386 - s4u::onPlatformCreated: right before the simulation starts
1387 - s4u::onSimulationEnd: right after the main simulation loop
1388 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1389 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1390 event from the availability_file changes the avail speed.
1391 - Links are now usable from s4u
1392 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1393 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1394 - New examples: The conversion of MSG examples to S4U has begun
1395 - Actors: create, kill, migration, and suspend
1396 - Applications: master-worker and token-ring
1397 - Action replay: communications and storage
1398 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1401 - Backwards Compatibility breaks
1402 - The SD_route_get_list and SD_route_get_size functions have been removed.
1403 They are replaced by the unique
1404 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1405 The route from src to dst is built in the links dynar whose size can
1406 be obtained with xbt_dynar_length.
1407 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1408 sg_host_route_bandwidth and sg_host_route_latency.
1409 Macros ensure the backwards compatibility, but you should fix your code
1412 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1413 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1414 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1415 We don't support 100% of it, but it should be enough. Contact us if not.
1416 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1417 naive and probably inaccurate)
1418 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1419 mapped onto a unique small file using some system magic.
1420 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1421 - Fortran: cleanups, fixes, support of user-added operations
1422 - RMA: MPI_Accumulate are applied in correct order.
1423 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1424 - Better support for MPI_IN_PLACE
1425 - Support for MPI_Win attrs and keyvals.
1426 - Support MPI_Comm_group_create, MPI_Type_size_x
1427 - MPI ops should be properly applied to all allowed types
1428 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1429 algorithm by a non failing one from mpich.
1432 - Kill the fifo data container: we don't use it anymore.
1435 - A <cluster> can now be created with different speed values to represent pstates.
1436 see examples/platforms/energy_cluster.xml for an example.
1438 SimGrid (3.14.159) Released December 28. 2016
1440 The previous release was only a crude approximation of a Pi release;
1441 this one should be a bit better.
1443 - Revert a last minute change that broke on Mac OSX.
1444 - Fix the changelog and NEWS files.
1445 - Improve the documentation, in particular of the routing module.
1446 - Integrate some patches flying around in Debian.
1448 SimGrid (3.14) Released December 24. 2016
1451 * The whole documentation was reworked and reorganized. There is
1452 still much room for improvement here, but we're on it.
1455 * We now need python3 for our testing infrastructure.
1456 * Model checking needs libevent
1459 * Now works on FreeBSD
1462 * Privatization now works on FreeBSD
1464 * Privatization is now activated by default in smpirun
1465 You can turn it off with -no-privatize if something goes wrong.
1467 * Call-location tracing for SMPI.
1468 You can add the exact location (filename / linenumber) of an MPI call to
1469 your trace files and slow down or speed up the simulation between two
1470 consecutive calls by using an adjustment file (see the documentation).
1472 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1473 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1474 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1475 values were ignored in some cases. The timings of these functions can now
1476 be significantly different.
1478 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1479 smpi/running-power is renamed to smpi/host-speed
1481 * smpi/grow-injected-times option to enable or disable multiplication of the
1482 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1483 make simulation less precise (but also much faster).
1485 * smpirun script should be (much) faster for large deployments.
1487 * SMPI tracing: fixed issue with poor matching of send/receives.
1489 * Replay: Fix broken waitall
1491 New functions and features
1492 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1494 Dropped / renamed functions and features
1495 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1496 - MSG_mailbox_is_empty() -> Mailbox::empty()
1497 - MSG_mailbox_front() -> Mailbox::front()
1498 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1499 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1500 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1501 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1502 * Don't pass the free_f parameter to property related functions:
1503 - MSG_host_set_property_value()
1504 - MSG_as_router_set_property_value()
1505 - MSG_storage_set_property_value()
1506 * VM properties. Since msg_vm_t are msg_host_t, just use
1507 MSG_host_get_property_value() and friends
1508 * VM I/O related things:
1509 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1510 - Unimplemented save/restore methods
1511 * MSG_as_router_get_property_value() was redundant with
1512 MSG_environment_as_get_property_value().
1513 - Removed MSG_as_router_*propert*() functions
1514 - Added MSG_environment_as_set_property_value() for consistency
1515 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1516 Well, they are still there for now, but deprecated with a warning.
1517 Please switch to xbt_dict_new_homogeneous() before this is removed
1519 * Task affinity. Its intended behavior (that was very badly tested
1520 and probably not really working) was deceiving what most users
1521 would have hoped here.
1522 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1523 * Option network/coordinates is now useless and should be dropped.
1526 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1529 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1530 * Added Dragonfly topology support
1533 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1534 EiBps, EBps, Eibps, Ebps for bandwidth.
1535 They may become useful to some lucky ones.
1538 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1539 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1540 * ex/async_waitAll: new example, on asynchronous communications
1543 * Memory usage should be decreased for simulations with a large number
1544 of processes. This also helps for SMPI.
1546 SimGrid (3.13) stable; urgency=low
1548 The Half Release, a.k.a. the Zealous Easter Trim.
1550 - We removed half of the lines, that were mostly experimental cruft.
1551 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1552 (not counting blanks and comments -- according to openhub.net)
1553 - The internals are now compiled in C++ (and will soon be clean C++)
1554 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1555 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1557 * Backwards Compatibility breaks
1558 - Removed Lua simulation bindings (switch to C or Java for that).
1559 Lua can still be used to describe platforms
1560 - Removed Java kernel plug-ins.
1561 Will be reintroduced after the ongoing major internals reorg.
1563 - the following functions were removed.
1564 They were too specific and should be reimplemented in a generic
1565 way, with filter function.
1566 - MSG_task_listen_from_host
1567 - MSG_mailbox_get_count_host_waiting_tasks
1568 - MSG_mailbox_put_with_timeout was removed.
1569 Please use MSG_task_send_with_timeout instead.
1571 - the SD_application_reinit function was removed. It has been a noop for a while.
1572 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1573 replaced by a more flexible mechanism.
1574 - The following functions thus do not exist anymore
1575 - SD_workstation_get_access_mode
1576 - SD_workstation_set_access_mode
1577 - SD_workstation_get_current_task
1578 - Basic estimation functions have been removed but can easily be replaced
1579 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1580 - SD_workstation_get_computation_time => amount / sg_host_speed()
1582 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1583 Use VM.getSpeed()*load/100 for the legacy behavior.
1585 - option enable_tracing was removed. It was not doing anything for a while.
1586 - In the ModelChecker:
1587 - the model-checker now ptraces the model-checked process which means
1588 you cannot use a debugger on the latter anymore (we might make this
1589 optional in the feature);
1590 - removed soft-dirty page tracking;
1591 - remove model-checked side snapshot management,
1592 MC_snapshot() and MC_compare_snapshot();
1593 - keep the MC_cut() function as a stub (it was not really working
1594 in the previous release).
1597 * All options are consistently in kebab-case. Old names are kept as alias.
1600 * Switch to platform v4 format.
1601 - Rename from 'power' to 'speed' the attributes describing the amount of
1602 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1603 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1604 - In <host> and <link>, attributes availability and state are gone.
1605 It was redundant with state and availability traces, and with peak values.
1606 - In <cluster>, attributes availability_file and state_file are gone.
1607 It was too complex and unused.
1608 - Kill <gpu>. Was not doing anything.
1609 - The DOCTYPE points to the right URL:
1610 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1611 (the file at this address now documents the changelog since its v1)
1612 - A warning is emitted for unit-less values (they are still accepted).
1613 - speed. Default: 'f' or 'flops'. Also defined:
1614 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1615 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1616 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1617 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1618 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1619 - latency. Default: 's' second. Also defined:
1620 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1622 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1623 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1626 * s4u::Host is now the preferred public interface to the Host features.
1627 sg_host_* functions are C bindings to the exact same behavior
1628 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1631 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1632 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1633 They were deprecated since a few years, and probably did not even compile anymore.
1636 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1637 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1638 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1639 encouraged to update their codes. A list of the modified functions can be found at the end of
1640 include/simgrid/simdag.h
1643 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1644 the lambda or closure passed as a parameter will run in kernel mode.
1645 All the callback functions should be rewritten to that interface at some point.
1648 * Reorganizing and cleaning the internals all around the place.
1651 * Remove old default barrier/bcast buggy algorithms (see #18407)
1652 * Various bug fixes to handle more codes
1653 * Remove the need for the --foreground option of smpirun (it is still
1654 accepted for backward compatibility).
1657 * Kill the setset data container: MC don't use it anymore.
1658 * Kill the queue data container: it made more sense with GRAS.
1659 * Kill the xbt_peer_t data type: it's useless without GRAS.
1660 * Kill rm_cb feature of config sets: it was never useful.
1661 * Kill graphxml parsing feature. It was not used.
1662 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1664 - xbt_dynar_sort_strings(), when the content is char*
1665 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1666 They throw exceptions on invalid input;
1667 * C++ support for declaring CLI flags (simgrid::config::Flag);
1668 * class for abstracting different signal backends (simgrid::xbt::signal).
1669 with no external dependencies (we need very simple signals).
1672 * refactoring and cleanup of the code;
1673 * ongoing process to cleanly separate the model-checking algorithms
1674 from the code model-checking support.
1676 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1678 SimGrid (3.12) stable; urgency=low
1680 The Facelift Release.
1683 * Require g++ v4.7 at least to not speak prehistorical C++.
1684 * Require Boost 1.48 (for signal2 component).
1685 * Java must be version 7 at least when activated.
1686 * Builds on Windows again (including Java bindings).
1687 * Tracing is now always enabled (no way to turn it out)
1688 * Remove GTNetS. It was not working anyway.
1689 * Various cleanups in the cmake scripts.
1690 * Move headers around to sort them out on installed systems:
1691 - instr/instr.h -> simgrid/instr.h
1692 - instr/jedule/* -> simgrid/jedule
1693 - simdag/datatypes.h was removed
1694 - simdag/simdag.h -> simgrid/simdag.h
1695 - msg/datatypes.h was removed
1696 - msg/msg.h -> simgrid/msg.h
1699 * Interface improvement:
1700 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1701 - Sanitize the interface in MSG_task_ module:
1702 - Merge two functions that were close enough but misleading:
1703 set_compute_duration(t) -> set_flops_amount(t)
1704 get_remaining_computation(t) -> get_flops_amount(t)
1705 - set_data_size(t) -> set_bytes_amount(t)
1706 get_data_size(t) -> get_bytes_amount(t)
1707 - Massive cleanups in the functions related to the energy
1708 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1709 - New: MSG_host_get_pstate()
1711 - msg/energy/onoff: switching hosts on and off
1713 * Interface improvement:
1714 - Rename Host.isAvail() to Host.isOn()
1715 - Rename Process.currentProcess() to Process.getCurrentProcess()
1716 - Rename Task.setDataSize() to Task.setBytesAmount()
1717 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1719 - #18874: Actually allows the GC to reclaim tasks
1723 - SIMIX_process_throw: raises an exception in a remote process
1724 * Refactoring: Separate sub-modules
1725 - libsmx: the public interface, as libc in a real system
1726 - popping: the strange dance that converts a user request into a kernel handling
1727 - smx_context_*: the virtualization mechanisms that embed the user code
1728 - smx_*: the handling of each simcalls
1729 * Interface cleanup:
1730 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1731 * Rename smx_action_t into smx_synchro_t, making explicit that these
1732 things are used to synchronize processes with their environment.
1733 For example, a communication is a sort of synchronization involving
1734 the communicating processes (that may block until the exchange) and
1735 the platform. The same can be said from computations, etc.
1737 - #18888: segfault when a process ends before its kill_time
1741 - 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.
1742 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1743 - MPI_Type_set_name, MPI_Type_get_name
1744 - MPI_*_c2f and MPI_*_f2c functions
1745 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1746 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1747 - Activate a lot of new tests from the mpich 3 testsuite
1749 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1750 - InfiniBand network model added: Based on the works of Jerome Vienne
1751 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1752 - When smpi/display_timing is set, also display global simulation time and application times
1753 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1754 * Collective communications
1755 - 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
1756 - 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/).
1757 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1759 - Replay now uses algorithms from wanted collective selector
1760 - Replay can be used with SMP-aware algorithms
1761 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1762 - 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.
1764 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1765 - Correctly use loopback on fat-tree clusters
1766 - Asynchronous small messages shouldn't trigger deadlocks anymore
1767 * Energy/DVFS cleanup and improvement
1768 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1769 - new: smpi_get_host_pstate()
1772 - "Full" network optimization flag was broken since Surf++
1773 - Better handling of precision flags in maxmin
1774 - Fix bug causing sometimes "Impossible" errors
1775 - Properly pass cluster properties to included hosts
1776 * Improvement of the Energy plugin.
1777 - Always update the consumption before returning that value
1778 - New property: watt_off to denote the dissipation when the host is off
1779 - New functions getWattMinAt and getWattMaxAt to retrieve the
1780 dissipation of pstates that we are not currently at.
1781 * Java: class NetworkLink renamed to Link
1782 * New function: simcall_process_get_kill_time()
1783 * Massive rename s/workstation/host/
1784 - That's intrusive, but that's good for the project consistency. Sorry.
1785 - Change config option "workstation/model" into "host/model"
1788 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1789 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1790 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1791 - xbt_dirname and xbt_basename for non-POSIX systems
1793 * The model checker now runs as a separate process.
1794 * The model checker runs must now be launched with the new simgrid-mc program.
1795 * Record/Replay: the MC can display a textual representation of a path in the
1796 execution graph. It can then be replayed outside of the model checker.
1798 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1800 SimGrid (3.11) stable; urgency=low
1805 * Normalizing pointers addresses tool for better diff between logs
1808 * Add cloud examples using new VMs
1809 - examples/msg/cloud/two_tasks_vm.tesh
1810 - examples/msg/cloud/simple_vm.tesh
1811 - examples/java/cloud/cloud.tesh
1812 - examples/java/cloud/migration/migration.tesh
1813 * Add java surf examples:
1814 - examples/java/surfPlugin/surf_plugin.tesh
1815 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1816 - examples/java/surfCpuModel/surf_cpu_model.tesh
1817 * Add SMPI+MSG example:
1818 - examples/smpi/smpi_msg_masterslave/
1823 - msg task destroy cancel test
1824 - msg_host on/off test
1825 * Move all tests in testsuite to teshsuite (adding tesh files)
1826 * Restructure teshsuites
1827 - one folder for each kind of test
1828 * Restructure AddTests.cmake
1830 - structure the order of tests (with sections)
1833 * Add virtual machine
1834 - creation of a VM on a PM
1835 - migration of a VM from a PM to another PM
1837 - MSG_process_join(msg_process_t process, double timeout)
1838 - msg_bar_t MSG_barrier_init(unsigned int count)
1839 - int MSG_barrier_wait(msg_bar_t barrier)
1840 - void MSG_barrier_destroy(msg_bar_t barrier)
1841 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1842 * New option "msg/debug_multiple_use" to help debugging when a task is used
1845 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1847 - Increase code coverage in test suites
1851 * Protect context stacks against stack overflow. The number of protected memory
1852 pages allocated on the top of each stack (1 by default) can be configured
1853 with the new command line option --cfg=contexts/guard_size:<value>.
1854 * Simcalls are now generated by a python script that
1855 - generates files included by SimGrid
1856 - checks that all the functions exist, and proposes prototypes
1858 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1859 host_set_data, host_get_data
1861 - simcall_process_join(smx_process_t process, double timeout)
1862 * Fix bug where sleeping processing could not be suspended.
1865 * Translate surf models from C to C++
1866 - Generic classes for all models: Model, Resource, Action
1867 - A generic interface for each kind of model (CPU, Network, Storage
1868 Workstation, WorkstationVM)
1870 * Translate surf routings from C to C++
1871 * Add callbacks using sigc++ or boost::signals2
1872 - Add callback functions for resource creation/destruction
1873 - Add callback functions for action state change
1874 - Handle Energy as a plugin
1875 * Replace swag by boost::intrusive
1876 * Add new routing models for clusters. For documentation, see
1877 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1878 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1879 parameters for cluster tag
1880 - Fat trees, with topology="FAT_TREE" and
1881 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1882 - see examples/platforms/torus_cluster.xml and
1883 examples/platforms/fat_tree_cluster.xml
1884 * More documentation
1887 * Hostfiles support host:nb_processes construct to deploy several processes on
1889 * Collective communication algorithms should not crash if used with
1890 improper number of nodes and report the error.
1891 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1892 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1893 MPI_Dims_create, MPI_Cart_sub are supported.
1894 * New interface to use SMPI programmatically (still depends on MSG for
1895 some parts, see examples/smpi/smpi_msg_masterslave):
1896 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1900 * Global variables privatization in MPI executables is now performed at runtime
1901 with the option smpi/privatize_global_variables (default:no).
1902 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1903 dynamic libraries loaded by the application are not privatized (static
1904 linking with these libraries is advised in this case)
1907 * Options defined in XML work correctly now.
1910 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1911 copied into simgrid.jar or not (ON by default). Use this option if you want
1912 to reduce the size of the installed simgrid.jar, *and* the native libraries
1913 are kept installed elsewhere.
1914 * Surf binding with SWIG (code generated in maintainer mode only):
1915 - plugin to handle callbacks
1916 - CPU model only for the moment
1919 * Supernovae build mode is definitively removed. It was used to improve
1920 inlining and inter-module optimizations. It is nowadays superseded by
1921 link-time optimizations commonly available in compilers.
1922 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1923 * Add boost dependency for surf++
1924 * Add new macro for tests
1925 - ADD_TESH(name <tesh_args>)
1926 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1930 - xbt_bar_t XBT_barrier_init(unsigned int count)
1931 - int XBT_barrier_wait(xbt_bar_t barrier)
1932 - void XBT_barrier_destroy(xbt_bar_t barrier)
1933 * Make the xbt_os_time module public
1935 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1937 SimGrid (3.10) stable; urgency=low
1939 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1942 * Reintegrate Java to the main archive as desynchronizing these
1943 package is not acceptable anymore (Java is now considered stable)
1944 * Add explicit synchronization facilities through semaphores
1945 * Bug fix: Task.setDataSize() only changed the C world, not the value
1946 cached in the Java world
1949 * Dramatically change the way files are handled. API and internals changed, but
1950 this part of MSG was not considered as production grade either.
1951 * Add explicit synchronization facilities through semaphores
1952 * Add a new function MSG_host_get_process_list()
1953 * Preliminary DVFS support (see examples/msg/energy/ for details)
1956 * SMPI is now included directly in the libsimgrid as the windows
1957 linker doesn't force us on splitting it anymore.
1958 * Improvements of the SMPI replay tool:
1959 - Most of the collective communications are now rooted in the same process as
1960 in the original application.
1961 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1962 used until now). Multiple data types can now be used in a trace.
1963 - The replay tool now supports traces produce either by TAU or a modified
1965 - Bug Fix: the compute part of the reduce action is now taken into account.
1966 - Gatherv collective is now supported
1967 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1968 of smpirun outputs time independent traces for the current run. One file
1969 is created per process. If too many processes are simulated, this behavior
1970 can be changed to one file for all processes by using the
1971 tracing/smpi/format/ti_one_file flag
1972 * smpirun generates the host file if needed (with given host count and platform)
1973 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1974 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1975 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1976 --cfg=smpi/coll_selector:(mpich/ompi)
1977 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1979 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1980 happening inside a collective SMPI call.
1981 * Fix the behavior of complex data types handling.
1982 * Make MPI_Wtime another synchronization point to take computations into
1984 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1985 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1986 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1987 (no privatization of global variables yet)
1988 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1991 * Verification of liveness properties is now available for SMPI applications
1992 (in addition to MSG applications)
1993 * Bugged examples using SMPI in examples/smpi/mc/
1994 * Add --cfg=model-check/visited option. Allows the verification of infinite
1995 programs. Detection of loops in the execution thanks to the system state
1996 comparison and reduction of the state space to explore. Can be combined with
1997 DPOR for safety properties.
2000 * Allow to change SimGrid configuration (see --help) within the code
2001 thanks to SD_config() as it can be done in MSG.
2002 * Add a new function SD_task_set_amount() upon user request.
2005 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
2006 * Remove rule based routing (no more PCRE dependency)
2007 * Add a limiter_link option to cluster tag, to specify a maximum reachable
2008 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
2009 * Add a loopback_bw and loopback_lat options to cluster tag.
2010 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
2011 routing. Now peers are encapsulated in an AS and have their own private
2012 router but this is transparent.
2015 * Our own implementation of getline is renamed xbt_getline, and gets
2016 used even if the OS provide a getline(). This should reduce the
2017 configuration complexity by using the same code on all platforms.
2018 * New type: xbt_cfg_elm_boolean.
2019 * Allow to use yes/no for boolean configuration options in the command line.
2020 * Allow to disable SimGrid cleanups at exit from command line option.
2021 There are situations where one may want a simulation to end with an exit.
2022 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
2023 annoying when scripting around the simulator. Adding a
2024 --cfg=clean_atexit:no allows to circumvent this issue.
2027 * Lots of memory leaks were corrected in this release.
2028 There is no known memory leaks anymore, in all of our 600+ tests.
2029 * New command line option --version, to get SimGrid version information.
2030 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
2032 * Supernovae builds are deprecated, and expected to be removed in the next
2035 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2037 SimGrid (3.9) stable; urgency=low
2039 The Grasgory release: GRAS is really dead now.
2041 * Complete overhaul of the internal host structures scheme.
2044 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
2045 was considered as experimental and badly maintained since then.
2046 * Keeping it was thus a trap to our potential users, that could take
2047 it instead of MSG or SMPI by mistake despite is pity state.
2048 * GRAS seems to have very few users (if any), and no one volunteered
2049 to maintain it further. It also induces a lot of XBT code (for
2050 portability sake), that must be maintained too.
2051 * For all these reasons, we killed GRAS. If someone wants to revive it
2052 in the future, don't cry, our git history still remembers of GRAS.
2055 * Major overhaul. Merge our documentation again as time proved that
2056 splitting it was really not helping our users.
2057 * Further improve the developer documentation to help newcomers
2058 hacking on SimGrid itself. The user documentation (and in
2059 particular, the beginner documentation) is still in a sorry state.
2062 * Now works on Windows too!
2063 * Much more extensive test suite, from MPICH
2066 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
2067 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
2068 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
2069 * Bug fix: let task be scheduled when the last dependency to be solved is
2070 a control dependency.
2071 * Remove SD_load_environment_script function.
2072 Use the C sg_platf function if you want to declare a platform
2076 * New function: MSG_process_get_number()
2077 * Old function documented: MSG_config()
2078 * Remove MSG_load_platform_script function
2079 Use the C sg_platf function if you want to declare a platform
2083 * Change the default value of the TCP_gamma constant (maximal size of TCP
2084 congestion window) to a more realistic 4MiB value. If you notice changes in
2085 your simulation results, you can fall back to the previous 20k tiny window
2086 by adding --cfg=network/TCP_gamma:20000 on command line.
2087 * (Hopefully) fix a bug wrt periodic availability and state traces
2088 * Bug fix: use default values at start when first event in availability/state
2089 trace is not at time 0.
2092 * remove the "new_" part of function name sg_platf_new_trace_connect
2093 (resulting in sg_platf_trace_connect), since it does not create
2097 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2098 use case where it's really mandatory, and maintaining it was a pain in
2100 * New: xbt_fifo_search(), search an item with a user-provided
2101 comparison function instead of dumb pointer comparison.
2104 * Fix the lua deployment:
2105 Use `simgrid.init_application()` before deployment instead of
2106 `simgrid.msg_register_application()` after.
2109 * Transfer the tracing files into the corresponding modules.
2111 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2113 SimGrid (3.8.1) stable; urgency=low
2115 The "we are told that some people want to also *install* the simgrid
2118 * Add missing file "tesh.1" to the archive.
2120 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2122 SimGrid (3.8) stable; urgency=low
2124 The Psssshiiiit release: SimGrid jumps into the Cloud.
2127 * Add an experimental interface to manipulate VMs. They are mainly
2128 process groups with very few intrinsic semantic, but they should
2129 allow you to build the semantic you want easily.
2130 * New function: MSG_host_set_property_value()
2131 * New function: MSG_process_on_exit(). To clean memory in all cases.
2132 * Bug fixes that made the host (and link) failures unusable.
2133 * Add a way to auto-restart process when the host in which they are
2134 executing comes back (ON_FAILURE="RESTART" on deployment file,
2135 MSG_process_auto_restart_set).
2136 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2137 please stop using the old ones, they are DEPRECATED.
2139 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2140 Please use MSG_init() instead. (reducing the amount of entry
2141 points in the library helps us).
2142 * Make it impossible to link against the wrong version of the lib
2143 * Deprecate MSG_clean(). No need to call it anymore.
2144 * Function MSG_get_host_number() is not deprecated anymore.
2147 * Split the doc into a user guide and a reference guide.
2148 * Start a developper guide to help people hacking on SimGrid.
2151 * Enable tracing by default. This modules rocks you should use it.
2152 * Remove option custom_flags. Now use environment variables CFLAGS
2154 * Use default cmake things to detect lua instead of home grown ones.
2155 * New option "enable_mallocators" to disable mallocators, for debugging
2156 purpose ("on" by default).
2159 * Bug fixes around the resource failures: don't let the processes
2160 survive the host they are running onto.
2161 * Add an interface to auto-restart processes when the host in which they are
2162 executing comes back.
2163 * Ensures that SIMIX_clean is called automatically. It's not part of
2164 the public interface anymore (bindings should be updated).
2167 * Bug fix for when SD_Simulate is called with a positive value: be careful
2168 when comparing doubles. Sometimes they are different for non significant
2170 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2171 parallel task whose initial work is distributed among host according
2172 to the Amdahl's law. Such tasks are created with a parameter alpha
2173 that corresponds to the non-parallelizable part of the computation.
2174 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2175 between two sets of workstations assuming a 1D block distribution (each
2176 workstation owns a similar share of data) on both sides.
2178 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2179 Data redistribution will be automatically scheduled once parent and child
2180 are both scheduled. The filling of computation_amount and
2181 communication_amount structures is now done seamlessly thanks to the chosen
2183 * New function SD_workstation_dump to display various information
2184 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2185 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2186 bandwidth on the route onto which the task is scheduled and the amount of
2188 To divide the nominal bandwidth by 2, the rate then has to be:
2189 rate = bandwidth/(2*amount)
2190 * Compute tasks that have failed can now be rescheduled and executed again
2191 (from their beginning)
2192 * Increasing source code coverage (src/simdag is now covered at 95.8%
2196 * Re-implement time-independent trace replay using SMPI (at the
2197 smpi_smp_* level) instead of MSG. This should replace
2198 examples/msg/actions/actions.c
2199 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2200 hindexed and structs)
2201 * Implement the exchange of non-contiguous data.
2202 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2203 * Correct behavior of smpi/sender_gap and set its default value to 0
2204 * Add option to asynchronously send small messages to allow better
2205 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2206 specifies the size in bytes under which messages will be asynchronously sent.
2207 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2208 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2211 * New C interface to define a platform: XML is now optional.
2212 For more info, please check include/simgrid/platf.h.
2213 * New interface to define random platforms from the C:
2214 For more info, please check include/simgrid/platf_generator.h and
2215 examples/msg/masterslave/masterslave_platfgen.c
2216 * Export a sg_cmdline dynar containing all the arguments we got from
2220 * Two new tracing options for adding comments to trace file so you
2221 can track your experiments (see --help-tracing for details).
2222 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2223 * Adding the SimGrid version that generated the trace file as a comment.
2224 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2225 * Fix to avoid key clashes on Paje links
2226 * Other minor fixes related to the Paje specification
2229 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2230 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2231 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2232 * Greatly improve the robustness of mmalloc to user errors (such as
2233 using an area after freeing it, or freeing it twice)
2235 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2237 SimGrid-java (3.8.1) stable; urgency=low
2239 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2240 Warning: all methods are visible, but only some of them are
2241 implemented so far. Check the source (src/jtrace.c)
2242 for further information.
2243 * New module: org.simgrid.msg.File (SimGrid File management functions)
2244 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2245 * Change the meaning of Process.restart: now restart the process from
2246 the beginning, like MSG_process_restart in C.
2247 * Add Process.setAutoRestart: handling of process restart when failed
2249 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2250 you to retrieve the properties of the processes/hosts
2251 * Deprecate Msg.clean(): you can just forget about it now.
2252 * New function Process.getCount(), that only works when compiling
2253 with the not yet released version 3.9 of the C library.
2255 * New context factory based on Coroutines. It mandates a modified JVM
2256 but then, the simulations run about five times faster, and there is
2257 no limit to the amount of processes (beside of the available memory).
2259 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2261 SimGrid (3.7.1) stable; urgency=low
2264 * Restore the prototype of MSG_process_create_with_environment() to
2265 the pre-3.7 situation by removing the kill_time argument.
2266 * Add a MSG_process_set_kill_time() function instead.
2269 * Fix weird behaviors when dealing with parallel tasks.
2272 * Simgrid is now built as a dll.
2273 * Simgrid-java now works on Windows.
2274 * Simgrid-Java is now included into Windows package.
2277 * First pre-build package for MacOSX.
2280 * Fix compilation when using MSG_USE_DEPRECATED.
2281 * Fix some compilation issues on Macs and Windows.
2282 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2283 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2285 We discovered that the Lua console is broken, but we are missing the
2286 manpower to fix it right now. The problem existed in 3.7 too, so we
2287 are not blocking the release for that. Sorry if you depended on this
2288 feature, any help would be really welcome.
2290 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2293 SimGrid-java (3.7.1) stable; urgency=low
2295 The "Java aint got to be bloated and slow" release
2298 * Various internal cleanups and performance improvement
2299 Simulations are expected to run up to twice faster or so
2300 * Make Process.kill(process) an instance method, not a static one
2301 * User processes are not java.lang.Thread subclasses.
2302 This breaks the compatibility (sorry), but previous API was
2303 brain-dead, making it impossible to have non-trivial
2304 initializations in the process constructor.
2305 * Require a full constructor per Process sub-class.
2306 Kinda breaks the compatibility (sorry), but this allows a much more
2307 efficient way to launch the processes at simulation startup.
2308 * Do not embeed our version of semaphores, java 1.5 can be considered
2309 as sufficiently prevalent for us to not dupplicate its features.
2313 * Add examples for almost every part of the API
2314 We spotted and fixed a lot of bugs in the process
2316 * New module: asynchronous communication API
2317 * New function: Process.sleep()
2318 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2319 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2320 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2322 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2324 SimGrid (3.7) stable; urgency=low
2326 The "spring cleanups (before the next Big Project kicks in)" release.
2329 * We can specify the SMPI latency/bandwidth factor with command line
2330 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2331 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2332 You can also use the "config tag" from platform file by adding this line
2333 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2334 (see "example/platforms/tag_config.xml" to use "config tag").
2335 Note that the command line supersedes the platform file configuration.
2336 * Change the correction factors used in LMM model, according to
2337 the latest experiments described in INRIA RR-7821.
2338 Accuracy should be improved this way.
2339 * Use the partial invalidation optimization by default for the
2340 network too. Should produce the exact same results, only faster.
2341 * Major cleanup in surf to merge models and split some optimization
2342 mechanisms from the core of the models. As a result you can now
2343 specify which model to use (e.g., --cfg=network/model:LV08
2344 --cfg=cpu/model:Cas01) and which optimization mode to use
2345 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2346 Incompatible combinations should err at initialization. See
2347 --help-models for the list of all models and optimization modes.
2348 * The CLM03 workstation model was dropped for simplicity because it
2349 used the deprecated CM02 network model. Use default instead.
2350 * Rename the TCP_gamma configuration option to network/TCP_gamma
2351 * Rename the coordinates configuration option to
2352 network/coordinates, and document it
2353 * Use now crosstraffic keyword instead of the terribly misleading
2354 fullduplex keyword. It is activated by default now in the current
2355 default model, use --cfg=network/crosstraffic:0 to turn it off.
2356 * Ongoing refactoring the model parsing to make XML files optional
2357 See include/simgrid/platf.h for details (still to be completed)
2360 * Major overhaul of the documentation. Almost instructive now :/
2361 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2362 functions and friends. This interface was considered as
2363 deprecated since over 2 years, it's time to inform our users that it is.
2364 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2365 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2366 use these (crappy) functions in your code.
2367 These functions will be removed soon. Stop using them now.
2368 * Deprecate MSG_get_host_{table,number}
2369 Implement MSG_hosts_as_dynar() instead.
2370 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2371 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2374 * Stabilize the parallel execution mode of user contexts
2375 * Introduce configuration variables to control parallel execution:
2376 - contexts/synchro: Synchronization mode to use when running
2377 contexts in parallel (either futex, posix or busy_wait)
2378 - contexts/parallel_threshold: Minimal number of user contexts
2379 that must be part of a scheduling round to switch to parallel
2380 execution mode (raw contexts only)
2381 * Fix bugs that prevented to use suspend/resume along with
2382 synchronization structures.
2383 * Fix bugs in process termination that lead to invalid memory access
2384 in very specific conditions.
2387 * Introduce a parallel mode for the models (controlled by surf/nthreads
2388 configuration item). In our tests, running the models in parallel
2389 never lead to any speedups because they are so fast that the gain
2390 of computing each model in parallel does not amortizes the
2391 synchronization costs, even when ultra fast futexes are used.
2392 This is released anyway because YMMV.
2395 * Performance boost by using a swag internally to compute the set of
2396 tasks that are finished and should constitute the return value of
2400 * Enable it by default now that it is considered rather stable.
2403 * Documentation of the tracing functions.
2404 * Performance gains when tracing categorized/uncategorized resource
2405 utilization by avoiding calls to get route when updating resource
2406 variables. LMM constraints are being used instead.
2407 * API changed to set task categories. Use MSG_task_set_category instead
2408 of TRACE_msg_set_task_category, and SD_task_set_category instead
2409 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2410 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2411 * Examples for link user variables added (see at examples/msg/tracing/)
2412 * Deprecated function TRACE_msg_set_process_category completely removed
2413 * Trace header updated according to the latest Paje file format
2414 * Tracing network lazy updates, no longer obligate users to use full updates
2415 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2416 * Experimental: let user code declare/set/push/pop application states for hosts
2417 * API changed to allow the manual creation of graph configuration files
2418 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2421 * Improve the API of Lua MSG bindings, using the Lua spirit.
2422 * Each simulated process now lives in its own Lua world (globals are
2423 automatically duplicated). It helps writing simulators. Will allow
2424 to run Splay programs within SimGrid in the future.
2425 * Add a Chord example in Lua, equivalent to the MSG one.
2428 * Start the implementation of a solution to express temporal
2429 properties, not only local assertions. This is still an
2430 experimental work in progress, stay clear from it to be safe.
2434 - Add new runtime parameters --help-logs and --help-log-categories
2435 to display information about supported logging parameters and
2437 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2438 don't exists anymore.
2439 * Mallocators: allow value NULL for the reset function.
2441 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2442 create homogeneous dictionaries, where all the elements share the
2443 same free function. Non homogeneous dictionaries will be
2444 deprecated in the next release.
2445 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2446 - Multi-level dictionaries are deprecated.
2448 - new function xbt_dynar_search_or_negative() that is useful when
2449 you have less than 2 million elements in your dynar and don't
2450 want of the extra complexity of catching exceptions when the
2451 element is not found.
2453 - Make xbt_os_thread module (for thread portability) public.
2454 Documentation is still to come, sorry.
2456 - Cleanups and simplifications to make it maintainable again.
2457 - Exotic features (such as memalign and valloc) were removed.
2458 - The metadata were extended and improved so that the
2459 model-checker becomes able to explore and inspect the heaps.
2460 - This may induce a performance drop when enable_model-checking is
2461 ON in cmake (even if it's not used in the simulation), but it is
2462 necessary at this point to get MC working.
2464 Turn model-checking OFF if simulation performance matters to you.
2465 Not enabling it at runtime is not enough, disable it in cmake.
2467 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2470 SimGrid (3.6.2) stable; urgency=low
2472 The "Not coding new stuff allows to polish old things" release.
2475 * New bindings to the NS3 packet level simulator (experimental)
2476 * Use the raw (efficient) execution contexts instead of the sysv
2477 (portable) ones when possible.
2478 * libpcre is now mandatory in any cases since not using it led to
2479 severe performance loss and possibly other issues
2480 * Update the XML platforms:
2481 - G5K: include the latest machine in Nancy
2482 - GridPP and LCG: new platforms
2483 * Documentation was partially updated, at least (more to come)
2485 Bug fixes, cosmetics and small improvements
2486 * Free terminated processes before the end of the simulation to avoid
2487 exhausting the memory of users having very dynamic amount of
2489 * Bug fix and cosmetics about canceling non-running tasks
2490 * Bug fix about the dot loader's issues when using libcgraph
2493 * Create an installer for windows with nsis (amd64 and win32)
2494 - Add an hello world project to illustrate simgrid project creation.
2495 - Embed libpcre into the Simgrid installer to avoid
2496 its compilation burden
2497 * The raw execution contexts should work on Apple now
2498 * Port to Windows 64 bits
2499 - Sysv contexts now have an implementation for this arch
2500 - GRAS communication features now support this arch
2501 * Drop support for borland compiler on windows
2502 - this code was not maintained, and we kinda depend on gcc nowadays
2503 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2504 * Fix portability issue on unstable ubuntu: linker became picky on
2507 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2509 SimGrid (3.6.1) stable; urgency=low
2511 The "Oops, we broke Macs too" release
2514 * Fixed contexts detection so that raw ones are used when possible
2515 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2516 leads to a strange error, with user code segfaulting sometimes when
2517 the generated code is not perfectly aligned (which is not
2518 controllable from the user side, depends on the amount of code)
2521 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2523 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2525 SimGrid-java (3.6) unstable; urgency=low
2528 * Split of every thing from simgrid v3.5 into a separate package.
2530 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2532 SimGrid (3.6) stable; urgency=medium
2534 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2537 * Bindings now constitute their own package, separated from the main one.
2538 Rationale: reduce our maintenance nightmare by reducing the module coupling
2539 They will soon be released on their own on gforge.
2541 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2542 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2544 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2545 * It's not quite deprecated for now because we have no replacement,
2546 but it may soon become the case.
2549 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2550 * Fortran: New user-level cache variable to store the rank of the running
2551 process. This improves performance by an order of magnitude.
2552 * C: New coccinelle script to automatically locate and modify global and
2553 local static variables.
2554 * Improved SMPI network model with a sender-side gap to account for multiple
2558 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2559 only say if a communication is finished, no matter whether it succeeded or
2560 failed. You can call MSG_comm_get_status() to know the status of a finished
2562 * New function MSG_task_dsend() to send a task and detach it. When a
2563 communication is detached, you are never notified of its success or failure
2564 and the memory is released automatically once it is finished. This function
2565 is useful when you don't care about the end nor the success of a
2567 * Change the prototypes of action replay. Sorry for inconvenience,
2568 but this is really more efficient this way (and to adapt your code,
2569 you just have to fix the initialization, that shouldn't be too long)
2570 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2571 ever managed to do anything useful with it.
2572 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2573 takes a dynar of communications. It returns immediately and gives the
2574 index of a finished communication (if any).
2575 * New example: a basic implementation of the Chord P2P algorithm.
2578 * New model for multi-core CPUs. You can now use the core attribute to
2579 precise the number of cores of a host. This is a basic model. Every
2580 process running on the host receives at most the power provided in
2581 the DTD (throughput<=power). Total throughput of process cannot exceed
2583 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2584 router linked by an up-link and a down-link (possibly asymmetrical).
2585 This kind of pattern allows to easily build last-mile model style platforms.
2586 Aggregating such patterns in a rule-based AS is thus the technique of
2587 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2588 * New model for Vivaldi routing. We transformed the Vivaldi network model
2589 into a Vivaldi routing model (based on the rule-based model). This allows to
2590 combine Vivaldi based latencies with last-mile platforms.
2593 * Added a check for NaN of IEEE754 infinite in the double entries of
2595 * Introduce a new context factory "raw", highly inspirated from the
2596 ucontext factory, but using manually crafted functions in assembly to
2597 do the work in an efficient manner.
2598 * Allow to change the used context factory at run time, not only at
2599 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2600 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2601 contexts at runtime (only with raw contexts or ucontexts).
2602 * Completely rewrote this module to allow parallel execution of user
2603 processes. Use --cfg=contexts/nthreads:N to execute user processes
2604 with N parallel threads (the default is 1, meaning no parallelism).
2605 * Allow to decide dynamically between sequential and parallel modes.
2606 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2607 processes in parallel only when their number is greater than or equal to P
2609 * Added a check for NaN of IEEE754 infinite in the double entries of
2613 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2614 won't output the state of processes when interrupted with Ctrl-C
2615 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2616 NULL-terminated array. This may solve backward compatibility issues
2617 due to the change to return type of SD_simulate. See also:
2618 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2619 * Add new macros with variable number of arguments.
2620 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2621 - in xbt/asserts.h: xbt_assert
2622 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2623 - in xbt/ex.h: THROWF and RETHROWF.
2624 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2626 * Change xbt_die() to accept a format string with arguments, just like printf.
2627 * New data structure: xbt_lib_t, like a dict but more general and with better
2631 * New configuration options
2632 Options triva/categorized and triva/uncategorized can be used to generate
2633 graph configuration files for Triva visualization tool.
2634 * Configuration option tracing/platform is renamed to tracing/categorized
2635 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2636 * New TRACE_user_link_variable interface:
2637 User provides the name of the link and the tracing variable to attach to it
2638 * the declaration of tracing categories must be done after the environment
2640 * simpler tracing interface, just one way to declare categories
2641 TRACE_category or TRACE_category_with_color, it is up to you
2642 * links in the trace file are again identified by their names
2643 * trace contains the full platform hierarchy exactly as declared using the ASes
2644 * Options tracing/msg/[task|process]:1 groups the process by hosts
2645 for both cases, tasks and processes must have names that are unique during the simulation
2646 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2647 * The experimental option tracing/msg/volume is deprecated
2648 its functionality may be reincorporated if needed
2650 The tracing generates a trace file with unordered timestamped events,
2651 because of the way the core simulator (surf) works. A script available
2652 at the tools directory (fix-paje-trace.sh) can be used to put the events
2653 in order. We have changed the tracing so it can generate ordered timestamped
2654 events in the final trace, but depending on the simulator (and how much time
2655 is simulated) that can lead to a huge memory utilization. It is deactivated
2656 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2658 Build Infrastructure
2659 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2660 - We are trying hard to keep the API stable, but it may happen that
2661 some things change (we're a research project after all, not a
2662 nuclear plant operating system). If such things should happen, you
2663 could rely on that macro to adapt.
2664 - current value: 30600 for 3.06.00, aka 3.6
2665 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2666 a number that can be compared with SIMGRID_VERSION.
2667 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2668 assertions and verbose/debug logging events are disabled at compile time.
2670 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2672 SimGrid (3.5) stable; urgency=medium
2675 * New feature to any SimGrid-based simulator: Model-Checking
2676 Check SIN#1 for more details.
2679 * New Model SMPI (three-interval linear regression for correction factors)
2680 See RR-7426, available at http://hal.inria.fr/inria-00527150
2681 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2682 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2683 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2684 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2685 * See also src/smpi/README
2690 - Tracing API changes: TRACE_start and TRACE_end should not be called
2691 by user-code. They are automatically called by simulators created
2692 with SimDAG, MSG and SMPI if the toolkit is compiled with
2693 tracing_enabled=ON. Categories declaration and utilization remain the
2694 same for MSG and SimDag.
2695 - A function was added to the tracing API to declare categories with
2697 - TRACE_category_with_color (char *category, char *color)
2698 where color must be in the following format
2699 "%f %f %f", red, green, blue
2700 and red, green, blue are float values in the interval [0, 1]
2701 - User can specify NULL as color parameter, or continue calling
2702 TRACE_category (cat)
2703 On that case, the tracing system will define random colors
2704 - The following command-line options are supported:
2705 --cfg=tracing/filename:msg.trace
2706 --cfg=tracing:1 (activate tracing, needed to use others)
2707 --cfg=tracing/platform:1 (categorized resource use)
2708 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2709 --cfg=tracing/msg/task:1 (task creation)
2710 --cfg=tracing/msg/process:1 (process creation, migration)
2711 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2712 --cfg=tracing/smpi:1 (SMPI interface tracing)
2713 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2714 - examples of examples/msg/tracing updated
2716 - DAXLoader and DOTLoader functions can generate tasks with categories
2717 - A new function to attribute a category to SD tasks:
2718 TRACE_sd_set_task_category (SD_task_t task, char *category)
2719 * Tracing the MPI interface implemented by SMPI
2720 - Collective operations are traced with states
2721 - Point-to-Point operations are traced with states/links
2722 - Tracing activated by a parameter "-trace filename" passed
2723 to smpirun during execution (considering that simgrid
2724 is compiled with tracing enabled)
2725 - To run the simulation with gdb, the simulator
2726 accepts --cfg=tracing/smpi:1 to trace SMPI
2727 - tesh files to check if smpi tracing is ok
2728 - See examples/smpi/NAS/DT-trace
2729 * GTNetS tracing re-worked
2730 - adaptation to the tracing system of GTNets to cope
2731 with modifications regarding the fullduplex mode
2732 - new tesh files to check if gtnets tracing is ok
2735 * Asynchronous communications through the functions:
2736 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2737 * New function: MSG_load_platform_script()
2738 to make possible using a lua script instead of XML files to set up platforms
2739 * New function: MSG_set_function
2740 to associate functions to processes, used when bypassing the parser
2741 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2743 Platforms: Add some more examples in examples/platforms
2744 * Grid'5000: see www.grid5000.fr
2745 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2748 * Change the XML format. This is a very important modification. SimGrid
2749 3.5 introduces a new hierarchical format based on the notion of
2750 Autonomous Systems. Compatibility with old format is ensured
2751 through the perl script provided in the install bin directory
2752 bin/simgrid_update_xml.
2753 It is now possible to build platforms with specific routing
2754 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2755 connect several platforms together. We will try to provide soon
2756 set of realistic platforms exploiting these properties (have a
2757 look at examples/platforms/ for the moment).
2758 * Take the opportunity of the XML format change to be a good XML citizen:
2759 rename link:ctn to link_ctn and similar changes (also dealt with by
2761 * Add a new routing scheme (rule-based) using regular expressions. It
2762 enables to have an extremely low memory footprint when the
2763 underlying routing is simple and can be compactly described. You
2764 need to have libpcre4-dev (perl regular expressions) installed if
2765 you want to use this routing scheme.
2766 * Revive the cluster TAG and allow to easily and efficiently (both in
2767 term of memory and speed) connect clusters together. Have a look
2768 at teshsuite/simdag/platforms/ to see how this can be done. With
2769 this tag, you can create clusters with thousands of tasks at no
2770 cost (have a look at examples/platforms/).
2771 Note: clusters are implemented as ASes, so there is no need for an
2772 enclosing AS tag if you have only one cluster in your platform.
2773 * Add new generic functions in the public interface that allows the user
2774 to call SURF 'create_resource' methods from your code (same
2775 functionality as the XML bypass mechanism but with a much lighter
2777 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2778 that uses a piecewise linear approximation to produce better
2779 results when exchanging small messages.
2780 * Add a new parameter to handle correctly full duplex link and account
2781 for interferences between uplink and downlink communications
2782 (activate with --cfg=fullduplex:1).
2785 * Rename the SD_READY (all dependencies are satisfied and task is
2786 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2787 dependencies are satisfied) state.
2788 This prevents a confusion between the notion of "ready to schedule"
2789 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2790 simulated" (SD_RUNNABLE) used by the simulation kernel.
2791 * Change the way a task is considered as ready. Instead of removing
2792 dependencies when a task is done, a counter is decreased. This way,
2793 it is always possible to reach ancestors thanks to the
2794 SD_taks_get_parents function (even after the end of the simulation.)
2795 * Change the return type of SD_Simulate from (SD_task_t*) into
2796 xbt_dynar_t. This function was in handling a dynar internally and
2797 converted it into a NULL terminated array for historical reasons.
2798 * New function SD_dotload(char*) to load a DAG described in dot
2799 format. This loader and the corresponding examples require the
2800 installation of the graphviz library.
2801 * Fix a bug in the management of tasks of size 0 in the surf network
2802 models. This problem was only visible with SIMDAG and you should
2803 thus disregard results produced with earlier versions if you
2804 relied on this feature (some tasks were blocked because of this).
2805 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2806 even though your applications were doing only point-to-point
2807 communications and sequential computations. Now you can really use any
2808 model you want (of course, if you create real parallel tasks, which are
2809 not implemented in most models beside ptaskL07, this will abort).
2810 * Add an example that schedules a DAX on an heterogeneous platform
2811 using a Min-Min strategy.
2812 * New function SD_workstation_get_current_task() that returns the kind
2813 of task currently running on a workstation in the sequential access
2815 * Raise some warnings when unexecuted tasks remains at the end of the
2816 simulation. This is usually caused by cycles in the DAG.
2819 * New function: SIMIX_process_set_function() called by MSG_set_function
2820 * Change the underlying waiting queue in semaphores so that a process
2821 can wait on several of them simultaneously (as in waitany).
2822 * Fix the way to handle tokens in semaphores so that all access patterns
2823 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2824 * kill the dirty pimple SIMIX_message_sizes_output()
2825 Please use (proper) visualization instead
2828 * New data container: setset (set of sets of elements)
2829 * New module: mmalloc (mapped malloc, allowing to have several
2830 independent segments of malloc)
2831 * New function: xbt_dict_cursor_set_data()
2832 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2833 * New function: xbt_dynar_is_empty()
2834 * New function: xbt_fifo_get_last_item()
2835 * Fix xbt_dynar_shrink(): use the right element size.
2836 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2837 uninitialized areas during expand.
2838 * Fix semaphores: previous implementation was severely broken.
2839 * Use library init/fini functions for our initialization.
2840 - you can use logs and other feature as soon as you want in your
2841 code (even before the xbt_init / MSG_init)
2842 - xbt_exit is now a no-op and produce a warning when used.
2845 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2846 benefit from the latest improvement to the simulation kernel.
2847 * Kill measurement sockets for now. If you rely on them, sorry. This
2848 release is not for you. This feature will be reintroduced in the
2849 future, but we cannot delay the release any further.
2850 * New function: gras_msgtype_get_name().
2851 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2852 * Fix (at last) the pmm example: it should not randomly fail anymore.
2854 Build chain: bug fixes and overall polishing
2855 * Cmake is now stable enough. Hence, we killed the autotools.
2856 * Port to windows ( TM :)
2857 * Fix the 'make install' target.
2858 No need to use 'make install-simgrid' anymore
2859 * Introduce a 'make dist' target compiling a *source* archive
2860 'make package' compiles a binary archive
2861 * Compile java files only on need
2862 * Add --cd and --setenv command line options to tesh
2863 * Out of source builds are not fully supported yet, but we are close
2864 * Enable supernovae and optimization flags by default for our users
2867 * Add layer to set up environment directly from lua, without XML.
2868 * The effect of gras_stub_generator can be achieved through
2869 lua too (check examples/gras/console/ping_generator.lua)
2871 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2873 SimGrid (3.4.1) stable; urgency=low
2875 The "Polishing easter eggs is probably a good idea" release.
2876 This is a bug fixes release only.
2880 * Fix a bug preventing the tasks from begin garbage collected.
2883 * Fix a bug occurring when a host involved in a communication fails.
2884 This was not detected properly by the other peer involved in the
2885 communication. Now, it's reported as a network error.
2888 * Warn the user about loop dependencies in data flow of DAX files
2889 * Obey the control-flow dependencies of DAX files
2892 * Add option "enable_smpi" allowing to not compile SMPI.
2893 Probably useful for the (Mac) users experiencing a build error here
2894 * Improve the detection of lua5.1 and ruby1.8
2896 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2898 SimGrid (3.4) stable; urgency=low
2900 The "Easter in Cargese" release. Also known as (major changes):
2902 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2903 ~> bindings were greatly improved
2904 ~> new tracing infrastructure for better visualization introduced
2906 * the "Welcome to configury modernity" release.
2907 ~> we switched from autotools to cmake, and improved our cdash
2910 A more detailled list of changes follow (full detail in svn log).
2912 Java Bindings: Various Cleanups
2913 * (install java-gcj-compat-dev on debian-like to use them)
2914 * Remove put/get: no need to export deprecated interface in Java
2915 Use send/receive instead.
2916 * Cleanup the examples and add a README per directory
2917 * Remove example autoDestination (that's the only way to go now)
2918 * Remove example explicitDestination (was a plain copy of basic)
2919 * Make JniException a runtime exception, so that there is no need to
2920 declare the fact that you may encounter such a beast. I guess that
2921 nobody will ever want to survive such error.
2922 * Create specific errors for each MSG case of failure:
2923 host failure, transfer failure, timeout, task cancelled
2924 * Cleanup the exceptions that may get thrown by each function
2925 * Other internal cleanups in Java bindings. Performance still bad :/
2926 Ruby and Lua Bindings: create them
2927 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2928 * That's new and great, you should try them out.
2929 Same functionalities than Java bindings, only even less polished
2931 * Kill the useless "rate" argument of SD_task_get_execution_time()
2932 Everyone used to provide -1 as a value, it was not used, and the
2933 semantic of a possible use wasn't even clear.
2934 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2935 either as comm costs or compute costs to mean that there is no
2936 such thing for that specific task.
2937 * Add a SD_task_set_name() function
2938 * Fix SD_task_unschedule() on typed tasks
2939 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2940 * In DAX loader, accept useless 'level' attributes to <job> since
2941 LIGO DAGs have them (seem to be to ease graphical representation).
2943 * Add an example masterslave_mailbox.c using send/receive and not
2944 the deprecated put/get interface.
2945 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2946 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2948 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2949 (declare MSG_USE_DEPRECATED to still have the old name)
2950 * Add a MSG_task_set_data() function
2951 * About trace replay (see examples/msg/actions):
2953 - Allow to work with splitted trace files for each process
2954 Give the specific trace file as argument of each process,
2955 and call MSG_action_trace_run(NULL)
2956 You can still have one merged file for all processes.
2957 - Fix implementation of collective operations
2958 * Allow task_execute() on 0-sized tasks (closes #10063)
2960 * This is the first release of SimGrid where SMPI is not considered
2961 beta anymore (even if some corners should still be improved)
2962 * Port over the new SIMIX_network submodule (internal refactoring)
2963 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2964 * Implement more missing elements of the standard:
2966 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2968 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2969 MPI_Scan() MPI_Get_processor_name()
2970 - Added implementation of missing case for Alltoall (warning: it's
2971 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2972 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2973 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2974 SMPI_MPI_Allgatherv()
2975 * Bug fixes include:
2976 - MPI_Waitsome() was broken
2977 - Allow relative includes in smpicc
2978 - Command line cfg argument 'reference_speed' was ignored...
2979 - Some functions did not properly lead to auto-benching of user code
2980 - smpicc passes -O2 by default (just like openmpi one)
2982 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2983 * Bug fixes about timeouts during communications
2984 * add SIMIX_message_sizes_output() as a pimple to write to file the
2985 amount of messages per size. Use gnuplot to get histogram.
2986 Pimple because that's the only user-visible function of simix,
2987 defined directly in xbt.h (irk, sorry)
2989 - Add a SIMIX_sem_get_capacity() function
2990 - Fix interactions with processe resume/suspende
2991 - release_forever() was stupidly broken
2992 - Fix SIMIX_display_process_status() for processes in a semaphore
2993 - Make SIMIX_sem_block_onto() user-visible
2994 * Refactoring context stuff:
2995 - Use pseudo-OOP for better modularity
2996 - reimplement SIMIX_process_kill() without process_schedule() so
2997 that the latter can take as invariant that it is called from
2999 - Merge context_start into context_new for sake of simplicity
3001 * Add a Vivaldi network model, coded live during SUD'10 ;)
3002 * Rename configuration variables to start a hierarchy:
3003 o cpu_model -> cpu/model
3004 o network_model -> network/model
3005 o workstation_model -> workstation/model
3006 * New configuration variables:
3007 o network/bandwidth_factor: correction to bandwidth
3008 o network/latency_factor: correction to latency
3009 o netwotk/weight_S: correction to the weight of competing streams
3010 * Add a long description to the models, that users can see with such
3011 argument on the command line: --cfg=cpu/model:help
3012 * --help-models display the long description of all known models
3014 * config: add the ability to set a default value after registration
3015 Does not override any previously set value (e.g. from cmd line)
3016 * dict: allow to have integer key and data.
3017 When so, you need to use the following functions
3018 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
3019 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
3020 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
3021 In contrary to regular dicts, the key is not malloced before copy.
3022 Mixing scalar and regular elements in the same dict is not tested
3024 * Allow to use xbt_dynar_shrink() to expend the dynar instead
3025 Tracing for Visualization:
3026 * SimGrid is now instrumented in order to generate a trace file for
3027 visualization analysis: to use it, need to compile SimGrid with the
3028 "tracing" option enabled, and instrument the program using SimGrid with
3029 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
3030 (among other functions).
3031 * The instrumentation only traces the platform utilization for now
3032 * Documentation to use the tracing functions and how to analyze the
3033 traces with the Triva tool is written.
3034 * More information about: SimGrid FAQ (in the section Tracing Simulations
3037 * We moved to cmake as default build system. Autotools support will
3038 be dropped soon. Check the FAQ for more info about how to use it.
3039 * Greatly improved our cdash/ctest interactions
3040 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
3041 * Added memory checking tests with valgrind; lot of memleak fixing.
3042 This may be the first release of simgrid with so few memory issues
3043 * Added code coverage tests.
3044 Our coverage is still improvable, but at least we see it on cdash.
3046 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
3048 SimGrid (3.3.4) stable; urgency=low
3050 The "Desktop Grid needs love too" release (also called Xmas release).
3052 Models improvements:
3053 * Major speedup in the maxmin system solving by using lazy evaluation
3054 Instead of solving completely the maxmin system at each iteration,
3055 only invalidate (and recompute) the modified parts.
3056 This new feature is enabled in default models but you can try to
3057 turn it on with "--cfg:maxmin-selective-update=1" for other models.
3058 * Cas01 IMproved as default CPU model
3059 This CPU model is the same Cas01 model, but it uses the
3060 maxmin-selective-update flag and a heap structure to manage
3061 actions on SURF kernel.
3062 It reduces the complexity to find the next action to finish and,
3063 consequently, it's faster than the old Cas01.
3064 This is the new default CPU model (Cas01).
3065 * Rename the old Cas01 model to Cas01_fullupdate
3066 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
3067 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
3068 * CpuTI (CPU Trace Integration)
3069 A new CPU model whose objective is simulate faster when using
3070 availability trace files.
3071 Instead of using a full featured, over engineered maxmin system for
3072 CPU modeling, this model does the pre-integration of traces files
3073 to calculate the amount of CPU power available, and so, executes
3074 faster than the old CPU models.
3075 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
3076 * Use LV08 as default network model since it gives better accuracy
3077 for small messages and shouldn't change things for big ones.
3078 Use --cfg=network_model:CM02 to get the previous behavior.
3081 ******************************************
3082 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
3083 ******************************************
3084 * The new CPU model may changes simulations!
3085 The point is that events occurring at the exact same timestamp
3086 are not scheduled in the same order with the old and new
3087 version. This may be enough to completely change the execution
3088 of simulations in some cases.
3089 * The new network model will change simulations!
3090 This new model is more realistic than the previous one, so you
3091 should consider redoing your old experiments with this model.
3092 Sorry for the inconvenience.
3095 * Introduce the supernovae compilation mode
3096 When compiled that way, the whole SimGrid (or almost) is put in a
3097 single compilation unit and compiled in one shoot.
3098 This is to help gcc which has difficulties to inline stuff from one
3100 The speedup seem to be above 15%, althrough more tests are needed on
3101 amd64 to confirm that gain.
3104 * Port of MSG's mailbox on top of SIMIX network
3105 The put/get mechanism was greatly simplified on the way.
3108 * New SIMIX network module. Provides:
3109 - Mailbox: rendez-vous mechanism to find with who you want to speak
3110 - Synchronous send/recv: easier and hopefully faster since the
3111 logic is handled in the maestro process directly now
3112 - Asynchronous send/recv: you dreamt of it? It's here now
3113 Too bad that nobody cared enough to propagate the change to MSG.
3114 * Add semaphores as SIMIX synchronization mechanism.
3117 * new function SD_daxload(char*) to load a DAX file
3118 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3119 * Introduce typed tasks. Specify its kind and cost at creation.
3120 At scheduling, just give where it should be placed, and the cost
3121 for each involved resource is automatically computed.
3122 Existing constructors so far (more to come of course):
3123 - SD_task_create_comm_e2e() for end-to-end communication
3124 - SD_task_create_comp_seq() for sequential computation
3125 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3126 * new function SD_task_dump() for debugging display
3127 * new function SD_task_dotty(task,FILE*) writing to file the info
3128 about the task in dotty format
3129 * SD_task_dependency_exists() can now cope with having one of its
3130 arguments NULL. If so, it tests whether the other argument has any
3132 * Add getters on list of preceding/following tasks:
3133 SD_task_get_parents(task) and SD_task_get_children(task)
3134 * Add getters on amount of workstations and list:
3135 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3136 * Add getter on task kind: SD_task_get_kind(task)
3137 * Update the start_time and finish_time of tasks on completion/failure
3138 * Bugfix: Remove task from state swags when destroyed
3141 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3144 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3145 dynar with one separate thread per value of the dynar.
3146 * Change the prototype of xbt_thread_create(), sorry.
3147 Added a boolean parameter indicating whether we want to join this
3148 thread (used in SG only for now)
3149 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3152 * GTNetS wrappers should now be usable again (and betterly tested too)
3153 * Fix a major regression from 3.2 where the timeout provided to
3154 MSG_task_put_with_timeout() was used as absolute time before which
3155 the comm should be done.
3156 * Start to fix the <cluster> tag.
3157 - Internal links should be good now (beside of the loopback, which
3158 use the private link instead)
3159 - paths to the external world is still rather broken
3160 - the <route:multi> tag is just broken. Actually that's brain-dead.
3161 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3162 to make it less stupid
3163 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3164 * Fix a source-level compatibility glitch from 3.2: after defining
3165 MSG_USE_DEPRECATED, you can use the old name
3166 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3167 * Allow to compile from the SVN with automake 1.11
3168 * Fix some problems when using the "start_time" tag in deployment XMLs.
3169 * Fix #8569: XBT/synchro.h has redundant declarations
3170 * Fix #8563: MSG return values and exceptions
3171 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3172 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3173 Thanks to Robson Peixoto.
3174 * Fix a vicious bug in dictionaries inducing that some elements were
3175 not freed on xbt_dict_free()
3177 Portability report of this version:
3178 * Main portability targets:
3179 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3180 on (amd64/i386/ia64)
3181 - mac leopard on i386
3182 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3183 but nothing critical.
3184 * Other platforms: windows, AIX and others were not tested for this release
3186 Timing report of this version:
3187 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3188 scenario parameters). From 8h to a few seconds in desktop grid settings.
3189 * Supernovae brings about 25% speedup on i386.
3191 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3193 SimGrid (3.3.3) stable; urgency=low
3195 The "Need for Speed" release.
3197 The timings done to validate the 3.3.2 were faulty.
3198 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3200 The problem was a conversion from a manually handled vector to
3201 xbt_dynar_t on the critical path.
3202 xbt_dynar_foreach calls functions, inducing stack management crap.
3204 We inlined these functions and xbt_dynar_foreach is now breath taking.
3205 We also inlined xbt_swag_belong on the way.
3207 Here are some approximate speedup measurements (on master/slaves
3208 simulations lasting between 10s and 20s each):
3209 3.3.1 -> 3.3.2: about same performance
3210 3.3.2 -> 3.3.3: 40% speedup
3211 3.3.1 -> 3.3.3: 40% speedup
3212 3.3.1 with inline patch -> 3.3.3: 30% speedup
3214 Our reading is that the refactoring which occurred in 3.3.2 made us
3215 suffer much more from the xbt_dynar_foreach low performance, but
3216 once we solved this, this refactoring proved to be very performance
3217 effective. From the 40% speedup, somehow, 10% are due to the
3218 inlining and 30% to the refactoring.
3220 That's a pitty that gcc cannot inline functions placed in other files
3221 alone. We have to choose between:
3222 - break the encapsulation (by putting private data structures and
3223 accessors in headers files to help gcc)
3224 - live with low performance
3225 - switch to a decent compiler such as icc (not quite possible).
3227 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3229 SimGrid (3.3.2) stable; urgency=low
3231 The "Simplicity does not preceed complexity, but follows it" release.
3233 The main contributors of this release were (lexical order):
3234 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3237 * Extract the routing logic into its own object.
3238 (was dupplicated in network.c and workstation_LV07.c;
3239 Allows to implement other ways of storing that info)
3240 => kill now useless network_card concept
3241 - Use dynar to represent routes (instead of void** + int*)
3242 - kill link_set (use surf_network_model->resource_set instead)
3243 - Add a command-line option to choose the routing schema to use
3244 - Add three new models:
3245 * Floyd (shortest path computed at initialization)
3246 * Dijikstra (shortest path recomputed all the time)
3247 * Cached Dijikstra (shortest path computed on need)
3248 All these models where contributed by Silas De Munck, and are
3249 described in his ICCS09 paper.
3251 * Simplify model declaration
3252 (less redirections, less function to write when defining a model)
3253 - Factorize stuff between models:
3256 surf_model_resource_set(model)
3257 surf_model_resource_by_name(model, name)
3258 - Unify the types of models in s_surf_model_t (using an union)
3259 - Embeed fields of common_public directly into s_surf_model_t
3260 - Rename model methods:
3261 action_free ~> action_unref
3262 action_change_state ~> action_state_set
3263 action_get_state ~> action_state_get
3264 - Change model methods into functions:
3265 (model)->common_public->action_use ~> surf_action_ref
3267 * Implement a generic resource; use it as ancestor to specific ones
3268 (allows to kill duplicated code in models)
3269 Drawback: timer command don't need no name nor properties;
3270 workstation_CLM03 don't need no properties
3271 (but I guess we can live with those few bytes wasted)
3273 * Improve the action object model
3274 - implement a constructor avoiding dupplicated code about field
3275 initialization in generic_action part.
3277 * Kill the SDP model: it has an external dependency, is deprecated
3278 in flavor of modern lmm models, and didn't compile since a while
3281 * Relocation of the context module from XBT to SIMIX.
3282 (the context were decoupled from the simix processes, duplicating a lot of code)
3283 => a lot of code was factorized
3284 - less overhead is introduced during scheduling
3285 - simpler API for the context factory
3286 - the logic for process creation,destruction and manipulation was simplified
3287 * Simplification of the s_smx_process_t data structure.
3288 => accesing the simix level data associated to a process is faster now,
3289 and the code is a lot more readable.
3292 * Implement some more MPI primitives:
3293 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3294 -implementation: Bcast: flat or 2-ary tree (default),
3295 Barrier: 4-ary tree,
3297 Allreduce: Reduce then Bcast
3298 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3299 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3300 Alltoallv: flat tree, like ompi
3302 * Add support for optimized collectives (Bcast is now binomial by default)
3303 * Port smpirun and smpicc to OS X
3306 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3307 Shout out if you used it.
3310 * Display the list of still queued messages in SG mode when existing
3314 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3315 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3316 * Add xbt_str_from_file(FILE*)
3317 * Add xbt_dict_get_key achieving a linear reverse search
3318 * Remove the context module
3320 Portability report of this version:
3321 * Main portability targets:
3322 - Linux(debian)/x86/context
3323 - Linux(debian)/x86/pthreads
3324 - Linux(debian)/amd64/context
3325 - Linux(debian)/amd64/pthreads
3326 On these, we still have the eratic breakages of gras/pmm and
3327 amok/saturate_sg reported in previous version. We still think
3328 that the tests are the cause of the fault, not the tested code.
3330 - Mac OSX Leopard/x86/context
3331 Still false negative in tesh autotesting.
3332 Smpi still fails, but this time because readlink does not accept -f
3333 Everything seems to work properly beside of that.
3336 - AIX version 5.3 (only tested contexts this time)
3337 Smpi still fails there because mktemp is not installed.
3338 Everything seems to work properly beside of that.
3340 I managed to compile it for the first time, but several breakages.
3341 Won't delay the release for this exotic platform.
3343 * Windows: it's still lagging behind. If you want to help, please
3346 Timing report of this version:
3347 This version seem to be more than 5% faster than 3.3.1 (on linux
3348 64bits with contextes). The gain is less than expected, we are
3349 investigating this for next release.
3351 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3353 SimGrid (3.3.1) stable; urgency=low
3356 * Implement a --cfg-help to show existing configuration variables
3357 * Build chain do not require doxygen in maintainer mode
3360 * fix a bug on struct sizeof computation, which prevented the
3361 exchange of arrays of structs in some conditions
3362 - added a regression test about this in datadesc_usage
3363 * Allow the exchange of 0-long dynamic vectors.
3364 - for that, use -1 as indicator of dynamic size instead of 0
3365 - This implied to change any size from unsigned long to long,
3366 reducing a bit communication abilities, but I guess that with
3367 64bits being quite common, this is more than enough.
3368 - This also induce a protocol change, thus bumping network protocol
3369 version from 0 to 1 (if we have external users, we have to get
3370 clean on that point too ;)
3371 - added two regression tests about this in datadesc_usage
3372 * Be more verbose when propagating local exceptions
3373 This helps debugging.
3374 * Display the status of simulated processes when receiving SIGINT in
3378 * Allow to control the simulation from a trace file.
3379 New functions MSG_action_register() and MSG_action_trace_run()
3380 The first one allows to associate a function execution to each
3381 kind of action while the second one parses a trace file and
3382 triggers the corresponding actions within the system.
3383 For now, only a toy example is provided in examples/msg/actions
3384 * Add an example of process migration in examples/msg/migration
3385 * Fix a bug in task exchange which broke MSG_task_get_sender()
3386 Add a teshsuite regression test for that.
3387 [Bug: if MSG_task_get_sender() is called after sender exit,
3389 * Fix a bug which prevented suspend/resume to work properly
3390 * Display the status of simulated processes when receiving SIGINT
3391 This fixes a regression of v3.3. due to the introduction of SIMIX
3392 * Bug fixing in failure management:
3393 - trace could not start by a failure at time 0
3394 - failure during communications were not working
3397 * Add SIMIX_process_set_name() to change the name of the current
3398 process in the log messages.
3399 * Store smx_hosts in a dict since we only retrieve them by name
3400 * Move the configuration infrastructure to surf
3403 * Move the configuration infrastructure to surf
3406 * Massive internal cleanups:
3407 - Store internal structures on processes instead of hosts (allows
3408 to have more than one process per host, in addition of being more
3410 - Cleanup the initialization/finalization process
3411 - Kill a whole bunch of unneeded synchronization:
3412 processes run in exclusive manner within the simulator
3413 - Move queues from global tables to process data fields
3415 - now accept -platform and -hostfile arguments
3416 - Pass the right rank value to processes according to the hostfile
3417 * Compile the examples by default, and use them as regression tests
3418 * Implement MPI_Wtime()
3419 * Change the reference speed to a command line option
3422 * TCP_gamma can now be specified as command line option using
3423 --cfg=TCP_gamma:10000000.0
3424 * Change the --surf-path cmd line option into --cfg=path:
3427 * Also include strbuff from xbt.h public header
3428 * xbt_ex_display(): do not free the exception after displaying
3429 This allows to do more with the given exception afterward.
3430 Users should call xbt_ex_free() themselves.
3434 Portability report of this version:
3435 * Main portability targets:
3436 - Linux(debian)/x86/context
3437 - Linux(debian)/x86/pthreads
3438 - Linux(debian)/amd64/context
3439 - Linux(debian)/amd64/pthreads
3440 These targets fail about 1/10 of times on gras/pmm, but we believe
3441 that this is because of the test, not because of simgrid.
3442 amok/saturate_sg fails even more rarely, and the test may not be
3445 - Mac OSX Leopard/x86/context
3446 The test suite still spits tons of errors because some obscure
3447 force prevents us from removing the temporary directories
3448 arguing that they still contain some metadata I've never heard of.
3449 Smpi fails because seq is not installed.
3450 Everything seems to work properly beside of that.
3453 - AIX version 5.3 (both contexts and pthread)
3454 Smpi still fails there because mktemp is not installed.
3455 XML inclusions seems rosty on AIX.
3457 * Windows: it's still lagging behind. If you want to help, please
3460 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3462 SimGrid (3.3) stable; urgency=high
3466 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3467 [Malek Cherier & Mt]
3469 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3472 Until now, GRAS were using MSG as an interface to SURF. It was
3473 quite difficult because both interface have several differences
3474 (MSG channels vs GRAS sockets were the most notable point).
3476 This also opens the gate to SMPI (which should occur soon) and speed
3477 up simulations by to 40% (even if it were not the main goal).
3479 **************************************
3480 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3481 **************************************
3482 The point is that events occurring at the exact same timestamp are
3483 not scheduled in the same order with the old and new version. This
3484 may be enough to completely change the execution of simulations in
3485 some cases. Sorry for the inconvenience.
3487 * Cleanup and upgrade the XML format to push further scalability
3488 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3490 * Improve the testing infrastructure with tesh. Now a very large part of
3491 the code is tested not only by being run but also by checking that the
3492 output match an expected output [Mt].
3494 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3495 is really less memory-demanding, which should allow you to use
3496 larger files in SimGrid [AL].
3498 * Inform valgrind about our contextes, so that it becomes usable
3499 with the default (and more effecient) version of SimGrid
3500 [contributed by Sékou Diakite, many thanks]
3503 * Introduce a listener thread in charge of receiving incoming
3504 messages from the network. It allows to overlap communication and
3505 computation but most notably, it removes some stupid deadlocks due
3506 to the fact that so far, a process could not send and receive at
3507 the same time. This made most non trivial communication schema
3509 * Convert the PIDs from long int to int to match the MSG ones (and
3510 linux ones too) [Mt]
3511 * New function: gras_agent_spawn() to launch a new process on
3512 current host. Only working in simulation for now. [Mt]
3513 * New function: gras_os_hostport() returning a constant form (ie,
3514 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3517 * Make the backtrace of exceptions more human readable [Mt]
3518 * New module: xbt/str [Mt]
3519 a ton of string utility functions (split, join, printf to a newly
3520 allocated buffer, trim, etc)
3521 * New module: xbt/hash [Mt]
3522 SHA1 hashing algorithm (more to come if needed)
3523 * New module: xbt/synchro [Mt]
3524 synchronization tools (mutex and conditions) working the same way
3525 in simulation and in real life (mainly useful for GRAS, but not
3527 * New module: xbt/queue [Mt]
3528 classical producer/consumer synchronization scheme
3529 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3530 (using the classical functions will get serialized) [Mt]
3531 * Make dictionary internal table dynamic. No need to specify its size
3532 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3534 * Make sure the log channels are organized as a tree under windows
3535 (because of ANSI C compatibility issue, any channel were child of
3536 root directly) [Mt].
3539 * Cleaned many thing in surf and fixed a few bugs [AL].
3540 * Add a nice command line configuration mechanism to compose models [AL].
3541 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3542 the previous one (KCCFLN05). It relies on something that looks like
3543 a max-min sharing mechanism but cannot be written as such. A new solver
3544 was thus designed [AL].
3545 * Add a new solver to lmm. Based on Lagrange optimization and
3546 gradient-based descent, it enables to efficiently maximise systems s.a
3548 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3550 This solver enables to propose two new network models for TCP Reno and
3551 TCP Vegas based on Low's work. These models still need to be fully
3552 tested though [Pedro Velho].
3555 * Bug fix in SD_simulate. Now the time bound given as argument is
3557 * Use the new parallel task model (ptask_L07) as default.
3558 * Use the SURF command line configuration mechanism.
3559 * 0-size tasks (for synchronization) should now work.
3561 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3563 SimGrid (3.2) stable; urgency=high
3567 We still experience issues on this platform, but we believe that at
3568 least MSG is usable.
3570 GRAS API BREAKAGE (for simplification purpose, sorry):
3571 * the gras_msgtype_by_name is not used anymore. Instead of
3572 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3573 you can write (and must)
3574 gras_msg_send(toserver, "request", &request);
3575 - If you still want to pass a gras_msgtype_t to the function (to cache
3576 the type and avoid the lookup time), use the gras_msg_send_() variant.
3577 - Impacted functions:
3578 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3579 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3580 * The callbacks are now expected to return 0 when everything went well
3581 (just like the main() function)
3583 GRAS new features and improvements:
3584 * New module mechanism where user code can use per process globals [Mt]
3585 This is similar to gras_userdata_*() functions, but for libraries. It
3586 factorize some code developped over and over in the examples and AMOK.
3587 It has still to be documented and used (only amok/peermanagement is
3589 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3590 loss when they were small enough to fit into the buffer and sent quickly
3591 enough so that they can all get received in one shoot.
3592 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3593 if not found. Use the *_or_null() variant for the old semantic.
3594 * In gras_msg_handle, do not discard messages without callback.
3595 They are probably messages to be explicitly awaited later (ie, proofs of
3596 mis-synchronization in userland since they are sent before being awaited)
3597 No big deal usually.
3598 * gras_socket_meas_send/recv: semantic changed!
3599 The numerical arguments used to be (1) the total amount of data to send
3600 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3601 messages. This was need for the fool willing to send more than MAXINT
3602 bytes on quite fat pipes.
3605 * Do really rename the hostmanagement module to peermanagement. [Mt]
3606 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3607 but this is rather new and this was documented in the module
3608 documentation (poor excuses, I admit)
3609 * Bandwidth measurement semantic changed! This follows the changes to
3610 gras_socket_meas_send/recv explained above.
3613 * A sequential mode has been added to the workstations. When a workstation
3614 is in sequential mode, it can execute only one task, and the other tasks
3615 are waiting in a FIFO. [Christophe Thiery]
3618 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3619 model for SIMDAG. [Christophe Thiery]
3620 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3621 equal to 0, which caused some bad behaviors in
3622 saturated_constraint_set_update. I now use a threshold mechanism like in
3626 * When running manually src/testall, you select specific units [Mt]
3627 testall is the result of our cunit mechanism, and should replace all
3628 the scripty thingy around since bash don't run easily on billware.
3630 * A mallocator system has been added. [Christophe Thiery]
3631 Mallocators allow you to recycle your unused objects instead of freeing them
3632 and allocating new ones.
3634 Documentation update:
3635 * FAQ reworking + New FAQs:
3636 - "Valgrind spits tons of errors!" [Mt]
3637 - "How to repport bugs" [Mt]
3638 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3639 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3640 - Communication time measurement within MSG [AL]
3641 - I experience weird communication times when I change the latency [AL]
3642 * GRAS tutorial [Mt]
3644 - an introduction to the framework and to the used communication model
3645 - an initiatic tour introducing the most proheminent features:
3647 . Lesson 0: Installing GRAS
3648 . Lesson 1: Setting up your own project
3649 o Part 2: Message passing
3650 . Lesson 2: Exchanging simple messages
3651 . Lesson 3: Passing arguments to the processes (in SG)
3652 . Lesson 4: Attaching callbacks to messages
3653 . Lesson 5: Using globals in processes
3654 . Lesson 6: Logging information properly
3655 . Lesson 7: Using internal timers
3656 . Lesson 8: Handling errors through exceptions
3657 . Lesson 9: Exchanging simple data
3658 . Lesson 10: Remote Procedure Calling (RPC)
3659 . Lesson 11: Explicitly waiting for messages
3660 . Recapping of message passing features in GRAS
3661 - A HOWTO section containing:
3662 o HOWTO design a GRAS application
3663 More are due, of course. They will come latter. In the meanwhile, you can
3664 check the examples which are still here.
3666 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3668 SimGrid (3.1) stable; urgency=high
3672 There was a stack corruption somewhere, visible only when optimizing
3673 with these versions. [Vince]
3676 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3677 module that enabled to play with Directed Acyclic Graphs. It is built
3678 directly on top of SURF and provides an API rather close to the old
3679 SG. Some old codes using SG are currently under rewrite to check that
3680 all needful functions are provided. [Christophe Thiery]
3683 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3684 extension of the classical CLM03 model that gracefully handles
3685 failures. This is now the default model for MSG and GRAS. It doesn't
3686 handle parallel tasks yet though. [AL]
3687 * Bug fix: Weights were not correctly set in the network part.
3688 WARNING: This may have resulted in incorrect results with simulations
3689 where there are more than one flow on a given link. [AL]
3692 * After a (long ?) discussion on simgrid-devel, we have decided that the
3693 convention we had on units was stupid. That is why it has been decided
3694 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3695 WARNING: This means that all previous platform files will not work as
3696 such with this version! A warning is issued to ask users to update
3698 A conversion script can be found in the contrib module of the CVS, under
3699 the name contrib/platform_generation/surfxml_update.pl [MQ]
3702 * Bug fix: Processes were started in reverse order, wrt deployment file.
3703 WARNING: if your code relies on this bug, please fix it. [AL]
3704 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3705 being executed on two different locations. [AL]
3706 * Bug fix: Failures are now better supported thanks to Derrick's tests
3707 (there was many failure situations I hadn't thought of and that weren't
3708 correctly handled). [AL]
3709 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3713 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3714 Exception occurring on server-side are propagated back to client (!).
3716 API CHANGE: the callback changed their prototype. Change:
3717 int my_handler(gras_socket_t expeditor, void *payload_data) {
3719 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3720 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3722 * New! function: gras_msg_handleall to deal with all messages arriving
3723 within a given period.
3724 * New! function: gras_socket_server_range to get a server socket in a
3725 range of port numbers (ease to avoid port number conflicts) [MQ]
3726 * New! gras processes display their backtrace when they get a SIGUSR1
3727 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3728 Sweet to debug RL processes [MQ]
3732 - Do not force experiment sizes to be expressed in kb, or it becomes
3733 impossible to measure the latency this way (needs one byte-long tests)
3734 WARNING: this changes the amok_bw_* function semantic. [MQ]
3735 - Implements the link saturation stuff. [MQ]
3736 * Peer management module:
3737 New! module factorizing code that we wrote over and over [MQ].
3740 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3741 - Test units are placed directly into the library code, they get extracted
3742 automatically and placed into the src/testall binary.
3743 - Convert most of the XBT tests to this system.
3744 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3745 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3746 live in the same C code. This required to change a little bit the API
3747 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3748 * New module: sparse graph structure with basic algorithms (this is work
3749 in progress and the API is not considered to be frozen yet). [AL]
3750 * Display more information on backtraces: source line & function names are
3751 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3752 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3753 * Reworked a little bit some #include statements to load only required
3754 headers. Some user code that relied on SimGrid to include stdlib or
3755 stdio may need to include it by themselves. [AL]
3756 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3758 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3759 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3760 but I guess that almost nobody used those parts. [MQ]
3762 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3764 SimGrid (3.0.1) stable; urgency=low
3767 * Unfortunately, I had missed 5 misnamed functions:
3768 xbt_fifo_item_t xbt_fifo_newitem(void);
3769 void xbt_fifo_freeitem(xbt_fifo_item_t);
3770 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3771 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3772 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3773 They're now deprecated. Please use their new versions:
3774 xbt_fifo_item_t xbt_fifo_new_item(void);
3775 void xbt_fifo_free_item(xbt_fifo_item_t);
3776 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3777 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3778 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3780 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3781 * Documentation: xbt_log module unmercifully reworked [MQ]
3782 * Bugfix: there was a problem with the ending of contexts with
3783 the pthread backend. It caused some weird deadlock or behavior
3784 depending on the pthread implementation. [AL]
3785 * Bugfix: get the exceptions raised in the simulator repport where
3786 and why they come from when they are not caught in time [AL, MQ]
3789 * Bugfix: Do repport the error when two non-connected hosts try to
3790 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3793 * Add additionnal checkings on communications. Assert that two
3794 communicating hosts are connected by a set of links... [AL]
3797 * Add additionnal checkings on channel values in communication [AL]
3798 * New: MSG_task_get_source to see on which host a task was generated [HC]
3799 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3800 the number of tasks waiting to be received on channel and sent
3802 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3803 waits for the first task coming from a given host.. [AL]
3805 GRAS new functionnalities: [MQ]
3806 * Enhance the parsing macro to allow the size of multidimentional objects
3807 to be given thru annotations.
3808 * New example (and documentation): Matrix Multiplication a la RPC
3809 (as when I was young!) and fix a bunch of bugs found on the way.
3811 GRAS performance improvements: [MQ]
3813 * Reduce the amount of cbps creation/destruction by making it static to
3814 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3817 * Change libdata to a set so that we can search for stuff by ID (and thus
3818 reduce the insane amount of dict lookups)
3821 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3822 RL communication on top of SG since it's so uneffective.
3823 It may also be used for inter-thread communication in RL, one day.
3824 * Use gras_datadesc_copy() to exchange messages on top of SG
3826 - improve message exchange performance on top of SG
3827 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3828 * Don't exchange on the network the size of the used part of buffer,
3829 instead, specify the possible buffer size to read().
3831 - reduces the amount of read/write calls (one pair per exchange)
3832 - reduces the amount of exchanged data (the size)
3833 - allows to retrieve all arrived data on receiver side, if we don't need
3834 it right now (subsequent read will peek the buffer)
3835 - allows the receiver to proceed with the beginning of the stream before
3836 everything is arrived
3837 - make it possible to build an iov transport (using readv/writev)
3839 - take care of the data with non-stable storage (like stacked data),
3841 * If possible, TCP send uses vector I/O (when writev() is here)
3842 - Don't use it for receive since we send data sizes and data on the
3843 same stream, so we wouldn't be able to chain large amount of chunks
3844 before having to flush the stuff to read the size.
3845 * Rework the transport plugin mechanism to simplify it and reduce the
3846 amount of pointer dereferencement when searching for the right function
3849 * I guess that now, we do almost as few system calls as possible while
3850 doing as few data copy as possible.
3852 To improve it further, we could try to send all the sizes first and then
3853 all the data (to use iov on receiving size), but it's only a partial
3854 solution: when you have 2 dimensional data, the sizes of the second
3855 dimension is data of the first dimension, so you need 3 streams.
3857 I'm not sure the potential performance gains justify the coding burden.
3859 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3861 SimGrid (3.00) stable; urgency=high
3864 * New! Give the possibility to hijack the surf parser and thus bypass
3865 MSG_create_environment and MSG_launch_application. Have a look at
3866 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3868 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3870 SimGrid (2.96) unstable; urgency=low
3875 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3876 This deprecates the xbt_error_t mechanisms.
3877 It modifies (simplifies) all XBT and GRAS API.
3878 MSG API keeps unchanged (exceptions raised by XBT are caught from
3879 within MSG and masked with existing error handling facilities)
3882 * New! Add a FATPIPE model. [AL]
3883 * New! Add a parallel task model. [AL]
3884 * New! Add automatically a loopback interface (in the default
3885 network model) if none was precised.
3888 * Bugfix: MSG_process_resume now works with the current running process.
3890 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3891 * Modification of MSG_task_get_compute_duration. Once a task has been
3892 processed, the value returned by this function is now equal to 0. [AL]
3893 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3894 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3895 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3896 of a m_task. For now, MSG_task_cancel only works with computation
3898 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3899 of the processor (in Mflop/s) regardless of the current load on the
3901 * API Change: use proper naming convention for MSG_getClock and
3902 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3904 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3905 This function changes the priority of a computation task. This priority
3906 doesn't affect the transfer rate. A priority of 2 will make a task
3907 receive two times more cpu power than the other ones. This function
3908 has been added to suit the needs of Nguyen The Loc and hasn't been that
3909 much tested yet. So if it fails, please report it and send me your code.
3911 * API Change: removed all functions and types that were marked "deprecated"
3912 since many months. Renamed MSG_global_init_args to MSG_global_init.
3914 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3916 SimGrid (2.95) unstable; urgency=low
3919 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3920 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3921 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3924 * Bugfix: Allow absolute paths for platform description files [MQ]
3925 * Bugfix: do free the variables after use. Leads to drastic performance
3927 * Implement max_duration (ie, timeouts) on resources [AL]
3930 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3932 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3933 the needed support (timeouts) [AL]
3936 * Implement measurement sockets. You can now get the bandwidth between two
3937 hosts thanks to AMOK (see below). [MQ]
3938 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3939 dynar over the network (yeah) [MQ]
3940 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3941 * simple/static token-ring example. [Alexandre Colucci and MQ]
3942 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3943 in place before [MQ]
3946 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3947 type parsing. "short" and "long" modifiers were also ignored; other
3948 modifier (such as reference level) are still ignored. [MQ]
3949 * Embeed the buffer size within the buffer itself on SG. [MQ]
3950 That way, send() are atomic and cannot get intermixed anymore (at least
3951 the ones which are less than 100k; bigger messages still have the issue)
3952 * Array size pushed by the field, not by the field type (or each
3953 and every long int will push stuff to the cbps) [MQ]
3954 * use select() to sleep since it allows to portably sleep less than one
3957 GRAS (minor cleanups)
3958 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3961 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3962 * type category 'ignored' killed as it was never used and were difficult
3964 * whether a type can cycle or not is now a flag, leaving room for more
3965 flags (as "ignored", if we feel the need one day ;)
3966 * Rename raw sockets to measurement sockets since "raw" has another
3967 meaning in networking community.
3970 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3971 and offers features not belonging to GRAS but that most applications
3972 need. One day, it may be a set of plugins loadable at runtime.
3973 * New module: bandwidth
3974 bandwidth measurement between arbitrary nodes running this module. [MQ]
3976 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3978 SimGrid (2.94) unstable; urgency=low
3980 The first beta release of SimGrid 3 !
3984 * Update the main page and the FAQ. Adding references to gforge.
3987 * Add a gras_os_getpid function.
3990 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3991 * Extend the logs so that they also print PID, hostname, date, ... if
3993 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3994 and kill the old version which were in testsuite/
3995 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3999 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
4000 for users. Instead, each programming environment (GRAS, MSG,...) use it
4001 when needed to provide such a feature to users.
4002 Don't shortcut the mechanism or you will also shortcut the virtualization
4003 you need on the simulator.
4007 * Cleanups in configury with regard to compile optimization/warning flags.
4008 Also add -fno-loop-optimize to any powerpc since it's the optimization
4009 killing gcc (< 3.4.0).
4010 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
4011 needed by MSG examples complications
4012 * Borrow configury beautifications from PHP
4015 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
4016 warning (thanks loris for stumbling into this one).
4017 * Bugfix: stop loading private headers (gras_config.h) from the public
4021 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
4022 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
4023 RL would imply a DNS resolver, which is *hard* to do in a portable way
4024 (and therefore delayed).
4025 * Implement a real timer mechanism and use it in timing macros. This allows
4026 to avoid rounding errors and get a 0.000005 sec precision in timing
4027 macros. While I was at it, various cleanups:
4028 - allow to declare more than one timed section per file (fix a stupid bug)
4029 - move some private declaration to the right place
4030 - merge conditional execution and timing macros into emulation module
4031 - document the module
4032 - make sure the module cleanups its mess on gras_exit
4033 * Documentation improvements:
4034 - (new) how to compile applications using GRAS
4035 - (new) emulation support (timing macros)
4037 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
4039 SimGrid (2.93) unstable; urgency=low
4041 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
4044 - Use Paje properly where used. Still to be sanitized properly.
4045 - Portability fix: Add an implementation of the contexts using pthread
4049 - Add xbt_procname(): returns the name of the current process.
4050 Use it to show the current process's name in all logging.
4052 - fix detection of older flex version and the reaction, since we do
4053 depend on modern ones (we use lex_destroy)
4054 - Better separation of SG and RL in the libs: remove all simulation code
4055 from libgras. As a result, this lib is now only 200k when stripped.
4056 Some of the xbt modules may also be duplicated (two sets and such) and
4057 should be cleaned/killed before SG3.
4058 - Insist on using xlC on AIX because of weird problems involving gcc there.
4059 - Cleanup the make remote stuff. This is now done by scripts
4060 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
4061 mainly for our private use, but we're working on changing them to user
4064 - Bugfix: flush the socket on close only if there is some *output*.
4065 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
4067 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
4070 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
4072 SimGrid (2.92) unstable; urgency=low
4074 Alpha 3 on the path to SimGrid 3
4078 - New! First try of benchmarking macros.
4079 - New! First try so that gras_stub_generator generate deployment and
4080 remote compilation helpers.
4082 - Bugfix: Initialization fix in msg_test.
4086 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
4088 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
4090 - API Change: gras_os_sleep to take the amount of seconds as a double.
4091 Accepting an int was error prone since it was the only location where
4092 seconds were coded as such. It leaded to damn rounding errors.
4093 - Bugfix: Hard to belive that timers ever worked before this.
4095 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4097 SimGrid (2.91) unstable; urgency=low
4099 Alpha 2 on the path to SimGrid 3
4103 - Bug fix in the lmm_solver.
4105 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4106 through the function MSG_paje_output.
4107 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4108 - It is possible to bound the rate of a communication in MSG with
4109 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4110 to write it in the changelog).
4111 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4116 - Various cleanups to the autotools stuff
4117 - Begin to move Gras examples to examples/gras/
4118 - Let make distcheck work again (yeah!)
4120 - documentation overhauled using doxygen.
4121 gtk-doc-tools is dead in SimGrid now.
4122 - Automatically extract all existing logging categories, and add the list
4123 to the documentation (long standing one, to say the less)
4125 - Cleanup the known architecture table. Reorder the entries to group what
4126 should be, and use a more consistent naming scheme.
4127 (some of the test dataset are still to be regenerated)
4128 - New! Allow library to register globals on each process just as userdata
4130 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4131 lookup time (for now).
4132 Use it in msg and trp.
4133 This cleans a lot the internals and helps enforcing privacy of the
4134 headers between the gras components.
4135 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4136 - Bugfix: gras_os_time was delirious in RL.
4137 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4139 - Reenable GRAS now that it works.
4141 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4143 SimGrid (2.90) unstable; urgency=low
4145 Alpha 1 on the path to SimGrid 3
4147 * It is a long time since the last release of SimGrid. I'm sorry about
4148 that but as I had told you, I was rewriting a lot of things. I apologize
4149 to those who had been reporting bugs to me and that I had not answered.
4150 If your bug is still in the new version, please tell me. Here is a
4151 summary of the main changes.
4153 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4154 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4155 portability, speed, and a lot more but you'll figure it out later.
4156 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4157 I wanted to release the new MSG as soon as possible and I have
4158 broken GRAS, which is the reason why, for now, only MSG is fully
4159 functional. A laconic description of these projects is available
4160 in the documentation.
4162 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4163 kernel optimized for our needs (called SURF but only the developers
4164 should use it). Hence, MSG is now roughly 30 times faster and I think
4165 that by rewriting a little bit MSG, I could event speed it up a little
4166 bit more. Beside the gain in speed, it is also much easier to encode a
4167 new platform model with SURF than it was with SG. More to come...
4169 * REVOLUTION 3: I have tried to change a little as possible the API of
4170 MSG but a few things really had to disappear. The main differences
4171 with the previous version are:
4172 1) no more m_links_t and the corresponding functions. Platforms are
4173 directly read from an XML description and cannot be hard-coded
4174 anymore. The same format is used for application deployment
4175 description. The new format is described in the documentation.
4176 Have a look in tools/platform_generation. There is a tiny script
4177 that converts from the old platform format to the new one. Concerning
4178 the application deployment format, parsing the old one is tricky.
4179 I think most of you should however be able to convert your files. If
4180 it is really an issue, I can write a C code that does the conversion.
4182 2) the toolbox tbx does not exist anymore. We now have a library
4183 with much more data-structures but without the hash-tables (we have
4184 dictionaries that are much faster).
4186 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4188 *****************************************************************************
4189 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4190 * don't feel like converting the oldies. (Mt) *
4191 *****************************************************************************
4194 Version 2.90: "the long awaited one"
4195 - Finished rewriting and debugging MSG. Rewrote the documentation.
4196 - disable GRAS for now since it needs to be ported to the newest SG
4199 - Finish the port to windows (using mingw32 for cross-compile)
4202 - Main loop and datastructures of SURF. A cpu resource object is
4203 functional. Surf can thus be used to create cpu's with variable
4204 performance on which you can execute some actions.
4206 2004-11-15 Martin Quinson
4207 - Port to ARM. Simply added the alignment and size descriptions. Should
4208 work, but the ARM machines are so slow that I didn't had the opportunity
4209 to 'make check' over there yet.
4211 2004-11-15 Arnaud Legrand
4212 - Trace manager now written. It uses a heap structure and is therefore
4213 expected to be efficient. It may however be speeded up (particularly
4214 when many events occur at the same date) by using red and black
4215 trees. One day maybe...
4216 - Max-min linear system solver written. It uses a sparse matrix
4217 structure taking advantage of its expected use. Most operations are
4218 O(1) and free/calloc are called as few as possible. The computation of
4219 the minimum could however be improved by using a red and black tree
4222 2004-11-03 Arnaud Legrand
4223 - Rename every gras_* function that was in xbt/ to its xbt_
4225 - Add a heap and a doubly-linked list to xbt
4226 - Added a dichotomy to the dictionaries. make check works as well before
4227 so I assume that the patch is correct. I do not know however if things
4228 run effectively faster than before now. :)
4230 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4231 SimGrid, and the version number is bumped to 2.x
4233 2004-10-29 Martin Quinson
4234 - Introduction of the remote errors.
4235 They are the result of a RMI/RPC on the remote machine.
4236 ErrCodes being scalar values, you can't get the host on which those
4237 errors did happen. Extending the error mechanism as in Gnome is possible.
4238 No idea yet whether it is a good idea.
4240 2004-10-28 Martin Quinson
4241 - Interface revolution: the Starred Structure Eradication.
4242 I used to do typedef struct {} toto_t; and then handle *toto_t.
4243 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4244 - ???_t is a valid type (builded with typedef)
4245 - s_toto_t is a structure (access to fields with .)
4246 - s_toto is a structure needing 'struct' keyword to be used
4247 - e_toto_t is an enum
4248 - toto_t is an 'object' (struct*)
4250 typedef struct s_toto {} s_toto_t, *toto_t;
4251 typedef enum {} e_toto_t;
4252 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4253 s_toto_t) is private.
4255 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4256 so that it can set the variable to NULL. It was so for dicts and sets,
4257 it changed for dynars.
4259 - Fix a bunch of memleaks in dict_remove
4260 - Fix a bug in sg/server_socket opening: it failed all the time.
4262 2004-10-07 Martin Quinson
4263 - Speed up dynar lookup operation a bit.
4265 gras_dynar_get is dead.
4267 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4268 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4269 (which gives you the address of the stored data).
4271 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4272 copy of the data using an affectation to do the job and not a memcpy.
4274 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4275 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4277 It does not really speedup the dynar test because they are
4278 setting elements all the time (and look them seldom). But the dict does
4279 far more lookup than setting.
4281 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4283 2004-10-05 Martin Quinson
4284 - Allow to (en/dis)able the cycle detection at run time.
4286 Whether we should check for cycle or not is now a property of each
4287 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4288 datadesc_cycle_unset allow to remove this property when previously set.
4290 Note that the cycle detection is off by default since it impacts the
4291 performance. Watch the data you feed GRAS with ;)
4293 This property is hereditary. Any element embedded in a structure having it
4294 set have it set for the time of this data exchange.
4296 You should set it both on sender and receiver side. If you don't set it on
4297 sender side, it will enter an endless loop. If you forget on receiver
4298 side, the cycles won't be recreated after communication.
4300 - Header reorganization.
4301 Kill gras_private.h, each submodule must load the headers it needs.
4303 2004-10-04 Martin Quinson
4304 - Interface revolution: do not try to survive to malloc failure.
4306 Now, gras_malloc and friends call gras_abort() on failure.
4307 As a conclusion, malloc_error is not a valid error anymore, and all
4308 functions for which it was the only gras_error_t return value are
4309 changed. They now return void, or there result directly.
4310 This simplify the API a lot.
4312 2004-09-29 Martin Quinson
4313 - Re-enable raw sockets.
4314 Created by gras_socket_{client,server}_ext;
4315 Used with gras_raw_{send,recv}
4318 It should allow to kill the last bits of gras first version soon.
4320 This is not completely satisfactory yet (duplicate code with
4321 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4324 - Simplify transport plugin (internal) interface by not passing any
4325 argument to _server and _client, but embedding them in the socket
4328 2004-09-28 Martin Quinson
4329 - Finish the port to AIX.
4330 autoconf was my problem (segfault within the malloc replacement
4331 function. No idea why)
4333 2004-09-16 Martin Quinson
4334 - Fix some size_t madness on 64bit architectures.
4336 2004-09-08 Martin Quinson
4337 - Reduce the number of system headers loaded, overload some more system
4338 calls (such as malloc to cast the result of the system one, and work
4340 - Fix and reintroduce the config support
4342 2004-09-07 Martin Quinson
4343 - Source code reorganization to allow Arnaud to surf all over there.
4344 - Allow to document the logging categories.
4345 - Remove all uppercase from logging categories and useless cleanup in names.
4347 2004-08-18 Martin Quinson
4348 Version 0.6.2 (protocol not changed; API changed)
4349 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4350 gras_error_t), and NULL when not found. Functions expecting a msgtype
4351 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4352 hopefully usefull message.
4353 - Portability to prehistoric sparcs again
4355 2004-08-17 Martin Quinson
4356 Version 0.6.1 (protocol not changed; ABI not changed)
4357 - prealloc some buffers to speed things up
4359 2004-08-11 Martin Quinson
4360 Version 0.6 (protocol not changed; ABI expended)
4361 - The parsing macro can deal with the references, provided that you add
4362 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4364 2004-08-09 Martin Quinson
4365 Version 0.5 (protocol not changed; ABI changed)
4366 - Allow to off turn the cycle detection code in data exchange at
4367 compilation time. It should be at run time, but I'm short of time (and
4368 the config stuff is still broken). That way, we keep dict out of the
4369 critical path, which is good because the performance is poor:
4370 - search not dichotomial yet
4371 - dynar give no way to access their content and memcpy everytime
4372 - In composed data description (struct, ref and so on), stop foolness of
4373 keeping the subtype's ID, but store the type itself. This keeps sets out
4374 of the critical path, which is good since they rely on dynar and
4375 dictionaries. The only loose of that is that we cannot detect the
4376 redeclaration of a structure/union with another content (but I'm not sure
4377 the code detected well this error before anyway). We still can detect
4378 the redefinition discrepancy for the other types.
4379 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4380 it breaks the code because of type-punning used all over the place).
4381 This breaks on all non-gcc architectures (for now).
4383 All those changes (plus the buffer of last time) allow me to gain 2 order
4384 of magnitude on cruel tests consisting of 800000 array of integers on two
4385 level of a hierarchical structure (200 secondes -> 4 secondes)
4388 - the selector of reference must now return the type it points to, not
4389 the ID of this type.
4391 2004-08-06 Martin Quinson
4392 Version 0.4 (protocol changed; ABI not changed)
4393 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4394 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4395 - Implement a buffer transport, and use it by default (it relies on tcp in
4396 real life and on sg in simulation).
4397 That's a bit hackish since I had a new field to the structure to store
4398 its data without interfering with the subtype ones. Inheritance
4399 is tricky in C. And that's a kind of reverse inheritance with one class
4400 derivating two classes. Or maybe a game with java interfaces. Anyway,
4401 that's damn hard in C (at least).
4402 Moreover, I got tired while trying to ensure plugin separation and
4403 genericity in SG mode. MSG wants me to do weird things, so let's go for
4404 cruel hacks (temporarily of course ;).
4405 See comment in transport_private.h:71
4406 - do not include all the _interface headers in private but in the files
4407 which really need them (to cut the compilation time when they are
4410 2004-07-26 Martin Quinson
4411 Version 0.3 (protocol not changed; ABI changed)
4412 - Major overhault of the datadesc interface to simplify it:
4413 - shorted the function names:
4414 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4415 - add a trivial way to push/pop integers into the cbps without malloc.
4416 This allows to make really generic sub_type description, which simply
4417 pop their size of the stack.
4418 - add a function gras_datadesc_ref_pop_arr() which does what users want
4419 most of the time: Declare a dynamic array (which pops its size of the
4420 stack) and declare a reference to it. Poor name, but anyway.
4421 - kill the post-send callback, add a post-receive one
4423 2004-07-23 Martin Quinson
4424 Version 0.2 (protocol changed; ABI changed)
4425 - add some testing for cpbs in the test cases, and fix some more bugs.
4426 This invalidate again the little64 data file, since I cannot regenerate
4428 - remove an awfull optimization in the logging stuff, allowing me to:
4429 - understand it again
4430 - learn gcc how to check that the argument match the provided format
4431 - fix all errors revealed by gcc after that
4432 - internal keys of dict are not \0 terminated. Deal with it properly in
4433 loggings instead of segfaulting when the user want to see the logs :-/
4435 2004-07-22 Martin Quinson
4436 - Fix some stupid bug preventing cbps (callback postit) from working
4438 2004-07-21 Martin Quinson
4439 - Some documentation cleanups
4440 - remove the useless last argument of msgtype_declare
4441 - rename the Virtu functions to fit into the 'os' namespace
4442 - move headers src/include -> src/include/gras/ and stop fooling with
4443 gras -> . symbolic link
4444 - make distcheck is now successful
4446 2004-07-19 Martin Quinson
4448 - Build shared library also
4449 - Install html doc to the right location
4450 - stop removing maintainer files in make clean
4451 - build tests only on make check
4453 2004-07-13 Martin Quinson
4455 - No major issue in previous version => change versioning schema
4456 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4457 corresponding dataset.
4459 2004-07-11 Martin Quinson
4461 - Get it working with any kind of structure (we can compute the padding
4462 bytes remotely for all the architectures I have access to)
4463 - Implement the structure parsing macro (still not quite robust/complete)
4464 - Improvement to the remote testing toysuite
4466 2004-07-10 Martin Quinson
4467 [autoconf mechanism]
4468 - get ride of a bunch of deprecated macros
4469 - actually run the test for two-compliment, not only compile it :-/
4470 - test whether the structures get packed (and bail out if yes. Damn.
4471 Alignment is a serious matter)
4472 - test whether the structures get compacted (but respecting the alignment
4473 constraints of each types)
4474 - test whether the array fields of structures can straddle alignment boundaries
4476 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4477 (took me 2 hours to find that bug, looking at the wrong place)
4478 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4479 before sending/receiving and are used to compute the offsets of fields
4480 - Given that padding size depend even on compiler options, keep track of
4481 alignment and aligned_size only for the current architecture. Not a big
4482 deal since we send structure fields one after the other (seems
4484 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4485 program, let it work on linux/gcc/little32. portability todo.
4487 2004-07-08 Martin Quinson
4488 - import and improve remote compilation support from FAST
4489 - make sure make check works on half a dozen of machines out there
4491 2004-07-07 Martin Quinson
4492 Let's say it's version 0.0.3 ;)
4493 - Implement conversions (yuhu!)
4494 - Let it work on solaris (beside conversion, of course)
4495 - Stupid me, using rand() to generate the conversion datatests in not wise.
4497 2004-07-06 Martin Quinson
4498 - Let make dist work, since I'm gonna need it to compile on remote hosts
4499 - Let Tests/datadesc_usage write the architecture on which the file was
4500 generated as first byte.
4501 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4502 architecture descriptions.
4503 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4504 on those architectures.
4505 - Optimization: send/recv array of scalar in one shoot
4507 2004-07-05 Martin Quinson
4508 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4510 - Plug a whole bunch of memleaks
4511 - each process now have to call gras_{init,exit}. One day, their log
4512 settings will be separated
4513 - Continue the code factorisation between SG, RL and common in Transport.
4515 2004-07-04 Martin Quinson
4517 - Redistribution between SG and RL.
4518 We wanna have to accept in SG, so move accepted related parts of RL in
4519 the common part. (more precisely, the dynar of all known sockets is no
4520 more a static in transport.c, but part of the process_data)
4522 [gras_stub_generator]
4523 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4525 2004-07-03 Martin Quinson
4526 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4527 them all up in one shot)
4529 - Fix the ugly case of reference to dynamic array.
4530 - New (semi-public) function gras_datadesc_size to allow the messaging
4531 layer to malloc the needed space for the buffer.
4533 - gras_socket_close now expect the socket to close (and not its address to
4534 put NULL in it after it). This is because the socket passed to handlers
4535 is one of their argument (=> not writable).
4537 - propagate the interface cleanup from last week in datadesc, ie remove a
4538 superfluous level of indirection. User pass adress of variable
4539 containing data (both when sending and receiving), and not of a variable
4540 being a pointer to the data. Let's say that I like it better ;)
4541 The price for that is constructs like "int msg=*(int*)payload" in
4542 handlers, but it's a fine price, IMHO.
4544 - Let it work in RL (yuhu)
4546 2004-06-21 Martin Quinson
4548 - porting SG plugin and SG select to new standards (works almost).
4549 - plug memleaks and fix bugs around.
4552 - cleanup the prototype of data recv and force users to specify when they
4553 want to handle references to objects. Test case working even for cycles.
4554 - plug memleaks. Valgrind is perfectly ok with this.
4556 2004-06-12 Martin Quinson
4558 - cleanup the separation between plugin and main code in plugin creation
4560 2004-06-11 Martin Quinson
4562 - Reput hook for raw sockets, needed for BW experiments
4563 - kill a few lines of dead code
4564 [Data description] Interface cleanup
4565 - gras_datadesc_by_name returns the searched type or NULL.
4566 That way, no variable is needed to use a type desc once, which makes
4568 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4569 two parameters were strings (field name, type name), leading to
4571 [Dicos] Interface cleanup
4572 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4573 This is consistant with the dynar API.
4575 2004-04-21 Martin Quinson
4577 - Porting to new standards.
4579 - interface cleanup.
4580 There is no bag anymore, no need to take extra provision to mask the
4581 pointers behind "ID".
4582 Better splitup of functions between files create/exchange/convert.
4583 This is still a bit artificial since convert and receive are so
4584 interleaved, but anyway.
4586 - add a queued message list to procdata (the ones not matching criteria
4588 - factorize some more code between SG and RL wrt procdata
4590 - use gras_exit in example to track memleaks
4591 - get rid of gs_example now that GS is properly integrated into gras
4592 - update run_test to integrate the latest tests (datadesc)
4594 - rename WARNINGn macros to WARNn since it prooved error-prone
4596 2004-04-19 Martin Quinson
4598 - register init/exit functions within gras module mechanism
4599 - send/receive function.
4600 Convertion is not implemented, but short-cutted if not needed.
4601 struct/array elements are sent one by one (instead of block-wise), but
4602 nobody really cares (yet). Get a prototype before optimizing.
4603 - tests (using a file socket) for DD send/receive on:
4604 - base types: int, float
4605 - array: fixed size, string (ie ref to dynamic string)
4606 - structure: homogeneous, heterogeneous
4607 - chained list, graph with cycle
4608 Believe it or not, valgrind is not too unhappy with the results. The
4609 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4613 - Bugfix when using a filename explicitly (instead of '-')
4615 2004-04-09 Martin Quinson
4617 - factorize more code between RL and SG in socket creation
4618 - Complete the implementation and tests of:
4620 o file (only in RL, and mainly for debugging)
4622 I lost 3 days to design a portable address resolver, and then decided
4623 that the prototype mainly have to run on my box.
4624 Addressing portability too early may be like optimizing too early :-/
4626 - use gras_init in the Tests instead of the crappy parse_log_opt
4627 (the latter function is removed)
4628 [Conditional execution]
4629 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4630 [Code reorganisation]
4631 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4632 Build examples against the RL library, since there is no way to disable
4633 its creation for now.
4635 For information, the beginning of coding on GRAS was back in june
4636 2003. I guess that every line has been rewritten at least twice since