1 SimGrid (3.30.1) NOT RELEASED YET (v3.31 expected March 20. 2022, 15:33 UTC)
3 The ненасильство release. We stand against war.
5 Against the agression by a sick system that forces peoples to take arms against each other.
9 - Rework the internals, for simpler and modern code. This shall unlock many future improvements.
10 - You can now define plugins onto the DFS explorer (previously called SafetyChecker), using the
11 declared signals. See CommunicationDeterminism for an example.
12 - Support mutex, semaphore and barrier in DPOR reduction
13 - Seems to work on Arm64 architectures too.
14 - Display a nice error message when ptrace is not usable.
15 - New test suite, imported from the MPI Bugs Initiative (MBI). Not all MBI generators are integrated yet.
16 - Remove the ISP test suite: it's not free software, and it's superseeded by MBI.
19 - fix for FG#100 by ensuring small asynchronous messages never overtake larger
20 ones, conforming to the standard.
21 - replay: fix waitall behaviour to avoid forgetting requests and leaking
23 - tracing: ensure that we dump the TI traces continuously during execution and
24 not just at the end, reducing memory cost and performance hit.
25 - Update OpenMPI collectives selection logic to match current one (4.1.2)
26 - Add a coherence check for collective operation order and root/MPI_Op
27 coherence. Potentially costly so not activated unless smpi:pendantic is set
31 - New signal: Engine::on_simulation_start_cb()
32 - Introduce a new execution mode with this_actor::thread_execute(). This simulate
33 the execution of a certain amount of flops by multiple threads ran by a host. Each
34 thread executes the same number of flops, given as argument. An example of this new
35 function can be found in examples/cpp/exec-threads.
36 - Reimplementation of barriers natively.
37 Previously, they were implemented on top of s4u::Mutex and s4u::ConditionVariable.
38 The new version should be faster (and can be used in the model-checker).
39 - Actor::get_restart_count(): Returns the number of reboots that this actor did.
42 - MSG_barrier_destroy now expects a non-const msg_barrier parameter.
44 New plugin: the Chaos Monkey (killing actors at any time)
45 - Along with the new simgrid-monkey script, it tests whether your simulation
46 resists resource failures at any possible timestamp in your simulation.
47 - It is mostly intended to test the simgrid core in extreme conditions,
48 but some users may find it interesting too.
51 - New solver for parallel task: BMF.
52 - More realistic sharing of heterogeneous resources compared to the fair
53 bottleneck solver used by ptask_L07.
54 - Implement the BMF (Bottleneck max fairness) fairness.
55 - Improved resource sharing for parallel tasks with sub-flows (parallel
56 communications between same source and destination inside the ptask).
58 - "--cfg=host/model:ptask_L07 --cfg=host/solver:bmf": enable the ptask
59 model with BMF solver.
60 - "--cfg=bmf/max-iterations: <N>": maximum number of iterations performed
61 by BMF solver (default: 1000).
62 - "--cfg=bmf/precision: <N>": numerical precision used when computing
63 resource sharing (default: 1e-12).
64 - This model requires Eigen3 library. Make sure Eigen3 is installed to use BMF.
67 - Modifications of the Profile mechanism, with some impact on users
68 - Addition of a new (S4U) method to init profiles from generic functions to improve versatility
69 - Fix initial behaviour of state_profiles
70 - Modify periodicity to behave like a period, and not like a loop delay
73 - Drop xbt_dynar_shrink().
76 - Made the following bindings static (previously member functions):
77 - Actor: Actor.kill_all(), Actor.by_pid()
78 - Host: Host.by_name(), Host.current(), Host.on_creation_cb()
79 - Mailbox: Mailbox.by_name()
80 - Added the following bindings:
81 - this_actor.warning()
82 - Mailbox.put_init() [example: examples/python/comm-waitallfor/]
83 - Comm.detach() [example: examples/python/comm-waitallfor/]
84 - Comm.wait_for() [example: examples/python/comm-waitfor/]
86 - Comm.wait_all_for() [example: examples/python/comm-waitallfor/]
87 - Mutex [example: examples/python/synchro-mutex/]
88 - Barrier [example: examples/python/synchro-barrier/]
89 - Semaphore [example: examples/python/synchro-semaphore/]
92 - Remove target "make uninstall" which was incomplete and no longer maintained.
94 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
95 (FG: issues on Framagit; GH: issues on GitHub)
96 - FG#57: Mc SimGrid should test whether ptrace is usable
97 - FG#87: Smpi scripts fail with spaces in paths
98 - FG#100: [SMPI] Order of the message matching is not guaranteed
99 - FG#101: LGPL 2.1 is deprecated license
100 - FG#104: "make uninstall" not up-to-date
101 - GH#151: Missing mutexes for DPOR.
103 ----------------------------------------------------------------------------
105 SimGrid (3.30) January 30. 2022.
107 The Sunday Bloody Sunday release.
109 Main user-visible changes:
110 - The SimDag API for the simulation of the scheduling of Directed Acyclic
111 Graphs has been dropped. It was marked as deprecated for a couple of years.
112 We finally complete the implementation of what has been called SimDag++
113 internally, i.e., porting the different features of SimDag on top of S4U.
114 The new way to simulate the execution of dependent activities directly by
115 maestro (without any other actor) is details in the examples/cpp/dag-* series
117 - The removal of SimDag led us to also remove the export to Jedule files that
118 was tightly coupled to SimDag. The instrumentation of DAG simulation is still
119 possible through the regular instrumentation API based on the Paje format.
120 - We also dropped the old and clumsy Lua bindings to create platforms in a
121 programmatic way. It can be done in C++ in a much cleaner way now, which
122 motivates this suppression.
125 - Introduce on_X_cb() functions for all signals, to attach a new
126 callback to the signal X. The signal variables are now hidden and
127 only these functions should be used.
128 Rationale: this enables the usual deprecation scheme where functions
129 remain for 4 releases if we need to modify the signals, while the
130 current code with the signal variables directly visible prevents any
132 - New function: Engine::run_until(date), to split the simulation.
133 - New signal: Activity::on_veto, to detect when an activity fails to start.
134 - Signal change: Comm::on_start(Comm&, bool) has been replaced by
135 Comm::on_send and Comm::recv. These two signals respectively correspond to
136 when the sending or receiving side of a Comm is ready. They are raised at
137 the same locations as the former Comm::on_start signal.
138 - New function: Engine::track_vetoed_activities() to interrupt run()
139 when an activity fails to start, and to keep track of such activities.
140 Please see the corresponding example for more info.
141 - New functions: s4u::Comm::{sendto_init, set_source, set_destination} to enable
142 the use of vetoers with direct host-to-host communications. Both source and
143 destination have to set for a comm to start. Each call to these setters check
144 if all vetoes are satisfied. When it is the case, the comm starts. A use case of
145 these functions is given in examples/cpp/dag-scheduling.
146 - New functions: {Exec, Io}::update_priority allow you to modify the priority of
147 these kinds of activities during their execution. Behavior is detailed in
148 examples/cpp/io-priority/.
151 - Dynamic costs for MPI operations: New API to allow users to dynamically
152 change injected costs for MPI_Recv, MPI_Send and MPI_Isend operations.
153 Alternative for smpi/or, smpi/os and smpi/ois configuration options.
154 - Fix some issues with the replay mechanism.
157 - Function xbt::Extendable::get_data() is now templated with the type of the
158 pointee. Untyped function is deprecated. Use get_data<void>() if you still
159 want to retrieve void*.
162 - New section: "SimGrid MPI calibration of a Grid5000 cluster"
163 presenting how to properly calibrate MPI communications in SimGrid.
164 - Complete and reword the platform section, which is now completed.
167 - Thread contexts are used by default with Python bindings. Other kinds of
168 contexts revealed unstable, specially starting with pybind11 v2.8.0.
170 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
171 (FG: issues on Framagit; GH: issues on GitHub)
172 - FG#95: Wrong computation time for multicore execution after pstate change
173 - FG#97: Wrong computation time for ptask+multicore+pstates
174 - FG#98: SMPI offline simulation is inconsistent with the online simulation
175 (deadlocks / message truncation)
176 - FG#99: Weird segfault when not sealing an host
178 ----------------------------------------------------------------------------
180 SimGrid (3.29) October 7. 2021
182 The "Ask a stupid question" release.
184 We wish that every user ask one question about SimGrid to celebrate.
185 On Mattermost, Stack Overflow or using the issues tracker.
188 New modeling features:
189 - Non-linear resource sharing, modeling resources whose performance heavily degrades with contention:
190 - The total capacity may be updated dynamically through a callback
191 and depends mainly on the number of concurrent flows.
192 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
194 - Dynamic factors: model variability in the speed of activities
195 - Each action can now have a factor that affects its progression.
196 This multiplicative factor is applied when updating the amount of work
197 remaining, thereby an activity with factor=0.5 only uses half of the
198 instantaneous power/bandwidth it is allocated and will appear twice
199 slower than what it actually consumes.
200 - This can be used to model a overhead (e.g., there is a 20 bytes
201 header in a 480 bytes TCP packet so the factor 0.9583) but the novelty
202 is this factor can now easily be adjusted depending on activity's and
203 resources characteristics.
204 - This existed for network (e.g., the effective bandwidth depends
205 on the message in SMPI piecewise-linear network model) but it is now
206 more general (the factor may depend on the source and destination and
207 thus account to different behaviors for intra-node communications and
208 extra-node communications) and is available for CPUs (e.g., if you
209 want to model an affinity as in the "Unrelated Machines" problem in
210 scheduling) and disks (e.g., if you want to model a stochastic
212 - For that, resources can be provided with a callback that computes
213 the activity factor when creating the action.
214 - Example: examples/cpp/exec-cpu-factors
215 - The same mechanism is also available for the latency, which
216 allows to easily introduce complex variability patterns.
219 - Added support to programmatic platform creation in Python.
220 Example: examples/python/clusters-multicpu
223 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
224 This can only be set through the API, not through XML files.
228 - Multiple fixes to ensure reproducibility of tracing
229 - scan/excan can now be replayed
230 - wait action now uses ranks and not pid, as the other ones.
231 - smpi/init and smpi/finalization-barrier are now valid for replays.
232 - exit() is now intercepted by SMPI to avoid premature shutdown of
233 simulation. First non 0 return codes is returned as simulation return
237 * New section "Release Notes" documenting recent and current developments.
238 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
239 * Improvements in API Reference for C++ and Python interfaces.
242 - Make wifi creation compatible with ns-3 version 3.34 too.
244 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
245 (FG: issues on Framagit; GH: issues on GitHub)
246 - FG#77: Search feature of doc is broken (update sphinx theme version)
247 - FG#78: Multiple fixes for SMPI replay:
248 - TI tracing of allotallv/w was outputting wrong values
249 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
251 ----------------------------------------------------------------------------
253 SimGrid (3.28) July 14. 2021
255 The Victoriadagarna Release.
258 - C++ platform interface: Users can now describe their platform directly in C++.
259 This provides greatly flexibility and performance improvement for complex
260 platforms. Main features:
261 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
262 "zones", instead of single hosts. This feature enables the description
263 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
264 - StarZone: new zone with a Star-like topology. The routes are defined
265 as a set of links used to communicate from node to everybody (node<->ALL).
266 - Split-Duplex links: auxiliary method to create split-duplex links in
267 the platform, easing its utilisation. It automatically creates both UP
268 and DOWN links (similarly as done in XML).
269 - Please refer to the documentation and the examples included:
270 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
271 - New plugin: Producer-Consumer with monitor. Just requires to include the
272 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
273 associated example (examples/cpp/plugin-prodcons).
276 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
277 s4u::Io::wait_any(), s4u::Io::wait_any_for().
278 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
279 parameter by reference, instead of a pointer.
280 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
281 Explicitly cancel the activity to get back to previous behavior.
282 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
285 - The default SMPI compiler flags are no more taken from the environment.
286 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
287 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
289 --cfg=smpi/finalization-barrier: which can be used to add
290 a barrier inside MPI_Finalize. This can help for some codes which cleanup
291 data attached to a process, but still used in other SMPI processes.
292 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
293 is active when set to false, to keep going after a small error
294 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
295 which may or may not be problematic in the end.
297 - fix behaviour, as maximum iteration count could be ignored
298 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
299 use sampling when the same kernel is called with a different set of parameters
300 which have an impact on the timing.
301 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
302 It should now work with smpi/auto-shared-malloc-thresh.
303 - Improve error handling and reporting in multiple places
304 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
305 validity checks, truncated messages are now an error, return errors
306 when explicitely deleted handles are reused, ...)
307 - RMA: multiple fixes and stability improvements.
308 - analysis (-analyze flag in smpirun):
309 - SMPI can now report buffer leaks as well as MPI handles leaks,
310 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
311 - if -trace-call-location is used when compiling, SMPI can report
312 origin of leaked handles/buffers
313 - group leaks by type/origin in output message if possible
314 - New implemented MPI calls: MPI_Comm_test_inter
317 - Changed internal implementation of bandwidth factors in network models.
318 Models affected: CM02, LV08 (default), SMPI, IB.
319 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
320 Bandwidth factors are applied to communications to describe that users
321 cannot use 100% of the available bandwidth. For example, the default network model,
322 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
323 behavior was implemented by limiting the bandwidth available in the LMM
324 system for this flow. This may give the false impression that there is
325 bandwidth available for other flows due to its underutilization, especially
326 for the dynamic bandwidth factors used in SMPI models.
327 To avoid this, we have modified the implementation so that each flow uses the
328 maximum physical bandwidth according to the LMM system.
329 However, the actual throughput of the flow seen by the user is defined by
330 the physical bandwidth multiplied by the bandwidth factor.
331 This change impacts on the simulation results for all network models on
332 which we have bandwidth factors configured.
333 *****************************************
334 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
335 *****************************************
336 This change may impact on the timing of your simulation results.
337 Take care when comparing simulations from different SimGrid's
338 versions. Sorry for the inconvenience.
339 - Dynamic network factors: users can configure a callback to define
340 the network factors dynamically. This API is available at
341 simgrid::kernel::resource::NetworkModelIntf.
342 - Users have access to complete information about the current communication
343 to decide which factor to apply. This includes: message size, source and
344 destination hosts, links and zones traversed.
345 - Dynamic factors for both latency and bandwidth.
346 - For more details, see the example in (examples/cpp/network-factors).
347 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
348 deprecated since version 3.24, are no longer supported. Instead, use
349 "wattage_off" and "wattage_per_state".
352 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
355 - New tutorial: Model-checking and formal assessment
356 - New sections: "Demystifying the routing" and "C++ platforms"
357 - Update and improve the part on visualization in MPI and Algo tutorials.
358 - Improve the section on routing: how to define it, how it's used internally
359 - Fix many issues, broken links and missing references in doxygen and Sphinx
362 - Lua platform files are deprecated. Their support will be dropped after v3.31.
365 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
367 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
368 (FG: issues on Framagit; GH: issues on GitHub)
369 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
370 - FG#64: Configuring smpi/IB-penalty-factors
371 - FG#67: Running computation concurrently with MPI_Iallreduce
372 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
373 - FG#71: Segmentation fault on invalid gw_src/gw_dst
374 - GH#322: Issue when an actor kills his host vm
376 ----------------------------------------------------------------------------
378 SimGrid (3.27) March 29. 2021
380 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
382 Main user-visible changes:
383 - All functions marked as deprecated for one year were removed.
384 - Removed all that was related to the "Storage" type in all APIs and bindings.
385 "Disks" have been introduced more than a year ago to replace them for the better.
386 It's time to finish this replacement.
387 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
388 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
389 avoiding "initializer element is not constant" errors with some codes. Please be
390 careful to recompile your MPI codes if switching to this new SimGrid version.
391 - Network model ns-3 now requires version ns-3.28 or higher.
392 - Removed some option aliases that were deprecated since v3.23 (June 2019)
393 - 'verbose-exit' is now 'debug/verbose-exit'
394 - 'simix/breakpoint' is now 'debug/breakpoint'
395 - 'clean-atexit' is now 'debug/clean-atexit'
398 - New functions: Comm::sendto_{init,async} to initiate a communication
399 on between two (possibly remote) hosts.
400 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
401 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
402 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
403 or Mailbox::get_async<void>() if you really want to play with void*.
404 - Unify the interface of Activity::on_{start/activity}
405 - New function: Comm::get_dst_data()
406 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
409 - ptask should now work better with multicore hosts (FG#37).
410 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
411 - Internal cleanups and refactoring around models, to prepare for programmatic
413 - Modernize our tests of the performance models
417 - New section: Configuring the logs from the command line.
418 - New section: Programming API of logs.
419 - New section: Getting realistic results
420 - New section: Examples of SMPI Usage
421 - New part: SimGrid's Internals, with a section on Design goals
422 - Move back to the classical Breathe tool chain, and fix many small glitches.
425 - Handle leak checking: SMPI will now display a message if MPI handles are still
426 present at the end of a simulation. Option smpi/list-leaks:n will display the n
427 first leaked pointers and their types.
428 - smpi/display-allocs option : Display Memory allocations analysis, with hints
429 about the largest allocations. This only works for allocations with malloc/calloc
430 and not compiled with SMPI_NO_OVERRIDE_MALLOC
431 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
432 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
433 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
434 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
435 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
436 - Replay now opens and parses the trace file only once for all ranks when possible.
438 - better handling of MPI_PROC_NULL
439 - MPI_Group_incl was broken in some cases
440 - PAPI tracing now fixed
441 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
442 - SMPI will now fail properly when receive size < sent size, instead of
443 just setting MPI_ERR_TRUNCATE and keep going.
444 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
445 - Remove useless build patches and split build/run patches on the page
446 - Build in parallel when possible
447 - Test on yum-based systems
448 - New nightly tested Proxy-apps:
449 - OpenMPI mpi-test-suite
450 - Coral MCB, LSMS, Qbox, CLOMP
451 - ECP/Mantevo Cloverleaf3D, Ember
452 - Codevault bhtree_mpi
455 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
456 - Start to overhaul the implementation to modernize it (still ongoing).
459 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
460 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
461 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
462 - Deprecate xbt_dynar_to_array().
463 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
465 C binding and interface:
466 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
467 - Constness requirement has been lowered for parameter argv of sg_actor_start()
468 and sg_actor_create() It's now (char* const*). The previous version is still
469 available as sg_actor_start_() and sg_actor_create_().
471 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
472 (FG: issues on Framagit; GH: issues on GitHub)
473 - FG#37: Parallel tasks are limited to 1 core per host
474 - FG#62: Running "smpirun -replay" on large networks
475 - FG!46: Fix a few potential memory leaks in SMPI colls
476 - GH#343: Wrong consumed energy when an exec timeouts
478 ----------------------------------------------------------------------------
480 SimGrid (3.26) December 15. 2020
482 The Release release (the French lockdown was eased today).
484 Important user-visible changes:
485 - SimGrid now requires a compiler with C++14 support.
486 Sibling projects should upgrade their FindSimgrid.cmake
487 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
488 several users had difficulties to understand issues when using high bandwidth or
489 small latency events. The new value was already the default for SMPI and
490 should not cause too much performance hit. It can change some simulation
491 timings if simulation was skipping a lot of small events (which was actually
492 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
493 - On some network models (L07, CM02, LV08), an implicit loopback link is created
494 for local communication. Default bandwidth was 500MBps and latency was 15µs,
495 which was rather bad. Default values have been changed to 10GBps and 0 latency.
496 Options network/loopback-lat and network/loopback-bw have been added to adapt
497 these values if needed. Network and routing models already providing separate
498 loopbacks (clusters and such) are not impacted by this change.
499 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
500 metadata that are added to each MPI message by MPI libraries (tag,
501 src/dest, and communicator information). This may change simulation times
502 for very small message sizes, but should be more realistic.
505 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
506 - New plugins: wifi_energy and link_load. See the associated examples.
507 - New examples: wifi networks, and communication suspend.
510 - LTO is now enabled for Intel/clang compilers.
511 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
512 Setting it to "auto" will use all available cores, while setting it to n will
513 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
514 - Remove obsolete runtime option 'contexts/parallel-threshold'.
515 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
516 has been kept for compatibility.
517 - Finally remove obsolete snake_case() aliases for runtime options.
518 - Further improve the documentation.
521 - the on_exit() of each actor is also executed when the simulation deadlocks.
522 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
523 An example is provided for s4u::Comm but it should work with Exec and Io.
526 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
527 https://framagit.org/simgrid/SMPI-proxy-apps
528 - MPI/IO: supports names starting with "./"
529 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
531 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
532 building shared libraries, this may cause issues, so environment variable
533 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
534 - most temporary files should now be created in /tmp dir (or equivalent).
535 If this one does not allow execution of code (noexec flag), this may cause issues.
536 Please use another tmp directory (using TMPDIR or equivalent system variable)
540 - A large refactoring is ongoing, but this should be transparent for users so far.
543 - Drop xbt_str_split().
545 C binding and interface:
546 - The return type of the sg_actor_on_exit() callbacks is now 'void'
547 (int value was previously ignored)
548 - Many C functions were renamed to follow a common naming style
549 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
550 names are kept until version 3.30.
551 - Many MSG tests were converted to the new S4U's interface in C, that
552 was extended for that.
554 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
555 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
556 - FG#41: Add sg_actor_create C interface
557 - FG#43: xbt::random needs some care
558 - FG#48: The Impossible Did Happen (yet again)
559 - FG#50: Suspending an actor executed at the current timestamp fails
560 - FG#52: Zero-seconds timeout : "That's in the past already"
561 - FG#53: Crash while using ns-3 network model
562 - FG#54: How to suspend a comm?
563 - FG!22: Stochastic Profiles
564 - FG!24: Documentation and fix for xbt/random
565 - FG!35: Add a modeling hint for parallel links in doc
566 - FG!36: [xbt/random] Read/Write the state of the RNG
567 - FG!42: Link cumulated load plugin
568 - FG!44: Test condition variable wait_until+notify_all
569 - GF#18137: Allow different stack sizes?
570 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
571 - GH#139: Allow pthread creation in SMPI
572 - GH#336: Packet-level simulation using SMPI?
573 - GH#345: Error in the LMM while migrating a VM
574 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
575 - GH#352: pip install / python setup.py install fail to find pybind11
576 - GH!337: Fix link_energy plugin for wifi platforms
577 - GH!339: Add Mailbox set_receiver method to python binding
578 - GH!344: Cast hugepages macros parameters to int64
580 ----------------------------------------------------------------------------
582 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
584 The Palindrom Day Release.
586 Important user-visible changes:
587 - Improve the Python usability (stability and documentation).
588 - A nasty synchronization bug was ironed out, see also below.
589 - Python's doc was organized as a separate tree, now integrated with C++.
590 - C bindings of S4U were not part of the doc.
591 - The C++ doc was also improved as methods are now split by theme.
592 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
593 - This is mandatory to use the Java bindings.
594 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
595 - SimDAG++: Automatic dependencies on S4U activities (experimental)
596 - Some features are already implemented but not all of them
597 - Cannot block an activity until it's scheduled on a resource
598 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
599 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
602 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
603 - Drop support for Viva/Triva (old visualization tools), hence removed
604 TRACE_get_node_types() and TRACE_get_edge_types() functions.
607 - Require pybind11 version 2.4.
608 - Greatly improve locking during context switches with Python.
609 - Declaring Python callbacks to this_actor::on_exit is now working.
610 - No more tests are failing currently, even on win32.
613 - Actor: Merge signals on_migration_start/end into on_host_change
614 - Actor: Rename migrate() into set_host()
615 - Disk: Allow users to get the read and write nominal bandwidth values
616 - Exec: Implement wait_for(timeout)
617 - Io: Implement wait_for(timeout)
620 - Parse errors now raise a simgrid::ParseError that you may want to catch.
623 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
624 been renamed to fix a spelling error. As usual, the previous definitions are
625 kept with a deprecation warning. However, to avoid runtime errors, it is now
626 forbidden to override these deprecated methods in a derived class. Please use
627 the new names immediately if you need to override them.
630 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
631 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
632 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
633 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
634 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
635 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
636 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
637 - New module: random, providing classical random numbers generators.
640 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
641 allocations' size above which they are considered "shared" by default (as if
642 they were performed through SMPI_SHARED_MALLOC macros).
643 Default = 0 = disabled feature.
644 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
645 This can cause some troubles if codes are already overriding these. If this
646 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
647 help, but will make this feature unusable.
648 - Input check is now performed much more consistently and outputs precise
649 warnings for each error found
650 - New supported calls : MPI_File_set_view, MPI_File_get_view
651 - MPI I/O now only supports disks, as storage is discontinued.
652 - MPI I/O allows opening files without specifying absolute path
653 - Note : SMPI configuration options are now initialized by smpi_init_options()
654 call, which should be called if SMPI is not being used through smpirun.
656 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
657 - FG#9: Python bindings crashing
658 - FG#39: Missing s4u::Comm::wait_any_for example
659 - FG#42: Add support for ThreadSanitizer (TSan)
660 - FG!19: Removing RngStream
661 - FG!20: A module for RNG calls
662 - FG!21: Choice between ad-hoc and standard distributions implementations
664 - FG!25: Fix link in Mutex doc
665 - FG!26: Fix links in SMPI interface doc
666 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
667 - GH#305: Unscheduled tasks are still excuted
668 - GH#313: smpirun: manual outdated w.r.t. --help
669 - GH#321: [S4U] Get task remaining work ratio
670 - GH#323: Crash when an actor turn off his physical host
671 - GH#335: Missing links on dragonfly example svg
672 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
673 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
675 ----------------------------------------------------------------------------
677 SimGrid (3.24) October 9. 2019
679 The Clean Disk Release.
682 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
683 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
684 thus becomes the default resource to execute I/O operations. Examples
685 and the file system plugin have been modified to use disks instead of
686 storage elements. Storage can still be used but is doomed to disappear.
687 Users are thus adviced to update their simulators accordingly.
688 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
689 for consistency with pthread_barrier_wait()
690 - Host::get_englobing_zone() returns the englobing netzone
691 - Actor::on_destruction is now called in the destructor
692 Actor::on_termination new signal called when the actor terminates
694 - Global signals are now part of the Engine:
695 - on_platform_creation: after config settings, before the XML parsing
696 - on_platform_created: right after the XML parsing
697 - on_time_advance: each time the clock advances
698 - on_simulation_end: after simulation, before cleanups
699 - on_deadlock: as the name implies.
701 - sg_{actor,host,link}_{data,data_set}() now all exist.
702 Use them to attach user data to the object and retrieve it.
705 - Introduce an experimental Wifi model. It sounds reasonable
706 according to the state of the art, but it still has to be properly
707 validated, at least against ns-3 if not against reality.
708 - Improved the usability of ns-3. Several bugs were ironed out.
709 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
710 It is now expressed as 'idle:epsilon:allCores' to properly model the
711 consumption of non-whole tasks on mono-core hosts. More info in the doc.
714 - convert a new set of functions to the S4U C interface and move the old MSG
715 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
718 - Fortran bindings for DVFS have been removed.
719 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
720 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
721 - SMPI can now be selected by cmake's find_module(MPI) with
722 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
723 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
724 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
725 - trace-call-location can be used with TI traces, and replayed, and is compatible
726 with smpi/comp-adjustment-file.
727 - sleep events are now correctly traced/replayed.
728 - Default for trace-call-location is now to use file names and not full paths.
729 To revert to previous behavior (in case of collision of filenames), option
730 "smpi/trace-call-use-absolute-path" can be set to yes.
733 - Use the included xxHash as an hash implem when C++14 is usable.
734 - Option model-checker/hash was removed. This is always activated now.
735 - New option smpi/buffering controls the MPI buffering in MC mode.
736 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
737 This is useful to check for MPI compliance.
740 - New section on plugins: how to define a new one, and existing ones.
743 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
744 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
747 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
748 and <mount> tags. The rationale is finish to decouple what should be handle
749 by SimGrid kernel (i.e., resources and activities) from plugins built in
750 user space (e.g., the file system plugin). This new tag currently accepts
751 only three attributes (id, read_bw, and write_bw). All the other information
752 that was declared with the storage related tags now has to be expressed as
753 properties. An example of platform using this new tag is available at
754 examples/platforms/hosts_with_disks.xml
757 - 'expect signal' can now accept more than one potential signal.
759 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
760 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
761 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
762 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
763 - FG#31: per-actor data
764 - FG#34: SG_BARRIER_SERIAL_THREAD?
765 - FG#35: model-checker does not like buster-produced binaries
766 - FG!13: MC: complete workaround in the error msg seen on modern systems
767 - FG!15: execute_flops now logs compute
768 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
769 - FG!17: ns-3: unblock the right number of communications + others issues
770 - FG!18: Improving the performance of the ns-3 bindings
771 - GH#189: Energy consumption of parallel tasks
772 - GH#219: Error in the throughput of TCP transfer
773 - GH!330: Minor polishing of English: Configuring_Simgrid
774 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
775 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
776 - GH!334: Rework energy plugin, again
778 ----------------------------------------------------------------------------
780 SimGrid (3.23.2) July 8. 2019
783 - Nicer introduction page.
784 - Migrate the "Deploy your application" page to the new doc.
785 - Move Java as a subtree of MSG.
788 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
791 - Drop xbt_dynar_sort_strings().
794 - Really fix FG#26: Turning off a link should raise NetworkFailureException
795 - FG#27: Wrong exception thrown to wait_any when link is turned off
796 - GH#328: Java: Canceling multiple tasks in a single vm/host
798 ----------------------------------------------------------------------------
800 SimGrid (3.23) June 25. 2019
802 The Exotic Solstice Release.
805 - SunOS and Haiku OS support. Because exotic platforms are fun.
806 - Stop setting random seed with srand48() at initialization.
807 - Use addr2line as a fallback for stacktraces when backtrace is not available.
808 - Build option -Denable_documentation is now OFF by default.
809 - Network model 'NS3' was renamed into 'ns-3'.
812 - Simgrid can now hopefully be installed with pip.
815 - wait_any can now be used for asynchronous executions too.
818 - New log appenders: stdout and stderr. Use stdout for xbt_help.
819 - Drop xbt_dict_dump.
822 - 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
823 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
824 - MPI interface is now const correct for input parameters
825 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
826 - Fortran bindings for async collectives.
827 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
830 - Remove option 'model-check/record': Paths are recorded in any cases now.
831 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
832 always sparse. That's an immense gain in memory, and can even be
833 faster because of cache effects. So there is no need to clutter the
834 code to allow the user to go for the unefficient mode.
837 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
838 models proved to be more accurate than these old experiments.
840 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
841 - FG#1: Broken link in error messages
842 - FG#2: missing installation documentation
843 - FG#3: missing documentation in smpirun
844 - FG#6: Python bindings not available on PyPI
845 - FG#7: simple cmake call requires doxygen
846 - FG#8: make python bindings an optional dependency
847 - FG#10: Can not use MSG_process_set_data from SMPI any more
848 - FG#11: Auto-restart actors forget their on_exit behavior
849 - FG#12: -Denable_lto=OFF doesn't disable LTO
850 - FG#13: Installs unstripped file 'bin/graphicator'
851 - FG#14: Installs the empty directory 'doc/simgrid/html'
852 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
853 - FG#17: Dead link in doc (pls_ns3)
854 - FG#20: 'tesh --help' should return 0
855 - FG#21: Documentation link on http://simgrid.org/ broken
856 - FG#22: Debian installation instruction are broken
857 - FG#26: Turning off a link should raise NetworkFailureException exceptions
858 - GH#133: Java: a process can run on a VM even if its host is off
859 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
860 - GH#326: Valgrind-detected error for join() when energy plugin is activated
862 ----------------------------------------------------------------------------
864 SimGrid (3.22) April 1. 2019
866 The Easter Christmas Release (this one is somewhat late).
869 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
870 - Not complete yet: asynchronous activities (amongst others) are still missing
871 - Still ongoing: the interface may change in the future. We need more
872 testers! Please report any glitches.
873 - No new project using Java should start now. Please switch to Python.
876 - Some of the internal cleanups may lead to speed improvements:
877 - The hard limitation on the amount of simulated actors with
878 Java+Mac was removed. Now, the available memory is the only limit.
879 - Our refcounting was tidyied, leading to 10% speedups in some cases.
880 - We are still working on making our code robust to the actor kills
881 and hosts' churn. Things are improving, but it's not perfect yet.
882 - Replaced our own code to display a backtrace (that was forking addr2line)
883 with the Boost.Stacktrace library.
884 You won't see your backtraces without this optional dependency.
885 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
886 - Stop setting random seed with srand() at initialization.
889 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
890 XML file version remains 4.2 since old files are still compatible.
893 - Process termination which was broken at version 3.21 has been repaired.
894 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
895 - Hide the examples into examples/deprecated. New users should use Python.
898 - Drop MSG_process_create_from_stdfunc() from the API.
899 This C++ function was a pimple in the C API, made necessary at some
900 point by the Java bindings. This is fixed now.
901 - Hide the examples into examples/deprecated. New users should use S4U.
902 - MSG_process_create and MSG_process_attach now crash if the host used
903 to run this process is off.
904 - Fix the protype of MSG_process_on_exit()
905 Now use (int,void*) callbacks instead of (void*,void*) ones.
906 The implementation was ways too messy to actually work, I guess.
909 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
910 too often and provide more accurate timings.
911 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
912 - MPI_Alltoallw support
913 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
914 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
915 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
916 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
918 - Basic implementation of generalized requests (SMPI doesn't
919 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
922 - Drop sg_cmdline. Please use xbt_cmdline instead.
923 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
924 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
925 OsSemaphore is implemented in a portable way with C++11 threads.
926 This should allow much more threads to be created at the same time,
927 allowing Mac and Java users to simulate many more actors.
928 - Implement the 'thread' factory with std::thread instead of xbt ones.
929 It is not possible to set the stack size with threads anymore, but
930 -fsplit-stack is the way to go nowadays when using threads.
931 - Drop the xbt_os_thread_t module (now unused)
932 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
933 - Drop xbt_str_join_array().
934 - Drop cunit, use Catch2 instead.
937 - Many cleanups in the kernel::activity namespace. This was long
938 overdue, and shall open the path to many future endeavors.
941 - #132: Java: a process can not shut down its own host
942 - #220: S4U: detached send causes exception if sender terminates soon after sending
943 - #261: Document the parameters of parallel execution's constructor
944 - #300: [s4u] BarrierPtr is missing
945 - #314: SMPI args internal cleanup
946 - #316: Fix a bug related to the CPU utilization of multi-core VM
947 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
948 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
949 - #325: Turning off a host has different behavior on sleeping actors and computing actors
951 ----------------------------------------------------------------------------
953 SimGrid (3.21) October 3. 2018
955 The Restarting Documentation (TRD) Release.
958 - Convert the user manual to Sphinx for improved usability.
959 Unstable version now hosted on https://simgrid.frama.io/simgrid
960 Many glitches remain; stay tuned.
961 - Restrict the Doxygen reference API to the installed header files.
962 (documenting all internals this way is overkill + doxygen scales badly)
963 - New tutorial on S4U (highly inspired from the old MSG one, but with
964 a git repository to fork as a starting point, and a docker image)
965 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
966 - The Developper manual is still to be converted (not compiled until then)
967 - Some parts are still missing in this conversion (such as the
968 platform and deployment sections) while others were blindly converted
969 and would need more love (such as the configuration flags).
970 Things will certainly further improve in the future releases.
973 - Fully reimplement auto_restart mechanism that was utterly broken
974 (fix #22, #131 and #281 that were all very related)
975 - Implement semaphores (fix #303 and #291)
976 - When creating an actor from a function and its parameters,
977 move-only parameters are not allowed anymore, as it would prevent
978 the actor to be restartable if its parameters are consumed on run.
979 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
980 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
981 IO operations that can be started, waited for, or canceled as a regular
983 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
987 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
988 - New host variable: 'core_count'
991 - Remove xbt_os_thread_specific features
992 - Remove portability wrapper to condition variables
993 - Remove xbt_os_thread_yield()
996 - MPICH collective selector now mimics MPICH 3.3b
997 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
998 - MPI_Init(NULL,NULL) is now allowed
999 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
1000 - Command-line arguments should now be handled properly in Fortran simulations
1001 - MPI Attributes and related callbacks should work in Fortran
1002 - Apps using C + Fortran should now work
1003 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
1004 - Support MPI_Error_String
1007 - Due to an internal bug, Msg.run() must now be your last line.
1008 We hope to fix it in a future release, and we are sorry for the inconvenience.
1011 - #22: Process autorestart seem to only work with CAS01 cpus
1012 - #56: Feature request: dynamic SMPI replay
1013 - #93: simgrid should not eat --help
1014 - #111: How to change the loopback link characteristics?
1015 - #116: Platform cluster radical
1016 - #117: Platform cluster/host order
1017 - #129: Replay_multiple on steroid
1018 - #131: Java : autostart does not work
1019 - #143: Setting a breakpoint at a given time
1020 - #235: Network energy models should be integrated
1021 - #264: Add ptask L07 resource tracing
1022 - #271: Dynamic and manual replay of SMPI traces
1023 - #279: Breakpoints option uses time deltas instead of absolute time
1024 - #281: Daemonized actor no longer daemonized after an auto-restart
1025 - #282: TIT convertor erase traces when using absolute path in the trace list file
1026 - #285: segfault when a s4u actor kills itself with msg process tracing activated
1027 - #286: Pajé traces are not exposing the number of cores
1028 - #287: Command-line parsing should stop when encountering '--'
1029 - #288: MPI_Init(NULL, NULL)
1030 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
1031 - #290: Method to check if a mailbox has a message ready to be consumed
1032 - #291: [s4u] semaphores do not seem available yet
1033 - #294: Alltoall collective implementation of mpich
1034 - #295: Starting a Host as OFF is not supported
1035 - #296: DTD too permissive
1036 - #299: Add a s4u barrier example
1038 ----------------------------------------------------------------------------
1040 SimGrid (3.20) Released June 24. 2018
1042 The proxy snake_case() release.
1045 - The whole API is now using snake_case() instead of mixing it with camelCase.
1046 Compatibility wrappers in place for one year.
1047 - We took the opportunity to increase the consistency of the API.
1048 Please report any remaining glitches.
1049 - New function simgrid::s4u::Host::get_actor_count:
1050 Returns the number of actors running on a specific host.
1053 - SMPI is now tested with ~45 proxy apps from various sources, with none or
1054 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
1055 - Replay: The replay file has been re-written in C++.
1056 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
1057 supported. They are stored in the trace and used when replayed.
1058 - Basic support of MPI_Cancel. Robustness not guaranteed.
1059 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
1060 (only for MPI_COMM_TYPE_SHARED).
1061 - New option: smpi/privatize-libs, to add external shared libs to be privatized
1062 by SMPI. They will be copied locally and loaded separately by each process.
1063 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
1064 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
1065 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
1066 with internal index optind with multiple processes. Only works if MPI_Init has
1067 already been called.
1068 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
1069 syntax without preprocessing tricks.
1072 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
1073 behavior tracing is no longer limited to MSG
1074 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
1075 this means that traces now need to be updated or re-obtained; both Irecv
1076 and Isend lines in your traces have now in total 5 mandatory fields:
1077 <rankid> <command> <to/from rankid> <tag> <size>
1078 To update your traces, it suffices to add a 0 for the tag here.
1079 - TIT now also supports waiting for a distinct request via MPI_Wait.
1080 Wait/Test now wait for a specific request, not just the last one that was
1081 issued. This unfortunately means another update, because we need to
1082 identify which request you want to wait for. We do this via the
1083 triplet (sender, receiver, tag), which needs to be added:
1084 <rankid> <command> <sender> <receiver> <tag>
1085 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
1086 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
1089 - The deprecation of MSG is ongoing (but this should not impact you).
1090 Many MSG functions are now simple wrappers on the C API of S4U. If
1091 you wish to convert your code to S4U, find the S4U counterparts of
1092 your MSG calls in src/msg/msg_legacy.cpp.
1093 - MSG can still be used, but won't evolve anymore.
1096 - Allow to run the Link energy plugin from the command line with
1097 --cfg=plugin:link_energy
1098 - Rename Energy plugin into host_energy
1099 - Rename Load plugin into host_load
1102 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
1103 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
1104 - Improve performance when handling timeouts of simix synchros.
1107 - Config: the C API is now deprecated (will be removed in 3.23), and
1108 the C++ API has been slightly improved.
1111 - Fix several build issues on OSX.
1112 - Move simgrid_config.h to simgrid/config.h (old header still working)
1115 - #143: Setting a breakpoint at a given time
1116 - #150: Inconsistent event names in SMPI replay
1117 - #258: daemonized actors hang after all non-daemonized actors have completed
1118 - #267: Linker error on unit_tmgr
1119 - #269: SMPI: tracing of MPI_Wait/all/any broken
1120 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
1122 ----------------------------------------------------------------------------
1124 SimGrid (3.19.1) Released March 22 2018
1126 The Fixed ABI Release.
1128 As you may know, we are currently refactoring SimGrid in deep.
1129 Upcoming SimGrid4 will be really different from SimGrid3: modular,
1130 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
1132 Our goal is to smooth this transition, with backward compatibility
1133 and automatic update paths, while still progressing toward SimGrid4.
1135 SimGrid remains open during works: The last pure SimGrid3 release was
1136 v3.12 while all subsequent versions are usable alpha versions of
1137 SimGrid4: Existing interfaces remain unchanged, but the new S4U
1138 interface is budding and the internals are deeply reorganized.
1140 Since 2015, we work hard to reduce the changes to public APIs. When
1141 we need to rename a public library symbol in S4U, we let your
1142 compiler issue an explicative warning when you use the deprecated
1143 function. These messages remain for four releases, i.e. for one full
1144 year, before turning into an error. Starting with v3.15, your can
1145 also adapt to API changes with the SIMGRID_VERSION macro, that is
1146 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
1148 Starting with this v3.19.1, our commitment to reduce the changes to
1149 the public interfaces is extended from the API to the ABI: a program
1150 using only MSG or SimDag and compiled against a given version of
1151 simgrid can probably be used with a later version of SimGrid without
1152 recompilation. We will do our best... but don't expect too much of
1153 it, that's a really difficult goal during such profund refactoring.
1155 The difference between v3.19 and v3.19.1 is that the former was
1156 accidentally breaking the ABI of MSG, while the later is restoring
1159 S4U and kernel APIs will still evolve until SimGrid4, with one-year
1160 deprecation warnings as currently. In fact, cleaning up these
1161 interfaces and converting them to snake_case() is one release goal of
1162 v3.20. But don't worry, we are working to smooth this upgrade path.
1164 In summary, new projects should start with S4U to benefit of the
1165 future, but old MSG projects should still be usable with no change.
1167 ----------------------------------------------------------------------------
1169 SimGrid (3.19) Released March 20 2018
1171 The Moscovitly-cold Spring release.
1173 SMPI user-visible improvements:
1174 - Support of MPI_Type_create_subarray
1175 - Unimplemented MPI calls now abort (instead of issuing a warning)
1176 Some harmless calls around errhandlers still only issue a warning.
1177 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
1178 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
1179 is now possible (mmap privatization was refactored to this end).
1180 Adding new SMPI instances should also be technically doable now.
1182 SMPI internal cleanups:
1183 - Removed index notion from SMPI -- no more "getPid() - 1"!
1184 In tracing files, actors are called rank-<PROCESS-ID> but these are
1185 NOT the real ranks (tracing several communicators is not supported yet).
1186 - Deprecate smpi_process_index() (will be removed in 3.22)
1187 - Better testing on classical proxy apps, through an external project:
1188 https://github.com/simgrid/SMPI-proxy-apps/
1191 - Use a graphical TOC to make it easier to find the documentation you need
1192 - Revamp the MSG tutorial
1195 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
1196 - Remove parameter of MSG_process_killall().
1197 Resetting the PID was bogus anyway (several actors could have the same PID).
1200 - Execution->setHost() can be called after start(), to migrate it.
1201 - Comm::test_any() is now implemented.
1202 - s4u::Actor now has onCreation() and onDestruction() signals.
1203 - Install some kernel header files for the users' plugins and more.
1204 Warning, their API is really not stable yet.
1205 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1206 include/simgrid/kernel/routing/*Zone.hpp
1209 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1211 Build System and other cleanups:
1212 - Remove unused run-time parameter "tracing/onelink-only".
1213 - SimGrid now works with intel compilers (MC still buggy)
1214 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1215 - tesh kills subprocesses on timeout (but not on windows)
1218 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1221 - #194: Feature request: simgrid::s4u::Comm::test_any()
1222 - #245: migrating an actor does not migrate its execution
1223 - #253: Feature Request: expose clusters as objects
1224 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1225 - #255: Tesh broken on Windows
1226 - #256: Modernize FindSimGrid.cmake
1227 - #257: Fix (ab)use of CMake install
1229 ----------------------------------------------------------------------------
1231 SimGrid (3.18) Released December 24 2017
1233 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1235 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1236 - Support for MSG should not be removed from SimGrid before 2020,
1237 but future evolutions will be limited to the integration of
1238 user-provided patches.
1241 - s4u::onDeadlock() signal, run before stopping the simulation.
1242 - s4u::Exec: asynchronous executions (abstraction of background
1244 - s4u::Host->getLoad() returns the achieved speed in flops/s
1247 - Link::name() is deprecated and will be removed in v3.21, use
1248 Link::getCname() instead.
1249 - Mailbox::getName() changed to return a std::string, use
1250 Mailbox::getCname() instead to get a char*.
1251 - Storage::getName() changed to return a std::string, use
1252 Storage::getCname() instead to get a char*.
1253 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1254 to have both allocation and deallocation of the map in user space.
1257 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1258 The amount of flops remaining to do is a vector, not a scalar.
1259 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1260 name implies on both sequential and parallel tasks.
1261 - Both changes fix GitHub's #223 using PR #237 as a basis.
1262 Thanks Michael Mercier.
1263 - Most examples were converted to S4U and hidden elsewhere at tests.
1264 Rationale: we still want MSG to work; we want newcomers to use S4U.
1267 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1268 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1271 - Switch to the faster dlopen privatization mechanism by default
1272 - Documentation improvments
1275 - Live migration is getting moved to a plugin. Dirty page tracking is
1276 the first part of this plugin. This imply that VM migration is now
1277 only possible if one this function is called:
1278 - C/MSG: MSG_vm_live_migration_plugin_init()
1279 - C/C++: sg_vm_live_migration_plugin_init()
1280 - Java: Msg.liveMigrationInit()
1281 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1285 - Define class simgrid::xbt::Path to manage file names.
1286 - Removed unused functions:
1287 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1288 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1289 xbt_graph_export_graphviz()
1290 - xbt/str.h: xbt_str_join()
1291 - Remove unused datatypes:
1292 - xbt/heap.h: use std::priority_queue or boost::heap instead
1293 - xbt/swag.h: use boost::intrusive::list instead
1296 - New link_energy plugin for the consumption of the links.
1297 - All of the operations on files and storage contents have been
1298 packaged into a plugin (src/plugins/file_system). The current
1299 public interface can be found in
1300 include/simgrid/plugins/file_system.h
1301 To use these functions you now have to initialize the plugin by
1302 calling MSG_storage_file_system_init() just after calling
1303 MSG_init() or sg_storage_file_system_init() just after creating
1307 - Remove the undocumented/untested tag <include>
1310 - Remove viva specific tracing as the tool is no longer maintained
1313 - #248: Kill of finished processes leads to segfault
1314 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1315 - #239: please implement signal s4u::onDeadlock()
1316 - #230: segfaults when exit() without run()
1317 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1318 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1319 - #222: Actor::kill() doesn't really kill and segfaults
1320 - #221: odd LMM warning when killing an actor
1321 - #120: Memory leak when the processes are forcefully killed
1323 ----------------------------------------------------------------------------
1325 SimGrid (3.17) Released October 8 2017
1327 The Drained Leaks release: (almost) no known leaks despite the tests.
1329 Even very long simulations will preserve your memory: our testsuite
1330 stresses SimGrid for over 45mn on fast machines for a coverage of
1331 over 80%, and there is only one single known leak, of about 4kb.
1335 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1336 handler may be given to free resources if the comm cannot be completed.
1337 - this_actor::send(mailbox) is now mailbox->put()
1338 - New: simgrid::s4u::Comm::wait_all()
1339 - New: Host.execute() for remote executions.
1342 - Unused option network/sender-gap is removed.
1345 - Deprecate MSG_task_isend_with_matching(): This unused feature
1346 really complicates our internals. Will be removed in v3.20.
1349 - Improved context termination. It is now possible to free resources, even
1350 when a process is forcibly killed.
1351 - ContextBoost: add support for Boost versions above 1.61.
1354 - Bring back run-time option --cfg=exception/cutpath to remove exception
1356 - Removed unused functions:
1357 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1358 xbt_str_rtrim(), xbt_str_trim().
1359 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1362 - Removed header files obsolete since SimGrid 3.12:
1363 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1365 - #3: SD_exit should be made optional
1366 - #120: Memory leak when the processes are forcefully killed
1367 - #159: Threading test regression in Actor refcounting
1368 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1369 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1370 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1371 - #191: VM migration and pstate
1372 - #192: Updating the energy consumptions of all hosts crashes with VMs
1373 - #195: All actors have PID=0 in the logs
1374 - #204: Sometimes segfault with thread contexts and mmap privatization
1375 - #222: Actor::kill() doesn't really kill and segfaults
1376 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1378 ----------------------------------------------------------------------------
1380 SimGrid (3.16) Released June 22. 2017.
1382 The Blooming Spring Release: developments are budding.
1385 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1386 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1387 - Java JAR file should be finally fully working on Mac OSX too.
1389 XML platforms: Switch to platform v4.1 format.
1390 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1391 - <zone> can be used as a synonym for the now deprecated <as>
1392 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1393 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1394 - <actor> can be used as a synonym for the now deprecated <process>
1395 - state_file and avail_file periodicity is now easier to express
1396 (check the documentation) the old behavior should still work.
1397 * Storage: not backward compatible, but it seems that nobody ever used it
1398 - Remove attribute 'content_type' of <storage_type>: was never used
1399 - Make attribute 'model' of <storage_type> optional: for future usage
1400 - Remove Bconnection model property: was never (in)validated. Replaced by
1401 the maximum of the read and write bandwidth as a resource constraint.
1404 - New and Backwards Compatibility break:
1405 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1406 When one wants to get the list of tasks whose states have changed during a
1407 simulation round, s/he has to allocate and free a dynar and use it as
1408 argument to this function. The former SD_simulate (double how_long)
1412 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1413 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1416 - New function to update the consumption of all hosts at once.
1417 - Fix the model for multi-core hosts, linear in the amount of busy
1418 cores with an abnormality for pIdle (see paper).
1419 BUG: Still not satisfactory for ptask on mono-cores.
1422 - The netzone are now available from the MSG API.
1423 The old names still work, but are now deprecated.
1426 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1427 - New option: smpi/keep-temps to not cleanup temp files
1428 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1429 mallocs mode is used, allows to change the size of the fake file used
1430 (default 1MB), to potentially limit the number of mappings for large runs.
1431 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1432 - Fortran ifort and flang compilers support
1433 - New RMA calls supported (experimental):
1434 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1435 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1436 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1437 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1438 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1439 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1442 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1443 - Behind the scene, ActivityImpl are now automatically refcounted.
1444 Removing this nasty bottleneck will greatly simplify our internals.
1447 - Replay: New function xbt_replay_action_get():
1448 Retrieve the function previously associated to an event type.
1449 - DROPPED FUNCTION: xbt_str_varsubst()
1450 - DROPPED MODULE: strbuff. We don't need it anymore.
1451 - DROPPED MODULE: matrix. We don't need it anymore.
1452 - DROPPED MODULE: lib. We don't need it anymore.
1454 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1456 SimGrid (3.15) Released March 22. 2017
1458 The Spring Release: continuous integration servers become green
1459 We fixed even the transient bugs on all target architectures:
1460 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1461 Windows, FreeBSD, NetBSD.
1463 - Rename NetCards to NetPoints.
1464 This was intended to help NS3 users, but that's not a netcard.
1465 That's a point in the routing algorithm, let's avoid wrong simplifications.
1468 - New: MSG_process_yield(). Stop and yield to other processes.
1469 - New: MSG_process_daemon(). Daemon processes are automatically killed
1470 when the last non-daemon process terminates
1471 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1472 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1473 to make room for the upcoming network energy plugin.
1474 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1477 - Ensure that an actor can kill itself with Process::exit()
1478 - Kill the obscure NativeException. Nobody want to survive the issues
1479 it denotes, so use JniException that is a RuntimeException (not to
1480 be caught explicitly).
1481 - Partial bug fix in initialization. SimGrid flags on command line were
1482 consumed at C level but stayed in the original Java String[] args.
1483 This could mess users' args[i] if SG flags were not put at the end of
1485 The SimGrid flags are now removed from the Java arguments. However,
1486 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1487 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1488 It might be if you have --log or --cfg flags in the command line.
1489 - Fix numerous memleaks all around the place. In particular, around VMs.
1493 - s4u::onPlatformCreated: right before the simulation starts
1494 - s4u::onSimulationEnd: right after the main simulation loop
1495 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1496 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1497 event from the availability_file changes the avail speed.
1498 - Links are now usable from s4u
1499 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1500 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1501 - New examples: The conversion of MSG examples to S4U has begun
1502 - Actors: create, kill, migration, and suspend
1503 - Applications: master-worker and token-ring
1504 - Action replay: communications and storage
1505 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1508 - Backwards Compatibility breaks
1509 - The SD_route_get_list and SD_route_get_size functions have been removed.
1510 They are replaced by the unique
1511 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1512 The route from src to dst is built in the links dynar whose size can
1513 be obtained with xbt_dynar_length.
1514 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1515 sg_host_route_bandwidth and sg_host_route_latency.
1516 Macros ensure the backwards compatibility, but you should fix your code
1519 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1520 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1521 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1522 We don't support 100% of it, but it should be enough. Contact us if not.
1523 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1524 naive and probably inaccurate)
1525 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1526 mapped onto a unique small file using some system magic.
1527 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1528 - Fortran: cleanups, fixes, support of user-added operations
1529 - RMA: MPI_Accumulate are applied in correct order.
1530 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1531 - Better support for MPI_IN_PLACE
1532 - Support for MPI_Win attrs and keyvals.
1533 - Support MPI_Comm_group_create, MPI_Type_size_x
1534 - MPI ops should be properly applied to all allowed types
1535 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1536 algorithm by a non failing one from mpich.
1539 - Kill the fifo data container: we don't use it anymore.
1542 - A <cluster> can now be created with different speed values to represent pstates.
1543 see examples/platforms/energy_cluster.xml for an example.
1545 SimGrid (3.14.159) Released December 28. 2016
1547 The previous release was only a crude approximation of a Pi release;
1548 this one should be a bit better.
1550 - Revert a last minute change that broke on Mac OSX.
1551 - Fix the changelog and NEWS files.
1552 - Improve the documentation, in particular of the routing module.
1553 - Integrate some patches flying around in Debian.
1555 SimGrid (3.14) Released December 24. 2016
1558 * The whole documentation was reworked and reorganized. There is
1559 still much room for improvement here, but we're on it.
1562 * We now need python3 for our testing infrastructure.
1563 * Model checking needs libevent
1566 * Now works on FreeBSD
1569 * Privatization now works on FreeBSD
1571 * Privatization is now activated by default in smpirun
1572 You can turn it off with -no-privatize if something goes wrong.
1574 * Call-location tracing for SMPI.
1575 You can add the exact location (filename / linenumber) of an MPI call to
1576 your trace files and slow down or speed up the simulation between two
1577 consecutive calls by using an adjustment file (see the documentation).
1579 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1580 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1581 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1582 values were ignored in some cases. The timings of these functions can now
1583 be significantly different.
1585 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1586 smpi/running-power is renamed to smpi/host-speed
1588 * smpi/grow-injected-times option to enable or disable multiplication of the
1589 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1590 make simulation less precise (but also much faster).
1592 * smpirun script should be (much) faster for large deployments.
1594 * SMPI tracing: fixed issue with poor matching of send/receives.
1596 * Replay: Fix broken waitall
1598 New functions and features
1599 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1601 Dropped / renamed functions and features
1602 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1603 - MSG_mailbox_is_empty() -> Mailbox::empty()
1604 - MSG_mailbox_front() -> Mailbox::front()
1605 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1606 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1607 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1608 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1609 * Don't pass the free_f parameter to property related functions:
1610 - MSG_host_set_property_value()
1611 - MSG_as_router_set_property_value()
1612 - MSG_storage_set_property_value()
1613 * VM properties. Since msg_vm_t are msg_host_t, just use
1614 MSG_host_get_property_value() and friends
1615 * VM I/O related things:
1616 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1617 - Unimplemented save/restore methods
1618 * MSG_as_router_get_property_value() was redundant with
1619 MSG_environment_as_get_property_value().
1620 - Removed MSG_as_router_*propert*() functions
1621 - Added MSG_environment_as_set_property_value() for consistency
1622 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1623 Well, they are still there for now, but deprecated with a warning.
1624 Please switch to xbt_dict_new_homogeneous() before this is removed
1626 * Task affinity. Its intended behavior (that was very badly tested
1627 and probably not really working) was deceiving what most users
1628 would have hoped here.
1629 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1630 * Option network/coordinates is now useless and should be dropped.
1633 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1636 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1637 * Added Dragonfly topology support
1640 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1641 EiBps, EBps, Eibps, Ebps for bandwidth.
1642 They may become useful to some lucky ones.
1645 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1646 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1647 * ex/async_waitAll: new example, on asynchronous communications
1650 * Memory usage should be decreased for simulations with a large number
1651 of processes. This also helps for SMPI.
1653 SimGrid (3.13) stable; urgency=low
1655 The Half Release, a.k.a. the Zealous Easter Trim.
1657 - We removed half of the lines, that were mostly experimental cruft.
1658 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1659 (not counting blanks and comments -- according to openhub.net)
1660 - The internals are now compiled in C++ (and will soon be clean C++)
1661 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1662 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1664 * Backwards Compatibility breaks
1665 - Removed Lua simulation bindings (switch to C or Java for that).
1666 Lua can still be used to describe platforms
1667 - Removed Java kernel plug-ins.
1668 Will be reintroduced after the ongoing major internals reorg.
1670 - the following functions were removed.
1671 They were too specific and should be reimplemented in a generic
1672 way, with filter function.
1673 - MSG_task_listen_from_host
1674 - MSG_mailbox_get_count_host_waiting_tasks
1675 - MSG_mailbox_put_with_timeout was removed.
1676 Please use MSG_task_send_with_timeout instead.
1678 - the SD_application_reinit function was removed. It has been a noop for a while.
1679 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1680 replaced by a more flexible mechanism.
1681 - The following functions thus do not exist anymore
1682 - SD_workstation_get_access_mode
1683 - SD_workstation_set_access_mode
1684 - SD_workstation_get_current_task
1685 - Basic estimation functions have been removed but can easily be replaced
1686 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1687 - SD_workstation_get_computation_time => amount / sg_host_speed()
1689 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1690 Use VM.getSpeed()*load/100 for the legacy behavior.
1692 - option enable_tracing was removed. It was not doing anything for a while.
1693 - In the ModelChecker:
1694 - the model-checker now ptraces the model-checked process which means
1695 you cannot use a debugger on the latter anymore (we might make this
1696 optional in the feature);
1697 - removed soft-dirty page tracking;
1698 - remove model-checked side snapshot management,
1699 MC_snapshot() and MC_compare_snapshot();
1700 - keep the MC_cut() function as a stub (it was not really working
1701 in the previous release).
1704 * All options are consistently in kebab-case. Old names are kept as alias.
1707 * Switch to platform v4 format.
1708 - Rename from 'power' to 'speed' the attributes describing the amount of
1709 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1710 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1711 - In <host> and <link>, attributes availability and state are gone.
1712 It was redundant with state and availability traces, and with peak values.
1713 - In <cluster>, attributes availability_file and state_file are gone.
1714 It was too complex and unused.
1715 - Kill <gpu>. Was not doing anything.
1716 - The DOCTYPE points to the right URL:
1717 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1718 (the file at this address now documents the changelog since its v1)
1719 - A warning is emitted for unit-less values (they are still accepted).
1720 - speed. Default: 'f' or 'flops'. Also defined:
1721 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1722 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1723 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1724 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1725 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1726 - latency. Default: 's' second. Also defined:
1727 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1729 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1730 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1733 * s4u::Host is now the preferred public interface to the Host features.
1734 sg_host_* functions are C bindings to the exact same behavior
1735 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1738 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1739 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1740 They were deprecated since a few years, and probably did not even compile anymore.
1743 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1744 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1745 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1746 encouraged to update their codes. A list of the modified functions can be found at the end of
1747 include/simgrid/simdag.h
1750 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1751 the lambda or closure passed as a parameter will run in kernel mode.
1752 All the callback functions should be rewritten to that interface at some point.
1755 * Reorganizing and cleaning the internals all around the place.
1758 * Remove old default barrier/bcast buggy algorithms (see #18407)
1759 * Various bug fixes to handle more codes
1760 * Remove the need for the --foreground option of smpirun (it is still
1761 accepted for backward compatibility).
1764 * Kill the setset data container: MC don't use it anymore.
1765 * Kill the queue data container: it made more sense with GRAS.
1766 * Kill the xbt_peer_t data type: it's useless without GRAS.
1767 * Kill rm_cb feature of config sets: it was never useful.
1768 * Kill graphxml parsing feature. It was not used.
1769 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1771 - xbt_dynar_sort_strings(), when the content is char*
1772 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1773 They throw exceptions on invalid input;
1774 * C++ support for declaring CLI flags (simgrid::config::Flag);
1775 * class for abstracting different signal backends (simgrid::xbt::signal).
1776 with no external dependencies (we need very simple signals).
1779 * refactoring and cleanup of the code;
1780 * ongoing process to cleanly separate the model-checking algorithms
1781 from the code model-checking support.
1783 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1785 SimGrid (3.12) stable; urgency=low
1787 The Facelift Release.
1790 * Require g++ v4.7 at least to not speak prehistorical C++.
1791 * Require Boost 1.48 (for signal2 component).
1792 * Java must be version 7 at least when activated.
1793 * Builds on Windows again (including Java bindings).
1794 * Tracing is now always enabled (no way to turn it out)
1795 * Remove GTNetS. It was not working anyway.
1796 * Various cleanups in the cmake scripts.
1797 * Move headers around to sort them out on installed systems:
1798 - instr/instr.h -> simgrid/instr.h
1799 - instr/jedule/* -> simgrid/jedule
1800 - simdag/datatypes.h was removed
1801 - simdag/simdag.h -> simgrid/simdag.h
1802 - msg/datatypes.h was removed
1803 - msg/msg.h -> simgrid/msg.h
1806 * Interface improvement:
1807 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1808 - Sanitize the interface in MSG_task_ module:
1809 - Merge two functions that were close enough but misleading:
1810 set_compute_duration(t) -> set_flops_amount(t)
1811 get_remaining_computation(t) -> get_flops_amount(t)
1812 - set_data_size(t) -> set_bytes_amount(t)
1813 get_data_size(t) -> get_bytes_amount(t)
1814 - Massive cleanups in the functions related to the energy
1815 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1816 - New: MSG_host_get_pstate()
1818 - msg/energy/onoff: switching hosts on and off
1820 * Interface improvement:
1821 - Rename Host.isAvail() to Host.isOn()
1822 - Rename Process.currentProcess() to Process.getCurrentProcess()
1823 - Rename Task.setDataSize() to Task.setBytesAmount()
1824 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1826 - #18874: Actually allows the GC to reclaim tasks
1830 - SIMIX_process_throw: raises an exception in a remote process
1831 * Refactoring: Separate sub-modules
1832 - libsmx: the public interface, as libc in a real system
1833 - popping: the strange dance that converts a user request into a kernel handling
1834 - smx_context_*: the virtualization mechanisms that embed the user code
1835 - smx_*: the handling of each simcalls
1836 * Interface cleanup:
1837 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1838 * Rename smx_action_t into smx_synchro_t, making explicit that these
1839 things are used to synchronize processes with their environment.
1840 For example, a communication is a sort of synchronization involving
1841 the communicating processes (that may block until the exchange) and
1842 the platform. The same can be said from computations, etc.
1844 - #18888: segfault when a process ends before its kill_time
1848 - 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.
1849 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1850 - MPI_Type_set_name, MPI_Type_get_name
1851 - MPI_*_c2f and MPI_*_f2c functions
1852 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1853 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1854 - Activate a lot of new tests from the mpich 3 testsuite
1856 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1857 - InfiniBand network model added: Based on the works of Jerome Vienne
1858 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1859 - When smpi/display_timing is set, also display global simulation time and application times
1860 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1861 * Collective communications
1862 - 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
1863 - 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/).
1864 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1866 - Replay now uses algorithms from wanted collective selector
1867 - Replay can be used with SMP-aware algorithms
1868 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1869 - 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.
1871 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1872 - Correctly use loopback on fat-tree clusters
1873 - Asynchronous small messages shouldn't trigger deadlocks anymore
1874 * Energy/DVFS cleanup and improvement
1875 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1876 - new: smpi_get_host_pstate()
1879 - "Full" network optimization flag was broken since Surf++
1880 - Better handling of precision flags in maxmin
1881 - Fix bug causing sometimes "Impossible" errors
1882 - Properly pass cluster properties to included hosts
1883 * Improvement of the Energy plugin.
1884 - Always update the consumption before returning that value
1885 - New property: watt_off to denote the dissipation when the host is off
1886 - New functions getWattMinAt and getWattMaxAt to retrieve the
1887 dissipation of pstates that we are not currently at.
1888 * Java: class NetworkLink renamed to Link
1889 * New function: simcall_process_get_kill_time()
1890 * Massive rename s/workstation/host/
1891 - That's intrusive, but that's good for the project consistency. Sorry.
1892 - Change config option "workstation/model" into "host/model"
1895 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1896 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1897 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1898 - xbt_dirname and xbt_basename for non-POSIX systems
1900 * The model checker now runs as a separate process.
1901 * The model checker runs must now be launched with the new simgrid-mc program.
1902 * Record/Replay: the MC can display a textual representation of a path in the
1903 execution graph. It can then be replayed outside of the model checker.
1905 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1907 SimGrid (3.11) stable; urgency=low
1912 * Normalizing pointers addresses tool for better diff between logs
1915 * Add cloud examples using new VMs
1916 - examples/msg/cloud/two_tasks_vm.tesh
1917 - examples/msg/cloud/simple_vm.tesh
1918 - examples/java/cloud/cloud.tesh
1919 - examples/java/cloud/migration/migration.tesh
1920 * Add java surf examples:
1921 - examples/java/surfPlugin/surf_plugin.tesh
1922 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1923 - examples/java/surfCpuModel/surf_cpu_model.tesh
1924 * Add SMPI+MSG example:
1925 - examples/smpi/smpi_msg_masterslave/
1930 - msg task destroy cancel test
1931 - msg_host on/off test
1932 * Move all tests in testsuite to teshsuite (adding tesh files)
1933 * Restructure teshsuites
1934 - one folder for each kind of test
1935 * Restructure AddTests.cmake
1937 - structure the order of tests (with sections)
1940 * Add virtual machine
1941 - creation of a VM on a PM
1942 - migration of a VM from a PM to another PM
1944 - MSG_process_join(msg_process_t process, double timeout)
1945 - msg_bar_t MSG_barrier_init(unsigned int count)
1946 - int MSG_barrier_wait(msg_bar_t barrier)
1947 - void MSG_barrier_destroy(msg_bar_t barrier)
1948 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1949 * New option "msg/debug_multiple_use" to help debugging when a task is used
1952 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1954 - Increase code coverage in test suites
1958 * Protect context stacks against stack overflow. The number of protected memory
1959 pages allocated on the top of each stack (1 by default) can be configured
1960 with the new command line option --cfg=contexts/guard_size:<value>.
1961 * Simcalls are now generated by a python script that
1962 - generates files included by SimGrid
1963 - checks that all the functions exist, and proposes prototypes
1965 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1966 host_set_data, host_get_data
1968 - simcall_process_join(smx_process_t process, double timeout)
1969 * Fix bug where sleeping processing could not be suspended.
1972 * Translate surf models from C to C++
1973 - Generic classes for all models: Model, Resource, Action
1974 - A generic interface for each kind of model (CPU, Network, Storage
1975 Workstation, WorkstationVM)
1977 * Translate surf routings from C to C++
1978 * Add callbacks using sigc++ or boost::signals2
1979 - Add callback functions for resource creation/destruction
1980 - Add callback functions for action state change
1981 - Handle Energy as a plugin
1982 * Replace swag by boost::intrusive
1983 * Add new routing models for clusters. For documentation, see
1984 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1985 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1986 parameters for cluster tag
1987 - Fat trees, with topology="FAT_TREE" and
1988 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1989 - see examples/platforms/torus_cluster.xml and
1990 examples/platforms/fat_tree_cluster.xml
1991 * More documentation
1994 * Hostfiles support host:nb_processes construct to deploy several processes on
1996 * Collective communication algorithms should not crash if used with
1997 improper number of nodes and report the error.
1998 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1999 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
2000 MPI_Dims_create, MPI_Cart_sub are supported.
2001 * New interface to use SMPI programmatically (still depends on MSG for
2002 some parts, see examples/smpi/smpi_msg_masterslave):
2003 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
2007 * Global variables privatization in MPI executables is now performed at runtime
2008 with the option smpi/privatize_global_variables (default:no).
2009 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
2010 dynamic libraries loaded by the application are not privatized (static
2011 linking with these libraries is advised in this case)
2014 * Options defined in XML work correctly now.
2017 * New cmake option, enable_lib_in_jar, to control whether native libraries are
2018 copied into simgrid.jar or not (ON by default). Use this option if you want
2019 to reduce the size of the installed simgrid.jar, *and* the native libraries
2020 are kept installed elsewhere.
2021 * Surf binding with SWIG (code generated in maintainer mode only):
2022 - plugin to handle callbacks
2023 - CPU model only for the moment
2026 * Supernovae build mode is definitively removed. It was used to improve
2027 inlining and inter-module optimizations. It is nowadays superseded by
2028 link-time optimizations commonly available in compilers.
2029 * Update ns-3 find lib. Bindings for ns-3 should work again now.
2030 * Add boost dependency for surf++
2031 * Add new macro for tests
2032 - ADD_TESH(name <tesh_args>)
2033 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
2037 - xbt_bar_t XBT_barrier_init(unsigned int count)
2038 - int XBT_barrier_wait(xbt_bar_t barrier)
2039 - void XBT_barrier_destroy(xbt_bar_t barrier)
2040 * Make the xbt_os_time module public
2042 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2044 SimGrid (3.10) stable; urgency=low
2046 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
2049 * Reintegrate Java to the main archive as desynchronizing these
2050 package is not acceptable anymore (Java is now considered stable)
2051 * Add explicit synchronization facilities through semaphores
2052 * Bug fix: Task.setDataSize() only changed the C world, not the value
2053 cached in the Java world
2056 * Dramatically change the way files are handled. API and internals changed, but
2057 this part of MSG was not considered as production grade either.
2058 * Add explicit synchronization facilities through semaphores
2059 * Add a new function MSG_host_get_process_list()
2060 * Preliminary DVFS support (see examples/msg/energy/ for details)
2063 * SMPI is now included directly in the libsimgrid as the windows
2064 linker doesn't force us on splitting it anymore.
2065 * Improvements of the SMPI replay tool:
2066 - Most of the collective communications are now rooted in the same process as
2067 in the original application.
2068 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
2069 used until now). Multiple data types can now be used in a trace.
2070 - The replay tool now supports traces produce either by TAU or a modified
2072 - Bug Fix: the compute part of the reduce action is now taken into account.
2073 - Gatherv collective is now supported
2074 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
2075 of smpirun outputs time independent traces for the current run. One file
2076 is created per process. If too many processes are simulated, this behavior
2077 can be changed to one file for all processes by using the
2078 tracing/smpi/format/ti_one_file flag
2079 * smpirun generates the host file if needed (with given host count and platform)
2080 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
2081 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
2082 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
2083 --cfg=smpi/coll_selector:(mpich/ompi)
2084 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
2086 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
2087 happening inside a collective SMPI call.
2088 * Fix the behavior of complex data types handling.
2089 * Make MPI_Wtime another synchronization point to take computations into
2091 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
2092 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
2093 * Add all missing Fortran bindings, SMPI should work with Fortran 90
2094 (no privatization of global variables yet)
2095 * Preliminary DVFS support (see examples/smpi/energy/ for details)
2098 * Verification of liveness properties is now available for SMPI applications
2099 (in addition to MSG applications)
2100 * Bugged examples using SMPI in examples/smpi/mc/
2101 * Add --cfg=model-check/visited option. Allows the verification of infinite
2102 programs. Detection of loops in the execution thanks to the system state
2103 comparison and reduction of the state space to explore. Can be combined with
2104 DPOR for safety properties.
2107 * Allow to change SimGrid configuration (see --help) within the code
2108 thanks to SD_config() as it can be done in MSG.
2109 * Add a new function SD_task_set_amount() upon user request.
2112 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
2113 * Remove rule based routing (no more PCRE dependency)
2114 * Add a limiter_link option to cluster tag, to specify a maximum reachable
2115 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
2116 * Add a loopback_bw and loopback_lat options to cluster tag.
2117 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
2118 routing. Now peers are encapsulated in an AS and have their own private
2119 router but this is transparent.
2122 * Our own implementation of getline is renamed xbt_getline, and gets
2123 used even if the OS provide a getline(). This should reduce the
2124 configuration complexity by using the same code on all platforms.
2125 * New type: xbt_cfg_elm_boolean.
2126 * Allow to use yes/no for boolean configuration options in the command line.
2127 * Allow to disable SimGrid cleanups at exit from command line option.
2128 There are situations where one may want a simulation to end with an exit.
2129 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
2130 annoying when scripting around the simulator. Adding a
2131 --cfg=clean_atexit:no allows to circumvent this issue.
2134 * Lots of memory leaks were corrected in this release.
2135 There is no known memory leaks anymore, in all of our 600+ tests.
2136 * New command line option --version, to get SimGrid version information.
2137 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
2139 * Supernovae builds are deprecated, and expected to be removed in the next
2142 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2144 SimGrid (3.9) stable; urgency=low
2146 The Grasgory release: GRAS is really dead now.
2148 * Complete overhaul of the internal host structures scheme.
2151 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
2152 was considered as experimental and badly maintained since then.
2153 * Keeping it was thus a trap to our potential users, that could take
2154 it instead of MSG or SMPI by mistake despite is pity state.
2155 * GRAS seems to have very few users (if any), and no one volunteered
2156 to maintain it further. It also induces a lot of XBT code (for
2157 portability sake), that must be maintained too.
2158 * For all these reasons, we killed GRAS. If someone wants to revive it
2159 in the future, don't cry, our git history still remembers of GRAS.
2162 * Major overhaul. Merge our documentation again as time proved that
2163 splitting it was really not helping our users.
2164 * Further improve the developer documentation to help newcomers
2165 hacking on SimGrid itself. The user documentation (and in
2166 particular, the beginner documentation) is still in a sorry state.
2169 * Now works on Windows too!
2170 * Much more extensive test suite, from MPICH
2173 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
2174 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
2175 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
2176 * Bug fix: let task be scheduled when the last dependency to be solved is
2177 a control dependency.
2178 * Remove SD_load_environment_script function.
2179 Use the C sg_platf function if you want to declare a platform
2183 * New function: MSG_process_get_number()
2184 * Old function documented: MSG_config()
2185 * Remove MSG_load_platform_script function
2186 Use the C sg_platf function if you want to declare a platform
2190 * Change the default value of the TCP_gamma constant (maximal size of TCP
2191 congestion window) to a more realistic 4MiB value. If you notice changes in
2192 your simulation results, you can fall back to the previous 20k tiny window
2193 by adding --cfg=network/TCP_gamma:20000 on command line.
2194 * (Hopefully) fix a bug wrt periodic availability and state traces
2195 * Bug fix: use default values at start when first event in availability/state
2196 trace is not at time 0.
2199 * remove the "new_" part of function name sg_platf_new_trace_connect
2200 (resulting in sg_platf_trace_connect), since it does not create
2204 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2205 use case where it's really mandatory, and maintaining it was a pain in
2207 * New: xbt_fifo_search(), search an item with a user-provided
2208 comparison function instead of dumb pointer comparison.
2211 * Fix the lua deployment:
2212 Use `simgrid.init_application()` before deployment instead of
2213 `simgrid.msg_register_application()` after.
2216 * Transfer the tracing files into the corresponding modules.
2218 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2220 SimGrid (3.8.1) stable; urgency=low
2222 The "we are told that some people want to also *install* the simgrid
2225 * Add missing file "tesh.1" to the archive.
2227 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2229 SimGrid (3.8) stable; urgency=low
2231 The Psssshiiiit release: SimGrid jumps into the Cloud.
2234 * Add an experimental interface to manipulate VMs. They are mainly
2235 process groups with very few intrinsic semantic, but they should
2236 allow you to build the semantic you want easily.
2237 * New function: MSG_host_set_property_value()
2238 * New function: MSG_process_on_exit(). To clean memory in all cases.
2239 * Bug fixes that made the host (and link) failures unusable.
2240 * Add a way to auto-restart process when the host in which they are
2241 executing comes back (ON_FAILURE="RESTART" on deployment file,
2242 MSG_process_auto_restart_set).
2243 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2244 please stop using the old ones, they are DEPRECATED.
2246 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2247 Please use MSG_init() instead. (reducing the amount of entry
2248 points in the library helps us).
2249 * Make it impossible to link against the wrong version of the lib
2250 * Deprecate MSG_clean(). No need to call it anymore.
2251 * Function MSG_get_host_number() is not deprecated anymore.
2254 * Split the doc into a user guide and a reference guide.
2255 * Start a developper guide to help people hacking on SimGrid.
2258 * Enable tracing by default. This modules rocks you should use it.
2259 * Remove option custom_flags. Now use environment variables CFLAGS
2261 * Use default cmake things to detect lua instead of home grown ones.
2262 * New option "enable_mallocators" to disable mallocators, for debugging
2263 purpose ("on" by default).
2266 * Bug fixes around the resource failures: don't let the processes
2267 survive the host they are running onto.
2268 * Add an interface to auto-restart processes when the host in which they are
2269 executing comes back.
2270 * Ensures that SIMIX_clean is called automatically. It's not part of
2271 the public interface anymore (bindings should be updated).
2274 * Bug fix for when SD_Simulate is called with a positive value: be careful
2275 when comparing doubles. Sometimes they are different for non significant
2277 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2278 parallel task whose initial work is distributed among host according
2279 to the Amdahl's law. Such tasks are created with a parameter alpha
2280 that corresponds to the non-parallelizable part of the computation.
2281 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2282 between two sets of workstations assuming a 1D block distribution (each
2283 workstation owns a similar share of data) on both sides.
2285 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2286 Data redistribution will be automatically scheduled once parent and child
2287 are both scheduled. The filling of computation_amount and
2288 communication_amount structures is now done seamlessly thanks to the chosen
2290 * New function SD_workstation_dump to display various information
2291 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2292 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2293 bandwidth on the route onto which the task is scheduled and the amount of
2295 To divide the nominal bandwidth by 2, the rate then has to be:
2296 rate = bandwidth/(2*amount)
2297 * Compute tasks that have failed can now be rescheduled and executed again
2298 (from their beginning)
2299 * Increasing source code coverage (src/simdag is now covered at 95.8%
2303 * Re-implement time-independent trace replay using SMPI (at the
2304 smpi_smp_* level) instead of MSG. This should replace
2305 examples/msg/actions/actions.c
2306 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2307 hindexed and structs)
2308 * Implement the exchange of non-contiguous data.
2309 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2310 * Correct behavior of smpi/sender_gap and set its default value to 0
2311 * Add option to asynchronously send small messages to allow better
2312 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2313 specifies the size in bytes under which messages will be asynchronously sent.
2314 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2315 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2318 * New C interface to define a platform: XML is now optional.
2319 For more info, please check include/simgrid/platf.h.
2320 * New interface to define random platforms from the C:
2321 For more info, please check include/simgrid/platf_generator.h and
2322 examples/msg/masterslave/masterslave_platfgen.c
2323 * Export a sg_cmdline dynar containing all the arguments we got from
2327 * Two new tracing options for adding comments to trace file so you
2328 can track your experiments (see --help-tracing for details).
2329 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2330 * Adding the SimGrid version that generated the trace file as a comment.
2331 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2332 * Fix to avoid key clashes on Paje links
2333 * Other minor fixes related to the Paje specification
2336 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2337 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2338 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2339 * Greatly improve the robustness of mmalloc to user errors (such as
2340 using an area after freeing it, or freeing it twice)
2342 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2344 SimGrid-java (3.8.1) stable; urgency=low
2346 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2347 Warning: all methods are visible, but only some of them are
2348 implemented so far. Check the source (src/jtrace.c)
2349 for further information.
2350 * New module: org.simgrid.msg.File (SimGrid File management functions)
2351 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2352 * Change the meaning of Process.restart: now restart the process from
2353 the beginning, like MSG_process_restart in C.
2354 * Add Process.setAutoRestart: handling of process restart when failed
2356 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2357 you to retrieve the properties of the processes/hosts
2358 * Deprecate Msg.clean(): you can just forget about it now.
2359 * New function Process.getCount(), that only works when compiling
2360 with the not yet released version 3.9 of the C library.
2362 * New context factory based on Coroutines. It mandates a modified JVM
2363 but then, the simulations run about five times faster, and there is
2364 no limit to the amount of processes (beside of the available memory).
2366 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2368 SimGrid (3.7.1) stable; urgency=low
2371 * Restore the prototype of MSG_process_create_with_environment() to
2372 the pre-3.7 situation by removing the kill_time argument.
2373 * Add a MSG_process_set_kill_time() function instead.
2376 * Fix weird behaviors when dealing with parallel tasks.
2379 * Simgrid is now built as a dll.
2380 * Simgrid-java now works on Windows.
2381 * Simgrid-Java is now included into Windows package.
2384 * First pre-build package for MacOSX.
2387 * Fix compilation when using MSG_USE_DEPRECATED.
2388 * Fix some compilation issues on Macs and Windows.
2389 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2390 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2392 We discovered that the Lua console is broken, but we are missing the
2393 manpower to fix it right now. The problem existed in 3.7 too, so we
2394 are not blocking the release for that. Sorry if you depended on this
2395 feature, any help would be really welcome.
2397 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2400 SimGrid-java (3.7.1) stable; urgency=low
2402 The "Java aint got to be bloated and slow" release
2405 * Various internal cleanups and performance improvement
2406 Simulations are expected to run up to twice faster or so
2407 * Make Process.kill(process) an instance method, not a static one
2408 * User processes are not java.lang.Thread subclasses.
2409 This breaks the compatibility (sorry), but previous API was
2410 brain-dead, making it impossible to have non-trivial
2411 initializations in the process constructor.
2412 * Require a full constructor per Process sub-class.
2413 Kinda breaks the compatibility (sorry), but this allows a much more
2414 efficient way to launch the processes at simulation startup.
2415 * Do not embeed our version of semaphores, java 1.5 can be considered
2416 as sufficiently prevalent for us to not dupplicate its features.
2420 * Add examples for almost every part of the API
2421 We spotted and fixed a lot of bugs in the process
2423 * New module: asynchronous communication API
2424 * New function: Process.sleep()
2425 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2426 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2427 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2429 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2431 SimGrid (3.7) stable; urgency=low
2433 The "spring cleanups (before the next Big Project kicks in)" release.
2436 * We can specify the SMPI latency/bandwidth factor with command line
2437 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2438 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2439 You can also use the "config tag" from platform file by adding this line
2440 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2441 (see "example/platforms/tag_config.xml" to use "config tag").
2442 Note that the command line supersedes the platform file configuration.
2443 * Change the correction factors used in LMM model, according to
2444 the latest experiments described in INRIA RR-7821.
2445 Accuracy should be improved this way.
2446 * Use the partial invalidation optimization by default for the
2447 network too. Should produce the exact same results, only faster.
2448 * Major cleanup in surf to merge models and split some optimization
2449 mechanisms from the core of the models. As a result you can now
2450 specify which model to use (e.g., --cfg=network/model:LV08
2451 --cfg=cpu/model:Cas01) and which optimization mode to use
2452 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2453 Incompatible combinations should err at initialization. See
2454 --help-models for the list of all models and optimization modes.
2455 * The CLM03 workstation model was dropped for simplicity because it
2456 used the deprecated CM02 network model. Use default instead.
2457 * Rename the TCP_gamma configuration option to network/TCP_gamma
2458 * Rename the coordinates configuration option to
2459 network/coordinates, and document it
2460 * Use now crosstraffic keyword instead of the terribly misleading
2461 fullduplex keyword. It is activated by default now in the current
2462 default model, use --cfg=network/crosstraffic:0 to turn it off.
2463 * Ongoing refactoring the model parsing to make XML files optional
2464 See include/simgrid/platf.h for details (still to be completed)
2467 * Major overhaul of the documentation. Almost instructive now :/
2468 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2469 functions and friends. This interface was considered as
2470 deprecated since over 2 years, it's time to inform our users that it is.
2471 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2472 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2473 use these (crappy) functions in your code.
2474 These functions will be removed soon. Stop using them now.
2475 * Deprecate MSG_get_host_{table,number}
2476 Implement MSG_hosts_as_dynar() instead.
2477 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2478 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2481 * Stabilize the parallel execution mode of user contexts
2482 * Introduce configuration variables to control parallel execution:
2483 - contexts/synchro: Synchronization mode to use when running
2484 contexts in parallel (either futex, posix or busy_wait)
2485 - contexts/parallel_threshold: Minimal number of user contexts
2486 that must be part of a scheduling round to switch to parallel
2487 execution mode (raw contexts only)
2488 * Fix bugs that prevented to use suspend/resume along with
2489 synchronization structures.
2490 * Fix bugs in process termination that lead to invalid memory access
2491 in very specific conditions.
2494 * Introduce a parallel mode for the models (controlled by surf/nthreads
2495 configuration item). In our tests, running the models in parallel
2496 never lead to any speedups because they are so fast that the gain
2497 of computing each model in parallel does not amortizes the
2498 synchronization costs, even when ultra fast futexes are used.
2499 This is released anyway because YMMV.
2502 * Performance boost by using a swag internally to compute the set of
2503 tasks that are finished and should constitute the return value of
2507 * Enable it by default now that it is considered rather stable.
2510 * Documentation of the tracing functions.
2511 * Performance gains when tracing categorized/uncategorized resource
2512 utilization by avoiding calls to get route when updating resource
2513 variables. LMM constraints are being used instead.
2514 * API changed to set task categories. Use MSG_task_set_category instead
2515 of TRACE_msg_set_task_category, and SD_task_set_category instead
2516 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2517 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2518 * Examples for link user variables added (see at examples/msg/tracing/)
2519 * Deprecated function TRACE_msg_set_process_category completely removed
2520 * Trace header updated according to the latest Paje file format
2521 * Tracing network lazy updates, no longer obligate users to use full updates
2522 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2523 * Experimental: let user code declare/set/push/pop application states for hosts
2524 * API changed to allow the manual creation of graph configuration files
2525 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2528 * Improve the API of Lua MSG bindings, using the Lua spirit.
2529 * Each simulated process now lives in its own Lua world (globals are
2530 automatically duplicated). It helps writing simulators. Will allow
2531 to run Splay programs within SimGrid in the future.
2532 * Add a Chord example in Lua, equivalent to the MSG one.
2535 * Start the implementation of a solution to express temporal
2536 properties, not only local assertions. This is still an
2537 experimental work in progress, stay clear from it to be safe.
2541 - Add new runtime parameters --help-logs and --help-log-categories
2542 to display information about supported logging parameters and
2544 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2545 don't exists anymore.
2546 * Mallocators: allow value NULL for the reset function.
2548 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2549 create homogeneous dictionaries, where all the elements share the
2550 same free function. Non homogeneous dictionaries will be
2551 deprecated in the next release.
2552 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2553 - Multi-level dictionaries are deprecated.
2555 - new function xbt_dynar_search_or_negative() that is useful when
2556 you have less than 2 million elements in your dynar and don't
2557 want of the extra complexity of catching exceptions when the
2558 element is not found.
2560 - Make xbt_os_thread module (for thread portability) public.
2561 Documentation is still to come, sorry.
2563 - Cleanups and simplifications to make it maintainable again.
2564 - Exotic features (such as memalign and valloc) were removed.
2565 - The metadata were extended and improved so that the
2566 model-checker becomes able to explore and inspect the heaps.
2567 - This may induce a performance drop when enable_model-checking is
2568 ON in cmake (even if it's not used in the simulation), but it is
2569 necessary at this point to get MC working.
2571 Turn model-checking OFF if simulation performance matters to you.
2572 Not enabling it at runtime is not enough, disable it in cmake.
2574 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2577 SimGrid (3.6.2) stable; urgency=low
2579 The "Not coding new stuff allows to polish old things" release.
2582 * New bindings to the NS3 packet level simulator (experimental)
2583 * Use the raw (efficient) execution contexts instead of the sysv
2584 (portable) ones when possible.
2585 * libpcre is now mandatory in any cases since not using it led to
2586 severe performance loss and possibly other issues
2587 * Update the XML platforms:
2588 - G5K: include the latest machine in Nancy
2589 - GridPP and LCG: new platforms
2590 * Documentation was partially updated, at least (more to come)
2592 Bug fixes, cosmetics and small improvements
2593 * Free terminated processes before the end of the simulation to avoid
2594 exhausting the memory of users having very dynamic amount of
2596 * Bug fix and cosmetics about canceling non-running tasks
2597 * Bug fix about the dot loader's issues when using libcgraph
2600 * Create an installer for windows with nsis (amd64 and win32)
2601 - Add an hello world project to illustrate simgrid project creation.
2602 - Embed libpcre into the Simgrid installer to avoid
2603 its compilation burden
2604 * The raw execution contexts should work on Apple now
2605 * Port to Windows 64 bits
2606 - Sysv contexts now have an implementation for this arch
2607 - GRAS communication features now support this arch
2608 * Drop support for borland compiler on windows
2609 - this code was not maintained, and we kinda depend on gcc nowadays
2610 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2611 * Fix portability issue on unstable ubuntu: linker became picky on
2614 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2616 SimGrid (3.6.1) stable; urgency=low
2618 The "Oops, we broke Macs too" release
2621 * Fixed contexts detection so that raw ones are used when possible
2622 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2623 leads to a strange error, with user code segfaulting sometimes when
2624 the generated code is not perfectly aligned (which is not
2625 controllable from the user side, depends on the amount of code)
2628 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2630 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2632 SimGrid-java (3.6) unstable; urgency=low
2635 * Split of every thing from simgrid v3.5 into a separate package.
2637 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2639 SimGrid (3.6) stable; urgency=medium
2641 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2644 * Bindings now constitute their own package, separated from the main one.
2645 Rationale: reduce our maintenance nightmare by reducing the module coupling
2646 They will soon be released on their own on gforge.
2648 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2649 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2651 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2652 * It's not quite deprecated for now because we have no replacement,
2653 but it may soon become the case.
2656 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2657 * Fortran: New user-level cache variable to store the rank of the running
2658 process. This improves performance by an order of magnitude.
2659 * C: New coccinelle script to automatically locate and modify global and
2660 local static variables.
2661 * Improved SMPI network model with a sender-side gap to account for multiple
2665 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2666 only say if a communication is finished, no matter whether it succeeded or
2667 failed. You can call MSG_comm_get_status() to know the status of a finished
2669 * New function MSG_task_dsend() to send a task and detach it. When a
2670 communication is detached, you are never notified of its success or failure
2671 and the memory is released automatically once it is finished. This function
2672 is useful when you don't care about the end nor the success of a
2674 * Change the prototypes of action replay. Sorry for inconvenience,
2675 but this is really more efficient this way (and to adapt your code,
2676 you just have to fix the initialization, that shouldn't be too long)
2677 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2678 ever managed to do anything useful with it.
2679 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2680 takes a dynar of communications. It returns immediately and gives the
2681 index of a finished communication (if any).
2682 * New example: a basic implementation of the Chord P2P algorithm.
2685 * New model for multi-core CPUs. You can now use the core attribute to
2686 precise the number of cores of a host. This is a basic model. Every
2687 process running on the host receives at most the power provided in
2688 the DTD (throughput<=power). Total throughput of process cannot exceed
2690 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2691 router linked by an up-link and a down-link (possibly asymmetrical).
2692 This kind of pattern allows to easily build last-mile model style platforms.
2693 Aggregating such patterns in a rule-based AS is thus the technique of
2694 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2695 * New model for Vivaldi routing. We transformed the Vivaldi network model
2696 into a Vivaldi routing model (based on the rule-based model). This allows to
2697 combine Vivaldi based latencies with last-mile platforms.
2700 * Added a check for NaN of IEEE754 infinite in the double entries of
2702 * Introduce a new context factory "raw", highly inspirated from the
2703 ucontext factory, but using manually crafted functions in assembly to
2704 do the work in an efficient manner.
2705 * Allow to change the used context factory at run time, not only at
2706 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2707 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2708 contexts at runtime (only with raw contexts or ucontexts).
2709 * Completely rewrote this module to allow parallel execution of user
2710 processes. Use --cfg=contexts/nthreads:N to execute user processes
2711 with N parallel threads (the default is 1, meaning no parallelism).
2712 * Allow to decide dynamically between sequential and parallel modes.
2713 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2714 processes in parallel only when their number is greater than or equal to P
2716 * Added a check for NaN of IEEE754 infinite in the double entries of
2720 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2721 won't output the state of processes when interrupted with Ctrl-C
2722 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2723 NULL-terminated array. This may solve backward compatibility issues
2724 due to the change to return type of SD_simulate. See also:
2725 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2726 * Add new macros with variable number of arguments.
2727 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2728 - in xbt/asserts.h: xbt_assert
2729 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2730 - in xbt/ex.h: THROWF and RETHROWF.
2731 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2733 * Change xbt_die() to accept a format string with arguments, just like printf.
2734 * New data structure: xbt_lib_t, like a dict but more general and with better
2738 * New configuration options
2739 Options triva/categorized and triva/uncategorized can be used to generate
2740 graph configuration files for Triva visualization tool.
2741 * Configuration option tracing/platform is renamed to tracing/categorized
2742 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2743 * New TRACE_user_link_variable interface:
2744 User provides the name of the link and the tracing variable to attach to it
2745 * the declaration of tracing categories must be done after the environment
2747 * simpler tracing interface, just one way to declare categories
2748 TRACE_category or TRACE_category_with_color, it is up to you
2749 * links in the trace file are again identified by their names
2750 * trace contains the full platform hierarchy exactly as declared using the ASes
2751 * Options tracing/msg/[task|process]:1 groups the process by hosts
2752 for both cases, tasks and processes must have names that are unique during the simulation
2753 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2754 * The experimental option tracing/msg/volume is deprecated
2755 its functionality may be reincorporated if needed
2757 The tracing generates a trace file with unordered timestamped events,
2758 because of the way the core simulator (surf) works. A script available
2759 at the tools directory (fix-paje-trace.sh) can be used to put the events
2760 in order. We have changed the tracing so it can generate ordered timestamped
2761 events in the final trace, but depending on the simulator (and how much time
2762 is simulated) that can lead to a huge memory utilization. It is deactivated
2763 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2765 Build Infrastructure
2766 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2767 - We are trying hard to keep the API stable, but it may happen that
2768 some things change (we're a research project after all, not a
2769 nuclear plant operating system). If such things should happen, you
2770 could rely on that macro to adapt.
2771 - current value: 30600 for 3.06.00, aka 3.6
2772 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2773 a number that can be compared with SIMGRID_VERSION.
2774 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2775 assertions and verbose/debug logging events are disabled at compile time.
2777 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2779 SimGrid (3.5) stable; urgency=medium
2782 * New feature to any SimGrid-based simulator: Model-Checking
2783 Check SIN#1 for more details.
2786 * New Model SMPI (three-interval linear regression for correction factors)
2787 See RR-7426, available at http://hal.inria.fr/inria-00527150
2788 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2789 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2790 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2791 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2792 * See also src/smpi/README
2797 - Tracing API changes: TRACE_start and TRACE_end should not be called
2798 by user-code. They are automatically called by simulators created
2799 with SimDAG, MSG and SMPI if the toolkit is compiled with
2800 tracing_enabled=ON. Categories declaration and utilization remain the
2801 same for MSG and SimDag.
2802 - A function was added to the tracing API to declare categories with
2804 - TRACE_category_with_color (char *category, char *color)
2805 where color must be in the following format
2806 "%f %f %f", red, green, blue
2807 and red, green, blue are float values in the interval [0, 1]
2808 - User can specify NULL as color parameter, or continue calling
2809 TRACE_category (cat)
2810 On that case, the tracing system will define random colors
2811 - The following command-line options are supported:
2812 --cfg=tracing/filename:msg.trace
2813 --cfg=tracing:1 (activate tracing, needed to use others)
2814 --cfg=tracing/platform:1 (categorized resource use)
2815 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2816 --cfg=tracing/msg/task:1 (task creation)
2817 --cfg=tracing/msg/process:1 (process creation, migration)
2818 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2819 --cfg=tracing/smpi:1 (SMPI interface tracing)
2820 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2821 - examples of examples/msg/tracing updated
2823 - DAXLoader and DOTLoader functions can generate tasks with categories
2824 - A new function to attribute a category to SD tasks:
2825 TRACE_sd_set_task_category (SD_task_t task, char *category)
2826 * Tracing the MPI interface implemented by SMPI
2827 - Collective operations are traced with states
2828 - Point-to-Point operations are traced with states/links
2829 - Tracing activated by a parameter "-trace filename" passed
2830 to smpirun during execution (considering that simgrid
2831 is compiled with tracing enabled)
2832 - To run the simulation with gdb, the simulator
2833 accepts --cfg=tracing/smpi:1 to trace SMPI
2834 - tesh files to check if smpi tracing is ok
2835 - See examples/smpi/NAS/DT-trace
2836 * GTNetS tracing re-worked
2837 - adaptation to the tracing system of GTNets to cope
2838 with modifications regarding the fullduplex mode
2839 - new tesh files to check if gtnets tracing is ok
2842 * Asynchronous communications through the functions:
2843 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2844 * New function: MSG_load_platform_script()
2845 to make possible using a lua script instead of XML files to set up platforms
2846 * New function: MSG_set_function
2847 to associate functions to processes, used when bypassing the parser
2848 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2850 Platforms: Add some more examples in examples/platforms
2851 * Grid'5000: see www.grid5000.fr
2852 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2855 * Change the XML format. This is a very important modification. SimGrid
2856 3.5 introduces a new hierarchical format based on the notion of
2857 Autonomous Systems. Compatibility with old format is ensured
2858 through the perl script provided in the install bin directory
2859 bin/simgrid_update_xml.
2860 It is now possible to build platforms with specific routing
2861 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2862 connect several platforms together. We will try to provide soon
2863 set of realistic platforms exploiting these properties (have a
2864 look at examples/platforms/ for the moment).
2865 * Take the opportunity of the XML format change to be a good XML citizen:
2866 rename link:ctn to link_ctn and similar changes (also dealt with by
2868 * Add a new routing scheme (rule-based) using regular expressions. It
2869 enables to have an extremely low memory footprint when the
2870 underlying routing is simple and can be compactly described. You
2871 need to have libpcre4-dev (perl regular expressions) installed if
2872 you want to use this routing scheme.
2873 * Revive the cluster TAG and allow to easily and efficiently (both in
2874 term of memory and speed) connect clusters together. Have a look
2875 at teshsuite/simdag/platforms/ to see how this can be done. With
2876 this tag, you can create clusters with thousands of tasks at no
2877 cost (have a look at examples/platforms/).
2878 Note: clusters are implemented as ASes, so there is no need for an
2879 enclosing AS tag if you have only one cluster in your platform.
2880 * Add new generic functions in the public interface that allows the user
2881 to call SURF 'create_resource' methods from your code (same
2882 functionality as the XML bypass mechanism but with a much lighter
2884 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2885 that uses a piecewise linear approximation to produce better
2886 results when exchanging small messages.
2887 * Add a new parameter to handle correctly full duplex link and account
2888 for interferences between uplink and downlink communications
2889 (activate with --cfg=fullduplex:1).
2892 * Rename the SD_READY (all dependencies are satisfied and task is
2893 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2894 dependencies are satisfied) state.
2895 This prevents a confusion between the notion of "ready to schedule"
2896 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2897 simulated" (SD_RUNNABLE) used by the simulation kernel.
2898 * Change the way a task is considered as ready. Instead of removing
2899 dependencies when a task is done, a counter is decreased. This way,
2900 it is always possible to reach ancestors thanks to the
2901 SD_taks_get_parents function (even after the end of the simulation.)
2902 * Change the return type of SD_Simulate from (SD_task_t*) into
2903 xbt_dynar_t. This function was in handling a dynar internally and
2904 converted it into a NULL terminated array for historical reasons.
2905 * New function SD_dotload(char*) to load a DAG described in dot
2906 format. This loader and the corresponding examples require the
2907 installation of the graphviz library.
2908 * Fix a bug in the management of tasks of size 0 in the surf network
2909 models. This problem was only visible with SIMDAG and you should
2910 thus disregard results produced with earlier versions if you
2911 relied on this feature (some tasks were blocked because of this).
2912 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2913 even though your applications were doing only point-to-point
2914 communications and sequential computations. Now you can really use any
2915 model you want (of course, if you create real parallel tasks, which are
2916 not implemented in most models beside ptaskL07, this will abort).
2917 * Add an example that schedules a DAX on an heterogeneous platform
2918 using a Min-Min strategy.
2919 * New function SD_workstation_get_current_task() that returns the kind
2920 of task currently running on a workstation in the sequential access
2922 * Raise some warnings when unexecuted tasks remains at the end of the
2923 simulation. This is usually caused by cycles in the DAG.
2926 * New function: SIMIX_process_set_function() called by MSG_set_function
2927 * Change the underlying waiting queue in semaphores so that a process
2928 can wait on several of them simultaneously (as in waitany).
2929 * Fix the way to handle tokens in semaphores so that all access patterns
2930 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2931 * kill the dirty pimple SIMIX_message_sizes_output()
2932 Please use (proper) visualization instead
2935 * New data container: setset (set of sets of elements)
2936 * New module: mmalloc (mapped malloc, allowing to have several
2937 independent segments of malloc)
2938 * New function: xbt_dict_cursor_set_data()
2939 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2940 * New function: xbt_dynar_is_empty()
2941 * New function: xbt_fifo_get_last_item()
2942 * Fix xbt_dynar_shrink(): use the right element size.
2943 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2944 uninitialized areas during expand.
2945 * Fix semaphores: previous implementation was severely broken.
2946 * Use library init/fini functions for our initialization.
2947 - you can use logs and other feature as soon as you want in your
2948 code (even before the xbt_init / MSG_init)
2949 - xbt_exit is now a no-op and produce a warning when used.
2952 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2953 benefit from the latest improvement to the simulation kernel.
2954 * Kill measurement sockets for now. If you rely on them, sorry. This
2955 release is not for you. This feature will be reintroduced in the
2956 future, but we cannot delay the release any further.
2957 * New function: gras_msgtype_get_name().
2958 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2959 * Fix (at last) the pmm example: it should not randomly fail anymore.
2961 Build chain: bug fixes and overall polishing
2962 * Cmake is now stable enough. Hence, we killed the autotools.
2963 * Port to windows ( TM :)
2964 * Fix the 'make install' target.
2965 No need to use 'make install-simgrid' anymore
2966 * Introduce a 'make dist' target compiling a *source* archive
2967 'make package' compiles a binary archive
2968 * Compile java files only on need
2969 * Add --cd and --setenv command line options to tesh
2970 * Out of source builds are not fully supported yet, but we are close
2971 * Enable supernovae and optimization flags by default for our users
2974 * Add layer to set up environment directly from lua, without XML.
2975 * The effect of gras_stub_generator can be achieved through
2976 lua too (check examples/gras/console/ping_generator.lua)
2978 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2980 SimGrid (3.4.1) stable; urgency=low
2982 The "Polishing easter eggs is probably a good idea" release.
2983 This is a bug fixes release only.
2987 * Fix a bug preventing the tasks from begin garbage collected.
2990 * Fix a bug occurring when a host involved in a communication fails.
2991 This was not detected properly by the other peer involved in the
2992 communication. Now, it's reported as a network error.
2995 * Warn the user about loop dependencies in data flow of DAX files
2996 * Obey the control-flow dependencies of DAX files
2999 * Add option "enable_smpi" allowing to not compile SMPI.
3000 Probably useful for the (Mac) users experiencing a build error here
3001 * Improve the detection of lua5.1 and ruby1.8
3003 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
3005 SimGrid (3.4) stable; urgency=low
3007 The "Easter in Cargese" release. Also known as (major changes):
3009 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
3010 ~> bindings were greatly improved
3011 ~> new tracing infrastructure for better visualization introduced
3013 * the "Welcome to configury modernity" release.
3014 ~> we switched from autotools to cmake, and improved our cdash
3017 A more detailled list of changes follow (full detail in svn log).
3019 Java Bindings: Various Cleanups
3020 * (install java-gcj-compat-dev on debian-like to use them)
3021 * Remove put/get: no need to export deprecated interface in Java
3022 Use send/receive instead.
3023 * Cleanup the examples and add a README per directory
3024 * Remove example autoDestination (that's the only way to go now)
3025 * Remove example explicitDestination (was a plain copy of basic)
3026 * Make JniException a runtime exception, so that there is no need to
3027 declare the fact that you may encounter such a beast. I guess that
3028 nobody will ever want to survive such error.
3029 * Create specific errors for each MSG case of failure:
3030 host failure, transfer failure, timeout, task cancelled
3031 * Cleanup the exceptions that may get thrown by each function
3032 * Other internal cleanups in Java bindings. Performance still bad :/
3033 Ruby and Lua Bindings: create them
3034 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
3035 * That's new and great, you should try them out.
3036 Same functionalities than Java bindings, only even less polished
3038 * Kill the useless "rate" argument of SD_task_get_execution_time()
3039 Everyone used to provide -1 as a value, it was not used, and the
3040 semantic of a possible use wasn't even clear.
3041 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
3042 either as comm costs or compute costs to mean that there is no
3043 such thing for that specific task.
3044 * Add a SD_task_set_name() function
3045 * Fix SD_task_unschedule() on typed tasks
3046 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
3047 * In DAX loader, accept useless 'level' attributes to <job> since
3048 LIGO DAGs have them (seem to be to ease graphical representation).
3050 * Add an example masterslave_mailbox.c using send/receive and not
3051 the deprecated put/get interface.
3052 * Kill the MSG_paje_output() function. It's a noop since 2 years.
3053 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
3055 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
3056 (declare MSG_USE_DEPRECATED to still have the old name)
3057 * Add a MSG_task_set_data() function
3058 * About trace replay (see examples/msg/actions):
3060 - Allow to work with splitted trace files for each process
3061 Give the specific trace file as argument of each process,
3062 and call MSG_action_trace_run(NULL)
3063 You can still have one merged file for all processes.
3064 - Fix implementation of collective operations
3065 * Allow task_execute() on 0-sized tasks (closes #10063)
3067 * This is the first release of SimGrid where SMPI is not considered
3068 beta anymore (even if some corners should still be improved)
3069 * Port over the new SIMIX_network submodule (internal refactoring)
3070 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
3071 * Implement more missing elements of the standard:
3073 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
3075 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
3076 MPI_Scan() MPI_Get_processor_name()
3077 - Added implementation of missing case for Alltoall (warning: it's
3078 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
3079 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
3080 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
3081 SMPI_MPI_Allgatherv()
3082 * Bug fixes include:
3083 - MPI_Waitsome() was broken
3084 - Allow relative includes in smpicc
3085 - Command line cfg argument 'reference_speed' was ignored...
3086 - Some functions did not properly lead to auto-benching of user code
3087 - smpicc passes -O2 by default (just like openmpi one)
3089 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
3090 * Bug fixes about timeouts during communications
3091 * add SIMIX_message_sizes_output() as a pimple to write to file the
3092 amount of messages per size. Use gnuplot to get histogram.
3093 Pimple because that's the only user-visible function of simix,
3094 defined directly in xbt.h (irk, sorry)
3096 - Add a SIMIX_sem_get_capacity() function
3097 - Fix interactions with processe resume/suspende
3098 - release_forever() was stupidly broken
3099 - Fix SIMIX_display_process_status() for processes in a semaphore
3100 - Make SIMIX_sem_block_onto() user-visible
3101 * Refactoring context stuff:
3102 - Use pseudo-OOP for better modularity
3103 - reimplement SIMIX_process_kill() without process_schedule() so
3104 that the latter can take as invariant that it is called from
3106 - Merge context_start into context_new for sake of simplicity
3108 * Add a Vivaldi network model, coded live during SUD'10 ;)
3109 * Rename configuration variables to start a hierarchy:
3110 o cpu_model -> cpu/model
3111 o network_model -> network/model
3112 o workstation_model -> workstation/model
3113 * New configuration variables:
3114 o network/bandwidth_factor: correction to bandwidth
3115 o network/latency_factor: correction to latency
3116 o netwotk/weight_S: correction to the weight of competing streams
3117 * Add a long description to the models, that users can see with such
3118 argument on the command line: --cfg=cpu/model:help
3119 * --help-models display the long description of all known models
3121 * config: add the ability to set a default value after registration
3122 Does not override any previously set value (e.g. from cmd line)
3123 * dict: allow to have integer key and data.
3124 When so, you need to use the following functions
3125 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
3126 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
3127 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
3128 In contrary to regular dicts, the key is not malloced before copy.
3129 Mixing scalar and regular elements in the same dict is not tested
3131 * Allow to use xbt_dynar_shrink() to expend the dynar instead
3132 Tracing for Visualization:
3133 * SimGrid is now instrumented in order to generate a trace file for
3134 visualization analysis: to use it, need to compile SimGrid with the
3135 "tracing" option enabled, and instrument the program using SimGrid with
3136 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
3137 (among other functions).
3138 * The instrumentation only traces the platform utilization for now
3139 * Documentation to use the tracing functions and how to analyze the
3140 traces with the Triva tool is written.
3141 * More information about: SimGrid FAQ (in the section Tracing Simulations
3144 * We moved to cmake as default build system. Autotools support will
3145 be dropped soon. Check the FAQ for more info about how to use it.
3146 * Greatly improved our cdash/ctest interactions
3147 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
3148 * Added memory checking tests with valgrind; lot of memleak fixing.
3149 This may be the first release of simgrid with so few memory issues
3150 * Added code coverage tests.
3151 Our coverage is still improvable, but at least we see it on cdash.
3153 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
3155 SimGrid (3.3.4) stable; urgency=low
3157 The "Desktop Grid needs love too" release (also called Xmas release).
3159 Models improvements:
3160 * Major speedup in the maxmin system solving by using lazy evaluation
3161 Instead of solving completely the maxmin system at each iteration,
3162 only invalidate (and recompute) the modified parts.
3163 This new feature is enabled in default models but you can try to
3164 turn it on with "--cfg:maxmin-selective-update=1" for other models.
3165 * Cas01 IMproved as default CPU model
3166 This CPU model is the same Cas01 model, but it uses the
3167 maxmin-selective-update flag and a heap structure to manage
3168 actions on SURF kernel.
3169 It reduces the complexity to find the next action to finish and,
3170 consequently, it's faster than the old Cas01.
3171 This is the new default CPU model (Cas01).
3172 * Rename the old Cas01 model to Cas01_fullupdate
3173 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
3174 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
3175 * CpuTI (CPU Trace Integration)
3176 A new CPU model whose objective is simulate faster when using
3177 availability trace files.
3178 Instead of using a full featured, over engineered maxmin system for
3179 CPU modeling, this model does the pre-integration of traces files
3180 to calculate the amount of CPU power available, and so, executes
3181 faster than the old CPU models.
3182 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
3183 * Use LV08 as default network model since it gives better accuracy
3184 for small messages and shouldn't change things for big ones.
3185 Use --cfg=network_model:CM02 to get the previous behavior.
3188 ******************************************
3189 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
3190 ******************************************
3191 * The new CPU model may changes simulations!
3192 The point is that events occurring at the exact same timestamp
3193 are not scheduled in the same order with the old and new
3194 version. This may be enough to completely change the execution
3195 of simulations in some cases.
3196 * The new network model will change simulations!
3197 This new model is more realistic than the previous one, so you
3198 should consider redoing your old experiments with this model.
3199 Sorry for the inconvenience.
3202 * Introduce the supernovae compilation mode
3203 When compiled that way, the whole SimGrid (or almost) is put in a
3204 single compilation unit and compiled in one shoot.
3205 This is to help gcc which has difficulties to inline stuff from one
3207 The speedup seem to be above 15%, althrough more tests are needed on
3208 amd64 to confirm that gain.
3211 * Port of MSG's mailbox on top of SIMIX network
3212 The put/get mechanism was greatly simplified on the way.
3215 * New SIMIX network module. Provides:
3216 - Mailbox: rendez-vous mechanism to find with who you want to speak
3217 - Synchronous send/recv: easier and hopefully faster since the
3218 logic is handled in the maestro process directly now
3219 - Asynchronous send/recv: you dreamt of it? It's here now
3220 Too bad that nobody cared enough to propagate the change to MSG.
3221 * Add semaphores as SIMIX synchronization mechanism.
3224 * new function SD_daxload(char*) to load a DAX file
3225 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3226 * Introduce typed tasks. Specify its kind and cost at creation.
3227 At scheduling, just give where it should be placed, and the cost
3228 for each involved resource is automatically computed.
3229 Existing constructors so far (more to come of course):
3230 - SD_task_create_comm_e2e() for end-to-end communication
3231 - SD_task_create_comp_seq() for sequential computation
3232 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3233 * new function SD_task_dump() for debugging display
3234 * new function SD_task_dotty(task,FILE*) writing to file the info
3235 about the task in dotty format
3236 * SD_task_dependency_exists() can now cope with having one of its
3237 arguments NULL. If so, it tests whether the other argument has any
3239 * Add getters on list of preceding/following tasks:
3240 SD_task_get_parents(task) and SD_task_get_children(task)
3241 * Add getters on amount of workstations and list:
3242 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3243 * Add getter on task kind: SD_task_get_kind(task)
3244 * Update the start_time and finish_time of tasks on completion/failure
3245 * Bugfix: Remove task from state swags when destroyed
3248 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3251 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3252 dynar with one separate thread per value of the dynar.
3253 * Change the prototype of xbt_thread_create(), sorry.
3254 Added a boolean parameter indicating whether we want to join this
3255 thread (used in SG only for now)
3256 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3259 * GTNetS wrappers should now be usable again (and betterly tested too)
3260 * Fix a major regression from 3.2 where the timeout provided to
3261 MSG_task_put_with_timeout() was used as absolute time before which
3262 the comm should be done.
3263 * Start to fix the <cluster> tag.
3264 - Internal links should be good now (beside of the loopback, which
3265 use the private link instead)
3266 - paths to the external world is still rather broken
3267 - the <route:multi> tag is just broken. Actually that's brain-dead.
3268 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3269 to make it less stupid
3270 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3271 * Fix a source-level compatibility glitch from 3.2: after defining
3272 MSG_USE_DEPRECATED, you can use the old name
3273 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3274 * Allow to compile from the SVN with automake 1.11
3275 * Fix some problems when using the "start_time" tag in deployment XMLs.
3276 * Fix #8569: XBT/synchro.h has redundant declarations
3277 * Fix #8563: MSG return values and exceptions
3278 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3279 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3280 Thanks to Robson Peixoto.
3281 * Fix a vicious bug in dictionaries inducing that some elements were
3282 not freed on xbt_dict_free()
3284 Portability report of this version:
3285 * Main portability targets:
3286 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3287 on (amd64/i386/ia64)
3288 - mac leopard on i386
3289 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3290 but nothing critical.
3291 * Other platforms: windows, AIX and others were not tested for this release
3293 Timing report of this version:
3294 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3295 scenario parameters). From 8h to a few seconds in desktop grid settings.
3296 * Supernovae brings about 25% speedup on i386.
3298 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3300 SimGrid (3.3.3) stable; urgency=low
3302 The "Need for Speed" release.
3304 The timings done to validate the 3.3.2 were faulty.
3305 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3307 The problem was a conversion from a manually handled vector to
3308 xbt_dynar_t on the critical path.
3309 xbt_dynar_foreach calls functions, inducing stack management crap.
3311 We inlined these functions and xbt_dynar_foreach is now breath taking.
3312 We also inlined xbt_swag_belong on the way.
3314 Here are some approximate speedup measurements (on master/slaves
3315 simulations lasting between 10s and 20s each):
3316 3.3.1 -> 3.3.2: about same performance
3317 3.3.2 -> 3.3.3: 40% speedup
3318 3.3.1 -> 3.3.3: 40% speedup
3319 3.3.1 with inline patch -> 3.3.3: 30% speedup
3321 Our reading is that the refactoring which occurred in 3.3.2 made us
3322 suffer much more from the xbt_dynar_foreach low performance, but
3323 once we solved this, this refactoring proved to be very performance
3324 effective. From the 40% speedup, somehow, 10% are due to the
3325 inlining and 30% to the refactoring.
3327 That's a pitty that gcc cannot inline functions placed in other files
3328 alone. We have to choose between:
3329 - break the encapsulation (by putting private data structures and
3330 accessors in headers files to help gcc)
3331 - live with low performance
3332 - switch to a decent compiler such as icc (not quite possible).
3334 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3336 SimGrid (3.3.2) stable; urgency=low
3338 The "Simplicity does not preceed complexity, but follows it" release.
3340 The main contributors of this release were (lexical order):
3341 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3344 * Extract the routing logic into its own object.
3345 (was dupplicated in network.c and workstation_LV07.c;
3346 Allows to implement other ways of storing that info)
3347 => kill now useless network_card concept
3348 - Use dynar to represent routes (instead of void** + int*)
3349 - kill link_set (use surf_network_model->resource_set instead)
3350 - Add a command-line option to choose the routing schema to use
3351 - Add three new models:
3352 * Floyd (shortest path computed at initialization)
3353 * Dijikstra (shortest path recomputed all the time)
3354 * Cached Dijikstra (shortest path computed on need)
3355 All these models where contributed by Silas De Munck, and are
3356 described in his ICCS09 paper.
3358 * Simplify model declaration
3359 (less redirections, less function to write when defining a model)
3360 - Factorize stuff between models:
3363 surf_model_resource_set(model)
3364 surf_model_resource_by_name(model, name)
3365 - Unify the types of models in s_surf_model_t (using an union)
3366 - Embeed fields of common_public directly into s_surf_model_t
3367 - Rename model methods:
3368 action_free ~> action_unref
3369 action_change_state ~> action_state_set
3370 action_get_state ~> action_state_get
3371 - Change model methods into functions:
3372 (model)->common_public->action_use ~> surf_action_ref
3374 * Implement a generic resource; use it as ancestor to specific ones
3375 (allows to kill duplicated code in models)
3376 Drawback: timer command don't need no name nor properties;
3377 workstation_CLM03 don't need no properties
3378 (but I guess we can live with those few bytes wasted)
3380 * Improve the action object model
3381 - implement a constructor avoiding dupplicated code about field
3382 initialization in generic_action part.
3384 * Kill the SDP model: it has an external dependency, is deprecated
3385 in flavor of modern lmm models, and didn't compile since a while
3388 * Relocation of the context module from XBT to SIMIX.
3389 (the context were decoupled from the simix processes, duplicating a lot of code)
3390 => a lot of code was factorized
3391 - less overhead is introduced during scheduling
3392 - simpler API for the context factory
3393 - the logic for process creation,destruction and manipulation was simplified
3394 * Simplification of the s_smx_process_t data structure.
3395 => accesing the simix level data associated to a process is faster now,
3396 and the code is a lot more readable.
3399 * Implement some more MPI primitives:
3400 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3401 -implementation: Bcast: flat or 2-ary tree (default),
3402 Barrier: 4-ary tree,
3404 Allreduce: Reduce then Bcast
3405 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3406 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3407 Alltoallv: flat tree, like ompi
3409 * Add support for optimized collectives (Bcast is now binomial by default)
3410 * Port smpirun and smpicc to OS X
3413 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3414 Shout out if you used it.
3417 * Display the list of still queued messages in SG mode when existing
3421 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3422 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3423 * Add xbt_str_from_file(FILE*)
3424 * Add xbt_dict_get_key achieving a linear reverse search
3425 * Remove the context module
3427 Portability report of this version:
3428 * Main portability targets:
3429 - Linux(debian)/x86/context
3430 - Linux(debian)/x86/pthreads
3431 - Linux(debian)/amd64/context
3432 - Linux(debian)/amd64/pthreads
3433 On these, we still have the eratic breakages of gras/pmm and
3434 amok/saturate_sg reported in previous version. We still think
3435 that the tests are the cause of the fault, not the tested code.
3437 - Mac OSX Leopard/x86/context
3438 Still false negative in tesh autotesting.
3439 Smpi still fails, but this time because readlink does not accept -f
3440 Everything seems to work properly beside of that.
3443 - AIX version 5.3 (only tested contexts this time)
3444 Smpi still fails there because mktemp is not installed.
3445 Everything seems to work properly beside of that.
3447 I managed to compile it for the first time, but several breakages.
3448 Won't delay the release for this exotic platform.
3450 * Windows: it's still lagging behind. If you want to help, please
3453 Timing report of this version:
3454 This version seem to be more than 5% faster than 3.3.1 (on linux
3455 64bits with contextes). The gain is less than expected, we are
3456 investigating this for next release.
3458 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3460 SimGrid (3.3.1) stable; urgency=low
3463 * Implement a --cfg-help to show existing configuration variables
3464 * Build chain do not require doxygen in maintainer mode
3467 * fix a bug on struct sizeof computation, which prevented the
3468 exchange of arrays of structs in some conditions
3469 - added a regression test about this in datadesc_usage
3470 * Allow the exchange of 0-long dynamic vectors.
3471 - for that, use -1 as indicator of dynamic size instead of 0
3472 - This implied to change any size from unsigned long to long,
3473 reducing a bit communication abilities, but I guess that with
3474 64bits being quite common, this is more than enough.
3475 - This also induce a protocol change, thus bumping network protocol
3476 version from 0 to 1 (if we have external users, we have to get
3477 clean on that point too ;)
3478 - added two regression tests about this in datadesc_usage
3479 * Be more verbose when propagating local exceptions
3480 This helps debugging.
3481 * Display the status of simulated processes when receiving SIGINT in
3485 * Allow to control the simulation from a trace file.
3486 New functions MSG_action_register() and MSG_action_trace_run()
3487 The first one allows to associate a function execution to each
3488 kind of action while the second one parses a trace file and
3489 triggers the corresponding actions within the system.
3490 For now, only a toy example is provided in examples/msg/actions
3491 * Add an example of process migration in examples/msg/migration
3492 * Fix a bug in task exchange which broke MSG_task_get_sender()
3493 Add a teshsuite regression test for that.
3494 [Bug: if MSG_task_get_sender() is called after sender exit,
3496 * Fix a bug which prevented suspend/resume to work properly
3497 * Display the status of simulated processes when receiving SIGINT
3498 This fixes a regression of v3.3. due to the introduction of SIMIX
3499 * Bug fixing in failure management:
3500 - trace could not start by a failure at time 0
3501 - failure during communications were not working
3504 * Add SIMIX_process_set_name() to change the name of the current
3505 process in the log messages.
3506 * Store smx_hosts in a dict since we only retrieve them by name
3507 * Move the configuration infrastructure to surf
3510 * Move the configuration infrastructure to surf
3513 * Massive internal cleanups:
3514 - Store internal structures on processes instead of hosts (allows
3515 to have more than one process per host, in addition of being more
3517 - Cleanup the initialization/finalization process
3518 - Kill a whole bunch of unneeded synchronization:
3519 processes run in exclusive manner within the simulator
3520 - Move queues from global tables to process data fields
3522 - now accept -platform and -hostfile arguments
3523 - Pass the right rank value to processes according to the hostfile
3524 * Compile the examples by default, and use them as regression tests
3525 * Implement MPI_Wtime()
3526 * Change the reference speed to a command line option
3529 * TCP_gamma can now be specified as command line option using
3530 --cfg=TCP_gamma:10000000.0
3531 * Change the --surf-path cmd line option into --cfg=path:
3534 * Also include strbuff from xbt.h public header
3535 * xbt_ex_display(): do not free the exception after displaying
3536 This allows to do more with the given exception afterward.
3537 Users should call xbt_ex_free() themselves.
3541 Portability report of this version:
3542 * Main portability targets:
3543 - Linux(debian)/x86/context
3544 - Linux(debian)/x86/pthreads
3545 - Linux(debian)/amd64/context
3546 - Linux(debian)/amd64/pthreads
3547 These targets fail about 1/10 of times on gras/pmm, but we believe
3548 that this is because of the test, not because of simgrid.
3549 amok/saturate_sg fails even more rarely, and the test may not be
3552 - Mac OSX Leopard/x86/context
3553 The test suite still spits tons of errors because some obscure
3554 force prevents us from removing the temporary directories
3555 arguing that they still contain some metadata I've never heard of.
3556 Smpi fails because seq is not installed.
3557 Everything seems to work properly beside of that.
3560 - AIX version 5.3 (both contexts and pthread)
3561 Smpi still fails there because mktemp is not installed.
3562 XML inclusions seems rosty on AIX.
3564 * Windows: it's still lagging behind. If you want to help, please
3567 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3569 SimGrid (3.3) stable; urgency=high
3573 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3574 [Malek Cherier & Mt]
3576 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3579 Until now, GRAS were using MSG as an interface to SURF. It was
3580 quite difficult because both interface have several differences
3581 (MSG channels vs GRAS sockets were the most notable point).
3583 This also opens the gate to SMPI (which should occur soon) and speed
3584 up simulations by to 40% (even if it were not the main goal).
3586 **************************************
3587 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3588 **************************************
3589 The point is that events occurring at the exact same timestamp are
3590 not scheduled in the same order with the old and new version. This
3591 may be enough to completely change the execution of simulations in
3592 some cases. Sorry for the inconvenience.
3594 * Cleanup and upgrade the XML format to push further scalability
3595 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3597 * Improve the testing infrastructure with tesh. Now a very large part of
3598 the code is tested not only by being run but also by checking that the
3599 output match an expected output [Mt].
3601 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3602 is really less memory-demanding, which should allow you to use
3603 larger files in SimGrid [AL].
3605 * Inform valgrind about our contextes, so that it becomes usable
3606 with the default (and more effecient) version of SimGrid
3607 [contributed by Sékou Diakite, many thanks]
3610 * Introduce a listener thread in charge of receiving incoming
3611 messages from the network. It allows to overlap communication and
3612 computation but most notably, it removes some stupid deadlocks due
3613 to the fact that so far, a process could not send and receive at
3614 the same time. This made most non trivial communication schema
3616 * Convert the PIDs from long int to int to match the MSG ones (and
3617 linux ones too) [Mt]
3618 * New function: gras_agent_spawn() to launch a new process on
3619 current host. Only working in simulation for now. [Mt]
3620 * New function: gras_os_hostport() returning a constant form (ie,
3621 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3624 * Make the backtrace of exceptions more human readable [Mt]
3625 * New module: xbt/str [Mt]
3626 a ton of string utility functions (split, join, printf to a newly
3627 allocated buffer, trim, etc)
3628 * New module: xbt/hash [Mt]
3629 SHA1 hashing algorithm (more to come if needed)
3630 * New module: xbt/synchro [Mt]
3631 synchronization tools (mutex and conditions) working the same way
3632 in simulation and in real life (mainly useful for GRAS, but not
3634 * New module: xbt/queue [Mt]
3635 classical producer/consumer synchronization scheme
3636 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3637 (using the classical functions will get serialized) [Mt]
3638 * Make dictionary internal table dynamic. No need to specify its size
3639 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3641 * Make sure the log channels are organized as a tree under windows
3642 (because of ANSI C compatibility issue, any channel were child of
3643 root directly) [Mt].
3646 * Cleaned many thing in surf and fixed a few bugs [AL].
3647 * Add a nice command line configuration mechanism to compose models [AL].
3648 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3649 the previous one (KCCFLN05). It relies on something that looks like
3650 a max-min sharing mechanism but cannot be written as such. A new solver
3651 was thus designed [AL].
3652 * Add a new solver to lmm. Based on Lagrange optimization and
3653 gradient-based descent, it enables to efficiently maximise systems s.a
3655 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3657 This solver enables to propose two new network models for TCP Reno and
3658 TCP Vegas based on Low's work. These models still need to be fully
3659 tested though [Pedro Velho].
3662 * Bug fix in SD_simulate. Now the time bound given as argument is
3664 * Use the new parallel task model (ptask_L07) as default.
3665 * Use the SURF command line configuration mechanism.
3666 * 0-size tasks (for synchronization) should now work.
3668 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3670 SimGrid (3.2) stable; urgency=high
3674 We still experience issues on this platform, but we believe that at
3675 least MSG is usable.
3677 GRAS API BREAKAGE (for simplification purpose, sorry):
3678 * the gras_msgtype_by_name is not used anymore. Instead of
3679 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3680 you can write (and must)
3681 gras_msg_send(toserver, "request", &request);
3682 - If you still want to pass a gras_msgtype_t to the function (to cache
3683 the type and avoid the lookup time), use the gras_msg_send_() variant.
3684 - Impacted functions:
3685 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3686 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3687 * The callbacks are now expected to return 0 when everything went well
3688 (just like the main() function)
3690 GRAS new features and improvements:
3691 * New module mechanism where user code can use per process globals [Mt]
3692 This is similar to gras_userdata_*() functions, but for libraries. It
3693 factorize some code developped over and over in the examples and AMOK.
3694 It has still to be documented and used (only amok/peermanagement is
3696 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3697 loss when they were small enough to fit into the buffer and sent quickly
3698 enough so that they can all get received in one shoot.
3699 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3700 if not found. Use the *_or_null() variant for the old semantic.
3701 * In gras_msg_handle, do not discard messages without callback.
3702 They are probably messages to be explicitly awaited later (ie, proofs of
3703 mis-synchronization in userland since they are sent before being awaited)
3704 No big deal usually.
3705 * gras_socket_meas_send/recv: semantic changed!
3706 The numerical arguments used to be (1) the total amount of data to send
3707 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3708 messages. This was need for the fool willing to send more than MAXINT
3709 bytes on quite fat pipes.
3712 * Do really rename the hostmanagement module to peermanagement. [Mt]
3713 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3714 but this is rather new and this was documented in the module
3715 documentation (poor excuses, I admit)
3716 * Bandwidth measurement semantic changed! This follows the changes to
3717 gras_socket_meas_send/recv explained above.
3720 * A sequential mode has been added to the workstations. When a workstation
3721 is in sequential mode, it can execute only one task, and the other tasks
3722 are waiting in a FIFO. [Christophe Thiery]
3725 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3726 model for SIMDAG. [Christophe Thiery]
3727 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3728 equal to 0, which caused some bad behaviors in
3729 saturated_constraint_set_update. I now use a threshold mechanism like in
3733 * When running manually src/testall, you select specific units [Mt]
3734 testall is the result of our cunit mechanism, and should replace all
3735 the scripty thingy around since bash don't run easily on billware.
3737 * A mallocator system has been added. [Christophe Thiery]
3738 Mallocators allow you to recycle your unused objects instead of freeing them
3739 and allocating new ones.
3741 Documentation update:
3742 * FAQ reworking + New FAQs:
3743 - "Valgrind spits tons of errors!" [Mt]
3744 - "How to repport bugs" [Mt]
3745 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3746 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3747 - Communication time measurement within MSG [AL]
3748 - I experience weird communication times when I change the latency [AL]
3749 * GRAS tutorial [Mt]
3751 - an introduction to the framework and to the used communication model
3752 - an initiatic tour introducing the most proheminent features:
3754 . Lesson 0: Installing GRAS
3755 . Lesson 1: Setting up your own project
3756 o Part 2: Message passing
3757 . Lesson 2: Exchanging simple messages
3758 . Lesson 3: Passing arguments to the processes (in SG)
3759 . Lesson 4: Attaching callbacks to messages
3760 . Lesson 5: Using globals in processes
3761 . Lesson 6: Logging information properly
3762 . Lesson 7: Using internal timers
3763 . Lesson 8: Handling errors through exceptions
3764 . Lesson 9: Exchanging simple data
3765 . Lesson 10: Remote Procedure Calling (RPC)
3766 . Lesson 11: Explicitly waiting for messages
3767 . Recapping of message passing features in GRAS
3768 - A HOWTO section containing:
3769 o HOWTO design a GRAS application
3770 More are due, of course. They will come latter. In the meanwhile, you can
3771 check the examples which are still here.
3773 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3775 SimGrid (3.1) stable; urgency=high
3779 There was a stack corruption somewhere, visible only when optimizing
3780 with these versions. [Vince]
3783 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3784 module that enabled to play with Directed Acyclic Graphs. It is built
3785 directly on top of SURF and provides an API rather close to the old
3786 SG. Some old codes using SG are currently under rewrite to check that
3787 all needful functions are provided. [Christophe Thiery]
3790 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3791 extension of the classical CLM03 model that gracefully handles
3792 failures. This is now the default model for MSG and GRAS. It doesn't
3793 handle parallel tasks yet though. [AL]
3794 * Bug fix: Weights were not correctly set in the network part.
3795 WARNING: This may have resulted in incorrect results with simulations
3796 where there are more than one flow on a given link. [AL]
3799 * After a (long ?) discussion on simgrid-devel, we have decided that the
3800 convention we had on units was stupid. That is why it has been decided
3801 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3802 WARNING: This means that all previous platform files will not work as
3803 such with this version! A warning is issued to ask users to update
3805 A conversion script can be found in the contrib module of the CVS, under
3806 the name contrib/platform_generation/surfxml_update.pl [MQ]
3809 * Bug fix: Processes were started in reverse order, wrt deployment file.
3810 WARNING: if your code relies on this bug, please fix it. [AL]
3811 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3812 being executed on two different locations. [AL]
3813 * Bug fix: Failures are now better supported thanks to Derrick's tests
3814 (there was many failure situations I hadn't thought of and that weren't
3815 correctly handled). [AL]
3816 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3820 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3821 Exception occurring on server-side are propagated back to client (!).
3823 API CHANGE: the callback changed their prototype. Change:
3824 int my_handler(gras_socket_t expeditor, void *payload_data) {
3826 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3827 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3829 * New! function: gras_msg_handleall to deal with all messages arriving
3830 within a given period.
3831 * New! function: gras_socket_server_range to get a server socket in a
3832 range of port numbers (ease to avoid port number conflicts) [MQ]
3833 * New! gras processes display their backtrace when they get a SIGUSR1
3834 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3835 Sweet to debug RL processes [MQ]
3839 - Do not force experiment sizes to be expressed in kb, or it becomes
3840 impossible to measure the latency this way (needs one byte-long tests)
3841 WARNING: this changes the amok_bw_* function semantic. [MQ]
3842 - Implements the link saturation stuff. [MQ]
3843 * Peer management module:
3844 New! module factorizing code that we wrote over and over [MQ].
3847 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3848 - Test units are placed directly into the library code, they get extracted
3849 automatically and placed into the src/testall binary.
3850 - Convert most of the XBT tests to this system.
3851 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3852 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3853 live in the same C code. This required to change a little bit the API
3854 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3855 * New module: sparse graph structure with basic algorithms (this is work
3856 in progress and the API is not considered to be frozen yet). [AL]
3857 * Display more information on backtraces: source line & function names are
3858 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3859 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3860 * Reworked a little bit some #include statements to load only required
3861 headers. Some user code that relied on SimGrid to include stdlib or
3862 stdio may need to include it by themselves. [AL]
3863 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3865 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3866 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3867 but I guess that almost nobody used those parts. [MQ]
3869 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3871 SimGrid (3.0.1) stable; urgency=low
3874 * Unfortunately, I had missed 5 misnamed functions:
3875 xbt_fifo_item_t xbt_fifo_newitem(void);
3876 void xbt_fifo_freeitem(xbt_fifo_item_t);
3877 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3878 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3879 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3880 They're now deprecated. Please use their new versions:
3881 xbt_fifo_item_t xbt_fifo_new_item(void);
3882 void xbt_fifo_free_item(xbt_fifo_item_t);
3883 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3884 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3885 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3887 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3888 * Documentation: xbt_log module unmercifully reworked [MQ]
3889 * Bugfix: there was a problem with the ending of contexts with
3890 the pthread backend. It caused some weird deadlock or behavior
3891 depending on the pthread implementation. [AL]
3892 * Bugfix: get the exceptions raised in the simulator repport where
3893 and why they come from when they are not caught in time [AL, MQ]
3896 * Bugfix: Do repport the error when two non-connected hosts try to
3897 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3900 * Add additionnal checkings on communications. Assert that two
3901 communicating hosts are connected by a set of links... [AL]
3904 * Add additionnal checkings on channel values in communication [AL]
3905 * New: MSG_task_get_source to see on which host a task was generated [HC]
3906 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3907 the number of tasks waiting to be received on channel and sent
3909 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3910 waits for the first task coming from a given host.. [AL]
3912 GRAS new functionnalities: [MQ]
3913 * Enhance the parsing macro to allow the size of multidimentional objects
3914 to be given thru annotations.
3915 * New example (and documentation): Matrix Multiplication a la RPC
3916 (as when I was young!) and fix a bunch of bugs found on the way.
3918 GRAS performance improvements: [MQ]
3920 * Reduce the amount of cbps creation/destruction by making it static to
3921 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3924 * Change libdata to a set so that we can search for stuff by ID (and thus
3925 reduce the insane amount of dict lookups)
3928 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3929 RL communication on top of SG since it's so uneffective.
3930 It may also be used for inter-thread communication in RL, one day.
3931 * Use gras_datadesc_copy() to exchange messages on top of SG
3933 - improve message exchange performance on top of SG
3934 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3935 * Don't exchange on the network the size of the used part of buffer,
3936 instead, specify the possible buffer size to read().
3938 - reduces the amount of read/write calls (one pair per exchange)
3939 - reduces the amount of exchanged data (the size)
3940 - allows to retrieve all arrived data on receiver side, if we don't need
3941 it right now (subsequent read will peek the buffer)
3942 - allows the receiver to proceed with the beginning of the stream before
3943 everything is arrived
3944 - make it possible to build an iov transport (using readv/writev)
3946 - take care of the data with non-stable storage (like stacked data),
3948 * If possible, TCP send uses vector I/O (when writev() is here)
3949 - Don't use it for receive since we send data sizes and data on the
3950 same stream, so we wouldn't be able to chain large amount of chunks
3951 before having to flush the stuff to read the size.
3952 * Rework the transport plugin mechanism to simplify it and reduce the
3953 amount of pointer dereferencement when searching for the right function
3956 * I guess that now, we do almost as few system calls as possible while
3957 doing as few data copy as possible.
3959 To improve it further, we could try to send all the sizes first and then
3960 all the data (to use iov on receiving size), but it's only a partial
3961 solution: when you have 2 dimensional data, the sizes of the second
3962 dimension is data of the first dimension, so you need 3 streams.
3964 I'm not sure the potential performance gains justify the coding burden.
3966 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3968 SimGrid (3.00) stable; urgency=high
3971 * New! Give the possibility to hijack the surf parser and thus bypass
3972 MSG_create_environment and MSG_launch_application. Have a look at
3973 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3975 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3977 SimGrid (2.96) unstable; urgency=low
3982 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3983 This deprecates the xbt_error_t mechanisms.
3984 It modifies (simplifies) all XBT and GRAS API.
3985 MSG API keeps unchanged (exceptions raised by XBT are caught from
3986 within MSG and masked with existing error handling facilities)
3989 * New! Add a FATPIPE model. [AL]
3990 * New! Add a parallel task model. [AL]
3991 * New! Add automatically a loopback interface (in the default
3992 network model) if none was precised.
3995 * Bugfix: MSG_process_resume now works with the current running process.
3997 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3998 * Modification of MSG_task_get_compute_duration. Once a task has been
3999 processed, the value returned by this function is now equal to 0. [AL]
4000 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
4001 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
4002 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
4003 of a m_task. For now, MSG_task_cancel only works with computation
4005 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
4006 of the processor (in Mflop/s) regardless of the current load on the
4008 * API Change: use proper naming convention for MSG_getClock and
4009 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
4011 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
4012 This function changes the priority of a computation task. This priority
4013 doesn't affect the transfer rate. A priority of 2 will make a task
4014 receive two times more cpu power than the other ones. This function
4015 has been added to suit the needs of Nguyen The Loc and hasn't been that
4016 much tested yet. So if it fails, please report it and send me your code.
4018 * API Change: removed all functions and types that were marked "deprecated"
4019 since many months. Renamed MSG_global_init_args to MSG_global_init.
4021 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
4023 SimGrid (2.95) unstable; urgency=low
4026 * Steal some nice code to GNU pth to fix context detection and usage [AL]
4027 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
4028 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
4031 * Bugfix: Allow absolute paths for platform description files [MQ]
4032 * Bugfix: do free the variables after use. Leads to drastic performance
4034 * Implement max_duration (ie, timeouts) on resources [AL]
4037 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
4039 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
4040 the needed support (timeouts) [AL]
4043 * Implement measurement sockets. You can now get the bandwidth between two
4044 hosts thanks to AMOK (see below). [MQ]
4045 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
4046 dynar over the network (yeah) [MQ]
4047 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
4048 * simple/static token-ring example. [Alexandre Colucci and MQ]
4049 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
4050 in place before [MQ]
4053 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
4054 type parsing. "short" and "long" modifiers were also ignored; other
4055 modifier (such as reference level) are still ignored. [MQ]
4056 * Embeed the buffer size within the buffer itself on SG. [MQ]
4057 That way, send() are atomic and cannot get intermixed anymore (at least
4058 the ones which are less than 100k; bigger messages still have the issue)
4059 * Array size pushed by the field, not by the field type (or each
4060 and every long int will push stuff to the cbps) [MQ]
4061 * use select() to sleep since it allows to portably sleep less than one
4064 GRAS (minor cleanups)
4065 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
4068 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
4069 * type category 'ignored' killed as it was never used and were difficult
4071 * whether a type can cycle or not is now a flag, leaving room for more
4072 flags (as "ignored", if we feel the need one day ;)
4073 * Rename raw sockets to measurement sockets since "raw" has another
4074 meaning in networking community.
4077 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
4078 and offers features not belonging to GRAS but that most applications
4079 need. One day, it may be a set of plugins loadable at runtime.
4080 * New module: bandwidth
4081 bandwidth measurement between arbitrary nodes running this module. [MQ]
4083 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
4085 SimGrid (2.94) unstable; urgency=low
4087 The first beta release of SimGrid 3 !
4091 * Update the main page and the FAQ. Adding references to gforge.
4094 * Add a gras_os_getpid function.
4097 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
4098 * Extend the logs so that they also print PID, hostname, date, ... if
4100 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
4101 and kill the old version which were in testsuite/
4102 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
4106 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
4107 for users. Instead, each programming environment (GRAS, MSG,...) use it
4108 when needed to provide such a feature to users.
4109 Don't shortcut the mechanism or you will also shortcut the virtualization
4110 you need on the simulator.
4114 * Cleanups in configury with regard to compile optimization/warning flags.
4115 Also add -fno-loop-optimize to any powerpc since it's the optimization
4116 killing gcc (< 3.4.0).
4117 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
4118 needed by MSG examples complications
4119 * Borrow configury beautifications from PHP
4122 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
4123 warning (thanks loris for stumbling into this one).
4124 * Bugfix: stop loading private headers (gras_config.h) from the public
4128 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
4129 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
4130 RL would imply a DNS resolver, which is *hard* to do in a portable way
4131 (and therefore delayed).
4132 * Implement a real timer mechanism and use it in timing macros. This allows
4133 to avoid rounding errors and get a 0.000005 sec precision in timing
4134 macros. While I was at it, various cleanups:
4135 - allow to declare more than one timed section per file (fix a stupid bug)
4136 - move some private declaration to the right place
4137 - merge conditional execution and timing macros into emulation module
4138 - document the module
4139 - make sure the module cleanups its mess on gras_exit
4140 * Documentation improvements:
4141 - (new) how to compile applications using GRAS
4142 - (new) emulation support (timing macros)
4144 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
4146 SimGrid (2.93) unstable; urgency=low
4148 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
4151 - Use Paje properly where used. Still to be sanitized properly.
4152 - Portability fix: Add an implementation of the contexts using pthread
4156 - Add xbt_procname(): returns the name of the current process.
4157 Use it to show the current process's name in all logging.
4159 - fix detection of older flex version and the reaction, since we do
4160 depend on modern ones (we use lex_destroy)
4161 - Better separation of SG and RL in the libs: remove all simulation code
4162 from libgras. As a result, this lib is now only 200k when stripped.
4163 Some of the xbt modules may also be duplicated (two sets and such) and
4164 should be cleaned/killed before SG3.
4165 - Insist on using xlC on AIX because of weird problems involving gcc there.
4166 - Cleanup the make remote stuff. This is now done by scripts
4167 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
4168 mainly for our private use, but we're working on changing them to user
4171 - Bugfix: flush the socket on close only if there is some *output*.
4172 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
4174 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
4177 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
4179 SimGrid (2.92) unstable; urgency=low
4181 Alpha 3 on the path to SimGrid 3
4185 - New! First try of benchmarking macros.
4186 - New! First try so that gras_stub_generator generate deployment and
4187 remote compilation helpers.
4189 - Bugfix: Initialization fix in msg_test.
4193 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
4195 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
4197 - API Change: gras_os_sleep to take the amount of seconds as a double.
4198 Accepting an int was error prone since it was the only location where
4199 seconds were coded as such. It leaded to damn rounding errors.
4200 - Bugfix: Hard to belive that timers ever worked before this.
4202 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4204 SimGrid (2.91) unstable; urgency=low
4206 Alpha 2 on the path to SimGrid 3
4210 - Bug fix in the lmm_solver.
4212 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4213 through the function MSG_paje_output.
4214 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4215 - It is possible to bound the rate of a communication in MSG with
4216 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4217 to write it in the changelog).
4218 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4223 - Various cleanups to the autotools stuff
4224 - Begin to move Gras examples to examples/gras/
4225 - Let make distcheck work again (yeah!)
4227 - documentation overhauled using doxygen.
4228 gtk-doc-tools is dead in SimGrid now.
4229 - Automatically extract all existing logging categories, and add the list
4230 to the documentation (long standing one, to say the less)
4232 - Cleanup the known architecture table. Reorder the entries to group what
4233 should be, and use a more consistent naming scheme.
4234 (some of the test dataset are still to be regenerated)
4235 - New! Allow library to register globals on each process just as userdata
4237 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4238 lookup time (for now).
4239 Use it in msg and trp.
4240 This cleans a lot the internals and helps enforcing privacy of the
4241 headers between the gras components.
4242 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4243 - Bugfix: gras_os_time was delirious in RL.
4244 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4246 - Reenable GRAS now that it works.
4248 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4250 SimGrid (2.90) unstable; urgency=low
4252 Alpha 1 on the path to SimGrid 3
4254 * It is a long time since the last release of SimGrid. I'm sorry about
4255 that but as I had told you, I was rewriting a lot of things. I apologize
4256 to those who had been reporting bugs to me and that I had not answered.
4257 If your bug is still in the new version, please tell me. Here is a
4258 summary of the main changes.
4260 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4261 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4262 portability, speed, and a lot more but you'll figure it out later.
4263 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4264 I wanted to release the new MSG as soon as possible and I have
4265 broken GRAS, which is the reason why, for now, only MSG is fully
4266 functional. A laconic description of these projects is available
4267 in the documentation.
4269 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4270 kernel optimized for our needs (called SURF but only the developers
4271 should use it). Hence, MSG is now roughly 30 times faster and I think
4272 that by rewriting a little bit MSG, I could event speed it up a little
4273 bit more. Beside the gain in speed, it is also much easier to encode a
4274 new platform model with SURF than it was with SG. More to come...
4276 * REVOLUTION 3: I have tried to change a little as possible the API of
4277 MSG but a few things really had to disappear. The main differences
4278 with the previous version are:
4279 1) no more m_links_t and the corresponding functions. Platforms are
4280 directly read from an XML description and cannot be hard-coded
4281 anymore. The same format is used for application deployment
4282 description. The new format is described in the documentation.
4283 Have a look in tools/platform_generation. There is a tiny script
4284 that converts from the old platform format to the new one. Concerning
4285 the application deployment format, parsing the old one is tricky.
4286 I think most of you should however be able to convert your files. If
4287 it is really an issue, I can write a C code that does the conversion.
4289 2) the toolbox tbx does not exist anymore. We now have a library
4290 with much more data-structures but without the hash-tables (we have
4291 dictionaries that are much faster).
4293 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4295 *****************************************************************************
4296 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4297 * don't feel like converting the oldies. (Mt) *
4298 *****************************************************************************
4301 Version 2.90: "the long awaited one"
4302 - Finished rewriting and debugging MSG. Rewrote the documentation.
4303 - disable GRAS for now since it needs to be ported to the newest SG
4306 - Finish the port to windows (using mingw32 for cross-compile)
4309 - Main loop and datastructures of SURF. A cpu resource object is
4310 functional. Surf can thus be used to create cpu's with variable
4311 performance on which you can execute some actions.
4313 2004-11-15 Martin Quinson
4314 - Port to ARM. Simply added the alignment and size descriptions. Should
4315 work, but the ARM machines are so slow that I didn't had the opportunity
4316 to 'make check' over there yet.
4318 2004-11-15 Arnaud Legrand
4319 - Trace manager now written. It uses a heap structure and is therefore
4320 expected to be efficient. It may however be speeded up (particularly
4321 when many events occur at the same date) by using red and black
4322 trees. One day maybe...
4323 - Max-min linear system solver written. It uses a sparse matrix
4324 structure taking advantage of its expected use. Most operations are
4325 O(1) and free/calloc are called as few as possible. The computation of
4326 the minimum could however be improved by using a red and black tree
4329 2004-11-03 Arnaud Legrand
4330 - Rename every gras_* function that was in xbt/ to its xbt_
4332 - Add a heap and a doubly-linked list to xbt
4333 - Added a dichotomy to the dictionaries. make check works as well before
4334 so I assume that the patch is correct. I do not know however if things
4335 run effectively faster than before now. :)
4337 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4338 SimGrid, and the version number is bumped to 2.x
4340 2004-10-29 Martin Quinson
4341 - Introduction of the remote errors.
4342 They are the result of a RMI/RPC on the remote machine.
4343 ErrCodes being scalar values, you can't get the host on which those
4344 errors did happen. Extending the error mechanism as in Gnome is possible.
4345 No idea yet whether it is a good idea.
4347 2004-10-28 Martin Quinson
4348 - Interface revolution: the Starred Structure Eradication.
4349 I used to do typedef struct {} toto_t; and then handle *toto_t.
4350 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4351 - ???_t is a valid type (builded with typedef)
4352 - s_toto_t is a structure (access to fields with .)
4353 - s_toto is a structure needing 'struct' keyword to be used
4354 - e_toto_t is an enum
4355 - toto_t is an 'object' (struct*)
4357 typedef struct s_toto {} s_toto_t, *toto_t;
4358 typedef enum {} e_toto_t;
4359 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4360 s_toto_t) is private.
4362 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4363 so that it can set the variable to NULL. It was so for dicts and sets,
4364 it changed for dynars.
4366 - Fix a bunch of memleaks in dict_remove
4367 - Fix a bug in sg/server_socket opening: it failed all the time.
4369 2004-10-07 Martin Quinson
4370 - Speed up dynar lookup operation a bit.
4372 gras_dynar_get is dead.
4374 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4375 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4376 (which gives you the address of the stored data).
4378 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4379 copy of the data using an affectation to do the job and not a memcpy.
4381 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4382 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4384 It does not really speedup the dynar test because they are
4385 setting elements all the time (and look them seldom). But the dict does
4386 far more lookup than setting.
4388 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4390 2004-10-05 Martin Quinson
4391 - Allow to (en/dis)able the cycle detection at run time.
4393 Whether we should check for cycle or not is now a property of each
4394 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4395 datadesc_cycle_unset allow to remove this property when previously set.
4397 Note that the cycle detection is off by default since it impacts the
4398 performance. Watch the data you feed GRAS with ;)
4400 This property is hereditary. Any element embedded in a structure having it
4401 set have it set for the time of this data exchange.
4403 You should set it both on sender and receiver side. If you don't set it on
4404 sender side, it will enter an endless loop. If you forget on receiver
4405 side, the cycles won't be recreated after communication.
4407 - Header reorganization.
4408 Kill gras_private.h, each submodule must load the headers it needs.
4410 2004-10-04 Martin Quinson
4411 - Interface revolution: do not try to survive to malloc failure.
4413 Now, gras_malloc and friends call gras_abort() on failure.
4414 As a conclusion, malloc_error is not a valid error anymore, and all
4415 functions for which it was the only gras_error_t return value are
4416 changed. They now return void, or there result directly.
4417 This simplify the API a lot.
4419 2004-09-29 Martin Quinson
4420 - Re-enable raw sockets.
4421 Created by gras_socket_{client,server}_ext;
4422 Used with gras_raw_{send,recv}
4425 It should allow to kill the last bits of gras first version soon.
4427 This is not completely satisfactory yet (duplicate code with
4428 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4431 - Simplify transport plugin (internal) interface by not passing any
4432 argument to _server and _client, but embedding them in the socket
4435 2004-09-28 Martin Quinson
4436 - Finish the port to AIX.
4437 autoconf was my problem (segfault within the malloc replacement
4438 function. No idea why)
4440 2004-09-16 Martin Quinson
4441 - Fix some size_t madness on 64bit architectures.
4443 2004-09-08 Martin Quinson
4444 - Reduce the number of system headers loaded, overload some more system
4445 calls (such as malloc to cast the result of the system one, and work
4447 - Fix and reintroduce the config support
4449 2004-09-07 Martin Quinson
4450 - Source code reorganization to allow Arnaud to surf all over there.
4451 - Allow to document the logging categories.
4452 - Remove all uppercase from logging categories and useless cleanup in names.
4454 2004-08-18 Martin Quinson
4455 Version 0.6.2 (protocol not changed; API changed)
4456 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4457 gras_error_t), and NULL when not found. Functions expecting a msgtype
4458 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4459 hopefully usefull message.
4460 - Portability to prehistoric sparcs again
4462 2004-08-17 Martin Quinson
4463 Version 0.6.1 (protocol not changed; ABI not changed)
4464 - prealloc some buffers to speed things up
4466 2004-08-11 Martin Quinson
4467 Version 0.6 (protocol not changed; ABI expended)
4468 - The parsing macro can deal with the references, provided that you add
4469 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4471 2004-08-09 Martin Quinson
4472 Version 0.5 (protocol not changed; ABI changed)
4473 - Allow to off turn the cycle detection code in data exchange at
4474 compilation time. It should be at run time, but I'm short of time (and
4475 the config stuff is still broken). That way, we keep dict out of the
4476 critical path, which is good because the performance is poor:
4477 - search not dichotomial yet
4478 - dynar give no way to access their content and memcpy everytime
4479 - In composed data description (struct, ref and so on), stop foolness of
4480 keeping the subtype's ID, but store the type itself. This keeps sets out
4481 of the critical path, which is good since they rely on dynar and
4482 dictionaries. The only loose of that is that we cannot detect the
4483 redeclaration of a structure/union with another content (but I'm not sure
4484 the code detected well this error before anyway). We still can detect
4485 the redefinition discrepancy for the other types.
4486 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4487 it breaks the code because of type-punning used all over the place).
4488 This breaks on all non-gcc architectures (for now).
4490 All those changes (plus the buffer of last time) allow me to gain 2 order
4491 of magnitude on cruel tests consisting of 800000 array of integers on two
4492 level of a hierarchical structure (200 secondes -> 4 secondes)
4495 - the selector of reference must now return the type it points to, not
4496 the ID of this type.
4498 2004-08-06 Martin Quinson
4499 Version 0.4 (protocol changed; ABI not changed)
4500 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4501 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4502 - Implement a buffer transport, and use it by default (it relies on tcp in
4503 real life and on sg in simulation).
4504 That's a bit hackish since I had a new field to the structure to store
4505 its data without interfering with the subtype ones. Inheritance
4506 is tricky in C. And that's a kind of reverse inheritance with one class
4507 derivating two classes. Or maybe a game with java interfaces. Anyway,
4508 that's damn hard in C (at least).
4509 Moreover, I got tired while trying to ensure plugin separation and
4510 genericity in SG mode. MSG wants me to do weird things, so let's go for
4511 cruel hacks (temporarily of course ;).
4512 See comment in transport_private.h:71
4513 - do not include all the _interface headers in private but in the files
4514 which really need them (to cut the compilation time when they are
4517 2004-07-26 Martin Quinson
4518 Version 0.3 (protocol not changed; ABI changed)
4519 - Major overhault of the datadesc interface to simplify it:
4520 - shorted the function names:
4521 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4522 - add a trivial way to push/pop integers into the cbps without malloc.
4523 This allows to make really generic sub_type description, which simply
4524 pop their size of the stack.
4525 - add a function gras_datadesc_ref_pop_arr() which does what users want
4526 most of the time: Declare a dynamic array (which pops its size of the
4527 stack) and declare a reference to it. Poor name, but anyway.
4528 - kill the post-send callback, add a post-receive one
4530 2004-07-23 Martin Quinson
4531 Version 0.2 (protocol changed; ABI changed)
4532 - add some testing for cpbs in the test cases, and fix some more bugs.
4533 This invalidate again the little64 data file, since I cannot regenerate
4535 - remove an awfull optimization in the logging stuff, allowing me to:
4536 - understand it again
4537 - learn gcc how to check that the argument match the provided format
4538 - fix all errors revealed by gcc after that
4539 - internal keys of dict are not \0 terminated. Deal with it properly in
4540 loggings instead of segfaulting when the user want to see the logs :-/
4542 2004-07-22 Martin Quinson
4543 - Fix some stupid bug preventing cbps (callback postit) from working
4545 2004-07-21 Martin Quinson
4546 - Some documentation cleanups
4547 - remove the useless last argument of msgtype_declare
4548 - rename the Virtu functions to fit into the 'os' namespace
4549 - move headers src/include -> src/include/gras/ and stop fooling with
4550 gras -> . symbolic link
4551 - make distcheck is now successful
4553 2004-07-19 Martin Quinson
4555 - Build shared library also
4556 - Install html doc to the right location
4557 - stop removing maintainer files in make clean
4558 - build tests only on make check
4560 2004-07-13 Martin Quinson
4562 - No major issue in previous version => change versioning schema
4563 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4564 corresponding dataset.
4566 2004-07-11 Martin Quinson
4568 - Get it working with any kind of structure (we can compute the padding
4569 bytes remotely for all the architectures I have access to)
4570 - Implement the structure parsing macro (still not quite robust/complete)
4571 - Improvement to the remote testing toysuite
4573 2004-07-10 Martin Quinson
4574 [autoconf mechanism]
4575 - get ride of a bunch of deprecated macros
4576 - actually run the test for two-compliment, not only compile it :-/
4577 - test whether the structures get packed (and bail out if yes. Damn.
4578 Alignment is a serious matter)
4579 - test whether the structures get compacted (but respecting the alignment
4580 constraints of each types)
4581 - test whether the array fields of structures can straddle alignment boundaries
4583 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4584 (took me 2 hours to find that bug, looking at the wrong place)
4585 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4586 before sending/receiving and are used to compute the offsets of fields
4587 - Given that padding size depend even on compiler options, keep track of
4588 alignment and aligned_size only for the current architecture. Not a big
4589 deal since we send structure fields one after the other (seems
4591 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4592 program, let it work on linux/gcc/little32. portability todo.
4594 2004-07-08 Martin Quinson
4595 - import and improve remote compilation support from FAST
4596 - make sure make check works on half a dozen of machines out there
4598 2004-07-07 Martin Quinson
4599 Let's say it's version 0.0.3 ;)
4600 - Implement conversions (yuhu!)
4601 - Let it work on solaris (beside conversion, of course)
4602 - Stupid me, using rand() to generate the conversion datatests in not wise.
4604 2004-07-06 Martin Quinson
4605 - Let make dist work, since I'm gonna need it to compile on remote hosts
4606 - Let Tests/datadesc_usage write the architecture on which the file was
4607 generated as first byte.
4608 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4609 architecture descriptions.
4610 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4611 on those architectures.
4612 - Optimization: send/recv array of scalar in one shoot
4614 2004-07-05 Martin Quinson
4615 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4617 - Plug a whole bunch of memleaks
4618 - each process now have to call gras_{init,exit}. One day, their log
4619 settings will be separated
4620 - Continue the code factorisation between SG, RL and common in Transport.
4622 2004-07-04 Martin Quinson
4624 - Redistribution between SG and RL.
4625 We wanna have to accept in SG, so move accepted related parts of RL in
4626 the common part. (more precisely, the dynar of all known sockets is no
4627 more a static in transport.c, but part of the process_data)
4629 [gras_stub_generator]
4630 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4632 2004-07-03 Martin Quinson
4633 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4634 them all up in one shot)
4636 - Fix the ugly case of reference to dynamic array.
4637 - New (semi-public) function gras_datadesc_size to allow the messaging
4638 layer to malloc the needed space for the buffer.
4640 - gras_socket_close now expect the socket to close (and not its address to
4641 put NULL in it after it). This is because the socket passed to handlers
4642 is one of their argument (=> not writable).
4644 - propagate the interface cleanup from last week in datadesc, ie remove a
4645 superfluous level of indirection. User pass adress of variable
4646 containing data (both when sending and receiving), and not of a variable
4647 being a pointer to the data. Let's say that I like it better ;)
4648 The price for that is constructs like "int msg=*(int*)payload" in
4649 handlers, but it's a fine price, IMHO.
4651 - Let it work in RL (yuhu)
4653 2004-06-21 Martin Quinson
4655 - porting SG plugin and SG select to new standards (works almost).
4656 - plug memleaks and fix bugs around.
4659 - cleanup the prototype of data recv and force users to specify when they
4660 want to handle references to objects. Test case working even for cycles.
4661 - plug memleaks. Valgrind is perfectly ok with this.
4663 2004-06-12 Martin Quinson
4665 - cleanup the separation between plugin and main code in plugin creation
4667 2004-06-11 Martin Quinson
4669 - Reput hook for raw sockets, needed for BW experiments
4670 - kill a few lines of dead code
4671 [Data description] Interface cleanup
4672 - gras_datadesc_by_name returns the searched type or NULL.
4673 That way, no variable is needed to use a type desc once, which makes
4675 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4676 two parameters were strings (field name, type name), leading to
4678 [Dicos] Interface cleanup
4679 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4680 This is consistant with the dynar API.
4682 2004-04-21 Martin Quinson
4684 - Porting to new standards.
4686 - interface cleanup.
4687 There is no bag anymore, no need to take extra provision to mask the
4688 pointers behind "ID".
4689 Better splitup of functions between files create/exchange/convert.
4690 This is still a bit artificial since convert and receive are so
4691 interleaved, but anyway.
4693 - add a queued message list to procdata (the ones not matching criteria
4695 - factorize some more code between SG and RL wrt procdata
4697 - use gras_exit in example to track memleaks
4698 - get rid of gs_example now that GS is properly integrated into gras
4699 - update run_test to integrate the latest tests (datadesc)
4701 - rename WARNINGn macros to WARNn since it prooved error-prone
4703 2004-04-19 Martin Quinson
4705 - register init/exit functions within gras module mechanism
4706 - send/receive function.
4707 Convertion is not implemented, but short-cutted if not needed.
4708 struct/array elements are sent one by one (instead of block-wise), but
4709 nobody really cares (yet). Get a prototype before optimizing.
4710 - tests (using a file socket) for DD send/receive on:
4711 - base types: int, float
4712 - array: fixed size, string (ie ref to dynamic string)
4713 - structure: homogeneous, heterogeneous
4714 - chained list, graph with cycle
4715 Believe it or not, valgrind is not too unhappy with the results. The
4716 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4720 - Bugfix when using a filename explicitly (instead of '-')
4722 2004-04-09 Martin Quinson
4724 - factorize more code between RL and SG in socket creation
4725 - Complete the implementation and tests of:
4727 o file (only in RL, and mainly for debugging)
4729 I lost 3 days to design a portable address resolver, and then decided
4730 that the prototype mainly have to run on my box.
4731 Addressing portability too early may be like optimizing too early :-/
4733 - use gras_init in the Tests instead of the crappy parse_log_opt
4734 (the latter function is removed)
4735 [Conditional execution]
4736 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4737 [Code reorganisation]
4738 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4739 Build examples against the RL library, since there is no way to disable
4740 its creation for now.
4742 For information, the beginning of coding on GRAS was back in june
4743 2003. I guess that every line has been rewritten at least twice since