1 SimGrid (3.31.1) NOT RELEASED YET (v3.32 expected June 21. 2022, 09:13 UTC)
4 - Added the s4u_enforce macro. This has the same syntax as xbt_assert. The only difference is that an AssertionError
5 exception is thrown if the condition is not satisfied, instead of calling abort().
8 - Added the following bindings / examples:
9 - Comm (now 100% covers the C++ interface):
10 - Comm.dst_data_size, Comm.mailbox, Comm.sender, Comm.start_time, Comm.finish_time
11 - Comm.state_str [examples: examples/python/comm-failure/, examples/python/comm-host2host/]
12 - Comm.remaining [examples: examples/python/comm-host2host/, examples/python/comm-suspend/]
13 - Comm.set_payload_size [example: examples/python/comm-host2host/]
14 - Comm.set_rate [example: examples/python/comm-throttling/]
15 - Comm.sendto, Comm.sendto_init, Comm.sendto_async [example: examples/python/comm-host2host/]
16 - Comm.start, Comm.suspend, Comm.resume [example: examples/python/comm-host2host/]
17 - Comm.test_any [example: examples/python/comm-testany/]
18 - Comm.wait_until [example: examples/python/comm-waituntil/]
20 - Engine.host_by_name [example: examples/python/comm-host2host/]
21 - Engine.mailbox_by_name_or_create [example: examples/python/comm-pingpong/]
22 - Mailbox: Mailbox.ready [example: examples/python/comm-ready/]
23 - Added an AssertionError exception that may be thrown in case of error. For instance, creating tow hosts with the
24 same name will now throw this exception instead of killing the interpreter.
26 Platform description & visualization:
27 - More robust sanity checks for platforms, to reject unallowed topologies with
28 a proper error message.
29 - One new C++ platform example, supernode.cpp. A Python script (supernode.py)
30 demonstrates how we can generate a nice graphical representation of the
34 - Enhancements to the graphicator tool:
35 - Allow to dump the platform topology as a CSV file representing the edges
36 of the graph (in addition to the DOT format).
37 - Fix graphicator for "cluster" topologies (e.g. fat-tree, dragonfly).
39 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
40 (FG: issues on Framagit; GH: issues on GitHub)
41 - FG#105: "Variable penalty should not be negative!" with in-flight messages and bandwidth profiles
42 - FG#109: Application time reported by --cfg=smpi/display-timing:yes is wrong
44 ----------------------------------------------------------------------------
47 - kill signal Comm::on_completion that was not working anyway.
48 - Expose signals Activity::on_suspend and Activity::on_resume
50 SimGrid (3.31) March 22. 2022.
52 The ненасильство release. We stand against war.
54 Against the agression by a sick system that forces peoples to take arms against each other.
57 - Rework the internals, for simpler and modern code. This shall unlock many future improvements.
58 - You can now define plugins onto the DFS explorer (previously called SafetyChecker), using the
59 declared signals. See CommunicationDeterminism for an example.
60 - Support mutex, semaphore and barrier in DPOR reduction
61 - Seems to work on Arm64 architectures too.
62 - Display a nice error message when ptrace is not usable.
63 - New test suite, imported from the MPI Bugs Initiative (MBI). Not all MBI generators are integrated yet.
64 - Remove the ISP test suite: it's not free software, and it's superseeded by MBI.
67 - fix for FG#100 by ensuring small asynchronous messages never overtake larger
68 ones, conforming to the standard.
69 - replay: fix waitall behaviour to avoid forgetting requests and leaking
71 - tracing: ensure that we dump the TI traces continuously during execution and
72 not just at the end, reducing memory cost and performance hit.
73 - Update OpenMPI collectives selection logic to match current one (4.1.2)
74 - Add a coherence check for collective operation order and root/MPI_Op
75 coherence. Potentially costly so not activated unless smpi:pendantic is set
79 - New signal: Engine::on_simulation_start_cb()
80 - Introduce a new execution mode with this_actor::thread_execute(). This simulate
81 the execution of a certain amount of flops by multiple threads ran by a host. Each
82 thread executes the same number of flops, given as argument. An example of this new
83 function can be found in examples/cpp/exec-threads.
84 - Reimplementation of barriers natively.
85 Previously, they were implemented on top of s4u::Mutex and s4u::ConditionVariable.
86 The new version should be faster (and can be used in the model-checker).
87 - Actor::get_restart_count(): Returns the number of reboots that this actor did.
90 - MSG_barrier_destroy now expects a non-const msg_barrier parameter.
92 New plugin: the Chaos Monkey (killing actors at any time)
93 - Along with the new simgrid-monkey script, it tests whether your simulation
94 resists resource failures at any possible timestamp in your simulation.
95 - It is mostly intended to test the simgrid core in extreme conditions,
96 but some users may find it interesting too.
99 - New solver for parallel task: BMF.
100 - More realistic sharing of heterogeneous resources compared to the fair
101 bottleneck solver used by ptask_L07.
102 - Implement the BMF (Bottleneck max fairness) fairness.
103 - Improved resource sharing for parallel tasks with sub-flows (parallel
104 communications between same source and destination inside the ptask).
106 - "--cfg=host/model:ptask_L07 --cfg=host/solver:bmf": enable the ptask
107 model with BMF solver.
108 - "--cfg=bmf/max-iterations: <N>": maximum number of iterations performed
109 by BMF solver (default: 1000).
110 - "--cfg=bmf/precision: <N>": numerical precision used when computing
111 resource sharing (default: 1e-12).
112 - This model requires Eigen3 library. Make sure Eigen3 is installed to use BMF.
115 - Modifications of the Profile mechanism, with some impact on users
116 - Addition of a new (S4U) method to init profiles from generic functions to improve versatility
117 - Fix initial behaviour of state_profiles
118 - Modify periodicity to behave like a period, and not like a loop delay
121 - Drop xbt_dynar_shrink().
124 - Made the following bindings static (previously member functions):
125 - Actor: Actor.kill_all(), Actor.by_pid()
126 - Host: Host.by_name(), Host.current(), Host.on_creation_cb()
127 - Mailbox: Mailbox.by_name()
128 - Added the following bindings:
129 - this_actor.warning()
130 - Mailbox.put_init() [example: examples/python/comm-waitallfor/]
131 - Comm.detach() [example: examples/python/comm-waitallfor/]
132 - Comm.wait_for() [example: examples/python/comm-waitfor/]
133 - Comm.wait_any_for()
134 - Comm.wait_all_for() [example: examples/python/comm-waitallfor/]
135 - Mutex [example: examples/python/synchro-mutex/]
136 - Barrier [example: examples/python/synchro-barrier/]
137 - Semaphore [example: examples/python/synchro-semaphore/]
140 - Remove target "make uninstall" which was incomplete and no longer maintained.
142 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
143 (FG: issues on Framagit; GH: issues on GitHub)
144 - FG#57: Mc SimGrid should test whether ptrace is usable
145 - FG#87: Smpi scripts fail with spaces in paths
146 - FG#100: [SMPI] Order of the message matching is not guaranteed
147 - FG#101: LGPL 2.1 is deprecated license
148 - FG#104: "make uninstall" not up-to-date
149 - GH#151: Missing mutexes for DPOR.
151 ----------------------------------------------------------------------------
153 SimGrid (3.30) January 30. 2022.
155 The Sunday Bloody Sunday release.
157 Main user-visible changes:
158 - The SimDag API for the simulation of the scheduling of Directed Acyclic
159 Graphs has been dropped. It was marked as deprecated for a couple of years.
160 We finally complete the implementation of what has been called SimDag++
161 internally, i.e., porting the different features of SimDag on top of S4U.
162 The new way to simulate the execution of dependent activities directly by
163 maestro (without any other actor) is details in the examples/cpp/dag-* series
165 - The removal of SimDag led us to also remove the export to Jedule files that
166 was tightly coupled to SimDag. The instrumentation of DAG simulation is still
167 possible through the regular instrumentation API based on the Paje format.
168 - We also dropped the old and clumsy Lua bindings to create platforms in a
169 programmatic way. It can be done in C++ in a much cleaner way now, which
170 motivates this suppression.
173 - Introduce on_X_cb() functions for all signals, to attach a new
174 callback to the signal X. The signal variables are now hidden and
175 only these functions should be used.
176 Rationale: this enables the usual deprecation scheme where functions
177 remain for 4 releases if we need to modify the signals, while the
178 current code with the signal variables directly visible prevents any
180 - New function: Engine::run_until(date), to split the simulation.
181 - New signal: Activity::on_veto, to detect when an activity fails to start.
182 - Signal change: Comm::on_start(Comm&, bool) has been replaced by
183 Comm::on_send and Comm::recv. These two signals respectively correspond to
184 when the sending or receiving side of a Comm is ready. They are raised at
185 the same locations as the former Comm::on_start signal.
186 - New function: Engine::track_vetoed_activities() to interrupt run()
187 when an activity fails to start, and to keep track of such activities.
188 Please see the corresponding example for more info.
189 - New functions: s4u::Comm::{sendto_init, set_source, set_destination} to enable
190 the use of vetoers with direct host-to-host communications. Both source and
191 destination have to set for a comm to start. Each call to these setters check
192 if all vetoes are satisfied. When it is the case, the comm starts. A use case of
193 these functions is given in examples/cpp/dag-scheduling.
194 - New functions: {Exec, Io}::update_priority allow you to modify the priority of
195 these kinds of activities during their execution. Behavior is detailed in
196 examples/cpp/io-priority/.
199 - Dynamic costs for MPI operations: New API to allow users to dynamically
200 change injected costs for MPI_Recv, MPI_Send and MPI_Isend operations.
201 Alternative for smpi/or, smpi/os and smpi/ois configuration options.
202 - Fix some issues with the replay mechanism.
205 - Function xbt::Extendable::get_data() is now templated with the type of the
206 pointee. Untyped function is deprecated. Use get_data<void>() if you still
207 want to retrieve void*.
210 - New section: "SimGrid MPI calibration of a Grid5000 cluster"
211 presenting how to properly calibrate MPI communications in SimGrid.
212 - Complete and reword the platform section, which is now completed.
215 - Thread contexts are used by default with Python bindings. Other kinds of
216 contexts revealed unstable, specially starting with pybind11 v2.8.0.
218 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
219 (FG: issues on Framagit; GH: issues on GitHub)
220 - FG#95: Wrong computation time for multicore execution after pstate change
221 - FG#97: Wrong computation time for ptask+multicore+pstates
222 - FG#98: SMPI offline simulation is inconsistent with the online simulation
223 (deadlocks / message truncation)
224 - FG#99: Weird segfault when not sealing an host
226 ----------------------------------------------------------------------------
228 SimGrid (3.29) October 7. 2021
230 The "Ask a stupid question" release.
232 We wish that every user ask one question about SimGrid to celebrate.
233 On Mattermost, Stack Overflow or using the issues tracker.
235 New modeling features:
236 - Non-linear resource sharing, modeling resources whose performance heavily degrades with contention:
237 - The total capacity may be updated dynamically through a callback
238 and depends mainly on the number of concurrent flows.
239 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
241 - Dynamic factors: model variability in the speed of activities
242 - Each action can now have a factor that affects its progression.
243 This multiplicative factor is applied when updating the amount of work
244 remaining, thereby an activity with factor=0.5 only uses half of the
245 instantaneous power/bandwidth it is allocated and will appear twice
246 slower than what it actually consumes.
247 - This can be used to model a overhead (e.g., there is a 20 bytes
248 header in a 480 bytes TCP packet so the factor 0.9583) but the novelty
249 is this factor can now easily be adjusted depending on activity's and
250 resources characteristics.
251 - This existed for network (e.g., the effective bandwidth depends
252 on the message in SMPI piecewise-linear network model) but it is now
253 more general (the factor may depend on the source and destination and
254 thus account to different behaviors for intra-node communications and
255 extra-node communications) and is available for CPUs (e.g., if you
256 want to model an affinity as in the "Unrelated Machines" problem in
257 scheduling) and disks (e.g., if you want to model a stochastic
259 - For that, resources can be provided with a callback that computes
260 the activity factor when creating the action.
261 - Example: examples/cpp/exec-cpu-factors
262 - The same mechanism is also available for the latency, which
263 allows to easily introduce complex variability patterns.
266 - Added support to programmatic platform creation in Python.
267 Example: examples/python/clusters-multicpu
270 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
271 This can only be set through the API, not through XML files.
275 - Multiple fixes to ensure reproducibility of tracing
276 - scan/excan can now be replayed
277 - wait action now uses ranks and not pid, as the other ones.
278 - smpi/init and smpi/finalization-barrier are now valid for replays.
279 - exit() is now intercepted by SMPI to avoid premature shutdown of
280 simulation. First non 0 return codes is returned as simulation return
284 * New section "Release Notes" documenting recent and current developments.
285 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
286 * Improvements in API Reference for C++ and Python interfaces.
289 - Make wifi creation compatible with ns-3 version 3.34 too.
291 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
292 (FG: issues on Framagit; GH: issues on GitHub)
293 - FG#77: Search feature of doc is broken (update sphinx theme version)
294 - FG#78: Multiple fixes for SMPI replay:
295 - TI tracing of allotallv/w was outputting wrong values
296 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
298 ----------------------------------------------------------------------------
300 SimGrid (3.28) July 14. 2021
302 The Victoriadagarna Release.
305 - C++ platform interface: Users can now describe their platform directly in C++.
306 This provides greatly flexibility and performance improvement for complex
307 platforms. Main features:
308 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
309 "zones", instead of single hosts. This feature enables the description
310 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
311 - StarZone: new zone with a Star-like topology. The routes are defined
312 as a set of links used to communicate from node to everybody (node<->ALL).
313 - Split-Duplex links: auxiliary method to create split-duplex links in
314 the platform, easing its utilisation. It automatically creates both UP
315 and DOWN links (similarly as done in XML).
316 - Please refer to the documentation and the examples included:
317 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
318 - New plugin: Producer-Consumer with monitor. Just requires to include the
319 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
320 associated example (examples/cpp/plugin-prodcons).
323 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
324 s4u::Io::wait_any(), s4u::Io::wait_any_for().
325 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
326 parameter by reference, instead of a pointer.
327 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
328 Explicitly cancel the activity to get back to previous behavior.
329 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
332 - The default SMPI compiler flags are no more taken from the environment.
333 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
334 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
336 --cfg=smpi/finalization-barrier: which can be used to add
337 a barrier inside MPI_Finalize. This can help for some codes which cleanup
338 data attached to a process, but still used in other SMPI processes.
339 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
340 is active when set to false, to keep going after a small error
341 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
342 which may or may not be problematic in the end.
344 - fix behaviour, as maximum iteration count could be ignored
345 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
346 use sampling when the same kernel is called with a different set of parameters
347 which have an impact on the timing.
348 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
349 It should now work with smpi/auto-shared-malloc-thresh.
350 - Improve error handling and reporting in multiple places
351 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
352 validity checks, truncated messages are now an error, return errors
353 when explicitely deleted handles are reused, ...)
354 - RMA: multiple fixes and stability improvements.
355 - analysis (-analyze flag in smpirun):
356 - SMPI can now report buffer leaks as well as MPI handles leaks,
357 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
358 - if -trace-call-location is used when compiling, SMPI can report
359 origin of leaked handles/buffers
360 - group leaks by type/origin in output message if possible
361 - New implemented MPI calls: MPI_Comm_test_inter
364 - Changed internal implementation of bandwidth factors in network models.
365 Models affected: CM02, LV08 (default), SMPI, IB.
366 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
367 Bandwidth factors are applied to communications to describe that users
368 cannot use 100% of the available bandwidth. For example, the default network model,
369 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
370 behavior was implemented by limiting the bandwidth available in the LMM
371 system for this flow. This may give the false impression that there is
372 bandwidth available for other flows due to its underutilization, especially
373 for the dynamic bandwidth factors used in SMPI models.
374 To avoid this, we have modified the implementation so that each flow uses the
375 maximum physical bandwidth according to the LMM system.
376 However, the actual throughput of the flow seen by the user is defined by
377 the physical bandwidth multiplied by the bandwidth factor.
378 This change impacts on the simulation results for all network models on
379 which we have bandwidth factors configured.
380 *****************************************
381 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
382 *****************************************
383 This change may impact on the timing of your simulation results.
384 Take care when comparing simulations from different SimGrid's
385 versions. Sorry for the inconvenience.
386 - Dynamic network factors: users can configure a callback to define
387 the network factors dynamically. This API is available at
388 simgrid::kernel::resource::NetworkModelIntf.
389 - Users have access to complete information about the current communication
390 to decide which factor to apply. This includes: message size, source and
391 destination hosts, links and zones traversed.
392 - Dynamic factors for both latency and bandwidth.
393 - For more details, see the example in (examples/cpp/network-factors).
394 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
395 deprecated since version 3.24, are no longer supported. Instead, use
396 "wattage_off" and "wattage_per_state".
399 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
402 - New tutorial: Model-checking and formal assessment
403 - New sections: "Demystifying the routing" and "C++ platforms"
404 - Update and improve the part on visualization in MPI and Algo tutorials.
405 - Improve the section on routing: how to define it, how it's used internally
406 - Fix many issues, broken links and missing references in doxygen and Sphinx
409 - Lua platform files are deprecated. Their support will be dropped after v3.31.
412 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
414 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
415 (FG: issues on Framagit; GH: issues on GitHub)
416 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
417 - FG#64: Configuring smpi/IB-penalty-factors
418 - FG#67: Running computation concurrently with MPI_Iallreduce
419 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
420 - FG#71: Segmentation fault on invalid gw_src/gw_dst
421 - GH#322: Issue when an actor kills his host vm
423 ----------------------------------------------------------------------------
425 SimGrid (3.27) March 29. 2021
427 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
429 Main user-visible changes:
430 - All functions marked as deprecated for one year were removed.
431 - Removed all that was related to the "Storage" type in all APIs and bindings.
432 "Disks" have been introduced more than a year ago to replace them for the better.
433 It's time to finish this replacement.
434 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
435 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
436 avoiding "initializer element is not constant" errors with some codes. Please be
437 careful to recompile your MPI codes if switching to this new SimGrid version.
438 - Network model ns-3 now requires version ns-3.28 or higher.
439 - Removed some option aliases that were deprecated since v3.23 (June 2019)
440 - 'verbose-exit' is now 'debug/verbose-exit'
441 - 'simix/breakpoint' is now 'debug/breakpoint'
442 - 'clean-atexit' is now 'debug/clean-atexit'
445 - New functions: Comm::sendto_{init,async} to initiate a communication
446 on between two (possibly remote) hosts.
447 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
448 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
449 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
450 or Mailbox::get_async<void>() if you really want to play with void*.
451 - Unify the interface of Activity::on_{start/activity}
452 - New function: Comm::get_dst_data()
453 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
456 - ptask should now work better with multicore hosts (FG#37).
457 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
458 - Internal cleanups and refactoring around models, to prepare for programmatic
460 - Modernize our tests of the performance models
464 - New section: Configuring the logs from the command line.
465 - New section: Programming API of logs.
466 - New section: Getting realistic results
467 - New section: Examples of SMPI Usage
468 - New part: SimGrid's Internals, with a section on Design goals
469 - Move back to the classical Breathe tool chain, and fix many small glitches.
472 - Handle leak checking: SMPI will now display a message if MPI handles are still
473 present at the end of a simulation. Option smpi/list-leaks:n will display the n
474 first leaked pointers and their types.
475 - smpi/display-allocs option : Display Memory allocations analysis, with hints
476 about the largest allocations. This only works for allocations with malloc/calloc
477 and not compiled with SMPI_NO_OVERRIDE_MALLOC
478 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
479 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
480 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
481 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
482 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
483 - Replay now opens and parses the trace file only once for all ranks when possible.
485 - better handling of MPI_PROC_NULL
486 - MPI_Group_incl was broken in some cases
487 - PAPI tracing now fixed
488 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
489 - SMPI will now fail properly when receive size < sent size, instead of
490 just setting MPI_ERR_TRUNCATE and keep going.
491 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
492 - Remove useless build patches and split build/run patches on the page
493 - Build in parallel when possible
494 - Test on yum-based systems
495 - New nightly tested Proxy-apps:
496 - OpenMPI mpi-test-suite
497 - Coral MCB, LSMS, Qbox, CLOMP
498 - ECP/Mantevo Cloverleaf3D, Ember
499 - Codevault bhtree_mpi
502 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
503 - Start to overhaul the implementation to modernize it (still ongoing).
506 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
507 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
508 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
509 - Deprecate xbt_dynar_to_array().
510 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
512 C binding and interface:
513 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
514 - Constness requirement has been lowered for parameter argv of sg_actor_start()
515 and sg_actor_create() It's now (char* const*). The previous version is still
516 available as sg_actor_start_() and sg_actor_create_().
518 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
519 (FG: issues on Framagit; GH: issues on GitHub)
520 - FG#37: Parallel tasks are limited to 1 core per host
521 - FG#62: Running "smpirun -replay" on large networks
522 - FG!46: Fix a few potential memory leaks in SMPI colls
523 - GH#343: Wrong consumed energy when an exec timeouts
525 ----------------------------------------------------------------------------
527 SimGrid (3.26) December 15. 2020
529 The Release release (the French lockdown was eased today).
531 Important user-visible changes:
532 - SimGrid now requires a compiler with C++14 support.
533 Sibling projects should upgrade their FindSimgrid.cmake
534 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
535 several users had difficulties to understand issues when using high bandwidth or
536 small latency events. The new value was already the default for SMPI and
537 should not cause too much performance hit. It can change some simulation
538 timings if simulation was skipping a lot of small events (which was actually
539 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
540 - On some network models (L07, CM02, LV08), an implicit loopback link is created
541 for local communication. Default bandwidth was 500MBps and latency was 15µs,
542 which was rather bad. Default values have been changed to 10GBps and 0 latency.
543 Options network/loopback-lat and network/loopback-bw have been added to adapt
544 these values if needed. Network and routing models already providing separate
545 loopbacks (clusters and such) are not impacted by this change.
546 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
547 metadata that are added to each MPI message by MPI libraries (tag,
548 src/dest, and communicator information). This may change simulation times
549 for very small message sizes, but should be more realistic.
552 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
553 - New plugins: wifi_energy and link_load. See the associated examples.
554 - New examples: wifi networks, and communication suspend.
557 - LTO is now enabled for Intel/clang compilers.
558 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
559 Setting it to "auto" will use all available cores, while setting it to n will
560 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
561 - Remove obsolete runtime option 'contexts/parallel-threshold'.
562 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
563 has been kept for compatibility.
564 - Finally remove obsolete snake_case() aliases for runtime options.
565 - Further improve the documentation.
568 - the on_exit() of each actor is also executed when the simulation deadlocks.
569 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
570 An example is provided for s4u::Comm but it should work with Exec and Io.
573 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
574 https://framagit.org/simgrid/SMPI-proxy-apps
575 - MPI/IO: supports names starting with "./"
576 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
578 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
579 building shared libraries, this may cause issues, so environment variable
580 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
581 - most temporary files should now be created in /tmp dir (or equivalent).
582 If this one does not allow execution of code (noexec flag), this may cause issues.
583 Please use another tmp directory (using TMPDIR or equivalent system variable)
587 - A large refactoring is ongoing, but this should be transparent for users so far.
590 - Drop xbt_str_split().
592 C binding and interface:
593 - The return type of the sg_actor_on_exit() callbacks is now 'void'
594 (int value was previously ignored)
595 - Many C functions were renamed to follow a common naming style
596 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
597 names are kept until version 3.30.
598 - Many MSG tests were converted to the new S4U's interface in C, that
599 was extended for that.
601 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
602 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
603 - FG#41: Add sg_actor_create C interface
604 - FG#43: xbt::random needs some care
605 - FG#48: The Impossible Did Happen (yet again)
606 - FG#50: Suspending an actor executed at the current timestamp fails
607 - FG#52: Zero-seconds timeout : "That's in the past already"
608 - FG#53: Crash while using ns-3 network model
609 - FG#54: How to suspend a comm?
610 - FG!22: Stochastic Profiles
611 - FG!24: Documentation and fix for xbt/random
612 - FG!35: Add a modeling hint for parallel links in doc
613 - FG!36: [xbt/random] Read/Write the state of the RNG
614 - FG!42: Link cumulated load plugin
615 - FG!44: Test condition variable wait_until+notify_all
616 - GF#18137: Allow different stack sizes?
617 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
618 - GH#139: Allow pthread creation in SMPI
619 - GH#336: Packet-level simulation using SMPI?
620 - GH#345: Error in the LMM while migrating a VM
621 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
622 - GH#352: pip install / python setup.py install fail to find pybind11
623 - GH!337: Fix link_energy plugin for wifi platforms
624 - GH!339: Add Mailbox set_receiver method to python binding
625 - GH!344: Cast hugepages macros parameters to int64
627 ----------------------------------------------------------------------------
629 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
631 The Palindrom Day Release.
633 Important user-visible changes:
634 - Improve the Python usability (stability and documentation).
635 - A nasty synchronization bug was ironed out, see also below.
636 - Python's doc was organized as a separate tree, now integrated with C++.
637 - C bindings of S4U were not part of the doc.
638 - The C++ doc was also improved as methods are now split by theme.
639 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
640 - This is mandatory to use the Java bindings.
641 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
642 - SimDAG++: Automatic dependencies on S4U activities (experimental)
643 - Some features are already implemented but not all of them
644 - Cannot block an activity until it's scheduled on a resource
645 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
646 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
649 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
650 - Drop support for Viva/Triva (old visualization tools), hence removed
651 TRACE_get_node_types() and TRACE_get_edge_types() functions.
654 - Require pybind11 version 2.4.
655 - Greatly improve locking during context switches with Python.
656 - Declaring Python callbacks to this_actor::on_exit is now working.
657 - No more tests are failing currently, even on win32.
660 - Actor: Merge signals on_migration_start/end into on_host_change
661 - Actor: Rename migrate() into set_host()
662 - Disk: Allow users to get the read and write nominal bandwidth values
663 - Exec: Implement wait_for(timeout)
664 - Io: Implement wait_for(timeout)
667 - Parse errors now raise a simgrid::ParseError that you may want to catch.
670 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
671 been renamed to fix a spelling error. As usual, the previous definitions are
672 kept with a deprecation warning. However, to avoid runtime errors, it is now
673 forbidden to override these deprecated methods in a derived class. Please use
674 the new names immediately if you need to override them.
677 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
678 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
679 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
680 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
681 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
682 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
683 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
684 - New module: random, providing classical random numbers generators.
687 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
688 allocations' size above which they are considered "shared" by default (as if
689 they were performed through SMPI_SHARED_MALLOC macros).
690 Default = 0 = disabled feature.
691 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
692 This can cause some troubles if codes are already overriding these. If this
693 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
694 help, but will make this feature unusable.
695 - Input check is now performed much more consistently and outputs precise
696 warnings for each error found
697 - New supported calls : MPI_File_set_view, MPI_File_get_view
698 - MPI I/O now only supports disks, as storage is discontinued.
699 - MPI I/O allows opening files without specifying absolute path
700 - Note : SMPI configuration options are now initialized by smpi_init_options()
701 call, which should be called if SMPI is not being used through smpirun.
703 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
704 - FG#9: Python bindings crashing
705 - FG#39: Missing s4u::Comm::wait_any_for example
706 - FG#42: Add support for ThreadSanitizer (TSan)
707 - FG!19: Removing RngStream
708 - FG!20: A module for RNG calls
709 - FG!21: Choice between ad-hoc and standard distributions implementations
711 - FG!25: Fix link in Mutex doc
712 - FG!26: Fix links in SMPI interface doc
713 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
714 - GH#305: Unscheduled tasks are still excuted
715 - GH#313: smpirun: manual outdated w.r.t. --help
716 - GH#321: [S4U] Get task remaining work ratio
717 - GH#323: Crash when an actor turn off his physical host
718 - GH#335: Missing links on dragonfly example svg
719 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
720 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
722 ----------------------------------------------------------------------------
724 SimGrid (3.24) October 9. 2019
726 The Clean Disk Release.
729 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
730 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
731 thus becomes the default resource to execute I/O operations. Examples
732 and the file system plugin have been modified to use disks instead of
733 storage elements. Storage can still be used but is doomed to disappear.
734 Users are thus adviced to update their simulators accordingly.
735 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
736 for consistency with pthread_barrier_wait()
737 - Host::get_englobing_zone() returns the englobing netzone
738 - Actor::on_destruction is now called in the destructor
739 Actor::on_termination new signal called when the actor terminates
741 - Global signals are now part of the Engine:
742 - on_platform_creation: after config settings, before the XML parsing
743 - on_platform_created: right after the XML parsing
744 - on_time_advance: each time the clock advances
745 - on_simulation_end: after simulation, before cleanups
746 - on_deadlock: as the name implies.
748 - sg_{actor,host,link}_{data,data_set}() now all exist.
749 Use them to attach user data to the object and retrieve it.
752 - Introduce an experimental Wifi model. It sounds reasonable
753 according to the state of the art, but it still has to be properly
754 validated, at least against ns-3 if not against reality.
755 - Improved the usability of ns-3. Several bugs were ironed out.
756 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
757 It is now expressed as 'idle:epsilon:allCores' to properly model the
758 consumption of non-whole tasks on mono-core hosts. More info in the doc.
761 - convert a new set of functions to the S4U C interface and move the old MSG
762 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
765 - Fortran bindings for DVFS have been removed.
766 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
767 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
768 - SMPI can now be selected by cmake's find_module(MPI) with
769 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
770 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
771 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
772 - trace-call-location can be used with TI traces, and replayed, and is compatible
773 with smpi/comp-adjustment-file.
774 - sleep events are now correctly traced/replayed.
775 - Default for trace-call-location is now to use file names and not full paths.
776 To revert to previous behavior (in case of collision of filenames), option
777 "smpi/trace-call-use-absolute-path" can be set to yes.
780 - Use the included xxHash as an hash implem when C++14 is usable.
781 - Option model-checker/hash was removed. This is always activated now.
782 - New option smpi/buffering controls the MPI buffering in MC mode.
783 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
784 This is useful to check for MPI compliance.
787 - New section on plugins: how to define a new one, and existing ones.
790 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
791 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
794 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
795 and <mount> tags. The rationale is finish to decouple what should be handle
796 by SimGrid kernel (i.e., resources and activities) from plugins built in
797 user space (e.g., the file system plugin). This new tag currently accepts
798 only three attributes (id, read_bw, and write_bw). All the other information
799 that was declared with the storage related tags now has to be expressed as
800 properties. An example of platform using this new tag is available at
801 examples/platforms/hosts_with_disks.xml
804 - 'expect signal' can now accept more than one potential signal.
806 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
807 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
808 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
809 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
810 - FG#31: per-actor data
811 - FG#34: SG_BARRIER_SERIAL_THREAD?
812 - FG#35: model-checker does not like buster-produced binaries
813 - FG!13: MC: complete workaround in the error msg seen on modern systems
814 - FG!15: execute_flops now logs compute
815 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
816 - FG!17: ns-3: unblock the right number of communications + others issues
817 - FG!18: Improving the performance of the ns-3 bindings
818 - GH#189: Energy consumption of parallel tasks
819 - GH#219: Error in the throughput of TCP transfer
820 - GH!330: Minor polishing of English: Configuring_Simgrid
821 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
822 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
823 - GH!334: Rework energy plugin, again
825 ----------------------------------------------------------------------------
827 SimGrid (3.23.2) July 8. 2019
830 - Nicer introduction page.
831 - Migrate the "Deploy your application" page to the new doc.
832 - Move Java as a subtree of MSG.
835 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
838 - Drop xbt_dynar_sort_strings().
841 - Really fix FG#26: Turning off a link should raise NetworkFailureException
842 - FG#27: Wrong exception thrown to wait_any when link is turned off
843 - GH#328: Java: Canceling multiple tasks in a single vm/host
845 ----------------------------------------------------------------------------
847 SimGrid (3.23) June 25. 2019
849 The Exotic Solstice Release.
852 - SunOS and Haiku OS support. Because exotic platforms are fun.
853 - Stop setting random seed with srand48() at initialization.
854 - Use addr2line as a fallback for stacktraces when backtrace is not available.
855 - Build option -Denable_documentation is now OFF by default.
856 - Network model 'NS3' was renamed into 'ns-3'.
859 - Simgrid can now hopefully be installed with pip.
862 - wait_any can now be used for asynchronous executions too.
865 - New log appenders: stdout and stderr. Use stdout for xbt_help.
866 - Drop xbt_dict_dump.
869 - 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
870 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
871 - MPI interface is now const correct for input parameters
872 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
873 - Fortran bindings for async collectives.
874 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
877 - Remove option 'model-check/record': Paths are recorded in any cases now.
878 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
879 always sparse. That's an immense gain in memory, and can even be
880 faster because of cache effects. So there is no need to clutter the
881 code to allow the user to go for the unefficient mode.
884 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
885 models proved to be more accurate than these old experiments.
887 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
888 - FG#1: Broken link in error messages
889 - FG#2: missing installation documentation
890 - FG#3: missing documentation in smpirun
891 - FG#6: Python bindings not available on PyPI
892 - FG#7: simple cmake call requires doxygen
893 - FG#8: make python bindings an optional dependency
894 - FG#10: Can not use MSG_process_set_data from SMPI any more
895 - FG#11: Auto-restart actors forget their on_exit behavior
896 - FG#12: -Denable_lto=OFF doesn't disable LTO
897 - FG#13: Installs unstripped file 'bin/graphicator'
898 - FG#14: Installs the empty directory 'doc/simgrid/html'
899 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
900 - FG#17: Dead link in doc (pls_ns3)
901 - FG#20: 'tesh --help' should return 0
902 - FG#21: Documentation link on http://simgrid.org/ broken
903 - FG#22: Debian installation instruction are broken
904 - FG#26: Turning off a link should raise NetworkFailureException exceptions
905 - GH#133: Java: a process can run on a VM even if its host is off
906 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
907 - GH#326: Valgrind-detected error for join() when energy plugin is activated
909 ----------------------------------------------------------------------------
911 SimGrid (3.22) April 1. 2019
913 The Easter Christmas Release (this one is somewhat late).
916 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
917 - Not complete yet: asynchronous activities (amongst others) are still missing
918 - Still ongoing: the interface may change in the future. We need more
919 testers! Please report any glitches.
920 - No new project using Java should start now. Please switch to Python.
923 - Some of the internal cleanups may lead to speed improvements:
924 - The hard limitation on the amount of simulated actors with
925 Java+Mac was removed. Now, the available memory is the only limit.
926 - Our refcounting was tidyied, leading to 10% speedups in some cases.
927 - We are still working on making our code robust to the actor kills
928 and hosts' churn. Things are improving, but it's not perfect yet.
929 - Replaced our own code to display a backtrace (that was forking addr2line)
930 with the Boost.Stacktrace library.
931 You won't see your backtraces without this optional dependency.
932 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
933 - Stop setting random seed with srand() at initialization.
936 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
937 XML file version remains 4.2 since old files are still compatible.
940 - Process termination which was broken at version 3.21 has been repaired.
941 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
942 - Hide the examples into examples/deprecated. New users should use Python.
945 - Drop MSG_process_create_from_stdfunc() from the API.
946 This C++ function was a pimple in the C API, made necessary at some
947 point by the Java bindings. This is fixed now.
948 - Hide the examples into examples/deprecated. New users should use S4U.
949 - MSG_process_create and MSG_process_attach now crash if the host used
950 to run this process is off.
951 - Fix the protype of MSG_process_on_exit()
952 Now use (int,void*) callbacks instead of (void*,void*) ones.
953 The implementation was ways too messy to actually work, I guess.
956 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
957 too often and provide more accurate timings.
958 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
959 - MPI_Alltoallw support
960 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
961 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
962 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
963 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
965 - Basic implementation of generalized requests (SMPI doesn't
966 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
969 - Drop sg_cmdline. Please use xbt_cmdline instead.
970 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
971 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
972 OsSemaphore is implemented in a portable way with C++11 threads.
973 This should allow much more threads to be created at the same time,
974 allowing Mac and Java users to simulate many more actors.
975 - Implement the 'thread' factory with std::thread instead of xbt ones.
976 It is not possible to set the stack size with threads anymore, but
977 -fsplit-stack is the way to go nowadays when using threads.
978 - Drop the xbt_os_thread_t module (now unused)
979 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
980 - Drop xbt_str_join_array().
981 - Drop cunit, use Catch2 instead.
984 - Many cleanups in the kernel::activity namespace. This was long
985 overdue, and shall open the path to many future endeavors.
988 - #132: Java: a process can not shut down its own host
989 - #220: S4U: detached send causes exception if sender terminates soon after sending
990 - #261: Document the parameters of parallel execution's constructor
991 - #300: [s4u] BarrierPtr is missing
992 - #314: SMPI args internal cleanup
993 - #316: Fix a bug related to the CPU utilization of multi-core VM
994 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
995 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
996 - #325: Turning off a host has different behavior on sleeping actors and computing actors
998 ----------------------------------------------------------------------------
1000 SimGrid (3.21) October 3. 2018
1002 The Restarting Documentation (TRD) Release.
1005 - Convert the user manual to Sphinx for improved usability.
1006 Unstable version now hosted on https://simgrid.frama.io/simgrid
1007 Many glitches remain; stay tuned.
1008 - Restrict the Doxygen reference API to the installed header files.
1009 (documenting all internals this way is overkill + doxygen scales badly)
1010 - New tutorial on S4U (highly inspired from the old MSG one, but with
1011 a git repository to fork as a starting point, and a docker image)
1012 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
1013 - The Developper manual is still to be converted (not compiled until then)
1014 - Some parts are still missing in this conversion (such as the
1015 platform and deployment sections) while others were blindly converted
1016 and would need more love (such as the configuration flags).
1017 Things will certainly further improve in the future releases.
1020 - Fully reimplement auto_restart mechanism that was utterly broken
1021 (fix #22, #131 and #281 that were all very related)
1022 - Implement semaphores (fix #303 and #291)
1023 - When creating an actor from a function and its parameters,
1024 move-only parameters are not allowed anymore, as it would prevent
1025 the actor to be restartable if its parameters are consumed on run.
1026 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
1027 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
1028 IO operations that can be started, waited for, or canceled as a regular
1030 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
1034 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
1035 - New host variable: 'core_count'
1038 - Remove xbt_os_thread_specific features
1039 - Remove portability wrapper to condition variables
1040 - Remove xbt_os_thread_yield()
1043 - MPICH collective selector now mimics MPICH 3.3b
1044 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
1045 - MPI_Init(NULL,NULL) is now allowed
1046 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
1047 - Command-line arguments should now be handled properly in Fortran simulations
1048 - MPI Attributes and related callbacks should work in Fortran
1049 - Apps using C + Fortran should now work
1050 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
1051 - Support MPI_Error_String
1054 - Due to an internal bug, Msg.run() must now be your last line.
1055 We hope to fix it in a future release, and we are sorry for the inconvenience.
1058 - #22: Process autorestart seem to only work with CAS01 cpus
1059 - #56: Feature request: dynamic SMPI replay
1060 - #93: simgrid should not eat --help
1061 - #111: How to change the loopback link characteristics?
1062 - #116: Platform cluster radical
1063 - #117: Platform cluster/host order
1064 - #129: Replay_multiple on steroid
1065 - #131: Java : autostart does not work
1066 - #143: Setting a breakpoint at a given time
1067 - #235: Network energy models should be integrated
1068 - #264: Add ptask L07 resource tracing
1069 - #271: Dynamic and manual replay of SMPI traces
1070 - #279: Breakpoints option uses time deltas instead of absolute time
1071 - #281: Daemonized actor no longer daemonized after an auto-restart
1072 - #282: TIT convertor erase traces when using absolute path in the trace list file
1073 - #285: segfault when a s4u actor kills itself with msg process tracing activated
1074 - #286: Pajé traces are not exposing the number of cores
1075 - #287: Command-line parsing should stop when encountering '--'
1076 - #288: MPI_Init(NULL, NULL)
1077 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
1078 - #290: Method to check if a mailbox has a message ready to be consumed
1079 - #291: [s4u] semaphores do not seem available yet
1080 - #294: Alltoall collective implementation of mpich
1081 - #295: Starting a Host as OFF is not supported
1082 - #296: DTD too permissive
1083 - #299: Add a s4u barrier example
1085 ----------------------------------------------------------------------------
1087 SimGrid (3.20) Released June 24. 2018
1089 The proxy snake_case() release.
1092 - The whole API is now using snake_case() instead of mixing it with camelCase.
1093 Compatibility wrappers in place for one year.
1094 - We took the opportunity to increase the consistency of the API.
1095 Please report any remaining glitches.
1096 - New function simgrid::s4u::Host::get_actor_count:
1097 Returns the number of actors running on a specific host.
1100 - SMPI is now tested with ~45 proxy apps from various sources, with none or
1101 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
1102 - Replay: The replay file has been re-written in C++.
1103 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
1104 supported. They are stored in the trace and used when replayed.
1105 - Basic support of MPI_Cancel. Robustness not guaranteed.
1106 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
1107 (only for MPI_COMM_TYPE_SHARED).
1108 - New option: smpi/privatize-libs, to add external shared libs to be privatized
1109 by SMPI. They will be copied locally and loaded separately by each process.
1110 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
1111 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
1112 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
1113 with internal index optind with multiple processes. Only works if MPI_Init has
1114 already been called.
1115 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
1116 syntax without preprocessing tricks.
1119 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
1120 behavior tracing is no longer limited to MSG
1121 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
1122 this means that traces now need to be updated or re-obtained; both Irecv
1123 and Isend lines in your traces have now in total 5 mandatory fields:
1124 <rankid> <command> <to/from rankid> <tag> <size>
1125 To update your traces, it suffices to add a 0 for the tag here.
1126 - TIT now also supports waiting for a distinct request via MPI_Wait.
1127 Wait/Test now wait for a specific request, not just the last one that was
1128 issued. This unfortunately means another update, because we need to
1129 identify which request you want to wait for. We do this via the
1130 triplet (sender, receiver, tag), which needs to be added:
1131 <rankid> <command> <sender> <receiver> <tag>
1132 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
1133 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
1136 - The deprecation of MSG is ongoing (but this should not impact you).
1137 Many MSG functions are now simple wrappers on the C API of S4U. If
1138 you wish to convert your code to S4U, find the S4U counterparts of
1139 your MSG calls in src/msg/msg_legacy.cpp.
1140 - MSG can still be used, but won't evolve anymore.
1143 - Allow to run the Link energy plugin from the command line with
1144 --cfg=plugin:link_energy
1145 - Rename Energy plugin into host_energy
1146 - Rename Load plugin into host_load
1149 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
1150 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
1151 - Improve performance when handling timeouts of simix synchros.
1154 - Config: the C API is now deprecated (will be removed in 3.23), and
1155 the C++ API has been slightly improved.
1158 - Fix several build issues on OSX.
1159 - Move simgrid_config.h to simgrid/config.h (old header still working)
1162 - #143: Setting a breakpoint at a given time
1163 - #150: Inconsistent event names in SMPI replay
1164 - #258: daemonized actors hang after all non-daemonized actors have completed
1165 - #267: Linker error on unit_tmgr
1166 - #269: SMPI: tracing of MPI_Wait/all/any broken
1167 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
1169 ----------------------------------------------------------------------------
1171 SimGrid (3.19.1) Released March 22 2018
1173 The Fixed ABI Release.
1175 As you may know, we are currently refactoring SimGrid in deep.
1176 Upcoming SimGrid4 will be really different from SimGrid3: modular,
1177 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
1179 Our goal is to smooth this transition, with backward compatibility
1180 and automatic update paths, while still progressing toward SimGrid4.
1182 SimGrid remains open during works: The last pure SimGrid3 release was
1183 v3.12 while all subsequent versions are usable alpha versions of
1184 SimGrid4: Existing interfaces remain unchanged, but the new S4U
1185 interface is budding and the internals are deeply reorganized.
1187 Since 2015, we work hard to reduce the changes to public APIs. When
1188 we need to rename a public library symbol in S4U, we let your
1189 compiler issue an explicative warning when you use the deprecated
1190 function. These messages remain for four releases, i.e. for one full
1191 year, before turning into an error. Starting with v3.15, your can
1192 also adapt to API changes with the SIMGRID_VERSION macro, that is
1193 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
1195 Starting with this v3.19.1, our commitment to reduce the changes to
1196 the public interfaces is extended from the API to the ABI: a program
1197 using only MSG or SimDag and compiled against a given version of
1198 simgrid can probably be used with a later version of SimGrid without
1199 recompilation. We will do our best... but don't expect too much of
1200 it, that's a really difficult goal during such profund refactoring.
1202 The difference between v3.19 and v3.19.1 is that the former was
1203 accidentally breaking the ABI of MSG, while the later is restoring
1206 S4U and kernel APIs will still evolve until SimGrid4, with one-year
1207 deprecation warnings as currently. In fact, cleaning up these
1208 interfaces and converting them to snake_case() is one release goal of
1209 v3.20. But don't worry, we are working to smooth this upgrade path.
1211 In summary, new projects should start with S4U to benefit of the
1212 future, but old MSG projects should still be usable with no change.
1214 ----------------------------------------------------------------------------
1216 SimGrid (3.19) Released March 20 2018
1218 The Moscovitly-cold Spring release.
1220 SMPI user-visible improvements:
1221 - Support of MPI_Type_create_subarray
1222 - Unimplemented MPI calls now abort (instead of issuing a warning)
1223 Some harmless calls around errhandlers still only issue a warning.
1224 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
1225 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
1226 is now possible (mmap privatization was refactored to this end).
1227 Adding new SMPI instances should also be technically doable now.
1229 SMPI internal cleanups:
1230 - Removed index notion from SMPI -- no more "getPid() - 1"!
1231 In tracing files, actors are called rank-<PROCESS-ID> but these are
1232 NOT the real ranks (tracing several communicators is not supported yet).
1233 - Deprecate smpi_process_index() (will be removed in 3.22)
1234 - Better testing on classical proxy apps, through an external project:
1235 https://github.com/simgrid/SMPI-proxy-apps/
1238 - Use a graphical TOC to make it easier to find the documentation you need
1239 - Revamp the MSG tutorial
1242 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
1243 - Remove parameter of MSG_process_killall().
1244 Resetting the PID was bogus anyway (several actors could have the same PID).
1247 - Execution->setHost() can be called after start(), to migrate it.
1248 - Comm::test_any() is now implemented.
1249 - s4u::Actor now has onCreation() and onDestruction() signals.
1250 - Install some kernel header files for the users' plugins and more.
1251 Warning, their API is really not stable yet.
1252 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1253 include/simgrid/kernel/routing/*Zone.hpp
1256 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1258 Build System and other cleanups:
1259 - Remove unused run-time parameter "tracing/onelink-only".
1260 - SimGrid now works with intel compilers (MC still buggy)
1261 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1262 - tesh kills subprocesses on timeout (but not on windows)
1265 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1268 - #194: Feature request: simgrid::s4u::Comm::test_any()
1269 - #245: migrating an actor does not migrate its execution
1270 - #253: Feature Request: expose clusters as objects
1271 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1272 - #255: Tesh broken on Windows
1273 - #256: Modernize FindSimGrid.cmake
1274 - #257: Fix (ab)use of CMake install
1276 ----------------------------------------------------------------------------
1278 SimGrid (3.18) Released December 24 2017
1280 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1282 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1283 - Support for MSG should not be removed from SimGrid before 2020,
1284 but future evolutions will be limited to the integration of
1285 user-provided patches.
1288 - s4u::onDeadlock() signal, run before stopping the simulation.
1289 - s4u::Exec: asynchronous executions (abstraction of background
1291 - s4u::Host->getLoad() returns the achieved speed in flops/s
1294 - Link::name() is deprecated and will be removed in v3.21, use
1295 Link::getCname() instead.
1296 - Mailbox::getName() changed to return a std::string, use
1297 Mailbox::getCname() instead to get a char*.
1298 - Storage::getName() changed to return a std::string, use
1299 Storage::getCname() instead to get a char*.
1300 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1301 to have both allocation and deallocation of the map in user space.
1304 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1305 The amount of flops remaining to do is a vector, not a scalar.
1306 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1307 name implies on both sequential and parallel tasks.
1308 - Both changes fix GitHub's #223 using PR #237 as a basis.
1309 Thanks Michael Mercier.
1310 - Most examples were converted to S4U and hidden elsewhere at tests.
1311 Rationale: we still want MSG to work; we want newcomers to use S4U.
1314 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1315 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1318 - Switch to the faster dlopen privatization mechanism by default
1319 - Documentation improvments
1322 - Live migration is getting moved to a plugin. Dirty page tracking is
1323 the first part of this plugin. This imply that VM migration is now
1324 only possible if one this function is called:
1325 - C/MSG: MSG_vm_live_migration_plugin_init()
1326 - C/C++: sg_vm_live_migration_plugin_init()
1327 - Java: Msg.liveMigrationInit()
1328 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1332 - Define class simgrid::xbt::Path to manage file names.
1333 - Removed unused functions:
1334 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1335 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1336 xbt_graph_export_graphviz()
1337 - xbt/str.h: xbt_str_join()
1338 - Remove unused datatypes:
1339 - xbt/heap.h: use std::priority_queue or boost::heap instead
1340 - xbt/swag.h: use boost::intrusive::list instead
1343 - New link_energy plugin for the consumption of the links.
1344 - All of the operations on files and storage contents have been
1345 packaged into a plugin (src/plugins/file_system). The current
1346 public interface can be found in
1347 include/simgrid/plugins/file_system.h
1348 To use these functions you now have to initialize the plugin by
1349 calling MSG_storage_file_system_init() just after calling
1350 MSG_init() or sg_storage_file_system_init() just after creating
1354 - Remove the undocumented/untested tag <include>
1357 - Remove viva specific tracing as the tool is no longer maintained
1360 - #248: Kill of finished processes leads to segfault
1361 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1362 - #239: please implement signal s4u::onDeadlock()
1363 - #230: segfaults when exit() without run()
1364 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1365 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1366 - #222: Actor::kill() doesn't really kill and segfaults
1367 - #221: odd LMM warning when killing an actor
1368 - #120: Memory leak when the processes are forcefully killed
1370 ----------------------------------------------------------------------------
1372 SimGrid (3.17) Released October 8 2017
1374 The Drained Leaks release: (almost) no known leaks despite the tests.
1376 Even very long simulations will preserve your memory: our testsuite
1377 stresses SimGrid for over 45mn on fast machines for a coverage of
1378 over 80%, and there is only one single known leak, of about 4kb.
1382 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1383 handler may be given to free resources if the comm cannot be completed.
1384 - this_actor::send(mailbox) is now mailbox->put()
1385 - New: simgrid::s4u::Comm::wait_all()
1386 - New: Host.execute() for remote executions.
1389 - Unused option network/sender-gap is removed.
1392 - Deprecate MSG_task_isend_with_matching(): This unused feature
1393 really complicates our internals. Will be removed in v3.20.
1396 - Improved context termination. It is now possible to free resources, even
1397 when a process is forcibly killed.
1398 - ContextBoost: add support for Boost versions above 1.61.
1401 - Bring back run-time option --cfg=exception/cutpath to remove exception
1403 - Removed unused functions:
1404 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1405 xbt_str_rtrim(), xbt_str_trim().
1406 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1409 - Removed header files obsolete since SimGrid 3.12:
1410 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1412 - #3: SD_exit should be made optional
1413 - #120: Memory leak when the processes are forcefully killed
1414 - #159: Threading test regression in Actor refcounting
1415 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1416 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1417 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1418 - #191: VM migration and pstate
1419 - #192: Updating the energy consumptions of all hosts crashes with VMs
1420 - #195: All actors have PID=0 in the logs
1421 - #204: Sometimes segfault with thread contexts and mmap privatization
1422 - #222: Actor::kill() doesn't really kill and segfaults
1423 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1425 ----------------------------------------------------------------------------
1427 SimGrid (3.16) Released June 22. 2017.
1429 The Blooming Spring Release: developments are budding.
1432 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1433 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1434 - Java JAR file should be finally fully working on Mac OSX too.
1436 XML platforms: Switch to platform v4.1 format.
1437 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1438 - <zone> can be used as a synonym for the now deprecated <as>
1439 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1440 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1441 - <actor> can be used as a synonym for the now deprecated <process>
1442 - state_file and avail_file periodicity is now easier to express
1443 (check the documentation) the old behavior should still work.
1444 * Storage: not backward compatible, but it seems that nobody ever used it
1445 - Remove attribute 'content_type' of <storage_type>: was never used
1446 - Make attribute 'model' of <storage_type> optional: for future usage
1447 - Remove Bconnection model property: was never (in)validated. Replaced by
1448 the maximum of the read and write bandwidth as a resource constraint.
1451 - New and Backwards Compatibility break:
1452 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1453 When one wants to get the list of tasks whose states have changed during a
1454 simulation round, s/he has to allocate and free a dynar and use it as
1455 argument to this function. The former SD_simulate (double how_long)
1459 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1460 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1463 - New function to update the consumption of all hosts at once.
1464 - Fix the model for multi-core hosts, linear in the amount of busy
1465 cores with an abnormality for pIdle (see paper).
1466 BUG: Still not satisfactory for ptask on mono-cores.
1469 - The netzone are now available from the MSG API.
1470 The old names still work, but are now deprecated.
1473 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1474 - New option: smpi/keep-temps to not cleanup temp files
1475 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1476 mallocs mode is used, allows to change the size of the fake file used
1477 (default 1MB), to potentially limit the number of mappings for large runs.
1478 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1479 - Fortran ifort and flang compilers support
1480 - New RMA calls supported (experimental):
1481 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1482 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1483 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1484 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1485 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1486 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1489 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1490 - Behind the scene, ActivityImpl are now automatically refcounted.
1491 Removing this nasty bottleneck will greatly simplify our internals.
1494 - Replay: New function xbt_replay_action_get():
1495 Retrieve the function previously associated to an event type.
1496 - DROPPED FUNCTION: xbt_str_varsubst()
1497 - DROPPED MODULE: strbuff. We don't need it anymore.
1498 - DROPPED MODULE: matrix. We don't need it anymore.
1499 - DROPPED MODULE: lib. We don't need it anymore.
1501 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1503 SimGrid (3.15) Released March 22. 2017
1505 The Spring Release: continuous integration servers become green
1506 We fixed even the transient bugs on all target architectures:
1507 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1508 Windows, FreeBSD, NetBSD.
1510 - Rename NetCards to NetPoints.
1511 This was intended to help NS3 users, but that's not a netcard.
1512 That's a point in the routing algorithm, let's avoid wrong simplifications.
1515 - New: MSG_process_yield(). Stop and yield to other processes.
1516 - New: MSG_process_daemon(). Daemon processes are automatically killed
1517 when the last non-daemon process terminates
1518 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1519 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1520 to make room for the upcoming network energy plugin.
1521 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1524 - Ensure that an actor can kill itself with Process::exit()
1525 - Kill the obscure NativeException. Nobody want to survive the issues
1526 it denotes, so use JniException that is a RuntimeException (not to
1527 be caught explicitly).
1528 - Partial bug fix in initialization. SimGrid flags on command line were
1529 consumed at C level but stayed in the original Java String[] args.
1530 This could mess users' args[i] if SG flags were not put at the end of
1532 The SimGrid flags are now removed from the Java arguments. However,
1533 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1534 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1535 It might be if you have --log or --cfg flags in the command line.
1536 - Fix numerous memleaks all around the place. In particular, around VMs.
1540 - s4u::onPlatformCreated: right before the simulation starts
1541 - s4u::onSimulationEnd: right after the main simulation loop
1542 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1543 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1544 event from the availability_file changes the avail speed.
1545 - Links are now usable from s4u
1546 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1547 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1548 - New examples: The conversion of MSG examples to S4U has begun
1549 - Actors: create, kill, migration, and suspend
1550 - Applications: master-worker and token-ring
1551 - Action replay: communications and storage
1552 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1555 - Backwards Compatibility breaks
1556 - The SD_route_get_list and SD_route_get_size functions have been removed.
1557 They are replaced by the unique
1558 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1559 The route from src to dst is built in the links dynar whose size can
1560 be obtained with xbt_dynar_length.
1561 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1562 sg_host_route_bandwidth and sg_host_route_latency.
1563 Macros ensure the backwards compatibility, but you should fix your code
1566 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1567 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1568 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1569 We don't support 100% of it, but it should be enough. Contact us if not.
1570 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1571 naive and probably inaccurate)
1572 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1573 mapped onto a unique small file using some system magic.
1574 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1575 - Fortran: cleanups, fixes, support of user-added operations
1576 - RMA: MPI_Accumulate are applied in correct order.
1577 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1578 - Better support for MPI_IN_PLACE
1579 - Support for MPI_Win attrs and keyvals.
1580 - Support MPI_Comm_group_create, MPI_Type_size_x
1581 - MPI ops should be properly applied to all allowed types
1582 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1583 algorithm by a non failing one from mpich.
1586 - Kill the fifo data container: we don't use it anymore.
1589 - A <cluster> can now be created with different speed values to represent pstates.
1590 see examples/platforms/energy_cluster.xml for an example.
1592 SimGrid (3.14.159) Released December 28. 2016
1594 The previous release was only a crude approximation of a Pi release;
1595 this one should be a bit better.
1597 - Revert a last minute change that broke on Mac OSX.
1598 - Fix the changelog and NEWS files.
1599 - Improve the documentation, in particular of the routing module.
1600 - Integrate some patches flying around in Debian.
1602 SimGrid (3.14) Released December 24. 2016
1605 * The whole documentation was reworked and reorganized. There is
1606 still much room for improvement here, but we're on it.
1609 * We now need python3 for our testing infrastructure.
1610 * Model checking needs libevent
1613 * Now works on FreeBSD
1616 * Privatization now works on FreeBSD
1618 * Privatization is now activated by default in smpirun
1619 You can turn it off with -no-privatize if something goes wrong.
1621 * Call-location tracing for SMPI.
1622 You can add the exact location (filename / linenumber) of an MPI call to
1623 your trace files and slow down or speed up the simulation between two
1624 consecutive calls by using an adjustment file (see the documentation).
1626 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1627 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1628 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1629 values were ignored in some cases. The timings of these functions can now
1630 be significantly different.
1632 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1633 smpi/running-power is renamed to smpi/host-speed
1635 * smpi/grow-injected-times option to enable or disable multiplication of the
1636 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1637 make simulation less precise (but also much faster).
1639 * smpirun script should be (much) faster for large deployments.
1641 * SMPI tracing: fixed issue with poor matching of send/receives.
1643 * Replay: Fix broken waitall
1645 New functions and features
1646 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1648 Dropped / renamed functions and features
1649 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1650 - MSG_mailbox_is_empty() -> Mailbox::empty()
1651 - MSG_mailbox_front() -> Mailbox::front()
1652 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1653 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1654 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1655 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1656 * Don't pass the free_f parameter to property related functions:
1657 - MSG_host_set_property_value()
1658 - MSG_as_router_set_property_value()
1659 - MSG_storage_set_property_value()
1660 * VM properties. Since msg_vm_t are msg_host_t, just use
1661 MSG_host_get_property_value() and friends
1662 * VM I/O related things:
1663 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1664 - Unimplemented save/restore methods
1665 * MSG_as_router_get_property_value() was redundant with
1666 MSG_environment_as_get_property_value().
1667 - Removed MSG_as_router_*propert*() functions
1668 - Added MSG_environment_as_set_property_value() for consistency
1669 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1670 Well, they are still there for now, but deprecated with a warning.
1671 Please switch to xbt_dict_new_homogeneous() before this is removed
1673 * Task affinity. Its intended behavior (that was very badly tested
1674 and probably not really working) was deceiving what most users
1675 would have hoped here.
1676 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1677 * Option network/coordinates is now useless and should be dropped.
1680 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1683 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1684 * Added Dragonfly topology support
1687 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1688 EiBps, EBps, Eibps, Ebps for bandwidth.
1689 They may become useful to some lucky ones.
1692 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1693 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1694 * ex/async_waitAll: new example, on asynchronous communications
1697 * Memory usage should be decreased for simulations with a large number
1698 of processes. This also helps for SMPI.
1700 SimGrid (3.13) stable; urgency=low
1702 The Half Release, a.k.a. the Zealous Easter Trim.
1704 - We removed half of the lines, that were mostly experimental cruft.
1705 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1706 (not counting blanks and comments -- according to openhub.net)
1707 - The internals are now compiled in C++ (and will soon be clean C++)
1708 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1709 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1711 * Backwards Compatibility breaks
1712 - Removed Lua simulation bindings (switch to C or Java for that).
1713 Lua can still be used to describe platforms
1714 - Removed Java kernel plug-ins.
1715 Will be reintroduced after the ongoing major internals reorg.
1717 - the following functions were removed.
1718 They were too specific and should be reimplemented in a generic
1719 way, with filter function.
1720 - MSG_task_listen_from_host
1721 - MSG_mailbox_get_count_host_waiting_tasks
1722 - MSG_mailbox_put_with_timeout was removed.
1723 Please use MSG_task_send_with_timeout instead.
1725 - the SD_application_reinit function was removed. It has been a noop for a while.
1726 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1727 replaced by a more flexible mechanism.
1728 - The following functions thus do not exist anymore
1729 - SD_workstation_get_access_mode
1730 - SD_workstation_set_access_mode
1731 - SD_workstation_get_current_task
1732 - Basic estimation functions have been removed but can easily be replaced
1733 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1734 - SD_workstation_get_computation_time => amount / sg_host_speed()
1736 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1737 Use VM.getSpeed()*load/100 for the legacy behavior.
1739 - option enable_tracing was removed. It was not doing anything for a while.
1740 - In the ModelChecker:
1741 - the model-checker now ptraces the model-checked process which means
1742 you cannot use a debugger on the latter anymore (we might make this
1743 optional in the feature);
1744 - removed soft-dirty page tracking;
1745 - remove model-checked side snapshot management,
1746 MC_snapshot() and MC_compare_snapshot();
1747 - keep the MC_cut() function as a stub (it was not really working
1748 in the previous release).
1751 * All options are consistently in kebab-case. Old names are kept as alias.
1754 * Switch to platform v4 format.
1755 - Rename from 'power' to 'speed' the attributes describing the amount of
1756 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1757 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1758 - In <host> and <link>, attributes availability and state are gone.
1759 It was redundant with state and availability traces, and with peak values.
1760 - In <cluster>, attributes availability_file and state_file are gone.
1761 It was too complex and unused.
1762 - Kill <gpu>. Was not doing anything.
1763 - The DOCTYPE points to the right URL:
1764 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1765 (the file at this address now documents the changelog since its v1)
1766 - A warning is emitted for unit-less values (they are still accepted).
1767 - speed. Default: 'f' or 'flops'. Also defined:
1768 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1769 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1770 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1771 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1772 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1773 - latency. Default: 's' second. Also defined:
1774 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1776 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1777 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1780 * s4u::Host is now the preferred public interface to the Host features.
1781 sg_host_* functions are C bindings to the exact same behavior
1782 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1785 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1786 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1787 They were deprecated since a few years, and probably did not even compile anymore.
1790 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1791 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1792 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1793 encouraged to update their codes. A list of the modified functions can be found at the end of
1794 include/simgrid/simdag.h
1797 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1798 the lambda or closure passed as a parameter will run in kernel mode.
1799 All the callback functions should be rewritten to that interface at some point.
1802 * Reorganizing and cleaning the internals all around the place.
1805 * Remove old default barrier/bcast buggy algorithms (see #18407)
1806 * Various bug fixes to handle more codes
1807 * Remove the need for the --foreground option of smpirun (it is still
1808 accepted for backward compatibility).
1811 * Kill the setset data container: MC don't use it anymore.
1812 * Kill the queue data container: it made more sense with GRAS.
1813 * Kill the xbt_peer_t data type: it's useless without GRAS.
1814 * Kill rm_cb feature of config sets: it was never useful.
1815 * Kill graphxml parsing feature. It was not used.
1816 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1818 - xbt_dynar_sort_strings(), when the content is char*
1819 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1820 They throw exceptions on invalid input;
1821 * C++ support for declaring CLI flags (simgrid::config::Flag);
1822 * class for abstracting different signal backends (simgrid::xbt::signal).
1823 with no external dependencies (we need very simple signals).
1826 * refactoring and cleanup of the code;
1827 * ongoing process to cleanly separate the model-checking algorithms
1828 from the code model-checking support.
1830 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1832 SimGrid (3.12) stable; urgency=low
1834 The Facelift Release.
1837 * Require g++ v4.7 at least to not speak prehistorical C++.
1838 * Require Boost 1.48 (for signal2 component).
1839 * Java must be version 7 at least when activated.
1840 * Builds on Windows again (including Java bindings).
1841 * Tracing is now always enabled (no way to turn it out)
1842 * Remove GTNetS. It was not working anyway.
1843 * Various cleanups in the cmake scripts.
1844 * Move headers around to sort them out on installed systems:
1845 - instr/instr.h -> simgrid/instr.h
1846 - instr/jedule/* -> simgrid/jedule
1847 - simdag/datatypes.h was removed
1848 - simdag/simdag.h -> simgrid/simdag.h
1849 - msg/datatypes.h was removed
1850 - msg/msg.h -> simgrid/msg.h
1853 * Interface improvement:
1854 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1855 - Sanitize the interface in MSG_task_ module:
1856 - Merge two functions that were close enough but misleading:
1857 set_compute_duration(t) -> set_flops_amount(t)
1858 get_remaining_computation(t) -> get_flops_amount(t)
1859 - set_data_size(t) -> set_bytes_amount(t)
1860 get_data_size(t) -> get_bytes_amount(t)
1861 - Massive cleanups in the functions related to the energy
1862 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1863 - New: MSG_host_get_pstate()
1865 - msg/energy/onoff: switching hosts on and off
1867 * Interface improvement:
1868 - Rename Host.isAvail() to Host.isOn()
1869 - Rename Process.currentProcess() to Process.getCurrentProcess()
1870 - Rename Task.setDataSize() to Task.setBytesAmount()
1871 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1873 - #18874: Actually allows the GC to reclaim tasks
1877 - SIMIX_process_throw: raises an exception in a remote process
1878 * Refactoring: Separate sub-modules
1879 - libsmx: the public interface, as libc in a real system
1880 - popping: the strange dance that converts a user request into a kernel handling
1881 - smx_context_*: the virtualization mechanisms that embed the user code
1882 - smx_*: the handling of each simcalls
1883 * Interface cleanup:
1884 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1885 * Rename smx_action_t into smx_synchro_t, making explicit that these
1886 things are used to synchronize processes with their environment.
1887 For example, a communication is a sort of synchronization involving
1888 the communicating processes (that may block until the exchange) and
1889 the platform. The same can be said from computations, etc.
1891 - #18888: segfault when a process ends before its kill_time
1895 - 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.
1896 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1897 - MPI_Type_set_name, MPI_Type_get_name
1898 - MPI_*_c2f and MPI_*_f2c functions
1899 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1900 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1901 - Activate a lot of new tests from the mpich 3 testsuite
1903 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1904 - InfiniBand network model added: Based on the works of Jerome Vienne
1905 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1906 - When smpi/display_timing is set, also display global simulation time and application times
1907 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1908 * Collective communications
1909 - 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
1910 - 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/).
1911 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1913 - Replay now uses algorithms from wanted collective selector
1914 - Replay can be used with SMP-aware algorithms
1915 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1916 - 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.
1918 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1919 - Correctly use loopback on fat-tree clusters
1920 - Asynchronous small messages shouldn't trigger deadlocks anymore
1921 * Energy/DVFS cleanup and improvement
1922 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1923 - new: smpi_get_host_pstate()
1926 - "Full" network optimization flag was broken since Surf++
1927 - Better handling of precision flags in maxmin
1928 - Fix bug causing sometimes "Impossible" errors
1929 - Properly pass cluster properties to included hosts
1930 * Improvement of the Energy plugin.
1931 - Always update the consumption before returning that value
1932 - New property: watt_off to denote the dissipation when the host is off
1933 - New functions getWattMinAt and getWattMaxAt to retrieve the
1934 dissipation of pstates that we are not currently at.
1935 * Java: class NetworkLink renamed to Link
1936 * New function: simcall_process_get_kill_time()
1937 * Massive rename s/workstation/host/
1938 - That's intrusive, but that's good for the project consistency. Sorry.
1939 - Change config option "workstation/model" into "host/model"
1942 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1943 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1944 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1945 - xbt_dirname and xbt_basename for non-POSIX systems
1947 * The model checker now runs as a separate process.
1948 * The model checker runs must now be launched with the new simgrid-mc program.
1949 * Record/Replay: the MC can display a textual representation of a path in the
1950 execution graph. It can then be replayed outside of the model checker.
1952 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1954 SimGrid (3.11) stable; urgency=low
1959 * Normalizing pointers addresses tool for better diff between logs
1962 * Add cloud examples using new VMs
1963 - examples/msg/cloud/two_tasks_vm.tesh
1964 - examples/msg/cloud/simple_vm.tesh
1965 - examples/java/cloud/cloud.tesh
1966 - examples/java/cloud/migration/migration.tesh
1967 * Add java surf examples:
1968 - examples/java/surfPlugin/surf_plugin.tesh
1969 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1970 - examples/java/surfCpuModel/surf_cpu_model.tesh
1971 * Add SMPI+MSG example:
1972 - examples/smpi/smpi_msg_masterslave/
1977 - msg task destroy cancel test
1978 - msg_host on/off test
1979 * Move all tests in testsuite to teshsuite (adding tesh files)
1980 * Restructure teshsuites
1981 - one folder for each kind of test
1982 * Restructure AddTests.cmake
1984 - structure the order of tests (with sections)
1987 * Add virtual machine
1988 - creation of a VM on a PM
1989 - migration of a VM from a PM to another PM
1991 - MSG_process_join(msg_process_t process, double timeout)
1992 - msg_bar_t MSG_barrier_init(unsigned int count)
1993 - int MSG_barrier_wait(msg_bar_t barrier)
1994 - void MSG_barrier_destroy(msg_bar_t barrier)
1995 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1996 * New option "msg/debug_multiple_use" to help debugging when a task is used
1999 - Improvements and finalization of MSG_storage, MSG_file APIs and their
2001 - Increase code coverage in test suites
2005 * Protect context stacks against stack overflow. The number of protected memory
2006 pages allocated on the top of each stack (1 by default) can be configured
2007 with the new command line option --cfg=contexts/guard_size:<value>.
2008 * Simcalls are now generated by a python script that
2009 - generates files included by SimGrid
2010 - checks that all the functions exist, and proposes prototypes
2012 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
2013 host_set_data, host_get_data
2015 - simcall_process_join(smx_process_t process, double timeout)
2016 * Fix bug where sleeping processing could not be suspended.
2019 * Translate surf models from C to C++
2020 - Generic classes for all models: Model, Resource, Action
2021 - A generic interface for each kind of model (CPU, Network, Storage
2022 Workstation, WorkstationVM)
2024 * Translate surf routings from C to C++
2025 * Add callbacks using sigc++ or boost::signals2
2026 - Add callback functions for resource creation/destruction
2027 - Add callback functions for action state change
2028 - Handle Energy as a plugin
2029 * Replace swag by boost::intrusive
2030 * Add new routing models for clusters. For documentation, see
2031 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
2032 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
2033 parameters for cluster tag
2034 - Fat trees, with topology="FAT_TREE" and
2035 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
2036 - see examples/platforms/torus_cluster.xml and
2037 examples/platforms/fat_tree_cluster.xml
2038 * More documentation
2041 * Hostfiles support host:nb_processes construct to deploy several processes on
2043 * Collective communication algorithms should not crash if used with
2044 improper number of nodes and report the error.
2045 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
2046 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
2047 MPI_Dims_create, MPI_Cart_sub are supported.
2048 * New interface to use SMPI programmatically (still depends on MSG for
2049 some parts, see examples/smpi/smpi_msg_masterslave):
2050 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
2054 * Global variables privatization in MPI executables is now performed at runtime
2055 with the option smpi/privatize_global_variables (default:no).
2056 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
2057 dynamic libraries loaded by the application are not privatized (static
2058 linking with these libraries is advised in this case)
2061 * Options defined in XML work correctly now.
2064 * New cmake option, enable_lib_in_jar, to control whether native libraries are
2065 copied into simgrid.jar or not (ON by default). Use this option if you want
2066 to reduce the size of the installed simgrid.jar, *and* the native libraries
2067 are kept installed elsewhere.
2068 * Surf binding with SWIG (code generated in maintainer mode only):
2069 - plugin to handle callbacks
2070 - CPU model only for the moment
2073 * Supernovae build mode is definitively removed. It was used to improve
2074 inlining and inter-module optimizations. It is nowadays superseded by
2075 link-time optimizations commonly available in compilers.
2076 * Update ns-3 find lib. Bindings for ns-3 should work again now.
2077 * Add boost dependency for surf++
2078 * Add new macro for tests
2079 - ADD_TESH(name <tesh_args>)
2080 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
2084 - xbt_bar_t XBT_barrier_init(unsigned int count)
2085 - int XBT_barrier_wait(xbt_bar_t barrier)
2086 - void XBT_barrier_destroy(xbt_bar_t barrier)
2087 * Make the xbt_os_time module public
2089 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2091 SimGrid (3.10) stable; urgency=low
2093 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
2096 * Reintegrate Java to the main archive as desynchronizing these
2097 package is not acceptable anymore (Java is now considered stable)
2098 * Add explicit synchronization facilities through semaphores
2099 * Bug fix: Task.setDataSize() only changed the C world, not the value
2100 cached in the Java world
2103 * Dramatically change the way files are handled. API and internals changed, but
2104 this part of MSG was not considered as production grade either.
2105 * Add explicit synchronization facilities through semaphores
2106 * Add a new function MSG_host_get_process_list()
2107 * Preliminary DVFS support (see examples/msg/energy/ for details)
2110 * SMPI is now included directly in the libsimgrid as the windows
2111 linker doesn't force us on splitting it anymore.
2112 * Improvements of the SMPI replay tool:
2113 - Most of the collective communications are now rooted in the same process as
2114 in the original application.
2115 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
2116 used until now). Multiple data types can now be used in a trace.
2117 - The replay tool now supports traces produce either by TAU or a modified
2119 - Bug Fix: the compute part of the reduce action is now taken into account.
2120 - Gatherv collective is now supported
2121 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
2122 of smpirun outputs time independent traces for the current run. One file
2123 is created per process. If too many processes are simulated, this behavior
2124 can be changed to one file for all processes by using the
2125 tracing/smpi/format/ti_one_file flag
2126 * smpirun generates the host file if needed (with given host count and platform)
2127 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
2128 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
2129 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
2130 --cfg=smpi/coll_selector:(mpich/ompi)
2131 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
2133 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
2134 happening inside a collective SMPI call.
2135 * Fix the behavior of complex data types handling.
2136 * Make MPI_Wtime another synchronization point to take computations into
2138 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
2139 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
2140 * Add all missing Fortran bindings, SMPI should work with Fortran 90
2141 (no privatization of global variables yet)
2142 * Preliminary DVFS support (see examples/smpi/energy/ for details)
2145 * Verification of liveness properties is now available for SMPI applications
2146 (in addition to MSG applications)
2147 * Bugged examples using SMPI in examples/smpi/mc/
2148 * Add --cfg=model-check/visited option. Allows the verification of infinite
2149 programs. Detection of loops in the execution thanks to the system state
2150 comparison and reduction of the state space to explore. Can be combined with
2151 DPOR for safety properties.
2154 * Allow to change SimGrid configuration (see --help) within the code
2155 thanks to SD_config() as it can be done in MSG.
2156 * Add a new function SD_task_set_amount() upon user request.
2159 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
2160 * Remove rule based routing (no more PCRE dependency)
2161 * Add a limiter_link option to cluster tag, to specify a maximum reachable
2162 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
2163 * Add a loopback_bw and loopback_lat options to cluster tag.
2164 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
2165 routing. Now peers are encapsulated in an AS and have their own private
2166 router but this is transparent.
2169 * Our own implementation of getline is renamed xbt_getline, and gets
2170 used even if the OS provide a getline(). This should reduce the
2171 configuration complexity by using the same code on all platforms.
2172 * New type: xbt_cfg_elm_boolean.
2173 * Allow to use yes/no for boolean configuration options in the command line.
2174 * Allow to disable SimGrid cleanups at exit from command line option.
2175 There are situations where one may want a simulation to end with an exit.
2176 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
2177 annoying when scripting around the simulator. Adding a
2178 --cfg=clean_atexit:no allows to circumvent this issue.
2181 * Lots of memory leaks were corrected in this release.
2182 There is no known memory leaks anymore, in all of our 600+ tests.
2183 * New command line option --version, to get SimGrid version information.
2184 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
2186 * Supernovae builds are deprecated, and expected to be removed in the next
2189 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2191 SimGrid (3.9) stable; urgency=low
2193 The Grasgory release: GRAS is really dead now.
2195 * Complete overhaul of the internal host structures scheme.
2198 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
2199 was considered as experimental and badly maintained since then.
2200 * Keeping it was thus a trap to our potential users, that could take
2201 it instead of MSG or SMPI by mistake despite is pity state.
2202 * GRAS seems to have very few users (if any), and no one volunteered
2203 to maintain it further. It also induces a lot of XBT code (for
2204 portability sake), that must be maintained too.
2205 * For all these reasons, we killed GRAS. If someone wants to revive it
2206 in the future, don't cry, our git history still remembers of GRAS.
2209 * Major overhaul. Merge our documentation again as time proved that
2210 splitting it was really not helping our users.
2211 * Further improve the developer documentation to help newcomers
2212 hacking on SimGrid itself. The user documentation (and in
2213 particular, the beginner documentation) is still in a sorry state.
2216 * Now works on Windows too!
2217 * Much more extensive test suite, from MPICH
2220 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
2221 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
2222 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
2223 * Bug fix: let task be scheduled when the last dependency to be solved is
2224 a control dependency.
2225 * Remove SD_load_environment_script function.
2226 Use the C sg_platf function if you want to declare a platform
2230 * New function: MSG_process_get_number()
2231 * Old function documented: MSG_config()
2232 * Remove MSG_load_platform_script function
2233 Use the C sg_platf function if you want to declare a platform
2237 * Change the default value of the TCP_gamma constant (maximal size of TCP
2238 congestion window) to a more realistic 4MiB value. If you notice changes in
2239 your simulation results, you can fall back to the previous 20k tiny window
2240 by adding --cfg=network/TCP_gamma:20000 on command line.
2241 * (Hopefully) fix a bug wrt periodic availability and state traces
2242 * Bug fix: use default values at start when first event in availability/state
2243 trace is not at time 0.
2246 * remove the "new_" part of function name sg_platf_new_trace_connect
2247 (resulting in sg_platf_trace_connect), since it does not create
2251 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2252 use case where it's really mandatory, and maintaining it was a pain in
2254 * New: xbt_fifo_search(), search an item with a user-provided
2255 comparison function instead of dumb pointer comparison.
2258 * Fix the lua deployment:
2259 Use `simgrid.init_application()` before deployment instead of
2260 `simgrid.msg_register_application()` after.
2263 * Transfer the tracing files into the corresponding modules.
2265 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2267 SimGrid (3.8.1) stable; urgency=low
2269 The "we are told that some people want to also *install* the simgrid
2272 * Add missing file "tesh.1" to the archive.
2274 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2276 SimGrid (3.8) stable; urgency=low
2278 The Psssshiiiit release: SimGrid jumps into the Cloud.
2281 * Add an experimental interface to manipulate VMs. They are mainly
2282 process groups with very few intrinsic semantic, but they should
2283 allow you to build the semantic you want easily.
2284 * New function: MSG_host_set_property_value()
2285 * New function: MSG_process_on_exit(). To clean memory in all cases.
2286 * Bug fixes that made the host (and link) failures unusable.
2287 * Add a way to auto-restart process when the host in which they are
2288 executing comes back (ON_FAILURE="RESTART" on deployment file,
2289 MSG_process_auto_restart_set).
2290 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2291 please stop using the old ones, they are DEPRECATED.
2293 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2294 Please use MSG_init() instead. (reducing the amount of entry
2295 points in the library helps us).
2296 * Make it impossible to link against the wrong version of the lib
2297 * Deprecate MSG_clean(). No need to call it anymore.
2298 * Function MSG_get_host_number() is not deprecated anymore.
2301 * Split the doc into a user guide and a reference guide.
2302 * Start a developper guide to help people hacking on SimGrid.
2305 * Enable tracing by default. This modules rocks you should use it.
2306 * Remove option custom_flags. Now use environment variables CFLAGS
2308 * Use default cmake things to detect lua instead of home grown ones.
2309 * New option "enable_mallocators" to disable mallocators, for debugging
2310 purpose ("on" by default).
2313 * Bug fixes around the resource failures: don't let the processes
2314 survive the host they are running onto.
2315 * Add an interface to auto-restart processes when the host in which they are
2316 executing comes back.
2317 * Ensures that SIMIX_clean is called automatically. It's not part of
2318 the public interface anymore (bindings should be updated).
2321 * Bug fix for when SD_Simulate is called with a positive value: be careful
2322 when comparing doubles. Sometimes they are different for non significant
2324 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2325 parallel task whose initial work is distributed among host according
2326 to the Amdahl's law. Such tasks are created with a parameter alpha
2327 that corresponds to the non-parallelizable part of the computation.
2328 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2329 between two sets of workstations assuming a 1D block distribution (each
2330 workstation owns a similar share of data) on both sides.
2332 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2333 Data redistribution will be automatically scheduled once parent and child
2334 are both scheduled. The filling of computation_amount and
2335 communication_amount structures is now done seamlessly thanks to the chosen
2337 * New function SD_workstation_dump to display various information
2338 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2339 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2340 bandwidth on the route onto which the task is scheduled and the amount of
2342 To divide the nominal bandwidth by 2, the rate then has to be:
2343 rate = bandwidth/(2*amount)
2344 * Compute tasks that have failed can now be rescheduled and executed again
2345 (from their beginning)
2346 * Increasing source code coverage (src/simdag is now covered at 95.8%
2350 * Re-implement time-independent trace replay using SMPI (at the
2351 smpi_smp_* level) instead of MSG. This should replace
2352 examples/msg/actions/actions.c
2353 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2354 hindexed and structs)
2355 * Implement the exchange of non-contiguous data.
2356 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2357 * Correct behavior of smpi/sender_gap and set its default value to 0
2358 * Add option to asynchronously send small messages to allow better
2359 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2360 specifies the size in bytes under which messages will be asynchronously sent.
2361 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2362 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2365 * New C interface to define a platform: XML is now optional.
2366 For more info, please check include/simgrid/platf.h.
2367 * New interface to define random platforms from the C:
2368 For more info, please check include/simgrid/platf_generator.h and
2369 examples/msg/masterslave/masterslave_platfgen.c
2370 * Export a sg_cmdline dynar containing all the arguments we got from
2374 * Two new tracing options for adding comments to trace file so you
2375 can track your experiments (see --help-tracing for details).
2376 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2377 * Adding the SimGrid version that generated the trace file as a comment.
2378 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2379 * Fix to avoid key clashes on Paje links
2380 * Other minor fixes related to the Paje specification
2383 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2384 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2385 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2386 * Greatly improve the robustness of mmalloc to user errors (such as
2387 using an area after freeing it, or freeing it twice)
2389 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2391 SimGrid-java (3.8.1) stable; urgency=low
2393 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2394 Warning: all methods are visible, but only some of them are
2395 implemented so far. Check the source (src/jtrace.c)
2396 for further information.
2397 * New module: org.simgrid.msg.File (SimGrid File management functions)
2398 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2399 * Change the meaning of Process.restart: now restart the process from
2400 the beginning, like MSG_process_restart in C.
2401 * Add Process.setAutoRestart: handling of process restart when failed
2403 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2404 you to retrieve the properties of the processes/hosts
2405 * Deprecate Msg.clean(): you can just forget about it now.
2406 * New function Process.getCount(), that only works when compiling
2407 with the not yet released version 3.9 of the C library.
2409 * New context factory based on Coroutines. It mandates a modified JVM
2410 but then, the simulations run about five times faster, and there is
2411 no limit to the amount of processes (beside of the available memory).
2413 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2415 SimGrid (3.7.1) stable; urgency=low
2418 * Restore the prototype of MSG_process_create_with_environment() to
2419 the pre-3.7 situation by removing the kill_time argument.
2420 * Add a MSG_process_set_kill_time() function instead.
2423 * Fix weird behaviors when dealing with parallel tasks.
2426 * Simgrid is now built as a dll.
2427 * Simgrid-java now works on Windows.
2428 * Simgrid-Java is now included into Windows package.
2431 * First pre-build package for MacOSX.
2434 * Fix compilation when using MSG_USE_DEPRECATED.
2435 * Fix some compilation issues on Macs and Windows.
2436 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2437 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2439 We discovered that the Lua console is broken, but we are missing the
2440 manpower to fix it right now. The problem existed in 3.7 too, so we
2441 are not blocking the release for that. Sorry if you depended on this
2442 feature, any help would be really welcome.
2444 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2447 SimGrid-java (3.7.1) stable; urgency=low
2449 The "Java aint got to be bloated and slow" release
2452 * Various internal cleanups and performance improvement
2453 Simulations are expected to run up to twice faster or so
2454 * Make Process.kill(process) an instance method, not a static one
2455 * User processes are not java.lang.Thread subclasses.
2456 This breaks the compatibility (sorry), but previous API was
2457 brain-dead, making it impossible to have non-trivial
2458 initializations in the process constructor.
2459 * Require a full constructor per Process sub-class.
2460 Kinda breaks the compatibility (sorry), but this allows a much more
2461 efficient way to launch the processes at simulation startup.
2462 * Do not embeed our version of semaphores, java 1.5 can be considered
2463 as sufficiently prevalent for us to not dupplicate its features.
2467 * Add examples for almost every part of the API
2468 We spotted and fixed a lot of bugs in the process
2470 * New module: asynchronous communication API
2471 * New function: Process.sleep()
2472 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2473 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2474 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2476 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2478 SimGrid (3.7) stable; urgency=low
2480 The "spring cleanups (before the next Big Project kicks in)" release.
2483 * We can specify the SMPI latency/bandwidth factor with command line
2484 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2485 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2486 You can also use the "config tag" from platform file by adding this line
2487 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2488 (see "example/platforms/tag_config.xml" to use "config tag").
2489 Note that the command line supersedes the platform file configuration.
2490 * Change the correction factors used in LMM model, according to
2491 the latest experiments described in INRIA RR-7821.
2492 Accuracy should be improved this way.
2493 * Use the partial invalidation optimization by default for the
2494 network too. Should produce the exact same results, only faster.
2495 * Major cleanup in surf to merge models and split some optimization
2496 mechanisms from the core of the models. As a result you can now
2497 specify which model to use (e.g., --cfg=network/model:LV08
2498 --cfg=cpu/model:Cas01) and which optimization mode to use
2499 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2500 Incompatible combinations should err at initialization. See
2501 --help-models for the list of all models and optimization modes.
2502 * The CLM03 workstation model was dropped for simplicity because it
2503 used the deprecated CM02 network model. Use default instead.
2504 * Rename the TCP_gamma configuration option to network/TCP_gamma
2505 * Rename the coordinates configuration option to
2506 network/coordinates, and document it
2507 * Use now crosstraffic keyword instead of the terribly misleading
2508 fullduplex keyword. It is activated by default now in the current
2509 default model, use --cfg=network/crosstraffic:0 to turn it off.
2510 * Ongoing refactoring the model parsing to make XML files optional
2511 See include/simgrid/platf.h for details (still to be completed)
2514 * Major overhaul of the documentation. Almost instructive now :/
2515 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2516 functions and friends. This interface was considered as
2517 deprecated since over 2 years, it's time to inform our users that it is.
2518 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2519 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2520 use these (crappy) functions in your code.
2521 These functions will be removed soon. Stop using them now.
2522 * Deprecate MSG_get_host_{table,number}
2523 Implement MSG_hosts_as_dynar() instead.
2524 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2525 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2528 * Stabilize the parallel execution mode of user contexts
2529 * Introduce configuration variables to control parallel execution:
2530 - contexts/synchro: Synchronization mode to use when running
2531 contexts in parallel (either futex, posix or busy_wait)
2532 - contexts/parallel_threshold: Minimal number of user contexts
2533 that must be part of a scheduling round to switch to parallel
2534 execution mode (raw contexts only)
2535 * Fix bugs that prevented to use suspend/resume along with
2536 synchronization structures.
2537 * Fix bugs in process termination that lead to invalid memory access
2538 in very specific conditions.
2541 * Introduce a parallel mode for the models (controlled by surf/nthreads
2542 configuration item). In our tests, running the models in parallel
2543 never lead to any speedups because they are so fast that the gain
2544 of computing each model in parallel does not amortizes the
2545 synchronization costs, even when ultra fast futexes are used.
2546 This is released anyway because YMMV.
2549 * Performance boost by using a swag internally to compute the set of
2550 tasks that are finished and should constitute the return value of
2554 * Enable it by default now that it is considered rather stable.
2557 * Documentation of the tracing functions.
2558 * Performance gains when tracing categorized/uncategorized resource
2559 utilization by avoiding calls to get route when updating resource
2560 variables. LMM constraints are being used instead.
2561 * API changed to set task categories. Use MSG_task_set_category instead
2562 of TRACE_msg_set_task_category, and SD_task_set_category instead
2563 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2564 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2565 * Examples for link user variables added (see at examples/msg/tracing/)
2566 * Deprecated function TRACE_msg_set_process_category completely removed
2567 * Trace header updated according to the latest Paje file format
2568 * Tracing network lazy updates, no longer obligate users to use full updates
2569 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2570 * Experimental: let user code declare/set/push/pop application states for hosts
2571 * API changed to allow the manual creation of graph configuration files
2572 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2575 * Improve the API of Lua MSG bindings, using the Lua spirit.
2576 * Each simulated process now lives in its own Lua world (globals are
2577 automatically duplicated). It helps writing simulators. Will allow
2578 to run Splay programs within SimGrid in the future.
2579 * Add a Chord example in Lua, equivalent to the MSG one.
2582 * Start the implementation of a solution to express temporal
2583 properties, not only local assertions. This is still an
2584 experimental work in progress, stay clear from it to be safe.
2588 - Add new runtime parameters --help-logs and --help-log-categories
2589 to display information about supported logging parameters and
2591 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2592 don't exists anymore.
2593 * Mallocators: allow value NULL for the reset function.
2595 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2596 create homogeneous dictionaries, where all the elements share the
2597 same free function. Non homogeneous dictionaries will be
2598 deprecated in the next release.
2599 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2600 - Multi-level dictionaries are deprecated.
2602 - new function xbt_dynar_search_or_negative() that is useful when
2603 you have less than 2 million elements in your dynar and don't
2604 want of the extra complexity of catching exceptions when the
2605 element is not found.
2607 - Make xbt_os_thread module (for thread portability) public.
2608 Documentation is still to come, sorry.
2610 - Cleanups and simplifications to make it maintainable again.
2611 - Exotic features (such as memalign and valloc) were removed.
2612 - The metadata were extended and improved so that the
2613 model-checker becomes able to explore and inspect the heaps.
2614 - This may induce a performance drop when enable_model-checking is
2615 ON in cmake (even if it's not used in the simulation), but it is
2616 necessary at this point to get MC working.
2618 Turn model-checking OFF if simulation performance matters to you.
2619 Not enabling it at runtime is not enough, disable it in cmake.
2621 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2624 SimGrid (3.6.2) stable; urgency=low
2626 The "Not coding new stuff allows to polish old things" release.
2629 * New bindings to the NS3 packet level simulator (experimental)
2630 * Use the raw (efficient) execution contexts instead of the sysv
2631 (portable) ones when possible.
2632 * libpcre is now mandatory in any cases since not using it led to
2633 severe performance loss and possibly other issues
2634 * Update the XML platforms:
2635 - G5K: include the latest machine in Nancy
2636 - GridPP and LCG: new platforms
2637 * Documentation was partially updated, at least (more to come)
2639 Bug fixes, cosmetics and small improvements
2640 * Free terminated processes before the end of the simulation to avoid
2641 exhausting the memory of users having very dynamic amount of
2643 * Bug fix and cosmetics about canceling non-running tasks
2644 * Bug fix about the dot loader's issues when using libcgraph
2647 * Create an installer for windows with nsis (amd64 and win32)
2648 - Add an hello world project to illustrate simgrid project creation.
2649 - Embed libpcre into the Simgrid installer to avoid
2650 its compilation burden
2651 * The raw execution contexts should work on Apple now
2652 * Port to Windows 64 bits
2653 - Sysv contexts now have an implementation for this arch
2654 - GRAS communication features now support this arch
2655 * Drop support for borland compiler on windows
2656 - this code was not maintained, and we kinda depend on gcc nowadays
2657 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2658 * Fix portability issue on unstable ubuntu: linker became picky on
2661 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2663 SimGrid (3.6.1) stable; urgency=low
2665 The "Oops, we broke Macs too" release
2668 * Fixed contexts detection so that raw ones are used when possible
2669 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2670 leads to a strange error, with user code segfaulting sometimes when
2671 the generated code is not perfectly aligned (which is not
2672 controllable from the user side, depends on the amount of code)
2675 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2677 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2679 SimGrid-java (3.6) unstable; urgency=low
2682 * Split of every thing from simgrid v3.5 into a separate package.
2684 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2686 SimGrid (3.6) stable; urgency=medium
2688 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2691 * Bindings now constitute their own package, separated from the main one.
2692 Rationale: reduce our maintenance nightmare by reducing the module coupling
2693 They will soon be released on their own on gforge.
2695 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2696 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2698 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2699 * It's not quite deprecated for now because we have no replacement,
2700 but it may soon become the case.
2703 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2704 * Fortran: New user-level cache variable to store the rank of the running
2705 process. This improves performance by an order of magnitude.
2706 * C: New coccinelle script to automatically locate and modify global and
2707 local static variables.
2708 * Improved SMPI network model with a sender-side gap to account for multiple
2712 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2713 only say if a communication is finished, no matter whether it succeeded or
2714 failed. You can call MSG_comm_get_status() to know the status of a finished
2716 * New function MSG_task_dsend() to send a task and detach it. When a
2717 communication is detached, you are never notified of its success or failure
2718 and the memory is released automatically once it is finished. This function
2719 is useful when you don't care about the end nor the success of a
2721 * Change the prototypes of action replay. Sorry for inconvenience,
2722 but this is really more efficient this way (and to adapt your code,
2723 you just have to fix the initialization, that shouldn't be too long)
2724 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2725 ever managed to do anything useful with it.
2726 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2727 takes a dynar of communications. It returns immediately and gives the
2728 index of a finished communication (if any).
2729 * New example: a basic implementation of the Chord P2P algorithm.
2732 * New model for multi-core CPUs. You can now use the core attribute to
2733 precise the number of cores of a host. This is a basic model. Every
2734 process running on the host receives at most the power provided in
2735 the DTD (throughput<=power). Total throughput of process cannot exceed
2737 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2738 router linked by an up-link and a down-link (possibly asymmetrical).
2739 This kind of pattern allows to easily build last-mile model style platforms.
2740 Aggregating such patterns in a rule-based AS is thus the technique of
2741 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2742 * New model for Vivaldi routing. We transformed the Vivaldi network model
2743 into a Vivaldi routing model (based on the rule-based model). This allows to
2744 combine Vivaldi based latencies with last-mile platforms.
2747 * Added a check for NaN of IEEE754 infinite in the double entries of
2749 * Introduce a new context factory "raw", highly inspirated from the
2750 ucontext factory, but using manually crafted functions in assembly to
2751 do the work in an efficient manner.
2752 * Allow to change the used context factory at run time, not only at
2753 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2754 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2755 contexts at runtime (only with raw contexts or ucontexts).
2756 * Completely rewrote this module to allow parallel execution of user
2757 processes. Use --cfg=contexts/nthreads:N to execute user processes
2758 with N parallel threads (the default is 1, meaning no parallelism).
2759 * Allow to decide dynamically between sequential and parallel modes.
2760 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2761 processes in parallel only when their number is greater than or equal to P
2763 * Added a check for NaN of IEEE754 infinite in the double entries of
2767 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2768 won't output the state of processes when interrupted with Ctrl-C
2769 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2770 NULL-terminated array. This may solve backward compatibility issues
2771 due to the change to return type of SD_simulate. See also:
2772 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2773 * Add new macros with variable number of arguments.
2774 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2775 - in xbt/asserts.h: xbt_assert
2776 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2777 - in xbt/ex.h: THROWF and RETHROWF.
2778 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2780 * Change xbt_die() to accept a format string with arguments, just like printf.
2781 * New data structure: xbt_lib_t, like a dict but more general and with better
2785 * New configuration options
2786 Options triva/categorized and triva/uncategorized can be used to generate
2787 graph configuration files for Triva visualization tool.
2788 * Configuration option tracing/platform is renamed to tracing/categorized
2789 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2790 * New TRACE_user_link_variable interface:
2791 User provides the name of the link and the tracing variable to attach to it
2792 * the declaration of tracing categories must be done after the environment
2794 * simpler tracing interface, just one way to declare categories
2795 TRACE_category or TRACE_category_with_color, it is up to you
2796 * links in the trace file are again identified by their names
2797 * trace contains the full platform hierarchy exactly as declared using the ASes
2798 * Options tracing/msg/[task|process]:1 groups the process by hosts
2799 for both cases, tasks and processes must have names that are unique during the simulation
2800 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2801 * The experimental option tracing/msg/volume is deprecated
2802 its functionality may be reincorporated if needed
2804 The tracing generates a trace file with unordered timestamped events,
2805 because of the way the core simulator (surf) works. A script available
2806 at the tools directory (fix-paje-trace.sh) can be used to put the events
2807 in order. We have changed the tracing so it can generate ordered timestamped
2808 events in the final trace, but depending on the simulator (and how much time
2809 is simulated) that can lead to a huge memory utilization. It is deactivated
2810 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2812 Build Infrastructure
2813 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2814 - We are trying hard to keep the API stable, but it may happen that
2815 some things change (we're a research project after all, not a
2816 nuclear plant operating system). If such things should happen, you
2817 could rely on that macro to adapt.
2818 - current value: 30600 for 3.06.00, aka 3.6
2819 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2820 a number that can be compared with SIMGRID_VERSION.
2821 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2822 assertions and verbose/debug logging events are disabled at compile time.
2824 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2826 SimGrid (3.5) stable; urgency=medium
2829 * New feature to any SimGrid-based simulator: Model-Checking
2830 Check SIN#1 for more details.
2833 * New Model SMPI (three-interval linear regression for correction factors)
2834 See RR-7426, available at http://hal.inria.fr/inria-00527150
2835 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2836 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2837 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2838 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2839 * See also src/smpi/README
2844 - Tracing API changes: TRACE_start and TRACE_end should not be called
2845 by user-code. They are automatically called by simulators created
2846 with SimDAG, MSG and SMPI if the toolkit is compiled with
2847 tracing_enabled=ON. Categories declaration and utilization remain the
2848 same for MSG and SimDag.
2849 - A function was added to the tracing API to declare categories with
2851 - TRACE_category_with_color (char *category, char *color)
2852 where color must be in the following format
2853 "%f %f %f", red, green, blue
2854 and red, green, blue are float values in the interval [0, 1]
2855 - User can specify NULL as color parameter, or continue calling
2856 TRACE_category (cat)
2857 On that case, the tracing system will define random colors
2858 - The following command-line options are supported:
2859 --cfg=tracing/filename:msg.trace
2860 --cfg=tracing:1 (activate tracing, needed to use others)
2861 --cfg=tracing/platform:1 (categorized resource use)
2862 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2863 --cfg=tracing/msg/task:1 (task creation)
2864 --cfg=tracing/msg/process:1 (process creation, migration)
2865 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2866 --cfg=tracing/smpi:1 (SMPI interface tracing)
2867 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2868 - examples of examples/msg/tracing updated
2870 - DAXLoader and DOTLoader functions can generate tasks with categories
2871 - A new function to attribute a category to SD tasks:
2872 TRACE_sd_set_task_category (SD_task_t task, char *category)
2873 * Tracing the MPI interface implemented by SMPI
2874 - Collective operations are traced with states
2875 - Point-to-Point operations are traced with states/links
2876 - Tracing activated by a parameter "-trace filename" passed
2877 to smpirun during execution (considering that simgrid
2878 is compiled with tracing enabled)
2879 - To run the simulation with gdb, the simulator
2880 accepts --cfg=tracing/smpi:1 to trace SMPI
2881 - tesh files to check if smpi tracing is ok
2882 - See examples/smpi/NAS/DT-trace
2883 * GTNetS tracing re-worked
2884 - adaptation to the tracing system of GTNets to cope
2885 with modifications regarding the fullduplex mode
2886 - new tesh files to check if gtnets tracing is ok
2889 * Asynchronous communications through the functions:
2890 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2891 * New function: MSG_load_platform_script()
2892 to make possible using a lua script instead of XML files to set up platforms
2893 * New function: MSG_set_function
2894 to associate functions to processes, used when bypassing the parser
2895 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2897 Platforms: Add some more examples in examples/platforms
2898 * Grid'5000: see www.grid5000.fr
2899 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2902 * Change the XML format. This is a very important modification. SimGrid
2903 3.5 introduces a new hierarchical format based on the notion of
2904 Autonomous Systems. Compatibility with old format is ensured
2905 through the perl script provided in the install bin directory
2906 bin/simgrid_update_xml.
2907 It is now possible to build platforms with specific routing
2908 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2909 connect several platforms together. We will try to provide soon
2910 set of realistic platforms exploiting these properties (have a
2911 look at examples/platforms/ for the moment).
2912 * Take the opportunity of the XML format change to be a good XML citizen:
2913 rename link:ctn to link_ctn and similar changes (also dealt with by
2915 * Add a new routing scheme (rule-based) using regular expressions. It
2916 enables to have an extremely low memory footprint when the
2917 underlying routing is simple and can be compactly described. You
2918 need to have libpcre4-dev (perl regular expressions) installed if
2919 you want to use this routing scheme.
2920 * Revive the cluster TAG and allow to easily and efficiently (both in
2921 term of memory and speed) connect clusters together. Have a look
2922 at teshsuite/simdag/platforms/ to see how this can be done. With
2923 this tag, you can create clusters with thousands of tasks at no
2924 cost (have a look at examples/platforms/).
2925 Note: clusters are implemented as ASes, so there is no need for an
2926 enclosing AS tag if you have only one cluster in your platform.
2927 * Add new generic functions in the public interface that allows the user
2928 to call SURF 'create_resource' methods from your code (same
2929 functionality as the XML bypass mechanism but with a much lighter
2931 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2932 that uses a piecewise linear approximation to produce better
2933 results when exchanging small messages.
2934 * Add a new parameter to handle correctly full duplex link and account
2935 for interferences between uplink and downlink communications
2936 (activate with --cfg=fullduplex:1).
2939 * Rename the SD_READY (all dependencies are satisfied and task is
2940 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2941 dependencies are satisfied) state.
2942 This prevents a confusion between the notion of "ready to schedule"
2943 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2944 simulated" (SD_RUNNABLE) used by the simulation kernel.
2945 * Change the way a task is considered as ready. Instead of removing
2946 dependencies when a task is done, a counter is decreased. This way,
2947 it is always possible to reach ancestors thanks to the
2948 SD_taks_get_parents function (even after the end of the simulation.)
2949 * Change the return type of SD_Simulate from (SD_task_t*) into
2950 xbt_dynar_t. This function was in handling a dynar internally and
2951 converted it into a NULL terminated array for historical reasons.
2952 * New function SD_dotload(char*) to load a DAG described in dot
2953 format. This loader and the corresponding examples require the
2954 installation of the graphviz library.
2955 * Fix a bug in the management of tasks of size 0 in the surf network
2956 models. This problem was only visible with SIMDAG and you should
2957 thus disregard results produced with earlier versions if you
2958 relied on this feature (some tasks were blocked because of this).
2959 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2960 even though your applications were doing only point-to-point
2961 communications and sequential computations. Now you can really use any
2962 model you want (of course, if you create real parallel tasks, which are
2963 not implemented in most models beside ptaskL07, this will abort).
2964 * Add an example that schedules a DAX on an heterogeneous platform
2965 using a Min-Min strategy.
2966 * New function SD_workstation_get_current_task() that returns the kind
2967 of task currently running on a workstation in the sequential access
2969 * Raise some warnings when unexecuted tasks remains at the end of the
2970 simulation. This is usually caused by cycles in the DAG.
2973 * New function: SIMIX_process_set_function() called by MSG_set_function
2974 * Change the underlying waiting queue in semaphores so that a process
2975 can wait on several of them simultaneously (as in waitany).
2976 * Fix the way to handle tokens in semaphores so that all access patterns
2977 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2978 * kill the dirty pimple SIMIX_message_sizes_output()
2979 Please use (proper) visualization instead
2982 * New data container: setset (set of sets of elements)
2983 * New module: mmalloc (mapped malloc, allowing to have several
2984 independent segments of malloc)
2985 * New function: xbt_dict_cursor_set_data()
2986 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2987 * New function: xbt_dynar_is_empty()
2988 * New function: xbt_fifo_get_last_item()
2989 * Fix xbt_dynar_shrink(): use the right element size.
2990 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2991 uninitialized areas during expand.
2992 * Fix semaphores: previous implementation was severely broken.
2993 * Use library init/fini functions for our initialization.
2994 - you can use logs and other feature as soon as you want in your
2995 code (even before the xbt_init / MSG_init)
2996 - xbt_exit is now a no-op and produce a warning when used.
2999 * Port GRAS to new SIMIX mechanisms. This allows gras users to
3000 benefit from the latest improvement to the simulation kernel.
3001 * Kill measurement sockets for now. If you rely on them, sorry. This
3002 release is not for you. This feature will be reintroduced in the
3003 future, but we cannot delay the release any further.
3004 * New function: gras_msgtype_get_name().
3005 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
3006 * Fix (at last) the pmm example: it should not randomly fail anymore.
3008 Build chain: bug fixes and overall polishing
3009 * Cmake is now stable enough. Hence, we killed the autotools.
3010 * Port to windows ( TM :)
3011 * Fix the 'make install' target.
3012 No need to use 'make install-simgrid' anymore
3013 * Introduce a 'make dist' target compiling a *source* archive
3014 'make package' compiles a binary archive
3015 * Compile java files only on need
3016 * Add --cd and --setenv command line options to tesh
3017 * Out of source builds are not fully supported yet, but we are close
3018 * Enable supernovae and optimization flags by default for our users
3021 * Add layer to set up environment directly from lua, without XML.
3022 * The effect of gras_stub_generator can be achieved through
3023 lua too (check examples/gras/console/ping_generator.lua)
3025 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
3027 SimGrid (3.4.1) stable; urgency=low
3029 The "Polishing easter eggs is probably a good idea" release.
3030 This is a bug fixes release only.
3034 * Fix a bug preventing the tasks from begin garbage collected.
3037 * Fix a bug occurring when a host involved in a communication fails.
3038 This was not detected properly by the other peer involved in the
3039 communication. Now, it's reported as a network error.
3042 * Warn the user about loop dependencies in data flow of DAX files
3043 * Obey the control-flow dependencies of DAX files
3046 * Add option "enable_smpi" allowing to not compile SMPI.
3047 Probably useful for the (Mac) users experiencing a build error here
3048 * Improve the detection of lua5.1 and ruby1.8
3050 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
3052 SimGrid (3.4) stable; urgency=low
3054 The "Easter in Cargese" release. Also known as (major changes):
3056 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
3057 ~> bindings were greatly improved
3058 ~> new tracing infrastructure for better visualization introduced
3060 * the "Welcome to configury modernity" release.
3061 ~> we switched from autotools to cmake, and improved our cdash
3064 A more detailled list of changes follow (full detail in svn log).
3066 Java Bindings: Various Cleanups
3067 * (install java-gcj-compat-dev on debian-like to use them)
3068 * Remove put/get: no need to export deprecated interface in Java
3069 Use send/receive instead.
3070 * Cleanup the examples and add a README per directory
3071 * Remove example autoDestination (that's the only way to go now)
3072 * Remove example explicitDestination (was a plain copy of basic)
3073 * Make JniException a runtime exception, so that there is no need to
3074 declare the fact that you may encounter such a beast. I guess that
3075 nobody will ever want to survive such error.
3076 * Create specific errors for each MSG case of failure:
3077 host failure, transfer failure, timeout, task cancelled
3078 * Cleanup the exceptions that may get thrown by each function
3079 * Other internal cleanups in Java bindings. Performance still bad :/
3080 Ruby and Lua Bindings: create them
3081 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
3082 * That's new and great, you should try them out.
3083 Same functionalities than Java bindings, only even less polished
3085 * Kill the useless "rate" argument of SD_task_get_execution_time()
3086 Everyone used to provide -1 as a value, it was not used, and the
3087 semantic of a possible use wasn't even clear.
3088 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
3089 either as comm costs or compute costs to mean that there is no
3090 such thing for that specific task.
3091 * Add a SD_task_set_name() function
3092 * Fix SD_task_unschedule() on typed tasks
3093 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
3094 * In DAX loader, accept useless 'level' attributes to <job> since
3095 LIGO DAGs have them (seem to be to ease graphical representation).
3097 * Add an example masterslave_mailbox.c using send/receive and not
3098 the deprecated put/get interface.
3099 * Kill the MSG_paje_output() function. It's a noop since 2 years.
3100 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
3102 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
3103 (declare MSG_USE_DEPRECATED to still have the old name)
3104 * Add a MSG_task_set_data() function
3105 * About trace replay (see examples/msg/actions):
3107 - Allow to work with splitted trace files for each process
3108 Give the specific trace file as argument of each process,
3109 and call MSG_action_trace_run(NULL)
3110 You can still have one merged file for all processes.
3111 - Fix implementation of collective operations
3112 * Allow task_execute() on 0-sized tasks (closes #10063)
3114 * This is the first release of SimGrid where SMPI is not considered
3115 beta anymore (even if some corners should still be improved)
3116 * Port over the new SIMIX_network submodule (internal refactoring)
3117 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
3118 * Implement more missing elements of the standard:
3120 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
3122 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
3123 MPI_Scan() MPI_Get_processor_name()
3124 - Added implementation of missing case for Alltoall (warning: it's
3125 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
3126 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
3127 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
3128 SMPI_MPI_Allgatherv()
3129 * Bug fixes include:
3130 - MPI_Waitsome() was broken
3131 - Allow relative includes in smpicc
3132 - Command line cfg argument 'reference_speed' was ignored...
3133 - Some functions did not properly lead to auto-benching of user code
3134 - smpicc passes -O2 by default (just like openmpi one)
3136 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
3137 * Bug fixes about timeouts during communications
3138 * add SIMIX_message_sizes_output() as a pimple to write to file the
3139 amount of messages per size. Use gnuplot to get histogram.
3140 Pimple because that's the only user-visible function of simix,
3141 defined directly in xbt.h (irk, sorry)
3143 - Add a SIMIX_sem_get_capacity() function
3144 - Fix interactions with processe resume/suspende
3145 - release_forever() was stupidly broken
3146 - Fix SIMIX_display_process_status() for processes in a semaphore
3147 - Make SIMIX_sem_block_onto() user-visible
3148 * Refactoring context stuff:
3149 - Use pseudo-OOP for better modularity
3150 - reimplement SIMIX_process_kill() without process_schedule() so
3151 that the latter can take as invariant that it is called from
3153 - Merge context_start into context_new for sake of simplicity
3155 * Add a Vivaldi network model, coded live during SUD'10 ;)
3156 * Rename configuration variables to start a hierarchy:
3157 o cpu_model -> cpu/model
3158 o network_model -> network/model
3159 o workstation_model -> workstation/model
3160 * New configuration variables:
3161 o network/bandwidth_factor: correction to bandwidth
3162 o network/latency_factor: correction to latency
3163 o netwotk/weight_S: correction to the weight of competing streams
3164 * Add a long description to the models, that users can see with such
3165 argument on the command line: --cfg=cpu/model:help
3166 * --help-models display the long description of all known models
3168 * config: add the ability to set a default value after registration
3169 Does not override any previously set value (e.g. from cmd line)
3170 * dict: allow to have integer key and data.
3171 When so, you need to use the following functions
3172 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
3173 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
3174 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
3175 In contrary to regular dicts, the key is not malloced before copy.
3176 Mixing scalar and regular elements in the same dict is not tested
3178 * Allow to use xbt_dynar_shrink() to expend the dynar instead
3179 Tracing for Visualization:
3180 * SimGrid is now instrumented in order to generate a trace file for
3181 visualization analysis: to use it, need to compile SimGrid with the
3182 "tracing" option enabled, and instrument the program using SimGrid with
3183 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
3184 (among other functions).
3185 * The instrumentation only traces the platform utilization for now
3186 * Documentation to use the tracing functions and how to analyze the
3187 traces with the Triva tool is written.
3188 * More information about: SimGrid FAQ (in the section Tracing Simulations
3191 * We moved to cmake as default build system. Autotools support will
3192 be dropped soon. Check the FAQ for more info about how to use it.
3193 * Greatly improved our cdash/ctest interactions
3194 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
3195 * Added memory checking tests with valgrind; lot of memleak fixing.
3196 This may be the first release of simgrid with so few memory issues
3197 * Added code coverage tests.
3198 Our coverage is still improvable, but at least we see it on cdash.
3200 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
3202 SimGrid (3.3.4) stable; urgency=low
3204 The "Desktop Grid needs love too" release (also called Xmas release).
3206 Models improvements:
3207 * Major speedup in the maxmin system solving by using lazy evaluation
3208 Instead of solving completely the maxmin system at each iteration,
3209 only invalidate (and recompute) the modified parts.
3210 This new feature is enabled in default models but you can try to
3211 turn it on with "--cfg:maxmin-selective-update=1" for other models.
3212 * Cas01 IMproved as default CPU model
3213 This CPU model is the same Cas01 model, but it uses the
3214 maxmin-selective-update flag and a heap structure to manage
3215 actions on SURF kernel.
3216 It reduces the complexity to find the next action to finish and,
3217 consequently, it's faster than the old Cas01.
3218 This is the new default CPU model (Cas01).
3219 * Rename the old Cas01 model to Cas01_fullupdate
3220 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
3221 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
3222 * CpuTI (CPU Trace Integration)
3223 A new CPU model whose objective is simulate faster when using
3224 availability trace files.
3225 Instead of using a full featured, over engineered maxmin system for
3226 CPU modeling, this model does the pre-integration of traces files
3227 to calculate the amount of CPU power available, and so, executes
3228 faster than the old CPU models.
3229 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
3230 * Use LV08 as default network model since it gives better accuracy
3231 for small messages and shouldn't change things for big ones.
3232 Use --cfg=network_model:CM02 to get the previous behavior.
3235 ******************************************
3236 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
3237 ******************************************
3238 * The new CPU model may changes simulations!
3239 The point is that events occurring at the exact same timestamp
3240 are not scheduled in the same order with the old and new
3241 version. This may be enough to completely change the execution
3242 of simulations in some cases.
3243 * The new network model will change simulations!
3244 This new model is more realistic than the previous one, so you
3245 should consider redoing your old experiments with this model.
3246 Sorry for the inconvenience.
3249 * Introduce the supernovae compilation mode
3250 When compiled that way, the whole SimGrid (or almost) is put in a
3251 single compilation unit and compiled in one shoot.
3252 This is to help gcc which has difficulties to inline stuff from one
3254 The speedup seem to be above 15%, althrough more tests are needed on
3255 amd64 to confirm that gain.
3258 * Port of MSG's mailbox on top of SIMIX network
3259 The put/get mechanism was greatly simplified on the way.
3262 * New SIMIX network module. Provides:
3263 - Mailbox: rendez-vous mechanism to find with who you want to speak
3264 - Synchronous send/recv: easier and hopefully faster since the
3265 logic is handled in the maestro process directly now
3266 - Asynchronous send/recv: you dreamt of it? It's here now
3267 Too bad that nobody cared enough to propagate the change to MSG.
3268 * Add semaphores as SIMIX synchronization mechanism.
3271 * new function SD_daxload(char*) to load a DAX file
3272 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3273 * Introduce typed tasks. Specify its kind and cost at creation.
3274 At scheduling, just give where it should be placed, and the cost
3275 for each involved resource is automatically computed.
3276 Existing constructors so far (more to come of course):
3277 - SD_task_create_comm_e2e() for end-to-end communication
3278 - SD_task_create_comp_seq() for sequential computation
3279 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3280 * new function SD_task_dump() for debugging display
3281 * new function SD_task_dotty(task,FILE*) writing to file the info
3282 about the task in dotty format
3283 * SD_task_dependency_exists() can now cope with having one of its
3284 arguments NULL. If so, it tests whether the other argument has any
3286 * Add getters on list of preceding/following tasks:
3287 SD_task_get_parents(task) and SD_task_get_children(task)
3288 * Add getters on amount of workstations and list:
3289 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3290 * Add getter on task kind: SD_task_get_kind(task)
3291 * Update the start_time and finish_time of tasks on completion/failure
3292 * Bugfix: Remove task from state swags when destroyed
3295 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3298 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3299 dynar with one separate thread per value of the dynar.
3300 * Change the prototype of xbt_thread_create(), sorry.
3301 Added a boolean parameter indicating whether we want to join this
3302 thread (used in SG only for now)
3303 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3306 * GTNetS wrappers should now be usable again (and betterly tested too)
3307 * Fix a major regression from 3.2 where the timeout provided to
3308 MSG_task_put_with_timeout() was used as absolute time before which
3309 the comm should be done.
3310 * Start to fix the <cluster> tag.
3311 - Internal links should be good now (beside of the loopback, which
3312 use the private link instead)
3313 - paths to the external world is still rather broken
3314 - the <route:multi> tag is just broken. Actually that's brain-dead.
3315 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3316 to make it less stupid
3317 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3318 * Fix a source-level compatibility glitch from 3.2: after defining
3319 MSG_USE_DEPRECATED, you can use the old name
3320 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3321 * Allow to compile from the SVN with automake 1.11
3322 * Fix some problems when using the "start_time" tag in deployment XMLs.
3323 * Fix #8569: XBT/synchro.h has redundant declarations
3324 * Fix #8563: MSG return values and exceptions
3325 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3326 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3327 Thanks to Robson Peixoto.
3328 * Fix a vicious bug in dictionaries inducing that some elements were
3329 not freed on xbt_dict_free()
3331 Portability report of this version:
3332 * Main portability targets:
3333 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3334 on (amd64/i386/ia64)
3335 - mac leopard on i386
3336 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3337 but nothing critical.
3338 * Other platforms: windows, AIX and others were not tested for this release
3340 Timing report of this version:
3341 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3342 scenario parameters). From 8h to a few seconds in desktop grid settings.
3343 * Supernovae brings about 25% speedup on i386.
3345 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3347 SimGrid (3.3.3) stable; urgency=low
3349 The "Need for Speed" release.
3351 The timings done to validate the 3.3.2 were faulty.
3352 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3354 The problem was a conversion from a manually handled vector to
3355 xbt_dynar_t on the critical path.
3356 xbt_dynar_foreach calls functions, inducing stack management crap.
3358 We inlined these functions and xbt_dynar_foreach is now breath taking.
3359 We also inlined xbt_swag_belong on the way.
3361 Here are some approximate speedup measurements (on master/slaves
3362 simulations lasting between 10s and 20s each):
3363 3.3.1 -> 3.3.2: about same performance
3364 3.3.2 -> 3.3.3: 40% speedup
3365 3.3.1 -> 3.3.3: 40% speedup
3366 3.3.1 with inline patch -> 3.3.3: 30% speedup
3368 Our reading is that the refactoring which occurred in 3.3.2 made us
3369 suffer much more from the xbt_dynar_foreach low performance, but
3370 once we solved this, this refactoring proved to be very performance
3371 effective. From the 40% speedup, somehow, 10% are due to the
3372 inlining and 30% to the refactoring.
3374 That's a pitty that gcc cannot inline functions placed in other files
3375 alone. We have to choose between:
3376 - break the encapsulation (by putting private data structures and
3377 accessors in headers files to help gcc)
3378 - live with low performance
3379 - switch to a decent compiler such as icc (not quite possible).
3381 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3383 SimGrid (3.3.2) stable; urgency=low
3385 The "Simplicity does not preceed complexity, but follows it" release.
3387 The main contributors of this release were (lexical order):
3388 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3391 * Extract the routing logic into its own object.
3392 (was dupplicated in network.c and workstation_LV07.c;
3393 Allows to implement other ways of storing that info)
3394 => kill now useless network_card concept
3395 - Use dynar to represent routes (instead of void** + int*)
3396 - kill link_set (use surf_network_model->resource_set instead)
3397 - Add a command-line option to choose the routing schema to use
3398 - Add three new models:
3399 * Floyd (shortest path computed at initialization)
3400 * Dijikstra (shortest path recomputed all the time)
3401 * Cached Dijikstra (shortest path computed on need)
3402 All these models where contributed by Silas De Munck, and are
3403 described in his ICCS09 paper.
3405 * Simplify model declaration
3406 (less redirections, less function to write when defining a model)
3407 - Factorize stuff between models:
3410 surf_model_resource_set(model)
3411 surf_model_resource_by_name(model, name)
3412 - Unify the types of models in s_surf_model_t (using an union)
3413 - Embeed fields of common_public directly into s_surf_model_t
3414 - Rename model methods:
3415 action_free ~> action_unref
3416 action_change_state ~> action_state_set
3417 action_get_state ~> action_state_get
3418 - Change model methods into functions:
3419 (model)->common_public->action_use ~> surf_action_ref
3421 * Implement a generic resource; use it as ancestor to specific ones
3422 (allows to kill duplicated code in models)
3423 Drawback: timer command don't need no name nor properties;
3424 workstation_CLM03 don't need no properties
3425 (but I guess we can live with those few bytes wasted)
3427 * Improve the action object model
3428 - implement a constructor avoiding dupplicated code about field
3429 initialization in generic_action part.
3431 * Kill the SDP model: it has an external dependency, is deprecated
3432 in flavor of modern lmm models, and didn't compile since a while
3435 * Relocation of the context module from XBT to SIMIX.
3436 (the context were decoupled from the simix processes, duplicating a lot of code)
3437 => a lot of code was factorized
3438 - less overhead is introduced during scheduling
3439 - simpler API for the context factory
3440 - the logic for process creation,destruction and manipulation was simplified
3441 * Simplification of the s_smx_process_t data structure.
3442 => accesing the simix level data associated to a process is faster now,
3443 and the code is a lot more readable.
3446 * Implement some more MPI primitives:
3447 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3448 -implementation: Bcast: flat or 2-ary tree (default),
3449 Barrier: 4-ary tree,
3451 Allreduce: Reduce then Bcast
3452 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3453 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3454 Alltoallv: flat tree, like ompi
3456 * Add support for optimized collectives (Bcast is now binomial by default)
3457 * Port smpirun and smpicc to OS X
3460 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3461 Shout out if you used it.
3464 * Display the list of still queued messages in SG mode when existing
3468 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3469 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3470 * Add xbt_str_from_file(FILE*)
3471 * Add xbt_dict_get_key achieving a linear reverse search
3472 * Remove the context module
3474 Portability report of this version:
3475 * Main portability targets:
3476 - Linux(debian)/x86/context
3477 - Linux(debian)/x86/pthreads
3478 - Linux(debian)/amd64/context
3479 - Linux(debian)/amd64/pthreads
3480 On these, we still have the eratic breakages of gras/pmm and
3481 amok/saturate_sg reported in previous version. We still think
3482 that the tests are the cause of the fault, not the tested code.
3484 - Mac OSX Leopard/x86/context
3485 Still false negative in tesh autotesting.
3486 Smpi still fails, but this time because readlink does not accept -f
3487 Everything seems to work properly beside of that.
3490 - AIX version 5.3 (only tested contexts this time)
3491 Smpi still fails there because mktemp is not installed.
3492 Everything seems to work properly beside of that.
3494 I managed to compile it for the first time, but several breakages.
3495 Won't delay the release for this exotic platform.
3497 * Windows: it's still lagging behind. If you want to help, please
3500 Timing report of this version:
3501 This version seem to be more than 5% faster than 3.3.1 (on linux
3502 64bits with contextes). The gain is less than expected, we are
3503 investigating this for next release.
3505 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3507 SimGrid (3.3.1) stable; urgency=low
3510 * Implement a --cfg-help to show existing configuration variables
3511 * Build chain do not require doxygen in maintainer mode
3514 * fix a bug on struct sizeof computation, which prevented the
3515 exchange of arrays of structs in some conditions
3516 - added a regression test about this in datadesc_usage
3517 * Allow the exchange of 0-long dynamic vectors.
3518 - for that, use -1 as indicator of dynamic size instead of 0
3519 - This implied to change any size from unsigned long to long,
3520 reducing a bit communication abilities, but I guess that with
3521 64bits being quite common, this is more than enough.
3522 - This also induce a protocol change, thus bumping network protocol
3523 version from 0 to 1 (if we have external users, we have to get
3524 clean on that point too ;)
3525 - added two regression tests about this in datadesc_usage
3526 * Be more verbose when propagating local exceptions
3527 This helps debugging.
3528 * Display the status of simulated processes when receiving SIGINT in
3532 * Allow to control the simulation from a trace file.
3533 New functions MSG_action_register() and MSG_action_trace_run()
3534 The first one allows to associate a function execution to each
3535 kind of action while the second one parses a trace file and
3536 triggers the corresponding actions within the system.
3537 For now, only a toy example is provided in examples/msg/actions
3538 * Add an example of process migration in examples/msg/migration
3539 * Fix a bug in task exchange which broke MSG_task_get_sender()
3540 Add a teshsuite regression test for that.
3541 [Bug: if MSG_task_get_sender() is called after sender exit,
3543 * Fix a bug which prevented suspend/resume to work properly
3544 * Display the status of simulated processes when receiving SIGINT
3545 This fixes a regression of v3.3. due to the introduction of SIMIX
3546 * Bug fixing in failure management:
3547 - trace could not start by a failure at time 0
3548 - failure during communications were not working
3551 * Add SIMIX_process_set_name() to change the name of the current
3552 process in the log messages.
3553 * Store smx_hosts in a dict since we only retrieve them by name
3554 * Move the configuration infrastructure to surf
3557 * Move the configuration infrastructure to surf
3560 * Massive internal cleanups:
3561 - Store internal structures on processes instead of hosts (allows
3562 to have more than one process per host, in addition of being more
3564 - Cleanup the initialization/finalization process
3565 - Kill a whole bunch of unneeded synchronization:
3566 processes run in exclusive manner within the simulator
3567 - Move queues from global tables to process data fields
3569 - now accept -platform and -hostfile arguments
3570 - Pass the right rank value to processes according to the hostfile
3571 * Compile the examples by default, and use them as regression tests
3572 * Implement MPI_Wtime()
3573 * Change the reference speed to a command line option
3576 * TCP_gamma can now be specified as command line option using
3577 --cfg=TCP_gamma:10000000.0
3578 * Change the --surf-path cmd line option into --cfg=path:
3581 * Also include strbuff from xbt.h public header
3582 * xbt_ex_display(): do not free the exception after displaying
3583 This allows to do more with the given exception afterward.
3584 Users should call xbt_ex_free() themselves.
3588 Portability report of this version:
3589 * Main portability targets:
3590 - Linux(debian)/x86/context
3591 - Linux(debian)/x86/pthreads
3592 - Linux(debian)/amd64/context
3593 - Linux(debian)/amd64/pthreads
3594 These targets fail about 1/10 of times on gras/pmm, but we believe
3595 that this is because of the test, not because of simgrid.
3596 amok/saturate_sg fails even more rarely, and the test may not be
3599 - Mac OSX Leopard/x86/context
3600 The test suite still spits tons of errors because some obscure
3601 force prevents us from removing the temporary directories
3602 arguing that they still contain some metadata I've never heard of.
3603 Smpi fails because seq is not installed.
3604 Everything seems to work properly beside of that.
3607 - AIX version 5.3 (both contexts and pthread)
3608 Smpi still fails there because mktemp is not installed.
3609 XML inclusions seems rosty on AIX.
3611 * Windows: it's still lagging behind. If you want to help, please
3614 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3616 SimGrid (3.3) stable; urgency=high
3620 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3621 [Malek Cherier & Mt]
3623 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3626 Until now, GRAS were using MSG as an interface to SURF. It was
3627 quite difficult because both interface have several differences
3628 (MSG channels vs GRAS sockets were the most notable point).
3630 This also opens the gate to SMPI (which should occur soon) and speed
3631 up simulations by to 40% (even if it were not the main goal).
3633 **************************************
3634 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3635 **************************************
3636 The point is that events occurring at the exact same timestamp are
3637 not scheduled in the same order with the old and new version. This
3638 may be enough to completely change the execution of simulations in
3639 some cases. Sorry for the inconvenience.
3641 * Cleanup and upgrade the XML format to push further scalability
3642 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3644 * Improve the testing infrastructure with tesh. Now a very large part of
3645 the code is tested not only by being run but also by checking that the
3646 output match an expected output [Mt].
3648 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3649 is really less memory-demanding, which should allow you to use
3650 larger files in SimGrid [AL].
3652 * Inform valgrind about our contextes, so that it becomes usable
3653 with the default (and more effecient) version of SimGrid
3654 [contributed by Sékou Diakite, many thanks]
3657 * Introduce a listener thread in charge of receiving incoming
3658 messages from the network. It allows to overlap communication and
3659 computation but most notably, it removes some stupid deadlocks due
3660 to the fact that so far, a process could not send and receive at
3661 the same time. This made most non trivial communication schema
3663 * Convert the PIDs from long int to int to match the MSG ones (and
3664 linux ones too) [Mt]
3665 * New function: gras_agent_spawn() to launch a new process on
3666 current host. Only working in simulation for now. [Mt]
3667 * New function: gras_os_hostport() returning a constant form (ie,
3668 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3671 * Make the backtrace of exceptions more human readable [Mt]
3672 * New module: xbt/str [Mt]
3673 a ton of string utility functions (split, join, printf to a newly
3674 allocated buffer, trim, etc)
3675 * New module: xbt/hash [Mt]
3676 SHA1 hashing algorithm (more to come if needed)
3677 * New module: xbt/synchro [Mt]
3678 synchronization tools (mutex and conditions) working the same way
3679 in simulation and in real life (mainly useful for GRAS, but not
3681 * New module: xbt/queue [Mt]
3682 classical producer/consumer synchronization scheme
3683 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3684 (using the classical functions will get serialized) [Mt]
3685 * Make dictionary internal table dynamic. No need to specify its size
3686 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3688 * Make sure the log channels are organized as a tree under windows
3689 (because of ANSI C compatibility issue, any channel were child of
3690 root directly) [Mt].
3693 * Cleaned many thing in surf and fixed a few bugs [AL].
3694 * Add a nice command line configuration mechanism to compose models [AL].
3695 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3696 the previous one (KCCFLN05). It relies on something that looks like
3697 a max-min sharing mechanism but cannot be written as such. A new solver
3698 was thus designed [AL].
3699 * Add a new solver to lmm. Based on Lagrange optimization and
3700 gradient-based descent, it enables to efficiently maximise systems s.a
3702 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3704 This solver enables to propose two new network models for TCP Reno and
3705 TCP Vegas based on Low's work. These models still need to be fully
3706 tested though [Pedro Velho].
3709 * Bug fix in SD_simulate. Now the time bound given as argument is
3711 * Use the new parallel task model (ptask_L07) as default.
3712 * Use the SURF command line configuration mechanism.
3713 * 0-size tasks (for synchronization) should now work.
3715 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3717 SimGrid (3.2) stable; urgency=high
3721 We still experience issues on this platform, but we believe that at
3722 least MSG is usable.
3724 GRAS API BREAKAGE (for simplification purpose, sorry):
3725 * the gras_msgtype_by_name is not used anymore. Instead of
3726 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3727 you can write (and must)
3728 gras_msg_send(toserver, "request", &request);
3729 - If you still want to pass a gras_msgtype_t to the function (to cache
3730 the type and avoid the lookup time), use the gras_msg_send_() variant.
3731 - Impacted functions:
3732 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3733 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3734 * The callbacks are now expected to return 0 when everything went well
3735 (just like the main() function)
3737 GRAS new features and improvements:
3738 * New module mechanism where user code can use per process globals [Mt]
3739 This is similar to gras_userdata_*() functions, but for libraries. It
3740 factorize some code developped over and over in the examples and AMOK.
3741 It has still to be documented and used (only amok/peermanagement is
3743 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3744 loss when they were small enough to fit into the buffer and sent quickly
3745 enough so that they can all get received in one shoot.
3746 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3747 if not found. Use the *_or_null() variant for the old semantic.
3748 * In gras_msg_handle, do not discard messages without callback.
3749 They are probably messages to be explicitly awaited later (ie, proofs of
3750 mis-synchronization in userland since they are sent before being awaited)
3751 No big deal usually.
3752 * gras_socket_meas_send/recv: semantic changed!
3753 The numerical arguments used to be (1) the total amount of data to send
3754 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3755 messages. This was need for the fool willing to send more than MAXINT
3756 bytes on quite fat pipes.
3759 * Do really rename the hostmanagement module to peermanagement. [Mt]
3760 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3761 but this is rather new and this was documented in the module
3762 documentation (poor excuses, I admit)
3763 * Bandwidth measurement semantic changed! This follows the changes to
3764 gras_socket_meas_send/recv explained above.
3767 * A sequential mode has been added to the workstations. When a workstation
3768 is in sequential mode, it can execute only one task, and the other tasks
3769 are waiting in a FIFO. [Christophe Thiery]
3772 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3773 model for SIMDAG. [Christophe Thiery]
3774 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3775 equal to 0, which caused some bad behaviors in
3776 saturated_constraint_set_update. I now use a threshold mechanism like in
3780 * When running manually src/testall, you select specific units [Mt]
3781 testall is the result of our cunit mechanism, and should replace all
3782 the scripty thingy around since bash don't run easily on billware.
3784 * A mallocator system has been added. [Christophe Thiery]
3785 Mallocators allow you to recycle your unused objects instead of freeing them
3786 and allocating new ones.
3788 Documentation update:
3789 * FAQ reworking + New FAQs:
3790 - "Valgrind spits tons of errors!" [Mt]
3791 - "How to repport bugs" [Mt]
3792 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3793 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3794 - Communication time measurement within MSG [AL]
3795 - I experience weird communication times when I change the latency [AL]
3796 * GRAS tutorial [Mt]
3798 - an introduction to the framework and to the used communication model
3799 - an initiatic tour introducing the most proheminent features:
3801 . Lesson 0: Installing GRAS
3802 . Lesson 1: Setting up your own project
3803 o Part 2: Message passing
3804 . Lesson 2: Exchanging simple messages
3805 . Lesson 3: Passing arguments to the processes (in SG)
3806 . Lesson 4: Attaching callbacks to messages
3807 . Lesson 5: Using globals in processes
3808 . Lesson 6: Logging information properly
3809 . Lesson 7: Using internal timers
3810 . Lesson 8: Handling errors through exceptions
3811 . Lesson 9: Exchanging simple data
3812 . Lesson 10: Remote Procedure Calling (RPC)
3813 . Lesson 11: Explicitly waiting for messages
3814 . Recapping of message passing features in GRAS
3815 - A HOWTO section containing:
3816 o HOWTO design a GRAS application
3817 More are due, of course. They will come latter. In the meanwhile, you can
3818 check the examples which are still here.
3820 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3822 SimGrid (3.1) stable; urgency=high
3826 There was a stack corruption somewhere, visible only when optimizing
3827 with these versions. [Vince]
3830 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3831 module that enabled to play with Directed Acyclic Graphs. It is built
3832 directly on top of SURF and provides an API rather close to the old
3833 SG. Some old codes using SG are currently under rewrite to check that
3834 all needful functions are provided. [Christophe Thiery]
3837 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3838 extension of the classical CLM03 model that gracefully handles
3839 failures. This is now the default model for MSG and GRAS. It doesn't
3840 handle parallel tasks yet though. [AL]
3841 * Bug fix: Weights were not correctly set in the network part.
3842 WARNING: This may have resulted in incorrect results with simulations
3843 where there are more than one flow on a given link. [AL]
3846 * After a (long ?) discussion on simgrid-devel, we have decided that the
3847 convention we had on units was stupid. That is why it has been decided
3848 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3849 WARNING: This means that all previous platform files will not work as
3850 such with this version! A warning is issued to ask users to update
3852 A conversion script can be found in the contrib module of the CVS, under
3853 the name contrib/platform_generation/surfxml_update.pl [MQ]
3856 * Bug fix: Processes were started in reverse order, wrt deployment file.
3857 WARNING: if your code relies on this bug, please fix it. [AL]
3858 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3859 being executed on two different locations. [AL]
3860 * Bug fix: Failures are now better supported thanks to Derrick's tests
3861 (there was many failure situations I hadn't thought of and that weren't
3862 correctly handled). [AL]
3863 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3867 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3868 Exception occurring on server-side are propagated back to client (!).
3870 API CHANGE: the callback changed their prototype. Change:
3871 int my_handler(gras_socket_t expeditor, void *payload_data) {
3873 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3874 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3876 * New! function: gras_msg_handleall to deal with all messages arriving
3877 within a given period.
3878 * New! function: gras_socket_server_range to get a server socket in a
3879 range of port numbers (ease to avoid port number conflicts) [MQ]
3880 * New! gras processes display their backtrace when they get a SIGUSR1
3881 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3882 Sweet to debug RL processes [MQ]
3886 - Do not force experiment sizes to be expressed in kb, or it becomes
3887 impossible to measure the latency this way (needs one byte-long tests)
3888 WARNING: this changes the amok_bw_* function semantic. [MQ]
3889 - Implements the link saturation stuff. [MQ]
3890 * Peer management module:
3891 New! module factorizing code that we wrote over and over [MQ].
3894 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3895 - Test units are placed directly into the library code, they get extracted
3896 automatically and placed into the src/testall binary.
3897 - Convert most of the XBT tests to this system.
3898 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3899 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3900 live in the same C code. This required to change a little bit the API
3901 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3902 * New module: sparse graph structure with basic algorithms (this is work
3903 in progress and the API is not considered to be frozen yet). [AL]
3904 * Display more information on backtraces: source line & function names are
3905 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3906 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3907 * Reworked a little bit some #include statements to load only required
3908 headers. Some user code that relied on SimGrid to include stdlib or
3909 stdio may need to include it by themselves. [AL]
3910 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3912 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3913 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3914 but I guess that almost nobody used those parts. [MQ]
3916 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3918 SimGrid (3.0.1) stable; urgency=low
3921 * Unfortunately, I had missed 5 misnamed functions:
3922 xbt_fifo_item_t xbt_fifo_newitem(void);
3923 void xbt_fifo_freeitem(xbt_fifo_item_t);
3924 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3925 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3926 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3927 They're now deprecated. Please use their new versions:
3928 xbt_fifo_item_t xbt_fifo_new_item(void);
3929 void xbt_fifo_free_item(xbt_fifo_item_t);
3930 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3931 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3932 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3934 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3935 * Documentation: xbt_log module unmercifully reworked [MQ]
3936 * Bugfix: there was a problem with the ending of contexts with
3937 the pthread backend. It caused some weird deadlock or behavior
3938 depending on the pthread implementation. [AL]
3939 * Bugfix: get the exceptions raised in the simulator repport where
3940 and why they come from when they are not caught in time [AL, MQ]
3943 * Bugfix: Do repport the error when two non-connected hosts try to
3944 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3947 * Add additionnal checkings on communications. Assert that two
3948 communicating hosts are connected by a set of links... [AL]
3951 * Add additionnal checkings on channel values in communication [AL]
3952 * New: MSG_task_get_source to see on which host a task was generated [HC]
3953 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3954 the number of tasks waiting to be received on channel and sent
3956 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3957 waits for the first task coming from a given host.. [AL]
3959 GRAS new functionnalities: [MQ]
3960 * Enhance the parsing macro to allow the size of multidimentional objects
3961 to be given thru annotations.
3962 * New example (and documentation): Matrix Multiplication a la RPC
3963 (as when I was young!) and fix a bunch of bugs found on the way.
3965 GRAS performance improvements: [MQ]
3967 * Reduce the amount of cbps creation/destruction by making it static to
3968 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3971 * Change libdata to a set so that we can search for stuff by ID (and thus
3972 reduce the insane amount of dict lookups)
3975 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3976 RL communication on top of SG since it's so uneffective.
3977 It may also be used for inter-thread communication in RL, one day.
3978 * Use gras_datadesc_copy() to exchange messages on top of SG
3980 - improve message exchange performance on top of SG
3981 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3982 * Don't exchange on the network the size of the used part of buffer,
3983 instead, specify the possible buffer size to read().
3985 - reduces the amount of read/write calls (one pair per exchange)
3986 - reduces the amount of exchanged data (the size)
3987 - allows to retrieve all arrived data on receiver side, if we don't need
3988 it right now (subsequent read will peek the buffer)
3989 - allows the receiver to proceed with the beginning of the stream before
3990 everything is arrived
3991 - make it possible to build an iov transport (using readv/writev)
3993 - take care of the data with non-stable storage (like stacked data),
3995 * If possible, TCP send uses vector I/O (when writev() is here)
3996 - Don't use it for receive since we send data sizes and data on the
3997 same stream, so we wouldn't be able to chain large amount of chunks
3998 before having to flush the stuff to read the size.
3999 * Rework the transport plugin mechanism to simplify it and reduce the
4000 amount of pointer dereferencement when searching for the right function
4003 * I guess that now, we do almost as few system calls as possible while
4004 doing as few data copy as possible.
4006 To improve it further, we could try to send all the sizes first and then
4007 all the data (to use iov on receiving size), but it's only a partial
4008 solution: when you have 2 dimensional data, the sizes of the second
4009 dimension is data of the first dimension, so you need 3 streams.
4011 I'm not sure the potential performance gains justify the coding burden.
4013 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
4015 SimGrid (3.00) stable; urgency=high
4018 * New! Give the possibility to hijack the surf parser and thus bypass
4019 MSG_create_environment and MSG_launch_application. Have a look at
4020 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
4022 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
4024 SimGrid (2.96) unstable; urgency=low
4029 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
4030 This deprecates the xbt_error_t mechanisms.
4031 It modifies (simplifies) all XBT and GRAS API.
4032 MSG API keeps unchanged (exceptions raised by XBT are caught from
4033 within MSG and masked with existing error handling facilities)
4036 * New! Add a FATPIPE model. [AL]
4037 * New! Add a parallel task model. [AL]
4038 * New! Add automatically a loopback interface (in the default
4039 network model) if none was precised.
4042 * Bugfix: MSG_process_resume now works with the current running process.
4044 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
4045 * Modification of MSG_task_get_compute_duration. Once a task has been
4046 processed, the value returned by this function is now equal to 0. [AL]
4047 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
4048 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
4049 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
4050 of a m_task. For now, MSG_task_cancel only works with computation
4052 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
4053 of the processor (in Mflop/s) regardless of the current load on the
4055 * API Change: use proper naming convention for MSG_getClock and
4056 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
4058 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
4059 This function changes the priority of a computation task. This priority
4060 doesn't affect the transfer rate. A priority of 2 will make a task
4061 receive two times more cpu power than the other ones. This function
4062 has been added to suit the needs of Nguyen The Loc and hasn't been that
4063 much tested yet. So if it fails, please report it and send me your code.
4065 * API Change: removed all functions and types that were marked "deprecated"
4066 since many months. Renamed MSG_global_init_args to MSG_global_init.
4068 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
4070 SimGrid (2.95) unstable; urgency=low
4073 * Steal some nice code to GNU pth to fix context detection and usage [AL]
4074 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
4075 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
4078 * Bugfix: Allow absolute paths for platform description files [MQ]
4079 * Bugfix: do free the variables after use. Leads to drastic performance
4081 * Implement max_duration (ie, timeouts) on resources [AL]
4084 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
4086 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
4087 the needed support (timeouts) [AL]
4090 * Implement measurement sockets. You can now get the bandwidth between two
4091 hosts thanks to AMOK (see below). [MQ]
4092 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
4093 dynar over the network (yeah) [MQ]
4094 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
4095 * simple/static token-ring example. [Alexandre Colucci and MQ]
4096 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
4097 in place before [MQ]
4100 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
4101 type parsing. "short" and "long" modifiers were also ignored; other
4102 modifier (such as reference level) are still ignored. [MQ]
4103 * Embeed the buffer size within the buffer itself on SG. [MQ]
4104 That way, send() are atomic and cannot get intermixed anymore (at least
4105 the ones which are less than 100k; bigger messages still have the issue)
4106 * Array size pushed by the field, not by the field type (or each
4107 and every long int will push stuff to the cbps) [MQ]
4108 * use select() to sleep since it allows to portably sleep less than one
4111 GRAS (minor cleanups)
4112 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
4115 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
4116 * type category 'ignored' killed as it was never used and were difficult
4118 * whether a type can cycle or not is now a flag, leaving room for more
4119 flags (as "ignored", if we feel the need one day ;)
4120 * Rename raw sockets to measurement sockets since "raw" has another
4121 meaning in networking community.
4124 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
4125 and offers features not belonging to GRAS but that most applications
4126 need. One day, it may be a set of plugins loadable at runtime.
4127 * New module: bandwidth
4128 bandwidth measurement between arbitrary nodes running this module. [MQ]
4130 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
4132 SimGrid (2.94) unstable; urgency=low
4134 The first beta release of SimGrid 3 !
4138 * Update the main page and the FAQ. Adding references to gforge.
4141 * Add a gras_os_getpid function.
4144 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
4145 * Extend the logs so that they also print PID, hostname, date, ... if
4147 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
4148 and kill the old version which were in testsuite/
4149 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
4153 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
4154 for users. Instead, each programming environment (GRAS, MSG,...) use it
4155 when needed to provide such a feature to users.
4156 Don't shortcut the mechanism or you will also shortcut the virtualization
4157 you need on the simulator.
4161 * Cleanups in configury with regard to compile optimization/warning flags.
4162 Also add -fno-loop-optimize to any powerpc since it's the optimization
4163 killing gcc (< 3.4.0).
4164 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
4165 needed by MSG examples complications
4166 * Borrow configury beautifications from PHP
4169 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
4170 warning (thanks loris for stumbling into this one).
4171 * Bugfix: stop loading private headers (gras_config.h) from the public
4175 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
4176 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
4177 RL would imply a DNS resolver, which is *hard* to do in a portable way
4178 (and therefore delayed).
4179 * Implement a real timer mechanism and use it in timing macros. This allows
4180 to avoid rounding errors and get a 0.000005 sec precision in timing
4181 macros. While I was at it, various cleanups:
4182 - allow to declare more than one timed section per file (fix a stupid bug)
4183 - move some private declaration to the right place
4184 - merge conditional execution and timing macros into emulation module
4185 - document the module
4186 - make sure the module cleanups its mess on gras_exit
4187 * Documentation improvements:
4188 - (new) how to compile applications using GRAS
4189 - (new) emulation support (timing macros)
4191 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
4193 SimGrid (2.93) unstable; urgency=low
4195 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
4198 - Use Paje properly where used. Still to be sanitized properly.
4199 - Portability fix: Add an implementation of the contexts using pthread
4203 - Add xbt_procname(): returns the name of the current process.
4204 Use it to show the current process's name in all logging.
4206 - fix detection of older flex version and the reaction, since we do
4207 depend on modern ones (we use lex_destroy)
4208 - Better separation of SG and RL in the libs: remove all simulation code
4209 from libgras. As a result, this lib is now only 200k when stripped.
4210 Some of the xbt modules may also be duplicated (two sets and such) and
4211 should be cleaned/killed before SG3.
4212 - Insist on using xlC on AIX because of weird problems involving gcc there.
4213 - Cleanup the make remote stuff. This is now done by scripts
4214 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
4215 mainly for our private use, but we're working on changing them to user
4218 - Bugfix: flush the socket on close only if there is some *output*.
4219 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
4221 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
4224 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
4226 SimGrid (2.92) unstable; urgency=low
4228 Alpha 3 on the path to SimGrid 3
4232 - New! First try of benchmarking macros.
4233 - New! First try so that gras_stub_generator generate deployment and
4234 remote compilation helpers.
4236 - Bugfix: Initialization fix in msg_test.
4240 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
4242 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
4244 - API Change: gras_os_sleep to take the amount of seconds as a double.
4245 Accepting an int was error prone since it was the only location where
4246 seconds were coded as such. It leaded to damn rounding errors.
4247 - Bugfix: Hard to belive that timers ever worked before this.
4249 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4251 SimGrid (2.91) unstable; urgency=low
4253 Alpha 2 on the path to SimGrid 3
4257 - Bug fix in the lmm_solver.
4259 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4260 through the function MSG_paje_output.
4261 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4262 - It is possible to bound the rate of a communication in MSG with
4263 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4264 to write it in the changelog).
4265 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4270 - Various cleanups to the autotools stuff
4271 - Begin to move Gras examples to examples/gras/
4272 - Let make distcheck work again (yeah!)
4274 - documentation overhauled using doxygen.
4275 gtk-doc-tools is dead in SimGrid now.
4276 - Automatically extract all existing logging categories, and add the list
4277 to the documentation (long standing one, to say the less)
4279 - Cleanup the known architecture table. Reorder the entries to group what
4280 should be, and use a more consistent naming scheme.
4281 (some of the test dataset are still to be regenerated)
4282 - New! Allow library to register globals on each process just as userdata
4284 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4285 lookup time (for now).
4286 Use it in msg and trp.
4287 This cleans a lot the internals and helps enforcing privacy of the
4288 headers between the gras components.
4289 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4290 - Bugfix: gras_os_time was delirious in RL.
4291 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4293 - Reenable GRAS now that it works.
4295 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4297 SimGrid (2.90) unstable; urgency=low
4299 Alpha 1 on the path to SimGrid 3
4301 * It is a long time since the last release of SimGrid. I'm sorry about
4302 that but as I had told you, I was rewriting a lot of things. I apologize
4303 to those who had been reporting bugs to me and that I had not answered.
4304 If your bug is still in the new version, please tell me. Here is a
4305 summary of the main changes.
4307 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4308 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4309 portability, speed, and a lot more but you'll figure it out later.
4310 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4311 I wanted to release the new MSG as soon as possible and I have
4312 broken GRAS, which is the reason why, for now, only MSG is fully
4313 functional. A laconic description of these projects is available
4314 in the documentation.
4316 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4317 kernel optimized for our needs (called SURF but only the developers
4318 should use it). Hence, MSG is now roughly 30 times faster and I think
4319 that by rewriting a little bit MSG, I could event speed it up a little
4320 bit more. Beside the gain in speed, it is also much easier to encode a
4321 new platform model with SURF than it was with SG. More to come...
4323 * REVOLUTION 3: I have tried to change a little as possible the API of
4324 MSG but a few things really had to disappear. The main differences
4325 with the previous version are:
4326 1) no more m_links_t and the corresponding functions. Platforms are
4327 directly read from an XML description and cannot be hard-coded
4328 anymore. The same format is used for application deployment
4329 description. The new format is described in the documentation.
4330 Have a look in tools/platform_generation. There is a tiny script
4331 that converts from the old platform format to the new one. Concerning
4332 the application deployment format, parsing the old one is tricky.
4333 I think most of you should however be able to convert your files. If
4334 it is really an issue, I can write a C code that does the conversion.
4336 2) the toolbox tbx does not exist anymore. We now have a library
4337 with much more data-structures but without the hash-tables (we have
4338 dictionaries that are much faster).
4340 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4342 *****************************************************************************
4343 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4344 * don't feel like converting the oldies. (Mt) *
4345 *****************************************************************************
4348 Version 2.90: "the long awaited one"
4349 - Finished rewriting and debugging MSG. Rewrote the documentation.
4350 - disable GRAS for now since it needs to be ported to the newest SG
4353 - Finish the port to windows (using mingw32 for cross-compile)
4356 - Main loop and datastructures of SURF. A cpu resource object is
4357 functional. Surf can thus be used to create cpu's with variable
4358 performance on which you can execute some actions.
4360 2004-11-15 Martin Quinson
4361 - Port to ARM. Simply added the alignment and size descriptions. Should
4362 work, but the ARM machines are so slow that I didn't had the opportunity
4363 to 'make check' over there yet.
4365 2004-11-15 Arnaud Legrand
4366 - Trace manager now written. It uses a heap structure and is therefore
4367 expected to be efficient. It may however be speeded up (particularly
4368 when many events occur at the same date) by using red and black
4369 trees. One day maybe...
4370 - Max-min linear system solver written. It uses a sparse matrix
4371 structure taking advantage of its expected use. Most operations are
4372 O(1) and free/calloc are called as few as possible. The computation of
4373 the minimum could however be improved by using a red and black tree
4376 2004-11-03 Arnaud Legrand
4377 - Rename every gras_* function that was in xbt/ to its xbt_
4379 - Add a heap and a doubly-linked list to xbt
4380 - Added a dichotomy to the dictionaries. make check works as well before
4381 so I assume that the patch is correct. I do not know however if things
4382 run effectively faster than before now. :)
4384 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4385 SimGrid, and the version number is bumped to 2.x
4387 2004-10-29 Martin Quinson
4388 - Introduction of the remote errors.
4389 They are the result of a RMI/RPC on the remote machine.
4390 ErrCodes being scalar values, you can't get the host on which those
4391 errors did happen. Extending the error mechanism as in Gnome is possible.
4392 No idea yet whether it is a good idea.
4394 2004-10-28 Martin Quinson
4395 - Interface revolution: the Starred Structure Eradication.
4396 I used to do typedef struct {} toto_t; and then handle *toto_t.
4397 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4398 - ???_t is a valid type (builded with typedef)
4399 - s_toto_t is a structure (access to fields with .)
4400 - s_toto is a structure needing 'struct' keyword to be used
4401 - e_toto_t is an enum
4402 - toto_t is an 'object' (struct*)
4404 typedef struct s_toto {} s_toto_t, *toto_t;
4405 typedef enum {} e_toto_t;
4406 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4407 s_toto_t) is private.
4409 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4410 so that it can set the variable to NULL. It was so for dicts and sets,
4411 it changed for dynars.
4413 - Fix a bunch of memleaks in dict_remove
4414 - Fix a bug in sg/server_socket opening: it failed all the time.
4416 2004-10-07 Martin Quinson
4417 - Speed up dynar lookup operation a bit.
4419 gras_dynar_get is dead.
4421 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4422 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4423 (which gives you the address of the stored data).
4425 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4426 copy of the data using an affectation to do the job and not a memcpy.
4428 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4429 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4431 It does not really speedup the dynar test because they are
4432 setting elements all the time (and look them seldom). But the dict does
4433 far more lookup than setting.
4435 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4437 2004-10-05 Martin Quinson
4438 - Allow to (en/dis)able the cycle detection at run time.
4440 Whether we should check for cycle or not is now a property of each
4441 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4442 datadesc_cycle_unset allow to remove this property when previously set.
4444 Note that the cycle detection is off by default since it impacts the
4445 performance. Watch the data you feed GRAS with ;)
4447 This property is hereditary. Any element embedded in a structure having it
4448 set have it set for the time of this data exchange.
4450 You should set it both on sender and receiver side. If you don't set it on
4451 sender side, it will enter an endless loop. If you forget on receiver
4452 side, the cycles won't be recreated after communication.
4454 - Header reorganization.
4455 Kill gras_private.h, each submodule must load the headers it needs.
4457 2004-10-04 Martin Quinson
4458 - Interface revolution: do not try to survive to malloc failure.
4460 Now, gras_malloc and friends call gras_abort() on failure.
4461 As a conclusion, malloc_error is not a valid error anymore, and all
4462 functions for which it was the only gras_error_t return value are
4463 changed. They now return void, or there result directly.
4464 This simplify the API a lot.
4466 2004-09-29 Martin Quinson
4467 - Re-enable raw sockets.
4468 Created by gras_socket_{client,server}_ext;
4469 Used with gras_raw_{send,recv}
4472 It should allow to kill the last bits of gras first version soon.
4474 This is not completely satisfactory yet (duplicate code with
4475 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4478 - Simplify transport plugin (internal) interface by not passing any
4479 argument to _server and _client, but embedding them in the socket
4482 2004-09-28 Martin Quinson
4483 - Finish the port to AIX.
4484 autoconf was my problem (segfault within the malloc replacement
4485 function. No idea why)
4487 2004-09-16 Martin Quinson
4488 - Fix some size_t madness on 64bit architectures.
4490 2004-09-08 Martin Quinson
4491 - Reduce the number of system headers loaded, overload some more system
4492 calls (such as malloc to cast the result of the system one, and work
4494 - Fix and reintroduce the config support
4496 2004-09-07 Martin Quinson
4497 - Source code reorganization to allow Arnaud to surf all over there.
4498 - Allow to document the logging categories.
4499 - Remove all uppercase from logging categories and useless cleanup in names.
4501 2004-08-18 Martin Quinson
4502 Version 0.6.2 (protocol not changed; API changed)
4503 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4504 gras_error_t), and NULL when not found. Functions expecting a msgtype
4505 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4506 hopefully usefull message.
4507 - Portability to prehistoric sparcs again
4509 2004-08-17 Martin Quinson
4510 Version 0.6.1 (protocol not changed; ABI not changed)
4511 - prealloc some buffers to speed things up
4513 2004-08-11 Martin Quinson
4514 Version 0.6 (protocol not changed; ABI expended)
4515 - The parsing macro can deal with the references, provided that you add
4516 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4518 2004-08-09 Martin Quinson
4519 Version 0.5 (protocol not changed; ABI changed)
4520 - Allow to off turn the cycle detection code in data exchange at
4521 compilation time. It should be at run time, but I'm short of time (and
4522 the config stuff is still broken). That way, we keep dict out of the
4523 critical path, which is good because the performance is poor:
4524 - search not dichotomial yet
4525 - dynar give no way to access their content and memcpy everytime
4526 - In composed data description (struct, ref and so on), stop foolness of
4527 keeping the subtype's ID, but store the type itself. This keeps sets out
4528 of the critical path, which is good since they rely on dynar and
4529 dictionaries. The only loose of that is that we cannot detect the
4530 redeclaration of a structure/union with another content (but I'm not sure
4531 the code detected well this error before anyway). We still can detect
4532 the redefinition discrepancy for the other types.
4533 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4534 it breaks the code because of type-punning used all over the place).
4535 This breaks on all non-gcc architectures (for now).
4537 All those changes (plus the buffer of last time) allow me to gain 2 order
4538 of magnitude on cruel tests consisting of 800000 array of integers on two
4539 level of a hierarchical structure (200 secondes -> 4 secondes)
4542 - the selector of reference must now return the type it points to, not
4543 the ID of this type.
4545 2004-08-06 Martin Quinson
4546 Version 0.4 (protocol changed; ABI not changed)
4547 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4548 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4549 - Implement a buffer transport, and use it by default (it relies on tcp in
4550 real life and on sg in simulation).
4551 That's a bit hackish since I had a new field to the structure to store
4552 its data without interfering with the subtype ones. Inheritance
4553 is tricky in C. And that's a kind of reverse inheritance with one class
4554 derivating two classes. Or maybe a game with java interfaces. Anyway,
4555 that's damn hard in C (at least).
4556 Moreover, I got tired while trying to ensure plugin separation and
4557 genericity in SG mode. MSG wants me to do weird things, so let's go for
4558 cruel hacks (temporarily of course ;).
4559 See comment in transport_private.h:71
4560 - do not include all the _interface headers in private but in the files
4561 which really need them (to cut the compilation time when they are
4564 2004-07-26 Martin Quinson
4565 Version 0.3 (protocol not changed; ABI changed)
4566 - Major overhault of the datadesc interface to simplify it:
4567 - shorted the function names:
4568 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4569 - add a trivial way to push/pop integers into the cbps without malloc.
4570 This allows to make really generic sub_type description, which simply
4571 pop their size of the stack.
4572 - add a function gras_datadesc_ref_pop_arr() which does what users want
4573 most of the time: Declare a dynamic array (which pops its size of the
4574 stack) and declare a reference to it. Poor name, but anyway.
4575 - kill the post-send callback, add a post-receive one
4577 2004-07-23 Martin Quinson
4578 Version 0.2 (protocol changed; ABI changed)
4579 - add some testing for cpbs in the test cases, and fix some more bugs.
4580 This invalidate again the little64 data file, since I cannot regenerate
4582 - remove an awfull optimization in the logging stuff, allowing me to:
4583 - understand it again
4584 - learn gcc how to check that the argument match the provided format
4585 - fix all errors revealed by gcc after that
4586 - internal keys of dict are not \0 terminated. Deal with it properly in
4587 loggings instead of segfaulting when the user want to see the logs :-/
4589 2004-07-22 Martin Quinson
4590 - Fix some stupid bug preventing cbps (callback postit) from working
4592 2004-07-21 Martin Quinson
4593 - Some documentation cleanups
4594 - remove the useless last argument of msgtype_declare
4595 - rename the Virtu functions to fit into the 'os' namespace
4596 - move headers src/include -> src/include/gras/ and stop fooling with
4597 gras -> . symbolic link
4598 - make distcheck is now successful
4600 2004-07-19 Martin Quinson
4602 - Build shared library also
4603 - Install html doc to the right location
4604 - stop removing maintainer files in make clean
4605 - build tests only on make check
4607 2004-07-13 Martin Quinson
4609 - No major issue in previous version => change versioning schema
4610 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4611 corresponding dataset.
4613 2004-07-11 Martin Quinson
4615 - Get it working with any kind of structure (we can compute the padding
4616 bytes remotely for all the architectures I have access to)
4617 - Implement the structure parsing macro (still not quite robust/complete)
4618 - Improvement to the remote testing toysuite
4620 2004-07-10 Martin Quinson
4621 [autoconf mechanism]
4622 - get ride of a bunch of deprecated macros
4623 - actually run the test for two-compliment, not only compile it :-/
4624 - test whether the structures get packed (and bail out if yes. Damn.
4625 Alignment is a serious matter)
4626 - test whether the structures get compacted (but respecting the alignment
4627 constraints of each types)
4628 - test whether the array fields of structures can straddle alignment boundaries
4630 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4631 (took me 2 hours to find that bug, looking at the wrong place)
4632 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4633 before sending/receiving and are used to compute the offsets of fields
4634 - Given that padding size depend even on compiler options, keep track of
4635 alignment and aligned_size only for the current architecture. Not a big
4636 deal since we send structure fields one after the other (seems
4638 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4639 program, let it work on linux/gcc/little32. portability todo.
4641 2004-07-08 Martin Quinson
4642 - import and improve remote compilation support from FAST
4643 - make sure make check works on half a dozen of machines out there
4645 2004-07-07 Martin Quinson
4646 Let's say it's version 0.0.3 ;)
4647 - Implement conversions (yuhu!)
4648 - Let it work on solaris (beside conversion, of course)
4649 - Stupid me, using rand() to generate the conversion datatests in not wise.
4651 2004-07-06 Martin Quinson
4652 - Let make dist work, since I'm gonna need it to compile on remote hosts
4653 - Let Tests/datadesc_usage write the architecture on which the file was
4654 generated as first byte.
4655 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4656 architecture descriptions.
4657 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4658 on those architectures.
4659 - Optimization: send/recv array of scalar in one shoot
4661 2004-07-05 Martin Quinson
4662 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4664 - Plug a whole bunch of memleaks
4665 - each process now have to call gras_{init,exit}. One day, their log
4666 settings will be separated
4667 - Continue the code factorisation between SG, RL and common in Transport.
4669 2004-07-04 Martin Quinson
4671 - Redistribution between SG and RL.
4672 We wanna have to accept in SG, so move accepted related parts of RL in
4673 the common part. (more precisely, the dynar of all known sockets is no
4674 more a static in transport.c, but part of the process_data)
4676 [gras_stub_generator]
4677 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4679 2004-07-03 Martin Quinson
4680 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4681 them all up in one shot)
4683 - Fix the ugly case of reference to dynamic array.
4684 - New (semi-public) function gras_datadesc_size to allow the messaging
4685 layer to malloc the needed space for the buffer.
4687 - gras_socket_close now expect the socket to close (and not its address to
4688 put NULL in it after it). This is because the socket passed to handlers
4689 is one of their argument (=> not writable).
4691 - propagate the interface cleanup from last week in datadesc, ie remove a
4692 superfluous level of indirection. User pass adress of variable
4693 containing data (both when sending and receiving), and not of a variable
4694 being a pointer to the data. Let's say that I like it better ;)
4695 The price for that is constructs like "int msg=*(int*)payload" in
4696 handlers, but it's a fine price, IMHO.
4698 - Let it work in RL (yuhu)
4700 2004-06-21 Martin Quinson
4702 - porting SG plugin and SG select to new standards (works almost).
4703 - plug memleaks and fix bugs around.
4706 - cleanup the prototype of data recv and force users to specify when they
4707 want to handle references to objects. Test case working even for cycles.
4708 - plug memleaks. Valgrind is perfectly ok with this.
4710 2004-06-12 Martin Quinson
4712 - cleanup the separation between plugin and main code in plugin creation
4714 2004-06-11 Martin Quinson
4716 - Reput hook for raw sockets, needed for BW experiments
4717 - kill a few lines of dead code
4718 [Data description] Interface cleanup
4719 - gras_datadesc_by_name returns the searched type or NULL.
4720 That way, no variable is needed to use a type desc once, which makes
4722 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4723 two parameters were strings (field name, type name), leading to
4725 [Dicos] Interface cleanup
4726 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4727 This is consistant with the dynar API.
4729 2004-04-21 Martin Quinson
4731 - Porting to new standards.
4733 - interface cleanup.
4734 There is no bag anymore, no need to take extra provision to mask the
4735 pointers behind "ID".
4736 Better splitup of functions between files create/exchange/convert.
4737 This is still a bit artificial since convert and receive are so
4738 interleaved, but anyway.
4740 - add a queued message list to procdata (the ones not matching criteria
4742 - factorize some more code between SG and RL wrt procdata
4744 - use gras_exit in example to track memleaks
4745 - get rid of gs_example now that GS is properly integrated into gras
4746 - update run_test to integrate the latest tests (datadesc)
4748 - rename WARNINGn macros to WARNn since it prooved error-prone
4750 2004-04-19 Martin Quinson
4752 - register init/exit functions within gras module mechanism
4753 - send/receive function.
4754 Convertion is not implemented, but short-cutted if not needed.
4755 struct/array elements are sent one by one (instead of block-wise), but
4756 nobody really cares (yet). Get a prototype before optimizing.
4757 - tests (using a file socket) for DD send/receive on:
4758 - base types: int, float
4759 - array: fixed size, string (ie ref to dynamic string)
4760 - structure: homogeneous, heterogeneous
4761 - chained list, graph with cycle
4762 Believe it or not, valgrind is not too unhappy with the results. The
4763 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4767 - Bugfix when using a filename explicitly (instead of '-')
4769 2004-04-09 Martin Quinson
4771 - factorize more code between RL and SG in socket creation
4772 - Complete the implementation and tests of:
4774 o file (only in RL, and mainly for debugging)
4776 I lost 3 days to design a portable address resolver, and then decided
4777 that the prototype mainly have to run on my box.
4778 Addressing portability too early may be like optimizing too early :-/
4780 - use gras_init in the Tests instead of the crappy parse_log_opt
4781 (the latter function is removed)
4782 [Conditional execution]
4783 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4784 [Code reorganisation]
4785 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4786 Build examples against the RL library, since there is no way to disable
4787 its creation for now.
4789 For information, the beginning of coding on GRAS was back in june
4790 2003. I guess that every line has been rewritten at least twice since