1 SimGrid (3.31.1) NOT RELEASED YET (v3.32 expected June 21. 2022, 09:13 UTC)
4 - Added the following bindings / examples:
5 - Comm (now 100% covers the C++ interface):
6 - Comm.dst_data_size, Comm.mailbox, Comm.sender, Comm.start_time, Comm.finish_time
7 - Comm.state_str [examples: examples/python/comm-failure/, examples/python/comm-host2host/]
8 - Comm.remaining [examples: examples/python/comm-host2host/, examples/python/comm-suspend/]
9 - Comm.set_payload_size [example: examples/python/comm-host2host/]
10 - Comm.set_rate [example: examples/python/comm-throttling/]
11 - Comm.sendto, Comm.sendto_init, Comm.sendto_async [example: examples/python/comm-host2host/]
12 - Comm.start, Comm.suspend, Comm.resume [example: examples/python/comm-host2host/]
13 - Comm.test_any [example: examples/python/comm-testany/]
14 - Comm.wait_until [example: examples/python/comm-waituntil/]
16 - Engine.host_by_name [example: examples/python/comm-host2host/]
17 - Engine.mailbox_by_name_or_create [example: examples/python/comm-pingpong/]
18 - Mailbox: Mailbox.ready [example: examples/python/comm-ready/]
20 Platform description & visualization:
21 - More robust sanity checks for platforms, to reject unallowed topologies with
22 a proper error message.
23 - One new C++ platform example, supernode.cpp. A Python script (supernode.py)
24 demonstrates how we can generate a nice graphical representation of the
28 - Enhancements to the graphicator tool:
29 - Allow to dump the platform topology as a CSV file representing the edges
30 of the graph (in addition to the DOT format).
31 - Fix graphicator for "cluster" topologies (e.g. fat-tree, dragonfly).
33 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
34 (FG: issues on Framagit; GH: issues on GitHub)
35 - FG#105: "Variable penalty should not be negative!" with in-flight messages and bandwidth profiles
36 - FG#109: Application time reported by --cfg=smpi/display-timing:yes is wrong
37 - FG!109: Trigger new engine solve upon host events such as host on/off
38 - FG#110: Wait_any does not trigger new model solve when host events occur
40 ----------------------------------------------------------------------------
43 - kill signal Comm::on_completion that was not working anyway.
44 - Expose signals Activity::on_suspend and Activity::on_resume
46 SimGrid (3.31) March 22. 2022.
48 The ненасильство release. We stand against war.
50 Against the agression by a sick system that forces peoples to take arms against each other.
53 - Rework the internals, for simpler and modern code. This shall unlock many future improvements.
54 - You can now define plugins onto the DFS explorer (previously called SafetyChecker), using the
55 declared signals. See CommunicationDeterminism for an example.
56 - Support mutex, semaphore and barrier in DPOR reduction
57 - Seems to work on Arm64 architectures too.
58 - Display a nice error message when ptrace is not usable.
59 - New test suite, imported from the MPI Bugs Initiative (MBI). Not all MBI generators are integrated yet.
60 - Remove the ISP test suite: it's not free software, and it's superseeded by MBI.
63 - fix for FG#100 by ensuring small asynchronous messages never overtake larger
64 ones, conforming to the standard.
65 - replay: fix waitall behaviour to avoid forgetting requests and leaking
67 - tracing: ensure that we dump the TI traces continuously during execution and
68 not just at the end, reducing memory cost and performance hit.
69 - Update OpenMPI collectives selection logic to match current one (4.1.2)
70 - Add a coherence check for collective operation order and root/MPI_Op
71 coherence. Potentially costly so not activated unless smpi:pendantic is set
75 - New signal: Engine::on_simulation_start_cb()
76 - Introduce a new execution mode with this_actor::thread_execute(). This simulate
77 the execution of a certain amount of flops by multiple threads ran by a host. Each
78 thread executes the same number of flops, given as argument. An example of this new
79 function can be found in examples/cpp/exec-threads.
80 - Reimplementation of barriers natively.
81 Previously, they were implemented on top of s4u::Mutex and s4u::ConditionVariable.
82 The new version should be faster (and can be used in the model-checker).
83 - Actor::get_restart_count(): Returns the number of reboots that this actor did.
86 - MSG_barrier_destroy now expects a non-const msg_barrier parameter.
88 New plugin: the Chaos Monkey (killing actors at any time)
89 - Along with the new simgrid-monkey script, it tests whether your simulation
90 resists resource failures at any possible timestamp in your simulation.
91 - It is mostly intended to test the simgrid core in extreme conditions,
92 but some users may find it interesting too.
95 - New solver for parallel task: BMF.
96 - More realistic sharing of heterogeneous resources compared to the fair
97 bottleneck solver used by ptask_L07.
98 - Implement the BMF (Bottleneck max fairness) fairness.
99 - Improved resource sharing for parallel tasks with sub-flows (parallel
100 communications between same source and destination inside the ptask).
102 - "--cfg=host/model:ptask_L07 --cfg=host/solver:bmf": enable the ptask
103 model with BMF solver.
104 - "--cfg=bmf/max-iterations: <N>": maximum number of iterations performed
105 by BMF solver (default: 1000).
106 - "--cfg=bmf/precision: <N>": numerical precision used when computing
107 resource sharing (default: 1e-12).
108 - This model requires Eigen3 library. Make sure Eigen3 is installed to use BMF.
111 - Modifications of the Profile mechanism, with some impact on users
112 - Addition of a new (S4U) method to init profiles from generic functions to improve versatility
113 - Fix initial behaviour of state_profiles
114 - Modify periodicity to behave like a period, and not like a loop delay
117 - Drop xbt_dynar_shrink().
120 - Made the following bindings static (previously member functions):
121 - Actor: Actor.kill_all(), Actor.by_pid()
122 - Host: Host.by_name(), Host.current(), Host.on_creation_cb()
123 - Mailbox: Mailbox.by_name()
124 - Added the following bindings:
125 - this_actor.warning()
126 - Mailbox.put_init() [example: examples/python/comm-waitallfor/]
127 - Comm.detach() [example: examples/python/comm-waitallfor/]
128 - Comm.wait_for() [example: examples/python/comm-waitfor/]
129 - Comm.wait_any_for()
130 - Comm.wait_all_for() [example: examples/python/comm-waitallfor/]
131 - Mutex [example: examples/python/synchro-mutex/]
132 - Barrier [example: examples/python/synchro-barrier/]
133 - Semaphore [example: examples/python/synchro-semaphore/]
136 - Remove target "make uninstall" which was incomplete and no longer maintained.
138 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
139 (FG: issues on Framagit; GH: issues on GitHub)
140 - FG#57: Mc SimGrid should test whether ptrace is usable
141 - FG#87: Smpi scripts fail with spaces in paths
142 - FG#100: [SMPI] Order of the message matching is not guaranteed
143 - FG#101: LGPL 2.1 is deprecated license
144 - FG#104: "make uninstall" not up-to-date
145 - GH#151: Missing mutexes for DPOR.
147 ----------------------------------------------------------------------------
149 SimGrid (3.30) January 30. 2022.
151 The Sunday Bloody Sunday release.
153 Main user-visible changes:
154 - The SimDag API for the simulation of the scheduling of Directed Acyclic
155 Graphs has been dropped. It was marked as deprecated for a couple of years.
156 We finally complete the implementation of what has been called SimDag++
157 internally, i.e., porting the different features of SimDag on top of S4U.
158 The new way to simulate the execution of dependent activities directly by
159 maestro (without any other actor) is details in the examples/cpp/dag-* series
161 - The removal of SimDag led us to also remove the export to Jedule files that
162 was tightly coupled to SimDag. The instrumentation of DAG simulation is still
163 possible through the regular instrumentation API based on the Paje format.
164 - We also dropped the old and clumsy Lua bindings to create platforms in a
165 programmatic way. It can be done in C++ in a much cleaner way now, which
166 motivates this suppression.
169 - Introduce on_X_cb() functions for all signals, to attach a new
170 callback to the signal X. The signal variables are now hidden and
171 only these functions should be used.
172 Rationale: this enables the usual deprecation scheme where functions
173 remain for 4 releases if we need to modify the signals, while the
174 current code with the signal variables directly visible prevents any
176 - New function: Engine::run_until(date), to split the simulation.
177 - New signal: Activity::on_veto, to detect when an activity fails to start.
178 - Signal change: Comm::on_start(Comm&, bool) has been replaced by
179 Comm::on_send and Comm::recv. These two signals respectively correspond to
180 when the sending or receiving side of a Comm is ready. They are raised at
181 the same locations as the former Comm::on_start signal.
182 - New function: Engine::track_vetoed_activities() to interrupt run()
183 when an activity fails to start, and to keep track of such activities.
184 Please see the corresponding example for more info.
185 - New functions: s4u::Comm::{sendto_init, set_source, set_destination} to enable
186 the use of vetoers with direct host-to-host communications. Both source and
187 destination have to set for a comm to start. Each call to these setters check
188 if all vetoes are satisfied. When it is the case, the comm starts. A use case of
189 these functions is given in examples/cpp/dag-scheduling.
190 - New functions: {Exec, Io}::update_priority allow you to modify the priority of
191 these kinds of activities during their execution. Behavior is detailed in
192 examples/cpp/io-priority/.
195 - Dynamic costs for MPI operations: New API to allow users to dynamically
196 change injected costs for MPI_Recv, MPI_Send and MPI_Isend operations.
197 Alternative for smpi/or, smpi/os and smpi/ois configuration options.
198 - Fix some issues with the replay mechanism.
201 - Function xbt::Extendable::get_data() is now templated with the type of the
202 pointee. Untyped function is deprecated. Use get_data<void>() if you still
203 want to retrieve void*.
206 - New section: "SimGrid MPI calibration of a Grid5000 cluster"
207 presenting how to properly calibrate MPI communications in SimGrid.
208 - Complete and reword the platform section, which is now completed.
211 - Thread contexts are used by default with Python bindings. Other kinds of
212 contexts revealed unstable, specially starting with pybind11 v2.8.0.
214 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
215 (FG: issues on Framagit; GH: issues on GitHub)
216 - FG#95: Wrong computation time for multicore execution after pstate change
217 - FG#97: Wrong computation time for ptask+multicore+pstates
218 - FG#98: SMPI offline simulation is inconsistent with the online simulation
219 (deadlocks / message truncation)
220 - FG#99: Weird segfault when not sealing an host
222 ----------------------------------------------------------------------------
224 SimGrid (3.29) October 7. 2021
226 The "Ask a stupid question" release.
228 We wish that every user ask one question about SimGrid to celebrate.
229 On Mattermost, Stack Overflow or using the issues tracker.
231 New modeling features:
232 - Non-linear resource sharing, modeling resources whose performance heavily degrades with contention:
233 - The total capacity may be updated dynamically through a callback
234 and depends mainly on the number of concurrent flows.
235 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
237 - Dynamic factors: model variability in the speed of activities
238 - Each action can now have a factor that affects its progression.
239 This multiplicative factor is applied when updating the amount of work
240 remaining, thereby an activity with factor=0.5 only uses half of the
241 instantaneous power/bandwidth it is allocated and will appear twice
242 slower than what it actually consumes.
243 - This can be used to model a overhead (e.g., there is a 20 bytes
244 header in a 480 bytes TCP packet so the factor 0.9583) but the novelty
245 is this factor can now easily be adjusted depending on activity's and
246 resources characteristics.
247 - This existed for network (e.g., the effective bandwidth depends
248 on the message in SMPI piecewise-linear network model) but it is now
249 more general (the factor may depend on the source and destination and
250 thus account to different behaviors for intra-node communications and
251 extra-node communications) and is available for CPUs (e.g., if you
252 want to model an affinity as in the "Unrelated Machines" problem in
253 scheduling) and disks (e.g., if you want to model a stochastic
255 - For that, resources can be provided with a callback that computes
256 the activity factor when creating the action.
257 - Example: examples/cpp/exec-cpu-factors
258 - The same mechanism is also available for the latency, which
259 allows to easily introduce complex variability patterns.
262 - Added support to programmatic platform creation in Python.
263 Example: examples/python/clusters-multicpu
266 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
267 This can only be set through the API, not through XML files.
271 - Multiple fixes to ensure reproducibility of tracing
272 - scan/excan can now be replayed
273 - wait action now uses ranks and not pid, as the other ones.
274 - smpi/init and smpi/finalization-barrier are now valid for replays.
275 - exit() is now intercepted by SMPI to avoid premature shutdown of
276 simulation. First non 0 return codes is returned as simulation return
280 * New section "Release Notes" documenting recent and current developments.
281 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
282 * Improvements in API Reference for C++ and Python interfaces.
285 - Make wifi creation compatible with ns-3 version 3.34 too.
287 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
288 (FG: issues on Framagit; GH: issues on GitHub)
289 - FG#77: Search feature of doc is broken (update sphinx theme version)
290 - FG#78: Multiple fixes for SMPI replay:
291 - TI tracing of allotallv/w was outputting wrong values
292 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
294 ----------------------------------------------------------------------------
296 SimGrid (3.28) July 14. 2021
298 The Victoriadagarna Release.
301 - C++ platform interface: Users can now describe their platform directly in C++.
302 This provides greatly flexibility and performance improvement for complex
303 platforms. Main features:
304 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
305 "zones", instead of single hosts. This feature enables the description
306 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
307 - StarZone: new zone with a Star-like topology. The routes are defined
308 as a set of links used to communicate from node to everybody (node<->ALL).
309 - Split-Duplex links: auxiliary method to create split-duplex links in
310 the platform, easing its utilisation. It automatically creates both UP
311 and DOWN links (similarly as done in XML).
312 - Please refer to the documentation and the examples included:
313 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
314 - New plugin: Producer-Consumer with monitor. Just requires to include the
315 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
316 associated example (examples/cpp/plugin-prodcons).
319 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
320 s4u::Io::wait_any(), s4u::Io::wait_any_for().
321 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
322 parameter by reference, instead of a pointer.
323 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
324 Explicitly cancel the activity to get back to previous behavior.
325 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
328 - The default SMPI compiler flags are no more taken from the environment.
329 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
330 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
332 --cfg=smpi/finalization-barrier: which can be used to add
333 a barrier inside MPI_Finalize. This can help for some codes which cleanup
334 data attached to a process, but still used in other SMPI processes.
335 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
336 is active when set to false, to keep going after a small error
337 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
338 which may or may not be problematic in the end.
340 - fix behaviour, as maximum iteration count could be ignored
341 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
342 use sampling when the same kernel is called with a different set of parameters
343 which have an impact on the timing.
344 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
345 It should now work with smpi/auto-shared-malloc-thresh.
346 - Improve error handling and reporting in multiple places
347 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
348 validity checks, truncated messages are now an error, return errors
349 when explicitely deleted handles are reused, ...)
350 - RMA: multiple fixes and stability improvements.
351 - analysis (-analyze flag in smpirun):
352 - SMPI can now report buffer leaks as well as MPI handles leaks,
353 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
354 - if -trace-call-location is used when compiling, SMPI can report
355 origin of leaked handles/buffers
356 - group leaks by type/origin in output message if possible
357 - New implemented MPI calls: MPI_Comm_test_inter
360 - Changed internal implementation of bandwidth factors in network models.
361 Models affected: CM02, LV08 (default), SMPI, IB.
362 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
363 Bandwidth factors are applied to communications to describe that users
364 cannot use 100% of the available bandwidth. For example, the default network model,
365 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
366 behavior was implemented by limiting the bandwidth available in the LMM
367 system for this flow. This may give the false impression that there is
368 bandwidth available for other flows due to its underutilization, especially
369 for the dynamic bandwidth factors used in SMPI models.
370 To avoid this, we have modified the implementation so that each flow uses the
371 maximum physical bandwidth according to the LMM system.
372 However, the actual throughput of the flow seen by the user is defined by
373 the physical bandwidth multiplied by the bandwidth factor.
374 This change impacts on the simulation results for all network models on
375 which we have bandwidth factors configured.
376 *****************************************
377 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
378 *****************************************
379 This change may impact on the timing of your simulation results.
380 Take care when comparing simulations from different SimGrid's
381 versions. Sorry for the inconvenience.
382 - Dynamic network factors: users can configure a callback to define
383 the network factors dynamically. This API is available at
384 simgrid::kernel::resource::NetworkModelIntf.
385 - Users have access to complete information about the current communication
386 to decide which factor to apply. This includes: message size, source and
387 destination hosts, links and zones traversed.
388 - Dynamic factors for both latency and bandwidth.
389 - For more details, see the example in (examples/cpp/network-factors).
390 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
391 deprecated since version 3.24, are no longer supported. Instead, use
392 "wattage_off" and "wattage_per_state".
395 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
398 - New tutorial: Model-checking and formal assessment
399 - New sections: "Demystifying the routing" and "C++ platforms"
400 - Update and improve the part on visualization in MPI and Algo tutorials.
401 - Improve the section on routing: how to define it, how it's used internally
402 - Fix many issues, broken links and missing references in doxygen and Sphinx
405 - Lua platform files are deprecated. Their support will be dropped after v3.31.
408 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
410 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
411 (FG: issues on Framagit; GH: issues on GitHub)
412 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
413 - FG#64: Configuring smpi/IB-penalty-factors
414 - FG#67: Running computation concurrently with MPI_Iallreduce
415 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
416 - FG#71: Segmentation fault on invalid gw_src/gw_dst
417 - GH#322: Issue when an actor kills his host vm
419 ----------------------------------------------------------------------------
421 SimGrid (3.27) March 29. 2021
423 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
425 Main user-visible changes:
426 - All functions marked as deprecated for one year were removed.
427 - Removed all that was related to the "Storage" type in all APIs and bindings.
428 "Disks" have been introduced more than a year ago to replace them for the better.
429 It's time to finish this replacement.
430 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
431 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
432 avoiding "initializer element is not constant" errors with some codes. Please be
433 careful to recompile your MPI codes if switching to this new SimGrid version.
434 - Network model ns-3 now requires version ns-3.28 or higher.
435 - Removed some option aliases that were deprecated since v3.23 (June 2019)
436 - 'verbose-exit' is now 'debug/verbose-exit'
437 - 'simix/breakpoint' is now 'debug/breakpoint'
438 - 'clean-atexit' is now 'debug/clean-atexit'
441 - New functions: Comm::sendto_{init,async} to initiate a communication
442 on between two (possibly remote) hosts.
443 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
444 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
445 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
446 or Mailbox::get_async<void>() if you really want to play with void*.
447 - Unify the interface of Activity::on_{start/activity}
448 - New function: Comm::get_dst_data()
449 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
452 - ptask should now work better with multicore hosts (FG#37).
453 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
454 - Internal cleanups and refactoring around models, to prepare for programmatic
456 - Modernize our tests of the performance models
460 - New section: Configuring the logs from the command line.
461 - New section: Programming API of logs.
462 - New section: Getting realistic results
463 - New section: Examples of SMPI Usage
464 - New part: SimGrid's Internals, with a section on Design goals
465 - Move back to the classical Breathe tool chain, and fix many small glitches.
468 - Handle leak checking: SMPI will now display a message if MPI handles are still
469 present at the end of a simulation. Option smpi/list-leaks:n will display the n
470 first leaked pointers and their types.
471 - smpi/display-allocs option : Display Memory allocations analysis, with hints
472 about the largest allocations. This only works for allocations with malloc/calloc
473 and not compiled with SMPI_NO_OVERRIDE_MALLOC
474 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
475 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
476 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
477 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
478 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
479 - Replay now opens and parses the trace file only once for all ranks when possible.
481 - better handling of MPI_PROC_NULL
482 - MPI_Group_incl was broken in some cases
483 - PAPI tracing now fixed
484 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
485 - SMPI will now fail properly when receive size < sent size, instead of
486 just setting MPI_ERR_TRUNCATE and keep going.
487 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
488 - Remove useless build patches and split build/run patches on the page
489 - Build in parallel when possible
490 - Test on yum-based systems
491 - New nightly tested Proxy-apps:
492 - OpenMPI mpi-test-suite
493 - Coral MCB, LSMS, Qbox, CLOMP
494 - ECP/Mantevo Cloverleaf3D, Ember
495 - Codevault bhtree_mpi
498 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
499 - Start to overhaul the implementation to modernize it (still ongoing).
502 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
503 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
504 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
505 - Deprecate xbt_dynar_to_array().
506 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
508 C binding and interface:
509 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
510 - Constness requirement has been lowered for parameter argv of sg_actor_start()
511 and sg_actor_create() It's now (char* const*). The previous version is still
512 available as sg_actor_start_() and sg_actor_create_().
514 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
515 (FG: issues on Framagit; GH: issues on GitHub)
516 - FG#37: Parallel tasks are limited to 1 core per host
517 - FG#62: Running "smpirun -replay" on large networks
518 - FG!46: Fix a few potential memory leaks in SMPI colls
519 - GH#343: Wrong consumed energy when an exec timeouts
521 ----------------------------------------------------------------------------
523 SimGrid (3.26) December 15. 2020
525 The Release release (the French lockdown was eased today).
527 Important user-visible changes:
528 - SimGrid now requires a compiler with C++14 support.
529 Sibling projects should upgrade their FindSimgrid.cmake
530 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
531 several users had difficulties to understand issues when using high bandwidth or
532 small latency events. The new value was already the default for SMPI and
533 should not cause too much performance hit. It can change some simulation
534 timings if simulation was skipping a lot of small events (which was actually
535 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
536 - On some network models (L07, CM02, LV08), an implicit loopback link is created
537 for local communication. Default bandwidth was 500MBps and latency was 15µs,
538 which was rather bad. Default values have been changed to 10GBps and 0 latency.
539 Options network/loopback-lat and network/loopback-bw have been added to adapt
540 these values if needed. Network and routing models already providing separate
541 loopbacks (clusters and such) are not impacted by this change.
542 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
543 metadata that are added to each MPI message by MPI libraries (tag,
544 src/dest, and communicator information). This may change simulation times
545 for very small message sizes, but should be more realistic.
548 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
549 - New plugins: wifi_energy and link_load. See the associated examples.
550 - New examples: wifi networks, and communication suspend.
553 - LTO is now enabled for Intel/clang compilers.
554 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
555 Setting it to "auto" will use all available cores, while setting it to n will
556 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
557 - Remove obsolete runtime option 'contexts/parallel-threshold'.
558 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
559 has been kept for compatibility.
560 - Finally remove obsolete snake_case() aliases for runtime options.
561 - Further improve the documentation.
564 - the on_exit() of each actor is also executed when the simulation deadlocks.
565 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
566 An example is provided for s4u::Comm but it should work with Exec and Io.
569 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
570 https://framagit.org/simgrid/SMPI-proxy-apps
571 - MPI/IO: supports names starting with "./"
572 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
574 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
575 building shared libraries, this may cause issues, so environment variable
576 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
577 - most temporary files should now be created in /tmp dir (or equivalent).
578 If this one does not allow execution of code (noexec flag), this may cause issues.
579 Please use another tmp directory (using TMPDIR or equivalent system variable)
583 - A large refactoring is ongoing, but this should be transparent for users so far.
586 - Drop xbt_str_split().
588 C binding and interface:
589 - The return type of the sg_actor_on_exit() callbacks is now 'void'
590 (int value was previously ignored)
591 - Many C functions were renamed to follow a common naming style
592 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
593 names are kept until version 3.30.
594 - Many MSG tests were converted to the new S4U's interface in C, that
595 was extended for that.
597 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
598 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
599 - FG#41: Add sg_actor_create C interface
600 - FG#43: xbt::random needs some care
601 - FG#48: The Impossible Did Happen (yet again)
602 - FG#50: Suspending an actor executed at the current timestamp fails
603 - FG#52: Zero-seconds timeout : "That's in the past already"
604 - FG#53: Crash while using ns-3 network model
605 - FG#54: How to suspend a comm?
606 - FG!22: Stochastic Profiles
607 - FG!24: Documentation and fix for xbt/random
608 - FG!35: Add a modeling hint for parallel links in doc
609 - FG!36: [xbt/random] Read/Write the state of the RNG
610 - FG!42: Link cumulated load plugin
611 - FG!44: Test condition variable wait_until+notify_all
612 - GF#18137: Allow different stack sizes?
613 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
614 - GH#139: Allow pthread creation in SMPI
615 - GH#336: Packet-level simulation using SMPI?
616 - GH#345: Error in the LMM while migrating a VM
617 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
618 - GH#352: pip install / python setup.py install fail to find pybind11
619 - GH!337: Fix link_energy plugin for wifi platforms
620 - GH!339: Add Mailbox set_receiver method to python binding
621 - GH!344: Cast hugepages macros parameters to int64
623 ----------------------------------------------------------------------------
625 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
627 The Palindrom Day Release.
629 Important user-visible changes:
630 - Improve the Python usability (stability and documentation).
631 - A nasty synchronization bug was ironed out, see also below.
632 - Python's doc was organized as a separate tree, now integrated with C++.
633 - C bindings of S4U were not part of the doc.
634 - The C++ doc was also improved as methods are now split by theme.
635 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
636 - This is mandatory to use the Java bindings.
637 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
638 - SimDAG++: Automatic dependencies on S4U activities (experimental)
639 - Some features are already implemented but not all of them
640 - Cannot block an activity until it's scheduled on a resource
641 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
642 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
645 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
646 - Drop support for Viva/Triva (old visualization tools), hence removed
647 TRACE_get_node_types() and TRACE_get_edge_types() functions.
650 - Require pybind11 version 2.4.
651 - Greatly improve locking during context switches with Python.
652 - Declaring Python callbacks to this_actor::on_exit is now working.
653 - No more tests are failing currently, even on win32.
656 - Actor: Merge signals on_migration_start/end into on_host_change
657 - Actor: Rename migrate() into set_host()
658 - Disk: Allow users to get the read and write nominal bandwidth values
659 - Exec: Implement wait_for(timeout)
660 - Io: Implement wait_for(timeout)
663 - Parse errors now raise a simgrid::ParseError that you may want to catch.
666 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
667 been renamed to fix a spelling error. As usual, the previous definitions are
668 kept with a deprecation warning. However, to avoid runtime errors, it is now
669 forbidden to override these deprecated methods in a derived class. Please use
670 the new names immediately if you need to override them.
673 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
674 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
675 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
676 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
677 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
678 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
679 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
680 - New module: random, providing classical random numbers generators.
683 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
684 allocations' size above which they are considered "shared" by default (as if
685 they were performed through SMPI_SHARED_MALLOC macros).
686 Default = 0 = disabled feature.
687 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
688 This can cause some troubles if codes are already overriding these. If this
689 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
690 help, but will make this feature unusable.
691 - Input check is now performed much more consistently and outputs precise
692 warnings for each error found
693 - New supported calls : MPI_File_set_view, MPI_File_get_view
694 - MPI I/O now only supports disks, as storage is discontinued.
695 - MPI I/O allows opening files without specifying absolute path
696 - Note : SMPI configuration options are now initialized by smpi_init_options()
697 call, which should be called if SMPI is not being used through smpirun.
699 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
700 - FG#9: Python bindings crashing
701 - FG#39: Missing s4u::Comm::wait_any_for example
702 - FG#42: Add support for ThreadSanitizer (TSan)
703 - FG!19: Removing RngStream
704 - FG!20: A module for RNG calls
705 - FG!21: Choice between ad-hoc and standard distributions implementations
707 - FG!25: Fix link in Mutex doc
708 - FG!26: Fix links in SMPI interface doc
709 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
710 - GH#305: Unscheduled tasks are still excuted
711 - GH#313: smpirun: manual outdated w.r.t. --help
712 - GH#321: [S4U] Get task remaining work ratio
713 - GH#323: Crash when an actor turn off his physical host
714 - GH#335: Missing links on dragonfly example svg
715 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
716 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
718 ----------------------------------------------------------------------------
720 SimGrid (3.24) October 9. 2019
722 The Clean Disk Release.
725 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
726 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
727 thus becomes the default resource to execute I/O operations. Examples
728 and the file system plugin have been modified to use disks instead of
729 storage elements. Storage can still be used but is doomed to disappear.
730 Users are thus adviced to update their simulators accordingly.
731 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
732 for consistency with pthread_barrier_wait()
733 - Host::get_englobing_zone() returns the englobing netzone
734 - Actor::on_destruction is now called in the destructor
735 Actor::on_termination new signal called when the actor terminates
737 - Global signals are now part of the Engine:
738 - on_platform_creation: after config settings, before the XML parsing
739 - on_platform_created: right after the XML parsing
740 - on_time_advance: each time the clock advances
741 - on_simulation_end: after simulation, before cleanups
742 - on_deadlock: as the name implies.
744 - sg_{actor,host,link}_{data,data_set}() now all exist.
745 Use them to attach user data to the object and retrieve it.
748 - Introduce an experimental Wifi model. It sounds reasonable
749 according to the state of the art, but it still has to be properly
750 validated, at least against ns-3 if not against reality.
751 - Improved the usability of ns-3. Several bugs were ironed out.
752 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
753 It is now expressed as 'idle:epsilon:allCores' to properly model the
754 consumption of non-whole tasks on mono-core hosts. More info in the doc.
757 - convert a new set of functions to the S4U C interface and move the old MSG
758 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
761 - Fortran bindings for DVFS have been removed.
762 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
763 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
764 - SMPI can now be selected by cmake's find_module(MPI) with
765 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
766 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
767 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
768 - trace-call-location can be used with TI traces, and replayed, and is compatible
769 with smpi/comp-adjustment-file.
770 - sleep events are now correctly traced/replayed.
771 - Default for trace-call-location is now to use file names and not full paths.
772 To revert to previous behavior (in case of collision of filenames), option
773 "smpi/trace-call-use-absolute-path" can be set to yes.
776 - Use the included xxHash as an hash implem when C++14 is usable.
777 - Option model-checker/hash was removed. This is always activated now.
778 - New option smpi/buffering controls the MPI buffering in MC mode.
779 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
780 This is useful to check for MPI compliance.
783 - New section on plugins: how to define a new one, and existing ones.
786 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
787 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
790 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
791 and <mount> tags. The rationale is finish to decouple what should be handle
792 by SimGrid kernel (i.e., resources and activities) from plugins built in
793 user space (e.g., the file system plugin). This new tag currently accepts
794 only three attributes (id, read_bw, and write_bw). All the other information
795 that was declared with the storage related tags now has to be expressed as
796 properties. An example of platform using this new tag is available at
797 examples/platforms/hosts_with_disks.xml
800 - 'expect signal' can now accept more than one potential signal.
802 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
803 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
804 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
805 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
806 - FG#31: per-actor data
807 - FG#34: SG_BARRIER_SERIAL_THREAD?
808 - FG#35: model-checker does not like buster-produced binaries
809 - FG!13: MC: complete workaround in the error msg seen on modern systems
810 - FG!15: execute_flops now logs compute
811 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
812 - FG!17: ns-3: unblock the right number of communications + others issues
813 - FG!18: Improving the performance of the ns-3 bindings
814 - GH#189: Energy consumption of parallel tasks
815 - GH#219: Error in the throughput of TCP transfer
816 - GH!330: Minor polishing of English: Configuring_Simgrid
817 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
818 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
819 - GH!334: Rework energy plugin, again
821 ----------------------------------------------------------------------------
823 SimGrid (3.23.2) July 8. 2019
826 - Nicer introduction page.
827 - Migrate the "Deploy your application" page to the new doc.
828 - Move Java as a subtree of MSG.
831 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
834 - Drop xbt_dynar_sort_strings().
837 - Really fix FG#26: Turning off a link should raise NetworkFailureException
838 - FG#27: Wrong exception thrown to wait_any when link is turned off
839 - GH#328: Java: Canceling multiple tasks in a single vm/host
841 ----------------------------------------------------------------------------
843 SimGrid (3.23) June 25. 2019
845 The Exotic Solstice Release.
848 - SunOS and Haiku OS support. Because exotic platforms are fun.
849 - Stop setting random seed with srand48() at initialization.
850 - Use addr2line as a fallback for stacktraces when backtrace is not available.
851 - Build option -Denable_documentation is now OFF by default.
852 - Network model 'NS3' was renamed into 'ns-3'.
855 - Simgrid can now hopefully be installed with pip.
858 - wait_any can now be used for asynchronous executions too.
861 - New log appenders: stdout and stderr. Use stdout for xbt_help.
862 - Drop xbt_dict_dump.
865 - 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
866 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
867 - MPI interface is now const correct for input parameters
868 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
869 - Fortran bindings for async collectives.
870 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
873 - Remove option 'model-check/record': Paths are recorded in any cases now.
874 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
875 always sparse. That's an immense gain in memory, and can even be
876 faster because of cache effects. So there is no need to clutter the
877 code to allow the user to go for the unefficient mode.
880 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
881 models proved to be more accurate than these old experiments.
883 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
884 - FG#1: Broken link in error messages
885 - FG#2: missing installation documentation
886 - FG#3: missing documentation in smpirun
887 - FG#6: Python bindings not available on PyPI
888 - FG#7: simple cmake call requires doxygen
889 - FG#8: make python bindings an optional dependency
890 - FG#10: Can not use MSG_process_set_data from SMPI any more
891 - FG#11: Auto-restart actors forget their on_exit behavior
892 - FG#12: -Denable_lto=OFF doesn't disable LTO
893 - FG#13: Installs unstripped file 'bin/graphicator'
894 - FG#14: Installs the empty directory 'doc/simgrid/html'
895 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
896 - FG#17: Dead link in doc (pls_ns3)
897 - FG#20: 'tesh --help' should return 0
898 - FG#21: Documentation link on http://simgrid.org/ broken
899 - FG#22: Debian installation instruction are broken
900 - FG#26: Turning off a link should raise NetworkFailureException exceptions
901 - GH#133: Java: a process can run on a VM even if its host is off
902 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
903 - GH#326: Valgrind-detected error for join() when energy plugin is activated
905 ----------------------------------------------------------------------------
907 SimGrid (3.22) April 1. 2019
909 The Easter Christmas Release (this one is somewhat late).
912 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
913 - Not complete yet: asynchronous activities (amongst others) are still missing
914 - Still ongoing: the interface may change in the future. We need more
915 testers! Please report any glitches.
916 - No new project using Java should start now. Please switch to Python.
919 - Some of the internal cleanups may lead to speed improvements:
920 - The hard limitation on the amount of simulated actors with
921 Java+Mac was removed. Now, the available memory is the only limit.
922 - Our refcounting was tidyied, leading to 10% speedups in some cases.
923 - We are still working on making our code robust to the actor kills
924 and hosts' churn. Things are improving, but it's not perfect yet.
925 - Replaced our own code to display a backtrace (that was forking addr2line)
926 with the Boost.Stacktrace library.
927 You won't see your backtraces without this optional dependency.
928 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
929 - Stop setting random seed with srand() at initialization.
932 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
933 XML file version remains 4.2 since old files are still compatible.
936 - Process termination which was broken at version 3.21 has been repaired.
937 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
938 - Hide the examples into examples/deprecated. New users should use Python.
941 - Drop MSG_process_create_from_stdfunc() from the API.
942 This C++ function was a pimple in the C API, made necessary at some
943 point by the Java bindings. This is fixed now.
944 - Hide the examples into examples/deprecated. New users should use S4U.
945 - MSG_process_create and MSG_process_attach now crash if the host used
946 to run this process is off.
947 - Fix the protype of MSG_process_on_exit()
948 Now use (int,void*) callbacks instead of (void*,void*) ones.
949 The implementation was ways too messy to actually work, I guess.
952 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
953 too often and provide more accurate timings.
954 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
955 - MPI_Alltoallw support
956 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
957 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
958 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
959 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
961 - Basic implementation of generalized requests (SMPI doesn't
962 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
965 - Drop sg_cmdline. Please use xbt_cmdline instead.
966 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
967 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
968 OsSemaphore is implemented in a portable way with C++11 threads.
969 This should allow much more threads to be created at the same time,
970 allowing Mac and Java users to simulate many more actors.
971 - Implement the 'thread' factory with std::thread instead of xbt ones.
972 It is not possible to set the stack size with threads anymore, but
973 -fsplit-stack is the way to go nowadays when using threads.
974 - Drop the xbt_os_thread_t module (now unused)
975 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
976 - Drop xbt_str_join_array().
977 - Drop cunit, use Catch2 instead.
980 - Many cleanups in the kernel::activity namespace. This was long
981 overdue, and shall open the path to many future endeavors.
984 - #132: Java: a process can not shut down its own host
985 - #220: S4U: detached send causes exception if sender terminates soon after sending
986 - #261: Document the parameters of parallel execution's constructor
987 - #300: [s4u] BarrierPtr is missing
988 - #314: SMPI args internal cleanup
989 - #316: Fix a bug related to the CPU utilization of multi-core VM
990 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
991 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
992 - #325: Turning off a host has different behavior on sleeping actors and computing actors
994 ----------------------------------------------------------------------------
996 SimGrid (3.21) October 3. 2018
998 The Restarting Documentation (TRD) Release.
1001 - Convert the user manual to Sphinx for improved usability.
1002 Unstable version now hosted on https://simgrid.frama.io/simgrid
1003 Many glitches remain; stay tuned.
1004 - Restrict the Doxygen reference API to the installed header files.
1005 (documenting all internals this way is overkill + doxygen scales badly)
1006 - New tutorial on S4U (highly inspired from the old MSG one, but with
1007 a git repository to fork as a starting point, and a docker image)
1008 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
1009 - The Developper manual is still to be converted (not compiled until then)
1010 - Some parts are still missing in this conversion (such as the
1011 platform and deployment sections) while others were blindly converted
1012 and would need more love (such as the configuration flags).
1013 Things will certainly further improve in the future releases.
1016 - Fully reimplement auto_restart mechanism that was utterly broken
1017 (fix #22, #131 and #281 that were all very related)
1018 - Implement semaphores (fix #303 and #291)
1019 - When creating an actor from a function and its parameters,
1020 move-only parameters are not allowed anymore, as it would prevent
1021 the actor to be restartable if its parameters are consumed on run.
1022 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
1023 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
1024 IO operations that can be started, waited for, or canceled as a regular
1026 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
1030 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
1031 - New host variable: 'core_count'
1034 - Remove xbt_os_thread_specific features
1035 - Remove portability wrapper to condition variables
1036 - Remove xbt_os_thread_yield()
1039 - MPICH collective selector now mimics MPICH 3.3b
1040 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
1041 - MPI_Init(NULL,NULL) is now allowed
1042 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
1043 - Command-line arguments should now be handled properly in Fortran simulations
1044 - MPI Attributes and related callbacks should work in Fortran
1045 - Apps using C + Fortran should now work
1046 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
1047 - Support MPI_Error_String
1050 - Due to an internal bug, Msg.run() must now be your last line.
1051 We hope to fix it in a future release, and we are sorry for the inconvenience.
1054 - #22: Process autorestart seem to only work with CAS01 cpus
1055 - #56: Feature request: dynamic SMPI replay
1056 - #93: simgrid should not eat --help
1057 - #111: How to change the loopback link characteristics?
1058 - #116: Platform cluster radical
1059 - #117: Platform cluster/host order
1060 - #129: Replay_multiple on steroid
1061 - #131: Java : autostart does not work
1062 - #143: Setting a breakpoint at a given time
1063 - #235: Network energy models should be integrated
1064 - #264: Add ptask L07 resource tracing
1065 - #271: Dynamic and manual replay of SMPI traces
1066 - #279: Breakpoints option uses time deltas instead of absolute time
1067 - #281: Daemonized actor no longer daemonized after an auto-restart
1068 - #282: TIT convertor erase traces when using absolute path in the trace list file
1069 - #285: segfault when a s4u actor kills itself with msg process tracing activated
1070 - #286: Pajé traces are not exposing the number of cores
1071 - #287: Command-line parsing should stop when encountering '--'
1072 - #288: MPI_Init(NULL, NULL)
1073 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
1074 - #290: Method to check if a mailbox has a message ready to be consumed
1075 - #291: [s4u] semaphores do not seem available yet
1076 - #294: Alltoall collective implementation of mpich
1077 - #295: Starting a Host as OFF is not supported
1078 - #296: DTD too permissive
1079 - #299: Add a s4u barrier example
1081 ----------------------------------------------------------------------------
1083 SimGrid (3.20) Released June 24. 2018
1085 The proxy snake_case() release.
1088 - The whole API is now using snake_case() instead of mixing it with camelCase.
1089 Compatibility wrappers in place for one year.
1090 - We took the opportunity to increase the consistency of the API.
1091 Please report any remaining glitches.
1092 - New function simgrid::s4u::Host::get_actor_count:
1093 Returns the number of actors running on a specific host.
1096 - SMPI is now tested with ~45 proxy apps from various sources, with none or
1097 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
1098 - Replay: The replay file has been re-written in C++.
1099 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
1100 supported. They are stored in the trace and used when replayed.
1101 - Basic support of MPI_Cancel. Robustness not guaranteed.
1102 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
1103 (only for MPI_COMM_TYPE_SHARED).
1104 - New option: smpi/privatize-libs, to add external shared libs to be privatized
1105 by SMPI. They will be copied locally and loaded separately by each process.
1106 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
1107 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
1108 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
1109 with internal index optind with multiple processes. Only works if MPI_Init has
1110 already been called.
1111 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
1112 syntax without preprocessing tricks.
1115 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
1116 behavior tracing is no longer limited to MSG
1117 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
1118 this means that traces now need to be updated or re-obtained; both Irecv
1119 and Isend lines in your traces have now in total 5 mandatory fields:
1120 <rankid> <command> <to/from rankid> <tag> <size>
1121 To update your traces, it suffices to add a 0 for the tag here.
1122 - TIT now also supports waiting for a distinct request via MPI_Wait.
1123 Wait/Test now wait for a specific request, not just the last one that was
1124 issued. This unfortunately means another update, because we need to
1125 identify which request you want to wait for. We do this via the
1126 triplet (sender, receiver, tag), which needs to be added:
1127 <rankid> <command> <sender> <receiver> <tag>
1128 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
1129 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
1132 - The deprecation of MSG is ongoing (but this should not impact you).
1133 Many MSG functions are now simple wrappers on the C API of S4U. If
1134 you wish to convert your code to S4U, find the S4U counterparts of
1135 your MSG calls in src/msg/msg_legacy.cpp.
1136 - MSG can still be used, but won't evolve anymore.
1139 - Allow to run the Link energy plugin from the command line with
1140 --cfg=plugin:link_energy
1141 - Rename Energy plugin into host_energy
1142 - Rename Load plugin into host_load
1145 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
1146 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
1147 - Improve performance when handling timeouts of simix synchros.
1150 - Config: the C API is now deprecated (will be removed in 3.23), and
1151 the C++ API has been slightly improved.
1154 - Fix several build issues on OSX.
1155 - Move simgrid_config.h to simgrid/config.h (old header still working)
1158 - #143: Setting a breakpoint at a given time
1159 - #150: Inconsistent event names in SMPI replay
1160 - #258: daemonized actors hang after all non-daemonized actors have completed
1161 - #267: Linker error on unit_tmgr
1162 - #269: SMPI: tracing of MPI_Wait/all/any broken
1163 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
1165 ----------------------------------------------------------------------------
1167 SimGrid (3.19.1) Released March 22 2018
1169 The Fixed ABI Release.
1171 As you may know, we are currently refactoring SimGrid in deep.
1172 Upcoming SimGrid4 will be really different from SimGrid3: modular,
1173 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
1175 Our goal is to smooth this transition, with backward compatibility
1176 and automatic update paths, while still progressing toward SimGrid4.
1178 SimGrid remains open during works: The last pure SimGrid3 release was
1179 v3.12 while all subsequent versions are usable alpha versions of
1180 SimGrid4: Existing interfaces remain unchanged, but the new S4U
1181 interface is budding and the internals are deeply reorganized.
1183 Since 2015, we work hard to reduce the changes to public APIs. When
1184 we need to rename a public library symbol in S4U, we let your
1185 compiler issue an explicative warning when you use the deprecated
1186 function. These messages remain for four releases, i.e. for one full
1187 year, before turning into an error. Starting with v3.15, your can
1188 also adapt to API changes with the SIMGRID_VERSION macro, that is
1189 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
1191 Starting with this v3.19.1, our commitment to reduce the changes to
1192 the public interfaces is extended from the API to the ABI: a program
1193 using only MSG or SimDag and compiled against a given version of
1194 simgrid can probably be used with a later version of SimGrid without
1195 recompilation. We will do our best... but don't expect too much of
1196 it, that's a really difficult goal during such profund refactoring.
1198 The difference between v3.19 and v3.19.1 is that the former was
1199 accidentally breaking the ABI of MSG, while the later is restoring
1202 S4U and kernel APIs will still evolve until SimGrid4, with one-year
1203 deprecation warnings as currently. In fact, cleaning up these
1204 interfaces and converting them to snake_case() is one release goal of
1205 v3.20. But don't worry, we are working to smooth this upgrade path.
1207 In summary, new projects should start with S4U to benefit of the
1208 future, but old MSG projects should still be usable with no change.
1210 ----------------------------------------------------------------------------
1212 SimGrid (3.19) Released March 20 2018
1214 The Moscovitly-cold Spring release.
1216 SMPI user-visible improvements:
1217 - Support of MPI_Type_create_subarray
1218 - Unimplemented MPI calls now abort (instead of issuing a warning)
1219 Some harmless calls around errhandlers still only issue a warning.
1220 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
1221 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
1222 is now possible (mmap privatization was refactored to this end).
1223 Adding new SMPI instances should also be technically doable now.
1225 SMPI internal cleanups:
1226 - Removed index notion from SMPI -- no more "getPid() - 1"!
1227 In tracing files, actors are called rank-<PROCESS-ID> but these are
1228 NOT the real ranks (tracing several communicators is not supported yet).
1229 - Deprecate smpi_process_index() (will be removed in 3.22)
1230 - Better testing on classical proxy apps, through an external project:
1231 https://github.com/simgrid/SMPI-proxy-apps/
1234 - Use a graphical TOC to make it easier to find the documentation you need
1235 - Revamp the MSG tutorial
1238 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
1239 - Remove parameter of MSG_process_killall().
1240 Resetting the PID was bogus anyway (several actors could have the same PID).
1243 - Execution->setHost() can be called after start(), to migrate it.
1244 - Comm::test_any() is now implemented.
1245 - s4u::Actor now has onCreation() and onDestruction() signals.
1246 - Install some kernel header files for the users' plugins and more.
1247 Warning, their API is really not stable yet.
1248 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1249 include/simgrid/kernel/routing/*Zone.hpp
1252 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1254 Build System and other cleanups:
1255 - Remove unused run-time parameter "tracing/onelink-only".
1256 - SimGrid now works with intel compilers (MC still buggy)
1257 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1258 - tesh kills subprocesses on timeout (but not on windows)
1261 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1264 - #194: Feature request: simgrid::s4u::Comm::test_any()
1265 - #245: migrating an actor does not migrate its execution
1266 - #253: Feature Request: expose clusters as objects
1267 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1268 - #255: Tesh broken on Windows
1269 - #256: Modernize FindSimGrid.cmake
1270 - #257: Fix (ab)use of CMake install
1272 ----------------------------------------------------------------------------
1274 SimGrid (3.18) Released December 24 2017
1276 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1278 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1279 - Support for MSG should not be removed from SimGrid before 2020,
1280 but future evolutions will be limited to the integration of
1281 user-provided patches.
1284 - s4u::onDeadlock() signal, run before stopping the simulation.
1285 - s4u::Exec: asynchronous executions (abstraction of background
1287 - s4u::Host->getLoad() returns the achieved speed in flops/s
1290 - Link::name() is deprecated and will be removed in v3.21, use
1291 Link::getCname() instead.
1292 - Mailbox::getName() changed to return a std::string, use
1293 Mailbox::getCname() instead to get a char*.
1294 - Storage::getName() changed to return a std::string, use
1295 Storage::getCname() instead to get a char*.
1296 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1297 to have both allocation and deallocation of the map in user space.
1300 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1301 The amount of flops remaining to do is a vector, not a scalar.
1302 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1303 name implies on both sequential and parallel tasks.
1304 - Both changes fix GitHub's #223 using PR #237 as a basis.
1305 Thanks Michael Mercier.
1306 - Most examples were converted to S4U and hidden elsewhere at tests.
1307 Rationale: we still want MSG to work; we want newcomers to use S4U.
1310 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1311 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1314 - Switch to the faster dlopen privatization mechanism by default
1315 - Documentation improvments
1318 - Live migration is getting moved to a plugin. Dirty page tracking is
1319 the first part of this plugin. This imply that VM migration is now
1320 only possible if one this function is called:
1321 - C/MSG: MSG_vm_live_migration_plugin_init()
1322 - C/C++: sg_vm_live_migration_plugin_init()
1323 - Java: Msg.liveMigrationInit()
1324 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1328 - Define class simgrid::xbt::Path to manage file names.
1329 - Removed unused functions:
1330 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1331 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1332 xbt_graph_export_graphviz()
1333 - xbt/str.h: xbt_str_join()
1334 - Remove unused datatypes:
1335 - xbt/heap.h: use std::priority_queue or boost::heap instead
1336 - xbt/swag.h: use boost::intrusive::list instead
1339 - New link_energy plugin for the consumption of the links.
1340 - All of the operations on files and storage contents have been
1341 packaged into a plugin (src/plugins/file_system). The current
1342 public interface can be found in
1343 include/simgrid/plugins/file_system.h
1344 To use these functions you now have to initialize the plugin by
1345 calling MSG_storage_file_system_init() just after calling
1346 MSG_init() or sg_storage_file_system_init() just after creating
1350 - Remove the undocumented/untested tag <include>
1353 - Remove viva specific tracing as the tool is no longer maintained
1356 - #248: Kill of finished processes leads to segfault
1357 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1358 - #239: please implement signal s4u::onDeadlock()
1359 - #230: segfaults when exit() without run()
1360 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1361 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1362 - #222: Actor::kill() doesn't really kill and segfaults
1363 - #221: odd LMM warning when killing an actor
1364 - #120: Memory leak when the processes are forcefully killed
1366 ----------------------------------------------------------------------------
1368 SimGrid (3.17) Released October 8 2017
1370 The Drained Leaks release: (almost) no known leaks despite the tests.
1372 Even very long simulations will preserve your memory: our testsuite
1373 stresses SimGrid for over 45mn on fast machines for a coverage of
1374 over 80%, and there is only one single known leak, of about 4kb.
1378 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1379 handler may be given to free resources if the comm cannot be completed.
1380 - this_actor::send(mailbox) is now mailbox->put()
1381 - New: simgrid::s4u::Comm::wait_all()
1382 - New: Host.execute() for remote executions.
1385 - Unused option network/sender-gap is removed.
1388 - Deprecate MSG_task_isend_with_matching(): This unused feature
1389 really complicates our internals. Will be removed in v3.20.
1392 - Improved context termination. It is now possible to free resources, even
1393 when a process is forcibly killed.
1394 - ContextBoost: add support for Boost versions above 1.61.
1397 - Bring back run-time option --cfg=exception/cutpath to remove exception
1399 - Removed unused functions:
1400 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1401 xbt_str_rtrim(), xbt_str_trim().
1402 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1405 - Removed header files obsolete since SimGrid 3.12:
1406 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1408 - #3: SD_exit should be made optional
1409 - #120: Memory leak when the processes are forcefully killed
1410 - #159: Threading test regression in Actor refcounting
1411 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1412 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1413 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1414 - #191: VM migration and pstate
1415 - #192: Updating the energy consumptions of all hosts crashes with VMs
1416 - #195: All actors have PID=0 in the logs
1417 - #204: Sometimes segfault with thread contexts and mmap privatization
1418 - #222: Actor::kill() doesn't really kill and segfaults
1419 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1421 ----------------------------------------------------------------------------
1423 SimGrid (3.16) Released June 22. 2017.
1425 The Blooming Spring Release: developments are budding.
1428 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1429 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1430 - Java JAR file should be finally fully working on Mac OSX too.
1432 XML platforms: Switch to platform v4.1 format.
1433 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1434 - <zone> can be used as a synonym for the now deprecated <as>
1435 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1436 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1437 - <actor> can be used as a synonym for the now deprecated <process>
1438 - state_file and avail_file periodicity is now easier to express
1439 (check the documentation) the old behavior should still work.
1440 * Storage: not backward compatible, but it seems that nobody ever used it
1441 - Remove attribute 'content_type' of <storage_type>: was never used
1442 - Make attribute 'model' of <storage_type> optional: for future usage
1443 - Remove Bconnection model property: was never (in)validated. Replaced by
1444 the maximum of the read and write bandwidth as a resource constraint.
1447 - New and Backwards Compatibility break:
1448 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1449 When one wants to get the list of tasks whose states have changed during a
1450 simulation round, s/he has to allocate and free a dynar and use it as
1451 argument to this function. The former SD_simulate (double how_long)
1455 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1456 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1459 - New function to update the consumption of all hosts at once.
1460 - Fix the model for multi-core hosts, linear in the amount of busy
1461 cores with an abnormality for pIdle (see paper).
1462 BUG: Still not satisfactory for ptask on mono-cores.
1465 - The netzone are now available from the MSG API.
1466 The old names still work, but are now deprecated.
1469 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1470 - New option: smpi/keep-temps to not cleanup temp files
1471 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1472 mallocs mode is used, allows to change the size of the fake file used
1473 (default 1MB), to potentially limit the number of mappings for large runs.
1474 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1475 - Fortran ifort and flang compilers support
1476 - New RMA calls supported (experimental):
1477 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1478 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1479 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1480 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1481 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1482 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1485 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1486 - Behind the scene, ActivityImpl are now automatically refcounted.
1487 Removing this nasty bottleneck will greatly simplify our internals.
1490 - Replay: New function xbt_replay_action_get():
1491 Retrieve the function previously associated to an event type.
1492 - DROPPED FUNCTION: xbt_str_varsubst()
1493 - DROPPED MODULE: strbuff. We don't need it anymore.
1494 - DROPPED MODULE: matrix. We don't need it anymore.
1495 - DROPPED MODULE: lib. We don't need it anymore.
1497 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1499 SimGrid (3.15) Released March 22. 2017
1501 The Spring Release: continuous integration servers become green
1502 We fixed even the transient bugs on all target architectures:
1503 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1504 Windows, FreeBSD, NetBSD.
1506 - Rename NetCards to NetPoints.
1507 This was intended to help NS3 users, but that's not a netcard.
1508 That's a point in the routing algorithm, let's avoid wrong simplifications.
1511 - New: MSG_process_yield(). Stop and yield to other processes.
1512 - New: MSG_process_daemon(). Daemon processes are automatically killed
1513 when the last non-daemon process terminates
1514 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1515 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1516 to make room for the upcoming network energy plugin.
1517 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1520 - Ensure that an actor can kill itself with Process::exit()
1521 - Kill the obscure NativeException. Nobody want to survive the issues
1522 it denotes, so use JniException that is a RuntimeException (not to
1523 be caught explicitly).
1524 - Partial bug fix in initialization. SimGrid flags on command line were
1525 consumed at C level but stayed in the original Java String[] args.
1526 This could mess users' args[i] if SG flags were not put at the end of
1528 The SimGrid flags are now removed from the Java arguments. However,
1529 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1530 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1531 It might be if you have --log or --cfg flags in the command line.
1532 - Fix numerous memleaks all around the place. In particular, around VMs.
1536 - s4u::onPlatformCreated: right before the simulation starts
1537 - s4u::onSimulationEnd: right after the main simulation loop
1538 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1539 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1540 event from the availability_file changes the avail speed.
1541 - Links are now usable from s4u
1542 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1543 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1544 - New examples: The conversion of MSG examples to S4U has begun
1545 - Actors: create, kill, migration, and suspend
1546 - Applications: master-worker and token-ring
1547 - Action replay: communications and storage
1548 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1551 - Backwards Compatibility breaks
1552 - The SD_route_get_list and SD_route_get_size functions have been removed.
1553 They are replaced by the unique
1554 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1555 The route from src to dst is built in the links dynar whose size can
1556 be obtained with xbt_dynar_length.
1557 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1558 sg_host_route_bandwidth and sg_host_route_latency.
1559 Macros ensure the backwards compatibility, but you should fix your code
1562 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1563 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1564 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1565 We don't support 100% of it, but it should be enough. Contact us if not.
1566 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1567 naive and probably inaccurate)
1568 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1569 mapped onto a unique small file using some system magic.
1570 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1571 - Fortran: cleanups, fixes, support of user-added operations
1572 - RMA: MPI_Accumulate are applied in correct order.
1573 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1574 - Better support for MPI_IN_PLACE
1575 - Support for MPI_Win attrs and keyvals.
1576 - Support MPI_Comm_group_create, MPI_Type_size_x
1577 - MPI ops should be properly applied to all allowed types
1578 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1579 algorithm by a non failing one from mpich.
1582 - Kill the fifo data container: we don't use it anymore.
1585 - A <cluster> can now be created with different speed values to represent pstates.
1586 see examples/platforms/energy_cluster.xml for an example.
1588 SimGrid (3.14.159) Released December 28. 2016
1590 The previous release was only a crude approximation of a Pi release;
1591 this one should be a bit better.
1593 - Revert a last minute change that broke on Mac OSX.
1594 - Fix the changelog and NEWS files.
1595 - Improve the documentation, in particular of the routing module.
1596 - Integrate some patches flying around in Debian.
1598 SimGrid (3.14) Released December 24. 2016
1601 * The whole documentation was reworked and reorganized. There is
1602 still much room for improvement here, but we're on it.
1605 * We now need python3 for our testing infrastructure.
1606 * Model checking needs libevent
1609 * Now works on FreeBSD
1612 * Privatization now works on FreeBSD
1614 * Privatization is now activated by default in smpirun
1615 You can turn it off with -no-privatize if something goes wrong.
1617 * Call-location tracing for SMPI.
1618 You can add the exact location (filename / linenumber) of an MPI call to
1619 your trace files and slow down or speed up the simulation between two
1620 consecutive calls by using an adjustment file (see the documentation).
1622 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1623 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1624 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1625 values were ignored in some cases. The timings of these functions can now
1626 be significantly different.
1628 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1629 smpi/running-power is renamed to smpi/host-speed
1631 * smpi/grow-injected-times option to enable or disable multiplication of the
1632 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1633 make simulation less precise (but also much faster).
1635 * smpirun script should be (much) faster for large deployments.
1637 * SMPI tracing: fixed issue with poor matching of send/receives.
1639 * Replay: Fix broken waitall
1641 New functions and features
1642 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1644 Dropped / renamed functions and features
1645 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1646 - MSG_mailbox_is_empty() -> Mailbox::empty()
1647 - MSG_mailbox_front() -> Mailbox::front()
1648 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1649 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1650 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1651 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1652 * Don't pass the free_f parameter to property related functions:
1653 - MSG_host_set_property_value()
1654 - MSG_as_router_set_property_value()
1655 - MSG_storage_set_property_value()
1656 * VM properties. Since msg_vm_t are msg_host_t, just use
1657 MSG_host_get_property_value() and friends
1658 * VM I/O related things:
1659 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1660 - Unimplemented save/restore methods
1661 * MSG_as_router_get_property_value() was redundant with
1662 MSG_environment_as_get_property_value().
1663 - Removed MSG_as_router_*propert*() functions
1664 - Added MSG_environment_as_set_property_value() for consistency
1665 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1666 Well, they are still there for now, but deprecated with a warning.
1667 Please switch to xbt_dict_new_homogeneous() before this is removed
1669 * Task affinity. Its intended behavior (that was very badly tested
1670 and probably not really working) was deceiving what most users
1671 would have hoped here.
1672 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1673 * Option network/coordinates is now useless and should be dropped.
1676 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1679 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1680 * Added Dragonfly topology support
1683 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1684 EiBps, EBps, Eibps, Ebps for bandwidth.
1685 They may become useful to some lucky ones.
1688 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1689 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1690 * ex/async_waitAll: new example, on asynchronous communications
1693 * Memory usage should be decreased for simulations with a large number
1694 of processes. This also helps for SMPI.
1696 SimGrid (3.13) stable; urgency=low
1698 The Half Release, a.k.a. the Zealous Easter Trim.
1700 - We removed half of the lines, that were mostly experimental cruft.
1701 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1702 (not counting blanks and comments -- according to openhub.net)
1703 - The internals are now compiled in C++ (and will soon be clean C++)
1704 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1705 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1707 * Backwards Compatibility breaks
1708 - Removed Lua simulation bindings (switch to C or Java for that).
1709 Lua can still be used to describe platforms
1710 - Removed Java kernel plug-ins.
1711 Will be reintroduced after the ongoing major internals reorg.
1713 - the following functions were removed.
1714 They were too specific and should be reimplemented in a generic
1715 way, with filter function.
1716 - MSG_task_listen_from_host
1717 - MSG_mailbox_get_count_host_waiting_tasks
1718 - MSG_mailbox_put_with_timeout was removed.
1719 Please use MSG_task_send_with_timeout instead.
1721 - the SD_application_reinit function was removed. It has been a noop for a while.
1722 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1723 replaced by a more flexible mechanism.
1724 - The following functions thus do not exist anymore
1725 - SD_workstation_get_access_mode
1726 - SD_workstation_set_access_mode
1727 - SD_workstation_get_current_task
1728 - Basic estimation functions have been removed but can easily be replaced
1729 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1730 - SD_workstation_get_computation_time => amount / sg_host_speed()
1732 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1733 Use VM.getSpeed()*load/100 for the legacy behavior.
1735 - option enable_tracing was removed. It was not doing anything for a while.
1736 - In the ModelChecker:
1737 - the model-checker now ptraces the model-checked process which means
1738 you cannot use a debugger on the latter anymore (we might make this
1739 optional in the feature);
1740 - removed soft-dirty page tracking;
1741 - remove model-checked side snapshot management,
1742 MC_snapshot() and MC_compare_snapshot();
1743 - keep the MC_cut() function as a stub (it was not really working
1744 in the previous release).
1747 * All options are consistently in kebab-case. Old names are kept as alias.
1750 * Switch to platform v4 format.
1751 - Rename from 'power' to 'speed' the attributes describing the amount of
1752 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1753 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1754 - In <host> and <link>, attributes availability and state are gone.
1755 It was redundant with state and availability traces, and with peak values.
1756 - In <cluster>, attributes availability_file and state_file are gone.
1757 It was too complex and unused.
1758 - Kill <gpu>. Was not doing anything.
1759 - The DOCTYPE points to the right URL:
1760 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1761 (the file at this address now documents the changelog since its v1)
1762 - A warning is emitted for unit-less values (they are still accepted).
1763 - speed. Default: 'f' or 'flops'. Also defined:
1764 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1765 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1766 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1767 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1768 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1769 - latency. Default: 's' second. Also defined:
1770 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1772 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1773 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1776 * s4u::Host is now the preferred public interface to the Host features.
1777 sg_host_* functions are C bindings to the exact same behavior
1778 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1781 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1782 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1783 They were deprecated since a few years, and probably did not even compile anymore.
1786 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1787 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1788 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1789 encouraged to update their codes. A list of the modified functions can be found at the end of
1790 include/simgrid/simdag.h
1793 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1794 the lambda or closure passed as a parameter will run in kernel mode.
1795 All the callback functions should be rewritten to that interface at some point.
1798 * Reorganizing and cleaning the internals all around the place.
1801 * Remove old default barrier/bcast buggy algorithms (see #18407)
1802 * Various bug fixes to handle more codes
1803 * Remove the need for the --foreground option of smpirun (it is still
1804 accepted for backward compatibility).
1807 * Kill the setset data container: MC don't use it anymore.
1808 * Kill the queue data container: it made more sense with GRAS.
1809 * Kill the xbt_peer_t data type: it's useless without GRAS.
1810 * Kill rm_cb feature of config sets: it was never useful.
1811 * Kill graphxml parsing feature. It was not used.
1812 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1814 - xbt_dynar_sort_strings(), when the content is char*
1815 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1816 They throw exceptions on invalid input;
1817 * C++ support for declaring CLI flags (simgrid::config::Flag);
1818 * class for abstracting different signal backends (simgrid::xbt::signal).
1819 with no external dependencies (we need very simple signals).
1822 * refactoring and cleanup of the code;
1823 * ongoing process to cleanly separate the model-checking algorithms
1824 from the code model-checking support.
1826 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1828 SimGrid (3.12) stable; urgency=low
1830 The Facelift Release.
1833 * Require g++ v4.7 at least to not speak prehistorical C++.
1834 * Require Boost 1.48 (for signal2 component).
1835 * Java must be version 7 at least when activated.
1836 * Builds on Windows again (including Java bindings).
1837 * Tracing is now always enabled (no way to turn it out)
1838 * Remove GTNetS. It was not working anyway.
1839 * Various cleanups in the cmake scripts.
1840 * Move headers around to sort them out on installed systems:
1841 - instr/instr.h -> simgrid/instr.h
1842 - instr/jedule/* -> simgrid/jedule
1843 - simdag/datatypes.h was removed
1844 - simdag/simdag.h -> simgrid/simdag.h
1845 - msg/datatypes.h was removed
1846 - msg/msg.h -> simgrid/msg.h
1849 * Interface improvement:
1850 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1851 - Sanitize the interface in MSG_task_ module:
1852 - Merge two functions that were close enough but misleading:
1853 set_compute_duration(t) -> set_flops_amount(t)
1854 get_remaining_computation(t) -> get_flops_amount(t)
1855 - set_data_size(t) -> set_bytes_amount(t)
1856 get_data_size(t) -> get_bytes_amount(t)
1857 - Massive cleanups in the functions related to the energy
1858 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1859 - New: MSG_host_get_pstate()
1861 - msg/energy/onoff: switching hosts on and off
1863 * Interface improvement:
1864 - Rename Host.isAvail() to Host.isOn()
1865 - Rename Process.currentProcess() to Process.getCurrentProcess()
1866 - Rename Task.setDataSize() to Task.setBytesAmount()
1867 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1869 - #18874: Actually allows the GC to reclaim tasks
1873 - SIMIX_process_throw: raises an exception in a remote process
1874 * Refactoring: Separate sub-modules
1875 - libsmx: the public interface, as libc in a real system
1876 - popping: the strange dance that converts a user request into a kernel handling
1877 - smx_context_*: the virtualization mechanisms that embed the user code
1878 - smx_*: the handling of each simcalls
1879 * Interface cleanup:
1880 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1881 * Rename smx_action_t into smx_synchro_t, making explicit that these
1882 things are used to synchronize processes with their environment.
1883 For example, a communication is a sort of synchronization involving
1884 the communicating processes (that may block until the exchange) and
1885 the platform. The same can be said from computations, etc.
1887 - #18888: segfault when a process ends before its kill_time
1891 - 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.
1892 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1893 - MPI_Type_set_name, MPI_Type_get_name
1894 - MPI_*_c2f and MPI_*_f2c functions
1895 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1896 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1897 - Activate a lot of new tests from the mpich 3 testsuite
1899 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1900 - InfiniBand network model added: Based on the works of Jerome Vienne
1901 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1902 - When smpi/display_timing is set, also display global simulation time and application times
1903 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1904 * Collective communications
1905 - 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
1906 - 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/).
1907 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1909 - Replay now uses algorithms from wanted collective selector
1910 - Replay can be used with SMP-aware algorithms
1911 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1912 - 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.
1914 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1915 - Correctly use loopback on fat-tree clusters
1916 - Asynchronous small messages shouldn't trigger deadlocks anymore
1917 * Energy/DVFS cleanup and improvement
1918 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1919 - new: smpi_get_host_pstate()
1922 - "Full" network optimization flag was broken since Surf++
1923 - Better handling of precision flags in maxmin
1924 - Fix bug causing sometimes "Impossible" errors
1925 - Properly pass cluster properties to included hosts
1926 * Improvement of the Energy plugin.
1927 - Always update the consumption before returning that value
1928 - New property: watt_off to denote the dissipation when the host is off
1929 - New functions getWattMinAt and getWattMaxAt to retrieve the
1930 dissipation of pstates that we are not currently at.
1931 * Java: class NetworkLink renamed to Link
1932 * New function: simcall_process_get_kill_time()
1933 * Massive rename s/workstation/host/
1934 - That's intrusive, but that's good for the project consistency. Sorry.
1935 - Change config option "workstation/model" into "host/model"
1938 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1939 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1940 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1941 - xbt_dirname and xbt_basename for non-POSIX systems
1943 * The model checker now runs as a separate process.
1944 * The model checker runs must now be launched with the new simgrid-mc program.
1945 * Record/Replay: the MC can display a textual representation of a path in the
1946 execution graph. It can then be replayed outside of the model checker.
1948 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1950 SimGrid (3.11) stable; urgency=low
1955 * Normalizing pointers addresses tool for better diff between logs
1958 * Add cloud examples using new VMs
1959 - examples/msg/cloud/two_tasks_vm.tesh
1960 - examples/msg/cloud/simple_vm.tesh
1961 - examples/java/cloud/cloud.tesh
1962 - examples/java/cloud/migration/migration.tesh
1963 * Add java surf examples:
1964 - examples/java/surfPlugin/surf_plugin.tesh
1965 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1966 - examples/java/surfCpuModel/surf_cpu_model.tesh
1967 * Add SMPI+MSG example:
1968 - examples/smpi/smpi_msg_masterslave/
1973 - msg task destroy cancel test
1974 - msg_host on/off test
1975 * Move all tests in testsuite to teshsuite (adding tesh files)
1976 * Restructure teshsuites
1977 - one folder for each kind of test
1978 * Restructure AddTests.cmake
1980 - structure the order of tests (with sections)
1983 * Add virtual machine
1984 - creation of a VM on a PM
1985 - migration of a VM from a PM to another PM
1987 - MSG_process_join(msg_process_t process, double timeout)
1988 - msg_bar_t MSG_barrier_init(unsigned int count)
1989 - int MSG_barrier_wait(msg_bar_t barrier)
1990 - void MSG_barrier_destroy(msg_bar_t barrier)
1991 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1992 * New option "msg/debug_multiple_use" to help debugging when a task is used
1995 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1997 - Increase code coverage in test suites
2001 * Protect context stacks against stack overflow. The number of protected memory
2002 pages allocated on the top of each stack (1 by default) can be configured
2003 with the new command line option --cfg=contexts/guard_size:<value>.
2004 * Simcalls are now generated by a python script that
2005 - generates files included by SimGrid
2006 - checks that all the functions exist, and proposes prototypes
2008 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
2009 host_set_data, host_get_data
2011 - simcall_process_join(smx_process_t process, double timeout)
2012 * Fix bug where sleeping processing could not be suspended.
2015 * Translate surf models from C to C++
2016 - Generic classes for all models: Model, Resource, Action
2017 - A generic interface for each kind of model (CPU, Network, Storage
2018 Workstation, WorkstationVM)
2020 * Translate surf routings from C to C++
2021 * Add callbacks using sigc++ or boost::signals2
2022 - Add callback functions for resource creation/destruction
2023 - Add callback functions for action state change
2024 - Handle Energy as a plugin
2025 * Replace swag by boost::intrusive
2026 * Add new routing models for clusters. For documentation, see
2027 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
2028 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
2029 parameters for cluster tag
2030 - Fat trees, with topology="FAT_TREE" and
2031 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
2032 - see examples/platforms/torus_cluster.xml and
2033 examples/platforms/fat_tree_cluster.xml
2034 * More documentation
2037 * Hostfiles support host:nb_processes construct to deploy several processes on
2039 * Collective communication algorithms should not crash if used with
2040 improper number of nodes and report the error.
2041 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
2042 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
2043 MPI_Dims_create, MPI_Cart_sub are supported.
2044 * New interface to use SMPI programmatically (still depends on MSG for
2045 some parts, see examples/smpi/smpi_msg_masterslave):
2046 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
2050 * Global variables privatization in MPI executables is now performed at runtime
2051 with the option smpi/privatize_global_variables (default:no).
2052 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
2053 dynamic libraries loaded by the application are not privatized (static
2054 linking with these libraries is advised in this case)
2057 * Options defined in XML work correctly now.
2060 * New cmake option, enable_lib_in_jar, to control whether native libraries are
2061 copied into simgrid.jar or not (ON by default). Use this option if you want
2062 to reduce the size of the installed simgrid.jar, *and* the native libraries
2063 are kept installed elsewhere.
2064 * Surf binding with SWIG (code generated in maintainer mode only):
2065 - plugin to handle callbacks
2066 - CPU model only for the moment
2069 * Supernovae build mode is definitively removed. It was used to improve
2070 inlining and inter-module optimizations. It is nowadays superseded by
2071 link-time optimizations commonly available in compilers.
2072 * Update ns-3 find lib. Bindings for ns-3 should work again now.
2073 * Add boost dependency for surf++
2074 * Add new macro for tests
2075 - ADD_TESH(name <tesh_args>)
2076 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
2080 - xbt_bar_t XBT_barrier_init(unsigned int count)
2081 - int XBT_barrier_wait(xbt_bar_t barrier)
2082 - void XBT_barrier_destroy(xbt_bar_t barrier)
2083 * Make the xbt_os_time module public
2085 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2087 SimGrid (3.10) stable; urgency=low
2089 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
2092 * Reintegrate Java to the main archive as desynchronizing these
2093 package is not acceptable anymore (Java is now considered stable)
2094 * Add explicit synchronization facilities through semaphores
2095 * Bug fix: Task.setDataSize() only changed the C world, not the value
2096 cached in the Java world
2099 * Dramatically change the way files are handled. API and internals changed, but
2100 this part of MSG was not considered as production grade either.
2101 * Add explicit synchronization facilities through semaphores
2102 * Add a new function MSG_host_get_process_list()
2103 * Preliminary DVFS support (see examples/msg/energy/ for details)
2106 * SMPI is now included directly in the libsimgrid as the windows
2107 linker doesn't force us on splitting it anymore.
2108 * Improvements of the SMPI replay tool:
2109 - Most of the collective communications are now rooted in the same process as
2110 in the original application.
2111 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
2112 used until now). Multiple data types can now be used in a trace.
2113 - The replay tool now supports traces produce either by TAU or a modified
2115 - Bug Fix: the compute part of the reduce action is now taken into account.
2116 - Gatherv collective is now supported
2117 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
2118 of smpirun outputs time independent traces for the current run. One file
2119 is created per process. If too many processes are simulated, this behavior
2120 can be changed to one file for all processes by using the
2121 tracing/smpi/format/ti_one_file flag
2122 * smpirun generates the host file if needed (with given host count and platform)
2123 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
2124 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
2125 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
2126 --cfg=smpi/coll_selector:(mpich/ompi)
2127 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
2129 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
2130 happening inside a collective SMPI call.
2131 * Fix the behavior of complex data types handling.
2132 * Make MPI_Wtime another synchronization point to take computations into
2134 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
2135 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
2136 * Add all missing Fortran bindings, SMPI should work with Fortran 90
2137 (no privatization of global variables yet)
2138 * Preliminary DVFS support (see examples/smpi/energy/ for details)
2141 * Verification of liveness properties is now available for SMPI applications
2142 (in addition to MSG applications)
2143 * Bugged examples using SMPI in examples/smpi/mc/
2144 * Add --cfg=model-check/visited option. Allows the verification of infinite
2145 programs. Detection of loops in the execution thanks to the system state
2146 comparison and reduction of the state space to explore. Can be combined with
2147 DPOR for safety properties.
2150 * Allow to change SimGrid configuration (see --help) within the code
2151 thanks to SD_config() as it can be done in MSG.
2152 * Add a new function SD_task_set_amount() upon user request.
2155 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
2156 * Remove rule based routing (no more PCRE dependency)
2157 * Add a limiter_link option to cluster tag, to specify a maximum reachable
2158 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
2159 * Add a loopback_bw and loopback_lat options to cluster tag.
2160 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
2161 routing. Now peers are encapsulated in an AS and have their own private
2162 router but this is transparent.
2165 * Our own implementation of getline is renamed xbt_getline, and gets
2166 used even if the OS provide a getline(). This should reduce the
2167 configuration complexity by using the same code on all platforms.
2168 * New type: xbt_cfg_elm_boolean.
2169 * Allow to use yes/no for boolean configuration options in the command line.
2170 * Allow to disable SimGrid cleanups at exit from command line option.
2171 There are situations where one may want a simulation to end with an exit.
2172 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
2173 annoying when scripting around the simulator. Adding a
2174 --cfg=clean_atexit:no allows to circumvent this issue.
2177 * Lots of memory leaks were corrected in this release.
2178 There is no known memory leaks anymore, in all of our 600+ tests.
2179 * New command line option --version, to get SimGrid version information.
2180 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
2182 * Supernovae builds are deprecated, and expected to be removed in the next
2185 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2187 SimGrid (3.9) stable; urgency=low
2189 The Grasgory release: GRAS is really dead now.
2191 * Complete overhaul of the internal host structures scheme.
2194 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
2195 was considered as experimental and badly maintained since then.
2196 * Keeping it was thus a trap to our potential users, that could take
2197 it instead of MSG or SMPI by mistake despite is pity state.
2198 * GRAS seems to have very few users (if any), and no one volunteered
2199 to maintain it further. It also induces a lot of XBT code (for
2200 portability sake), that must be maintained too.
2201 * For all these reasons, we killed GRAS. If someone wants to revive it
2202 in the future, don't cry, our git history still remembers of GRAS.
2205 * Major overhaul. Merge our documentation again as time proved that
2206 splitting it was really not helping our users.
2207 * Further improve the developer documentation to help newcomers
2208 hacking on SimGrid itself. The user documentation (and in
2209 particular, the beginner documentation) is still in a sorry state.
2212 * Now works on Windows too!
2213 * Much more extensive test suite, from MPICH
2216 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
2217 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
2218 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
2219 * Bug fix: let task be scheduled when the last dependency to be solved is
2220 a control dependency.
2221 * Remove SD_load_environment_script function.
2222 Use the C sg_platf function if you want to declare a platform
2226 * New function: MSG_process_get_number()
2227 * Old function documented: MSG_config()
2228 * Remove MSG_load_platform_script function
2229 Use the C sg_platf function if you want to declare a platform
2233 * Change the default value of the TCP_gamma constant (maximal size of TCP
2234 congestion window) to a more realistic 4MiB value. If you notice changes in
2235 your simulation results, you can fall back to the previous 20k tiny window
2236 by adding --cfg=network/TCP_gamma:20000 on command line.
2237 * (Hopefully) fix a bug wrt periodic availability and state traces
2238 * Bug fix: use default values at start when first event in availability/state
2239 trace is not at time 0.
2242 * remove the "new_" part of function name sg_platf_new_trace_connect
2243 (resulting in sg_platf_trace_connect), since it does not create
2247 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2248 use case where it's really mandatory, and maintaining it was a pain in
2250 * New: xbt_fifo_search(), search an item with a user-provided
2251 comparison function instead of dumb pointer comparison.
2254 * Fix the lua deployment:
2255 Use `simgrid.init_application()` before deployment instead of
2256 `simgrid.msg_register_application()` after.
2259 * Transfer the tracing files into the corresponding modules.
2261 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2263 SimGrid (3.8.1) stable; urgency=low
2265 The "we are told that some people want to also *install* the simgrid
2268 * Add missing file "tesh.1" to the archive.
2270 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2272 SimGrid (3.8) stable; urgency=low
2274 The Psssshiiiit release: SimGrid jumps into the Cloud.
2277 * Add an experimental interface to manipulate VMs. They are mainly
2278 process groups with very few intrinsic semantic, but they should
2279 allow you to build the semantic you want easily.
2280 * New function: MSG_host_set_property_value()
2281 * New function: MSG_process_on_exit(). To clean memory in all cases.
2282 * Bug fixes that made the host (and link) failures unusable.
2283 * Add a way to auto-restart process when the host in which they are
2284 executing comes back (ON_FAILURE="RESTART" on deployment file,
2285 MSG_process_auto_restart_set).
2286 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2287 please stop using the old ones, they are DEPRECATED.
2289 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2290 Please use MSG_init() instead. (reducing the amount of entry
2291 points in the library helps us).
2292 * Make it impossible to link against the wrong version of the lib
2293 * Deprecate MSG_clean(). No need to call it anymore.
2294 * Function MSG_get_host_number() is not deprecated anymore.
2297 * Split the doc into a user guide and a reference guide.
2298 * Start a developper guide to help people hacking on SimGrid.
2301 * Enable tracing by default. This modules rocks you should use it.
2302 * Remove option custom_flags. Now use environment variables CFLAGS
2304 * Use default cmake things to detect lua instead of home grown ones.
2305 * New option "enable_mallocators" to disable mallocators, for debugging
2306 purpose ("on" by default).
2309 * Bug fixes around the resource failures: don't let the processes
2310 survive the host they are running onto.
2311 * Add an interface to auto-restart processes when the host in which they are
2312 executing comes back.
2313 * Ensures that SIMIX_clean is called automatically. It's not part of
2314 the public interface anymore (bindings should be updated).
2317 * Bug fix for when SD_Simulate is called with a positive value: be careful
2318 when comparing doubles. Sometimes they are different for non significant
2320 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2321 parallel task whose initial work is distributed among host according
2322 to the Amdahl's law. Such tasks are created with a parameter alpha
2323 that corresponds to the non-parallelizable part of the computation.
2324 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2325 between two sets of workstations assuming a 1D block distribution (each
2326 workstation owns a similar share of data) on both sides.
2328 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2329 Data redistribution will be automatically scheduled once parent and child
2330 are both scheduled. The filling of computation_amount and
2331 communication_amount structures is now done seamlessly thanks to the chosen
2333 * New function SD_workstation_dump to display various information
2334 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2335 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2336 bandwidth on the route onto which the task is scheduled and the amount of
2338 To divide the nominal bandwidth by 2, the rate then has to be:
2339 rate = bandwidth/(2*amount)
2340 * Compute tasks that have failed can now be rescheduled and executed again
2341 (from their beginning)
2342 * Increasing source code coverage (src/simdag is now covered at 95.8%
2346 * Re-implement time-independent trace replay using SMPI (at the
2347 smpi_smp_* level) instead of MSG. This should replace
2348 examples/msg/actions/actions.c
2349 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2350 hindexed and structs)
2351 * Implement the exchange of non-contiguous data.
2352 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2353 * Correct behavior of smpi/sender_gap and set its default value to 0
2354 * Add option to asynchronously send small messages to allow better
2355 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2356 specifies the size in bytes under which messages will be asynchronously sent.
2357 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2358 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2361 * New C interface to define a platform: XML is now optional.
2362 For more info, please check include/simgrid/platf.h.
2363 * New interface to define random platforms from the C:
2364 For more info, please check include/simgrid/platf_generator.h and
2365 examples/msg/masterslave/masterslave_platfgen.c
2366 * Export a sg_cmdline dynar containing all the arguments we got from
2370 * Two new tracing options for adding comments to trace file so you
2371 can track your experiments (see --help-tracing for details).
2372 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2373 * Adding the SimGrid version that generated the trace file as a comment.
2374 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2375 * Fix to avoid key clashes on Paje links
2376 * Other minor fixes related to the Paje specification
2379 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2380 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2381 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2382 * Greatly improve the robustness of mmalloc to user errors (such as
2383 using an area after freeing it, or freeing it twice)
2385 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2387 SimGrid-java (3.8.1) stable; urgency=low
2389 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2390 Warning: all methods are visible, but only some of them are
2391 implemented so far. Check the source (src/jtrace.c)
2392 for further information.
2393 * New module: org.simgrid.msg.File (SimGrid File management functions)
2394 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2395 * Change the meaning of Process.restart: now restart the process from
2396 the beginning, like MSG_process_restart in C.
2397 * Add Process.setAutoRestart: handling of process restart when failed
2399 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2400 you to retrieve the properties of the processes/hosts
2401 * Deprecate Msg.clean(): you can just forget about it now.
2402 * New function Process.getCount(), that only works when compiling
2403 with the not yet released version 3.9 of the C library.
2405 * New context factory based on Coroutines. It mandates a modified JVM
2406 but then, the simulations run about five times faster, and there is
2407 no limit to the amount of processes (beside of the available memory).
2409 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2411 SimGrid (3.7.1) stable; urgency=low
2414 * Restore the prototype of MSG_process_create_with_environment() to
2415 the pre-3.7 situation by removing the kill_time argument.
2416 * Add a MSG_process_set_kill_time() function instead.
2419 * Fix weird behaviors when dealing with parallel tasks.
2422 * Simgrid is now built as a dll.
2423 * Simgrid-java now works on Windows.
2424 * Simgrid-Java is now included into Windows package.
2427 * First pre-build package for MacOSX.
2430 * Fix compilation when using MSG_USE_DEPRECATED.
2431 * Fix some compilation issues on Macs and Windows.
2432 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2433 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2435 We discovered that the Lua console is broken, but we are missing the
2436 manpower to fix it right now. The problem existed in 3.7 too, so we
2437 are not blocking the release for that. Sorry if you depended on this
2438 feature, any help would be really welcome.
2440 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2443 SimGrid-java (3.7.1) stable; urgency=low
2445 The "Java aint got to be bloated and slow" release
2448 * Various internal cleanups and performance improvement
2449 Simulations are expected to run up to twice faster or so
2450 * Make Process.kill(process) an instance method, not a static one
2451 * User processes are not java.lang.Thread subclasses.
2452 This breaks the compatibility (sorry), but previous API was
2453 brain-dead, making it impossible to have non-trivial
2454 initializations in the process constructor.
2455 * Require a full constructor per Process sub-class.
2456 Kinda breaks the compatibility (sorry), but this allows a much more
2457 efficient way to launch the processes at simulation startup.
2458 * Do not embeed our version of semaphores, java 1.5 can be considered
2459 as sufficiently prevalent for us to not dupplicate its features.
2463 * Add examples for almost every part of the API
2464 We spotted and fixed a lot of bugs in the process
2466 * New module: asynchronous communication API
2467 * New function: Process.sleep()
2468 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2469 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2470 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2472 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2474 SimGrid (3.7) stable; urgency=low
2476 The "spring cleanups (before the next Big Project kicks in)" release.
2479 * We can specify the SMPI latency/bandwidth factor with command line
2480 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2481 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2482 You can also use the "config tag" from platform file by adding this line
2483 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2484 (see "example/platforms/tag_config.xml" to use "config tag").
2485 Note that the command line supersedes the platform file configuration.
2486 * Change the correction factors used in LMM model, according to
2487 the latest experiments described in INRIA RR-7821.
2488 Accuracy should be improved this way.
2489 * Use the partial invalidation optimization by default for the
2490 network too. Should produce the exact same results, only faster.
2491 * Major cleanup in surf to merge models and split some optimization
2492 mechanisms from the core of the models. As a result you can now
2493 specify which model to use (e.g., --cfg=network/model:LV08
2494 --cfg=cpu/model:Cas01) and which optimization mode to use
2495 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2496 Incompatible combinations should err at initialization. See
2497 --help-models for the list of all models and optimization modes.
2498 * The CLM03 workstation model was dropped for simplicity because it
2499 used the deprecated CM02 network model. Use default instead.
2500 * Rename the TCP_gamma configuration option to network/TCP_gamma
2501 * Rename the coordinates configuration option to
2502 network/coordinates, and document it
2503 * Use now crosstraffic keyword instead of the terribly misleading
2504 fullduplex keyword. It is activated by default now in the current
2505 default model, use --cfg=network/crosstraffic:0 to turn it off.
2506 * Ongoing refactoring the model parsing to make XML files optional
2507 See include/simgrid/platf.h for details (still to be completed)
2510 * Major overhaul of the documentation. Almost instructive now :/
2511 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2512 functions and friends. This interface was considered as
2513 deprecated since over 2 years, it's time to inform our users that it is.
2514 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2515 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2516 use these (crappy) functions in your code.
2517 These functions will be removed soon. Stop using them now.
2518 * Deprecate MSG_get_host_{table,number}
2519 Implement MSG_hosts_as_dynar() instead.
2520 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2521 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2524 * Stabilize the parallel execution mode of user contexts
2525 * Introduce configuration variables to control parallel execution:
2526 - contexts/synchro: Synchronization mode to use when running
2527 contexts in parallel (either futex, posix or busy_wait)
2528 - contexts/parallel_threshold: Minimal number of user contexts
2529 that must be part of a scheduling round to switch to parallel
2530 execution mode (raw contexts only)
2531 * Fix bugs that prevented to use suspend/resume along with
2532 synchronization structures.
2533 * Fix bugs in process termination that lead to invalid memory access
2534 in very specific conditions.
2537 * Introduce a parallel mode for the models (controlled by surf/nthreads
2538 configuration item). In our tests, running the models in parallel
2539 never lead to any speedups because they are so fast that the gain
2540 of computing each model in parallel does not amortizes the
2541 synchronization costs, even when ultra fast futexes are used.
2542 This is released anyway because YMMV.
2545 * Performance boost by using a swag internally to compute the set of
2546 tasks that are finished and should constitute the return value of
2550 * Enable it by default now that it is considered rather stable.
2553 * Documentation of the tracing functions.
2554 * Performance gains when tracing categorized/uncategorized resource
2555 utilization by avoiding calls to get route when updating resource
2556 variables. LMM constraints are being used instead.
2557 * API changed to set task categories. Use MSG_task_set_category instead
2558 of TRACE_msg_set_task_category, and SD_task_set_category instead
2559 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2560 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2561 * Examples for link user variables added (see at examples/msg/tracing/)
2562 * Deprecated function TRACE_msg_set_process_category completely removed
2563 * Trace header updated according to the latest Paje file format
2564 * Tracing network lazy updates, no longer obligate users to use full updates
2565 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2566 * Experimental: let user code declare/set/push/pop application states for hosts
2567 * API changed to allow the manual creation of graph configuration files
2568 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2571 * Improve the API of Lua MSG bindings, using the Lua spirit.
2572 * Each simulated process now lives in its own Lua world (globals are
2573 automatically duplicated). It helps writing simulators. Will allow
2574 to run Splay programs within SimGrid in the future.
2575 * Add a Chord example in Lua, equivalent to the MSG one.
2578 * Start the implementation of a solution to express temporal
2579 properties, not only local assertions. This is still an
2580 experimental work in progress, stay clear from it to be safe.
2584 - Add new runtime parameters --help-logs and --help-log-categories
2585 to display information about supported logging parameters and
2587 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2588 don't exists anymore.
2589 * Mallocators: allow value NULL for the reset function.
2591 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2592 create homogeneous dictionaries, where all the elements share the
2593 same free function. Non homogeneous dictionaries will be
2594 deprecated in the next release.
2595 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2596 - Multi-level dictionaries are deprecated.
2598 - new function xbt_dynar_search_or_negative() that is useful when
2599 you have less than 2 million elements in your dynar and don't
2600 want of the extra complexity of catching exceptions when the
2601 element is not found.
2603 - Make xbt_os_thread module (for thread portability) public.
2604 Documentation is still to come, sorry.
2606 - Cleanups and simplifications to make it maintainable again.
2607 - Exotic features (such as memalign and valloc) were removed.
2608 - The metadata were extended and improved so that the
2609 model-checker becomes able to explore and inspect the heaps.
2610 - This may induce a performance drop when enable_model-checking is
2611 ON in cmake (even if it's not used in the simulation), but it is
2612 necessary at this point to get MC working.
2614 Turn model-checking OFF if simulation performance matters to you.
2615 Not enabling it at runtime is not enough, disable it in cmake.
2617 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2620 SimGrid (3.6.2) stable; urgency=low
2622 The "Not coding new stuff allows to polish old things" release.
2625 * New bindings to the NS3 packet level simulator (experimental)
2626 * Use the raw (efficient) execution contexts instead of the sysv
2627 (portable) ones when possible.
2628 * libpcre is now mandatory in any cases since not using it led to
2629 severe performance loss and possibly other issues
2630 * Update the XML platforms:
2631 - G5K: include the latest machine in Nancy
2632 - GridPP and LCG: new platforms
2633 * Documentation was partially updated, at least (more to come)
2635 Bug fixes, cosmetics and small improvements
2636 * Free terminated processes before the end of the simulation to avoid
2637 exhausting the memory of users having very dynamic amount of
2639 * Bug fix and cosmetics about canceling non-running tasks
2640 * Bug fix about the dot loader's issues when using libcgraph
2643 * Create an installer for windows with nsis (amd64 and win32)
2644 - Add an hello world project to illustrate simgrid project creation.
2645 - Embed libpcre into the Simgrid installer to avoid
2646 its compilation burden
2647 * The raw execution contexts should work on Apple now
2648 * Port to Windows 64 bits
2649 - Sysv contexts now have an implementation for this arch
2650 - GRAS communication features now support this arch
2651 * Drop support for borland compiler on windows
2652 - this code was not maintained, and we kinda depend on gcc nowadays
2653 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2654 * Fix portability issue on unstable ubuntu: linker became picky on
2657 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2659 SimGrid (3.6.1) stable; urgency=low
2661 The "Oops, we broke Macs too" release
2664 * Fixed contexts detection so that raw ones are used when possible
2665 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2666 leads to a strange error, with user code segfaulting sometimes when
2667 the generated code is not perfectly aligned (which is not
2668 controllable from the user side, depends on the amount of code)
2671 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2673 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2675 SimGrid-java (3.6) unstable; urgency=low
2678 * Split of every thing from simgrid v3.5 into a separate package.
2680 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2682 SimGrid (3.6) stable; urgency=medium
2684 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2687 * Bindings now constitute their own package, separated from the main one.
2688 Rationale: reduce our maintenance nightmare by reducing the module coupling
2689 They will soon be released on their own on gforge.
2691 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2692 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2694 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2695 * It's not quite deprecated for now because we have no replacement,
2696 but it may soon become the case.
2699 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2700 * Fortran: New user-level cache variable to store the rank of the running
2701 process. This improves performance by an order of magnitude.
2702 * C: New coccinelle script to automatically locate and modify global and
2703 local static variables.
2704 * Improved SMPI network model with a sender-side gap to account for multiple
2708 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2709 only say if a communication is finished, no matter whether it succeeded or
2710 failed. You can call MSG_comm_get_status() to know the status of a finished
2712 * New function MSG_task_dsend() to send a task and detach it. When a
2713 communication is detached, you are never notified of its success or failure
2714 and the memory is released automatically once it is finished. This function
2715 is useful when you don't care about the end nor the success of a
2717 * Change the prototypes of action replay. Sorry for inconvenience,
2718 but this is really more efficient this way (and to adapt your code,
2719 you just have to fix the initialization, that shouldn't be too long)
2720 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2721 ever managed to do anything useful with it.
2722 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2723 takes a dynar of communications. It returns immediately and gives the
2724 index of a finished communication (if any).
2725 * New example: a basic implementation of the Chord P2P algorithm.
2728 * New model for multi-core CPUs. You can now use the core attribute to
2729 precise the number of cores of a host. This is a basic model. Every
2730 process running on the host receives at most the power provided in
2731 the DTD (throughput<=power). Total throughput of process cannot exceed
2733 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2734 router linked by an up-link and a down-link (possibly asymmetrical).
2735 This kind of pattern allows to easily build last-mile model style platforms.
2736 Aggregating such patterns in a rule-based AS is thus the technique of
2737 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2738 * New model for Vivaldi routing. We transformed the Vivaldi network model
2739 into a Vivaldi routing model (based on the rule-based model). This allows to
2740 combine Vivaldi based latencies with last-mile platforms.
2743 * Added a check for NaN of IEEE754 infinite in the double entries of
2745 * Introduce a new context factory "raw", highly inspirated from the
2746 ucontext factory, but using manually crafted functions in assembly to
2747 do the work in an efficient manner.
2748 * Allow to change the used context factory at run time, not only at
2749 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2750 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2751 contexts at runtime (only with raw contexts or ucontexts).
2752 * Completely rewrote this module to allow parallel execution of user
2753 processes. Use --cfg=contexts/nthreads:N to execute user processes
2754 with N parallel threads (the default is 1, meaning no parallelism).
2755 * Allow to decide dynamically between sequential and parallel modes.
2756 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2757 processes in parallel only when their number is greater than or equal to P
2759 * Added a check for NaN of IEEE754 infinite in the double entries of
2763 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2764 won't output the state of processes when interrupted with Ctrl-C
2765 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2766 NULL-terminated array. This may solve backward compatibility issues
2767 due to the change to return type of SD_simulate. See also:
2768 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2769 * Add new macros with variable number of arguments.
2770 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2771 - in xbt/asserts.h: xbt_assert
2772 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2773 - in xbt/ex.h: THROWF and RETHROWF.
2774 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2776 * Change xbt_die() to accept a format string with arguments, just like printf.
2777 * New data structure: xbt_lib_t, like a dict but more general and with better
2781 * New configuration options
2782 Options triva/categorized and triva/uncategorized can be used to generate
2783 graph configuration files for Triva visualization tool.
2784 * Configuration option tracing/platform is renamed to tracing/categorized
2785 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2786 * New TRACE_user_link_variable interface:
2787 User provides the name of the link and the tracing variable to attach to it
2788 * the declaration of tracing categories must be done after the environment
2790 * simpler tracing interface, just one way to declare categories
2791 TRACE_category or TRACE_category_with_color, it is up to you
2792 * links in the trace file are again identified by their names
2793 * trace contains the full platform hierarchy exactly as declared using the ASes
2794 * Options tracing/msg/[task|process]:1 groups the process by hosts
2795 for both cases, tasks and processes must have names that are unique during the simulation
2796 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2797 * The experimental option tracing/msg/volume is deprecated
2798 its functionality may be reincorporated if needed
2800 The tracing generates a trace file with unordered timestamped events,
2801 because of the way the core simulator (surf) works. A script available
2802 at the tools directory (fix-paje-trace.sh) can be used to put the events
2803 in order. We have changed the tracing so it can generate ordered timestamped
2804 events in the final trace, but depending on the simulator (and how much time
2805 is simulated) that can lead to a huge memory utilization. It is deactivated
2806 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2808 Build Infrastructure
2809 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2810 - We are trying hard to keep the API stable, but it may happen that
2811 some things change (we're a research project after all, not a
2812 nuclear plant operating system). If such things should happen, you
2813 could rely on that macro to adapt.
2814 - current value: 30600 for 3.06.00, aka 3.6
2815 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2816 a number that can be compared with SIMGRID_VERSION.
2817 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2818 assertions and verbose/debug logging events are disabled at compile time.
2820 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2822 SimGrid (3.5) stable; urgency=medium
2825 * New feature to any SimGrid-based simulator: Model-Checking
2826 Check SIN#1 for more details.
2829 * New Model SMPI (three-interval linear regression for correction factors)
2830 See RR-7426, available at http://hal.inria.fr/inria-00527150
2831 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2832 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2833 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2834 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2835 * See also src/smpi/README
2840 - Tracing API changes: TRACE_start and TRACE_end should not be called
2841 by user-code. They are automatically called by simulators created
2842 with SimDAG, MSG and SMPI if the toolkit is compiled with
2843 tracing_enabled=ON. Categories declaration and utilization remain the
2844 same for MSG and SimDag.
2845 - A function was added to the tracing API to declare categories with
2847 - TRACE_category_with_color (char *category, char *color)
2848 where color must be in the following format
2849 "%f %f %f", red, green, blue
2850 and red, green, blue are float values in the interval [0, 1]
2851 - User can specify NULL as color parameter, or continue calling
2852 TRACE_category (cat)
2853 On that case, the tracing system will define random colors
2854 - The following command-line options are supported:
2855 --cfg=tracing/filename:msg.trace
2856 --cfg=tracing:1 (activate tracing, needed to use others)
2857 --cfg=tracing/platform:1 (categorized resource use)
2858 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2859 --cfg=tracing/msg/task:1 (task creation)
2860 --cfg=tracing/msg/process:1 (process creation, migration)
2861 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2862 --cfg=tracing/smpi:1 (SMPI interface tracing)
2863 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2864 - examples of examples/msg/tracing updated
2866 - DAXLoader and DOTLoader functions can generate tasks with categories
2867 - A new function to attribute a category to SD tasks:
2868 TRACE_sd_set_task_category (SD_task_t task, char *category)
2869 * Tracing the MPI interface implemented by SMPI
2870 - Collective operations are traced with states
2871 - Point-to-Point operations are traced with states/links
2872 - Tracing activated by a parameter "-trace filename" passed
2873 to smpirun during execution (considering that simgrid
2874 is compiled with tracing enabled)
2875 - To run the simulation with gdb, the simulator
2876 accepts --cfg=tracing/smpi:1 to trace SMPI
2877 - tesh files to check if smpi tracing is ok
2878 - See examples/smpi/NAS/DT-trace
2879 * GTNetS tracing re-worked
2880 - adaptation to the tracing system of GTNets to cope
2881 with modifications regarding the fullduplex mode
2882 - new tesh files to check if gtnets tracing is ok
2885 * Asynchronous communications through the functions:
2886 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2887 * New function: MSG_load_platform_script()
2888 to make possible using a lua script instead of XML files to set up platforms
2889 * New function: MSG_set_function
2890 to associate functions to processes, used when bypassing the parser
2891 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2893 Platforms: Add some more examples in examples/platforms
2894 * Grid'5000: see www.grid5000.fr
2895 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2898 * Change the XML format. This is a very important modification. SimGrid
2899 3.5 introduces a new hierarchical format based on the notion of
2900 Autonomous Systems. Compatibility with old format is ensured
2901 through the perl script provided in the install bin directory
2902 bin/simgrid_update_xml.
2903 It is now possible to build platforms with specific routing
2904 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2905 connect several platforms together. We will try to provide soon
2906 set of realistic platforms exploiting these properties (have a
2907 look at examples/platforms/ for the moment).
2908 * Take the opportunity of the XML format change to be a good XML citizen:
2909 rename link:ctn to link_ctn and similar changes (also dealt with by
2911 * Add a new routing scheme (rule-based) using regular expressions. It
2912 enables to have an extremely low memory footprint when the
2913 underlying routing is simple and can be compactly described. You
2914 need to have libpcre4-dev (perl regular expressions) installed if
2915 you want to use this routing scheme.
2916 * Revive the cluster TAG and allow to easily and efficiently (both in
2917 term of memory and speed) connect clusters together. Have a look
2918 at teshsuite/simdag/platforms/ to see how this can be done. With
2919 this tag, you can create clusters with thousands of tasks at no
2920 cost (have a look at examples/platforms/).
2921 Note: clusters are implemented as ASes, so there is no need for an
2922 enclosing AS tag if you have only one cluster in your platform.
2923 * Add new generic functions in the public interface that allows the user
2924 to call SURF 'create_resource' methods from your code (same
2925 functionality as the XML bypass mechanism but with a much lighter
2927 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2928 that uses a piecewise linear approximation to produce better
2929 results when exchanging small messages.
2930 * Add a new parameter to handle correctly full duplex link and account
2931 for interferences between uplink and downlink communications
2932 (activate with --cfg=fullduplex:1).
2935 * Rename the SD_READY (all dependencies are satisfied and task is
2936 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2937 dependencies are satisfied) state.
2938 This prevents a confusion between the notion of "ready to schedule"
2939 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2940 simulated" (SD_RUNNABLE) used by the simulation kernel.
2941 * Change the way a task is considered as ready. Instead of removing
2942 dependencies when a task is done, a counter is decreased. This way,
2943 it is always possible to reach ancestors thanks to the
2944 SD_taks_get_parents function (even after the end of the simulation.)
2945 * Change the return type of SD_Simulate from (SD_task_t*) into
2946 xbt_dynar_t. This function was in handling a dynar internally and
2947 converted it into a NULL terminated array for historical reasons.
2948 * New function SD_dotload(char*) to load a DAG described in dot
2949 format. This loader and the corresponding examples require the
2950 installation of the graphviz library.
2951 * Fix a bug in the management of tasks of size 0 in the surf network
2952 models. This problem was only visible with SIMDAG and you should
2953 thus disregard results produced with earlier versions if you
2954 relied on this feature (some tasks were blocked because of this).
2955 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2956 even though your applications were doing only point-to-point
2957 communications and sequential computations. Now you can really use any
2958 model you want (of course, if you create real parallel tasks, which are
2959 not implemented in most models beside ptaskL07, this will abort).
2960 * Add an example that schedules a DAX on an heterogeneous platform
2961 using a Min-Min strategy.
2962 * New function SD_workstation_get_current_task() that returns the kind
2963 of task currently running on a workstation in the sequential access
2965 * Raise some warnings when unexecuted tasks remains at the end of the
2966 simulation. This is usually caused by cycles in the DAG.
2969 * New function: SIMIX_process_set_function() called by MSG_set_function
2970 * Change the underlying waiting queue in semaphores so that a process
2971 can wait on several of them simultaneously (as in waitany).
2972 * Fix the way to handle tokens in semaphores so that all access patterns
2973 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2974 * kill the dirty pimple SIMIX_message_sizes_output()
2975 Please use (proper) visualization instead
2978 * New data container: setset (set of sets of elements)
2979 * New module: mmalloc (mapped malloc, allowing to have several
2980 independent segments of malloc)
2981 * New function: xbt_dict_cursor_set_data()
2982 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2983 * New function: xbt_dynar_is_empty()
2984 * New function: xbt_fifo_get_last_item()
2985 * Fix xbt_dynar_shrink(): use the right element size.
2986 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2987 uninitialized areas during expand.
2988 * Fix semaphores: previous implementation was severely broken.
2989 * Use library init/fini functions for our initialization.
2990 - you can use logs and other feature as soon as you want in your
2991 code (even before the xbt_init / MSG_init)
2992 - xbt_exit is now a no-op and produce a warning when used.
2995 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2996 benefit from the latest improvement to the simulation kernel.
2997 * Kill measurement sockets for now. If you rely on them, sorry. This
2998 release is not for you. This feature will be reintroduced in the
2999 future, but we cannot delay the release any further.
3000 * New function: gras_msgtype_get_name().
3001 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
3002 * Fix (at last) the pmm example: it should not randomly fail anymore.
3004 Build chain: bug fixes and overall polishing
3005 * Cmake is now stable enough. Hence, we killed the autotools.
3006 * Port to windows ( TM :)
3007 * Fix the 'make install' target.
3008 No need to use 'make install-simgrid' anymore
3009 * Introduce a 'make dist' target compiling a *source* archive
3010 'make package' compiles a binary archive
3011 * Compile java files only on need
3012 * Add --cd and --setenv command line options to tesh
3013 * Out of source builds are not fully supported yet, but we are close
3014 * Enable supernovae and optimization flags by default for our users
3017 * Add layer to set up environment directly from lua, without XML.
3018 * The effect of gras_stub_generator can be achieved through
3019 lua too (check examples/gras/console/ping_generator.lua)
3021 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
3023 SimGrid (3.4.1) stable; urgency=low
3025 The "Polishing easter eggs is probably a good idea" release.
3026 This is a bug fixes release only.
3030 * Fix a bug preventing the tasks from begin garbage collected.
3033 * Fix a bug occurring when a host involved in a communication fails.
3034 This was not detected properly by the other peer involved in the
3035 communication. Now, it's reported as a network error.
3038 * Warn the user about loop dependencies in data flow of DAX files
3039 * Obey the control-flow dependencies of DAX files
3042 * Add option "enable_smpi" allowing to not compile SMPI.
3043 Probably useful for the (Mac) users experiencing a build error here
3044 * Improve the detection of lua5.1 and ruby1.8
3046 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
3048 SimGrid (3.4) stable; urgency=low
3050 The "Easter in Cargese" release. Also known as (major changes):
3052 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
3053 ~> bindings were greatly improved
3054 ~> new tracing infrastructure for better visualization introduced
3056 * the "Welcome to configury modernity" release.
3057 ~> we switched from autotools to cmake, and improved our cdash
3060 A more detailled list of changes follow (full detail in svn log).
3062 Java Bindings: Various Cleanups
3063 * (install java-gcj-compat-dev on debian-like to use them)
3064 * Remove put/get: no need to export deprecated interface in Java
3065 Use send/receive instead.
3066 * Cleanup the examples and add a README per directory
3067 * Remove example autoDestination (that's the only way to go now)
3068 * Remove example explicitDestination (was a plain copy of basic)
3069 * Make JniException a runtime exception, so that there is no need to
3070 declare the fact that you may encounter such a beast. I guess that
3071 nobody will ever want to survive such error.
3072 * Create specific errors for each MSG case of failure:
3073 host failure, transfer failure, timeout, task cancelled
3074 * Cleanup the exceptions that may get thrown by each function
3075 * Other internal cleanups in Java bindings. Performance still bad :/
3076 Ruby and Lua Bindings: create them
3077 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
3078 * That's new and great, you should try them out.
3079 Same functionalities than Java bindings, only even less polished
3081 * Kill the useless "rate" argument of SD_task_get_execution_time()
3082 Everyone used to provide -1 as a value, it was not used, and the
3083 semantic of a possible use wasn't even clear.
3084 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
3085 either as comm costs or compute costs to mean that there is no
3086 such thing for that specific task.
3087 * Add a SD_task_set_name() function
3088 * Fix SD_task_unschedule() on typed tasks
3089 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
3090 * In DAX loader, accept useless 'level' attributes to <job> since
3091 LIGO DAGs have them (seem to be to ease graphical representation).
3093 * Add an example masterslave_mailbox.c using send/receive and not
3094 the deprecated put/get interface.
3095 * Kill the MSG_paje_output() function. It's a noop since 2 years.
3096 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
3098 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
3099 (declare MSG_USE_DEPRECATED to still have the old name)
3100 * Add a MSG_task_set_data() function
3101 * About trace replay (see examples/msg/actions):
3103 - Allow to work with splitted trace files for each process
3104 Give the specific trace file as argument of each process,
3105 and call MSG_action_trace_run(NULL)
3106 You can still have one merged file for all processes.
3107 - Fix implementation of collective operations
3108 * Allow task_execute() on 0-sized tasks (closes #10063)
3110 * This is the first release of SimGrid where SMPI is not considered
3111 beta anymore (even if some corners should still be improved)
3112 * Port over the new SIMIX_network submodule (internal refactoring)
3113 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
3114 * Implement more missing elements of the standard:
3116 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
3118 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
3119 MPI_Scan() MPI_Get_processor_name()
3120 - Added implementation of missing case for Alltoall (warning: it's
3121 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
3122 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
3123 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
3124 SMPI_MPI_Allgatherv()
3125 * Bug fixes include:
3126 - MPI_Waitsome() was broken
3127 - Allow relative includes in smpicc
3128 - Command line cfg argument 'reference_speed' was ignored...
3129 - Some functions did not properly lead to auto-benching of user code
3130 - smpicc passes -O2 by default (just like openmpi one)
3132 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
3133 * Bug fixes about timeouts during communications
3134 * add SIMIX_message_sizes_output() as a pimple to write to file the
3135 amount of messages per size. Use gnuplot to get histogram.
3136 Pimple because that's the only user-visible function of simix,
3137 defined directly in xbt.h (irk, sorry)
3139 - Add a SIMIX_sem_get_capacity() function
3140 - Fix interactions with processe resume/suspende
3141 - release_forever() was stupidly broken
3142 - Fix SIMIX_display_process_status() for processes in a semaphore
3143 - Make SIMIX_sem_block_onto() user-visible
3144 * Refactoring context stuff:
3145 - Use pseudo-OOP for better modularity
3146 - reimplement SIMIX_process_kill() without process_schedule() so
3147 that the latter can take as invariant that it is called from
3149 - Merge context_start into context_new for sake of simplicity
3151 * Add a Vivaldi network model, coded live during SUD'10 ;)
3152 * Rename configuration variables to start a hierarchy:
3153 o cpu_model -> cpu/model
3154 o network_model -> network/model
3155 o workstation_model -> workstation/model
3156 * New configuration variables:
3157 o network/bandwidth_factor: correction to bandwidth
3158 o network/latency_factor: correction to latency
3159 o netwotk/weight_S: correction to the weight of competing streams
3160 * Add a long description to the models, that users can see with such
3161 argument on the command line: --cfg=cpu/model:help
3162 * --help-models display the long description of all known models
3164 * config: add the ability to set a default value after registration
3165 Does not override any previously set value (e.g. from cmd line)
3166 * dict: allow to have integer key and data.
3167 When so, you need to use the following functions
3168 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
3169 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
3170 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
3171 In contrary to regular dicts, the key is not malloced before copy.
3172 Mixing scalar and regular elements in the same dict is not tested
3174 * Allow to use xbt_dynar_shrink() to expend the dynar instead
3175 Tracing for Visualization:
3176 * SimGrid is now instrumented in order to generate a trace file for
3177 visualization analysis: to use it, need to compile SimGrid with the
3178 "tracing" option enabled, and instrument the program using SimGrid with
3179 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
3180 (among other functions).
3181 * The instrumentation only traces the platform utilization for now
3182 * Documentation to use the tracing functions and how to analyze the
3183 traces with the Triva tool is written.
3184 * More information about: SimGrid FAQ (in the section Tracing Simulations
3187 * We moved to cmake as default build system. Autotools support will
3188 be dropped soon. Check the FAQ for more info about how to use it.
3189 * Greatly improved our cdash/ctest interactions
3190 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
3191 * Added memory checking tests with valgrind; lot of memleak fixing.
3192 This may be the first release of simgrid with so few memory issues
3193 * Added code coverage tests.
3194 Our coverage is still improvable, but at least we see it on cdash.
3196 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
3198 SimGrid (3.3.4) stable; urgency=low
3200 The "Desktop Grid needs love too" release (also called Xmas release).
3202 Models improvements:
3203 * Major speedup in the maxmin system solving by using lazy evaluation
3204 Instead of solving completely the maxmin system at each iteration,
3205 only invalidate (and recompute) the modified parts.
3206 This new feature is enabled in default models but you can try to
3207 turn it on with "--cfg:maxmin-selective-update=1" for other models.
3208 * Cas01 IMproved as default CPU model
3209 This CPU model is the same Cas01 model, but it uses the
3210 maxmin-selective-update flag and a heap structure to manage
3211 actions on SURF kernel.
3212 It reduces the complexity to find the next action to finish and,
3213 consequently, it's faster than the old Cas01.
3214 This is the new default CPU model (Cas01).
3215 * Rename the old Cas01 model to Cas01_fullupdate
3216 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
3217 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
3218 * CpuTI (CPU Trace Integration)
3219 A new CPU model whose objective is simulate faster when using
3220 availability trace files.
3221 Instead of using a full featured, over engineered maxmin system for
3222 CPU modeling, this model does the pre-integration of traces files
3223 to calculate the amount of CPU power available, and so, executes
3224 faster than the old CPU models.
3225 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
3226 * Use LV08 as default network model since it gives better accuracy
3227 for small messages and shouldn't change things for big ones.
3228 Use --cfg=network_model:CM02 to get the previous behavior.
3231 ******************************************
3232 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
3233 ******************************************
3234 * The new CPU model may changes simulations!
3235 The point is that events occurring at the exact same timestamp
3236 are not scheduled in the same order with the old and new
3237 version. This may be enough to completely change the execution
3238 of simulations in some cases.
3239 * The new network model will change simulations!
3240 This new model is more realistic than the previous one, so you
3241 should consider redoing your old experiments with this model.
3242 Sorry for the inconvenience.
3245 * Introduce the supernovae compilation mode
3246 When compiled that way, the whole SimGrid (or almost) is put in a
3247 single compilation unit and compiled in one shoot.
3248 This is to help gcc which has difficulties to inline stuff from one
3250 The speedup seem to be above 15%, althrough more tests are needed on
3251 amd64 to confirm that gain.
3254 * Port of MSG's mailbox on top of SIMIX network
3255 The put/get mechanism was greatly simplified on the way.
3258 * New SIMIX network module. Provides:
3259 - Mailbox: rendez-vous mechanism to find with who you want to speak
3260 - Synchronous send/recv: easier and hopefully faster since the
3261 logic is handled in the maestro process directly now
3262 - Asynchronous send/recv: you dreamt of it? It's here now
3263 Too bad that nobody cared enough to propagate the change to MSG.
3264 * Add semaphores as SIMIX synchronization mechanism.
3267 * new function SD_daxload(char*) to load a DAX file
3268 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3269 * Introduce typed tasks. Specify its kind and cost at creation.
3270 At scheduling, just give where it should be placed, and the cost
3271 for each involved resource is automatically computed.
3272 Existing constructors so far (more to come of course):
3273 - SD_task_create_comm_e2e() for end-to-end communication
3274 - SD_task_create_comp_seq() for sequential computation
3275 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3276 * new function SD_task_dump() for debugging display
3277 * new function SD_task_dotty(task,FILE*) writing to file the info
3278 about the task in dotty format
3279 * SD_task_dependency_exists() can now cope with having one of its
3280 arguments NULL. If so, it tests whether the other argument has any
3282 * Add getters on list of preceding/following tasks:
3283 SD_task_get_parents(task) and SD_task_get_children(task)
3284 * Add getters on amount of workstations and list:
3285 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3286 * Add getter on task kind: SD_task_get_kind(task)
3287 * Update the start_time and finish_time of tasks on completion/failure
3288 * Bugfix: Remove task from state swags when destroyed
3291 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3294 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3295 dynar with one separate thread per value of the dynar.
3296 * Change the prototype of xbt_thread_create(), sorry.
3297 Added a boolean parameter indicating whether we want to join this
3298 thread (used in SG only for now)
3299 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3302 * GTNetS wrappers should now be usable again (and betterly tested too)
3303 * Fix a major regression from 3.2 where the timeout provided to
3304 MSG_task_put_with_timeout() was used as absolute time before which
3305 the comm should be done.
3306 * Start to fix the <cluster> tag.
3307 - Internal links should be good now (beside of the loopback, which
3308 use the private link instead)
3309 - paths to the external world is still rather broken
3310 - the <route:multi> tag is just broken. Actually that's brain-dead.
3311 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3312 to make it less stupid
3313 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3314 * Fix a source-level compatibility glitch from 3.2: after defining
3315 MSG_USE_DEPRECATED, you can use the old name
3316 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3317 * Allow to compile from the SVN with automake 1.11
3318 * Fix some problems when using the "start_time" tag in deployment XMLs.
3319 * Fix #8569: XBT/synchro.h has redundant declarations
3320 * Fix #8563: MSG return values and exceptions
3321 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3322 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3323 Thanks to Robson Peixoto.
3324 * Fix a vicious bug in dictionaries inducing that some elements were
3325 not freed on xbt_dict_free()
3327 Portability report of this version:
3328 * Main portability targets:
3329 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3330 on (amd64/i386/ia64)
3331 - mac leopard on i386
3332 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3333 but nothing critical.
3334 * Other platforms: windows, AIX and others were not tested for this release
3336 Timing report of this version:
3337 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3338 scenario parameters). From 8h to a few seconds in desktop grid settings.
3339 * Supernovae brings about 25% speedup on i386.
3341 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3343 SimGrid (3.3.3) stable; urgency=low
3345 The "Need for Speed" release.
3347 The timings done to validate the 3.3.2 were faulty.
3348 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3350 The problem was a conversion from a manually handled vector to
3351 xbt_dynar_t on the critical path.
3352 xbt_dynar_foreach calls functions, inducing stack management crap.
3354 We inlined these functions and xbt_dynar_foreach is now breath taking.
3355 We also inlined xbt_swag_belong on the way.
3357 Here are some approximate speedup measurements (on master/slaves
3358 simulations lasting between 10s and 20s each):
3359 3.3.1 -> 3.3.2: about same performance
3360 3.3.2 -> 3.3.3: 40% speedup
3361 3.3.1 -> 3.3.3: 40% speedup
3362 3.3.1 with inline patch -> 3.3.3: 30% speedup
3364 Our reading is that the refactoring which occurred in 3.3.2 made us
3365 suffer much more from the xbt_dynar_foreach low performance, but
3366 once we solved this, this refactoring proved to be very performance
3367 effective. From the 40% speedup, somehow, 10% are due to the
3368 inlining and 30% to the refactoring.
3370 That's a pitty that gcc cannot inline functions placed in other files
3371 alone. We have to choose between:
3372 - break the encapsulation (by putting private data structures and
3373 accessors in headers files to help gcc)
3374 - live with low performance
3375 - switch to a decent compiler such as icc (not quite possible).
3377 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3379 SimGrid (3.3.2) stable; urgency=low
3381 The "Simplicity does not preceed complexity, but follows it" release.
3383 The main contributors of this release were (lexical order):
3384 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3387 * Extract the routing logic into its own object.
3388 (was dupplicated in network.c and workstation_LV07.c;
3389 Allows to implement other ways of storing that info)
3390 => kill now useless network_card concept
3391 - Use dynar to represent routes (instead of void** + int*)
3392 - kill link_set (use surf_network_model->resource_set instead)
3393 - Add a command-line option to choose the routing schema to use
3394 - Add three new models:
3395 * Floyd (shortest path computed at initialization)
3396 * Dijikstra (shortest path recomputed all the time)
3397 * Cached Dijikstra (shortest path computed on need)
3398 All these models where contributed by Silas De Munck, and are
3399 described in his ICCS09 paper.
3401 * Simplify model declaration
3402 (less redirections, less function to write when defining a model)
3403 - Factorize stuff between models:
3406 surf_model_resource_set(model)
3407 surf_model_resource_by_name(model, name)
3408 - Unify the types of models in s_surf_model_t (using an union)
3409 - Embeed fields of common_public directly into s_surf_model_t
3410 - Rename model methods:
3411 action_free ~> action_unref
3412 action_change_state ~> action_state_set
3413 action_get_state ~> action_state_get
3414 - Change model methods into functions:
3415 (model)->common_public->action_use ~> surf_action_ref
3417 * Implement a generic resource; use it as ancestor to specific ones
3418 (allows to kill duplicated code in models)
3419 Drawback: timer command don't need no name nor properties;
3420 workstation_CLM03 don't need no properties
3421 (but I guess we can live with those few bytes wasted)
3423 * Improve the action object model
3424 - implement a constructor avoiding dupplicated code about field
3425 initialization in generic_action part.
3427 * Kill the SDP model: it has an external dependency, is deprecated
3428 in flavor of modern lmm models, and didn't compile since a while
3431 * Relocation of the context module from XBT to SIMIX.
3432 (the context were decoupled from the simix processes, duplicating a lot of code)
3433 => a lot of code was factorized
3434 - less overhead is introduced during scheduling
3435 - simpler API for the context factory
3436 - the logic for process creation,destruction and manipulation was simplified
3437 * Simplification of the s_smx_process_t data structure.
3438 => accesing the simix level data associated to a process is faster now,
3439 and the code is a lot more readable.
3442 * Implement some more MPI primitives:
3443 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3444 -implementation: Bcast: flat or 2-ary tree (default),
3445 Barrier: 4-ary tree,
3447 Allreduce: Reduce then Bcast
3448 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3449 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3450 Alltoallv: flat tree, like ompi
3452 * Add support for optimized collectives (Bcast is now binomial by default)
3453 * Port smpirun and smpicc to OS X
3456 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3457 Shout out if you used it.
3460 * Display the list of still queued messages in SG mode when existing
3464 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3465 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3466 * Add xbt_str_from_file(FILE*)
3467 * Add xbt_dict_get_key achieving a linear reverse search
3468 * Remove the context module
3470 Portability report of this version:
3471 * Main portability targets:
3472 - Linux(debian)/x86/context
3473 - Linux(debian)/x86/pthreads
3474 - Linux(debian)/amd64/context
3475 - Linux(debian)/amd64/pthreads
3476 On these, we still have the eratic breakages of gras/pmm and
3477 amok/saturate_sg reported in previous version. We still think
3478 that the tests are the cause of the fault, not the tested code.
3480 - Mac OSX Leopard/x86/context
3481 Still false negative in tesh autotesting.
3482 Smpi still fails, but this time because readlink does not accept -f
3483 Everything seems to work properly beside of that.
3486 - AIX version 5.3 (only tested contexts this time)
3487 Smpi still fails there because mktemp is not installed.
3488 Everything seems to work properly beside of that.
3490 I managed to compile it for the first time, but several breakages.
3491 Won't delay the release for this exotic platform.
3493 * Windows: it's still lagging behind. If you want to help, please
3496 Timing report of this version:
3497 This version seem to be more than 5% faster than 3.3.1 (on linux
3498 64bits with contextes). The gain is less than expected, we are
3499 investigating this for next release.
3501 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3503 SimGrid (3.3.1) stable; urgency=low
3506 * Implement a --cfg-help to show existing configuration variables
3507 * Build chain do not require doxygen in maintainer mode
3510 * fix a bug on struct sizeof computation, which prevented the
3511 exchange of arrays of structs in some conditions
3512 - added a regression test about this in datadesc_usage
3513 * Allow the exchange of 0-long dynamic vectors.
3514 - for that, use -1 as indicator of dynamic size instead of 0
3515 - This implied to change any size from unsigned long to long,
3516 reducing a bit communication abilities, but I guess that with
3517 64bits being quite common, this is more than enough.
3518 - This also induce a protocol change, thus bumping network protocol
3519 version from 0 to 1 (if we have external users, we have to get
3520 clean on that point too ;)
3521 - added two regression tests about this in datadesc_usage
3522 * Be more verbose when propagating local exceptions
3523 This helps debugging.
3524 * Display the status of simulated processes when receiving SIGINT in
3528 * Allow to control the simulation from a trace file.
3529 New functions MSG_action_register() and MSG_action_trace_run()
3530 The first one allows to associate a function execution to each
3531 kind of action while the second one parses a trace file and
3532 triggers the corresponding actions within the system.
3533 For now, only a toy example is provided in examples/msg/actions
3534 * Add an example of process migration in examples/msg/migration
3535 * Fix a bug in task exchange which broke MSG_task_get_sender()
3536 Add a teshsuite regression test for that.
3537 [Bug: if MSG_task_get_sender() is called after sender exit,
3539 * Fix a bug which prevented suspend/resume to work properly
3540 * Display the status of simulated processes when receiving SIGINT
3541 This fixes a regression of v3.3. due to the introduction of SIMIX
3542 * Bug fixing in failure management:
3543 - trace could not start by a failure at time 0
3544 - failure during communications were not working
3547 * Add SIMIX_process_set_name() to change the name of the current
3548 process in the log messages.
3549 * Store smx_hosts in a dict since we only retrieve them by name
3550 * Move the configuration infrastructure to surf
3553 * Move the configuration infrastructure to surf
3556 * Massive internal cleanups:
3557 - Store internal structures on processes instead of hosts (allows
3558 to have more than one process per host, in addition of being more
3560 - Cleanup the initialization/finalization process
3561 - Kill a whole bunch of unneeded synchronization:
3562 processes run in exclusive manner within the simulator
3563 - Move queues from global tables to process data fields
3565 - now accept -platform and -hostfile arguments
3566 - Pass the right rank value to processes according to the hostfile
3567 * Compile the examples by default, and use them as regression tests
3568 * Implement MPI_Wtime()
3569 * Change the reference speed to a command line option
3572 * TCP_gamma can now be specified as command line option using
3573 --cfg=TCP_gamma:10000000.0
3574 * Change the --surf-path cmd line option into --cfg=path:
3577 * Also include strbuff from xbt.h public header
3578 * xbt_ex_display(): do not free the exception after displaying
3579 This allows to do more with the given exception afterward.
3580 Users should call xbt_ex_free() themselves.
3584 Portability report of this version:
3585 * Main portability targets:
3586 - Linux(debian)/x86/context
3587 - Linux(debian)/x86/pthreads
3588 - Linux(debian)/amd64/context
3589 - Linux(debian)/amd64/pthreads
3590 These targets fail about 1/10 of times on gras/pmm, but we believe
3591 that this is because of the test, not because of simgrid.
3592 amok/saturate_sg fails even more rarely, and the test may not be
3595 - Mac OSX Leopard/x86/context
3596 The test suite still spits tons of errors because some obscure
3597 force prevents us from removing the temporary directories
3598 arguing that they still contain some metadata I've never heard of.
3599 Smpi fails because seq is not installed.
3600 Everything seems to work properly beside of that.
3603 - AIX version 5.3 (both contexts and pthread)
3604 Smpi still fails there because mktemp is not installed.
3605 XML inclusions seems rosty on AIX.
3607 * Windows: it's still lagging behind. If you want to help, please
3610 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3612 SimGrid (3.3) stable; urgency=high
3616 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3617 [Malek Cherier & Mt]
3619 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3622 Until now, GRAS were using MSG as an interface to SURF. It was
3623 quite difficult because both interface have several differences
3624 (MSG channels vs GRAS sockets were the most notable point).
3626 This also opens the gate to SMPI (which should occur soon) and speed
3627 up simulations by to 40% (even if it were not the main goal).
3629 **************************************
3630 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3631 **************************************
3632 The point is that events occurring at the exact same timestamp are
3633 not scheduled in the same order with the old and new version. This
3634 may be enough to completely change the execution of simulations in
3635 some cases. Sorry for the inconvenience.
3637 * Cleanup and upgrade the XML format to push further scalability
3638 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3640 * Improve the testing infrastructure with tesh. Now a very large part of
3641 the code is tested not only by being run but also by checking that the
3642 output match an expected output [Mt].
3644 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3645 is really less memory-demanding, which should allow you to use
3646 larger files in SimGrid [AL].
3648 * Inform valgrind about our contextes, so that it becomes usable
3649 with the default (and more effecient) version of SimGrid
3650 [contributed by Sékou Diakite, many thanks]
3653 * Introduce a listener thread in charge of receiving incoming
3654 messages from the network. It allows to overlap communication and
3655 computation but most notably, it removes some stupid deadlocks due
3656 to the fact that so far, a process could not send and receive at
3657 the same time. This made most non trivial communication schema
3659 * Convert the PIDs from long int to int to match the MSG ones (and
3660 linux ones too) [Mt]
3661 * New function: gras_agent_spawn() to launch a new process on
3662 current host. Only working in simulation for now. [Mt]
3663 * New function: gras_os_hostport() returning a constant form (ie,
3664 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3667 * Make the backtrace of exceptions more human readable [Mt]
3668 * New module: xbt/str [Mt]
3669 a ton of string utility functions (split, join, printf to a newly
3670 allocated buffer, trim, etc)
3671 * New module: xbt/hash [Mt]
3672 SHA1 hashing algorithm (more to come if needed)
3673 * New module: xbt/synchro [Mt]
3674 synchronization tools (mutex and conditions) working the same way
3675 in simulation and in real life (mainly useful for GRAS, but not
3677 * New module: xbt/queue [Mt]
3678 classical producer/consumer synchronization scheme
3679 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3680 (using the classical functions will get serialized) [Mt]
3681 * Make dictionary internal table dynamic. No need to specify its size
3682 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3684 * Make sure the log channels are organized as a tree under windows
3685 (because of ANSI C compatibility issue, any channel were child of
3686 root directly) [Mt].
3689 * Cleaned many thing in surf and fixed a few bugs [AL].
3690 * Add a nice command line configuration mechanism to compose models [AL].
3691 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3692 the previous one (KCCFLN05). It relies on something that looks like
3693 a max-min sharing mechanism but cannot be written as such. A new solver
3694 was thus designed [AL].
3695 * Add a new solver to lmm. Based on Lagrange optimization and
3696 gradient-based descent, it enables to efficiently maximise systems s.a
3698 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3700 This solver enables to propose two new network models for TCP Reno and
3701 TCP Vegas based on Low's work. These models still need to be fully
3702 tested though [Pedro Velho].
3705 * Bug fix in SD_simulate. Now the time bound given as argument is
3707 * Use the new parallel task model (ptask_L07) as default.
3708 * Use the SURF command line configuration mechanism.
3709 * 0-size tasks (for synchronization) should now work.
3711 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3713 SimGrid (3.2) stable; urgency=high
3717 We still experience issues on this platform, but we believe that at
3718 least MSG is usable.
3720 GRAS API BREAKAGE (for simplification purpose, sorry):
3721 * the gras_msgtype_by_name is not used anymore. Instead of
3722 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3723 you can write (and must)
3724 gras_msg_send(toserver, "request", &request);
3725 - If you still want to pass a gras_msgtype_t to the function (to cache
3726 the type and avoid the lookup time), use the gras_msg_send_() variant.
3727 - Impacted functions:
3728 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3729 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3730 * The callbacks are now expected to return 0 when everything went well
3731 (just like the main() function)
3733 GRAS new features and improvements:
3734 * New module mechanism where user code can use per process globals [Mt]
3735 This is similar to gras_userdata_*() functions, but for libraries. It
3736 factorize some code developped over and over in the examples and AMOK.
3737 It has still to be documented and used (only amok/peermanagement is
3739 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3740 loss when they were small enough to fit into the buffer and sent quickly
3741 enough so that they can all get received in one shoot.
3742 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3743 if not found. Use the *_or_null() variant for the old semantic.
3744 * In gras_msg_handle, do not discard messages without callback.
3745 They are probably messages to be explicitly awaited later (ie, proofs of
3746 mis-synchronization in userland since they are sent before being awaited)
3747 No big deal usually.
3748 * gras_socket_meas_send/recv: semantic changed!
3749 The numerical arguments used to be (1) the total amount of data to send
3750 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3751 messages. This was need for the fool willing to send more than MAXINT
3752 bytes on quite fat pipes.
3755 * Do really rename the hostmanagement module to peermanagement. [Mt]
3756 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3757 but this is rather new and this was documented in the module
3758 documentation (poor excuses, I admit)
3759 * Bandwidth measurement semantic changed! This follows the changes to
3760 gras_socket_meas_send/recv explained above.
3763 * A sequential mode has been added to the workstations. When a workstation
3764 is in sequential mode, it can execute only one task, and the other tasks
3765 are waiting in a FIFO. [Christophe Thiery]
3768 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3769 model for SIMDAG. [Christophe Thiery]
3770 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3771 equal to 0, which caused some bad behaviors in
3772 saturated_constraint_set_update. I now use a threshold mechanism like in
3776 * When running manually src/testall, you select specific units [Mt]
3777 testall is the result of our cunit mechanism, and should replace all
3778 the scripty thingy around since bash don't run easily on billware.
3780 * A mallocator system has been added. [Christophe Thiery]
3781 Mallocators allow you to recycle your unused objects instead of freeing them
3782 and allocating new ones.
3784 Documentation update:
3785 * FAQ reworking + New FAQs:
3786 - "Valgrind spits tons of errors!" [Mt]
3787 - "How to repport bugs" [Mt]
3788 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3789 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3790 - Communication time measurement within MSG [AL]
3791 - I experience weird communication times when I change the latency [AL]
3792 * GRAS tutorial [Mt]
3794 - an introduction to the framework and to the used communication model
3795 - an initiatic tour introducing the most proheminent features:
3797 . Lesson 0: Installing GRAS
3798 . Lesson 1: Setting up your own project
3799 o Part 2: Message passing
3800 . Lesson 2: Exchanging simple messages
3801 . Lesson 3: Passing arguments to the processes (in SG)
3802 . Lesson 4: Attaching callbacks to messages
3803 . Lesson 5: Using globals in processes
3804 . Lesson 6: Logging information properly
3805 . Lesson 7: Using internal timers
3806 . Lesson 8: Handling errors through exceptions
3807 . Lesson 9: Exchanging simple data
3808 . Lesson 10: Remote Procedure Calling (RPC)
3809 . Lesson 11: Explicitly waiting for messages
3810 . Recapping of message passing features in GRAS
3811 - A HOWTO section containing:
3812 o HOWTO design a GRAS application
3813 More are due, of course. They will come latter. In the meanwhile, you can
3814 check the examples which are still here.
3816 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3818 SimGrid (3.1) stable; urgency=high
3822 There was a stack corruption somewhere, visible only when optimizing
3823 with these versions. [Vince]
3826 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3827 module that enabled to play with Directed Acyclic Graphs. It is built
3828 directly on top of SURF and provides an API rather close to the old
3829 SG. Some old codes using SG are currently under rewrite to check that
3830 all needful functions are provided. [Christophe Thiery]
3833 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3834 extension of the classical CLM03 model that gracefully handles
3835 failures. This is now the default model for MSG and GRAS. It doesn't
3836 handle parallel tasks yet though. [AL]
3837 * Bug fix: Weights were not correctly set in the network part.
3838 WARNING: This may have resulted in incorrect results with simulations
3839 where there are more than one flow on a given link. [AL]
3842 * After a (long ?) discussion on simgrid-devel, we have decided that the
3843 convention we had on units was stupid. That is why it has been decided
3844 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3845 WARNING: This means that all previous platform files will not work as
3846 such with this version! A warning is issued to ask users to update
3848 A conversion script can be found in the contrib module of the CVS, under
3849 the name contrib/platform_generation/surfxml_update.pl [MQ]
3852 * Bug fix: Processes were started in reverse order, wrt deployment file.
3853 WARNING: if your code relies on this bug, please fix it. [AL]
3854 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3855 being executed on two different locations. [AL]
3856 * Bug fix: Failures are now better supported thanks to Derrick's tests
3857 (there was many failure situations I hadn't thought of and that weren't
3858 correctly handled). [AL]
3859 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3863 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3864 Exception occurring on server-side are propagated back to client (!).
3866 API CHANGE: the callback changed their prototype. Change:
3867 int my_handler(gras_socket_t expeditor, void *payload_data) {
3869 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3870 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3872 * New! function: gras_msg_handleall to deal with all messages arriving
3873 within a given period.
3874 * New! function: gras_socket_server_range to get a server socket in a
3875 range of port numbers (ease to avoid port number conflicts) [MQ]
3876 * New! gras processes display their backtrace when they get a SIGUSR1
3877 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3878 Sweet to debug RL processes [MQ]
3882 - Do not force experiment sizes to be expressed in kb, or it becomes
3883 impossible to measure the latency this way (needs one byte-long tests)
3884 WARNING: this changes the amok_bw_* function semantic. [MQ]
3885 - Implements the link saturation stuff. [MQ]
3886 * Peer management module:
3887 New! module factorizing code that we wrote over and over [MQ].
3890 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3891 - Test units are placed directly into the library code, they get extracted
3892 automatically and placed into the src/testall binary.
3893 - Convert most of the XBT tests to this system.
3894 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3895 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3896 live in the same C code. This required to change a little bit the API
3897 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3898 * New module: sparse graph structure with basic algorithms (this is work
3899 in progress and the API is not considered to be frozen yet). [AL]
3900 * Display more information on backtraces: source line & function names are
3901 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3902 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3903 * Reworked a little bit some #include statements to load only required
3904 headers. Some user code that relied on SimGrid to include stdlib or
3905 stdio may need to include it by themselves. [AL]
3906 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3908 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3909 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3910 but I guess that almost nobody used those parts. [MQ]
3912 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3914 SimGrid (3.0.1) stable; urgency=low
3917 * Unfortunately, I had missed 5 misnamed functions:
3918 xbt_fifo_item_t xbt_fifo_newitem(void);
3919 void xbt_fifo_freeitem(xbt_fifo_item_t);
3920 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3921 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3922 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3923 They're now deprecated. Please use their new versions:
3924 xbt_fifo_item_t xbt_fifo_new_item(void);
3925 void xbt_fifo_free_item(xbt_fifo_item_t);
3926 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3927 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3928 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3930 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3931 * Documentation: xbt_log module unmercifully reworked [MQ]
3932 * Bugfix: there was a problem with the ending of contexts with
3933 the pthread backend. It caused some weird deadlock or behavior
3934 depending on the pthread implementation. [AL]
3935 * Bugfix: get the exceptions raised in the simulator repport where
3936 and why they come from when they are not caught in time [AL, MQ]
3939 * Bugfix: Do repport the error when two non-connected hosts try to
3940 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3943 * Add additionnal checkings on communications. Assert that two
3944 communicating hosts are connected by a set of links... [AL]
3947 * Add additionnal checkings on channel values in communication [AL]
3948 * New: MSG_task_get_source to see on which host a task was generated [HC]
3949 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3950 the number of tasks waiting to be received on channel and sent
3952 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3953 waits for the first task coming from a given host.. [AL]
3955 GRAS new functionnalities: [MQ]
3956 * Enhance the parsing macro to allow the size of multidimentional objects
3957 to be given thru annotations.
3958 * New example (and documentation): Matrix Multiplication a la RPC
3959 (as when I was young!) and fix a bunch of bugs found on the way.
3961 GRAS performance improvements: [MQ]
3963 * Reduce the amount of cbps creation/destruction by making it static to
3964 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3967 * Change libdata to a set so that we can search for stuff by ID (and thus
3968 reduce the insane amount of dict lookups)
3971 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3972 RL communication on top of SG since it's so uneffective.
3973 It may also be used for inter-thread communication in RL, one day.
3974 * Use gras_datadesc_copy() to exchange messages on top of SG
3976 - improve message exchange performance on top of SG
3977 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3978 * Don't exchange on the network the size of the used part of buffer,
3979 instead, specify the possible buffer size to read().
3981 - reduces the amount of read/write calls (one pair per exchange)
3982 - reduces the amount of exchanged data (the size)
3983 - allows to retrieve all arrived data on receiver side, if we don't need
3984 it right now (subsequent read will peek the buffer)
3985 - allows the receiver to proceed with the beginning of the stream before
3986 everything is arrived
3987 - make it possible to build an iov transport (using readv/writev)
3989 - take care of the data with non-stable storage (like stacked data),
3991 * If possible, TCP send uses vector I/O (when writev() is here)
3992 - Don't use it for receive since we send data sizes and data on the
3993 same stream, so we wouldn't be able to chain large amount of chunks
3994 before having to flush the stuff to read the size.
3995 * Rework the transport plugin mechanism to simplify it and reduce the
3996 amount of pointer dereferencement when searching for the right function
3999 * I guess that now, we do almost as few system calls as possible while
4000 doing as few data copy as possible.
4002 To improve it further, we could try to send all the sizes first and then
4003 all the data (to use iov on receiving size), but it's only a partial
4004 solution: when you have 2 dimensional data, the sizes of the second
4005 dimension is data of the first dimension, so you need 3 streams.
4007 I'm not sure the potential performance gains justify the coding burden.
4009 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
4011 SimGrid (3.00) stable; urgency=high
4014 * New! Give the possibility to hijack the surf parser and thus bypass
4015 MSG_create_environment and MSG_launch_application. Have a look at
4016 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
4018 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
4020 SimGrid (2.96) unstable; urgency=low
4025 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
4026 This deprecates the xbt_error_t mechanisms.
4027 It modifies (simplifies) all XBT and GRAS API.
4028 MSG API keeps unchanged (exceptions raised by XBT are caught from
4029 within MSG and masked with existing error handling facilities)
4032 * New! Add a FATPIPE model. [AL]
4033 * New! Add a parallel task model. [AL]
4034 * New! Add automatically a loopback interface (in the default
4035 network model) if none was precised.
4038 * Bugfix: MSG_process_resume now works with the current running process.
4040 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
4041 * Modification of MSG_task_get_compute_duration. Once a task has been
4042 processed, the value returned by this function is now equal to 0. [AL]
4043 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
4044 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
4045 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
4046 of a m_task. For now, MSG_task_cancel only works with computation
4048 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
4049 of the processor (in Mflop/s) regardless of the current load on the
4051 * API Change: use proper naming convention for MSG_getClock and
4052 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
4054 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
4055 This function changes the priority of a computation task. This priority
4056 doesn't affect the transfer rate. A priority of 2 will make a task
4057 receive two times more cpu power than the other ones. This function
4058 has been added to suit the needs of Nguyen The Loc and hasn't been that
4059 much tested yet. So if it fails, please report it and send me your code.
4061 * API Change: removed all functions and types that were marked "deprecated"
4062 since many months. Renamed MSG_global_init_args to MSG_global_init.
4064 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
4066 SimGrid (2.95) unstable; urgency=low
4069 * Steal some nice code to GNU pth to fix context detection and usage [AL]
4070 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
4071 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
4074 * Bugfix: Allow absolute paths for platform description files [MQ]
4075 * Bugfix: do free the variables after use. Leads to drastic performance
4077 * Implement max_duration (ie, timeouts) on resources [AL]
4080 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
4082 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
4083 the needed support (timeouts) [AL]
4086 * Implement measurement sockets. You can now get the bandwidth between two
4087 hosts thanks to AMOK (see below). [MQ]
4088 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
4089 dynar over the network (yeah) [MQ]
4090 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
4091 * simple/static token-ring example. [Alexandre Colucci and MQ]
4092 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
4093 in place before [MQ]
4096 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
4097 type parsing. "short" and "long" modifiers were also ignored; other
4098 modifier (such as reference level) are still ignored. [MQ]
4099 * Embeed the buffer size within the buffer itself on SG. [MQ]
4100 That way, send() are atomic and cannot get intermixed anymore (at least
4101 the ones which are less than 100k; bigger messages still have the issue)
4102 * Array size pushed by the field, not by the field type (or each
4103 and every long int will push stuff to the cbps) [MQ]
4104 * use select() to sleep since it allows to portably sleep less than one
4107 GRAS (minor cleanups)
4108 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
4111 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
4112 * type category 'ignored' killed as it was never used and were difficult
4114 * whether a type can cycle or not is now a flag, leaving room for more
4115 flags (as "ignored", if we feel the need one day ;)
4116 * Rename raw sockets to measurement sockets since "raw" has another
4117 meaning in networking community.
4120 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
4121 and offers features not belonging to GRAS but that most applications
4122 need. One day, it may be a set of plugins loadable at runtime.
4123 * New module: bandwidth
4124 bandwidth measurement between arbitrary nodes running this module. [MQ]
4126 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
4128 SimGrid (2.94) unstable; urgency=low
4130 The first beta release of SimGrid 3 !
4134 * Update the main page and the FAQ. Adding references to gforge.
4137 * Add a gras_os_getpid function.
4140 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
4141 * Extend the logs so that they also print PID, hostname, date, ... if
4143 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
4144 and kill the old version which were in testsuite/
4145 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
4149 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
4150 for users. Instead, each programming environment (GRAS, MSG,...) use it
4151 when needed to provide such a feature to users.
4152 Don't shortcut the mechanism or you will also shortcut the virtualization
4153 you need on the simulator.
4157 * Cleanups in configury with regard to compile optimization/warning flags.
4158 Also add -fno-loop-optimize to any powerpc since it's the optimization
4159 killing gcc (< 3.4.0).
4160 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
4161 needed by MSG examples complications
4162 * Borrow configury beautifications from PHP
4165 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
4166 warning (thanks loris for stumbling into this one).
4167 * Bugfix: stop loading private headers (gras_config.h) from the public
4171 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
4172 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
4173 RL would imply a DNS resolver, which is *hard* to do in a portable way
4174 (and therefore delayed).
4175 * Implement a real timer mechanism and use it in timing macros. This allows
4176 to avoid rounding errors and get a 0.000005 sec precision in timing
4177 macros. While I was at it, various cleanups:
4178 - allow to declare more than one timed section per file (fix a stupid bug)
4179 - move some private declaration to the right place
4180 - merge conditional execution and timing macros into emulation module
4181 - document the module
4182 - make sure the module cleanups its mess on gras_exit
4183 * Documentation improvements:
4184 - (new) how to compile applications using GRAS
4185 - (new) emulation support (timing macros)
4187 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
4189 SimGrid (2.93) unstable; urgency=low
4191 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
4194 - Use Paje properly where used. Still to be sanitized properly.
4195 - Portability fix: Add an implementation of the contexts using pthread
4199 - Add xbt_procname(): returns the name of the current process.
4200 Use it to show the current process's name in all logging.
4202 - fix detection of older flex version and the reaction, since we do
4203 depend on modern ones (we use lex_destroy)
4204 - Better separation of SG and RL in the libs: remove all simulation code
4205 from libgras. As a result, this lib is now only 200k when stripped.
4206 Some of the xbt modules may also be duplicated (two sets and such) and
4207 should be cleaned/killed before SG3.
4208 - Insist on using xlC on AIX because of weird problems involving gcc there.
4209 - Cleanup the make remote stuff. This is now done by scripts
4210 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
4211 mainly for our private use, but we're working on changing them to user
4214 - Bugfix: flush the socket on close only if there is some *output*.
4215 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
4217 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
4220 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
4222 SimGrid (2.92) unstable; urgency=low
4224 Alpha 3 on the path to SimGrid 3
4228 - New! First try of benchmarking macros.
4229 - New! First try so that gras_stub_generator generate deployment and
4230 remote compilation helpers.
4232 - Bugfix: Initialization fix in msg_test.
4236 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
4238 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
4240 - API Change: gras_os_sleep to take the amount of seconds as a double.
4241 Accepting an int was error prone since it was the only location where
4242 seconds were coded as such. It leaded to damn rounding errors.
4243 - Bugfix: Hard to belive that timers ever worked before this.
4245 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4247 SimGrid (2.91) unstable; urgency=low
4249 Alpha 2 on the path to SimGrid 3
4253 - Bug fix in the lmm_solver.
4255 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4256 through the function MSG_paje_output.
4257 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4258 - It is possible to bound the rate of a communication in MSG with
4259 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4260 to write it in the changelog).
4261 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4266 - Various cleanups to the autotools stuff
4267 - Begin to move Gras examples to examples/gras/
4268 - Let make distcheck work again (yeah!)
4270 - documentation overhauled using doxygen.
4271 gtk-doc-tools is dead in SimGrid now.
4272 - Automatically extract all existing logging categories, and add the list
4273 to the documentation (long standing one, to say the less)
4275 - Cleanup the known architecture table. Reorder the entries to group what
4276 should be, and use a more consistent naming scheme.
4277 (some of the test dataset are still to be regenerated)
4278 - New! Allow library to register globals on each process just as userdata
4280 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4281 lookup time (for now).
4282 Use it in msg and trp.
4283 This cleans a lot the internals and helps enforcing privacy of the
4284 headers between the gras components.
4285 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4286 - Bugfix: gras_os_time was delirious in RL.
4287 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4289 - Reenable GRAS now that it works.
4291 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4293 SimGrid (2.90) unstable; urgency=low
4295 Alpha 1 on the path to SimGrid 3
4297 * It is a long time since the last release of SimGrid. I'm sorry about
4298 that but as I had told you, I was rewriting a lot of things. I apologize
4299 to those who had been reporting bugs to me and that I had not answered.
4300 If your bug is still in the new version, please tell me. Here is a
4301 summary of the main changes.
4303 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4304 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4305 portability, speed, and a lot more but you'll figure it out later.
4306 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4307 I wanted to release the new MSG as soon as possible and I have
4308 broken GRAS, which is the reason why, for now, only MSG is fully
4309 functional. A laconic description of these projects is available
4310 in the documentation.
4312 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4313 kernel optimized for our needs (called SURF but only the developers
4314 should use it). Hence, MSG is now roughly 30 times faster and I think
4315 that by rewriting a little bit MSG, I could event speed it up a little
4316 bit more. Beside the gain in speed, it is also much easier to encode a
4317 new platform model with SURF than it was with SG. More to come...
4319 * REVOLUTION 3: I have tried to change a little as possible the API of
4320 MSG but a few things really had to disappear. The main differences
4321 with the previous version are:
4322 1) no more m_links_t and the corresponding functions. Platforms are
4323 directly read from an XML description and cannot be hard-coded
4324 anymore. The same format is used for application deployment
4325 description. The new format is described in the documentation.
4326 Have a look in tools/platform_generation. There is a tiny script
4327 that converts from the old platform format to the new one. Concerning
4328 the application deployment format, parsing the old one is tricky.
4329 I think most of you should however be able to convert your files. If
4330 it is really an issue, I can write a C code that does the conversion.
4332 2) the toolbox tbx does not exist anymore. We now have a library
4333 with much more data-structures but without the hash-tables (we have
4334 dictionaries that are much faster).
4336 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4338 *****************************************************************************
4339 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4340 * don't feel like converting the oldies. (Mt) *
4341 *****************************************************************************
4344 Version 2.90: "the long awaited one"
4345 - Finished rewriting and debugging MSG. Rewrote the documentation.
4346 - disable GRAS for now since it needs to be ported to the newest SG
4349 - Finish the port to windows (using mingw32 for cross-compile)
4352 - Main loop and datastructures of SURF. A cpu resource object is
4353 functional. Surf can thus be used to create cpu's with variable
4354 performance on which you can execute some actions.
4356 2004-11-15 Martin Quinson
4357 - Port to ARM. Simply added the alignment and size descriptions. Should
4358 work, but the ARM machines are so slow that I didn't had the opportunity
4359 to 'make check' over there yet.
4361 2004-11-15 Arnaud Legrand
4362 - Trace manager now written. It uses a heap structure and is therefore
4363 expected to be efficient. It may however be speeded up (particularly
4364 when many events occur at the same date) by using red and black
4365 trees. One day maybe...
4366 - Max-min linear system solver written. It uses a sparse matrix
4367 structure taking advantage of its expected use. Most operations are
4368 O(1) and free/calloc are called as few as possible. The computation of
4369 the minimum could however be improved by using a red and black tree
4372 2004-11-03 Arnaud Legrand
4373 - Rename every gras_* function that was in xbt/ to its xbt_
4375 - Add a heap and a doubly-linked list to xbt
4376 - Added a dichotomy to the dictionaries. make check works as well before
4377 so I assume that the patch is correct. I do not know however if things
4378 run effectively faster than before now. :)
4380 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4381 SimGrid, and the version number is bumped to 2.x
4383 2004-10-29 Martin Quinson
4384 - Introduction of the remote errors.
4385 They are the result of a RMI/RPC on the remote machine.
4386 ErrCodes being scalar values, you can't get the host on which those
4387 errors did happen. Extending the error mechanism as in Gnome is possible.
4388 No idea yet whether it is a good idea.
4390 2004-10-28 Martin Quinson
4391 - Interface revolution: the Starred Structure Eradication.
4392 I used to do typedef struct {} toto_t; and then handle *toto_t.
4393 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4394 - ???_t is a valid type (builded with typedef)
4395 - s_toto_t is a structure (access to fields with .)
4396 - s_toto is a structure needing 'struct' keyword to be used
4397 - e_toto_t is an enum
4398 - toto_t is an 'object' (struct*)
4400 typedef struct s_toto {} s_toto_t, *toto_t;
4401 typedef enum {} e_toto_t;
4402 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4403 s_toto_t) is private.
4405 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4406 so that it can set the variable to NULL. It was so for dicts and sets,
4407 it changed for dynars.
4409 - Fix a bunch of memleaks in dict_remove
4410 - Fix a bug in sg/server_socket opening: it failed all the time.
4412 2004-10-07 Martin Quinson
4413 - Speed up dynar lookup operation a bit.
4415 gras_dynar_get is dead.
4417 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4418 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4419 (which gives you the address of the stored data).
4421 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4422 copy of the data using an affectation to do the job and not a memcpy.
4424 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4425 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4427 It does not really speedup the dynar test because they are
4428 setting elements all the time (and look them seldom). But the dict does
4429 far more lookup than setting.
4431 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4433 2004-10-05 Martin Quinson
4434 - Allow to (en/dis)able the cycle detection at run time.
4436 Whether we should check for cycle or not is now a property of each
4437 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4438 datadesc_cycle_unset allow to remove this property when previously set.
4440 Note that the cycle detection is off by default since it impacts the
4441 performance. Watch the data you feed GRAS with ;)
4443 This property is hereditary. Any element embedded in a structure having it
4444 set have it set for the time of this data exchange.
4446 You should set it both on sender and receiver side. If you don't set it on
4447 sender side, it will enter an endless loop. If you forget on receiver
4448 side, the cycles won't be recreated after communication.
4450 - Header reorganization.
4451 Kill gras_private.h, each submodule must load the headers it needs.
4453 2004-10-04 Martin Quinson
4454 - Interface revolution: do not try to survive to malloc failure.
4456 Now, gras_malloc and friends call gras_abort() on failure.
4457 As a conclusion, malloc_error is not a valid error anymore, and all
4458 functions for which it was the only gras_error_t return value are
4459 changed. They now return void, or there result directly.
4460 This simplify the API a lot.
4462 2004-09-29 Martin Quinson
4463 - Re-enable raw sockets.
4464 Created by gras_socket_{client,server}_ext;
4465 Used with gras_raw_{send,recv}
4468 It should allow to kill the last bits of gras first version soon.
4470 This is not completely satisfactory yet (duplicate code with
4471 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4474 - Simplify transport plugin (internal) interface by not passing any
4475 argument to _server and _client, but embedding them in the socket
4478 2004-09-28 Martin Quinson
4479 - Finish the port to AIX.
4480 autoconf was my problem (segfault within the malloc replacement
4481 function. No idea why)
4483 2004-09-16 Martin Quinson
4484 - Fix some size_t madness on 64bit architectures.
4486 2004-09-08 Martin Quinson
4487 - Reduce the number of system headers loaded, overload some more system
4488 calls (such as malloc to cast the result of the system one, and work
4490 - Fix and reintroduce the config support
4492 2004-09-07 Martin Quinson
4493 - Source code reorganization to allow Arnaud to surf all over there.
4494 - Allow to document the logging categories.
4495 - Remove all uppercase from logging categories and useless cleanup in names.
4497 2004-08-18 Martin Quinson
4498 Version 0.6.2 (protocol not changed; API changed)
4499 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4500 gras_error_t), and NULL when not found. Functions expecting a msgtype
4501 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4502 hopefully usefull message.
4503 - Portability to prehistoric sparcs again
4505 2004-08-17 Martin Quinson
4506 Version 0.6.1 (protocol not changed; ABI not changed)
4507 - prealloc some buffers to speed things up
4509 2004-08-11 Martin Quinson
4510 Version 0.6 (protocol not changed; ABI expended)
4511 - The parsing macro can deal with the references, provided that you add
4512 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4514 2004-08-09 Martin Quinson
4515 Version 0.5 (protocol not changed; ABI changed)
4516 - Allow to off turn the cycle detection code in data exchange at
4517 compilation time. It should be at run time, but I'm short of time (and
4518 the config stuff is still broken). That way, we keep dict out of the
4519 critical path, which is good because the performance is poor:
4520 - search not dichotomial yet
4521 - dynar give no way to access their content and memcpy everytime
4522 - In composed data description (struct, ref and so on), stop foolness of
4523 keeping the subtype's ID, but store the type itself. This keeps sets out
4524 of the critical path, which is good since they rely on dynar and
4525 dictionaries. The only loose of that is that we cannot detect the
4526 redeclaration of a structure/union with another content (but I'm not sure
4527 the code detected well this error before anyway). We still can detect
4528 the redefinition discrepancy for the other types.
4529 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4530 it breaks the code because of type-punning used all over the place).
4531 This breaks on all non-gcc architectures (for now).
4533 All those changes (plus the buffer of last time) allow me to gain 2 order
4534 of magnitude on cruel tests consisting of 800000 array of integers on two
4535 level of a hierarchical structure (200 secondes -> 4 secondes)
4538 - the selector of reference must now return the type it points to, not
4539 the ID of this type.
4541 2004-08-06 Martin Quinson
4542 Version 0.4 (protocol changed; ABI not changed)
4543 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4544 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4545 - Implement a buffer transport, and use it by default (it relies on tcp in
4546 real life and on sg in simulation).
4547 That's a bit hackish since I had a new field to the structure to store
4548 its data without interfering with the subtype ones. Inheritance
4549 is tricky in C. And that's a kind of reverse inheritance with one class
4550 derivating two classes. Or maybe a game with java interfaces. Anyway,
4551 that's damn hard in C (at least).
4552 Moreover, I got tired while trying to ensure plugin separation and
4553 genericity in SG mode. MSG wants me to do weird things, so let's go for
4554 cruel hacks (temporarily of course ;).
4555 See comment in transport_private.h:71
4556 - do not include all the _interface headers in private but in the files
4557 which really need them (to cut the compilation time when they are
4560 2004-07-26 Martin Quinson
4561 Version 0.3 (protocol not changed; ABI changed)
4562 - Major overhault of the datadesc interface to simplify it:
4563 - shorted the function names:
4564 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4565 - add a trivial way to push/pop integers into the cbps without malloc.
4566 This allows to make really generic sub_type description, which simply
4567 pop their size of the stack.
4568 - add a function gras_datadesc_ref_pop_arr() which does what users want
4569 most of the time: Declare a dynamic array (which pops its size of the
4570 stack) and declare a reference to it. Poor name, but anyway.
4571 - kill the post-send callback, add a post-receive one
4573 2004-07-23 Martin Quinson
4574 Version 0.2 (protocol changed; ABI changed)
4575 - add some testing for cpbs in the test cases, and fix some more bugs.
4576 This invalidate again the little64 data file, since I cannot regenerate
4578 - remove an awfull optimization in the logging stuff, allowing me to:
4579 - understand it again
4580 - learn gcc how to check that the argument match the provided format
4581 - fix all errors revealed by gcc after that
4582 - internal keys of dict are not \0 terminated. Deal with it properly in
4583 loggings instead of segfaulting when the user want to see the logs :-/
4585 2004-07-22 Martin Quinson
4586 - Fix some stupid bug preventing cbps (callback postit) from working
4588 2004-07-21 Martin Quinson
4589 - Some documentation cleanups
4590 - remove the useless last argument of msgtype_declare
4591 - rename the Virtu functions to fit into the 'os' namespace
4592 - move headers src/include -> src/include/gras/ and stop fooling with
4593 gras -> . symbolic link
4594 - make distcheck is now successful
4596 2004-07-19 Martin Quinson
4598 - Build shared library also
4599 - Install html doc to the right location
4600 - stop removing maintainer files in make clean
4601 - build tests only on make check
4603 2004-07-13 Martin Quinson
4605 - No major issue in previous version => change versioning schema
4606 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4607 corresponding dataset.
4609 2004-07-11 Martin Quinson
4611 - Get it working with any kind of structure (we can compute the padding
4612 bytes remotely for all the architectures I have access to)
4613 - Implement the structure parsing macro (still not quite robust/complete)
4614 - Improvement to the remote testing toysuite
4616 2004-07-10 Martin Quinson
4617 [autoconf mechanism]
4618 - get ride of a bunch of deprecated macros
4619 - actually run the test for two-compliment, not only compile it :-/
4620 - test whether the structures get packed (and bail out if yes. Damn.
4621 Alignment is a serious matter)
4622 - test whether the structures get compacted (but respecting the alignment
4623 constraints of each types)
4624 - test whether the array fields of structures can straddle alignment boundaries
4626 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4627 (took me 2 hours to find that bug, looking at the wrong place)
4628 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4629 before sending/receiving and are used to compute the offsets of fields
4630 - Given that padding size depend even on compiler options, keep track of
4631 alignment and aligned_size only for the current architecture. Not a big
4632 deal since we send structure fields one after the other (seems
4634 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4635 program, let it work on linux/gcc/little32. portability todo.
4637 2004-07-08 Martin Quinson
4638 - import and improve remote compilation support from FAST
4639 - make sure make check works on half a dozen of machines out there
4641 2004-07-07 Martin Quinson
4642 Let's say it's version 0.0.3 ;)
4643 - Implement conversions (yuhu!)
4644 - Let it work on solaris (beside conversion, of course)
4645 - Stupid me, using rand() to generate the conversion datatests in not wise.
4647 2004-07-06 Martin Quinson
4648 - Let make dist work, since I'm gonna need it to compile on remote hosts
4649 - Let Tests/datadesc_usage write the architecture on which the file was
4650 generated as first byte.
4651 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4652 architecture descriptions.
4653 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4654 on those architectures.
4655 - Optimization: send/recv array of scalar in one shoot
4657 2004-07-05 Martin Quinson
4658 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4660 - Plug a whole bunch of memleaks
4661 - each process now have to call gras_{init,exit}. One day, their log
4662 settings will be separated
4663 - Continue the code factorisation between SG, RL and common in Transport.
4665 2004-07-04 Martin Quinson
4667 - Redistribution between SG and RL.
4668 We wanna have to accept in SG, so move accepted related parts of RL in
4669 the common part. (more precisely, the dynar of all known sockets is no
4670 more a static in transport.c, but part of the process_data)
4672 [gras_stub_generator]
4673 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4675 2004-07-03 Martin Quinson
4676 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4677 them all up in one shot)
4679 - Fix the ugly case of reference to dynamic array.
4680 - New (semi-public) function gras_datadesc_size to allow the messaging
4681 layer to malloc the needed space for the buffer.
4683 - gras_socket_close now expect the socket to close (and not its address to
4684 put NULL in it after it). This is because the socket passed to handlers
4685 is one of their argument (=> not writable).
4687 - propagate the interface cleanup from last week in datadesc, ie remove a
4688 superfluous level of indirection. User pass adress of variable
4689 containing data (both when sending and receiving), and not of a variable
4690 being a pointer to the data. Let's say that I like it better ;)
4691 The price for that is constructs like "int msg=*(int*)payload" in
4692 handlers, but it's a fine price, IMHO.
4694 - Let it work in RL (yuhu)
4696 2004-06-21 Martin Quinson
4698 - porting SG plugin and SG select to new standards (works almost).
4699 - plug memleaks and fix bugs around.
4702 - cleanup the prototype of data recv and force users to specify when they
4703 want to handle references to objects. Test case working even for cycles.
4704 - plug memleaks. Valgrind is perfectly ok with this.
4706 2004-06-12 Martin Quinson
4708 - cleanup the separation between plugin and main code in plugin creation
4710 2004-06-11 Martin Quinson
4712 - Reput hook for raw sockets, needed for BW experiments
4713 - kill a few lines of dead code
4714 [Data description] Interface cleanup
4715 - gras_datadesc_by_name returns the searched type or NULL.
4716 That way, no variable is needed to use a type desc once, which makes
4718 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4719 two parameters were strings (field name, type name), leading to
4721 [Dicos] Interface cleanup
4722 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4723 This is consistant with the dynar API.
4725 2004-04-21 Martin Quinson
4727 - Porting to new standards.
4729 - interface cleanup.
4730 There is no bag anymore, no need to take extra provision to mask the
4731 pointers behind "ID".
4732 Better splitup of functions between files create/exchange/convert.
4733 This is still a bit artificial since convert and receive are so
4734 interleaved, but anyway.
4736 - add a queued message list to procdata (the ones not matching criteria
4738 - factorize some more code between SG and RL wrt procdata
4740 - use gras_exit in example to track memleaks
4741 - get rid of gs_example now that GS is properly integrated into gras
4742 - update run_test to integrate the latest tests (datadesc)
4744 - rename WARNINGn macros to WARNn since it prooved error-prone
4746 2004-04-19 Martin Quinson
4748 - register init/exit functions within gras module mechanism
4749 - send/receive function.
4750 Convertion is not implemented, but short-cutted if not needed.
4751 struct/array elements are sent one by one (instead of block-wise), but
4752 nobody really cares (yet). Get a prototype before optimizing.
4753 - tests (using a file socket) for DD send/receive on:
4754 - base types: int, float
4755 - array: fixed size, string (ie ref to dynamic string)
4756 - structure: homogeneous, heterogeneous
4757 - chained list, graph with cycle
4758 Believe it or not, valgrind is not too unhappy with the results. The
4759 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4763 - Bugfix when using a filename explicitly (instead of '-')
4765 2004-04-09 Martin Quinson
4767 - factorize more code between RL and SG in socket creation
4768 - Complete the implementation and tests of:
4770 o file (only in RL, and mainly for debugging)
4772 I lost 3 days to design a portable address resolver, and then decided
4773 that the prototype mainly have to run on my box.
4774 Addressing portability too early may be like optimizing too early :-/
4776 - use gras_init in the Tests instead of the crappy parse_log_opt
4777 (the latter function is removed)
4778 [Conditional execution]
4779 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4780 [Code reorganisation]
4781 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4782 Build examples against the RL library, since there is no way to disable
4783 its creation for now.
4785 For information, the beginning of coding on GRAS was back in june
4786 2003. I guess that every line has been rewritten at least twice since