1 SimGrid (3.30.1) NOT RELEASED YET (v3.31 expected March 20. 2022, 15:33 UTC)
4 - Rework the internals, for simpler and modern code. This shall unlock many future improvements.
5 - You can now define plugins onto SafetyChecker (a simple DFS explorer), using the declared signals.
6 See CommunicationDeterminism for an example.
7 - Support mutex and semaphore in DPOR reduction
8 - Seems to work on Arm64 architectures too.
9 - Display a nice error message when ptrace is not usable.
12 - fix for FG#100 by ensuring small asynchronous messages never overtake larger
13 ones, conforming to the standard.
14 - replay: fix waitall behaviour to avoid forgetting requests and leaking
16 - tracing: ensure that we dump the TI traces continuously during execution and
17 not just at the end, reducing memory cost and performance hit.
18 - Update OpenMPI collectives selection logic to match current one (4.1.2)
21 - New signal: Engine::on_simulation_start_cb()
23 New plugin: the Chaos Monkey (killing actors at any time)
24 - Along with the new simgrid-monkey script, it tests whether your simulation
25 resists resource failures at any possible timestamp in your simulation.
26 - It is mostly intended to test the simgrid core in extreme conditions,
27 but users may find it interesting too.
30 - Drop xbt_dynar_shrink().
32 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
33 (FG: issues on Framagit; GH: issues on GitHub)
34 - FG#57: Mc SimGrid should test whether ptrace is usable
35 - FG#87: Smpi scripts fail with spaces in paths
36 - FG#100: [SMPI] Order of the message matching is not guaranteed
37 - FG#101: LGPL 2.1 is deprecated license
38 - GH#151: Missing mutexes for DPOR.
40 ----------------------------------------------------------------------------
42 SimGrid (3.30) January 30. 2022.
44 The Sunday Bloody Sunday release.
46 Main user-visible changes:
47 - The SimDag API for the simulation of the scheduling of Directed Acyclic
48 Graphs has been dropped. It was marked as deprecated for a couple of years.
49 We finally complete the implementation of what has been called SimDag++
50 internally, i.e., porting the different features of SimDag on top of S4U.
51 The new way to simulate the execution of dependent activities directly by
52 maestro (without any other actor) is details in the examples/cpp/dag-* series
54 - The removal of SimDag led us to also remove the export to Jedule files that
55 was tightly coupled to SimDag. The instrumentation of DAG simulation is still
56 possible through the regular instrumentation API based on the Paje format.
57 - We also dropped the old and clumsy Lua bindings to create platforms in a
58 programmatic way. It can be done in C++ in a much cleaner way now, which
59 motivates this suppression.
62 - Introduce on_X_cb() functions for all signals, to attach a new
63 callback to the signal X. The signal variables are now hidden and
64 only these functions should be used.
65 Rationale: this enables the usual deprecation scheme where functions
66 remain for 4 releases if we need to modify the signals, while the
67 current code with the signal variables directly visible prevents any
69 - New function: Engine::run_until(date), to split the simulation.
70 - New signal: Activity::on_veto, to detect when an activity fails to start.
71 - Signal change: Comm::on_start(Comm&, bool) has been replaced by
72 Comm::on_send and Comm::recv. These two signals respectively correspond to
73 when the sending or receiving side of a Comm is ready. They are raised at
74 the same locations as the former Comm::on_start signal.
75 - New function: Engine::track_vetoed_activities() to interrupt run()
76 when an activity fails to start, and to keep track of such activities.
77 Please see the corresponding example for more info.
78 - New functions: s4u::Comm::{sendto_init, set_source, set_destination} to enable
79 the use of vetoers with direct host-to-host communications. Both source and
80 destination have to set for a comm to start. Each call to these setters check
81 if all vetoes are satisfied. When it is the case, the comm starts. A use case of
82 these functions is given in examples/cpp/dag-scheduling.
83 - New functions: {Exec, Io}::update_priority allow you to modify the priority of
84 these kinds of activities during their execution. Behavior is detailed in
85 examples/cpp/io-priority/.
88 - Dynamic costs for MPI operations: New API to allow users to dynamically
89 change injected costs for MPI_Recv, MPI_Send and MPI_Isend operations.
90 Alternative for smpi/or, smpi/os and smpi/ois configuration options.
91 - Fix some issues with the replay mechanism.
94 - Function xbt::Extendable::get_data() is now templated with the type of the
95 pointee. Untyped function is deprecated. Use get_data<void>() if you still
96 want to retrieve void*.
99 - New section: "SimGrid MPI calibration of a Grid5000 cluster"
100 presenting how to properly calibrate MPI communications in SimGrid.
101 - Complete and reword the platform section, which is now completed.
104 - Thread contexts are used by default with Python bindings. Other kinds of
105 contexts revealed unstable, specially starting with pybind11 v2.8.0.
107 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
108 (FG: issues on Framagit; GH: issues on GitHub)
109 - FG#95: Wrong computation time for multicore execution after pstate change
110 - FG#97: Wrong computation time for ptask+multicore+pstates
111 - FG#98: SMPI offline simulation is inconsistent with the online simulation
112 (deadlocks / message truncation)
113 - FG#99: Weird segfault when not sealing an host
115 ----------------------------------------------------------------------------
117 SimGrid (3.29) October 7. 2021
119 The "Ask a stupid question" release.
121 We wish that every user ask one question about SimGrid to celebrate.
122 On Mattermost, Stack Overflow or using the issues tracker.
125 New modeling features:
126 - Non-linear resource sharing, modeling resources whose performance heavily degrades with contention:
127 - The total capacity may be updated dynamically through a callback
128 and depends mainly on the number of concurrent flows.
129 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
131 - Dynamic factors: model variability in the speed of activities
132 - Each action can now have a factor that affects its progression.
133 This multiplicative factor is applied when updating the amount of work
134 remaining, thereby an activity with factor=0.5 only uses half of the
135 instantaneous power/bandwidth it is allocated and will appear twice
136 slower than what it actually consumes.
137 - This can be used to model a overhead (e.g., there is a 20 bytes
138 header in a 480 bytes TCP packet so the factor 0.9583) but the novelty
139 is this factor can now easily be adjusted depending on activity's and
140 resources characteristics.
141 - This existed for network (e.g., the effective bandwidth depends
142 on the message in SMPI piecewise-linear network model) but it is now
143 more general (the factor may depend on the source and destination and
144 thus account to different behaviors for intra-node communications and
145 extra-node communications) and is available for CPUs (e.g., if you
146 want to model an affinity as in the "Unrelated Machines" problem in
147 scheduling) and disks (e.g., if you want to model a stochastic
149 - For that, resources can be provided with a callback that computes
150 the activity factor when creating the action.
151 - Example: examples/cpp/exec-cpu-factors
152 - The same mechanism is also available for the latency, which
153 allows to easily introduce complex variability patterns.
156 - Added support to programmatic platform creation in Python.
157 Example: examples/python/clusters-multicpu
160 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
161 This can only be set through the API, not through XML files.
165 - Multiple fixes to ensure reproducibility of tracing
166 - scan/excan can now be replayed
167 - wait action now uses ranks and not pid, as the other ones.
168 - smpi/init and smpi/finalization-barrier are now valid for replays.
169 - exit() is now intercepted by SMPI to avoid premature shutdown of
170 simulation. First non 0 return codes is returned as simulation return
174 * New section "Release Notes" documenting recent and current developments.
175 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
176 * Improvements in API Reference for C++ and Python interfaces.
179 - Make wifi creation compatible with ns-3 version 3.34 too.
181 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
182 (FG: issues on Framagit; GH: issues on GitHub)
183 - FG#77: Search feature of doc is broken (update sphinx theme version)
184 - FG#78: Multiple fixes for SMPI replay:
185 - TI tracing of allotallv/w was outputting wrong values
186 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
188 ----------------------------------------------------------------------------
190 SimGrid (3.28) July 14. 2021
192 The Victoriadagarna Release.
195 - C++ platform interface: Users can now describe their platform directly in C++.
196 This provides greatly flexibility and performance improvement for complex
197 platforms. Main features:
198 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
199 "zones", instead of single hosts. This feature enables the description
200 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
201 - StarZone: new zone with a Star-like topology. The routes are defined
202 as a set of links used to communicate from node to everybody (node<->ALL).
203 - Split-Duplex links: auxiliary method to create split-duplex links in
204 the platform, easing its utilisation. It automatically creates both UP
205 and DOWN links (similarly as done in XML).
206 - Please refer to the documentation and the examples included:
207 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
208 - New plugin: Producer-Consumer with monitor. Just requires to include the
209 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
210 associated example (examples/cpp/plugin-prodcons).
213 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
214 s4u::Io::wait_any(), s4u::Io::wait_any_for().
215 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
216 parameter by reference, instead of a pointer.
217 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
218 Explicitly cancel the activity to get back to previous behavior.
219 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
222 - The default SMPI compiler flags are no more taken from the environment.
223 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
224 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
226 --cfg=smpi/finalization-barrier: which can be used to add
227 a barrier inside MPI_Finalize. This can help for some codes which cleanup
228 data attached to a process, but still used in other SMPI processes.
229 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
230 is active when set to false, to keep going after a small error
231 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
232 which may or may not be problematic in the end.
234 - fix behaviour, as maximum iteration count could be ignored
235 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
236 use sampling when the same kernel is called with a different set of parameters
237 which have an impact on the timing.
238 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
239 It should now work with smpi/auto-shared-malloc-thresh.
240 - Improve error handling and reporting in multiple places
241 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
242 validity checks, truncated messages are now an error, return errors
243 when explicitely deleted handles are reused, ...)
244 - RMA: multiple fixes and stability improvements.
245 - analysis (-analyze flag in smpirun):
246 - SMPI can now report buffer leaks as well as MPI handles leaks,
247 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
248 - if -trace-call-location is used when compiling, SMPI can report
249 origin of leaked handles/buffers
250 - group leaks by type/origin in output message if possible
251 - New implemented MPI calls: MPI_Comm_test_inter
254 - Changed internal implementation of bandwidth factors in network models.
255 Models affected: CM02, LV08 (default), SMPI, IB.
256 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
257 Bandwidth factors are applied to communications to describe that users
258 cannot use 100% of the available bandwidth. For example, the default network model,
259 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
260 behavior was implemented by limiting the bandwidth available in the LMM
261 system for this flow. This may give the false impression that there is
262 bandwidth available for other flows due to its underutilization, especially
263 for the dynamic bandwidth factors used in SMPI models.
264 To avoid this, we have modified the implementation so that each flow uses the
265 maximum physical bandwidth according to the LMM system.
266 However, the actual throughput of the flow seen by the user is defined by
267 the physical bandwidth multiplied by the bandwidth factor.
268 This change impacts on the simulation results for all network models on
269 which we have bandwidth factors configured.
270 *****************************************
271 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
272 *****************************************
273 This change may impact on the timing of your simulation results.
274 Take care when comparing simulations from different SimGrid's
275 versions. Sorry for the inconvenience.
276 - Dynamic network factors: users can configure a callback to define
277 the network factors dynamically. This API is available at
278 simgrid::kernel::resource::NetworkModelIntf.
279 - Users have access to complete information about the current communication
280 to decide which factor to apply. This includes: message size, source and
281 destination hosts, links and zones traversed.
282 - Dynamic factors for both latency and bandwidth.
283 - For more details, see the example in (examples/cpp/network-factors).
284 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
285 deprecated since version 3.24, are no longer supported. Instead, use
286 "wattage_off" and "wattage_per_state".
289 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
292 - New tutorial: Model-checking and formal assessment
293 - New sections: "Demystifying the routing" and "C++ platforms"
294 - Update and improve the part on visualization in MPI and Algo tutorials.
295 - Improve the section on routing: how to define it, how it's used internally
296 - Fix many issues, broken links and missing references in doxygen and Sphinx
299 - Lua platform files are deprecated. Their support will be dropped after v3.31.
302 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
304 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
305 (FG: issues on Framagit; GH: issues on GitHub)
306 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
307 - FG#64: Configuring smpi/IB-penalty-factors
308 - FG#67: Running computation concurrently with MPI_Iallreduce
309 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
310 - FG#71: Segmentation fault on invalid gw_src/gw_dst
311 - GH#322: Issue when an actor kills his host vm
313 ----------------------------------------------------------------------------
315 SimGrid (3.27) March 29. 2021
317 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
319 Main user-visible changes:
320 - All functions marked as deprecated for one year were removed.
321 - Removed all that was related to the "Storage" type in all APIs and bindings.
322 "Disks" have been introduced more than a year ago to replace them for the better.
323 It's time to finish this replacement.
324 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
325 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
326 avoiding "initializer element is not constant" errors with some codes. Please be
327 careful to recompile your MPI codes if switching to this new SimGrid version.
328 - Network model ns-3 now requires version ns-3.28 or higher.
329 - Removed some option aliases that were deprecated since v3.23 (June 2019)
330 - 'verbose-exit' is now 'debug/verbose-exit'
331 - 'simix/breakpoint' is now 'debug/breakpoint'
332 - 'clean-atexit' is now 'debug/clean-atexit'
335 - New functions: Comm::sendto_{init,async} to initiate a communication
336 on between two (possibly remote) hosts.
337 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
338 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
339 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
340 or Mailbox::get_async<void>() if you really want to play with void*.
341 - Unify the interface of Activity::on_{start/activity}
342 - New function: Comm::get_dst_data()
343 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
346 - ptask should now work better with multicore hosts (FG#37).
347 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
348 - Internal cleanups and refactoring around models, to prepare for programmatic
350 - Modernize our tests of the performance models
354 - New section: Configuring the logs from the command line.
355 - New section: Programming API of logs.
356 - New section: Getting realistic results
357 - New section: Examples of SMPI Usage
358 - New part: SimGrid's Internals, with a section on Design goals
359 - Move back to the classical Breathe tool chain, and fix many small glitches.
362 - Handle leak checking: SMPI will now display a message if MPI handles are still
363 present at the end of a simulation. Option smpi/list-leaks:n will display the n
364 first leaked pointers and their types.
365 - smpi/display-allocs option : Display Memory allocations analysis, with hints
366 about the largest allocations. This only works for allocations with malloc/calloc
367 and not compiled with SMPI_NO_OVERRIDE_MALLOC
368 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
369 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
370 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
371 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
372 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
373 - Replay now opens and parses the trace file only once for all ranks when possible.
375 - better handling of MPI_PROC_NULL
376 - MPI_Group_incl was broken in some cases
377 - PAPI tracing now fixed
378 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
379 - SMPI will now fail properly when receive size < sent size, instead of
380 just setting MPI_ERR_TRUNCATE and keep going.
381 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
382 - Remove useless build patches and split build/run patches on the page
383 - Build in parallel when possible
384 - Test on yum-based systems
385 - New nightly tested Proxy-apps:
386 - OpenMPI mpi-test-suite
387 - Coral MCB, LSMS, Qbox, CLOMP
388 - ECP/Mantevo Cloverleaf3D, Ember
389 - Codevault bhtree_mpi
392 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
393 - Start to overhaul the implementation to modernize it (still ongoing).
396 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
397 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
398 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
399 - Deprecate xbt_dynar_to_array().
400 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
402 C binding and interface:
403 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
404 - Constness requirement has been lowered for parameter argv of sg_actor_start()
405 and sg_actor_create() It's now (char* const*). The previous version is still
406 available as sg_actor_start_() and sg_actor_create_().
408 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
409 (FG: issues on Framagit; GH: issues on GitHub)
410 - FG#37: Parallel tasks are limited to 1 core per host
411 - FG#62: Running "smpirun -replay" on large networks
412 - FG!46: Fix a few potential memory leaks in SMPI colls
413 - GH#343: Wrong consumed energy when an exec timeouts
415 ----------------------------------------------------------------------------
417 SimGrid (3.26) December 15. 2020
419 The Release release (the French lockdown was eased today).
421 Important user-visible changes:
422 - SimGrid now requires a compiler with C++14 support.
423 Sibling projects should upgrade their FindSimgrid.cmake
424 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
425 several users had difficulties to understand issues when using high bandwidth or
426 small latency events. The new value was already the default for SMPI and
427 should not cause too much performance hit. It can change some simulation
428 timings if simulation was skipping a lot of small events (which was actually
429 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
430 - On some network models (L07, CM02, LV08), an implicit loopback link is created
431 for local communication. Default bandwidth was 500MBps and latency was 15µs,
432 which was rather bad. Default values have been changed to 10GBps and 0 latency.
433 Options network/loopback-lat and network/loopback-bw have been added to adapt
434 these values if needed. Network and routing models already providing separate
435 loopbacks (clusters and such) are not impacted by this change.
436 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
437 metadata that are added to each MPI message by MPI libraries (tag,
438 src/dest, and communicator information). This may change simulation times
439 for very small message sizes, but should be more realistic.
442 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
443 - New plugins: wifi_energy and link_load. See the associated examples.
444 - New examples: wifi networks, and communication suspend.
447 - LTO is now enabled for Intel/clang compilers.
448 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
449 Setting it to "auto" will use all available cores, while setting it to n will
450 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
451 - Remove obsolete runtime option 'contexts/parallel-threshold'.
452 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
453 has been kept for compatibility.
454 - Finally remove obsolete snake_case() aliases for runtime options.
455 - Further improve the documentation.
458 - the on_exit() of each actor is also executed when the simulation deadlocks.
459 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
460 An example is provided for s4u::Comm but it should work with Exec and Io.
463 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
464 https://framagit.org/simgrid/SMPI-proxy-apps
465 - MPI/IO: supports names starting with "./"
466 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
468 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
469 building shared libraries, this may cause issues, so environment variable
470 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
471 - most temporary files should now be created in /tmp dir (or equivalent).
472 If this one does not allow execution of code (noexec flag), this may cause issues.
473 Please use another tmp directory (using TMPDIR or equivalent system variable)
477 - A large refactoring is ongoing, but this should be transparent for users so far.
480 - Drop xbt_str_split().
482 C binding and interface:
483 - The return type of the sg_actor_on_exit() callbacks is now 'void'
484 (int value was previously ignored)
485 - Many C functions were renamed to follow a common naming style
486 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
487 names are kept until version 3.30.
488 - Many MSG tests were converted to the new S4U's interface in C, that
489 was extended for that.
491 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
492 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
493 - FG#41: Add sg_actor_create C interface
494 - FG#43: xbt::random needs some care
495 - FG#48: The Impossible Did Happen (yet again)
496 - FG#50: Suspending an actor executed at the current timestamp fails
497 - FG#52: Zero-seconds timeout : "That's in the past already"
498 - FG#53: Crash while using ns-3 network model
499 - FG#54: How to suspend a comm?
500 - FG!22: Stochastic Profiles
501 - FG!24: Documentation and fix for xbt/random
502 - FG!35: Add a modeling hint for parallel links in doc
503 - FG!36: [xbt/random] Read/Write the state of the RNG
504 - FG!42: Link cumulated load plugin
505 - FG!44: Test condition variable wait_until+notify_all
506 - GF#18137: Allow different stack sizes?
507 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
508 - GH#139: Allow pthread creation in SMPI
509 - GH#336: Packet-level simulation using SMPI?
510 - GH#345: Error in the LMM while migrating a VM
511 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
512 - GH#352: pip install / python setup.py install fail to find pybind11
513 - GH!337: Fix link_energy plugin for wifi platforms
514 - GH!339: Add Mailbox set_receiver method to python binding
515 - GH!344: Cast hugepages macros parameters to int64
517 ----------------------------------------------------------------------------
519 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
521 The Palindrom Day Release.
523 Important user-visible changes:
524 - Improve the Python usability (stability and documentation).
525 - A nasty synchronization bug was ironed out, see also below.
526 - Python's doc was organized as a separate tree, now integrated with C++.
527 - C bindings of S4U were not part of the doc.
528 - The C++ doc was also improved as methods are now split by theme.
529 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
530 - This is mandatory to use the Java bindings.
531 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
532 - SimDAG++: Automatic dependencies on S4U activities (experimental)
533 - Some features are already implemented but not all of them
534 - Cannot block an activity until it's scheduled on a resource
535 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
536 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
539 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
540 - Drop support for Viva/Triva (old visualization tools), hence removed
541 TRACE_get_node_types() and TRACE_get_edge_types() functions.
544 - Require pybind11 version 2.4.
545 - Greatly improve locking during context switches with Python.
546 - Declaring Python callbacks to this_actor::on_exit is now working.
547 - No more tests are failing currently, even on win32.
550 - Actor: Merge signals on_migration_start/end into on_host_change
551 - Actor: Rename migrate() into set_host()
552 - Disk: Allow users to get the read and write nominal bandwidth values
553 - Exec: Implement wait_for(timeout)
554 - Io: Implement wait_for(timeout)
557 - Parse errors now raise a simgrid::ParseError that you may want to catch.
560 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
561 been renamed to fix a spelling error. As usual, the previous definitions are
562 kept with a deprecation warning. However, to avoid runtime errors, it is now
563 forbidden to override these deprecated methods in a derived class. Please use
564 the new names immediately if you need to override them.
567 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
568 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
569 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
570 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
571 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
572 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
573 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
574 - New module: random, providing classical random numbers generators.
577 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
578 allocations' size above which they are considered "shared" by default (as if
579 they were performed through SMPI_SHARED_MALLOC macros).
580 Default = 0 = disabled feature.
581 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
582 This can cause some troubles if codes are already overriding these. If this
583 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
584 help, but will make this feature unusable.
585 - Input check is now performed much more consistently and outputs precise
586 warnings for each error found
587 - New supported calls : MPI_File_set_view, MPI_File_get_view
588 - MPI I/O now only supports disks, as storage is discontinued.
589 - MPI I/O allows opening files without specifying absolute path
590 - Note : SMPI configuration options are now initialized by smpi_init_options()
591 call, which should be called if SMPI is not being used through smpirun.
593 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
594 - FG#9: Python bindings crashing
595 - FG#39: Missing s4u::Comm::wait_any_for example
596 - FG#42: Add support for ThreadSanitizer (TSan)
597 - FG!19: Removing RngStream
598 - FG!20: A module for RNG calls
599 - FG!21: Choice between ad-hoc and standard distributions implementations
601 - FG!25: Fix link in Mutex doc
602 - FG!26: Fix links in SMPI interface doc
603 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
604 - GH#305: Unscheduled tasks are still excuted
605 - GH#313: smpirun: manual outdated w.r.t. --help
606 - GH#321: [S4U] Get task remaining work ratio
607 - GH#323: Crash when an actor turn off his physical host
608 - GH#335: Missing links on dragonfly example svg
609 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
610 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
612 ----------------------------------------------------------------------------
614 SimGrid (3.24) October 9. 2019
616 The Clean Disk Release.
619 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
620 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
621 thus becomes the default resource to execute I/O operations. Examples
622 and the file system plugin have been modified to use disks instead of
623 storage elements. Storage can still be used but is doomed to disappear.
624 Users are thus adviced to update their simulators accordingly.
625 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
626 for consistency with pthread_barrier_wait()
627 - Host::get_englobing_zone() returns the englobing netzone
628 - Actor::on_destruction is now called in the destructor
629 Actor::on_termination new signal called when the actor terminates
631 - Global signals are now part of the Engine:
632 - on_platform_creation: after config settings, before the XML parsing
633 - on_platform_created: right after the XML parsing
634 - on_time_advance: each time the clock advances
635 - on_simulation_end: after simulation, before cleanups
636 - on_deadlock: as the name implies.
638 - sg_{actor,host,link}_{data,data_set}() now all exist.
639 Use them to attach user data to the object and retrieve it.
642 - Introduce an experimental Wifi model. It sounds reasonable
643 according to the state of the art, but it still has to be properly
644 validated, at least against ns-3 if not against reality.
645 - Improved the usability of ns-3. Several bugs were ironed out.
646 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
647 It is now expressed as 'idle:epsilon:allCores' to properly model the
648 consumption of non-whole tasks on mono-core hosts. More info in the doc.
651 - convert a new set of functions to the S4U C interface and move the old MSG
652 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
655 - Fortran bindings for DVFS have been removed.
656 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
657 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
658 - SMPI can now be selected by cmake's find_module(MPI) with
659 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
660 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
661 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
662 - trace-call-location can be used with TI traces, and replayed, and is compatible
663 with smpi/comp-adjustment-file.
664 - sleep events are now correctly traced/replayed.
665 - Default for trace-call-location is now to use file names and not full paths.
666 To revert to previous behavior (in case of collision of filenames), option
667 "smpi/trace-call-use-absolute-path" can be set to yes.
670 - Use the included xxHash as an hash implem when C++14 is usable.
671 - Option model-checker/hash was removed. This is always activated now.
672 - New option smpi/buffering controls the MPI buffering in MC mode.
673 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
674 This is useful to check for MPI compliance.
677 - New section on plugins: how to define a new one, and existing ones.
680 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
681 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
684 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
685 and <mount> tags. The rationale is finish to decouple what should be handle
686 by SimGrid kernel (i.e., resources and activities) from plugins built in
687 user space (e.g., the file system plugin). This new tag currently accepts
688 only three attributes (id, read_bw, and write_bw). All the other information
689 that was declared with the storage related tags now has to be expressed as
690 properties. An example of platform using this new tag is available at
691 examples/platforms/hosts_with_disks.xml
694 - 'expect signal' can now accept more than one potential signal.
696 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
697 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
698 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
699 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
700 - FG#31: per-actor data
701 - FG#34: SG_BARRIER_SERIAL_THREAD?
702 - FG#35: model-checker does not like buster-produced binaries
703 - FG!13: MC: complete workaround in the error msg seen on modern systems
704 - FG!15: execute_flops now logs compute
705 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
706 - FG!17: ns-3: unblock the right number of communications + others issues
707 - FG!18: Improving the performance of the ns-3 bindings
708 - GH#189: Energy consumption of parallel tasks
709 - GH#219: Error in the throughput of TCP transfer
710 - GH!330: Minor polishing of English: Configuring_Simgrid
711 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
712 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
713 - GH!334: Rework energy plugin, again
715 ----------------------------------------------------------------------------
717 SimGrid (3.23.2) July 8. 2019
720 - Nicer introduction page.
721 - Migrate the "Deploy your application" page to the new doc.
722 - Move Java as a subtree of MSG.
725 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
728 - Drop xbt_dynar_sort_strings().
731 - Really fix FG#26: Turning off a link should raise NetworkFailureException
732 - FG#27: Wrong exception thrown to wait_any when link is turned off
733 - GH#328: Java: Canceling multiple tasks in a single vm/host
735 ----------------------------------------------------------------------------
737 SimGrid (3.23) June 25. 2019
739 The Exotic Solstice Release.
742 - SunOS and Haiku OS support. Because exotic platforms are fun.
743 - Stop setting random seed with srand48() at initialization.
744 - Use addr2line as a fallback for stacktraces when backtrace is not available.
745 - Build option -Denable_documentation is now OFF by default.
746 - Network model 'NS3' was renamed into 'ns-3'.
749 - Simgrid can now hopefully be installed with pip.
752 - wait_any can now be used for asynchronous executions too.
755 - New log appenders: stdout and stderr. Use stdout for xbt_help.
756 - Drop xbt_dict_dump.
759 - 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
760 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
761 - MPI interface is now const correct for input parameters
762 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
763 - Fortran bindings for async collectives.
764 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
767 - Remove option 'model-check/record': Paths are recorded in any cases now.
768 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
769 always sparse. That's an immense gain in memory, and can even be
770 faster because of cache effects. So there is no need to clutter the
771 code to allow the user to go for the unefficient mode.
774 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
775 models proved to be more accurate than these old experiments.
777 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
778 - FG#1: Broken link in error messages
779 - FG#2: missing installation documentation
780 - FG#3: missing documentation in smpirun
781 - FG#6: Python bindings not available on PyPI
782 - FG#7: simple cmake call requires doxygen
783 - FG#8: make python bindings an optional dependency
784 - FG#10: Can not use MSG_process_set_data from SMPI any more
785 - FG#11: Auto-restart actors forget their on_exit behavior
786 - FG#12: -Denable_lto=OFF doesn't disable LTO
787 - FG#13: Installs unstripped file 'bin/graphicator'
788 - FG#14: Installs the empty directory 'doc/simgrid/html'
789 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
790 - FG#17: Dead link in doc (pls_ns3)
791 - FG#20: 'tesh --help' should return 0
792 - FG#21: Documentation link on http://simgrid.org/ broken
793 - FG#22: Debian installation instruction are broken
794 - FG#26: Turning off a link should raise NetworkFailureException exceptions
795 - GH#133: Java: a process can run on a VM even if its host is off
796 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
797 - GH#326: Valgrind-detected error for join() when energy plugin is activated
799 ----------------------------------------------------------------------------
801 SimGrid (3.22) April 1. 2019
803 The Easter Christmas Release (this one is somewhat late).
806 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
807 - Not complete yet: asynchronous activities (amongst others) are still missing
808 - Still ongoing: the interface may change in the future. We need more
809 testers! Please report any glitches.
810 - No new project using Java should start now. Please switch to Python.
813 - Some of the internal cleanups may lead to speed improvements:
814 - The hard limitation on the amount of simulated actors with
815 Java+Mac was removed. Now, the available memory is the only limit.
816 - Our refcounting was tidyied, leading to 10% speedups in some cases.
817 - We are still working on making our code robust to the actor kills
818 and hosts' churn. Things are improving, but it's not perfect yet.
819 - Replaced our own code to display a backtrace (that was forking addr2line)
820 with the Boost.Stacktrace library.
821 You won't see your backtraces without this optional dependency.
822 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
823 - Stop setting random seed with srand() at initialization.
826 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
827 XML file version remains 4.2 since old files are still compatible.
830 - Process termination which was broken at version 3.21 has been repaired.
831 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
832 - Hide the examples into examples/deprecated. New users should use Python.
835 - Drop MSG_process_create_from_stdfunc() from the API.
836 This C++ function was a pimple in the C API, made necessary at some
837 point by the Java bindings. This is fixed now.
838 - Hide the examples into examples/deprecated. New users should use S4U.
839 - MSG_process_create and MSG_process_attach now crash if the host used
840 to run this process is off.
841 - Fix the protype of MSG_process_on_exit()
842 Now use (int,void*) callbacks instead of (void*,void*) ones.
843 The implementation was ways too messy to actually work, I guess.
846 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
847 too often and provide more accurate timings.
848 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
849 - MPI_Alltoallw support
850 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
851 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
852 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
853 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
855 - Basic implementation of generalized requests (SMPI doesn't
856 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
859 - Drop sg_cmdline. Please use xbt_cmdline instead.
860 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
861 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
862 OsSemaphore is implemented in a portable way with C++11 threads.
863 This should allow much more threads to be created at the same time,
864 allowing Mac and Java users to simulate many more actors.
865 - Implement the 'thread' factory with std::thread instead of xbt ones.
866 It is not possible to set the stack size with threads anymore, but
867 -fsplit-stack is the way to go nowadays when using threads.
868 - Drop the xbt_os_thread_t module (now unused)
869 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
870 - Drop xbt_str_join_array().
871 - Drop cunit, use Catch2 instead.
874 - Many cleanups in the kernel::activity namespace. This was long
875 overdue, and shall open the path to many future endeavors.
878 - #132: Java: a process can not shut down its own host
879 - #220: S4U: detached send causes exception if sender terminates soon after sending
880 - #261: Document the parameters of parallel execution's constructor
881 - #300: [s4u] BarrierPtr is missing
882 - #314: SMPI args internal cleanup
883 - #316: Fix a bug related to the CPU utilization of multi-core VM
884 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
885 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
886 - #325: Turning off a host has different behavior on sleeping actors and computing actors
888 ----------------------------------------------------------------------------
890 SimGrid (3.21) October 3. 2018
892 The Restarting Documentation (TRD) Release.
895 - Convert the user manual to Sphinx for improved usability.
896 Unstable version now hosted on https://simgrid.frama.io/simgrid
897 Many glitches remain; stay tuned.
898 - Restrict the Doxygen reference API to the installed header files.
899 (documenting all internals this way is overkill + doxygen scales badly)
900 - New tutorial on S4U (highly inspired from the old MSG one, but with
901 a git repository to fork as a starting point, and a docker image)
902 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
903 - The Developper manual is still to be converted (not compiled until then)
904 - Some parts are still missing in this conversion (such as the
905 platform and deployment sections) while others were blindly converted
906 and would need more love (such as the configuration flags).
907 Things will certainly further improve in the future releases.
910 - Fully reimplement auto_restart mechanism that was utterly broken
911 (fix #22, #131 and #281 that were all very related)
912 - Implement semaphores (fix #303 and #291)
913 - When creating an actor from a function and its parameters,
914 move-only parameters are not allowed anymore, as it would prevent
915 the actor to be restartable if its parameters are consumed on run.
916 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
917 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
918 IO operations that can be started, waited for, or canceled as a regular
920 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
924 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
925 - New host variable: 'core_count'
928 - Remove xbt_os_thread_specific features
929 - Remove portability wrapper to condition variables
930 - Remove xbt_os_thread_yield()
933 - MPICH collective selector now mimics MPICH 3.3b
934 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
935 - MPI_Init(NULL,NULL) is now allowed
936 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
937 - Command-line arguments should now be handled properly in Fortran simulations
938 - MPI Attributes and related callbacks should work in Fortran
939 - Apps using C + Fortran should now work
940 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
941 - Support MPI_Error_String
944 - Due to an internal bug, Msg.run() must now be your last line.
945 We hope to fix it in a future release, and we are sorry for the inconvenience.
948 - #22: Process autorestart seem to only work with CAS01 cpus
949 - #56: Feature request: dynamic SMPI replay
950 - #93: simgrid should not eat --help
951 - #111: How to change the loopback link characteristics?
952 - #116: Platform cluster radical
953 - #117: Platform cluster/host order
954 - #129: Replay_multiple on steroid
955 - #131: Java : autostart does not work
956 - #143: Setting a breakpoint at a given time
957 - #235: Network energy models should be integrated
958 - #264: Add ptask L07 resource tracing
959 - #271: Dynamic and manual replay of SMPI traces
960 - #279: Breakpoints option uses time deltas instead of absolute time
961 - #281: Daemonized actor no longer daemonized after an auto-restart
962 - #282: TIT convertor erase traces when using absolute path in the trace list file
963 - #285: segfault when a s4u actor kills itself with msg process tracing activated
964 - #286: Pajé traces are not exposing the number of cores
965 - #287: Command-line parsing should stop when encountering '--'
966 - #288: MPI_Init(NULL, NULL)
967 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
968 - #290: Method to check if a mailbox has a message ready to be consumed
969 - #291: [s4u] semaphores do not seem available yet
970 - #294: Alltoall collective implementation of mpich
971 - #295: Starting a Host as OFF is not supported
972 - #296: DTD too permissive
973 - #299: Add a s4u barrier example
975 ----------------------------------------------------------------------------
977 SimGrid (3.20) Released June 24. 2018
979 The proxy snake_case() release.
982 - The whole API is now using snake_case() instead of mixing it with camelCase.
983 Compatibility wrappers in place for one year.
984 - We took the opportunity to increase the consistency of the API.
985 Please report any remaining glitches.
986 - New function simgrid::s4u::Host::get_actor_count:
987 Returns the number of actors running on a specific host.
990 - SMPI is now tested with ~45 proxy apps from various sources, with none or
991 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
992 - Replay: The replay file has been re-written in C++.
993 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
994 supported. They are stored in the trace and used when replayed.
995 - Basic support of MPI_Cancel. Robustness not guaranteed.
996 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
997 (only for MPI_COMM_TYPE_SHARED).
998 - New option: smpi/privatize-libs, to add external shared libs to be privatized
999 by SMPI. They will be copied locally and loaded separately by each process.
1000 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
1001 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
1002 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
1003 with internal index optind with multiple processes. Only works if MPI_Init has
1004 already been called.
1005 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
1006 syntax without preprocessing tricks.
1009 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
1010 behavior tracing is no longer limited to MSG
1011 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
1012 this means that traces now need to be updated or re-obtained; both Irecv
1013 and Isend lines in your traces have now in total 5 mandatory fields:
1014 <rankid> <command> <to/from rankid> <tag> <size>
1015 To update your traces, it suffices to add a 0 for the tag here.
1016 - TIT now also supports waiting for a distinct request via MPI_Wait.
1017 Wait/Test now wait for a specific request, not just the last one that was
1018 issued. This unfortunately means another update, because we need to
1019 identify which request you want to wait for. We do this via the
1020 triplet (sender, receiver, tag), which needs to be added:
1021 <rankid> <command> <sender> <receiver> <tag>
1022 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
1023 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
1026 - The deprecation of MSG is ongoing (but this should not impact you).
1027 Many MSG functions are now simple wrappers on the C API of S4U. If
1028 you wish to convert your code to S4U, find the S4U counterparts of
1029 your MSG calls in src/msg/msg_legacy.cpp.
1030 - MSG can still be used, but won't evolve anymore.
1033 - Allow to run the Link energy plugin from the command line with
1034 --cfg=plugin:link_energy
1035 - Rename Energy plugin into host_energy
1036 - Rename Load plugin into host_load
1039 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
1040 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
1041 - Improve performance when handling timeouts of simix synchros.
1044 - Config: the C API is now deprecated (will be removed in 3.23), and
1045 the C++ API has been slightly improved.
1048 - Fix several build issues on OSX.
1049 - Move simgrid_config.h to simgrid/config.h (old header still working)
1052 - #143: Setting a breakpoint at a given time
1053 - #150: Inconsistent event names in SMPI replay
1054 - #258: daemonized actors hang after all non-daemonized actors have completed
1055 - #267: Linker error on unit_tmgr
1056 - #269: SMPI: tracing of MPI_Wait/all/any broken
1057 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
1059 ----------------------------------------------------------------------------
1061 SimGrid (3.19.1) Released March 22 2018
1063 The Fixed ABI Release.
1065 As you may know, we are currently refactoring SimGrid in deep.
1066 Upcoming SimGrid4 will be really different from SimGrid3: modular,
1067 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
1069 Our goal is to smooth this transition, with backward compatibility
1070 and automatic update paths, while still progressing toward SimGrid4.
1072 SimGrid remains open during works: The last pure SimGrid3 release was
1073 v3.12 while all subsequent versions are usable alpha versions of
1074 SimGrid4: Existing interfaces remain unchanged, but the new S4U
1075 interface is budding and the internals are deeply reorganized.
1077 Since 2015, we work hard to reduce the changes to public APIs. When
1078 we need to rename a public library symbol in S4U, we let your
1079 compiler issue an explicative warning when you use the deprecated
1080 function. These messages remain for four releases, i.e. for one full
1081 year, before turning into an error. Starting with v3.15, your can
1082 also adapt to API changes with the SIMGRID_VERSION macro, that is
1083 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
1085 Starting with this v3.19.1, our commitment to reduce the changes to
1086 the public interfaces is extended from the API to the ABI: a program
1087 using only MSG or SimDag and compiled against a given version of
1088 simgrid can probably be used with a later version of SimGrid without
1089 recompilation. We will do our best... but don't expect too much of
1090 it, that's a really difficult goal during such profund refactoring.
1092 The difference between v3.19 and v3.19.1 is that the former was
1093 accidentally breaking the ABI of MSG, while the later is restoring
1096 S4U and kernel APIs will still evolve until SimGrid4, with one-year
1097 deprecation warnings as currently. In fact, cleaning up these
1098 interfaces and converting them to snake_case() is one release goal of
1099 v3.20. But don't worry, we are working to smooth this upgrade path.
1101 In summary, new projects should start with S4U to benefit of the
1102 future, but old MSG projects should still be usable with no change.
1104 ----------------------------------------------------------------------------
1106 SimGrid (3.19) Released March 20 2018
1108 The Moscovitly-cold Spring release.
1110 SMPI user-visible improvements:
1111 - Support of MPI_Type_create_subarray
1112 - Unimplemented MPI calls now abort (instead of issuing a warning)
1113 Some harmless calls around errhandlers still only issue a warning.
1114 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
1115 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
1116 is now possible (mmap privatization was refactored to this end).
1117 Adding new SMPI instances should also be technically doable now.
1119 SMPI internal cleanups:
1120 - Removed index notion from SMPI -- no more "getPid() - 1"!
1121 In tracing files, actors are called rank-<PROCESS-ID> but these are
1122 NOT the real ranks (tracing several communicators is not supported yet).
1123 - Deprecate smpi_process_index() (will be removed in 3.22)
1124 - Better testing on classical proxy apps, through an external project:
1125 https://github.com/simgrid/SMPI-proxy-apps/
1128 - Use a graphical TOC to make it easier to find the documentation you need
1129 - Revamp the MSG tutorial
1132 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
1133 - Remove parameter of MSG_process_killall().
1134 Resetting the PID was bogus anyway (several actors could have the same PID).
1137 - Execution->setHost() can be called after start(), to migrate it.
1138 - Comm::test_any() is now implemented.
1139 - s4u::Actor now has onCreation() and onDestruction() signals.
1140 - Install some kernel header files for the users' plugins and more.
1141 Warning, their API is really not stable yet.
1142 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1143 include/simgrid/kernel/routing/*Zone.hpp
1146 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1148 Build System and other cleanups:
1149 - Remove unused run-time parameter "tracing/onelink-only".
1150 - SimGrid now works with intel compilers (MC still buggy)
1151 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1152 - tesh kills subprocesses on timeout (but not on windows)
1155 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1158 - #194: Feature request: simgrid::s4u::Comm::test_any()
1159 - #245: migrating an actor does not migrate its execution
1160 - #253: Feature Request: expose clusters as objects
1161 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1162 - #255: Tesh broken on Windows
1163 - #256: Modernize FindSimGrid.cmake
1164 - #257: Fix (ab)use of CMake install
1166 ----------------------------------------------------------------------------
1168 SimGrid (3.18) Released December 24 2017
1170 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1172 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1173 - Support for MSG should not be removed from SimGrid before 2020,
1174 but future evolutions will be limited to the integration of
1175 user-provided patches.
1178 - s4u::onDeadlock() signal, run before stopping the simulation.
1179 - s4u::Exec: asynchronous executions (abstraction of background
1181 - s4u::Host->getLoad() returns the achieved speed in flops/s
1184 - Link::name() is deprecated and will be removed in v3.21, use
1185 Link::getCname() instead.
1186 - Mailbox::getName() changed to return a std::string, use
1187 Mailbox::getCname() instead to get a char*.
1188 - Storage::getName() changed to return a std::string, use
1189 Storage::getCname() instead to get a char*.
1190 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1191 to have both allocation and deallocation of the map in user space.
1194 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1195 The amount of flops remaining to do is a vector, not a scalar.
1196 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1197 name implies on both sequential and parallel tasks.
1198 - Both changes fix GitHub's #223 using PR #237 as a basis.
1199 Thanks Michael Mercier.
1200 - Most examples were converted to S4U and hidden elsewhere at tests.
1201 Rationale: we still want MSG to work; we want newcomers to use S4U.
1204 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1205 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1208 - Switch to the faster dlopen privatization mechanism by default
1209 - Documentation improvments
1212 - Live migration is getting moved to a plugin. Dirty page tracking is
1213 the first part of this plugin. This imply that VM migration is now
1214 only possible if one this function is called:
1215 - C/MSG: MSG_vm_live_migration_plugin_init()
1216 - C/C++: sg_vm_live_migration_plugin_init()
1217 - Java: Msg.liveMigrationInit()
1218 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1222 - Define class simgrid::xbt::Path to manage file names.
1223 - Removed unused functions:
1224 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1225 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1226 xbt_graph_export_graphviz()
1227 - xbt/str.h: xbt_str_join()
1228 - Remove unused datatypes:
1229 - xbt/heap.h: use std::priority_queue or boost::heap instead
1230 - xbt/swag.h: use boost::intrusive::list instead
1233 - New link_energy plugin for the consumption of the links.
1234 - All of the operations on files and storage contents have been
1235 packaged into a plugin (src/plugins/file_system). The current
1236 public interface can be found in
1237 include/simgrid/plugins/file_system.h
1238 To use these functions you now have to initialize the plugin by
1239 calling MSG_storage_file_system_init() just after calling
1240 MSG_init() or sg_storage_file_system_init() just after creating
1244 - Remove the undocumented/untested tag <include>
1247 - Remove viva specific tracing as the tool is no longer maintained
1250 - #248: Kill of finished processes leads to segfault
1251 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1252 - #239: please implement signal s4u::onDeadlock()
1253 - #230: segfaults when exit() without run()
1254 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1255 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1256 - #222: Actor::kill() doesn't really kill and segfaults
1257 - #221: odd LMM warning when killing an actor
1258 - #120: Memory leak when the processes are forcefully killed
1260 ----------------------------------------------------------------------------
1262 SimGrid (3.17) Released October 8 2017
1264 The Drained Leaks release: (almost) no known leaks despite the tests.
1266 Even very long simulations will preserve your memory: our testsuite
1267 stresses SimGrid for over 45mn on fast machines for a coverage of
1268 over 80%, and there is only one single known leak, of about 4kb.
1272 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1273 handler may be given to free resources if the comm cannot be completed.
1274 - this_actor::send(mailbox) is now mailbox->put()
1275 - New: simgrid::s4u::Comm::wait_all()
1276 - New: Host.execute() for remote executions.
1279 - Unused option network/sender-gap is removed.
1282 - Deprecate MSG_task_isend_with_matching(): This unused feature
1283 really complicates our internals. Will be removed in v3.20.
1286 - Improved context termination. It is now possible to free resources, even
1287 when a process is forcibly killed.
1288 - ContextBoost: add support for Boost versions above 1.61.
1291 - Bring back run-time option --cfg=exception/cutpath to remove exception
1293 - Removed unused functions:
1294 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1295 xbt_str_rtrim(), xbt_str_trim().
1296 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1299 - Removed header files obsolete since SimGrid 3.12:
1300 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1302 - #3: SD_exit should be made optional
1303 - #120: Memory leak when the processes are forcefully killed
1304 - #159: Threading test regression in Actor refcounting
1305 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1306 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1307 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1308 - #191: VM migration and pstate
1309 - #192: Updating the energy consumptions of all hosts crashes with VMs
1310 - #195: All actors have PID=0 in the logs
1311 - #204: Sometimes segfault with thread contexts and mmap privatization
1312 - #222: Actor::kill() doesn't really kill and segfaults
1313 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1315 ----------------------------------------------------------------------------
1317 SimGrid (3.16) Released June 22. 2017.
1319 The Blooming Spring Release: developments are budding.
1322 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1323 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1324 - Java JAR file should be finally fully working on Mac OSX too.
1326 XML platforms: Switch to platform v4.1 format.
1327 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1328 - <zone> can be used as a synonym for the now deprecated <as>
1329 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1330 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1331 - <actor> can be used as a synonym for the now deprecated <process>
1332 - state_file and avail_file periodicity is now easier to express
1333 (check the documentation) the old behavior should still work.
1334 * Storage: not backward compatible, but it seems that nobody ever used it
1335 - Remove attribute 'content_type' of <storage_type>: was never used
1336 - Make attribute 'model' of <storage_type> optional: for future usage
1337 - Remove Bconnection model property: was never (in)validated. Replaced by
1338 the maximum of the read and write bandwidth as a resource constraint.
1341 - New and Backwards Compatibility break:
1342 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1343 When one wants to get the list of tasks whose states have changed during a
1344 simulation round, s/he has to allocate and free a dynar and use it as
1345 argument to this function. The former SD_simulate (double how_long)
1349 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1350 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1353 - New function to update the consumption of all hosts at once.
1354 - Fix the model for multi-core hosts, linear in the amount of busy
1355 cores with an abnormality for pIdle (see paper).
1356 BUG: Still not satisfactory for ptask on mono-cores.
1359 - The netzone are now available from the MSG API.
1360 The old names still work, but are now deprecated.
1363 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1364 - New option: smpi/keep-temps to not cleanup temp files
1365 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1366 mallocs mode is used, allows to change the size of the fake file used
1367 (default 1MB), to potentially limit the number of mappings for large runs.
1368 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1369 - Fortran ifort and flang compilers support
1370 - New RMA calls supported (experimental):
1371 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1372 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1373 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1374 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1375 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1376 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1379 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1380 - Behind the scene, ActivityImpl are now automatically refcounted.
1381 Removing this nasty bottleneck will greatly simplify our internals.
1384 - Replay: New function xbt_replay_action_get():
1385 Retrieve the function previously associated to an event type.
1386 - DROPPED FUNCTION: xbt_str_varsubst()
1387 - DROPPED MODULE: strbuff. We don't need it anymore.
1388 - DROPPED MODULE: matrix. We don't need it anymore.
1389 - DROPPED MODULE: lib. We don't need it anymore.
1391 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1393 SimGrid (3.15) Released March 22. 2017
1395 The Spring Release: continuous integration servers become green
1396 We fixed even the transient bugs on all target architectures:
1397 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1398 Windows, FreeBSD, NetBSD.
1400 - Rename NetCards to NetPoints.
1401 This was intended to help NS3 users, but that's not a netcard.
1402 That's a point in the routing algorithm, let's avoid wrong simplifications.
1405 - New: MSG_process_yield(). Stop and yield to other processes.
1406 - New: MSG_process_daemon(). Daemon processes are automatically killed
1407 when the last non-daemon process terminates
1408 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1409 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1410 to make room for the upcoming network energy plugin.
1411 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1414 - Ensure that an actor can kill itself with Process::exit()
1415 - Kill the obscure NativeException. Nobody want to survive the issues
1416 it denotes, so use JniException that is a RuntimeException (not to
1417 be caught explicitly).
1418 - Partial bug fix in initialization. SimGrid flags on command line were
1419 consumed at C level but stayed in the original Java String[] args.
1420 This could mess users' args[i] if SG flags were not put at the end of
1422 The SimGrid flags are now removed from the Java arguments. However,
1423 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1424 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1425 It might be if you have --log or --cfg flags in the command line.
1426 - Fix numerous memleaks all around the place. In particular, around VMs.
1430 - s4u::onPlatformCreated: right before the simulation starts
1431 - s4u::onSimulationEnd: right after the main simulation loop
1432 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1433 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1434 event from the availability_file changes the avail speed.
1435 - Links are now usable from s4u
1436 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1437 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1438 - New examples: The conversion of MSG examples to S4U has begun
1439 - Actors: create, kill, migration, and suspend
1440 - Applications: master-worker and token-ring
1441 - Action replay: communications and storage
1442 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1445 - Backwards Compatibility breaks
1446 - The SD_route_get_list and SD_route_get_size functions have been removed.
1447 They are replaced by the unique
1448 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1449 The route from src to dst is built in the links dynar whose size can
1450 be obtained with xbt_dynar_length.
1451 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1452 sg_host_route_bandwidth and sg_host_route_latency.
1453 Macros ensure the backwards compatibility, but you should fix your code
1456 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1457 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1458 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1459 We don't support 100% of it, but it should be enough. Contact us if not.
1460 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1461 naive and probably inaccurate)
1462 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1463 mapped onto a unique small file using some system magic.
1464 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1465 - Fortran: cleanups, fixes, support of user-added operations
1466 - RMA: MPI_Accumulate are applied in correct order.
1467 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1468 - Better support for MPI_IN_PLACE
1469 - Support for MPI_Win attrs and keyvals.
1470 - Support MPI_Comm_group_create, MPI_Type_size_x
1471 - MPI ops should be properly applied to all allowed types
1472 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1473 algorithm by a non failing one from mpich.
1476 - Kill the fifo data container: we don't use it anymore.
1479 - A <cluster> can now be created with different speed values to represent pstates.
1480 see examples/platforms/energy_cluster.xml for an example.
1482 SimGrid (3.14.159) Released December 28. 2016
1484 The previous release was only a crude approximation of a Pi release;
1485 this one should be a bit better.
1487 - Revert a last minute change that broke on Mac OSX.
1488 - Fix the changelog and NEWS files.
1489 - Improve the documentation, in particular of the routing module.
1490 - Integrate some patches flying around in Debian.
1492 SimGrid (3.14) Released December 24. 2016
1495 * The whole documentation was reworked and reorganized. There is
1496 still much room for improvement here, but we're on it.
1499 * We now need python3 for our testing infrastructure.
1500 * Model checking needs libevent
1503 * Now works on FreeBSD
1506 * Privatization now works on FreeBSD
1508 * Privatization is now activated by default in smpirun
1509 You can turn it off with -no-privatize if something goes wrong.
1511 * Call-location tracing for SMPI.
1512 You can add the exact location (filename / linenumber) of an MPI call to
1513 your trace files and slow down or speed up the simulation between two
1514 consecutive calls by using an adjustment file (see the documentation).
1516 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1517 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1518 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1519 values were ignored in some cases. The timings of these functions can now
1520 be significantly different.
1522 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1523 smpi/running-power is renamed to smpi/host-speed
1525 * smpi/grow-injected-times option to enable or disable multiplication of the
1526 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1527 make simulation less precise (but also much faster).
1529 * smpirun script should be (much) faster for large deployments.
1531 * SMPI tracing: fixed issue with poor matching of send/receives.
1533 * Replay: Fix broken waitall
1535 New functions and features
1536 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1538 Dropped / renamed functions and features
1539 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1540 - MSG_mailbox_is_empty() -> Mailbox::empty()
1541 - MSG_mailbox_front() -> Mailbox::front()
1542 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1543 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1544 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1545 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1546 * Don't pass the free_f parameter to property related functions:
1547 - MSG_host_set_property_value()
1548 - MSG_as_router_set_property_value()
1549 - MSG_storage_set_property_value()
1550 * VM properties. Since msg_vm_t are msg_host_t, just use
1551 MSG_host_get_property_value() and friends
1552 * VM I/O related things:
1553 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1554 - Unimplemented save/restore methods
1555 * MSG_as_router_get_property_value() was redundant with
1556 MSG_environment_as_get_property_value().
1557 - Removed MSG_as_router_*propert*() functions
1558 - Added MSG_environment_as_set_property_value() for consistency
1559 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1560 Well, they are still there for now, but deprecated with a warning.
1561 Please switch to xbt_dict_new_homogeneous() before this is removed
1563 * Task affinity. Its intended behavior (that was very badly tested
1564 and probably not really working) was deceiving what most users
1565 would have hoped here.
1566 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1567 * Option network/coordinates is now useless and should be dropped.
1570 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1573 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1574 * Added Dragonfly topology support
1577 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1578 EiBps, EBps, Eibps, Ebps for bandwidth.
1579 They may become useful to some lucky ones.
1582 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1583 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1584 * ex/async_waitAll: new example, on asynchronous communications
1587 * Memory usage should be decreased for simulations with a large number
1588 of processes. This also helps for SMPI.
1590 SimGrid (3.13) stable; urgency=low
1592 The Half Release, a.k.a. the Zealous Easter Trim.
1594 - We removed half of the lines, that were mostly experimental cruft.
1595 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1596 (not counting blanks and comments -- according to openhub.net)
1597 - The internals are now compiled in C++ (and will soon be clean C++)
1598 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1599 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1601 * Backwards Compatibility breaks
1602 - Removed Lua simulation bindings (switch to C or Java for that).
1603 Lua can still be used to describe platforms
1604 - Removed Java kernel plug-ins.
1605 Will be reintroduced after the ongoing major internals reorg.
1607 - the following functions were removed.
1608 They were too specific and should be reimplemented in a generic
1609 way, with filter function.
1610 - MSG_task_listen_from_host
1611 - MSG_mailbox_get_count_host_waiting_tasks
1612 - MSG_mailbox_put_with_timeout was removed.
1613 Please use MSG_task_send_with_timeout instead.
1615 - the SD_application_reinit function was removed. It has been a noop for a while.
1616 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1617 replaced by a more flexible mechanism.
1618 - The following functions thus do not exist anymore
1619 - SD_workstation_get_access_mode
1620 - SD_workstation_set_access_mode
1621 - SD_workstation_get_current_task
1622 - Basic estimation functions have been removed but can easily be replaced
1623 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1624 - SD_workstation_get_computation_time => amount / sg_host_speed()
1626 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1627 Use VM.getSpeed()*load/100 for the legacy behavior.
1629 - option enable_tracing was removed. It was not doing anything for a while.
1630 - In the ModelChecker:
1631 - the model-checker now ptraces the model-checked process which means
1632 you cannot use a debugger on the latter anymore (we might make this
1633 optional in the feature);
1634 - removed soft-dirty page tracking;
1635 - remove model-checked side snapshot management,
1636 MC_snapshot() and MC_compare_snapshot();
1637 - keep the MC_cut() function as a stub (it was not really working
1638 in the previous release).
1641 * All options are consistently in kebab-case. Old names are kept as alias.
1644 * Switch to platform v4 format.
1645 - Rename from 'power' to 'speed' the attributes describing the amount of
1646 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1647 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1648 - In <host> and <link>, attributes availability and state are gone.
1649 It was redundant with state and availability traces, and with peak values.
1650 - In <cluster>, attributes availability_file and state_file are gone.
1651 It was too complex and unused.
1652 - Kill <gpu>. Was not doing anything.
1653 - The DOCTYPE points to the right URL:
1654 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1655 (the file at this address now documents the changelog since its v1)
1656 - A warning is emitted for unit-less values (they are still accepted).
1657 - speed. Default: 'f' or 'flops'. Also defined:
1658 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1659 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1660 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1661 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1662 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1663 - latency. Default: 's' second. Also defined:
1664 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1666 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1667 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1670 * s4u::Host is now the preferred public interface to the Host features.
1671 sg_host_* functions are C bindings to the exact same behavior
1672 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1675 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1676 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1677 They were deprecated since a few years, and probably did not even compile anymore.
1680 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1681 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1682 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1683 encouraged to update their codes. A list of the modified functions can be found at the end of
1684 include/simgrid/simdag.h
1687 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1688 the lambda or closure passed as a parameter will run in kernel mode.
1689 All the callback functions should be rewritten to that interface at some point.
1692 * Reorganizing and cleaning the internals all around the place.
1695 * Remove old default barrier/bcast buggy algorithms (see #18407)
1696 * Various bug fixes to handle more codes
1697 * Remove the need for the --foreground option of smpirun (it is still
1698 accepted for backward compatibility).
1701 * Kill the setset data container: MC don't use it anymore.
1702 * Kill the queue data container: it made more sense with GRAS.
1703 * Kill the xbt_peer_t data type: it's useless without GRAS.
1704 * Kill rm_cb feature of config sets: it was never useful.
1705 * Kill graphxml parsing feature. It was not used.
1706 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1708 - xbt_dynar_sort_strings(), when the content is char*
1709 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1710 They throw exceptions on invalid input;
1711 * C++ support for declaring CLI flags (simgrid::config::Flag);
1712 * class for abstracting different signal backends (simgrid::xbt::signal).
1713 with no external dependencies (we need very simple signals).
1716 * refactoring and cleanup of the code;
1717 * ongoing process to cleanly separate the model-checking algorithms
1718 from the code model-checking support.
1720 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1722 SimGrid (3.12) stable; urgency=low
1724 The Facelift Release.
1727 * Require g++ v4.7 at least to not speak prehistorical C++.
1728 * Require Boost 1.48 (for signal2 component).
1729 * Java must be version 7 at least when activated.
1730 * Builds on Windows again (including Java bindings).
1731 * Tracing is now always enabled (no way to turn it out)
1732 * Remove GTNetS. It was not working anyway.
1733 * Various cleanups in the cmake scripts.
1734 * Move headers around to sort them out on installed systems:
1735 - instr/instr.h -> simgrid/instr.h
1736 - instr/jedule/* -> simgrid/jedule
1737 - simdag/datatypes.h was removed
1738 - simdag/simdag.h -> simgrid/simdag.h
1739 - msg/datatypes.h was removed
1740 - msg/msg.h -> simgrid/msg.h
1743 * Interface improvement:
1744 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1745 - Sanitize the interface in MSG_task_ module:
1746 - Merge two functions that were close enough but misleading:
1747 set_compute_duration(t) -> set_flops_amount(t)
1748 get_remaining_computation(t) -> get_flops_amount(t)
1749 - set_data_size(t) -> set_bytes_amount(t)
1750 get_data_size(t) -> get_bytes_amount(t)
1751 - Massive cleanups in the functions related to the energy
1752 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1753 - New: MSG_host_get_pstate()
1755 - msg/energy/onoff: switching hosts on and off
1757 * Interface improvement:
1758 - Rename Host.isAvail() to Host.isOn()
1759 - Rename Process.currentProcess() to Process.getCurrentProcess()
1760 - Rename Task.setDataSize() to Task.setBytesAmount()
1761 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1763 - #18874: Actually allows the GC to reclaim tasks
1767 - SIMIX_process_throw: raises an exception in a remote process
1768 * Refactoring: Separate sub-modules
1769 - libsmx: the public interface, as libc in a real system
1770 - popping: the strange dance that converts a user request into a kernel handling
1771 - smx_context_*: the virtualization mechanisms that embed the user code
1772 - smx_*: the handling of each simcalls
1773 * Interface cleanup:
1774 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1775 * Rename smx_action_t into smx_synchro_t, making explicit that these
1776 things are used to synchronize processes with their environment.
1777 For example, a communication is a sort of synchronization involving
1778 the communicating processes (that may block until the exchange) and
1779 the platform. The same can be said from computations, etc.
1781 - #18888: segfault when a process ends before its kill_time
1785 - 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.
1786 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1787 - MPI_Type_set_name, MPI_Type_get_name
1788 - MPI_*_c2f and MPI_*_f2c functions
1789 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1790 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1791 - Activate a lot of new tests from the mpich 3 testsuite
1793 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1794 - InfiniBand network model added: Based on the works of Jerome Vienne
1795 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1796 - When smpi/display_timing is set, also display global simulation time and application times
1797 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1798 * Collective communications
1799 - 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
1800 - 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/).
1801 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1803 - Replay now uses algorithms from wanted collective selector
1804 - Replay can be used with SMP-aware algorithms
1805 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1806 - 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.
1808 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1809 - Correctly use loopback on fat-tree clusters
1810 - Asynchronous small messages shouldn't trigger deadlocks anymore
1811 * Energy/DVFS cleanup and improvement
1812 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1813 - new: smpi_get_host_pstate()
1816 - "Full" network optimization flag was broken since Surf++
1817 - Better handling of precision flags in maxmin
1818 - Fix bug causing sometimes "Impossible" errors
1819 - Properly pass cluster properties to included hosts
1820 * Improvement of the Energy plugin.
1821 - Always update the consumption before returning that value
1822 - New property: watt_off to denote the dissipation when the host is off
1823 - New functions getWattMinAt and getWattMaxAt to retrieve the
1824 dissipation of pstates that we are not currently at.
1825 * Java: class NetworkLink renamed to Link
1826 * New function: simcall_process_get_kill_time()
1827 * Massive rename s/workstation/host/
1828 - That's intrusive, but that's good for the project consistency. Sorry.
1829 - Change config option "workstation/model" into "host/model"
1832 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1833 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1834 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1835 - xbt_dirname and xbt_basename for non-POSIX systems
1837 * The model checker now runs as a separate process.
1838 * The model checker runs must now be launched with the new simgrid-mc program.
1839 * Record/Replay: the MC can display a textual representation of a path in the
1840 execution graph. It can then be replayed outside of the model checker.
1842 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1844 SimGrid (3.11) stable; urgency=low
1849 * Normalizing pointers addresses tool for better diff between logs
1852 * Add cloud examples using new VMs
1853 - examples/msg/cloud/two_tasks_vm.tesh
1854 - examples/msg/cloud/simple_vm.tesh
1855 - examples/java/cloud/cloud.tesh
1856 - examples/java/cloud/migration/migration.tesh
1857 * Add java surf examples:
1858 - examples/java/surfPlugin/surf_plugin.tesh
1859 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1860 - examples/java/surfCpuModel/surf_cpu_model.tesh
1861 * Add SMPI+MSG example:
1862 - examples/smpi/smpi_msg_masterslave/
1867 - msg task destroy cancel test
1868 - msg_host on/off test
1869 * Move all tests in testsuite to teshsuite (adding tesh files)
1870 * Restructure teshsuites
1871 - one folder for each kind of test
1872 * Restructure AddTests.cmake
1874 - structure the order of tests (with sections)
1877 * Add virtual machine
1878 - creation of a VM on a PM
1879 - migration of a VM from a PM to another PM
1881 - MSG_process_join(msg_process_t process, double timeout)
1882 - msg_bar_t MSG_barrier_init(unsigned int count)
1883 - int MSG_barrier_wait(msg_bar_t barrier)
1884 - void MSG_barrier_destroy(msg_bar_t barrier)
1885 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1886 * New option "msg/debug_multiple_use" to help debugging when a task is used
1889 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1891 - Increase code coverage in test suites
1895 * Protect context stacks against stack overflow. The number of protected memory
1896 pages allocated on the top of each stack (1 by default) can be configured
1897 with the new command line option --cfg=contexts/guard_size:<value>.
1898 * Simcalls are now generated by a python script that
1899 - generates files included by SimGrid
1900 - checks that all the functions exist, and proposes prototypes
1902 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1903 host_set_data, host_get_data
1905 - simcall_process_join(smx_process_t process, double timeout)
1906 * Fix bug where sleeping processing could not be suspended.
1909 * Translate surf models from C to C++
1910 - Generic classes for all models: Model, Resource, Action
1911 - A generic interface for each kind of model (CPU, Network, Storage
1912 Workstation, WorkstationVM)
1914 * Translate surf routings from C to C++
1915 * Add callbacks using sigc++ or boost::signals2
1916 - Add callback functions for resource creation/destruction
1917 - Add callback functions for action state change
1918 - Handle Energy as a plugin
1919 * Replace swag by boost::intrusive
1920 * Add new routing models for clusters. For documentation, see
1921 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1922 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1923 parameters for cluster tag
1924 - Fat trees, with topology="FAT_TREE" and
1925 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1926 - see examples/platforms/torus_cluster.xml and
1927 examples/platforms/fat_tree_cluster.xml
1928 * More documentation
1931 * Hostfiles support host:nb_processes construct to deploy several processes on
1933 * Collective communication algorithms should not crash if used with
1934 improper number of nodes and report the error.
1935 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1936 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1937 MPI_Dims_create, MPI_Cart_sub are supported.
1938 * New interface to use SMPI programmatically (still depends on MSG for
1939 some parts, see examples/smpi/smpi_msg_masterslave):
1940 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1944 * Global variables privatization in MPI executables is now performed at runtime
1945 with the option smpi/privatize_global_variables (default:no).
1946 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1947 dynamic libraries loaded by the application are not privatized (static
1948 linking with these libraries is advised in this case)
1951 * Options defined in XML work correctly now.
1954 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1955 copied into simgrid.jar or not (ON by default). Use this option if you want
1956 to reduce the size of the installed simgrid.jar, *and* the native libraries
1957 are kept installed elsewhere.
1958 * Surf binding with SWIG (code generated in maintainer mode only):
1959 - plugin to handle callbacks
1960 - CPU model only for the moment
1963 * Supernovae build mode is definitively removed. It was used to improve
1964 inlining and inter-module optimizations. It is nowadays superseded by
1965 link-time optimizations commonly available in compilers.
1966 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1967 * Add boost dependency for surf++
1968 * Add new macro for tests
1969 - ADD_TESH(name <tesh_args>)
1970 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1974 - xbt_bar_t XBT_barrier_init(unsigned int count)
1975 - int XBT_barrier_wait(xbt_bar_t barrier)
1976 - void XBT_barrier_destroy(xbt_bar_t barrier)
1977 * Make the xbt_os_time module public
1979 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1981 SimGrid (3.10) stable; urgency=low
1983 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1986 * Reintegrate Java to the main archive as desynchronizing these
1987 package is not acceptable anymore (Java is now considered stable)
1988 * Add explicit synchronization facilities through semaphores
1989 * Bug fix: Task.setDataSize() only changed the C world, not the value
1990 cached in the Java world
1993 * Dramatically change the way files are handled. API and internals changed, but
1994 this part of MSG was not considered as production grade either.
1995 * Add explicit synchronization facilities through semaphores
1996 * Add a new function MSG_host_get_process_list()
1997 * Preliminary DVFS support (see examples/msg/energy/ for details)
2000 * SMPI is now included directly in the libsimgrid as the windows
2001 linker doesn't force us on splitting it anymore.
2002 * Improvements of the SMPI replay tool:
2003 - Most of the collective communications are now rooted in the same process as
2004 in the original application.
2005 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
2006 used until now). Multiple data types can now be used in a trace.
2007 - The replay tool now supports traces produce either by TAU or a modified
2009 - Bug Fix: the compute part of the reduce action is now taken into account.
2010 - Gatherv collective is now supported
2011 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
2012 of smpirun outputs time independent traces for the current run. One file
2013 is created per process. If too many processes are simulated, this behavior
2014 can be changed to one file for all processes by using the
2015 tracing/smpi/format/ti_one_file flag
2016 * smpirun generates the host file if needed (with given host count and platform)
2017 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
2018 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
2019 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
2020 --cfg=smpi/coll_selector:(mpich/ompi)
2021 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
2023 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
2024 happening inside a collective SMPI call.
2025 * Fix the behavior of complex data types handling.
2026 * Make MPI_Wtime another synchronization point to take computations into
2028 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
2029 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
2030 * Add all missing Fortran bindings, SMPI should work with Fortran 90
2031 (no privatization of global variables yet)
2032 * Preliminary DVFS support (see examples/smpi/energy/ for details)
2035 * Verification of liveness properties is now available for SMPI applications
2036 (in addition to MSG applications)
2037 * Bugged examples using SMPI in examples/smpi/mc/
2038 * Add --cfg=model-check/visited option. Allows the verification of infinite
2039 programs. Detection of loops in the execution thanks to the system state
2040 comparison and reduction of the state space to explore. Can be combined with
2041 DPOR for safety properties.
2044 * Allow to change SimGrid configuration (see --help) within the code
2045 thanks to SD_config() as it can be done in MSG.
2046 * Add a new function SD_task_set_amount() upon user request.
2049 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
2050 * Remove rule based routing (no more PCRE dependency)
2051 * Add a limiter_link option to cluster tag, to specify a maximum reachable
2052 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
2053 * Add a loopback_bw and loopback_lat options to cluster tag.
2054 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
2055 routing. Now peers are encapsulated in an AS and have their own private
2056 router but this is transparent.
2059 * Our own implementation of getline is renamed xbt_getline, and gets
2060 used even if the OS provide a getline(). This should reduce the
2061 configuration complexity by using the same code on all platforms.
2062 * New type: xbt_cfg_elm_boolean.
2063 * Allow to use yes/no for boolean configuration options in the command line.
2064 * Allow to disable SimGrid cleanups at exit from command line option.
2065 There are situations where one may want a simulation to end with an exit.
2066 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
2067 annoying when scripting around the simulator. Adding a
2068 --cfg=clean_atexit:no allows to circumvent this issue.
2071 * Lots of memory leaks were corrected in this release.
2072 There is no known memory leaks anymore, in all of our 600+ tests.
2073 * New command line option --version, to get SimGrid version information.
2074 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
2076 * Supernovae builds are deprecated, and expected to be removed in the next
2079 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2081 SimGrid (3.9) stable; urgency=low
2083 The Grasgory release: GRAS is really dead now.
2085 * Complete overhaul of the internal host structures scheme.
2088 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
2089 was considered as experimental and badly maintained since then.
2090 * Keeping it was thus a trap to our potential users, that could take
2091 it instead of MSG or SMPI by mistake despite is pity state.
2092 * GRAS seems to have very few users (if any), and no one volunteered
2093 to maintain it further. It also induces a lot of XBT code (for
2094 portability sake), that must be maintained too.
2095 * For all these reasons, we killed GRAS. If someone wants to revive it
2096 in the future, don't cry, our git history still remembers of GRAS.
2099 * Major overhaul. Merge our documentation again as time proved that
2100 splitting it was really not helping our users.
2101 * Further improve the developer documentation to help newcomers
2102 hacking on SimGrid itself. The user documentation (and in
2103 particular, the beginner documentation) is still in a sorry state.
2106 * Now works on Windows too!
2107 * Much more extensive test suite, from MPICH
2110 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
2111 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
2112 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
2113 * Bug fix: let task be scheduled when the last dependency to be solved is
2114 a control dependency.
2115 * Remove SD_load_environment_script function.
2116 Use the C sg_platf function if you want to declare a platform
2120 * New function: MSG_process_get_number()
2121 * Old function documented: MSG_config()
2122 * Remove MSG_load_platform_script function
2123 Use the C sg_platf function if you want to declare a platform
2127 * Change the default value of the TCP_gamma constant (maximal size of TCP
2128 congestion window) to a more realistic 4MiB value. If you notice changes in
2129 your simulation results, you can fall back to the previous 20k tiny window
2130 by adding --cfg=network/TCP_gamma:20000 on command line.
2131 * (Hopefully) fix a bug wrt periodic availability and state traces
2132 * Bug fix: use default values at start when first event in availability/state
2133 trace is not at time 0.
2136 * remove the "new_" part of function name sg_platf_new_trace_connect
2137 (resulting in sg_platf_trace_connect), since it does not create
2141 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2142 use case where it's really mandatory, and maintaining it was a pain in
2144 * New: xbt_fifo_search(), search an item with a user-provided
2145 comparison function instead of dumb pointer comparison.
2148 * Fix the lua deployment:
2149 Use `simgrid.init_application()` before deployment instead of
2150 `simgrid.msg_register_application()` after.
2153 * Transfer the tracing files into the corresponding modules.
2155 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2157 SimGrid (3.8.1) stable; urgency=low
2159 The "we are told that some people want to also *install* the simgrid
2162 * Add missing file "tesh.1" to the archive.
2164 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2166 SimGrid (3.8) stable; urgency=low
2168 The Psssshiiiit release: SimGrid jumps into the Cloud.
2171 * Add an experimental interface to manipulate VMs. They are mainly
2172 process groups with very few intrinsic semantic, but they should
2173 allow you to build the semantic you want easily.
2174 * New function: MSG_host_set_property_value()
2175 * New function: MSG_process_on_exit(). To clean memory in all cases.
2176 * Bug fixes that made the host (and link) failures unusable.
2177 * Add a way to auto-restart process when the host in which they are
2178 executing comes back (ON_FAILURE="RESTART" on deployment file,
2179 MSG_process_auto_restart_set).
2180 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2181 please stop using the old ones, they are DEPRECATED.
2183 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2184 Please use MSG_init() instead. (reducing the amount of entry
2185 points in the library helps us).
2186 * Make it impossible to link against the wrong version of the lib
2187 * Deprecate MSG_clean(). No need to call it anymore.
2188 * Function MSG_get_host_number() is not deprecated anymore.
2191 * Split the doc into a user guide and a reference guide.
2192 * Start a developper guide to help people hacking on SimGrid.
2195 * Enable tracing by default. This modules rocks you should use it.
2196 * Remove option custom_flags. Now use environment variables CFLAGS
2198 * Use default cmake things to detect lua instead of home grown ones.
2199 * New option "enable_mallocators" to disable mallocators, for debugging
2200 purpose ("on" by default).
2203 * Bug fixes around the resource failures: don't let the processes
2204 survive the host they are running onto.
2205 * Add an interface to auto-restart processes when the host in which they are
2206 executing comes back.
2207 * Ensures that SIMIX_clean is called automatically. It's not part of
2208 the public interface anymore (bindings should be updated).
2211 * Bug fix for when SD_Simulate is called with a positive value: be careful
2212 when comparing doubles. Sometimes they are different for non significant
2214 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2215 parallel task whose initial work is distributed among host according
2216 to the Amdahl's law. Such tasks are created with a parameter alpha
2217 that corresponds to the non-parallelizable part of the computation.
2218 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2219 between two sets of workstations assuming a 1D block distribution (each
2220 workstation owns a similar share of data) on both sides.
2222 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2223 Data redistribution will be automatically scheduled once parent and child
2224 are both scheduled. The filling of computation_amount and
2225 communication_amount structures is now done seamlessly thanks to the chosen
2227 * New function SD_workstation_dump to display various information
2228 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2229 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2230 bandwidth on the route onto which the task is scheduled and the amount of
2232 To divide the nominal bandwidth by 2, the rate then has to be:
2233 rate = bandwidth/(2*amount)
2234 * Compute tasks that have failed can now be rescheduled and executed again
2235 (from their beginning)
2236 * Increasing source code coverage (src/simdag is now covered at 95.8%
2240 * Re-implement time-independent trace replay using SMPI (at the
2241 smpi_smp_* level) instead of MSG. This should replace
2242 examples/msg/actions/actions.c
2243 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2244 hindexed and structs)
2245 * Implement the exchange of non-contiguous data.
2246 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2247 * Correct behavior of smpi/sender_gap and set its default value to 0
2248 * Add option to asynchronously send small messages to allow better
2249 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2250 specifies the size in bytes under which messages will be asynchronously sent.
2251 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2252 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2255 * New C interface to define a platform: XML is now optional.
2256 For more info, please check include/simgrid/platf.h.
2257 * New interface to define random platforms from the C:
2258 For more info, please check include/simgrid/platf_generator.h and
2259 examples/msg/masterslave/masterslave_platfgen.c
2260 * Export a sg_cmdline dynar containing all the arguments we got from
2264 * Two new tracing options for adding comments to trace file so you
2265 can track your experiments (see --help-tracing for details).
2266 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2267 * Adding the SimGrid version that generated the trace file as a comment.
2268 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2269 * Fix to avoid key clashes on Paje links
2270 * Other minor fixes related to the Paje specification
2273 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2274 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2275 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2276 * Greatly improve the robustness of mmalloc to user errors (such as
2277 using an area after freeing it, or freeing it twice)
2279 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2281 SimGrid-java (3.8.1) stable; urgency=low
2283 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2284 Warning: all methods are visible, but only some of them are
2285 implemented so far. Check the source (src/jtrace.c)
2286 for further information.
2287 * New module: org.simgrid.msg.File (SimGrid File management functions)
2288 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2289 * Change the meaning of Process.restart: now restart the process from
2290 the beginning, like MSG_process_restart in C.
2291 * Add Process.setAutoRestart: handling of process restart when failed
2293 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2294 you to retrieve the properties of the processes/hosts
2295 * Deprecate Msg.clean(): you can just forget about it now.
2296 * New function Process.getCount(), that only works when compiling
2297 with the not yet released version 3.9 of the C library.
2299 * New context factory based on Coroutines. It mandates a modified JVM
2300 but then, the simulations run about five times faster, and there is
2301 no limit to the amount of processes (beside of the available memory).
2303 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2305 SimGrid (3.7.1) stable; urgency=low
2308 * Restore the prototype of MSG_process_create_with_environment() to
2309 the pre-3.7 situation by removing the kill_time argument.
2310 * Add a MSG_process_set_kill_time() function instead.
2313 * Fix weird behaviors when dealing with parallel tasks.
2316 * Simgrid is now built as a dll.
2317 * Simgrid-java now works on Windows.
2318 * Simgrid-Java is now included into Windows package.
2321 * First pre-build package for MacOSX.
2324 * Fix compilation when using MSG_USE_DEPRECATED.
2325 * Fix some compilation issues on Macs and Windows.
2326 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2327 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2329 We discovered that the Lua console is broken, but we are missing the
2330 manpower to fix it right now. The problem existed in 3.7 too, so we
2331 are not blocking the release for that. Sorry if you depended on this
2332 feature, any help would be really welcome.
2334 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2337 SimGrid-java (3.7.1) stable; urgency=low
2339 The "Java aint got to be bloated and slow" release
2342 * Various internal cleanups and performance improvement
2343 Simulations are expected to run up to twice faster or so
2344 * Make Process.kill(process) an instance method, not a static one
2345 * User processes are not java.lang.Thread subclasses.
2346 This breaks the compatibility (sorry), but previous API was
2347 brain-dead, making it impossible to have non-trivial
2348 initializations in the process constructor.
2349 * Require a full constructor per Process sub-class.
2350 Kinda breaks the compatibility (sorry), but this allows a much more
2351 efficient way to launch the processes at simulation startup.
2352 * Do not embeed our version of semaphores, java 1.5 can be considered
2353 as sufficiently prevalent for us to not dupplicate its features.
2357 * Add examples for almost every part of the API
2358 We spotted and fixed a lot of bugs in the process
2360 * New module: asynchronous communication API
2361 * New function: Process.sleep()
2362 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2363 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2364 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2366 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2368 SimGrid (3.7) stable; urgency=low
2370 The "spring cleanups (before the next Big Project kicks in)" release.
2373 * We can specify the SMPI latency/bandwidth factor with command line
2374 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2375 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2376 You can also use the "config tag" from platform file by adding this line
2377 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2378 (see "example/platforms/tag_config.xml" to use "config tag").
2379 Note that the command line supersedes the platform file configuration.
2380 * Change the correction factors used in LMM model, according to
2381 the latest experiments described in INRIA RR-7821.
2382 Accuracy should be improved this way.
2383 * Use the partial invalidation optimization by default for the
2384 network too. Should produce the exact same results, only faster.
2385 * Major cleanup in surf to merge models and split some optimization
2386 mechanisms from the core of the models. As a result you can now
2387 specify which model to use (e.g., --cfg=network/model:LV08
2388 --cfg=cpu/model:Cas01) and which optimization mode to use
2389 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2390 Incompatible combinations should err at initialization. See
2391 --help-models for the list of all models and optimization modes.
2392 * The CLM03 workstation model was dropped for simplicity because it
2393 used the deprecated CM02 network model. Use default instead.
2394 * Rename the TCP_gamma configuration option to network/TCP_gamma
2395 * Rename the coordinates configuration option to
2396 network/coordinates, and document it
2397 * Use now crosstraffic keyword instead of the terribly misleading
2398 fullduplex keyword. It is activated by default now in the current
2399 default model, use --cfg=network/crosstraffic:0 to turn it off.
2400 * Ongoing refactoring the model parsing to make XML files optional
2401 See include/simgrid/platf.h for details (still to be completed)
2404 * Major overhaul of the documentation. Almost instructive now :/
2405 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2406 functions and friends. This interface was considered as
2407 deprecated since over 2 years, it's time to inform our users that it is.
2408 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2409 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2410 use these (crappy) functions in your code.
2411 These functions will be removed soon. Stop using them now.
2412 * Deprecate MSG_get_host_{table,number}
2413 Implement MSG_hosts_as_dynar() instead.
2414 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2415 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2418 * Stabilize the parallel execution mode of user contexts
2419 * Introduce configuration variables to control parallel execution:
2420 - contexts/synchro: Synchronization mode to use when running
2421 contexts in parallel (either futex, posix or busy_wait)
2422 - contexts/parallel_threshold: Minimal number of user contexts
2423 that must be part of a scheduling round to switch to parallel
2424 execution mode (raw contexts only)
2425 * Fix bugs that prevented to use suspend/resume along with
2426 synchronization structures.
2427 * Fix bugs in process termination that lead to invalid memory access
2428 in very specific conditions.
2431 * Introduce a parallel mode for the models (controlled by surf/nthreads
2432 configuration item). In our tests, running the models in parallel
2433 never lead to any speedups because they are so fast that the gain
2434 of computing each model in parallel does not amortizes the
2435 synchronization costs, even when ultra fast futexes are used.
2436 This is released anyway because YMMV.
2439 * Performance boost by using a swag internally to compute the set of
2440 tasks that are finished and should constitute the return value of
2444 * Enable it by default now that it is considered rather stable.
2447 * Documentation of the tracing functions.
2448 * Performance gains when tracing categorized/uncategorized resource
2449 utilization by avoiding calls to get route when updating resource
2450 variables. LMM constraints are being used instead.
2451 * API changed to set task categories. Use MSG_task_set_category instead
2452 of TRACE_msg_set_task_category, and SD_task_set_category instead
2453 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2454 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2455 * Examples for link user variables added (see at examples/msg/tracing/)
2456 * Deprecated function TRACE_msg_set_process_category completely removed
2457 * Trace header updated according to the latest Paje file format
2458 * Tracing network lazy updates, no longer obligate users to use full updates
2459 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2460 * Experimental: let user code declare/set/push/pop application states for hosts
2461 * API changed to allow the manual creation of graph configuration files
2462 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2465 * Improve the API of Lua MSG bindings, using the Lua spirit.
2466 * Each simulated process now lives in its own Lua world (globals are
2467 automatically duplicated). It helps writing simulators. Will allow
2468 to run Splay programs within SimGrid in the future.
2469 * Add a Chord example in Lua, equivalent to the MSG one.
2472 * Start the implementation of a solution to express temporal
2473 properties, not only local assertions. This is still an
2474 experimental work in progress, stay clear from it to be safe.
2478 - Add new runtime parameters --help-logs and --help-log-categories
2479 to display information about supported logging parameters and
2481 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2482 don't exists anymore.
2483 * Mallocators: allow value NULL for the reset function.
2485 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2486 create homogeneous dictionaries, where all the elements share the
2487 same free function. Non homogeneous dictionaries will be
2488 deprecated in the next release.
2489 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2490 - Multi-level dictionaries are deprecated.
2492 - new function xbt_dynar_search_or_negative() that is useful when
2493 you have less than 2 million elements in your dynar and don't
2494 want of the extra complexity of catching exceptions when the
2495 element is not found.
2497 - Make xbt_os_thread module (for thread portability) public.
2498 Documentation is still to come, sorry.
2500 - Cleanups and simplifications to make it maintainable again.
2501 - Exotic features (such as memalign and valloc) were removed.
2502 - The metadata were extended and improved so that the
2503 model-checker becomes able to explore and inspect the heaps.
2504 - This may induce a performance drop when enable_model-checking is
2505 ON in cmake (even if it's not used in the simulation), but it is
2506 necessary at this point to get MC working.
2508 Turn model-checking OFF if simulation performance matters to you.
2509 Not enabling it at runtime is not enough, disable it in cmake.
2511 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2514 SimGrid (3.6.2) stable; urgency=low
2516 The "Not coding new stuff allows to polish old things" release.
2519 * New bindings to the NS3 packet level simulator (experimental)
2520 * Use the raw (efficient) execution contexts instead of the sysv
2521 (portable) ones when possible.
2522 * libpcre is now mandatory in any cases since not using it led to
2523 severe performance loss and possibly other issues
2524 * Update the XML platforms:
2525 - G5K: include the latest machine in Nancy
2526 - GridPP and LCG: new platforms
2527 * Documentation was partially updated, at least (more to come)
2529 Bug fixes, cosmetics and small improvements
2530 * Free terminated processes before the end of the simulation to avoid
2531 exhausting the memory of users having very dynamic amount of
2533 * Bug fix and cosmetics about canceling non-running tasks
2534 * Bug fix about the dot loader's issues when using libcgraph
2537 * Create an installer for windows with nsis (amd64 and win32)
2538 - Add an hello world project to illustrate simgrid project creation.
2539 - Embed libpcre into the Simgrid installer to avoid
2540 its compilation burden
2541 * The raw execution contexts should work on Apple now
2542 * Port to Windows 64 bits
2543 - Sysv contexts now have an implementation for this arch
2544 - GRAS communication features now support this arch
2545 * Drop support for borland compiler on windows
2546 - this code was not maintained, and we kinda depend on gcc nowadays
2547 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2548 * Fix portability issue on unstable ubuntu: linker became picky on
2551 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2553 SimGrid (3.6.1) stable; urgency=low
2555 The "Oops, we broke Macs too" release
2558 * Fixed contexts detection so that raw ones are used when possible
2559 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2560 leads to a strange error, with user code segfaulting sometimes when
2561 the generated code is not perfectly aligned (which is not
2562 controllable from the user side, depends on the amount of code)
2565 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2567 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2569 SimGrid-java (3.6) unstable; urgency=low
2572 * Split of every thing from simgrid v3.5 into a separate package.
2574 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2576 SimGrid (3.6) stable; urgency=medium
2578 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2581 * Bindings now constitute their own package, separated from the main one.
2582 Rationale: reduce our maintenance nightmare by reducing the module coupling
2583 They will soon be released on their own on gforge.
2585 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2586 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2588 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2589 * It's not quite deprecated for now because we have no replacement,
2590 but it may soon become the case.
2593 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2594 * Fortran: New user-level cache variable to store the rank of the running
2595 process. This improves performance by an order of magnitude.
2596 * C: New coccinelle script to automatically locate and modify global and
2597 local static variables.
2598 * Improved SMPI network model with a sender-side gap to account for multiple
2602 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2603 only say if a communication is finished, no matter whether it succeeded or
2604 failed. You can call MSG_comm_get_status() to know the status of a finished
2606 * New function MSG_task_dsend() to send a task and detach it. When a
2607 communication is detached, you are never notified of its success or failure
2608 and the memory is released automatically once it is finished. This function
2609 is useful when you don't care about the end nor the success of a
2611 * Change the prototypes of action replay. Sorry for inconvenience,
2612 but this is really more efficient this way (and to adapt your code,
2613 you just have to fix the initialization, that shouldn't be too long)
2614 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2615 ever managed to do anything useful with it.
2616 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2617 takes a dynar of communications. It returns immediately and gives the
2618 index of a finished communication (if any).
2619 * New example: a basic implementation of the Chord P2P algorithm.
2622 * New model for multi-core CPUs. You can now use the core attribute to
2623 precise the number of cores of a host. This is a basic model. Every
2624 process running on the host receives at most the power provided in
2625 the DTD (throughput<=power). Total throughput of process cannot exceed
2627 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2628 router linked by an up-link and a down-link (possibly asymmetrical).
2629 This kind of pattern allows to easily build last-mile model style platforms.
2630 Aggregating such patterns in a rule-based AS is thus the technique of
2631 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2632 * New model for Vivaldi routing. We transformed the Vivaldi network model
2633 into a Vivaldi routing model (based on the rule-based model). This allows to
2634 combine Vivaldi based latencies with last-mile platforms.
2637 * Added a check for NaN of IEEE754 infinite in the double entries of
2639 * Introduce a new context factory "raw", highly inspirated from the
2640 ucontext factory, but using manually crafted functions in assembly to
2641 do the work in an efficient manner.
2642 * Allow to change the used context factory at run time, not only at
2643 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2644 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2645 contexts at runtime (only with raw contexts or ucontexts).
2646 * Completely rewrote this module to allow parallel execution of user
2647 processes. Use --cfg=contexts/nthreads:N to execute user processes
2648 with N parallel threads (the default is 1, meaning no parallelism).
2649 * Allow to decide dynamically between sequential and parallel modes.
2650 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2651 processes in parallel only when their number is greater than or equal to P
2653 * Added a check for NaN of IEEE754 infinite in the double entries of
2657 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2658 won't output the state of processes when interrupted with Ctrl-C
2659 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2660 NULL-terminated array. This may solve backward compatibility issues
2661 due to the change to return type of SD_simulate. See also:
2662 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2663 * Add new macros with variable number of arguments.
2664 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2665 - in xbt/asserts.h: xbt_assert
2666 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2667 - in xbt/ex.h: THROWF and RETHROWF.
2668 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2670 * Change xbt_die() to accept a format string with arguments, just like printf.
2671 * New data structure: xbt_lib_t, like a dict but more general and with better
2675 * New configuration options
2676 Options triva/categorized and triva/uncategorized can be used to generate
2677 graph configuration files for Triva visualization tool.
2678 * Configuration option tracing/platform is renamed to tracing/categorized
2679 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2680 * New TRACE_user_link_variable interface:
2681 User provides the name of the link and the tracing variable to attach to it
2682 * the declaration of tracing categories must be done after the environment
2684 * simpler tracing interface, just one way to declare categories
2685 TRACE_category or TRACE_category_with_color, it is up to you
2686 * links in the trace file are again identified by their names
2687 * trace contains the full platform hierarchy exactly as declared using the ASes
2688 * Options tracing/msg/[task|process]:1 groups the process by hosts
2689 for both cases, tasks and processes must have names that are unique during the simulation
2690 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2691 * The experimental option tracing/msg/volume is deprecated
2692 its functionality may be reincorporated if needed
2694 The tracing generates a trace file with unordered timestamped events,
2695 because of the way the core simulator (surf) works. A script available
2696 at the tools directory (fix-paje-trace.sh) can be used to put the events
2697 in order. We have changed the tracing so it can generate ordered timestamped
2698 events in the final trace, but depending on the simulator (and how much time
2699 is simulated) that can lead to a huge memory utilization. It is deactivated
2700 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2702 Build Infrastructure
2703 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2704 - We are trying hard to keep the API stable, but it may happen that
2705 some things change (we're a research project after all, not a
2706 nuclear plant operating system). If such things should happen, you
2707 could rely on that macro to adapt.
2708 - current value: 30600 for 3.06.00, aka 3.6
2709 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2710 a number that can be compared with SIMGRID_VERSION.
2711 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2712 assertions and verbose/debug logging events are disabled at compile time.
2714 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2716 SimGrid (3.5) stable; urgency=medium
2719 * New feature to any SimGrid-based simulator: Model-Checking
2720 Check SIN#1 for more details.
2723 * New Model SMPI (three-interval linear regression for correction factors)
2724 See RR-7426, available at http://hal.inria.fr/inria-00527150
2725 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2726 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2727 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2728 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2729 * See also src/smpi/README
2734 - Tracing API changes: TRACE_start and TRACE_end should not be called
2735 by user-code. They are automatically called by simulators created
2736 with SimDAG, MSG and SMPI if the toolkit is compiled with
2737 tracing_enabled=ON. Categories declaration and utilization remain the
2738 same for MSG and SimDag.
2739 - A function was added to the tracing API to declare categories with
2741 - TRACE_category_with_color (char *category, char *color)
2742 where color must be in the following format
2743 "%f %f %f", red, green, blue
2744 and red, green, blue are float values in the interval [0, 1]
2745 - User can specify NULL as color parameter, or continue calling
2746 TRACE_category (cat)
2747 On that case, the tracing system will define random colors
2748 - The following command-line options are supported:
2749 --cfg=tracing/filename:msg.trace
2750 --cfg=tracing:1 (activate tracing, needed to use others)
2751 --cfg=tracing/platform:1 (categorized resource use)
2752 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2753 --cfg=tracing/msg/task:1 (task creation)
2754 --cfg=tracing/msg/process:1 (process creation, migration)
2755 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2756 --cfg=tracing/smpi:1 (SMPI interface tracing)
2757 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2758 - examples of examples/msg/tracing updated
2760 - DAXLoader and DOTLoader functions can generate tasks with categories
2761 - A new function to attribute a category to SD tasks:
2762 TRACE_sd_set_task_category (SD_task_t task, char *category)
2763 * Tracing the MPI interface implemented by SMPI
2764 - Collective operations are traced with states
2765 - Point-to-Point operations are traced with states/links
2766 - Tracing activated by a parameter "-trace filename" passed
2767 to smpirun during execution (considering that simgrid
2768 is compiled with tracing enabled)
2769 - To run the simulation with gdb, the simulator
2770 accepts --cfg=tracing/smpi:1 to trace SMPI
2771 - tesh files to check if smpi tracing is ok
2772 - See examples/smpi/NAS/DT-trace
2773 * GTNetS tracing re-worked
2774 - adaptation to the tracing system of GTNets to cope
2775 with modifications regarding the fullduplex mode
2776 - new tesh files to check if gtnets tracing is ok
2779 * Asynchronous communications through the functions:
2780 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2781 * New function: MSG_load_platform_script()
2782 to make possible using a lua script instead of XML files to set up platforms
2783 * New function: MSG_set_function
2784 to associate functions to processes, used when bypassing the parser
2785 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2787 Platforms: Add some more examples in examples/platforms
2788 * Grid'5000: see www.grid5000.fr
2789 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2792 * Change the XML format. This is a very important modification. SimGrid
2793 3.5 introduces a new hierarchical format based on the notion of
2794 Autonomous Systems. Compatibility with old format is ensured
2795 through the perl script provided in the install bin directory
2796 bin/simgrid_update_xml.
2797 It is now possible to build platforms with specific routing
2798 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2799 connect several platforms together. We will try to provide soon
2800 set of realistic platforms exploiting these properties (have a
2801 look at examples/platforms/ for the moment).
2802 * Take the opportunity of the XML format change to be a good XML citizen:
2803 rename link:ctn to link_ctn and similar changes (also dealt with by
2805 * Add a new routing scheme (rule-based) using regular expressions. It
2806 enables to have an extremely low memory footprint when the
2807 underlying routing is simple and can be compactly described. You
2808 need to have libpcre4-dev (perl regular expressions) installed if
2809 you want to use this routing scheme.
2810 * Revive the cluster TAG and allow to easily and efficiently (both in
2811 term of memory and speed) connect clusters together. Have a look
2812 at teshsuite/simdag/platforms/ to see how this can be done. With
2813 this tag, you can create clusters with thousands of tasks at no
2814 cost (have a look at examples/platforms/).
2815 Note: clusters are implemented as ASes, so there is no need for an
2816 enclosing AS tag if you have only one cluster in your platform.
2817 * Add new generic functions in the public interface that allows the user
2818 to call SURF 'create_resource' methods from your code (same
2819 functionality as the XML bypass mechanism but with a much lighter
2821 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2822 that uses a piecewise linear approximation to produce better
2823 results when exchanging small messages.
2824 * Add a new parameter to handle correctly full duplex link and account
2825 for interferences between uplink and downlink communications
2826 (activate with --cfg=fullduplex:1).
2829 * Rename the SD_READY (all dependencies are satisfied and task is
2830 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2831 dependencies are satisfied) state.
2832 This prevents a confusion between the notion of "ready to schedule"
2833 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2834 simulated" (SD_RUNNABLE) used by the simulation kernel.
2835 * Change the way a task is considered as ready. Instead of removing
2836 dependencies when a task is done, a counter is decreased. This way,
2837 it is always possible to reach ancestors thanks to the
2838 SD_taks_get_parents function (even after the end of the simulation.)
2839 * Change the return type of SD_Simulate from (SD_task_t*) into
2840 xbt_dynar_t. This function was in handling a dynar internally and
2841 converted it into a NULL terminated array for historical reasons.
2842 * New function SD_dotload(char*) to load a DAG described in dot
2843 format. This loader and the corresponding examples require the
2844 installation of the graphviz library.
2845 * Fix a bug in the management of tasks of size 0 in the surf network
2846 models. This problem was only visible with SIMDAG and you should
2847 thus disregard results produced with earlier versions if you
2848 relied on this feature (some tasks were blocked because of this).
2849 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2850 even though your applications were doing only point-to-point
2851 communications and sequential computations. Now you can really use any
2852 model you want (of course, if you create real parallel tasks, which are
2853 not implemented in most models beside ptaskL07, this will abort).
2854 * Add an example that schedules a DAX on an heterogeneous platform
2855 using a Min-Min strategy.
2856 * New function SD_workstation_get_current_task() that returns the kind
2857 of task currently running on a workstation in the sequential access
2859 * Raise some warnings when unexecuted tasks remains at the end of the
2860 simulation. This is usually caused by cycles in the DAG.
2863 * New function: SIMIX_process_set_function() called by MSG_set_function
2864 * Change the underlying waiting queue in semaphores so that a process
2865 can wait on several of them simultaneously (as in waitany).
2866 * Fix the way to handle tokens in semaphores so that all access patterns
2867 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2868 * kill the dirty pimple SIMIX_message_sizes_output()
2869 Please use (proper) visualization instead
2872 * New data container: setset (set of sets of elements)
2873 * New module: mmalloc (mapped malloc, allowing to have several
2874 independent segments of malloc)
2875 * New function: xbt_dict_cursor_set_data()
2876 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2877 * New function: xbt_dynar_is_empty()
2878 * New function: xbt_fifo_get_last_item()
2879 * Fix xbt_dynar_shrink(): use the right element size.
2880 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2881 uninitialized areas during expand.
2882 * Fix semaphores: previous implementation was severely broken.
2883 * Use library init/fini functions for our initialization.
2884 - you can use logs and other feature as soon as you want in your
2885 code (even before the xbt_init / MSG_init)
2886 - xbt_exit is now a no-op and produce a warning when used.
2889 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2890 benefit from the latest improvement to the simulation kernel.
2891 * Kill measurement sockets for now. If you rely on them, sorry. This
2892 release is not for you. This feature will be reintroduced in the
2893 future, but we cannot delay the release any further.
2894 * New function: gras_msgtype_get_name().
2895 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2896 * Fix (at last) the pmm example: it should not randomly fail anymore.
2898 Build chain: bug fixes and overall polishing
2899 * Cmake is now stable enough. Hence, we killed the autotools.
2900 * Port to windows ( TM :)
2901 * Fix the 'make install' target.
2902 No need to use 'make install-simgrid' anymore
2903 * Introduce a 'make dist' target compiling a *source* archive
2904 'make package' compiles a binary archive
2905 * Compile java files only on need
2906 * Add --cd and --setenv command line options to tesh
2907 * Out of source builds are not fully supported yet, but we are close
2908 * Enable supernovae and optimization flags by default for our users
2911 * Add layer to set up environment directly from lua, without XML.
2912 * The effect of gras_stub_generator can be achieved through
2913 lua too (check examples/gras/console/ping_generator.lua)
2915 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2917 SimGrid (3.4.1) stable; urgency=low
2919 The "Polishing easter eggs is probably a good idea" release.
2920 This is a bug fixes release only.
2924 * Fix a bug preventing the tasks from begin garbage collected.
2927 * Fix a bug occurring when a host involved in a communication fails.
2928 This was not detected properly by the other peer involved in the
2929 communication. Now, it's reported as a network error.
2932 * Warn the user about loop dependencies in data flow of DAX files
2933 * Obey the control-flow dependencies of DAX files
2936 * Add option "enable_smpi" allowing to not compile SMPI.
2937 Probably useful for the (Mac) users experiencing a build error here
2938 * Improve the detection of lua5.1 and ruby1.8
2940 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2942 SimGrid (3.4) stable; urgency=low
2944 The "Easter in Cargese" release. Also known as (major changes):
2946 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2947 ~> bindings were greatly improved
2948 ~> new tracing infrastructure for better visualization introduced
2950 * the "Welcome to configury modernity" release.
2951 ~> we switched from autotools to cmake, and improved our cdash
2954 A more detailled list of changes follow (full detail in svn log).
2956 Java Bindings: Various Cleanups
2957 * (install java-gcj-compat-dev on debian-like to use them)
2958 * Remove put/get: no need to export deprecated interface in Java
2959 Use send/receive instead.
2960 * Cleanup the examples and add a README per directory
2961 * Remove example autoDestination (that's the only way to go now)
2962 * Remove example explicitDestination (was a plain copy of basic)
2963 * Make JniException a runtime exception, so that there is no need to
2964 declare the fact that you may encounter such a beast. I guess that
2965 nobody will ever want to survive such error.
2966 * Create specific errors for each MSG case of failure:
2967 host failure, transfer failure, timeout, task cancelled
2968 * Cleanup the exceptions that may get thrown by each function
2969 * Other internal cleanups in Java bindings. Performance still bad :/
2970 Ruby and Lua Bindings: create them
2971 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2972 * That's new and great, you should try them out.
2973 Same functionalities than Java bindings, only even less polished
2975 * Kill the useless "rate" argument of SD_task_get_execution_time()
2976 Everyone used to provide -1 as a value, it was not used, and the
2977 semantic of a possible use wasn't even clear.
2978 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2979 either as comm costs or compute costs to mean that there is no
2980 such thing for that specific task.
2981 * Add a SD_task_set_name() function
2982 * Fix SD_task_unschedule() on typed tasks
2983 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2984 * In DAX loader, accept useless 'level' attributes to <job> since
2985 LIGO DAGs have them (seem to be to ease graphical representation).
2987 * Add an example masterslave_mailbox.c using send/receive and not
2988 the deprecated put/get interface.
2989 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2990 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2992 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2993 (declare MSG_USE_DEPRECATED to still have the old name)
2994 * Add a MSG_task_set_data() function
2995 * About trace replay (see examples/msg/actions):
2997 - Allow to work with splitted trace files for each process
2998 Give the specific trace file as argument of each process,
2999 and call MSG_action_trace_run(NULL)
3000 You can still have one merged file for all processes.
3001 - Fix implementation of collective operations
3002 * Allow task_execute() on 0-sized tasks (closes #10063)
3004 * This is the first release of SimGrid where SMPI is not considered
3005 beta anymore (even if some corners should still be improved)
3006 * Port over the new SIMIX_network submodule (internal refactoring)
3007 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
3008 * Implement more missing elements of the standard:
3010 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
3012 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
3013 MPI_Scan() MPI_Get_processor_name()
3014 - Added implementation of missing case for Alltoall (warning: it's
3015 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
3016 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
3017 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
3018 SMPI_MPI_Allgatherv()
3019 * Bug fixes include:
3020 - MPI_Waitsome() was broken
3021 - Allow relative includes in smpicc
3022 - Command line cfg argument 'reference_speed' was ignored...
3023 - Some functions did not properly lead to auto-benching of user code
3024 - smpicc passes -O2 by default (just like openmpi one)
3026 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
3027 * Bug fixes about timeouts during communications
3028 * add SIMIX_message_sizes_output() as a pimple to write to file the
3029 amount of messages per size. Use gnuplot to get histogram.
3030 Pimple because that's the only user-visible function of simix,
3031 defined directly in xbt.h (irk, sorry)
3033 - Add a SIMIX_sem_get_capacity() function
3034 - Fix interactions with processe resume/suspende
3035 - release_forever() was stupidly broken
3036 - Fix SIMIX_display_process_status() for processes in a semaphore
3037 - Make SIMIX_sem_block_onto() user-visible
3038 * Refactoring context stuff:
3039 - Use pseudo-OOP for better modularity
3040 - reimplement SIMIX_process_kill() without process_schedule() so
3041 that the latter can take as invariant that it is called from
3043 - Merge context_start into context_new for sake of simplicity
3045 * Add a Vivaldi network model, coded live during SUD'10 ;)
3046 * Rename configuration variables to start a hierarchy:
3047 o cpu_model -> cpu/model
3048 o network_model -> network/model
3049 o workstation_model -> workstation/model
3050 * New configuration variables:
3051 o network/bandwidth_factor: correction to bandwidth
3052 o network/latency_factor: correction to latency
3053 o netwotk/weight_S: correction to the weight of competing streams
3054 * Add a long description to the models, that users can see with such
3055 argument on the command line: --cfg=cpu/model:help
3056 * --help-models display the long description of all known models
3058 * config: add the ability to set a default value after registration
3059 Does not override any previously set value (e.g. from cmd line)
3060 * dict: allow to have integer key and data.
3061 When so, you need to use the following functions
3062 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
3063 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
3064 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
3065 In contrary to regular dicts, the key is not malloced before copy.
3066 Mixing scalar and regular elements in the same dict is not tested
3068 * Allow to use xbt_dynar_shrink() to expend the dynar instead
3069 Tracing for Visualization:
3070 * SimGrid is now instrumented in order to generate a trace file for
3071 visualization analysis: to use it, need to compile SimGrid with the
3072 "tracing" option enabled, and instrument the program using SimGrid with
3073 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
3074 (among other functions).
3075 * The instrumentation only traces the platform utilization for now
3076 * Documentation to use the tracing functions and how to analyze the
3077 traces with the Triva tool is written.
3078 * More information about: SimGrid FAQ (in the section Tracing Simulations
3081 * We moved to cmake as default build system. Autotools support will
3082 be dropped soon. Check the FAQ for more info about how to use it.
3083 * Greatly improved our cdash/ctest interactions
3084 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
3085 * Added memory checking tests with valgrind; lot of memleak fixing.
3086 This may be the first release of simgrid with so few memory issues
3087 * Added code coverage tests.
3088 Our coverage is still improvable, but at least we see it on cdash.
3090 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
3092 SimGrid (3.3.4) stable; urgency=low
3094 The "Desktop Grid needs love too" release (also called Xmas release).
3096 Models improvements:
3097 * Major speedup in the maxmin system solving by using lazy evaluation
3098 Instead of solving completely the maxmin system at each iteration,
3099 only invalidate (and recompute) the modified parts.
3100 This new feature is enabled in default models but you can try to
3101 turn it on with "--cfg:maxmin-selective-update=1" for other models.
3102 * Cas01 IMproved as default CPU model
3103 This CPU model is the same Cas01 model, but it uses the
3104 maxmin-selective-update flag and a heap structure to manage
3105 actions on SURF kernel.
3106 It reduces the complexity to find the next action to finish and,
3107 consequently, it's faster than the old Cas01.
3108 This is the new default CPU model (Cas01).
3109 * Rename the old Cas01 model to Cas01_fullupdate
3110 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
3111 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
3112 * CpuTI (CPU Trace Integration)
3113 A new CPU model whose objective is simulate faster when using
3114 availability trace files.
3115 Instead of using a full featured, over engineered maxmin system for
3116 CPU modeling, this model does the pre-integration of traces files
3117 to calculate the amount of CPU power available, and so, executes
3118 faster than the old CPU models.
3119 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
3120 * Use LV08 as default network model since it gives better accuracy
3121 for small messages and shouldn't change things for big ones.
3122 Use --cfg=network_model:CM02 to get the previous behavior.
3125 ******************************************
3126 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
3127 ******************************************
3128 * The new CPU model may changes simulations!
3129 The point is that events occurring at the exact same timestamp
3130 are not scheduled in the same order with the old and new
3131 version. This may be enough to completely change the execution
3132 of simulations in some cases.
3133 * The new network model will change simulations!
3134 This new model is more realistic than the previous one, so you
3135 should consider redoing your old experiments with this model.
3136 Sorry for the inconvenience.
3139 * Introduce the supernovae compilation mode
3140 When compiled that way, the whole SimGrid (or almost) is put in a
3141 single compilation unit and compiled in one shoot.
3142 This is to help gcc which has difficulties to inline stuff from one
3144 The speedup seem to be above 15%, althrough more tests are needed on
3145 amd64 to confirm that gain.
3148 * Port of MSG's mailbox on top of SIMIX network
3149 The put/get mechanism was greatly simplified on the way.
3152 * New SIMIX network module. Provides:
3153 - Mailbox: rendez-vous mechanism to find with who you want to speak
3154 - Synchronous send/recv: easier and hopefully faster since the
3155 logic is handled in the maestro process directly now
3156 - Asynchronous send/recv: you dreamt of it? It's here now
3157 Too bad that nobody cared enough to propagate the change to MSG.
3158 * Add semaphores as SIMIX synchronization mechanism.
3161 * new function SD_daxload(char*) to load a DAX file
3162 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3163 * Introduce typed tasks. Specify its kind and cost at creation.
3164 At scheduling, just give where it should be placed, and the cost
3165 for each involved resource is automatically computed.
3166 Existing constructors so far (more to come of course):
3167 - SD_task_create_comm_e2e() for end-to-end communication
3168 - SD_task_create_comp_seq() for sequential computation
3169 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3170 * new function SD_task_dump() for debugging display
3171 * new function SD_task_dotty(task,FILE*) writing to file the info
3172 about the task in dotty format
3173 * SD_task_dependency_exists() can now cope with having one of its
3174 arguments NULL. If so, it tests whether the other argument has any
3176 * Add getters on list of preceding/following tasks:
3177 SD_task_get_parents(task) and SD_task_get_children(task)
3178 * Add getters on amount of workstations and list:
3179 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3180 * Add getter on task kind: SD_task_get_kind(task)
3181 * Update the start_time and finish_time of tasks on completion/failure
3182 * Bugfix: Remove task from state swags when destroyed
3185 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3188 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3189 dynar with one separate thread per value of the dynar.
3190 * Change the prototype of xbt_thread_create(), sorry.
3191 Added a boolean parameter indicating whether we want to join this
3192 thread (used in SG only for now)
3193 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3196 * GTNetS wrappers should now be usable again (and betterly tested too)
3197 * Fix a major regression from 3.2 where the timeout provided to
3198 MSG_task_put_with_timeout() was used as absolute time before which
3199 the comm should be done.
3200 * Start to fix the <cluster> tag.
3201 - Internal links should be good now (beside of the loopback, which
3202 use the private link instead)
3203 - paths to the external world is still rather broken
3204 - the <route:multi> tag is just broken. Actually that's brain-dead.
3205 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3206 to make it less stupid
3207 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3208 * Fix a source-level compatibility glitch from 3.2: after defining
3209 MSG_USE_DEPRECATED, you can use the old name
3210 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3211 * Allow to compile from the SVN with automake 1.11
3212 * Fix some problems when using the "start_time" tag in deployment XMLs.
3213 * Fix #8569: XBT/synchro.h has redundant declarations
3214 * Fix #8563: MSG return values and exceptions
3215 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3216 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3217 Thanks to Robson Peixoto.
3218 * Fix a vicious bug in dictionaries inducing that some elements were
3219 not freed on xbt_dict_free()
3221 Portability report of this version:
3222 * Main portability targets:
3223 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3224 on (amd64/i386/ia64)
3225 - mac leopard on i386
3226 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3227 but nothing critical.
3228 * Other platforms: windows, AIX and others were not tested for this release
3230 Timing report of this version:
3231 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3232 scenario parameters). From 8h to a few seconds in desktop grid settings.
3233 * Supernovae brings about 25% speedup on i386.
3235 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3237 SimGrid (3.3.3) stable; urgency=low
3239 The "Need for Speed" release.
3241 The timings done to validate the 3.3.2 were faulty.
3242 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3244 The problem was a conversion from a manually handled vector to
3245 xbt_dynar_t on the critical path.
3246 xbt_dynar_foreach calls functions, inducing stack management crap.
3248 We inlined these functions and xbt_dynar_foreach is now breath taking.
3249 We also inlined xbt_swag_belong on the way.
3251 Here are some approximate speedup measurements (on master/slaves
3252 simulations lasting between 10s and 20s each):
3253 3.3.1 -> 3.3.2: about same performance
3254 3.3.2 -> 3.3.3: 40% speedup
3255 3.3.1 -> 3.3.3: 40% speedup
3256 3.3.1 with inline patch -> 3.3.3: 30% speedup
3258 Our reading is that the refactoring which occurred in 3.3.2 made us
3259 suffer much more from the xbt_dynar_foreach low performance, but
3260 once we solved this, this refactoring proved to be very performance
3261 effective. From the 40% speedup, somehow, 10% are due to the
3262 inlining and 30% to the refactoring.
3264 That's a pitty that gcc cannot inline functions placed in other files
3265 alone. We have to choose between:
3266 - break the encapsulation (by putting private data structures and
3267 accessors in headers files to help gcc)
3268 - live with low performance
3269 - switch to a decent compiler such as icc (not quite possible).
3271 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3273 SimGrid (3.3.2) stable; urgency=low
3275 The "Simplicity does not preceed complexity, but follows it" release.
3277 The main contributors of this release were (lexical order):
3278 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3281 * Extract the routing logic into its own object.
3282 (was dupplicated in network.c and workstation_LV07.c;
3283 Allows to implement other ways of storing that info)
3284 => kill now useless network_card concept
3285 - Use dynar to represent routes (instead of void** + int*)
3286 - kill link_set (use surf_network_model->resource_set instead)
3287 - Add a command-line option to choose the routing schema to use
3288 - Add three new models:
3289 * Floyd (shortest path computed at initialization)
3290 * Dijikstra (shortest path recomputed all the time)
3291 * Cached Dijikstra (shortest path computed on need)
3292 All these models where contributed by Silas De Munck, and are
3293 described in his ICCS09 paper.
3295 * Simplify model declaration
3296 (less redirections, less function to write when defining a model)
3297 - Factorize stuff between models:
3300 surf_model_resource_set(model)
3301 surf_model_resource_by_name(model, name)
3302 - Unify the types of models in s_surf_model_t (using an union)
3303 - Embeed fields of common_public directly into s_surf_model_t
3304 - Rename model methods:
3305 action_free ~> action_unref
3306 action_change_state ~> action_state_set
3307 action_get_state ~> action_state_get
3308 - Change model methods into functions:
3309 (model)->common_public->action_use ~> surf_action_ref
3311 * Implement a generic resource; use it as ancestor to specific ones
3312 (allows to kill duplicated code in models)
3313 Drawback: timer command don't need no name nor properties;
3314 workstation_CLM03 don't need no properties
3315 (but I guess we can live with those few bytes wasted)
3317 * Improve the action object model
3318 - implement a constructor avoiding dupplicated code about field
3319 initialization in generic_action part.
3321 * Kill the SDP model: it has an external dependency, is deprecated
3322 in flavor of modern lmm models, and didn't compile since a while
3325 * Relocation of the context module from XBT to SIMIX.
3326 (the context were decoupled from the simix processes, duplicating a lot of code)
3327 => a lot of code was factorized
3328 - less overhead is introduced during scheduling
3329 - simpler API for the context factory
3330 - the logic for process creation,destruction and manipulation was simplified
3331 * Simplification of the s_smx_process_t data structure.
3332 => accesing the simix level data associated to a process is faster now,
3333 and the code is a lot more readable.
3336 * Implement some more MPI primitives:
3337 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3338 -implementation: Bcast: flat or 2-ary tree (default),
3339 Barrier: 4-ary tree,
3341 Allreduce: Reduce then Bcast
3342 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3343 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3344 Alltoallv: flat tree, like ompi
3346 * Add support for optimized collectives (Bcast is now binomial by default)
3347 * Port smpirun and smpicc to OS X
3350 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3351 Shout out if you used it.
3354 * Display the list of still queued messages in SG mode when existing
3358 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3359 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3360 * Add xbt_str_from_file(FILE*)
3361 * Add xbt_dict_get_key achieving a linear reverse search
3362 * Remove the context module
3364 Portability report of this version:
3365 * Main portability targets:
3366 - Linux(debian)/x86/context
3367 - Linux(debian)/x86/pthreads
3368 - Linux(debian)/amd64/context
3369 - Linux(debian)/amd64/pthreads
3370 On these, we still have the eratic breakages of gras/pmm and
3371 amok/saturate_sg reported in previous version. We still think
3372 that the tests are the cause of the fault, not the tested code.
3374 - Mac OSX Leopard/x86/context
3375 Still false negative in tesh autotesting.
3376 Smpi still fails, but this time because readlink does not accept -f
3377 Everything seems to work properly beside of that.
3380 - AIX version 5.3 (only tested contexts this time)
3381 Smpi still fails there because mktemp is not installed.
3382 Everything seems to work properly beside of that.
3384 I managed to compile it for the first time, but several breakages.
3385 Won't delay the release for this exotic platform.
3387 * Windows: it's still lagging behind. If you want to help, please
3390 Timing report of this version:
3391 This version seem to be more than 5% faster than 3.3.1 (on linux
3392 64bits with contextes). The gain is less than expected, we are
3393 investigating this for next release.
3395 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3397 SimGrid (3.3.1) stable; urgency=low
3400 * Implement a --cfg-help to show existing configuration variables
3401 * Build chain do not require doxygen in maintainer mode
3404 * fix a bug on struct sizeof computation, which prevented the
3405 exchange of arrays of structs in some conditions
3406 - added a regression test about this in datadesc_usage
3407 * Allow the exchange of 0-long dynamic vectors.
3408 - for that, use -1 as indicator of dynamic size instead of 0
3409 - This implied to change any size from unsigned long to long,
3410 reducing a bit communication abilities, but I guess that with
3411 64bits being quite common, this is more than enough.
3412 - This also induce a protocol change, thus bumping network protocol
3413 version from 0 to 1 (if we have external users, we have to get
3414 clean on that point too ;)
3415 - added two regression tests about this in datadesc_usage
3416 * Be more verbose when propagating local exceptions
3417 This helps debugging.
3418 * Display the status of simulated processes when receiving SIGINT in
3422 * Allow to control the simulation from a trace file.
3423 New functions MSG_action_register() and MSG_action_trace_run()
3424 The first one allows to associate a function execution to each
3425 kind of action while the second one parses a trace file and
3426 triggers the corresponding actions within the system.
3427 For now, only a toy example is provided in examples/msg/actions
3428 * Add an example of process migration in examples/msg/migration
3429 * Fix a bug in task exchange which broke MSG_task_get_sender()
3430 Add a teshsuite regression test for that.
3431 [Bug: if MSG_task_get_sender() is called after sender exit,
3433 * Fix a bug which prevented suspend/resume to work properly
3434 * Display the status of simulated processes when receiving SIGINT
3435 This fixes a regression of v3.3. due to the introduction of SIMIX
3436 * Bug fixing in failure management:
3437 - trace could not start by a failure at time 0
3438 - failure during communications were not working
3441 * Add SIMIX_process_set_name() to change the name of the current
3442 process in the log messages.
3443 * Store smx_hosts in a dict since we only retrieve them by name
3444 * Move the configuration infrastructure to surf
3447 * Move the configuration infrastructure to surf
3450 * Massive internal cleanups:
3451 - Store internal structures on processes instead of hosts (allows
3452 to have more than one process per host, in addition of being more
3454 - Cleanup the initialization/finalization process
3455 - Kill a whole bunch of unneeded synchronization:
3456 processes run in exclusive manner within the simulator
3457 - Move queues from global tables to process data fields
3459 - now accept -platform and -hostfile arguments
3460 - Pass the right rank value to processes according to the hostfile
3461 * Compile the examples by default, and use them as regression tests
3462 * Implement MPI_Wtime()
3463 * Change the reference speed to a command line option
3466 * TCP_gamma can now be specified as command line option using
3467 --cfg=TCP_gamma:10000000.0
3468 * Change the --surf-path cmd line option into --cfg=path:
3471 * Also include strbuff from xbt.h public header
3472 * xbt_ex_display(): do not free the exception after displaying
3473 This allows to do more with the given exception afterward.
3474 Users should call xbt_ex_free() themselves.
3478 Portability report of this version:
3479 * Main portability targets:
3480 - Linux(debian)/x86/context
3481 - Linux(debian)/x86/pthreads
3482 - Linux(debian)/amd64/context
3483 - Linux(debian)/amd64/pthreads
3484 These targets fail about 1/10 of times on gras/pmm, but we believe
3485 that this is because of the test, not because of simgrid.
3486 amok/saturate_sg fails even more rarely, and the test may not be
3489 - Mac OSX Leopard/x86/context
3490 The test suite still spits tons of errors because some obscure
3491 force prevents us from removing the temporary directories
3492 arguing that they still contain some metadata I've never heard of.
3493 Smpi fails because seq is not installed.
3494 Everything seems to work properly beside of that.
3497 - AIX version 5.3 (both contexts and pthread)
3498 Smpi still fails there because mktemp is not installed.
3499 XML inclusions seems rosty on AIX.
3501 * Windows: it's still lagging behind. If you want to help, please
3504 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3506 SimGrid (3.3) stable; urgency=high
3510 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3511 [Malek Cherier & Mt]
3513 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3516 Until now, GRAS were using MSG as an interface to SURF. It was
3517 quite difficult because both interface have several differences
3518 (MSG channels vs GRAS sockets were the most notable point).
3520 This also opens the gate to SMPI (which should occur soon) and speed
3521 up simulations by to 40% (even if it were not the main goal).
3523 **************************************
3524 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3525 **************************************
3526 The point is that events occurring at the exact same timestamp are
3527 not scheduled in the same order with the old and new version. This
3528 may be enough to completely change the execution of simulations in
3529 some cases. Sorry for the inconvenience.
3531 * Cleanup and upgrade the XML format to push further scalability
3532 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3534 * Improve the testing infrastructure with tesh. Now a very large part of
3535 the code is tested not only by being run but also by checking that the
3536 output match an expected output [Mt].
3538 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3539 is really less memory-demanding, which should allow you to use
3540 larger files in SimGrid [AL].
3542 * Inform valgrind about our contextes, so that it becomes usable
3543 with the default (and more effecient) version of SimGrid
3544 [contributed by Sékou Diakite, many thanks]
3547 * Introduce a listener thread in charge of receiving incoming
3548 messages from the network. It allows to overlap communication and
3549 computation but most notably, it removes some stupid deadlocks due
3550 to the fact that so far, a process could not send and receive at
3551 the same time. This made most non trivial communication schema
3553 * Convert the PIDs from long int to int to match the MSG ones (and
3554 linux ones too) [Mt]
3555 * New function: gras_agent_spawn() to launch a new process on
3556 current host. Only working in simulation for now. [Mt]
3557 * New function: gras_os_hostport() returning a constant form (ie,
3558 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3561 * Make the backtrace of exceptions more human readable [Mt]
3562 * New module: xbt/str [Mt]
3563 a ton of string utility functions (split, join, printf to a newly
3564 allocated buffer, trim, etc)
3565 * New module: xbt/hash [Mt]
3566 SHA1 hashing algorithm (more to come if needed)
3567 * New module: xbt/synchro [Mt]
3568 synchronization tools (mutex and conditions) working the same way
3569 in simulation and in real life (mainly useful for GRAS, but not
3571 * New module: xbt/queue [Mt]
3572 classical producer/consumer synchronization scheme
3573 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3574 (using the classical functions will get serialized) [Mt]
3575 * Make dictionary internal table dynamic. No need to specify its size
3576 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3578 * Make sure the log channels are organized as a tree under windows
3579 (because of ANSI C compatibility issue, any channel were child of
3580 root directly) [Mt].
3583 * Cleaned many thing in surf and fixed a few bugs [AL].
3584 * Add a nice command line configuration mechanism to compose models [AL].
3585 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3586 the previous one (KCCFLN05). It relies on something that looks like
3587 a max-min sharing mechanism but cannot be written as such. A new solver
3588 was thus designed [AL].
3589 * Add a new solver to lmm. Based on Lagrange optimization and
3590 gradient-based descent, it enables to efficiently maximise systems s.a
3592 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3594 This solver enables to propose two new network models for TCP Reno and
3595 TCP Vegas based on Low's work. These models still need to be fully
3596 tested though [Pedro Velho].
3599 * Bug fix in SD_simulate. Now the time bound given as argument is
3601 * Use the new parallel task model (ptask_L07) as default.
3602 * Use the SURF command line configuration mechanism.
3603 * 0-size tasks (for synchronization) should now work.
3605 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3607 SimGrid (3.2) stable; urgency=high
3611 We still experience issues on this platform, but we believe that at
3612 least MSG is usable.
3614 GRAS API BREAKAGE (for simplification purpose, sorry):
3615 * the gras_msgtype_by_name is not used anymore. Instead of
3616 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3617 you can write (and must)
3618 gras_msg_send(toserver, "request", &request);
3619 - If you still want to pass a gras_msgtype_t to the function (to cache
3620 the type and avoid the lookup time), use the gras_msg_send_() variant.
3621 - Impacted functions:
3622 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3623 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3624 * The callbacks are now expected to return 0 when everything went well
3625 (just like the main() function)
3627 GRAS new features and improvements:
3628 * New module mechanism where user code can use per process globals [Mt]
3629 This is similar to gras_userdata_*() functions, but for libraries. It
3630 factorize some code developped over and over in the examples and AMOK.
3631 It has still to be documented and used (only amok/peermanagement is
3633 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3634 loss when they were small enough to fit into the buffer and sent quickly
3635 enough so that they can all get received in one shoot.
3636 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3637 if not found. Use the *_or_null() variant for the old semantic.
3638 * In gras_msg_handle, do not discard messages without callback.
3639 They are probably messages to be explicitly awaited later (ie, proofs of
3640 mis-synchronization in userland since they are sent before being awaited)
3641 No big deal usually.
3642 * gras_socket_meas_send/recv: semantic changed!
3643 The numerical arguments used to be (1) the total amount of data to send
3644 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3645 messages. This was need for the fool willing to send more than MAXINT
3646 bytes on quite fat pipes.
3649 * Do really rename the hostmanagement module to peermanagement. [Mt]
3650 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3651 but this is rather new and this was documented in the module
3652 documentation (poor excuses, I admit)
3653 * Bandwidth measurement semantic changed! This follows the changes to
3654 gras_socket_meas_send/recv explained above.
3657 * A sequential mode has been added to the workstations. When a workstation
3658 is in sequential mode, it can execute only one task, and the other tasks
3659 are waiting in a FIFO. [Christophe Thiery]
3662 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3663 model for SIMDAG. [Christophe Thiery]
3664 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3665 equal to 0, which caused some bad behaviors in
3666 saturated_constraint_set_update. I now use a threshold mechanism like in
3670 * When running manually src/testall, you select specific units [Mt]
3671 testall is the result of our cunit mechanism, and should replace all
3672 the scripty thingy around since bash don't run easily on billware.
3674 * A mallocator system has been added. [Christophe Thiery]
3675 Mallocators allow you to recycle your unused objects instead of freeing them
3676 and allocating new ones.
3678 Documentation update:
3679 * FAQ reworking + New FAQs:
3680 - "Valgrind spits tons of errors!" [Mt]
3681 - "How to repport bugs" [Mt]
3682 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3683 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3684 - Communication time measurement within MSG [AL]
3685 - I experience weird communication times when I change the latency [AL]
3686 * GRAS tutorial [Mt]
3688 - an introduction to the framework and to the used communication model
3689 - an initiatic tour introducing the most proheminent features:
3691 . Lesson 0: Installing GRAS
3692 . Lesson 1: Setting up your own project
3693 o Part 2: Message passing
3694 . Lesson 2: Exchanging simple messages
3695 . Lesson 3: Passing arguments to the processes (in SG)
3696 . Lesson 4: Attaching callbacks to messages
3697 . Lesson 5: Using globals in processes
3698 . Lesson 6: Logging information properly
3699 . Lesson 7: Using internal timers
3700 . Lesson 8: Handling errors through exceptions
3701 . Lesson 9: Exchanging simple data
3702 . Lesson 10: Remote Procedure Calling (RPC)
3703 . Lesson 11: Explicitly waiting for messages
3704 . Recapping of message passing features in GRAS
3705 - A HOWTO section containing:
3706 o HOWTO design a GRAS application
3707 More are due, of course. They will come latter. In the meanwhile, you can
3708 check the examples which are still here.
3710 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3712 SimGrid (3.1) stable; urgency=high
3716 There was a stack corruption somewhere, visible only when optimizing
3717 with these versions. [Vince]
3720 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3721 module that enabled to play with Directed Acyclic Graphs. It is built
3722 directly on top of SURF and provides an API rather close to the old
3723 SG. Some old codes using SG are currently under rewrite to check that
3724 all needful functions are provided. [Christophe Thiery]
3727 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3728 extension of the classical CLM03 model that gracefully handles
3729 failures. This is now the default model for MSG and GRAS. It doesn't
3730 handle parallel tasks yet though. [AL]
3731 * Bug fix: Weights were not correctly set in the network part.
3732 WARNING: This may have resulted in incorrect results with simulations
3733 where there are more than one flow on a given link. [AL]
3736 * After a (long ?) discussion on simgrid-devel, we have decided that the
3737 convention we had on units was stupid. That is why it has been decided
3738 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3739 WARNING: This means that all previous platform files will not work as
3740 such with this version! A warning is issued to ask users to update
3742 A conversion script can be found in the contrib module of the CVS, under
3743 the name contrib/platform_generation/surfxml_update.pl [MQ]
3746 * Bug fix: Processes were started in reverse order, wrt deployment file.
3747 WARNING: if your code relies on this bug, please fix it. [AL]
3748 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3749 being executed on two different locations. [AL]
3750 * Bug fix: Failures are now better supported thanks to Derrick's tests
3751 (there was many failure situations I hadn't thought of and that weren't
3752 correctly handled). [AL]
3753 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3757 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3758 Exception occurring on server-side are propagated back to client (!).
3760 API CHANGE: the callback changed their prototype. Change:
3761 int my_handler(gras_socket_t expeditor, void *payload_data) {
3763 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3764 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3766 * New! function: gras_msg_handleall to deal with all messages arriving
3767 within a given period.
3768 * New! function: gras_socket_server_range to get a server socket in a
3769 range of port numbers (ease to avoid port number conflicts) [MQ]
3770 * New! gras processes display their backtrace when they get a SIGUSR1
3771 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3772 Sweet to debug RL processes [MQ]
3776 - Do not force experiment sizes to be expressed in kb, or it becomes
3777 impossible to measure the latency this way (needs one byte-long tests)
3778 WARNING: this changes the amok_bw_* function semantic. [MQ]
3779 - Implements the link saturation stuff. [MQ]
3780 * Peer management module:
3781 New! module factorizing code that we wrote over and over [MQ].
3784 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3785 - Test units are placed directly into the library code, they get extracted
3786 automatically and placed into the src/testall binary.
3787 - Convert most of the XBT tests to this system.
3788 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3789 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3790 live in the same C code. This required to change a little bit the API
3791 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3792 * New module: sparse graph structure with basic algorithms (this is work
3793 in progress and the API is not considered to be frozen yet). [AL]
3794 * Display more information on backtraces: source line & function names are
3795 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3796 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3797 * Reworked a little bit some #include statements to load only required
3798 headers. Some user code that relied on SimGrid to include stdlib or
3799 stdio may need to include it by themselves. [AL]
3800 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3802 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3803 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3804 but I guess that almost nobody used those parts. [MQ]
3806 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3808 SimGrid (3.0.1) stable; urgency=low
3811 * Unfortunately, I had missed 5 misnamed functions:
3812 xbt_fifo_item_t xbt_fifo_newitem(void);
3813 void xbt_fifo_freeitem(xbt_fifo_item_t);
3814 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3815 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3816 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3817 They're now deprecated. Please use their new versions:
3818 xbt_fifo_item_t xbt_fifo_new_item(void);
3819 void xbt_fifo_free_item(xbt_fifo_item_t);
3820 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3821 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3822 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3824 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3825 * Documentation: xbt_log module unmercifully reworked [MQ]
3826 * Bugfix: there was a problem with the ending of contexts with
3827 the pthread backend. It caused some weird deadlock or behavior
3828 depending on the pthread implementation. [AL]
3829 * Bugfix: get the exceptions raised in the simulator repport where
3830 and why they come from when they are not caught in time [AL, MQ]
3833 * Bugfix: Do repport the error when two non-connected hosts try to
3834 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3837 * Add additionnal checkings on communications. Assert that two
3838 communicating hosts are connected by a set of links... [AL]
3841 * Add additionnal checkings on channel values in communication [AL]
3842 * New: MSG_task_get_source to see on which host a task was generated [HC]
3843 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3844 the number of tasks waiting to be received on channel and sent
3846 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3847 waits for the first task coming from a given host.. [AL]
3849 GRAS new functionnalities: [MQ]
3850 * Enhance the parsing macro to allow the size of multidimentional objects
3851 to be given thru annotations.
3852 * New example (and documentation): Matrix Multiplication a la RPC
3853 (as when I was young!) and fix a bunch of bugs found on the way.
3855 GRAS performance improvements: [MQ]
3857 * Reduce the amount of cbps creation/destruction by making it static to
3858 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3861 * Change libdata to a set so that we can search for stuff by ID (and thus
3862 reduce the insane amount of dict lookups)
3865 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3866 RL communication on top of SG since it's so uneffective.
3867 It may also be used for inter-thread communication in RL, one day.
3868 * Use gras_datadesc_copy() to exchange messages on top of SG
3870 - improve message exchange performance on top of SG
3871 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3872 * Don't exchange on the network the size of the used part of buffer,
3873 instead, specify the possible buffer size to read().
3875 - reduces the amount of read/write calls (one pair per exchange)
3876 - reduces the amount of exchanged data (the size)
3877 - allows to retrieve all arrived data on receiver side, if we don't need
3878 it right now (subsequent read will peek the buffer)
3879 - allows the receiver to proceed with the beginning of the stream before
3880 everything is arrived
3881 - make it possible to build an iov transport (using readv/writev)
3883 - take care of the data with non-stable storage (like stacked data),
3885 * If possible, TCP send uses vector I/O (when writev() is here)
3886 - Don't use it for receive since we send data sizes and data on the
3887 same stream, so we wouldn't be able to chain large amount of chunks
3888 before having to flush the stuff to read the size.
3889 * Rework the transport plugin mechanism to simplify it and reduce the
3890 amount of pointer dereferencement when searching for the right function
3893 * I guess that now, we do almost as few system calls as possible while
3894 doing as few data copy as possible.
3896 To improve it further, we could try to send all the sizes first and then
3897 all the data (to use iov on receiving size), but it's only a partial
3898 solution: when you have 2 dimensional data, the sizes of the second
3899 dimension is data of the first dimension, so you need 3 streams.
3901 I'm not sure the potential performance gains justify the coding burden.
3903 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3905 SimGrid (3.00) stable; urgency=high
3908 * New! Give the possibility to hijack the surf parser and thus bypass
3909 MSG_create_environment and MSG_launch_application. Have a look at
3910 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3912 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3914 SimGrid (2.96) unstable; urgency=low
3919 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3920 This deprecates the xbt_error_t mechanisms.
3921 It modifies (simplifies) all XBT and GRAS API.
3922 MSG API keeps unchanged (exceptions raised by XBT are caught from
3923 within MSG and masked with existing error handling facilities)
3926 * New! Add a FATPIPE model. [AL]
3927 * New! Add a parallel task model. [AL]
3928 * New! Add automatically a loopback interface (in the default
3929 network model) if none was precised.
3932 * Bugfix: MSG_process_resume now works with the current running process.
3934 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3935 * Modification of MSG_task_get_compute_duration. Once a task has been
3936 processed, the value returned by this function is now equal to 0. [AL]
3937 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3938 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3939 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3940 of a m_task. For now, MSG_task_cancel only works with computation
3942 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3943 of the processor (in Mflop/s) regardless of the current load on the
3945 * API Change: use proper naming convention for MSG_getClock and
3946 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3948 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3949 This function changes the priority of a computation task. This priority
3950 doesn't affect the transfer rate. A priority of 2 will make a task
3951 receive two times more cpu power than the other ones. This function
3952 has been added to suit the needs of Nguyen The Loc and hasn't been that
3953 much tested yet. So if it fails, please report it and send me your code.
3955 * API Change: removed all functions and types that were marked "deprecated"
3956 since many months. Renamed MSG_global_init_args to MSG_global_init.
3958 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3960 SimGrid (2.95) unstable; urgency=low
3963 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3964 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3965 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3968 * Bugfix: Allow absolute paths for platform description files [MQ]
3969 * Bugfix: do free the variables after use. Leads to drastic performance
3971 * Implement max_duration (ie, timeouts) on resources [AL]
3974 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3976 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3977 the needed support (timeouts) [AL]
3980 * Implement measurement sockets. You can now get the bandwidth between two
3981 hosts thanks to AMOK (see below). [MQ]
3982 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3983 dynar over the network (yeah) [MQ]
3984 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3985 * simple/static token-ring example. [Alexandre Colucci and MQ]
3986 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3987 in place before [MQ]
3990 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3991 type parsing. "short" and "long" modifiers were also ignored; other
3992 modifier (such as reference level) are still ignored. [MQ]
3993 * Embeed the buffer size within the buffer itself on SG. [MQ]
3994 That way, send() are atomic and cannot get intermixed anymore (at least
3995 the ones which are less than 100k; bigger messages still have the issue)
3996 * Array size pushed by the field, not by the field type (or each
3997 and every long int will push stuff to the cbps) [MQ]
3998 * use select() to sleep since it allows to portably sleep less than one
4001 GRAS (minor cleanups)
4002 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
4005 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
4006 * type category 'ignored' killed as it was never used and were difficult
4008 * whether a type can cycle or not is now a flag, leaving room for more
4009 flags (as "ignored", if we feel the need one day ;)
4010 * Rename raw sockets to measurement sockets since "raw" has another
4011 meaning in networking community.
4014 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
4015 and offers features not belonging to GRAS but that most applications
4016 need. One day, it may be a set of plugins loadable at runtime.
4017 * New module: bandwidth
4018 bandwidth measurement between arbitrary nodes running this module. [MQ]
4020 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
4022 SimGrid (2.94) unstable; urgency=low
4024 The first beta release of SimGrid 3 !
4028 * Update the main page and the FAQ. Adding references to gforge.
4031 * Add a gras_os_getpid function.
4034 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
4035 * Extend the logs so that they also print PID, hostname, date, ... if
4037 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
4038 and kill the old version which were in testsuite/
4039 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
4043 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
4044 for users. Instead, each programming environment (GRAS, MSG,...) use it
4045 when needed to provide such a feature to users.
4046 Don't shortcut the mechanism or you will also shortcut the virtualization
4047 you need on the simulator.
4051 * Cleanups in configury with regard to compile optimization/warning flags.
4052 Also add -fno-loop-optimize to any powerpc since it's the optimization
4053 killing gcc (< 3.4.0).
4054 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
4055 needed by MSG examples complications
4056 * Borrow configury beautifications from PHP
4059 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
4060 warning (thanks loris for stumbling into this one).
4061 * Bugfix: stop loading private headers (gras_config.h) from the public
4065 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
4066 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
4067 RL would imply a DNS resolver, which is *hard* to do in a portable way
4068 (and therefore delayed).
4069 * Implement a real timer mechanism and use it in timing macros. This allows
4070 to avoid rounding errors and get a 0.000005 sec precision in timing
4071 macros. While I was at it, various cleanups:
4072 - allow to declare more than one timed section per file (fix a stupid bug)
4073 - move some private declaration to the right place
4074 - merge conditional execution and timing macros into emulation module
4075 - document the module
4076 - make sure the module cleanups its mess on gras_exit
4077 * Documentation improvements:
4078 - (new) how to compile applications using GRAS
4079 - (new) emulation support (timing macros)
4081 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
4083 SimGrid (2.93) unstable; urgency=low
4085 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
4088 - Use Paje properly where used. Still to be sanitized properly.
4089 - Portability fix: Add an implementation of the contexts using pthread
4093 - Add xbt_procname(): returns the name of the current process.
4094 Use it to show the current process's name in all logging.
4096 - fix detection of older flex version and the reaction, since we do
4097 depend on modern ones (we use lex_destroy)
4098 - Better separation of SG and RL in the libs: remove all simulation code
4099 from libgras. As a result, this lib is now only 200k when stripped.
4100 Some of the xbt modules may also be duplicated (two sets and such) and
4101 should be cleaned/killed before SG3.
4102 - Insist on using xlC on AIX because of weird problems involving gcc there.
4103 - Cleanup the make remote stuff. This is now done by scripts
4104 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
4105 mainly for our private use, but we're working on changing them to user
4108 - Bugfix: flush the socket on close only if there is some *output*.
4109 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
4111 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
4114 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
4116 SimGrid (2.92) unstable; urgency=low
4118 Alpha 3 on the path to SimGrid 3
4122 - New! First try of benchmarking macros.
4123 - New! First try so that gras_stub_generator generate deployment and
4124 remote compilation helpers.
4126 - Bugfix: Initialization fix in msg_test.
4130 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
4132 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
4134 - API Change: gras_os_sleep to take the amount of seconds as a double.
4135 Accepting an int was error prone since it was the only location where
4136 seconds were coded as such. It leaded to damn rounding errors.
4137 - Bugfix: Hard to belive that timers ever worked before this.
4139 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4141 SimGrid (2.91) unstable; urgency=low
4143 Alpha 2 on the path to SimGrid 3
4147 - Bug fix in the lmm_solver.
4149 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4150 through the function MSG_paje_output.
4151 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4152 - It is possible to bound the rate of a communication in MSG with
4153 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4154 to write it in the changelog).
4155 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4160 - Various cleanups to the autotools stuff
4161 - Begin to move Gras examples to examples/gras/
4162 - Let make distcheck work again (yeah!)
4164 - documentation overhauled using doxygen.
4165 gtk-doc-tools is dead in SimGrid now.
4166 - Automatically extract all existing logging categories, and add the list
4167 to the documentation (long standing one, to say the less)
4169 - Cleanup the known architecture table. Reorder the entries to group what
4170 should be, and use a more consistent naming scheme.
4171 (some of the test dataset are still to be regenerated)
4172 - New! Allow library to register globals on each process just as userdata
4174 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4175 lookup time (for now).
4176 Use it in msg and trp.
4177 This cleans a lot the internals and helps enforcing privacy of the
4178 headers between the gras components.
4179 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4180 - Bugfix: gras_os_time was delirious in RL.
4181 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4183 - Reenable GRAS now that it works.
4185 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4187 SimGrid (2.90) unstable; urgency=low
4189 Alpha 1 on the path to SimGrid 3
4191 * It is a long time since the last release of SimGrid. I'm sorry about
4192 that but as I had told you, I was rewriting a lot of things. I apologize
4193 to those who had been reporting bugs to me and that I had not answered.
4194 If your bug is still in the new version, please tell me. Here is a
4195 summary of the main changes.
4197 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4198 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4199 portability, speed, and a lot more but you'll figure it out later.
4200 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4201 I wanted to release the new MSG as soon as possible and I have
4202 broken GRAS, which is the reason why, for now, only MSG is fully
4203 functional. A laconic description of these projects is available
4204 in the documentation.
4206 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4207 kernel optimized for our needs (called SURF but only the developers
4208 should use it). Hence, MSG is now roughly 30 times faster and I think
4209 that by rewriting a little bit MSG, I could event speed it up a little
4210 bit more. Beside the gain in speed, it is also much easier to encode a
4211 new platform model with SURF than it was with SG. More to come...
4213 * REVOLUTION 3: I have tried to change a little as possible the API of
4214 MSG but a few things really had to disappear. The main differences
4215 with the previous version are:
4216 1) no more m_links_t and the corresponding functions. Platforms are
4217 directly read from an XML description and cannot be hard-coded
4218 anymore. The same format is used for application deployment
4219 description. The new format is described in the documentation.
4220 Have a look in tools/platform_generation. There is a tiny script
4221 that converts from the old platform format to the new one. Concerning
4222 the application deployment format, parsing the old one is tricky.
4223 I think most of you should however be able to convert your files. If
4224 it is really an issue, I can write a C code that does the conversion.
4226 2) the toolbox tbx does not exist anymore. We now have a library
4227 with much more data-structures but without the hash-tables (we have
4228 dictionaries that are much faster).
4230 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4232 *****************************************************************************
4233 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4234 * don't feel like converting the oldies. (Mt) *
4235 *****************************************************************************
4238 Version 2.90: "the long awaited one"
4239 - Finished rewriting and debugging MSG. Rewrote the documentation.
4240 - disable GRAS for now since it needs to be ported to the newest SG
4243 - Finish the port to windows (using mingw32 for cross-compile)
4246 - Main loop and datastructures of SURF. A cpu resource object is
4247 functional. Surf can thus be used to create cpu's with variable
4248 performance on which you can execute some actions.
4250 2004-11-15 Martin Quinson
4251 - Port to ARM. Simply added the alignment and size descriptions. Should
4252 work, but the ARM machines are so slow that I didn't had the opportunity
4253 to 'make check' over there yet.
4255 2004-11-15 Arnaud Legrand
4256 - Trace manager now written. It uses a heap structure and is therefore
4257 expected to be efficient. It may however be speeded up (particularly
4258 when many events occur at the same date) by using red and black
4259 trees. One day maybe...
4260 - Max-min linear system solver written. It uses a sparse matrix
4261 structure taking advantage of its expected use. Most operations are
4262 O(1) and free/calloc are called as few as possible. The computation of
4263 the minimum could however be improved by using a red and black tree
4266 2004-11-03 Arnaud Legrand
4267 - Rename every gras_* function that was in xbt/ to its xbt_
4269 - Add a heap and a doubly-linked list to xbt
4270 - Added a dichotomy to the dictionaries. make check works as well before
4271 so I assume that the patch is correct. I do not know however if things
4272 run effectively faster than before now. :)
4274 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4275 SimGrid, and the version number is bumped to 2.x
4277 2004-10-29 Martin Quinson
4278 - Introduction of the remote errors.
4279 They are the result of a RMI/RPC on the remote machine.
4280 ErrCodes being scalar values, you can't get the host on which those
4281 errors did happen. Extending the error mechanism as in Gnome is possible.
4282 No idea yet whether it is a good idea.
4284 2004-10-28 Martin Quinson
4285 - Interface revolution: the Starred Structure Eradication.
4286 I used to do typedef struct {} toto_t; and then handle *toto_t.
4287 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4288 - ???_t is a valid type (builded with typedef)
4289 - s_toto_t is a structure (access to fields with .)
4290 - s_toto is a structure needing 'struct' keyword to be used
4291 - e_toto_t is an enum
4292 - toto_t is an 'object' (struct*)
4294 typedef struct s_toto {} s_toto_t, *toto_t;
4295 typedef enum {} e_toto_t;
4296 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4297 s_toto_t) is private.
4299 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4300 so that it can set the variable to NULL. It was so for dicts and sets,
4301 it changed for dynars.
4303 - Fix a bunch of memleaks in dict_remove
4304 - Fix a bug in sg/server_socket opening: it failed all the time.
4306 2004-10-07 Martin Quinson
4307 - Speed up dynar lookup operation a bit.
4309 gras_dynar_get is dead.
4311 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4312 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4313 (which gives you the address of the stored data).
4315 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4316 copy of the data using an affectation to do the job and not a memcpy.
4318 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4319 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4321 It does not really speedup the dynar test because they are
4322 setting elements all the time (and look them seldom). But the dict does
4323 far more lookup than setting.
4325 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4327 2004-10-05 Martin Quinson
4328 - Allow to (en/dis)able the cycle detection at run time.
4330 Whether we should check for cycle or not is now a property of each
4331 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4332 datadesc_cycle_unset allow to remove this property when previously set.
4334 Note that the cycle detection is off by default since it impacts the
4335 performance. Watch the data you feed GRAS with ;)
4337 This property is hereditary. Any element embedded in a structure having it
4338 set have it set for the time of this data exchange.
4340 You should set it both on sender and receiver side. If you don't set it on
4341 sender side, it will enter an endless loop. If you forget on receiver
4342 side, the cycles won't be recreated after communication.
4344 - Header reorganization.
4345 Kill gras_private.h, each submodule must load the headers it needs.
4347 2004-10-04 Martin Quinson
4348 - Interface revolution: do not try to survive to malloc failure.
4350 Now, gras_malloc and friends call gras_abort() on failure.
4351 As a conclusion, malloc_error is not a valid error anymore, and all
4352 functions for which it was the only gras_error_t return value are
4353 changed. They now return void, or there result directly.
4354 This simplify the API a lot.
4356 2004-09-29 Martin Quinson
4357 - Re-enable raw sockets.
4358 Created by gras_socket_{client,server}_ext;
4359 Used with gras_raw_{send,recv}
4362 It should allow to kill the last bits of gras first version soon.
4364 This is not completely satisfactory yet (duplicate code with
4365 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4368 - Simplify transport plugin (internal) interface by not passing any
4369 argument to _server and _client, but embedding them in the socket
4372 2004-09-28 Martin Quinson
4373 - Finish the port to AIX.
4374 autoconf was my problem (segfault within the malloc replacement
4375 function. No idea why)
4377 2004-09-16 Martin Quinson
4378 - Fix some size_t madness on 64bit architectures.
4380 2004-09-08 Martin Quinson
4381 - Reduce the number of system headers loaded, overload some more system
4382 calls (such as malloc to cast the result of the system one, and work
4384 - Fix and reintroduce the config support
4386 2004-09-07 Martin Quinson
4387 - Source code reorganization to allow Arnaud to surf all over there.
4388 - Allow to document the logging categories.
4389 - Remove all uppercase from logging categories and useless cleanup in names.
4391 2004-08-18 Martin Quinson
4392 Version 0.6.2 (protocol not changed; API changed)
4393 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4394 gras_error_t), and NULL when not found. Functions expecting a msgtype
4395 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4396 hopefully usefull message.
4397 - Portability to prehistoric sparcs again
4399 2004-08-17 Martin Quinson
4400 Version 0.6.1 (protocol not changed; ABI not changed)
4401 - prealloc some buffers to speed things up
4403 2004-08-11 Martin Quinson
4404 Version 0.6 (protocol not changed; ABI expended)
4405 - The parsing macro can deal with the references, provided that you add
4406 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4408 2004-08-09 Martin Quinson
4409 Version 0.5 (protocol not changed; ABI changed)
4410 - Allow to off turn the cycle detection code in data exchange at
4411 compilation time. It should be at run time, but I'm short of time (and
4412 the config stuff is still broken). That way, we keep dict out of the
4413 critical path, which is good because the performance is poor:
4414 - search not dichotomial yet
4415 - dynar give no way to access their content and memcpy everytime
4416 - In composed data description (struct, ref and so on), stop foolness of
4417 keeping the subtype's ID, but store the type itself. This keeps sets out
4418 of the critical path, which is good since they rely on dynar and
4419 dictionaries. The only loose of that is that we cannot detect the
4420 redeclaration of a structure/union with another content (but I'm not sure
4421 the code detected well this error before anyway). We still can detect
4422 the redefinition discrepancy for the other types.
4423 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4424 it breaks the code because of type-punning used all over the place).
4425 This breaks on all non-gcc architectures (for now).
4427 All those changes (plus the buffer of last time) allow me to gain 2 order
4428 of magnitude on cruel tests consisting of 800000 array of integers on two
4429 level of a hierarchical structure (200 secondes -> 4 secondes)
4432 - the selector of reference must now return the type it points to, not
4433 the ID of this type.
4435 2004-08-06 Martin Quinson
4436 Version 0.4 (protocol changed; ABI not changed)
4437 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4438 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4439 - Implement a buffer transport, and use it by default (it relies on tcp in
4440 real life and on sg in simulation).
4441 That's a bit hackish since I had a new field to the structure to store
4442 its data without interfering with the subtype ones. Inheritance
4443 is tricky in C. And that's a kind of reverse inheritance with one class
4444 derivating two classes. Or maybe a game with java interfaces. Anyway,
4445 that's damn hard in C (at least).
4446 Moreover, I got tired while trying to ensure plugin separation and
4447 genericity in SG mode. MSG wants me to do weird things, so let's go for
4448 cruel hacks (temporarily of course ;).
4449 See comment in transport_private.h:71
4450 - do not include all the _interface headers in private but in the files
4451 which really need them (to cut the compilation time when they are
4454 2004-07-26 Martin Quinson
4455 Version 0.3 (protocol not changed; ABI changed)
4456 - Major overhault of the datadesc interface to simplify it:
4457 - shorted the function names:
4458 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4459 - add a trivial way to push/pop integers into the cbps without malloc.
4460 This allows to make really generic sub_type description, which simply
4461 pop their size of the stack.
4462 - add a function gras_datadesc_ref_pop_arr() which does what users want
4463 most of the time: Declare a dynamic array (which pops its size of the
4464 stack) and declare a reference to it. Poor name, but anyway.
4465 - kill the post-send callback, add a post-receive one
4467 2004-07-23 Martin Quinson
4468 Version 0.2 (protocol changed; ABI changed)
4469 - add some testing for cpbs in the test cases, and fix some more bugs.
4470 This invalidate again the little64 data file, since I cannot regenerate
4472 - remove an awfull optimization in the logging stuff, allowing me to:
4473 - understand it again
4474 - learn gcc how to check that the argument match the provided format
4475 - fix all errors revealed by gcc after that
4476 - internal keys of dict are not \0 terminated. Deal with it properly in
4477 loggings instead of segfaulting when the user want to see the logs :-/
4479 2004-07-22 Martin Quinson
4480 - Fix some stupid bug preventing cbps (callback postit) from working
4482 2004-07-21 Martin Quinson
4483 - Some documentation cleanups
4484 - remove the useless last argument of msgtype_declare
4485 - rename the Virtu functions to fit into the 'os' namespace
4486 - move headers src/include -> src/include/gras/ and stop fooling with
4487 gras -> . symbolic link
4488 - make distcheck is now successful
4490 2004-07-19 Martin Quinson
4492 - Build shared library also
4493 - Install html doc to the right location
4494 - stop removing maintainer files in make clean
4495 - build tests only on make check
4497 2004-07-13 Martin Quinson
4499 - No major issue in previous version => change versioning schema
4500 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4501 corresponding dataset.
4503 2004-07-11 Martin Quinson
4505 - Get it working with any kind of structure (we can compute the padding
4506 bytes remotely for all the architectures I have access to)
4507 - Implement the structure parsing macro (still not quite robust/complete)
4508 - Improvement to the remote testing toysuite
4510 2004-07-10 Martin Quinson
4511 [autoconf mechanism]
4512 - get ride of a bunch of deprecated macros
4513 - actually run the test for two-compliment, not only compile it :-/
4514 - test whether the structures get packed (and bail out if yes. Damn.
4515 Alignment is a serious matter)
4516 - test whether the structures get compacted (but respecting the alignment
4517 constraints of each types)
4518 - test whether the array fields of structures can straddle alignment boundaries
4520 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4521 (took me 2 hours to find that bug, looking at the wrong place)
4522 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4523 before sending/receiving and are used to compute the offsets of fields
4524 - Given that padding size depend even on compiler options, keep track of
4525 alignment and aligned_size only for the current architecture. Not a big
4526 deal since we send structure fields one after the other (seems
4528 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4529 program, let it work on linux/gcc/little32. portability todo.
4531 2004-07-08 Martin Quinson
4532 - import and improve remote compilation support from FAST
4533 - make sure make check works on half a dozen of machines out there
4535 2004-07-07 Martin Quinson
4536 Let's say it's version 0.0.3 ;)
4537 - Implement conversions (yuhu!)
4538 - Let it work on solaris (beside conversion, of course)
4539 - Stupid me, using rand() to generate the conversion datatests in not wise.
4541 2004-07-06 Martin Quinson
4542 - Let make dist work, since I'm gonna need it to compile on remote hosts
4543 - Let Tests/datadesc_usage write the architecture on which the file was
4544 generated as first byte.
4545 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4546 architecture descriptions.
4547 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4548 on those architectures.
4549 - Optimization: send/recv array of scalar in one shoot
4551 2004-07-05 Martin Quinson
4552 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4554 - Plug a whole bunch of memleaks
4555 - each process now have to call gras_{init,exit}. One day, their log
4556 settings will be separated
4557 - Continue the code factorisation between SG, RL and common in Transport.
4559 2004-07-04 Martin Quinson
4561 - Redistribution between SG and RL.
4562 We wanna have to accept in SG, so move accepted related parts of RL in
4563 the common part. (more precisely, the dynar of all known sockets is no
4564 more a static in transport.c, but part of the process_data)
4566 [gras_stub_generator]
4567 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4569 2004-07-03 Martin Quinson
4570 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4571 them all up in one shot)
4573 - Fix the ugly case of reference to dynamic array.
4574 - New (semi-public) function gras_datadesc_size to allow the messaging
4575 layer to malloc the needed space for the buffer.
4577 - gras_socket_close now expect the socket to close (and not its address to
4578 put NULL in it after it). This is because the socket passed to handlers
4579 is one of their argument (=> not writable).
4581 - propagate the interface cleanup from last week in datadesc, ie remove a
4582 superfluous level of indirection. User pass adress of variable
4583 containing data (both when sending and receiving), and not of a variable
4584 being a pointer to the data. Let's say that I like it better ;)
4585 The price for that is constructs like "int msg=*(int*)payload" in
4586 handlers, but it's a fine price, IMHO.
4588 - Let it work in RL (yuhu)
4590 2004-06-21 Martin Quinson
4592 - porting SG plugin and SG select to new standards (works almost).
4593 - plug memleaks and fix bugs around.
4596 - cleanup the prototype of data recv and force users to specify when they
4597 want to handle references to objects. Test case working even for cycles.
4598 - plug memleaks. Valgrind is perfectly ok with this.
4600 2004-06-12 Martin Quinson
4602 - cleanup the separation between plugin and main code in plugin creation
4604 2004-06-11 Martin Quinson
4606 - Reput hook for raw sockets, needed for BW experiments
4607 - kill a few lines of dead code
4608 [Data description] Interface cleanup
4609 - gras_datadesc_by_name returns the searched type or NULL.
4610 That way, no variable is needed to use a type desc once, which makes
4612 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4613 two parameters were strings (field name, type name), leading to
4615 [Dicos] Interface cleanup
4616 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4617 This is consistant with the dynar API.
4619 2004-04-21 Martin Quinson
4621 - Porting to new standards.
4623 - interface cleanup.
4624 There is no bag anymore, no need to take extra provision to mask the
4625 pointers behind "ID".
4626 Better splitup of functions between files create/exchange/convert.
4627 This is still a bit artificial since convert and receive are so
4628 interleaved, but anyway.
4630 - add a queued message list to procdata (the ones not matching criteria
4632 - factorize some more code between SG and RL wrt procdata
4634 - use gras_exit in example to track memleaks
4635 - get rid of gs_example now that GS is properly integrated into gras
4636 - update run_test to integrate the latest tests (datadesc)
4638 - rename WARNINGn macros to WARNn since it prooved error-prone
4640 2004-04-19 Martin Quinson
4642 - register init/exit functions within gras module mechanism
4643 - send/receive function.
4644 Convertion is not implemented, but short-cutted if not needed.
4645 struct/array elements are sent one by one (instead of block-wise), but
4646 nobody really cares (yet). Get a prototype before optimizing.
4647 - tests (using a file socket) for DD send/receive on:
4648 - base types: int, float
4649 - array: fixed size, string (ie ref to dynamic string)
4650 - structure: homogeneous, heterogeneous
4651 - chained list, graph with cycle
4652 Believe it or not, valgrind is not too unhappy with the results. The
4653 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4657 - Bugfix when using a filename explicitly (instead of '-')
4659 2004-04-09 Martin Quinson
4661 - factorize more code between RL and SG in socket creation
4662 - Complete the implementation and tests of:
4664 o file (only in RL, and mainly for debugging)
4666 I lost 3 days to design a portable address resolver, and then decided
4667 that the prototype mainly have to run on my box.
4668 Addressing portability too early may be like optimizing too early :-/
4670 - use gras_init in the Tests instead of the crappy parse_log_opt
4671 (the latter function is removed)
4672 [Conditional execution]
4673 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4674 [Code reorganisation]
4675 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4676 Build examples against the RL library, since there is no way to disable
4677 its creation for now.
4679 For information, the beginning of coding on GRAS was back in june
4680 2003. I guess that every line has been rewritten at least twice since