1 SimGrid (3.31.1) NOT RELEASED YET (v3.32 expected June 21. 2022, 09:13 UTC)
4 - Added the xbt_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/]
23 - Mailbox: Mailbox.ready [example: examples/python/comm-ready/]
24 - Ptask [example: examples/python/exec-ptask/]:
25 - this_actor.exec_init
26 - this_actor.parallel_execute
29 - Added an AssertionError exception that may be thrown in case of error. For instance, creating tow hosts with the
30 same name will now throw this exception instead of killing the interpreter.
32 Platform description & visualization:
33 - More robust sanity checks for platforms, to reject unallowed topologies with
34 a proper error message.
35 - One new C++ platform example, supernode.cpp. A Python script (supernode.py)
36 demonstrates how we can generate a nice graphical representation of the
40 - Modified the host_by_name functions:
41 - Now, they return only hosts, not VMs, and in a much more efficient way.
42 - If one wants to find a VM by name, he now needs to know the host on
43 which it runs and call vm_by_name (or manually iterate over the list of hosts)
46 - Enhancements to the graphicator tool:
47 - Allow to dump the platform topology as a CSV file representing the edges
48 of the graph (in addition to the DOT format).
49 - Fix graphicator for "cluster" topologies (e.g. fat-tree, dragonfly).
51 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
52 (FG: issues on Framagit; GH: issues on GitHub)
53 - FG#105: "Variable penalty should not be negative!" with in-flight messages and bandwidth profiles
54 - FG#109: Application time reported by --cfg=smpi/display-timing:yes is wrong
55 - FG!109: Trigger new engine solve upon host events such as host on/off
56 - FG#110: Wait_any does not trigger new model solve when host events occur
58 ----------------------------------------------------------------------------
61 - kill signal Comm::on_completion that was not working anyway.
62 - Expose signals Activity::on_suspend and Activity::on_resume
64 SimGrid (3.31) March 22. 2022.
66 The ненасильство release. We stand against war.
68 Against the agression by a sick system that forces peoples to take arms against each other.
71 - Rework the internals, for simpler and modern code. This shall unlock many future improvements.
72 - You can now define plugins onto the DFS explorer (previously called SafetyChecker), using the
73 declared signals. See CommunicationDeterminism for an example.
74 - Support mutex, semaphore and barrier in DPOR reduction
75 - Seems to work on Arm64 architectures too.
76 - Display a nice error message when ptrace is not usable.
77 - New test suite, imported from the MPI Bugs Initiative (MBI). Not all MBI generators are integrated yet.
78 - Remove the ISP test suite: it's not free software, and it's superseeded by MBI.
81 - fix for FG#100 by ensuring small asynchronous messages never overtake larger
82 ones, conforming to the standard.
83 - replay: fix waitall behaviour to avoid forgetting requests and leaking
85 - tracing: ensure that we dump the TI traces continuously during execution and
86 not just at the end, reducing memory cost and performance hit.
87 - Update OpenMPI collectives selection logic to match current one (4.1.2)
88 - Add a coherence check for collective operation order and root/MPI_Op
89 coherence. Potentially costly so not activated unless smpi:pendantic is set
93 - New signal: Engine::on_simulation_start_cb()
94 - Introduce a new execution mode with this_actor::thread_execute(). This simulate
95 the execution of a certain amount of flops by multiple threads ran by a host. Each
96 thread executes the same number of flops, given as argument. An example of this new
97 function can be found in examples/cpp/exec-threads.
98 - Reimplementation of barriers natively.
99 Previously, they were implemented on top of s4u::Mutex and s4u::ConditionVariable.
100 The new version should be faster (and can be used in the model-checker).
101 - Actor::get_restart_count(): Returns the number of reboots that this actor did.
104 - MSG_barrier_destroy now expects a non-const msg_barrier parameter.
106 New plugin: the Chaos Monkey (killing actors at any time)
107 - Along with the new simgrid-monkey script, it tests whether your simulation
108 resists resource failures at any possible timestamp in your simulation.
109 - It is mostly intended to test the simgrid core in extreme conditions,
110 but some users may find it interesting too.
113 - New solver for parallel task: BMF.
114 - More realistic sharing of heterogeneous resources compared to the fair
115 bottleneck solver used by ptask_L07.
116 - Implement the BMF (Bottleneck max fairness) fairness.
117 - Improved resource sharing for parallel tasks with sub-flows (parallel
118 communications between same source and destination inside the ptask).
120 - "--cfg=host/model:ptask_L07 --cfg=host/solver:bmf": enable the ptask
121 model with BMF solver.
122 - "--cfg=bmf/max-iterations: <N>": maximum number of iterations performed
123 by BMF solver (default: 1000).
124 - "--cfg=bmf/precision: <N>": numerical precision used when computing
125 resource sharing (default: 1e-12).
126 - This model requires Eigen3 library. Make sure Eigen3 is installed to use BMF.
129 - Modifications of the Profile mechanism, with some impact on users
130 - Addition of a new (S4U) method to init profiles from generic functions to improve versatility
131 - Fix initial behaviour of state_profiles
132 - Modify periodicity to behave like a period, and not like a loop delay
135 - Drop xbt_dynar_shrink().
138 - Made the following bindings static (previously member functions):
139 - Actor: Actor.kill_all(), Actor.by_pid()
140 - Host: Host.by_name(), Host.current(), Host.on_creation_cb()
141 - Mailbox: Mailbox.by_name()
142 - Added the following bindings:
143 - this_actor.warning()
144 - Mailbox.put_init() [example: examples/python/comm-waitallfor/]
145 - Comm.detach() [example: examples/python/comm-waitallfor/]
146 - Comm.wait_for() [example: examples/python/comm-waitfor/]
147 - Comm.wait_any_for()
148 - Comm.wait_all_for() [example: examples/python/comm-waitallfor/]
149 - Mutex [example: examples/python/synchro-mutex/]
150 - Barrier [example: examples/python/synchro-barrier/]
151 - Semaphore [example: examples/python/synchro-semaphore/]
154 - Remove target "make uninstall" which was incomplete and no longer maintained.
156 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
157 (FG: issues on Framagit; GH: issues on GitHub)
158 - FG#57: Mc SimGrid should test whether ptrace is usable
159 - FG#87: Smpi scripts fail with spaces in paths
160 - FG#100: [SMPI] Order of the message matching is not guaranteed
161 - FG#101: LGPL 2.1 is deprecated license
162 - FG#104: "make uninstall" not up-to-date
163 - GH#151: Missing mutexes for DPOR.
165 ----------------------------------------------------------------------------
167 SimGrid (3.30) January 30. 2022.
169 The Sunday Bloody Sunday release.
171 Main user-visible changes:
172 - The SimDag API for the simulation of the scheduling of Directed Acyclic
173 Graphs has been dropped. It was marked as deprecated for a couple of years.
174 We finally complete the implementation of what has been called SimDag++
175 internally, i.e., porting the different features of SimDag on top of S4U.
176 The new way to simulate the execution of dependent activities directly by
177 maestro (without any other actor) is details in the examples/cpp/dag-* series
179 - The removal of SimDag led us to also remove the export to Jedule files that
180 was tightly coupled to SimDag. The instrumentation of DAG simulation is still
181 possible through the regular instrumentation API based on the Paje format.
182 - We also dropped the old and clumsy Lua bindings to create platforms in a
183 programmatic way. It can be done in C++ in a much cleaner way now, which
184 motivates this suppression.
187 - Introduce on_X_cb() functions for all signals, to attach a new
188 callback to the signal X. The signal variables are now hidden and
189 only these functions should be used.
190 Rationale: this enables the usual deprecation scheme where functions
191 remain for 4 releases if we need to modify the signals, while the
192 current code with the signal variables directly visible prevents any
194 - New function: Engine::run_until(date), to split the simulation.
195 - New signal: Activity::on_veto, to detect when an activity fails to start.
196 - Signal change: Comm::on_start(Comm&, bool) has been replaced by
197 Comm::on_send and Comm::recv. These two signals respectively correspond to
198 when the sending or receiving side of a Comm is ready. They are raised at
199 the same locations as the former Comm::on_start signal.
200 - New function: Engine::track_vetoed_activities() to interrupt run()
201 when an activity fails to start, and to keep track of such activities.
202 Please see the corresponding example for more info.
203 - New functions: s4u::Comm::{sendto_init, set_source, set_destination} to enable
204 the use of vetoers with direct host-to-host communications. Both source and
205 destination have to set for a comm to start. Each call to these setters check
206 if all vetoes are satisfied. When it is the case, the comm starts. A use case of
207 these functions is given in examples/cpp/dag-scheduling.
208 - New functions: {Exec, Io}::update_priority allow you to modify the priority of
209 these kinds of activities during their execution. Behavior is detailed in
210 examples/cpp/io-priority/.
213 - Dynamic costs for MPI operations: New API to allow users to dynamically
214 change injected costs for MPI_Recv, MPI_Send and MPI_Isend operations.
215 Alternative for smpi/or, smpi/os and smpi/ois configuration options.
216 - Fix some issues with the replay mechanism.
219 - Function xbt::Extendable::get_data() is now templated with the type of the
220 pointee. Untyped function is deprecated. Use get_data<void>() if you still
221 want to retrieve void*.
224 - New section: "SimGrid MPI calibration of a Grid5000 cluster"
225 presenting how to properly calibrate MPI communications in SimGrid.
226 - Complete and reword the platform section, which is now completed.
229 - Thread contexts are used by default with Python bindings. Other kinds of
230 contexts revealed unstable, specially starting with pybind11 v2.8.0.
232 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
233 (FG: issues on Framagit; GH: issues on GitHub)
234 - FG#95: Wrong computation time for multicore execution after pstate change
235 - FG#97: Wrong computation time for ptask+multicore+pstates
236 - FG#98: SMPI offline simulation is inconsistent with the online simulation
237 (deadlocks / message truncation)
238 - FG#99: Weird segfault when not sealing an host
240 ----------------------------------------------------------------------------
242 SimGrid (3.29) October 7. 2021
244 The "Ask a stupid question" release.
246 We wish that every user ask one question about SimGrid to celebrate.
247 On Mattermost, Stack Overflow or using the issues tracker.
249 New modeling features:
250 - Non-linear resource sharing, modeling resources whose performance heavily degrades with contention:
251 - The total capacity may be updated dynamically through a callback
252 and depends mainly on the number of concurrent flows.
253 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
255 - Dynamic factors: model variability in the speed of activities
256 - Each action can now have a factor that affects its progression.
257 This multiplicative factor is applied when updating the amount of work
258 remaining, thereby an activity with factor=0.5 only uses half of the
259 instantaneous power/bandwidth it is allocated and will appear twice
260 slower than what it actually consumes.
261 - This can be used to model a overhead (e.g., there is a 20 bytes
262 header in a 480 bytes TCP packet so the factor 0.9583) but the novelty
263 is this factor can now easily be adjusted depending on activity's and
264 resources characteristics.
265 - This existed for network (e.g., the effective bandwidth depends
266 on the message in SMPI piecewise-linear network model) but it is now
267 more general (the factor may depend on the source and destination and
268 thus account to different behaviors for intra-node communications and
269 extra-node communications) and is available for CPUs (e.g., if you
270 want to model an affinity as in the "Unrelated Machines" problem in
271 scheduling) and disks (e.g., if you want to model a stochastic
273 - For that, resources can be provided with a callback that computes
274 the activity factor when creating the action.
275 - Example: examples/cpp/exec-cpu-factors
276 - The same mechanism is also available for the latency, which
277 allows to easily introduce complex variability patterns.
280 - Added support to programmatic platform creation in Python.
281 Example: examples/python/clusters-multicpu
284 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
285 This can only be set through the API, not through XML files.
289 - Multiple fixes to ensure reproducibility of tracing
290 - scan/excan can now be replayed
291 - wait action now uses ranks and not pid, as the other ones.
292 - smpi/init and smpi/finalization-barrier are now valid for replays.
293 - exit() is now intercepted by SMPI to avoid premature shutdown of
294 simulation. First non 0 return codes is returned as simulation return
298 * New section "Release Notes" documenting recent and current developments.
299 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
300 * Improvements in API Reference for C++ and Python interfaces.
303 - Make wifi creation compatible with ns-3 version 3.34 too.
305 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
306 (FG: issues on Framagit; GH: issues on GitHub)
307 - FG#77: Search feature of doc is broken (update sphinx theme version)
308 - FG#78: Multiple fixes for SMPI replay:
309 - TI tracing of allotallv/w was outputting wrong values
310 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
312 ----------------------------------------------------------------------------
314 SimGrid (3.28) July 14. 2021
316 The Victoriadagarna Release.
319 - C++ platform interface: Users can now describe their platform directly in C++.
320 This provides greatly flexibility and performance improvement for complex
321 platforms. Main features:
322 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
323 "zones", instead of single hosts. This feature enables the description
324 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
325 - StarZone: new zone with a Star-like topology. The routes are defined
326 as a set of links used to communicate from node to everybody (node<->ALL).
327 - Split-Duplex links: auxiliary method to create split-duplex links in
328 the platform, easing its utilisation. It automatically creates both UP
329 and DOWN links (similarly as done in XML).
330 - Please refer to the documentation and the examples included:
331 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
332 - New plugin: Producer-Consumer with monitor. Just requires to include the
333 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
334 associated example (examples/cpp/plugin-prodcons).
337 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
338 s4u::Io::wait_any(), s4u::Io::wait_any_for().
339 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
340 parameter by reference, instead of a pointer.
341 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
342 Explicitly cancel the activity to get back to previous behavior.
343 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
346 - The default SMPI compiler flags are no more taken from the environment.
347 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
348 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
350 --cfg=smpi/finalization-barrier: which can be used to add
351 a barrier inside MPI_Finalize. This can help for some codes which cleanup
352 data attached to a process, but still used in other SMPI processes.
353 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
354 is active when set to false, to keep going after a small error
355 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
356 which may or may not be problematic in the end.
358 - fix behaviour, as maximum iteration count could be ignored
359 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
360 use sampling when the same kernel is called with a different set of parameters
361 which have an impact on the timing.
362 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
363 It should now work with smpi/auto-shared-malloc-thresh.
364 - Improve error handling and reporting in multiple places
365 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
366 validity checks, truncated messages are now an error, return errors
367 when explicitely deleted handles are reused, ...)
368 - RMA: multiple fixes and stability improvements.
369 - analysis (-analyze flag in smpirun):
370 - SMPI can now report buffer leaks as well as MPI handles leaks,
371 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
372 - if -trace-call-location is used when compiling, SMPI can report
373 origin of leaked handles/buffers
374 - group leaks by type/origin in output message if possible
375 - New implemented MPI calls: MPI_Comm_test_inter
378 - Changed internal implementation of bandwidth factors in network models.
379 Models affected: CM02, LV08 (default), SMPI, IB.
380 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
381 Bandwidth factors are applied to communications to describe that users
382 cannot use 100% of the available bandwidth. For example, the default network model,
383 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
384 behavior was implemented by limiting the bandwidth available in the LMM
385 system for this flow. This may give the false impression that there is
386 bandwidth available for other flows due to its underutilization, especially
387 for the dynamic bandwidth factors used in SMPI models.
388 To avoid this, we have modified the implementation so that each flow uses the
389 maximum physical bandwidth according to the LMM system.
390 However, the actual throughput of the flow seen by the user is defined by
391 the physical bandwidth multiplied by the bandwidth factor.
392 This change impacts on the simulation results for all network models on
393 which we have bandwidth factors configured.
394 *****************************************
395 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
396 *****************************************
397 This change may impact on the timing of your simulation results.
398 Take care when comparing simulations from different SimGrid's
399 versions. Sorry for the inconvenience.
400 - Dynamic network factors: users can configure a callback to define
401 the network factors dynamically. This API is available at
402 simgrid::kernel::resource::NetworkModelIntf.
403 - Users have access to complete information about the current communication
404 to decide which factor to apply. This includes: message size, source and
405 destination hosts, links and zones traversed.
406 - Dynamic factors for both latency and bandwidth.
407 - For more details, see the example in (examples/cpp/network-factors).
408 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
409 deprecated since version 3.24, are no longer supported. Instead, use
410 "wattage_off" and "wattage_per_state".
413 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
416 - New tutorial: Model-checking and formal assessment
417 - New sections: "Demystifying the routing" and "C++ platforms"
418 - Update and improve the part on visualization in MPI and Algo tutorials.
419 - Improve the section on routing: how to define it, how it's used internally
420 - Fix many issues, broken links and missing references in doxygen and Sphinx
423 - Lua platform files are deprecated. Their support will be dropped after v3.31.
426 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
428 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
429 (FG: issues on Framagit; GH: issues on GitHub)
430 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
431 - FG#64: Configuring smpi/IB-penalty-factors
432 - FG#67: Running computation concurrently with MPI_Iallreduce
433 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
434 - FG#71: Segmentation fault on invalid gw_src/gw_dst
435 - GH#322: Issue when an actor kills his host vm
437 ----------------------------------------------------------------------------
439 SimGrid (3.27) March 29. 2021
441 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
443 Main user-visible changes:
444 - All functions marked as deprecated for one year were removed.
445 - Removed all that was related to the "Storage" type in all APIs and bindings.
446 "Disks" have been introduced more than a year ago to replace them for the better.
447 It's time to finish this replacement.
448 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
449 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
450 avoiding "initializer element is not constant" errors with some codes. Please be
451 careful to recompile your MPI codes if switching to this new SimGrid version.
452 - Network model ns-3 now requires version ns-3.28 or higher.
453 - Removed some option aliases that were deprecated since v3.23 (June 2019)
454 - 'verbose-exit' is now 'debug/verbose-exit'
455 - 'simix/breakpoint' is now 'debug/breakpoint'
456 - 'clean-atexit' is now 'debug/clean-atexit'
459 - New functions: Comm::sendto_{init,async} to initiate a communication
460 on between two (possibly remote) hosts.
461 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
462 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
463 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
464 or Mailbox::get_async<void>() if you really want to play with void*.
465 - Unify the interface of Activity::on_{start/activity}
466 - New function: Comm::get_dst_data()
467 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
470 - ptask should now work better with multicore hosts (FG#37).
471 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
472 - Internal cleanups and refactoring around models, to prepare for programmatic
474 - Modernize our tests of the performance models
478 - New section: Configuring the logs from the command line.
479 - New section: Programming API of logs.
480 - New section: Getting realistic results
481 - New section: Examples of SMPI Usage
482 - New part: SimGrid's Internals, with a section on Design goals
483 - Move back to the classical Breathe tool chain, and fix many small glitches.
486 - Handle leak checking: SMPI will now display a message if MPI handles are still
487 present at the end of a simulation. Option smpi/list-leaks:n will display the n
488 first leaked pointers and their types.
489 - smpi/display-allocs option : Display Memory allocations analysis, with hints
490 about the largest allocations. This only works for allocations with malloc/calloc
491 and not compiled with SMPI_NO_OVERRIDE_MALLOC
492 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
493 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
494 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
495 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
496 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
497 - Replay now opens and parses the trace file only once for all ranks when possible.
499 - better handling of MPI_PROC_NULL
500 - MPI_Group_incl was broken in some cases
501 - PAPI tracing now fixed
502 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
503 - SMPI will now fail properly when receive size < sent size, instead of
504 just setting MPI_ERR_TRUNCATE and keep going.
505 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
506 - Remove useless build patches and split build/run patches on the page
507 - Build in parallel when possible
508 - Test on yum-based systems
509 - New nightly tested Proxy-apps:
510 - OpenMPI mpi-test-suite
511 - Coral MCB, LSMS, Qbox, CLOMP
512 - ECP/Mantevo Cloverleaf3D, Ember
513 - Codevault bhtree_mpi
516 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
517 - Start to overhaul the implementation to modernize it (still ongoing).
520 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
521 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
522 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
523 - Deprecate xbt_dynar_to_array().
524 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
526 C binding and interface:
527 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
528 - Constness requirement has been lowered for parameter argv of sg_actor_start()
529 and sg_actor_create() It's now (char* const*). The previous version is still
530 available as sg_actor_start_() and sg_actor_create_().
532 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
533 (FG: issues on Framagit; GH: issues on GitHub)
534 - FG#37: Parallel tasks are limited to 1 core per host
535 - FG#62: Running "smpirun -replay" on large networks
536 - FG!46: Fix a few potential memory leaks in SMPI colls
537 - GH#343: Wrong consumed energy when an exec timeouts
539 ----------------------------------------------------------------------------
541 SimGrid (3.26) December 15. 2020
543 The Release release (the French lockdown was eased today).
545 Important user-visible changes:
546 - SimGrid now requires a compiler with C++14 support.
547 Sibling projects should upgrade their FindSimgrid.cmake
548 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
549 several users had difficulties to understand issues when using high bandwidth or
550 small latency events. The new value was already the default for SMPI and
551 should not cause too much performance hit. It can change some simulation
552 timings if simulation was skipping a lot of small events (which was actually
553 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
554 - On some network models (L07, CM02, LV08), an implicit loopback link is created
555 for local communication. Default bandwidth was 500MBps and latency was 15µs,
556 which was rather bad. Default values have been changed to 10GBps and 0 latency.
557 Options network/loopback-lat and network/loopback-bw have been added to adapt
558 these values if needed. Network and routing models already providing separate
559 loopbacks (clusters and such) are not impacted by this change.
560 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
561 metadata that are added to each MPI message by MPI libraries (tag,
562 src/dest, and communicator information). This may change simulation times
563 for very small message sizes, but should be more realistic.
566 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
567 - New plugins: wifi_energy and link_load. See the associated examples.
568 - New examples: wifi networks, and communication suspend.
571 - LTO is now enabled for Intel/clang compilers.
572 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
573 Setting it to "auto" will use all available cores, while setting it to n will
574 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
575 - Remove obsolete runtime option 'contexts/parallel-threshold'.
576 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
577 has been kept for compatibility.
578 - Finally remove obsolete snake_case() aliases for runtime options.
579 - Further improve the documentation.
582 - the on_exit() of each actor is also executed when the simulation deadlocks.
583 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
584 An example is provided for s4u::Comm but it should work with Exec and Io.
587 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
588 https://framagit.org/simgrid/SMPI-proxy-apps
589 - MPI/IO: supports names starting with "./"
590 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
592 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
593 building shared libraries, this may cause issues, so environment variable
594 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
595 - most temporary files should now be created in /tmp dir (or equivalent).
596 If this one does not allow execution of code (noexec flag), this may cause issues.
597 Please use another tmp directory (using TMPDIR or equivalent system variable)
601 - A large refactoring is ongoing, but this should be transparent for users so far.
604 - Drop xbt_str_split().
606 C binding and interface:
607 - The return type of the sg_actor_on_exit() callbacks is now 'void'
608 (int value was previously ignored)
609 - Many C functions were renamed to follow a common naming style
610 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
611 names are kept until version 3.30.
612 - Many MSG tests were converted to the new S4U's interface in C, that
613 was extended for that.
615 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
616 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
617 - FG#41: Add sg_actor_create C interface
618 - FG#43: xbt::random needs some care
619 - FG#48: The Impossible Did Happen (yet again)
620 - FG#50: Suspending an actor executed at the current timestamp fails
621 - FG#52: Zero-seconds timeout : "That's in the past already"
622 - FG#53: Crash while using ns-3 network model
623 - FG#54: How to suspend a comm?
624 - FG!22: Stochastic Profiles
625 - FG!24: Documentation and fix for xbt/random
626 - FG!35: Add a modeling hint for parallel links in doc
627 - FG!36: [xbt/random] Read/Write the state of the RNG
628 - FG!42: Link cumulated load plugin
629 - FG!44: Test condition variable wait_until+notify_all
630 - GF#18137: Allow different stack sizes?
631 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
632 - GH#139: Allow pthread creation in SMPI
633 - GH#336: Packet-level simulation using SMPI?
634 - GH#345: Error in the LMM while migrating a VM
635 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
636 - GH#352: pip install / python setup.py install fail to find pybind11
637 - GH!337: Fix link_energy plugin for wifi platforms
638 - GH!339: Add Mailbox set_receiver method to python binding
639 - GH!344: Cast hugepages macros parameters to int64
641 ----------------------------------------------------------------------------
643 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
645 The Palindrom Day Release.
647 Important user-visible changes:
648 - Improve the Python usability (stability and documentation).
649 - A nasty synchronization bug was ironed out, see also below.
650 - Python's doc was organized as a separate tree, now integrated with C++.
651 - C bindings of S4U were not part of the doc.
652 - The C++ doc was also improved as methods are now split by theme.
653 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
654 - This is mandatory to use the Java bindings.
655 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
656 - SimDAG++: Automatic dependencies on S4U activities (experimental)
657 - Some features are already implemented but not all of them
658 - Cannot block an activity until it's scheduled on a resource
659 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
660 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
663 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
664 - Drop support for Viva/Triva (old visualization tools), hence removed
665 TRACE_get_node_types() and TRACE_get_edge_types() functions.
668 - Require pybind11 version 2.4.
669 - Greatly improve locking during context switches with Python.
670 - Declaring Python callbacks to this_actor::on_exit is now working.
671 - No more tests are failing currently, even on win32.
674 - Actor: Merge signals on_migration_start/end into on_host_change
675 - Actor: Rename migrate() into set_host()
676 - Disk: Allow users to get the read and write nominal bandwidth values
677 - Exec: Implement wait_for(timeout)
678 - Io: Implement wait_for(timeout)
681 - Parse errors now raise a simgrid::ParseError that you may want to catch.
684 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
685 been renamed to fix a spelling error. As usual, the previous definitions are
686 kept with a deprecation warning. However, to avoid runtime errors, it is now
687 forbidden to override these deprecated methods in a derived class. Please use
688 the new names immediately if you need to override them.
691 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
692 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
693 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
694 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
695 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
696 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
697 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
698 - New module: random, providing classical random numbers generators.
701 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
702 allocations' size above which they are considered "shared" by default (as if
703 they were performed through SMPI_SHARED_MALLOC macros).
704 Default = 0 = disabled feature.
705 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
706 This can cause some troubles if codes are already overriding these. If this
707 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
708 help, but will make this feature unusable.
709 - Input check is now performed much more consistently and outputs precise
710 warnings for each error found
711 - New supported calls : MPI_File_set_view, MPI_File_get_view
712 - MPI I/O now only supports disks, as storage is discontinued.
713 - MPI I/O allows opening files without specifying absolute path
714 - Note : SMPI configuration options are now initialized by smpi_init_options()
715 call, which should be called if SMPI is not being used through smpirun.
717 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
718 - FG#9: Python bindings crashing
719 - FG#39: Missing s4u::Comm::wait_any_for example
720 - FG#42: Add support for ThreadSanitizer (TSan)
721 - FG!19: Removing RngStream
722 - FG!20: A module for RNG calls
723 - FG!21: Choice between ad-hoc and standard distributions implementations
725 - FG!25: Fix link in Mutex doc
726 - FG!26: Fix links in SMPI interface doc
727 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
728 - GH#305: Unscheduled tasks are still excuted
729 - GH#313: smpirun: manual outdated w.r.t. --help
730 - GH#321: [S4U] Get task remaining work ratio
731 - GH#323: Crash when an actor turn off his physical host
732 - GH#335: Missing links on dragonfly example svg
733 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
734 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
736 ----------------------------------------------------------------------------
738 SimGrid (3.24) October 9. 2019
740 The Clean Disk Release.
743 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
744 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
745 thus becomes the default resource to execute I/O operations. Examples
746 and the file system plugin have been modified to use disks instead of
747 storage elements. Storage can still be used but is doomed to disappear.
748 Users are thus adviced to update their simulators accordingly.
749 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
750 for consistency with pthread_barrier_wait()
751 - Host::get_englobing_zone() returns the englobing netzone
752 - Actor::on_destruction is now called in the destructor
753 Actor::on_termination new signal called when the actor terminates
755 - Global signals are now part of the Engine:
756 - on_platform_creation: after config settings, before the XML parsing
757 - on_platform_created: right after the XML parsing
758 - on_time_advance: each time the clock advances
759 - on_simulation_end: after simulation, before cleanups
760 - on_deadlock: as the name implies.
762 - sg_{actor,host,link}_{data,data_set}() now all exist.
763 Use them to attach user data to the object and retrieve it.
766 - Introduce an experimental Wifi model. It sounds reasonable
767 according to the state of the art, but it still has to be properly
768 validated, at least against ns-3 if not against reality.
769 - Improved the usability of ns-3. Several bugs were ironed out.
770 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
771 It is now expressed as 'idle:epsilon:allCores' to properly model the
772 consumption of non-whole tasks on mono-core hosts. More info in the doc.
775 - convert a new set of functions to the S4U C interface and move the old MSG
776 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
779 - Fortran bindings for DVFS have been removed.
780 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
781 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
782 - SMPI can now be selected by cmake's find_module(MPI) with
783 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
784 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
785 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
786 - trace-call-location can be used with TI traces, and replayed, and is compatible
787 with smpi/comp-adjustment-file.
788 - sleep events are now correctly traced/replayed.
789 - Default for trace-call-location is now to use file names and not full paths.
790 To revert to previous behavior (in case of collision of filenames), option
791 "smpi/trace-call-use-absolute-path" can be set to yes.
794 - Use the included xxHash as an hash implem when C++14 is usable.
795 - Option model-checker/hash was removed. This is always activated now.
796 - New option smpi/buffering controls the MPI buffering in MC mode.
797 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
798 This is useful to check for MPI compliance.
801 - New section on plugins: how to define a new one, and existing ones.
804 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
805 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
808 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
809 and <mount> tags. The rationale is finish to decouple what should be handle
810 by SimGrid kernel (i.e., resources and activities) from plugins built in
811 user space (e.g., the file system plugin). This new tag currently accepts
812 only three attributes (id, read_bw, and write_bw). All the other information
813 that was declared with the storage related tags now has to be expressed as
814 properties. An example of platform using this new tag is available at
815 examples/platforms/hosts_with_disks.xml
818 - 'expect signal' can now accept more than one potential signal.
820 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
821 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
822 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
823 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
824 - FG#31: per-actor data
825 - FG#34: SG_BARRIER_SERIAL_THREAD?
826 - FG#35: model-checker does not like buster-produced binaries
827 - FG!13: MC: complete workaround in the error msg seen on modern systems
828 - FG!15: execute_flops now logs compute
829 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
830 - FG!17: ns-3: unblock the right number of communications + others issues
831 - FG!18: Improving the performance of the ns-3 bindings
832 - GH#189: Energy consumption of parallel tasks
833 - GH#219: Error in the throughput of TCP transfer
834 - GH!330: Minor polishing of English: Configuring_Simgrid
835 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
836 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
837 - GH!334: Rework energy plugin, again
839 ----------------------------------------------------------------------------
841 SimGrid (3.23.2) July 8. 2019
844 - Nicer introduction page.
845 - Migrate the "Deploy your application" page to the new doc.
846 - Move Java as a subtree of MSG.
849 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
852 - Drop xbt_dynar_sort_strings().
855 - Really fix FG#26: Turning off a link should raise NetworkFailureException
856 - FG#27: Wrong exception thrown to wait_any when link is turned off
857 - GH#328: Java: Canceling multiple tasks in a single vm/host
859 ----------------------------------------------------------------------------
861 SimGrid (3.23) June 25. 2019
863 The Exotic Solstice Release.
866 - SunOS and Haiku OS support. Because exotic platforms are fun.
867 - Stop setting random seed with srand48() at initialization.
868 - Use addr2line as a fallback for stacktraces when backtrace is not available.
869 - Build option -Denable_documentation is now OFF by default.
870 - Network model 'NS3' was renamed into 'ns-3'.
873 - Simgrid can now hopefully be installed with pip.
876 - wait_any can now be used for asynchronous executions too.
879 - New log appenders: stdout and stderr. Use stdout for xbt_help.
880 - Drop xbt_dict_dump.
883 - 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
884 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
885 - MPI interface is now const correct for input parameters
886 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
887 - Fortran bindings for async collectives.
888 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
891 - Remove option 'model-check/record': Paths are recorded in any cases now.
892 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
893 always sparse. That's an immense gain in memory, and can even be
894 faster because of cache effects. So there is no need to clutter the
895 code to allow the user to go for the unefficient mode.
898 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
899 models proved to be more accurate than these old experiments.
901 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
902 - FG#1: Broken link in error messages
903 - FG#2: missing installation documentation
904 - FG#3: missing documentation in smpirun
905 - FG#6: Python bindings not available on PyPI
906 - FG#7: simple cmake call requires doxygen
907 - FG#8: make python bindings an optional dependency
908 - FG#10: Can not use MSG_process_set_data from SMPI any more
909 - FG#11: Auto-restart actors forget their on_exit behavior
910 - FG#12: -Denable_lto=OFF doesn't disable LTO
911 - FG#13: Installs unstripped file 'bin/graphicator'
912 - FG#14: Installs the empty directory 'doc/simgrid/html'
913 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
914 - FG#17: Dead link in doc (pls_ns3)
915 - FG#20: 'tesh --help' should return 0
916 - FG#21: Documentation link on http://simgrid.org/ broken
917 - FG#22: Debian installation instruction are broken
918 - FG#26: Turning off a link should raise NetworkFailureException exceptions
919 - GH#133: Java: a process can run on a VM even if its host is off
920 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
921 - GH#326: Valgrind-detected error for join() when energy plugin is activated
923 ----------------------------------------------------------------------------
925 SimGrid (3.22) April 1. 2019
927 The Easter Christmas Release (this one is somewhat late).
930 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
931 - Not complete yet: asynchronous activities (amongst others) are still missing
932 - Still ongoing: the interface may change in the future. We need more
933 testers! Please report any glitches.
934 - No new project using Java should start now. Please switch to Python.
937 - Some of the internal cleanups may lead to speed improvements:
938 - The hard limitation on the amount of simulated actors with
939 Java+Mac was removed. Now, the available memory is the only limit.
940 - Our refcounting was tidyied, leading to 10% speedups in some cases.
941 - We are still working on making our code robust to the actor kills
942 and hosts' churn. Things are improving, but it's not perfect yet.
943 - Replaced our own code to display a backtrace (that was forking addr2line)
944 with the Boost.Stacktrace library.
945 You won't see your backtraces without this optional dependency.
946 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
947 - Stop setting random seed with srand() at initialization.
950 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
951 XML file version remains 4.2 since old files are still compatible.
954 - Process termination which was broken at version 3.21 has been repaired.
955 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
956 - Hide the examples into examples/deprecated. New users should use Python.
959 - Drop MSG_process_create_from_stdfunc() from the API.
960 This C++ function was a pimple in the C API, made necessary at some
961 point by the Java bindings. This is fixed now.
962 - Hide the examples into examples/deprecated. New users should use S4U.
963 - MSG_process_create and MSG_process_attach now crash if the host used
964 to run this process is off.
965 - Fix the protype of MSG_process_on_exit()
966 Now use (int,void*) callbacks instead of (void*,void*) ones.
967 The implementation was ways too messy to actually work, I guess.
970 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
971 too often and provide more accurate timings.
972 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
973 - MPI_Alltoallw support
974 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
975 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
976 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
977 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
979 - Basic implementation of generalized requests (SMPI doesn't
980 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
983 - Drop sg_cmdline. Please use xbt_cmdline instead.
984 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
985 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
986 OsSemaphore is implemented in a portable way with C++11 threads.
987 This should allow much more threads to be created at the same time,
988 allowing Mac and Java users to simulate many more actors.
989 - Implement the 'thread' factory with std::thread instead of xbt ones.
990 It is not possible to set the stack size with threads anymore, but
991 -fsplit-stack is the way to go nowadays when using threads.
992 - Drop the xbt_os_thread_t module (now unused)
993 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
994 - Drop xbt_str_join_array().
995 - Drop cunit, use Catch2 instead.
998 - Many cleanups in the kernel::activity namespace. This was long
999 overdue, and shall open the path to many future endeavors.
1002 - #132: Java: a process can not shut down its own host
1003 - #220: S4U: detached send causes exception if sender terminates soon after sending
1004 - #261: Document the parameters of parallel execution's constructor
1005 - #300: [s4u] BarrierPtr is missing
1006 - #314: SMPI args internal cleanup
1007 - #316: Fix a bug related to the CPU utilization of multi-core VM
1008 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
1009 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
1010 - #325: Turning off a host has different behavior on sleeping actors and computing actors
1012 ----------------------------------------------------------------------------
1014 SimGrid (3.21) October 3. 2018
1016 The Restarting Documentation (TRD) Release.
1019 - Convert the user manual to Sphinx for improved usability.
1020 Unstable version now hosted on https://simgrid.frama.io/simgrid
1021 Many glitches remain; stay tuned.
1022 - Restrict the Doxygen reference API to the installed header files.
1023 (documenting all internals this way is overkill + doxygen scales badly)
1024 - New tutorial on S4U (highly inspired from the old MSG one, but with
1025 a git repository to fork as a starting point, and a docker image)
1026 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
1027 - The Developper manual is still to be converted (not compiled until then)
1028 - Some parts are still missing in this conversion (such as the
1029 platform and deployment sections) while others were blindly converted
1030 and would need more love (such as the configuration flags).
1031 Things will certainly further improve in the future releases.
1034 - Fully reimplement auto_restart mechanism that was utterly broken
1035 (fix #22, #131 and #281 that were all very related)
1036 - Implement semaphores (fix #303 and #291)
1037 - When creating an actor from a function and its parameters,
1038 move-only parameters are not allowed anymore, as it would prevent
1039 the actor to be restartable if its parameters are consumed on run.
1040 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
1041 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
1042 IO operations that can be started, waited for, or canceled as a regular
1044 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
1048 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
1049 - New host variable: 'core_count'
1052 - Remove xbt_os_thread_specific features
1053 - Remove portability wrapper to condition variables
1054 - Remove xbt_os_thread_yield()
1057 - MPICH collective selector now mimics MPICH 3.3b
1058 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
1059 - MPI_Init(NULL,NULL) is now allowed
1060 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
1061 - Command-line arguments should now be handled properly in Fortran simulations
1062 - MPI Attributes and related callbacks should work in Fortran
1063 - Apps using C + Fortran should now work
1064 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
1065 - Support MPI_Error_String
1068 - Due to an internal bug, Msg.run() must now be your last line.
1069 We hope to fix it in a future release, and we are sorry for the inconvenience.
1072 - #22: Process autorestart seem to only work with CAS01 cpus
1073 - #56: Feature request: dynamic SMPI replay
1074 - #93: simgrid should not eat --help
1075 - #111: How to change the loopback link characteristics?
1076 - #116: Platform cluster radical
1077 - #117: Platform cluster/host order
1078 - #129: Replay_multiple on steroid
1079 - #131: Java : autostart does not work
1080 - #143: Setting a breakpoint at a given time
1081 - #235: Network energy models should be integrated
1082 - #264: Add ptask L07 resource tracing
1083 - #271: Dynamic and manual replay of SMPI traces
1084 - #279: Breakpoints option uses time deltas instead of absolute time
1085 - #281: Daemonized actor no longer daemonized after an auto-restart
1086 - #282: TIT convertor erase traces when using absolute path in the trace list file
1087 - #285: segfault when a s4u actor kills itself with msg process tracing activated
1088 - #286: Pajé traces are not exposing the number of cores
1089 - #287: Command-line parsing should stop when encountering '--'
1090 - #288: MPI_Init(NULL, NULL)
1091 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
1092 - #290: Method to check if a mailbox has a message ready to be consumed
1093 - #291: [s4u] semaphores do not seem available yet
1094 - #294: Alltoall collective implementation of mpich
1095 - #295: Starting a Host as OFF is not supported
1096 - #296: DTD too permissive
1097 - #299: Add a s4u barrier example
1099 ----------------------------------------------------------------------------
1101 SimGrid (3.20) Released June 24. 2018
1103 The proxy snake_case() release.
1106 - The whole API is now using snake_case() instead of mixing it with camelCase.
1107 Compatibility wrappers in place for one year.
1108 - We took the opportunity to increase the consistency of the API.
1109 Please report any remaining glitches.
1110 - New function simgrid::s4u::Host::get_actor_count:
1111 Returns the number of actors running on a specific host.
1114 - SMPI is now tested with ~45 proxy apps from various sources, with none or
1115 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
1116 - Replay: The replay file has been re-written in C++.
1117 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
1118 supported. They are stored in the trace and used when replayed.
1119 - Basic support of MPI_Cancel. Robustness not guaranteed.
1120 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
1121 (only for MPI_COMM_TYPE_SHARED).
1122 - New option: smpi/privatize-libs, to add external shared libs to be privatized
1123 by SMPI. They will be copied locally and loaded separately by each process.
1124 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
1125 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
1126 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
1127 with internal index optind with multiple processes. Only works if MPI_Init has
1128 already been called.
1129 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
1130 syntax without preprocessing tricks.
1133 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
1134 behavior tracing is no longer limited to MSG
1135 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
1136 this means that traces now need to be updated or re-obtained; both Irecv
1137 and Isend lines in your traces have now in total 5 mandatory fields:
1138 <rankid> <command> <to/from rankid> <tag> <size>
1139 To update your traces, it suffices to add a 0 for the tag here.
1140 - TIT now also supports waiting for a distinct request via MPI_Wait.
1141 Wait/Test now wait for a specific request, not just the last one that was
1142 issued. This unfortunately means another update, because we need to
1143 identify which request you want to wait for. We do this via the
1144 triplet (sender, receiver, tag), which needs to be added:
1145 <rankid> <command> <sender> <receiver> <tag>
1146 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
1147 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
1150 - The deprecation of MSG is ongoing (but this should not impact you).
1151 Many MSG functions are now simple wrappers on the C API of S4U. If
1152 you wish to convert your code to S4U, find the S4U counterparts of
1153 your MSG calls in src/msg/msg_legacy.cpp.
1154 - MSG can still be used, but won't evolve anymore.
1157 - Allow to run the Link energy plugin from the command line with
1158 --cfg=plugin:link_energy
1159 - Rename Energy plugin into host_energy
1160 - Rename Load plugin into host_load
1163 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
1164 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
1165 - Improve performance when handling timeouts of simix synchros.
1168 - Config: the C API is now deprecated (will be removed in 3.23), and
1169 the C++ API has been slightly improved.
1172 - Fix several build issues on OSX.
1173 - Move simgrid_config.h to simgrid/config.h (old header still working)
1176 - #143: Setting a breakpoint at a given time
1177 - #150: Inconsistent event names in SMPI replay
1178 - #258: daemonized actors hang after all non-daemonized actors have completed
1179 - #267: Linker error on unit_tmgr
1180 - #269: SMPI: tracing of MPI_Wait/all/any broken
1181 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
1183 ----------------------------------------------------------------------------
1185 SimGrid (3.19.1) Released March 22 2018
1187 The Fixed ABI Release.
1189 As you may know, we are currently refactoring SimGrid in deep.
1190 Upcoming SimGrid4 will be really different from SimGrid3: modular,
1191 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
1193 Our goal is to smooth this transition, with backward compatibility
1194 and automatic update paths, while still progressing toward SimGrid4.
1196 SimGrid remains open during works: The last pure SimGrid3 release was
1197 v3.12 while all subsequent versions are usable alpha versions of
1198 SimGrid4: Existing interfaces remain unchanged, but the new S4U
1199 interface is budding and the internals are deeply reorganized.
1201 Since 2015, we work hard to reduce the changes to public APIs. When
1202 we need to rename a public library symbol in S4U, we let your
1203 compiler issue an explicative warning when you use the deprecated
1204 function. These messages remain for four releases, i.e. for one full
1205 year, before turning into an error. Starting with v3.15, your can
1206 also adapt to API changes with the SIMGRID_VERSION macro, that is
1207 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
1209 Starting with this v3.19.1, our commitment to reduce the changes to
1210 the public interfaces is extended from the API to the ABI: a program
1211 using only MSG or SimDag and compiled against a given version of
1212 simgrid can probably be used with a later version of SimGrid without
1213 recompilation. We will do our best... but don't expect too much of
1214 it, that's a really difficult goal during such profund refactoring.
1216 The difference between v3.19 and v3.19.1 is that the former was
1217 accidentally breaking the ABI of MSG, while the later is restoring
1220 S4U and kernel APIs will still evolve until SimGrid4, with one-year
1221 deprecation warnings as currently. In fact, cleaning up these
1222 interfaces and converting them to snake_case() is one release goal of
1223 v3.20. But don't worry, we are working to smooth this upgrade path.
1225 In summary, new projects should start with S4U to benefit of the
1226 future, but old MSG projects should still be usable with no change.
1228 ----------------------------------------------------------------------------
1230 SimGrid (3.19) Released March 20 2018
1232 The Moscovitly-cold Spring release.
1234 SMPI user-visible improvements:
1235 - Support of MPI_Type_create_subarray
1236 - Unimplemented MPI calls now abort (instead of issuing a warning)
1237 Some harmless calls around errhandlers still only issue a warning.
1238 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
1239 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
1240 is now possible (mmap privatization was refactored to this end).
1241 Adding new SMPI instances should also be technically doable now.
1243 SMPI internal cleanups:
1244 - Removed index notion from SMPI -- no more "getPid() - 1"!
1245 In tracing files, actors are called rank-<PROCESS-ID> but these are
1246 NOT the real ranks (tracing several communicators is not supported yet).
1247 - Deprecate smpi_process_index() (will be removed in 3.22)
1248 - Better testing on classical proxy apps, through an external project:
1249 https://github.com/simgrid/SMPI-proxy-apps/
1252 - Use a graphical TOC to make it easier to find the documentation you need
1253 - Revamp the MSG tutorial
1256 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
1257 - Remove parameter of MSG_process_killall().
1258 Resetting the PID was bogus anyway (several actors could have the same PID).
1261 - Execution->setHost() can be called after start(), to migrate it.
1262 - Comm::test_any() is now implemented.
1263 - s4u::Actor now has onCreation() and onDestruction() signals.
1264 - Install some kernel header files for the users' plugins and more.
1265 Warning, their API is really not stable yet.
1266 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1267 include/simgrid/kernel/routing/*Zone.hpp
1270 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1272 Build System and other cleanups:
1273 - Remove unused run-time parameter "tracing/onelink-only".
1274 - SimGrid now works with intel compilers (MC still buggy)
1275 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1276 - tesh kills subprocesses on timeout (but not on windows)
1279 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1282 - #194: Feature request: simgrid::s4u::Comm::test_any()
1283 - #245: migrating an actor does not migrate its execution
1284 - #253: Feature Request: expose clusters as objects
1285 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1286 - #255: Tesh broken on Windows
1287 - #256: Modernize FindSimGrid.cmake
1288 - #257: Fix (ab)use of CMake install
1290 ----------------------------------------------------------------------------
1292 SimGrid (3.18) Released December 24 2017
1294 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1296 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1297 - Support for MSG should not be removed from SimGrid before 2020,
1298 but future evolutions will be limited to the integration of
1299 user-provided patches.
1302 - s4u::onDeadlock() signal, run before stopping the simulation.
1303 - s4u::Exec: asynchronous executions (abstraction of background
1305 - s4u::Host->getLoad() returns the achieved speed in flops/s
1308 - Link::name() is deprecated and will be removed in v3.21, use
1309 Link::getCname() instead.
1310 - Mailbox::getName() changed to return a std::string, use
1311 Mailbox::getCname() instead to get a char*.
1312 - Storage::getName() changed to return a std::string, use
1313 Storage::getCname() instead to get a char*.
1314 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1315 to have both allocation and deallocation of the map in user space.
1318 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1319 The amount of flops remaining to do is a vector, not a scalar.
1320 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1321 name implies on both sequential and parallel tasks.
1322 - Both changes fix GitHub's #223 using PR #237 as a basis.
1323 Thanks Michael Mercier.
1324 - Most examples were converted to S4U and hidden elsewhere at tests.
1325 Rationale: we still want MSG to work; we want newcomers to use S4U.
1328 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1329 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1332 - Switch to the faster dlopen privatization mechanism by default
1333 - Documentation improvments
1336 - Live migration is getting moved to a plugin. Dirty page tracking is
1337 the first part of this plugin. This imply that VM migration is now
1338 only possible if one this function is called:
1339 - C/MSG: MSG_vm_live_migration_plugin_init()
1340 - C/C++: sg_vm_live_migration_plugin_init()
1341 - Java: Msg.liveMigrationInit()
1342 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1346 - Define class simgrid::xbt::Path to manage file names.
1347 - Removed unused functions:
1348 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1349 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1350 xbt_graph_export_graphviz()
1351 - xbt/str.h: xbt_str_join()
1352 - Remove unused datatypes:
1353 - xbt/heap.h: use std::priority_queue or boost::heap instead
1354 - xbt/swag.h: use boost::intrusive::list instead
1357 - New link_energy plugin for the consumption of the links.
1358 - All of the operations on files and storage contents have been
1359 packaged into a plugin (src/plugins/file_system). The current
1360 public interface can be found in
1361 include/simgrid/plugins/file_system.h
1362 To use these functions you now have to initialize the plugin by
1363 calling MSG_storage_file_system_init() just after calling
1364 MSG_init() or sg_storage_file_system_init() just after creating
1368 - Remove the undocumented/untested tag <include>
1371 - Remove viva specific tracing as the tool is no longer maintained
1374 - #248: Kill of finished processes leads to segfault
1375 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1376 - #239: please implement signal s4u::onDeadlock()
1377 - #230: segfaults when exit() without run()
1378 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1379 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1380 - #222: Actor::kill() doesn't really kill and segfaults
1381 - #221: odd LMM warning when killing an actor
1382 - #120: Memory leak when the processes are forcefully killed
1384 ----------------------------------------------------------------------------
1386 SimGrid (3.17) Released October 8 2017
1388 The Drained Leaks release: (almost) no known leaks despite the tests.
1390 Even very long simulations will preserve your memory: our testsuite
1391 stresses SimGrid for over 45mn on fast machines for a coverage of
1392 over 80%, and there is only one single known leak, of about 4kb.
1396 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1397 handler may be given to free resources if the comm cannot be completed.
1398 - this_actor::send(mailbox) is now mailbox->put()
1399 - New: simgrid::s4u::Comm::wait_all()
1400 - New: Host.execute() for remote executions.
1403 - Unused option network/sender-gap is removed.
1406 - Deprecate MSG_task_isend_with_matching(): This unused feature
1407 really complicates our internals. Will be removed in v3.20.
1410 - Improved context termination. It is now possible to free resources, even
1411 when a process is forcibly killed.
1412 - ContextBoost: add support for Boost versions above 1.61.
1415 - Bring back run-time option --cfg=exception/cutpath to remove exception
1417 - Removed unused functions:
1418 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1419 xbt_str_rtrim(), xbt_str_trim().
1420 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1423 - Removed header files obsolete since SimGrid 3.12:
1424 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1426 - #3: SD_exit should be made optional
1427 - #120: Memory leak when the processes are forcefully killed
1428 - #159: Threading test regression in Actor refcounting
1429 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1430 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1431 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1432 - #191: VM migration and pstate
1433 - #192: Updating the energy consumptions of all hosts crashes with VMs
1434 - #195: All actors have PID=0 in the logs
1435 - #204: Sometimes segfault with thread contexts and mmap privatization
1436 - #222: Actor::kill() doesn't really kill and segfaults
1437 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1439 ----------------------------------------------------------------------------
1441 SimGrid (3.16) Released June 22. 2017.
1443 The Blooming Spring Release: developments are budding.
1446 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1447 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1448 - Java JAR file should be finally fully working on Mac OSX too.
1450 XML platforms: Switch to platform v4.1 format.
1451 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1452 - <zone> can be used as a synonym for the now deprecated <as>
1453 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1454 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1455 - <actor> can be used as a synonym for the now deprecated <process>
1456 - state_file and avail_file periodicity is now easier to express
1457 (check the documentation) the old behavior should still work.
1458 * Storage: not backward compatible, but it seems that nobody ever used it
1459 - Remove attribute 'content_type' of <storage_type>: was never used
1460 - Make attribute 'model' of <storage_type> optional: for future usage
1461 - Remove Bconnection model property: was never (in)validated. Replaced by
1462 the maximum of the read and write bandwidth as a resource constraint.
1465 - New and Backwards Compatibility break:
1466 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1467 When one wants to get the list of tasks whose states have changed during a
1468 simulation round, s/he has to allocate and free a dynar and use it as
1469 argument to this function. The former SD_simulate (double how_long)
1473 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1474 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1477 - New function to update the consumption of all hosts at once.
1478 - Fix the model for multi-core hosts, linear in the amount of busy
1479 cores with an abnormality for pIdle (see paper).
1480 BUG: Still not satisfactory for ptask on mono-cores.
1483 - The netzone are now available from the MSG API.
1484 The old names still work, but are now deprecated.
1487 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1488 - New option: smpi/keep-temps to not cleanup temp files
1489 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1490 mallocs mode is used, allows to change the size of the fake file used
1491 (default 1MB), to potentially limit the number of mappings for large runs.
1492 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1493 - Fortran ifort and flang compilers support
1494 - New RMA calls supported (experimental):
1495 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1496 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1497 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1498 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1499 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1500 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1503 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1504 - Behind the scene, ActivityImpl are now automatically refcounted.
1505 Removing this nasty bottleneck will greatly simplify our internals.
1508 - Replay: New function xbt_replay_action_get():
1509 Retrieve the function previously associated to an event type.
1510 - DROPPED FUNCTION: xbt_str_varsubst()
1511 - DROPPED MODULE: strbuff. We don't need it anymore.
1512 - DROPPED MODULE: matrix. We don't need it anymore.
1513 - DROPPED MODULE: lib. We don't need it anymore.
1515 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1517 SimGrid (3.15) Released March 22. 2017
1519 The Spring Release: continuous integration servers become green
1520 We fixed even the transient bugs on all target architectures:
1521 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1522 Windows, FreeBSD, NetBSD.
1524 - Rename NetCards to NetPoints.
1525 This was intended to help NS3 users, but that's not a netcard.
1526 That's a point in the routing algorithm, let's avoid wrong simplifications.
1529 - New: MSG_process_yield(). Stop and yield to other processes.
1530 - New: MSG_process_daemon(). Daemon processes are automatically killed
1531 when the last non-daemon process terminates
1532 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1533 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1534 to make room for the upcoming network energy plugin.
1535 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1538 - Ensure that an actor can kill itself with Process::exit()
1539 - Kill the obscure NativeException. Nobody want to survive the issues
1540 it denotes, so use JniException that is a RuntimeException (not to
1541 be caught explicitly).
1542 - Partial bug fix in initialization. SimGrid flags on command line were
1543 consumed at C level but stayed in the original Java String[] args.
1544 This could mess users' args[i] if SG flags were not put at the end of
1546 The SimGrid flags are now removed from the Java arguments. However,
1547 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1548 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1549 It might be if you have --log or --cfg flags in the command line.
1550 - Fix numerous memleaks all around the place. In particular, around VMs.
1554 - s4u::onPlatformCreated: right before the simulation starts
1555 - s4u::onSimulationEnd: right after the main simulation loop
1556 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1557 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1558 event from the availability_file changes the avail speed.
1559 - Links are now usable from s4u
1560 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1561 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1562 - New examples: The conversion of MSG examples to S4U has begun
1563 - Actors: create, kill, migration, and suspend
1564 - Applications: master-worker and token-ring
1565 - Action replay: communications and storage
1566 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1569 - Backwards Compatibility breaks
1570 - The SD_route_get_list and SD_route_get_size functions have been removed.
1571 They are replaced by the unique
1572 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1573 The route from src to dst is built in the links dynar whose size can
1574 be obtained with xbt_dynar_length.
1575 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1576 sg_host_route_bandwidth and sg_host_route_latency.
1577 Macros ensure the backwards compatibility, but you should fix your code
1580 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1581 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1582 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1583 We don't support 100% of it, but it should be enough. Contact us if not.
1584 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1585 naive and probably inaccurate)
1586 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1587 mapped onto a unique small file using some system magic.
1588 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1589 - Fortran: cleanups, fixes, support of user-added operations
1590 - RMA: MPI_Accumulate are applied in correct order.
1591 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1592 - Better support for MPI_IN_PLACE
1593 - Support for MPI_Win attrs and keyvals.
1594 - Support MPI_Comm_group_create, MPI_Type_size_x
1595 - MPI ops should be properly applied to all allowed types
1596 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1597 algorithm by a non failing one from mpich.
1600 - Kill the fifo data container: we don't use it anymore.
1603 - A <cluster> can now be created with different speed values to represent pstates.
1604 see examples/platforms/energy_cluster.xml for an example.
1606 SimGrid (3.14.159) Released December 28. 2016
1608 The previous release was only a crude approximation of a Pi release;
1609 this one should be a bit better.
1611 - Revert a last minute change that broke on Mac OSX.
1612 - Fix the changelog and NEWS files.
1613 - Improve the documentation, in particular of the routing module.
1614 - Integrate some patches flying around in Debian.
1616 SimGrid (3.14) Released December 24. 2016
1619 * The whole documentation was reworked and reorganized. There is
1620 still much room for improvement here, but we're on it.
1623 * We now need python3 for our testing infrastructure.
1624 * Model checking needs libevent
1627 * Now works on FreeBSD
1630 * Privatization now works on FreeBSD
1632 * Privatization is now activated by default in smpirun
1633 You can turn it off with -no-privatize if something goes wrong.
1635 * Call-location tracing for SMPI.
1636 You can add the exact location (filename / linenumber) of an MPI call to
1637 your trace files and slow down or speed up the simulation between two
1638 consecutive calls by using an adjustment file (see the documentation).
1640 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1641 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1642 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1643 values were ignored in some cases. The timings of these functions can now
1644 be significantly different.
1646 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1647 smpi/running-power is renamed to smpi/host-speed
1649 * smpi/grow-injected-times option to enable or disable multiplication of the
1650 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1651 make simulation less precise (but also much faster).
1653 * smpirun script should be (much) faster for large deployments.
1655 * SMPI tracing: fixed issue with poor matching of send/receives.
1657 * Replay: Fix broken waitall
1659 New functions and features
1660 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1662 Dropped / renamed functions and features
1663 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1664 - MSG_mailbox_is_empty() -> Mailbox::empty()
1665 - MSG_mailbox_front() -> Mailbox::front()
1666 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1667 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1668 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1669 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1670 * Don't pass the free_f parameter to property related functions:
1671 - MSG_host_set_property_value()
1672 - MSG_as_router_set_property_value()
1673 - MSG_storage_set_property_value()
1674 * VM properties. Since msg_vm_t are msg_host_t, just use
1675 MSG_host_get_property_value() and friends
1676 * VM I/O related things:
1677 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1678 - Unimplemented save/restore methods
1679 * MSG_as_router_get_property_value() was redundant with
1680 MSG_environment_as_get_property_value().
1681 - Removed MSG_as_router_*propert*() functions
1682 - Added MSG_environment_as_set_property_value() for consistency
1683 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1684 Well, they are still there for now, but deprecated with a warning.
1685 Please switch to xbt_dict_new_homogeneous() before this is removed
1687 * Task affinity. Its intended behavior (that was very badly tested
1688 and probably not really working) was deceiving what most users
1689 would have hoped here.
1690 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1691 * Option network/coordinates is now useless and should be dropped.
1694 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1697 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1698 * Added Dragonfly topology support
1701 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1702 EiBps, EBps, Eibps, Ebps for bandwidth.
1703 They may become useful to some lucky ones.
1706 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1707 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1708 * ex/async_waitAll: new example, on asynchronous communications
1711 * Memory usage should be decreased for simulations with a large number
1712 of processes. This also helps for SMPI.
1714 SimGrid (3.13) stable; urgency=low
1716 The Half Release, a.k.a. the Zealous Easter Trim.
1718 - We removed half of the lines, that were mostly experimental cruft.
1719 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1720 (not counting blanks and comments -- according to openhub.net)
1721 - The internals are now compiled in C++ (and will soon be clean C++)
1722 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1723 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1725 * Backwards Compatibility breaks
1726 - Removed Lua simulation bindings (switch to C or Java for that).
1727 Lua can still be used to describe platforms
1728 - Removed Java kernel plug-ins.
1729 Will be reintroduced after the ongoing major internals reorg.
1731 - the following functions were removed.
1732 They were too specific and should be reimplemented in a generic
1733 way, with filter function.
1734 - MSG_task_listen_from_host
1735 - MSG_mailbox_get_count_host_waiting_tasks
1736 - MSG_mailbox_put_with_timeout was removed.
1737 Please use MSG_task_send_with_timeout instead.
1739 - the SD_application_reinit function was removed. It has been a noop for a while.
1740 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1741 replaced by a more flexible mechanism.
1742 - The following functions thus do not exist anymore
1743 - SD_workstation_get_access_mode
1744 - SD_workstation_set_access_mode
1745 - SD_workstation_get_current_task
1746 - Basic estimation functions have been removed but can easily be replaced
1747 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1748 - SD_workstation_get_computation_time => amount / sg_host_speed()
1750 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1751 Use VM.getSpeed()*load/100 for the legacy behavior.
1753 - option enable_tracing was removed. It was not doing anything for a while.
1754 - In the ModelChecker:
1755 - the model-checker now ptraces the model-checked process which means
1756 you cannot use a debugger on the latter anymore (we might make this
1757 optional in the feature);
1758 - removed soft-dirty page tracking;
1759 - remove model-checked side snapshot management,
1760 MC_snapshot() and MC_compare_snapshot();
1761 - keep the MC_cut() function as a stub (it was not really working
1762 in the previous release).
1765 * All options are consistently in kebab-case. Old names are kept as alias.
1768 * Switch to platform v4 format.
1769 - Rename from 'power' to 'speed' the attributes describing the amount of
1770 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1771 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1772 - In <host> and <link>, attributes availability and state are gone.
1773 It was redundant with state and availability traces, and with peak values.
1774 - In <cluster>, attributes availability_file and state_file are gone.
1775 It was too complex and unused.
1776 - Kill <gpu>. Was not doing anything.
1777 - The DOCTYPE points to the right URL:
1778 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1779 (the file at this address now documents the changelog since its v1)
1780 - A warning is emitted for unit-less values (they are still accepted).
1781 - speed. Default: 'f' or 'flops'. Also defined:
1782 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1783 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1784 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1785 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1786 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1787 - latency. Default: 's' second. Also defined:
1788 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1790 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1791 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1794 * s4u::Host is now the preferred public interface to the Host features.
1795 sg_host_* functions are C bindings to the exact same behavior
1796 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1799 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1800 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1801 They were deprecated since a few years, and probably did not even compile anymore.
1804 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1805 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1806 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1807 encouraged to update their codes. A list of the modified functions can be found at the end of
1808 include/simgrid/simdag.h
1811 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1812 the lambda or closure passed as a parameter will run in kernel mode.
1813 All the callback functions should be rewritten to that interface at some point.
1816 * Reorganizing and cleaning the internals all around the place.
1819 * Remove old default barrier/bcast buggy algorithms (see #18407)
1820 * Various bug fixes to handle more codes
1821 * Remove the need for the --foreground option of smpirun (it is still
1822 accepted for backward compatibility).
1825 * Kill the setset data container: MC don't use it anymore.
1826 * Kill the queue data container: it made more sense with GRAS.
1827 * Kill the xbt_peer_t data type: it's useless without GRAS.
1828 * Kill rm_cb feature of config sets: it was never useful.
1829 * Kill graphxml parsing feature. It was not used.
1830 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1832 - xbt_dynar_sort_strings(), when the content is char*
1833 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1834 They throw exceptions on invalid input;
1835 * C++ support for declaring CLI flags (simgrid::config::Flag);
1836 * class for abstracting different signal backends (simgrid::xbt::signal).
1837 with no external dependencies (we need very simple signals).
1840 * refactoring and cleanup of the code;
1841 * ongoing process to cleanly separate the model-checking algorithms
1842 from the code model-checking support.
1844 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1846 SimGrid (3.12) stable; urgency=low
1848 The Facelift Release.
1851 * Require g++ v4.7 at least to not speak prehistorical C++.
1852 * Require Boost 1.48 (for signal2 component).
1853 * Java must be version 7 at least when activated.
1854 * Builds on Windows again (including Java bindings).
1855 * Tracing is now always enabled (no way to turn it out)
1856 * Remove GTNetS. It was not working anyway.
1857 * Various cleanups in the cmake scripts.
1858 * Move headers around to sort them out on installed systems:
1859 - instr/instr.h -> simgrid/instr.h
1860 - instr/jedule/* -> simgrid/jedule
1861 - simdag/datatypes.h was removed
1862 - simdag/simdag.h -> simgrid/simdag.h
1863 - msg/datatypes.h was removed
1864 - msg/msg.h -> simgrid/msg.h
1867 * Interface improvement:
1868 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1869 - Sanitize the interface in MSG_task_ module:
1870 - Merge two functions that were close enough but misleading:
1871 set_compute_duration(t) -> set_flops_amount(t)
1872 get_remaining_computation(t) -> get_flops_amount(t)
1873 - set_data_size(t) -> set_bytes_amount(t)
1874 get_data_size(t) -> get_bytes_amount(t)
1875 - Massive cleanups in the functions related to the energy
1876 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1877 - New: MSG_host_get_pstate()
1879 - msg/energy/onoff: switching hosts on and off
1881 * Interface improvement:
1882 - Rename Host.isAvail() to Host.isOn()
1883 - Rename Process.currentProcess() to Process.getCurrentProcess()
1884 - Rename Task.setDataSize() to Task.setBytesAmount()
1885 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1887 - #18874: Actually allows the GC to reclaim tasks
1891 - SIMIX_process_throw: raises an exception in a remote process
1892 * Refactoring: Separate sub-modules
1893 - libsmx: the public interface, as libc in a real system
1894 - popping: the strange dance that converts a user request into a kernel handling
1895 - smx_context_*: the virtualization mechanisms that embed the user code
1896 - smx_*: the handling of each simcalls
1897 * Interface cleanup:
1898 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1899 * Rename smx_action_t into smx_synchro_t, making explicit that these
1900 things are used to synchronize processes with their environment.
1901 For example, a communication is a sort of synchronization involving
1902 the communicating processes (that may block until the exchange) and
1903 the platform. The same can be said from computations, etc.
1905 - #18888: segfault when a process ends before its kill_time
1909 - 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.
1910 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1911 - MPI_Type_set_name, MPI_Type_get_name
1912 - MPI_*_c2f and MPI_*_f2c functions
1913 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1914 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1915 - Activate a lot of new tests from the mpich 3 testsuite
1917 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1918 - InfiniBand network model added: Based on the works of Jerome Vienne
1919 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1920 - When smpi/display_timing is set, also display global simulation time and application times
1921 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1922 * Collective communications
1923 - 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
1924 - 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/).
1925 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1927 - Replay now uses algorithms from wanted collective selector
1928 - Replay can be used with SMP-aware algorithms
1929 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1930 - 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.
1932 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1933 - Correctly use loopback on fat-tree clusters
1934 - Asynchronous small messages shouldn't trigger deadlocks anymore
1935 * Energy/DVFS cleanup and improvement
1936 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1937 - new: smpi_get_host_pstate()
1940 - "Full" network optimization flag was broken since Surf++
1941 - Better handling of precision flags in maxmin
1942 - Fix bug causing sometimes "Impossible" errors
1943 - Properly pass cluster properties to included hosts
1944 * Improvement of the Energy plugin.
1945 - Always update the consumption before returning that value
1946 - New property: watt_off to denote the dissipation when the host is off
1947 - New functions getWattMinAt and getWattMaxAt to retrieve the
1948 dissipation of pstates that we are not currently at.
1949 * Java: class NetworkLink renamed to Link
1950 * New function: simcall_process_get_kill_time()
1951 * Massive rename s/workstation/host/
1952 - That's intrusive, but that's good for the project consistency. Sorry.
1953 - Change config option "workstation/model" into "host/model"
1956 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1957 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1958 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1959 - xbt_dirname and xbt_basename for non-POSIX systems
1961 * The model checker now runs as a separate process.
1962 * The model checker runs must now be launched with the new simgrid-mc program.
1963 * Record/Replay: the MC can display a textual representation of a path in the
1964 execution graph. It can then be replayed outside of the model checker.
1966 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1968 SimGrid (3.11) stable; urgency=low
1973 * Normalizing pointers addresses tool for better diff between logs
1976 * Add cloud examples using new VMs
1977 - examples/msg/cloud/two_tasks_vm.tesh
1978 - examples/msg/cloud/simple_vm.tesh
1979 - examples/java/cloud/cloud.tesh
1980 - examples/java/cloud/migration/migration.tesh
1981 * Add java surf examples:
1982 - examples/java/surfPlugin/surf_plugin.tesh
1983 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1984 - examples/java/surfCpuModel/surf_cpu_model.tesh
1985 * Add SMPI+MSG example:
1986 - examples/smpi/smpi_msg_masterslave/
1991 - msg task destroy cancel test
1992 - msg_host on/off test
1993 * Move all tests in testsuite to teshsuite (adding tesh files)
1994 * Restructure teshsuites
1995 - one folder for each kind of test
1996 * Restructure AddTests.cmake
1998 - structure the order of tests (with sections)
2001 * Add virtual machine
2002 - creation of a VM on a PM
2003 - migration of a VM from a PM to another PM
2005 - MSG_process_join(msg_process_t process, double timeout)
2006 - msg_bar_t MSG_barrier_init(unsigned int count)
2007 - int MSG_barrier_wait(msg_bar_t barrier)
2008 - void MSG_barrier_destroy(msg_bar_t barrier)
2009 - msg_as_t MSG_environment_as_get_by_name(const char * name)
2010 * New option "msg/debug_multiple_use" to help debugging when a task is used
2013 - Improvements and finalization of MSG_storage, MSG_file APIs and their
2015 - Increase code coverage in test suites
2019 * Protect context stacks against stack overflow. The number of protected memory
2020 pages allocated on the top of each stack (1 by default) can be configured
2021 with the new command line option --cfg=contexts/guard_size:<value>.
2022 * Simcalls are now generated by a python script that
2023 - generates files included by SimGrid
2024 - checks that all the functions exist, and proposes prototypes
2026 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
2027 host_set_data, host_get_data
2029 - simcall_process_join(smx_process_t process, double timeout)
2030 * Fix bug where sleeping processing could not be suspended.
2033 * Translate surf models from C to C++
2034 - Generic classes for all models: Model, Resource, Action
2035 - A generic interface for each kind of model (CPU, Network, Storage
2036 Workstation, WorkstationVM)
2038 * Translate surf routings from C to C++
2039 * Add callbacks using sigc++ or boost::signals2
2040 - Add callback functions for resource creation/destruction
2041 - Add callback functions for action state change
2042 - Handle Energy as a plugin
2043 * Replace swag by boost::intrusive
2044 * Add new routing models for clusters. For documentation, see
2045 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
2046 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
2047 parameters for cluster tag
2048 - Fat trees, with topology="FAT_TREE" and
2049 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
2050 - see examples/platforms/torus_cluster.xml and
2051 examples/platforms/fat_tree_cluster.xml
2052 * More documentation
2055 * Hostfiles support host:nb_processes construct to deploy several processes on
2057 * Collective communication algorithms should not crash if used with
2058 improper number of nodes and report the error.
2059 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
2060 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
2061 MPI_Dims_create, MPI_Cart_sub are supported.
2062 * New interface to use SMPI programmatically (still depends on MSG for
2063 some parts, see examples/smpi/smpi_msg_masterslave):
2064 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
2068 * Global variables privatization in MPI executables is now performed at runtime
2069 with the option smpi/privatize_global_variables (default:no).
2070 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
2071 dynamic libraries loaded by the application are not privatized (static
2072 linking with these libraries is advised in this case)
2075 * Options defined in XML work correctly now.
2078 * New cmake option, enable_lib_in_jar, to control whether native libraries are
2079 copied into simgrid.jar or not (ON by default). Use this option if you want
2080 to reduce the size of the installed simgrid.jar, *and* the native libraries
2081 are kept installed elsewhere.
2082 * Surf binding with SWIG (code generated in maintainer mode only):
2083 - plugin to handle callbacks
2084 - CPU model only for the moment
2087 * Supernovae build mode is definitively removed. It was used to improve
2088 inlining and inter-module optimizations. It is nowadays superseded by
2089 link-time optimizations commonly available in compilers.
2090 * Update ns-3 find lib. Bindings for ns-3 should work again now.
2091 * Add boost dependency for surf++
2092 * Add new macro for tests
2093 - ADD_TESH(name <tesh_args>)
2094 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
2098 - xbt_bar_t XBT_barrier_init(unsigned int count)
2099 - int XBT_barrier_wait(xbt_bar_t barrier)
2100 - void XBT_barrier_destroy(xbt_bar_t barrier)
2101 * Make the xbt_os_time module public
2103 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2105 SimGrid (3.10) stable; urgency=low
2107 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
2110 * Reintegrate Java to the main archive as desynchronizing these
2111 package is not acceptable anymore (Java is now considered stable)
2112 * Add explicit synchronization facilities through semaphores
2113 * Bug fix: Task.setDataSize() only changed the C world, not the value
2114 cached in the Java world
2117 * Dramatically change the way files are handled. API and internals changed, but
2118 this part of MSG was not considered as production grade either.
2119 * Add explicit synchronization facilities through semaphores
2120 * Add a new function MSG_host_get_process_list()
2121 * Preliminary DVFS support (see examples/msg/energy/ for details)
2124 * SMPI is now included directly in the libsimgrid as the windows
2125 linker doesn't force us on splitting it anymore.
2126 * Improvements of the SMPI replay tool:
2127 - Most of the collective communications are now rooted in the same process as
2128 in the original application.
2129 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
2130 used until now). Multiple data types can now be used in a trace.
2131 - The replay tool now supports traces produce either by TAU or a modified
2133 - Bug Fix: the compute part of the reduce action is now taken into account.
2134 - Gatherv collective is now supported
2135 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
2136 of smpirun outputs time independent traces for the current run. One file
2137 is created per process. If too many processes are simulated, this behavior
2138 can be changed to one file for all processes by using the
2139 tracing/smpi/format/ti_one_file flag
2140 * smpirun generates the host file if needed (with given host count and platform)
2141 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
2142 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
2143 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
2144 --cfg=smpi/coll_selector:(mpich/ompi)
2145 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
2147 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
2148 happening inside a collective SMPI call.
2149 * Fix the behavior of complex data types handling.
2150 * Make MPI_Wtime another synchronization point to take computations into
2152 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
2153 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
2154 * Add all missing Fortran bindings, SMPI should work with Fortran 90
2155 (no privatization of global variables yet)
2156 * Preliminary DVFS support (see examples/smpi/energy/ for details)
2159 * Verification of liveness properties is now available for SMPI applications
2160 (in addition to MSG applications)
2161 * Bugged examples using SMPI in examples/smpi/mc/
2162 * Add --cfg=model-check/visited option. Allows the verification of infinite
2163 programs. Detection of loops in the execution thanks to the system state
2164 comparison and reduction of the state space to explore. Can be combined with
2165 DPOR for safety properties.
2168 * Allow to change SimGrid configuration (see --help) within the code
2169 thanks to SD_config() as it can be done in MSG.
2170 * Add a new function SD_task_set_amount() upon user request.
2173 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
2174 * Remove rule based routing (no more PCRE dependency)
2175 * Add a limiter_link option to cluster tag, to specify a maximum reachable
2176 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
2177 * Add a loopback_bw and loopback_lat options to cluster tag.
2178 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
2179 routing. Now peers are encapsulated in an AS and have their own private
2180 router but this is transparent.
2183 * Our own implementation of getline is renamed xbt_getline, and gets
2184 used even if the OS provide a getline(). This should reduce the
2185 configuration complexity by using the same code on all platforms.
2186 * New type: xbt_cfg_elm_boolean.
2187 * Allow to use yes/no for boolean configuration options in the command line.
2188 * Allow to disable SimGrid cleanups at exit from command line option.
2189 There are situations where one may want a simulation to end with an exit.
2190 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
2191 annoying when scripting around the simulator. Adding a
2192 --cfg=clean_atexit:no allows to circumvent this issue.
2195 * Lots of memory leaks were corrected in this release.
2196 There is no known memory leaks anymore, in all of our 600+ tests.
2197 * New command line option --version, to get SimGrid version information.
2198 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
2200 * Supernovae builds are deprecated, and expected to be removed in the next
2203 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2205 SimGrid (3.9) stable; urgency=low
2207 The Grasgory release: GRAS is really dead now.
2209 * Complete overhaul of the internal host structures scheme.
2212 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
2213 was considered as experimental and badly maintained since then.
2214 * Keeping it was thus a trap to our potential users, that could take
2215 it instead of MSG or SMPI by mistake despite is pity state.
2216 * GRAS seems to have very few users (if any), and no one volunteered
2217 to maintain it further. It also induces a lot of XBT code (for
2218 portability sake), that must be maintained too.
2219 * For all these reasons, we killed GRAS. If someone wants to revive it
2220 in the future, don't cry, our git history still remembers of GRAS.
2223 * Major overhaul. Merge our documentation again as time proved that
2224 splitting it was really not helping our users.
2225 * Further improve the developer documentation to help newcomers
2226 hacking on SimGrid itself. The user documentation (and in
2227 particular, the beginner documentation) is still in a sorry state.
2230 * Now works on Windows too!
2231 * Much more extensive test suite, from MPICH
2234 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
2235 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
2236 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
2237 * Bug fix: let task be scheduled when the last dependency to be solved is
2238 a control dependency.
2239 * Remove SD_load_environment_script function.
2240 Use the C sg_platf function if you want to declare a platform
2244 * New function: MSG_process_get_number()
2245 * Old function documented: MSG_config()
2246 * Remove MSG_load_platform_script function
2247 Use the C sg_platf function if you want to declare a platform
2251 * Change the default value of the TCP_gamma constant (maximal size of TCP
2252 congestion window) to a more realistic 4MiB value. If you notice changes in
2253 your simulation results, you can fall back to the previous 20k tiny window
2254 by adding --cfg=network/TCP_gamma:20000 on command line.
2255 * (Hopefully) fix a bug wrt periodic availability and state traces
2256 * Bug fix: use default values at start when first event in availability/state
2257 trace is not at time 0.
2260 * remove the "new_" part of function name sg_platf_new_trace_connect
2261 (resulting in sg_platf_trace_connect), since it does not create
2265 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2266 use case where it's really mandatory, and maintaining it was a pain in
2268 * New: xbt_fifo_search(), search an item with a user-provided
2269 comparison function instead of dumb pointer comparison.
2272 * Fix the lua deployment:
2273 Use `simgrid.init_application()` before deployment instead of
2274 `simgrid.msg_register_application()` after.
2277 * Transfer the tracing files into the corresponding modules.
2279 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2281 SimGrid (3.8.1) stable; urgency=low
2283 The "we are told that some people want to also *install* the simgrid
2286 * Add missing file "tesh.1" to the archive.
2288 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2290 SimGrid (3.8) stable; urgency=low
2292 The Psssshiiiit release: SimGrid jumps into the Cloud.
2295 * Add an experimental interface to manipulate VMs. They are mainly
2296 process groups with very few intrinsic semantic, but they should
2297 allow you to build the semantic you want easily.
2298 * New function: MSG_host_set_property_value()
2299 * New function: MSG_process_on_exit(). To clean memory in all cases.
2300 * Bug fixes that made the host (and link) failures unusable.
2301 * Add a way to auto-restart process when the host in which they are
2302 executing comes back (ON_FAILURE="RESTART" on deployment file,
2303 MSG_process_auto_restart_set).
2304 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2305 please stop using the old ones, they are DEPRECATED.
2307 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2308 Please use MSG_init() instead. (reducing the amount of entry
2309 points in the library helps us).
2310 * Make it impossible to link against the wrong version of the lib
2311 * Deprecate MSG_clean(). No need to call it anymore.
2312 * Function MSG_get_host_number() is not deprecated anymore.
2315 * Split the doc into a user guide and a reference guide.
2316 * Start a developper guide to help people hacking on SimGrid.
2319 * Enable tracing by default. This modules rocks you should use it.
2320 * Remove option custom_flags. Now use environment variables CFLAGS
2322 * Use default cmake things to detect lua instead of home grown ones.
2323 * New option "enable_mallocators" to disable mallocators, for debugging
2324 purpose ("on" by default).
2327 * Bug fixes around the resource failures: don't let the processes
2328 survive the host they are running onto.
2329 * Add an interface to auto-restart processes when the host in which they are
2330 executing comes back.
2331 * Ensures that SIMIX_clean is called automatically. It's not part of
2332 the public interface anymore (bindings should be updated).
2335 * Bug fix for when SD_Simulate is called with a positive value: be careful
2336 when comparing doubles. Sometimes they are different for non significant
2338 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2339 parallel task whose initial work is distributed among host according
2340 to the Amdahl's law. Such tasks are created with a parameter alpha
2341 that corresponds to the non-parallelizable part of the computation.
2342 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2343 between two sets of workstations assuming a 1D block distribution (each
2344 workstation owns a similar share of data) on both sides.
2346 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2347 Data redistribution will be automatically scheduled once parent and child
2348 are both scheduled. The filling of computation_amount and
2349 communication_amount structures is now done seamlessly thanks to the chosen
2351 * New function SD_workstation_dump to display various information
2352 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2353 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2354 bandwidth on the route onto which the task is scheduled and the amount of
2356 To divide the nominal bandwidth by 2, the rate then has to be:
2357 rate = bandwidth/(2*amount)
2358 * Compute tasks that have failed can now be rescheduled and executed again
2359 (from their beginning)
2360 * Increasing source code coverage (src/simdag is now covered at 95.8%
2364 * Re-implement time-independent trace replay using SMPI (at the
2365 smpi_smp_* level) instead of MSG. This should replace
2366 examples/msg/actions/actions.c
2367 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2368 hindexed and structs)
2369 * Implement the exchange of non-contiguous data.
2370 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2371 * Correct behavior of smpi/sender_gap and set its default value to 0
2372 * Add option to asynchronously send small messages to allow better
2373 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2374 specifies the size in bytes under which messages will be asynchronously sent.
2375 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2376 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2379 * New C interface to define a platform: XML is now optional.
2380 For more info, please check include/simgrid/platf.h.
2381 * New interface to define random platforms from the C:
2382 For more info, please check include/simgrid/platf_generator.h and
2383 examples/msg/masterslave/masterslave_platfgen.c
2384 * Export a sg_cmdline dynar containing all the arguments we got from
2388 * Two new tracing options for adding comments to trace file so you
2389 can track your experiments (see --help-tracing for details).
2390 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2391 * Adding the SimGrid version that generated the trace file as a comment.
2392 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2393 * Fix to avoid key clashes on Paje links
2394 * Other minor fixes related to the Paje specification
2397 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2398 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2399 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2400 * Greatly improve the robustness of mmalloc to user errors (such as
2401 using an area after freeing it, or freeing it twice)
2403 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2405 SimGrid-java (3.8.1) stable; urgency=low
2407 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2408 Warning: all methods are visible, but only some of them are
2409 implemented so far. Check the source (src/jtrace.c)
2410 for further information.
2411 * New module: org.simgrid.msg.File (SimGrid File management functions)
2412 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2413 * Change the meaning of Process.restart: now restart the process from
2414 the beginning, like MSG_process_restart in C.
2415 * Add Process.setAutoRestart: handling of process restart when failed
2417 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2418 you to retrieve the properties of the processes/hosts
2419 * Deprecate Msg.clean(): you can just forget about it now.
2420 * New function Process.getCount(), that only works when compiling
2421 with the not yet released version 3.9 of the C library.
2423 * New context factory based on Coroutines. It mandates a modified JVM
2424 but then, the simulations run about five times faster, and there is
2425 no limit to the amount of processes (beside of the available memory).
2427 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2429 SimGrid (3.7.1) stable; urgency=low
2432 * Restore the prototype of MSG_process_create_with_environment() to
2433 the pre-3.7 situation by removing the kill_time argument.
2434 * Add a MSG_process_set_kill_time() function instead.
2437 * Fix weird behaviors when dealing with parallel tasks.
2440 * Simgrid is now built as a dll.
2441 * Simgrid-java now works on Windows.
2442 * Simgrid-Java is now included into Windows package.
2445 * First pre-build package for MacOSX.
2448 * Fix compilation when using MSG_USE_DEPRECATED.
2449 * Fix some compilation issues on Macs and Windows.
2450 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2451 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2453 We discovered that the Lua console is broken, but we are missing the
2454 manpower to fix it right now. The problem existed in 3.7 too, so we
2455 are not blocking the release for that. Sorry if you depended on this
2456 feature, any help would be really welcome.
2458 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2461 SimGrid-java (3.7.1) stable; urgency=low
2463 The "Java aint got to be bloated and slow" release
2466 * Various internal cleanups and performance improvement
2467 Simulations are expected to run up to twice faster or so
2468 * Make Process.kill(process) an instance method, not a static one
2469 * User processes are not java.lang.Thread subclasses.
2470 This breaks the compatibility (sorry), but previous API was
2471 brain-dead, making it impossible to have non-trivial
2472 initializations in the process constructor.
2473 * Require a full constructor per Process sub-class.
2474 Kinda breaks the compatibility (sorry), but this allows a much more
2475 efficient way to launch the processes at simulation startup.
2476 * Do not embeed our version of semaphores, java 1.5 can be considered
2477 as sufficiently prevalent for us to not dupplicate its features.
2481 * Add examples for almost every part of the API
2482 We spotted and fixed a lot of bugs in the process
2484 * New module: asynchronous communication API
2485 * New function: Process.sleep()
2486 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2487 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2488 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2490 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2492 SimGrid (3.7) stable; urgency=low
2494 The "spring cleanups (before the next Big Project kicks in)" release.
2497 * We can specify the SMPI latency/bandwidth factor with command line
2498 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2499 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2500 You can also use the "config tag" from platform file by adding this line
2501 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2502 (see "example/platforms/tag_config.xml" to use "config tag").
2503 Note that the command line supersedes the platform file configuration.
2504 * Change the correction factors used in LMM model, according to
2505 the latest experiments described in INRIA RR-7821.
2506 Accuracy should be improved this way.
2507 * Use the partial invalidation optimization by default for the
2508 network too. Should produce the exact same results, only faster.
2509 * Major cleanup in surf to merge models and split some optimization
2510 mechanisms from the core of the models. As a result you can now
2511 specify which model to use (e.g., --cfg=network/model:LV08
2512 --cfg=cpu/model:Cas01) and which optimization mode to use
2513 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2514 Incompatible combinations should err at initialization. See
2515 --help-models for the list of all models and optimization modes.
2516 * The CLM03 workstation model was dropped for simplicity because it
2517 used the deprecated CM02 network model. Use default instead.
2518 * Rename the TCP_gamma configuration option to network/TCP_gamma
2519 * Rename the coordinates configuration option to
2520 network/coordinates, and document it
2521 * Use now crosstraffic keyword instead of the terribly misleading
2522 fullduplex keyword. It is activated by default now in the current
2523 default model, use --cfg=network/crosstraffic:0 to turn it off.
2524 * Ongoing refactoring the model parsing to make XML files optional
2525 See include/simgrid/platf.h for details (still to be completed)
2528 * Major overhaul of the documentation. Almost instructive now :/
2529 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2530 functions and friends. This interface was considered as
2531 deprecated since over 2 years, it's time to inform our users that it is.
2532 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2533 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2534 use these (crappy) functions in your code.
2535 These functions will be removed soon. Stop using them now.
2536 * Deprecate MSG_get_host_{table,number}
2537 Implement MSG_hosts_as_dynar() instead.
2538 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2539 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2542 * Stabilize the parallel execution mode of user contexts
2543 * Introduce configuration variables to control parallel execution:
2544 - contexts/synchro: Synchronization mode to use when running
2545 contexts in parallel (either futex, posix or busy_wait)
2546 - contexts/parallel_threshold: Minimal number of user contexts
2547 that must be part of a scheduling round to switch to parallel
2548 execution mode (raw contexts only)
2549 * Fix bugs that prevented to use suspend/resume along with
2550 synchronization structures.
2551 * Fix bugs in process termination that lead to invalid memory access
2552 in very specific conditions.
2555 * Introduce a parallel mode for the models (controlled by surf/nthreads
2556 configuration item). In our tests, running the models in parallel
2557 never lead to any speedups because they are so fast that the gain
2558 of computing each model in parallel does not amortizes the
2559 synchronization costs, even when ultra fast futexes are used.
2560 This is released anyway because YMMV.
2563 * Performance boost by using a swag internally to compute the set of
2564 tasks that are finished and should constitute the return value of
2568 * Enable it by default now that it is considered rather stable.
2571 * Documentation of the tracing functions.
2572 * Performance gains when tracing categorized/uncategorized resource
2573 utilization by avoiding calls to get route when updating resource
2574 variables. LMM constraints are being used instead.
2575 * API changed to set task categories. Use MSG_task_set_category instead
2576 of TRACE_msg_set_task_category, and SD_task_set_category instead
2577 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2578 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2579 * Examples for link user variables added (see at examples/msg/tracing/)
2580 * Deprecated function TRACE_msg_set_process_category completely removed
2581 * Trace header updated according to the latest Paje file format
2582 * Tracing network lazy updates, no longer obligate users to use full updates
2583 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2584 * Experimental: let user code declare/set/push/pop application states for hosts
2585 * API changed to allow the manual creation of graph configuration files
2586 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2589 * Improve the API of Lua MSG bindings, using the Lua spirit.
2590 * Each simulated process now lives in its own Lua world (globals are
2591 automatically duplicated). It helps writing simulators. Will allow
2592 to run Splay programs within SimGrid in the future.
2593 * Add a Chord example in Lua, equivalent to the MSG one.
2596 * Start the implementation of a solution to express temporal
2597 properties, not only local assertions. This is still an
2598 experimental work in progress, stay clear from it to be safe.
2602 - Add new runtime parameters --help-logs and --help-log-categories
2603 to display information about supported logging parameters and
2605 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2606 don't exists anymore.
2607 * Mallocators: allow value NULL for the reset function.
2609 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2610 create homogeneous dictionaries, where all the elements share the
2611 same free function. Non homogeneous dictionaries will be
2612 deprecated in the next release.
2613 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2614 - Multi-level dictionaries are deprecated.
2616 - new function xbt_dynar_search_or_negative() that is useful when
2617 you have less than 2 million elements in your dynar and don't
2618 want of the extra complexity of catching exceptions when the
2619 element is not found.
2621 - Make xbt_os_thread module (for thread portability) public.
2622 Documentation is still to come, sorry.
2624 - Cleanups and simplifications to make it maintainable again.
2625 - Exotic features (such as memalign and valloc) were removed.
2626 - The metadata were extended and improved so that the
2627 model-checker becomes able to explore and inspect the heaps.
2628 - This may induce a performance drop when enable_model-checking is
2629 ON in cmake (even if it's not used in the simulation), but it is
2630 necessary at this point to get MC working.
2632 Turn model-checking OFF if simulation performance matters to you.
2633 Not enabling it at runtime is not enough, disable it in cmake.
2635 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2638 SimGrid (3.6.2) stable; urgency=low
2640 The "Not coding new stuff allows to polish old things" release.
2643 * New bindings to the NS3 packet level simulator (experimental)
2644 * Use the raw (efficient) execution contexts instead of the sysv
2645 (portable) ones when possible.
2646 * libpcre is now mandatory in any cases since not using it led to
2647 severe performance loss and possibly other issues
2648 * Update the XML platforms:
2649 - G5K: include the latest machine in Nancy
2650 - GridPP and LCG: new platforms
2651 * Documentation was partially updated, at least (more to come)
2653 Bug fixes, cosmetics and small improvements
2654 * Free terminated processes before the end of the simulation to avoid
2655 exhausting the memory of users having very dynamic amount of
2657 * Bug fix and cosmetics about canceling non-running tasks
2658 * Bug fix about the dot loader's issues when using libcgraph
2661 * Create an installer for windows with nsis (amd64 and win32)
2662 - Add an hello world project to illustrate simgrid project creation.
2663 - Embed libpcre into the Simgrid installer to avoid
2664 its compilation burden
2665 * The raw execution contexts should work on Apple now
2666 * Port to Windows 64 bits
2667 - Sysv contexts now have an implementation for this arch
2668 - GRAS communication features now support this arch
2669 * Drop support for borland compiler on windows
2670 - this code was not maintained, and we kinda depend on gcc nowadays
2671 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2672 * Fix portability issue on unstable ubuntu: linker became picky on
2675 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2677 SimGrid (3.6.1) stable; urgency=low
2679 The "Oops, we broke Macs too" release
2682 * Fixed contexts detection so that raw ones are used when possible
2683 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2684 leads to a strange error, with user code segfaulting sometimes when
2685 the generated code is not perfectly aligned (which is not
2686 controllable from the user side, depends on the amount of code)
2689 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2691 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2693 SimGrid-java (3.6) unstable; urgency=low
2696 * Split of every thing from simgrid v3.5 into a separate package.
2698 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2700 SimGrid (3.6) stable; urgency=medium
2702 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2705 * Bindings now constitute their own package, separated from the main one.
2706 Rationale: reduce our maintenance nightmare by reducing the module coupling
2707 They will soon be released on their own on gforge.
2709 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2710 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2712 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2713 * It's not quite deprecated for now because we have no replacement,
2714 but it may soon become the case.
2717 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2718 * Fortran: New user-level cache variable to store the rank of the running
2719 process. This improves performance by an order of magnitude.
2720 * C: New coccinelle script to automatically locate and modify global and
2721 local static variables.
2722 * Improved SMPI network model with a sender-side gap to account for multiple
2726 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2727 only say if a communication is finished, no matter whether it succeeded or
2728 failed. You can call MSG_comm_get_status() to know the status of a finished
2730 * New function MSG_task_dsend() to send a task and detach it. When a
2731 communication is detached, you are never notified of its success or failure
2732 and the memory is released automatically once it is finished. This function
2733 is useful when you don't care about the end nor the success of a
2735 * Change the prototypes of action replay. Sorry for inconvenience,
2736 but this is really more efficient this way (and to adapt your code,
2737 you just have to fix the initialization, that shouldn't be too long)
2738 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2739 ever managed to do anything useful with it.
2740 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2741 takes a dynar of communications. It returns immediately and gives the
2742 index of a finished communication (if any).
2743 * New example: a basic implementation of the Chord P2P algorithm.
2746 * New model for multi-core CPUs. You can now use the core attribute to
2747 precise the number of cores of a host. This is a basic model. Every
2748 process running on the host receives at most the power provided in
2749 the DTD (throughput<=power). Total throughput of process cannot exceed
2751 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2752 router linked by an up-link and a down-link (possibly asymmetrical).
2753 This kind of pattern allows to easily build last-mile model style platforms.
2754 Aggregating such patterns in a rule-based AS is thus the technique of
2755 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2756 * New model for Vivaldi routing. We transformed the Vivaldi network model
2757 into a Vivaldi routing model (based on the rule-based model). This allows to
2758 combine Vivaldi based latencies with last-mile platforms.
2761 * Added a check for NaN of IEEE754 infinite in the double entries of
2763 * Introduce a new context factory "raw", highly inspirated from the
2764 ucontext factory, but using manually crafted functions in assembly to
2765 do the work in an efficient manner.
2766 * Allow to change the used context factory at run time, not only at
2767 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2768 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2769 contexts at runtime (only with raw contexts or ucontexts).
2770 * Completely rewrote this module to allow parallel execution of user
2771 processes. Use --cfg=contexts/nthreads:N to execute user processes
2772 with N parallel threads (the default is 1, meaning no parallelism).
2773 * Allow to decide dynamically between sequential and parallel modes.
2774 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2775 processes in parallel only when their number is greater than or equal to P
2777 * Added a check for NaN of IEEE754 infinite in the double entries of
2781 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2782 won't output the state of processes when interrupted with Ctrl-C
2783 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2784 NULL-terminated array. This may solve backward compatibility issues
2785 due to the change to return type of SD_simulate. See also:
2786 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2787 * Add new macros with variable number of arguments.
2788 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2789 - in xbt/asserts.h: xbt_assert
2790 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2791 - in xbt/ex.h: THROWF and RETHROWF.
2792 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2794 * Change xbt_die() to accept a format string with arguments, just like printf.
2795 * New data structure: xbt_lib_t, like a dict but more general and with better
2799 * New configuration options
2800 Options triva/categorized and triva/uncategorized can be used to generate
2801 graph configuration files for Triva visualization tool.
2802 * Configuration option tracing/platform is renamed to tracing/categorized
2803 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2804 * New TRACE_user_link_variable interface:
2805 User provides the name of the link and the tracing variable to attach to it
2806 * the declaration of tracing categories must be done after the environment
2808 * simpler tracing interface, just one way to declare categories
2809 TRACE_category or TRACE_category_with_color, it is up to you
2810 * links in the trace file are again identified by their names
2811 * trace contains the full platform hierarchy exactly as declared using the ASes
2812 * Options tracing/msg/[task|process]:1 groups the process by hosts
2813 for both cases, tasks and processes must have names that are unique during the simulation
2814 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2815 * The experimental option tracing/msg/volume is deprecated
2816 its functionality may be reincorporated if needed
2818 The tracing generates a trace file with unordered timestamped events,
2819 because of the way the core simulator (surf) works. A script available
2820 at the tools directory (fix-paje-trace.sh) can be used to put the events
2821 in order. We have changed the tracing so it can generate ordered timestamped
2822 events in the final trace, but depending on the simulator (and how much time
2823 is simulated) that can lead to a huge memory utilization. It is deactivated
2824 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2826 Build Infrastructure
2827 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2828 - We are trying hard to keep the API stable, but it may happen that
2829 some things change (we're a research project after all, not a
2830 nuclear plant operating system). If such things should happen, you
2831 could rely on that macro to adapt.
2832 - current value: 30600 for 3.06.00, aka 3.6
2833 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2834 a number that can be compared with SIMGRID_VERSION.
2835 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2836 assertions and verbose/debug logging events are disabled at compile time.
2838 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2840 SimGrid (3.5) stable; urgency=medium
2843 * New feature to any SimGrid-based simulator: Model-Checking
2844 Check SIN#1 for more details.
2847 * New Model SMPI (three-interval linear regression for correction factors)
2848 See RR-7426, available at http://hal.inria.fr/inria-00527150
2849 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2850 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2851 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2852 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2853 * See also src/smpi/README
2858 - Tracing API changes: TRACE_start and TRACE_end should not be called
2859 by user-code. They are automatically called by simulators created
2860 with SimDAG, MSG and SMPI if the toolkit is compiled with
2861 tracing_enabled=ON. Categories declaration and utilization remain the
2862 same for MSG and SimDag.
2863 - A function was added to the tracing API to declare categories with
2865 - TRACE_category_with_color (char *category, char *color)
2866 where color must be in the following format
2867 "%f %f %f", red, green, blue
2868 and red, green, blue are float values in the interval [0, 1]
2869 - User can specify NULL as color parameter, or continue calling
2870 TRACE_category (cat)
2871 On that case, the tracing system will define random colors
2872 - The following command-line options are supported:
2873 --cfg=tracing/filename:msg.trace
2874 --cfg=tracing:1 (activate tracing, needed to use others)
2875 --cfg=tracing/platform:1 (categorized resource use)
2876 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2877 --cfg=tracing/msg/task:1 (task creation)
2878 --cfg=tracing/msg/process:1 (process creation, migration)
2879 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2880 --cfg=tracing/smpi:1 (SMPI interface tracing)
2881 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2882 - examples of examples/msg/tracing updated
2884 - DAXLoader and DOTLoader functions can generate tasks with categories
2885 - A new function to attribute a category to SD tasks:
2886 TRACE_sd_set_task_category (SD_task_t task, char *category)
2887 * Tracing the MPI interface implemented by SMPI
2888 - Collective operations are traced with states
2889 - Point-to-Point operations are traced with states/links
2890 - Tracing activated by a parameter "-trace filename" passed
2891 to smpirun during execution (considering that simgrid
2892 is compiled with tracing enabled)
2893 - To run the simulation with gdb, the simulator
2894 accepts --cfg=tracing/smpi:1 to trace SMPI
2895 - tesh files to check if smpi tracing is ok
2896 - See examples/smpi/NAS/DT-trace
2897 * GTNetS tracing re-worked
2898 - adaptation to the tracing system of GTNets to cope
2899 with modifications regarding the fullduplex mode
2900 - new tesh files to check if gtnets tracing is ok
2903 * Asynchronous communications through the functions:
2904 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2905 * New function: MSG_load_platform_script()
2906 to make possible using a lua script instead of XML files to set up platforms
2907 * New function: MSG_set_function
2908 to associate functions to processes, used when bypassing the parser
2909 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2911 Platforms: Add some more examples in examples/platforms
2912 * Grid'5000: see www.grid5000.fr
2913 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2916 * Change the XML format. This is a very important modification. SimGrid
2917 3.5 introduces a new hierarchical format based on the notion of
2918 Autonomous Systems. Compatibility with old format is ensured
2919 through the perl script provided in the install bin directory
2920 bin/simgrid_update_xml.
2921 It is now possible to build platforms with specific routing
2922 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2923 connect several platforms together. We will try to provide soon
2924 set of realistic platforms exploiting these properties (have a
2925 look at examples/platforms/ for the moment).
2926 * Take the opportunity of the XML format change to be a good XML citizen:
2927 rename link:ctn to link_ctn and similar changes (also dealt with by
2929 * Add a new routing scheme (rule-based) using regular expressions. It
2930 enables to have an extremely low memory footprint when the
2931 underlying routing is simple and can be compactly described. You
2932 need to have libpcre4-dev (perl regular expressions) installed if
2933 you want to use this routing scheme.
2934 * Revive the cluster TAG and allow to easily and efficiently (both in
2935 term of memory and speed) connect clusters together. Have a look
2936 at teshsuite/simdag/platforms/ to see how this can be done. With
2937 this tag, you can create clusters with thousands of tasks at no
2938 cost (have a look at examples/platforms/).
2939 Note: clusters are implemented as ASes, so there is no need for an
2940 enclosing AS tag if you have only one cluster in your platform.
2941 * Add new generic functions in the public interface that allows the user
2942 to call SURF 'create_resource' methods from your code (same
2943 functionality as the XML bypass mechanism but with a much lighter
2945 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2946 that uses a piecewise linear approximation to produce better
2947 results when exchanging small messages.
2948 * Add a new parameter to handle correctly full duplex link and account
2949 for interferences between uplink and downlink communications
2950 (activate with --cfg=fullduplex:1).
2953 * Rename the SD_READY (all dependencies are satisfied and task is
2954 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2955 dependencies are satisfied) state.
2956 This prevents a confusion between the notion of "ready to schedule"
2957 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2958 simulated" (SD_RUNNABLE) used by the simulation kernel.
2959 * Change the way a task is considered as ready. Instead of removing
2960 dependencies when a task is done, a counter is decreased. This way,
2961 it is always possible to reach ancestors thanks to the
2962 SD_taks_get_parents function (even after the end of the simulation.)
2963 * Change the return type of SD_Simulate from (SD_task_t*) into
2964 xbt_dynar_t. This function was in handling a dynar internally and
2965 converted it into a NULL terminated array for historical reasons.
2966 * New function SD_dotload(char*) to load a DAG described in dot
2967 format. This loader and the corresponding examples require the
2968 installation of the graphviz library.
2969 * Fix a bug in the management of tasks of size 0 in the surf network
2970 models. This problem was only visible with SIMDAG and you should
2971 thus disregard results produced with earlier versions if you
2972 relied on this feature (some tasks were blocked because of this).
2973 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2974 even though your applications were doing only point-to-point
2975 communications and sequential computations. Now you can really use any
2976 model you want (of course, if you create real parallel tasks, which are
2977 not implemented in most models beside ptaskL07, this will abort).
2978 * Add an example that schedules a DAX on an heterogeneous platform
2979 using a Min-Min strategy.
2980 * New function SD_workstation_get_current_task() that returns the kind
2981 of task currently running on a workstation in the sequential access
2983 * Raise some warnings when unexecuted tasks remains at the end of the
2984 simulation. This is usually caused by cycles in the DAG.
2987 * New function: SIMIX_process_set_function() called by MSG_set_function
2988 * Change the underlying waiting queue in semaphores so that a process
2989 can wait on several of them simultaneously (as in waitany).
2990 * Fix the way to handle tokens in semaphores so that all access patterns
2991 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2992 * kill the dirty pimple SIMIX_message_sizes_output()
2993 Please use (proper) visualization instead
2996 * New data container: setset (set of sets of elements)
2997 * New module: mmalloc (mapped malloc, allowing to have several
2998 independent segments of malloc)
2999 * New function: xbt_dict_cursor_set_data()
3000 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
3001 * New function: xbt_dynar_is_empty()
3002 * New function: xbt_fifo_get_last_item()
3003 * Fix xbt_dynar_shrink(): use the right element size.
3004 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
3005 uninitialized areas during expand.
3006 * Fix semaphores: previous implementation was severely broken.
3007 * Use library init/fini functions for our initialization.
3008 - you can use logs and other feature as soon as you want in your
3009 code (even before the xbt_init / MSG_init)
3010 - xbt_exit is now a no-op and produce a warning when used.
3013 * Port GRAS to new SIMIX mechanisms. This allows gras users to
3014 benefit from the latest improvement to the simulation kernel.
3015 * Kill measurement sockets for now. If you rely on them, sorry. This
3016 release is not for you. This feature will be reintroduced in the
3017 future, but we cannot delay the release any further.
3018 * New function: gras_msgtype_get_name().
3019 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
3020 * Fix (at last) the pmm example: it should not randomly fail anymore.
3022 Build chain: bug fixes and overall polishing
3023 * Cmake is now stable enough. Hence, we killed the autotools.
3024 * Port to windows ( TM :)
3025 * Fix the 'make install' target.
3026 No need to use 'make install-simgrid' anymore
3027 * Introduce a 'make dist' target compiling a *source* archive
3028 'make package' compiles a binary archive
3029 * Compile java files only on need
3030 * Add --cd and --setenv command line options to tesh
3031 * Out of source builds are not fully supported yet, but we are close
3032 * Enable supernovae and optimization flags by default for our users
3035 * Add layer to set up environment directly from lua, without XML.
3036 * The effect of gras_stub_generator can be achieved through
3037 lua too (check examples/gras/console/ping_generator.lua)
3039 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
3041 SimGrid (3.4.1) stable; urgency=low
3043 The "Polishing easter eggs is probably a good idea" release.
3044 This is a bug fixes release only.
3048 * Fix a bug preventing the tasks from begin garbage collected.
3051 * Fix a bug occurring when a host involved in a communication fails.
3052 This was not detected properly by the other peer involved in the
3053 communication. Now, it's reported as a network error.
3056 * Warn the user about loop dependencies in data flow of DAX files
3057 * Obey the control-flow dependencies of DAX files
3060 * Add option "enable_smpi" allowing to not compile SMPI.
3061 Probably useful for the (Mac) users experiencing a build error here
3062 * Improve the detection of lua5.1 and ruby1.8
3064 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
3066 SimGrid (3.4) stable; urgency=low
3068 The "Easter in Cargese" release. Also known as (major changes):
3070 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
3071 ~> bindings were greatly improved
3072 ~> new tracing infrastructure for better visualization introduced
3074 * the "Welcome to configury modernity" release.
3075 ~> we switched from autotools to cmake, and improved our cdash
3078 A more detailled list of changes follow (full detail in svn log).
3080 Java Bindings: Various Cleanups
3081 * (install java-gcj-compat-dev on debian-like to use them)
3082 * Remove put/get: no need to export deprecated interface in Java
3083 Use send/receive instead.
3084 * Cleanup the examples and add a README per directory
3085 * Remove example autoDestination (that's the only way to go now)
3086 * Remove example explicitDestination (was a plain copy of basic)
3087 * Make JniException a runtime exception, so that there is no need to
3088 declare the fact that you may encounter such a beast. I guess that
3089 nobody will ever want to survive such error.
3090 * Create specific errors for each MSG case of failure:
3091 host failure, transfer failure, timeout, task cancelled
3092 * Cleanup the exceptions that may get thrown by each function
3093 * Other internal cleanups in Java bindings. Performance still bad :/
3094 Ruby and Lua Bindings: create them
3095 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
3096 * That's new and great, you should try them out.
3097 Same functionalities than Java bindings, only even less polished
3099 * Kill the useless "rate" argument of SD_task_get_execution_time()
3100 Everyone used to provide -1 as a value, it was not used, and the
3101 semantic of a possible use wasn't even clear.
3102 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
3103 either as comm costs or compute costs to mean that there is no
3104 such thing for that specific task.
3105 * Add a SD_task_set_name() function
3106 * Fix SD_task_unschedule() on typed tasks
3107 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
3108 * In DAX loader, accept useless 'level' attributes to <job> since
3109 LIGO DAGs have them (seem to be to ease graphical representation).
3111 * Add an example masterslave_mailbox.c using send/receive and not
3112 the deprecated put/get interface.
3113 * Kill the MSG_paje_output() function. It's a noop since 2 years.
3114 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
3116 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
3117 (declare MSG_USE_DEPRECATED to still have the old name)
3118 * Add a MSG_task_set_data() function
3119 * About trace replay (see examples/msg/actions):
3121 - Allow to work with splitted trace files for each process
3122 Give the specific trace file as argument of each process,
3123 and call MSG_action_trace_run(NULL)
3124 You can still have one merged file for all processes.
3125 - Fix implementation of collective operations
3126 * Allow task_execute() on 0-sized tasks (closes #10063)
3128 * This is the first release of SimGrid where SMPI is not considered
3129 beta anymore (even if some corners should still be improved)
3130 * Port over the new SIMIX_network submodule (internal refactoring)
3131 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
3132 * Implement more missing elements of the standard:
3134 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
3136 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
3137 MPI_Scan() MPI_Get_processor_name()
3138 - Added implementation of missing case for Alltoall (warning: it's
3139 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
3140 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
3141 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
3142 SMPI_MPI_Allgatherv()
3143 * Bug fixes include:
3144 - MPI_Waitsome() was broken
3145 - Allow relative includes in smpicc
3146 - Command line cfg argument 'reference_speed' was ignored...
3147 - Some functions did not properly lead to auto-benching of user code
3148 - smpicc passes -O2 by default (just like openmpi one)
3150 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
3151 * Bug fixes about timeouts during communications
3152 * add SIMIX_message_sizes_output() as a pimple to write to file the
3153 amount of messages per size. Use gnuplot to get histogram.
3154 Pimple because that's the only user-visible function of simix,
3155 defined directly in xbt.h (irk, sorry)
3157 - Add a SIMIX_sem_get_capacity() function
3158 - Fix interactions with processe resume/suspende
3159 - release_forever() was stupidly broken
3160 - Fix SIMIX_display_process_status() for processes in a semaphore
3161 - Make SIMIX_sem_block_onto() user-visible
3162 * Refactoring context stuff:
3163 - Use pseudo-OOP for better modularity
3164 - reimplement SIMIX_process_kill() without process_schedule() so
3165 that the latter can take as invariant that it is called from
3167 - Merge context_start into context_new for sake of simplicity
3169 * Add a Vivaldi network model, coded live during SUD'10 ;)
3170 * Rename configuration variables to start a hierarchy:
3171 o cpu_model -> cpu/model
3172 o network_model -> network/model
3173 o workstation_model -> workstation/model
3174 * New configuration variables:
3175 o network/bandwidth_factor: correction to bandwidth
3176 o network/latency_factor: correction to latency
3177 o netwotk/weight_S: correction to the weight of competing streams
3178 * Add a long description to the models, that users can see with such
3179 argument on the command line: --cfg=cpu/model:help
3180 * --help-models display the long description of all known models
3182 * config: add the ability to set a default value after registration
3183 Does not override any previously set value (e.g. from cmd line)
3184 * dict: allow to have integer key and data.
3185 When so, you need to use the following functions
3186 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
3187 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
3188 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
3189 In contrary to regular dicts, the key is not malloced before copy.
3190 Mixing scalar and regular elements in the same dict is not tested
3192 * Allow to use xbt_dynar_shrink() to expend the dynar instead
3193 Tracing for Visualization:
3194 * SimGrid is now instrumented in order to generate a trace file for
3195 visualization analysis: to use it, need to compile SimGrid with the
3196 "tracing" option enabled, and instrument the program using SimGrid with
3197 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
3198 (among other functions).
3199 * The instrumentation only traces the platform utilization for now
3200 * Documentation to use the tracing functions and how to analyze the
3201 traces with the Triva tool is written.
3202 * More information about: SimGrid FAQ (in the section Tracing Simulations
3205 * We moved to cmake as default build system. Autotools support will
3206 be dropped soon. Check the FAQ for more info about how to use it.
3207 * Greatly improved our cdash/ctest interactions
3208 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
3209 * Added memory checking tests with valgrind; lot of memleak fixing.
3210 This may be the first release of simgrid with so few memory issues
3211 * Added code coverage tests.
3212 Our coverage is still improvable, but at least we see it on cdash.
3214 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
3216 SimGrid (3.3.4) stable; urgency=low
3218 The "Desktop Grid needs love too" release (also called Xmas release).
3220 Models improvements:
3221 * Major speedup in the maxmin system solving by using lazy evaluation
3222 Instead of solving completely the maxmin system at each iteration,
3223 only invalidate (and recompute) the modified parts.
3224 This new feature is enabled in default models but you can try to
3225 turn it on with "--cfg:maxmin-selective-update=1" for other models.
3226 * Cas01 IMproved as default CPU model
3227 This CPU model is the same Cas01 model, but it uses the
3228 maxmin-selective-update flag and a heap structure to manage
3229 actions on SURF kernel.
3230 It reduces the complexity to find the next action to finish and,
3231 consequently, it's faster than the old Cas01.
3232 This is the new default CPU model (Cas01).
3233 * Rename the old Cas01 model to Cas01_fullupdate
3234 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
3235 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
3236 * CpuTI (CPU Trace Integration)
3237 A new CPU model whose objective is simulate faster when using
3238 availability trace files.
3239 Instead of using a full featured, over engineered maxmin system for
3240 CPU modeling, this model does the pre-integration of traces files
3241 to calculate the amount of CPU power available, and so, executes
3242 faster than the old CPU models.
3243 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
3244 * Use LV08 as default network model since it gives better accuracy
3245 for small messages and shouldn't change things for big ones.
3246 Use --cfg=network_model:CM02 to get the previous behavior.
3249 ******************************************
3250 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
3251 ******************************************
3252 * The new CPU model may changes simulations!
3253 The point is that events occurring at the exact same timestamp
3254 are not scheduled in the same order with the old and new
3255 version. This may be enough to completely change the execution
3256 of simulations in some cases.
3257 * The new network model will change simulations!
3258 This new model is more realistic than the previous one, so you
3259 should consider redoing your old experiments with this model.
3260 Sorry for the inconvenience.
3263 * Introduce the supernovae compilation mode
3264 When compiled that way, the whole SimGrid (or almost) is put in a
3265 single compilation unit and compiled in one shoot.
3266 This is to help gcc which has difficulties to inline stuff from one
3268 The speedup seem to be above 15%, althrough more tests are needed on
3269 amd64 to confirm that gain.
3272 * Port of MSG's mailbox on top of SIMIX network
3273 The put/get mechanism was greatly simplified on the way.
3276 * New SIMIX network module. Provides:
3277 - Mailbox: rendez-vous mechanism to find with who you want to speak
3278 - Synchronous send/recv: easier and hopefully faster since the
3279 logic is handled in the maestro process directly now
3280 - Asynchronous send/recv: you dreamt of it? It's here now
3281 Too bad that nobody cared enough to propagate the change to MSG.
3282 * Add semaphores as SIMIX synchronization mechanism.
3285 * new function SD_daxload(char*) to load a DAX file
3286 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3287 * Introduce typed tasks. Specify its kind and cost at creation.
3288 At scheduling, just give where it should be placed, and the cost
3289 for each involved resource is automatically computed.
3290 Existing constructors so far (more to come of course):
3291 - SD_task_create_comm_e2e() for end-to-end communication
3292 - SD_task_create_comp_seq() for sequential computation
3293 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3294 * new function SD_task_dump() for debugging display
3295 * new function SD_task_dotty(task,FILE*) writing to file the info
3296 about the task in dotty format
3297 * SD_task_dependency_exists() can now cope with having one of its
3298 arguments NULL. If so, it tests whether the other argument has any
3300 * Add getters on list of preceding/following tasks:
3301 SD_task_get_parents(task) and SD_task_get_children(task)
3302 * Add getters on amount of workstations and list:
3303 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3304 * Add getter on task kind: SD_task_get_kind(task)
3305 * Update the start_time and finish_time of tasks on completion/failure
3306 * Bugfix: Remove task from state swags when destroyed
3309 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3312 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3313 dynar with one separate thread per value of the dynar.
3314 * Change the prototype of xbt_thread_create(), sorry.
3315 Added a boolean parameter indicating whether we want to join this
3316 thread (used in SG only for now)
3317 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3320 * GTNetS wrappers should now be usable again (and betterly tested too)
3321 * Fix a major regression from 3.2 where the timeout provided to
3322 MSG_task_put_with_timeout() was used as absolute time before which
3323 the comm should be done.
3324 * Start to fix the <cluster> tag.
3325 - Internal links should be good now (beside of the loopback, which
3326 use the private link instead)
3327 - paths to the external world is still rather broken
3328 - the <route:multi> tag is just broken. Actually that's brain-dead.
3329 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3330 to make it less stupid
3331 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3332 * Fix a source-level compatibility glitch from 3.2: after defining
3333 MSG_USE_DEPRECATED, you can use the old name
3334 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3335 * Allow to compile from the SVN with automake 1.11
3336 * Fix some problems when using the "start_time" tag in deployment XMLs.
3337 * Fix #8569: XBT/synchro.h has redundant declarations
3338 * Fix #8563: MSG return values and exceptions
3339 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3340 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3341 Thanks to Robson Peixoto.
3342 * Fix a vicious bug in dictionaries inducing that some elements were
3343 not freed on xbt_dict_free()
3345 Portability report of this version:
3346 * Main portability targets:
3347 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3348 on (amd64/i386/ia64)
3349 - mac leopard on i386
3350 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3351 but nothing critical.
3352 * Other platforms: windows, AIX and others were not tested for this release
3354 Timing report of this version:
3355 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3356 scenario parameters). From 8h to a few seconds in desktop grid settings.
3357 * Supernovae brings about 25% speedup on i386.
3359 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3361 SimGrid (3.3.3) stable; urgency=low
3363 The "Need for Speed" release.
3365 The timings done to validate the 3.3.2 were faulty.
3366 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3368 The problem was a conversion from a manually handled vector to
3369 xbt_dynar_t on the critical path.
3370 xbt_dynar_foreach calls functions, inducing stack management crap.
3372 We inlined these functions and xbt_dynar_foreach is now breath taking.
3373 We also inlined xbt_swag_belong on the way.
3375 Here are some approximate speedup measurements (on master/slaves
3376 simulations lasting between 10s and 20s each):
3377 3.3.1 -> 3.3.2: about same performance
3378 3.3.2 -> 3.3.3: 40% speedup
3379 3.3.1 -> 3.3.3: 40% speedup
3380 3.3.1 with inline patch -> 3.3.3: 30% speedup
3382 Our reading is that the refactoring which occurred in 3.3.2 made us
3383 suffer much more from the xbt_dynar_foreach low performance, but
3384 once we solved this, this refactoring proved to be very performance
3385 effective. From the 40% speedup, somehow, 10% are due to the
3386 inlining and 30% to the refactoring.
3388 That's a pitty that gcc cannot inline functions placed in other files
3389 alone. We have to choose between:
3390 - break the encapsulation (by putting private data structures and
3391 accessors in headers files to help gcc)
3392 - live with low performance
3393 - switch to a decent compiler such as icc (not quite possible).
3395 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3397 SimGrid (3.3.2) stable; urgency=low
3399 The "Simplicity does not preceed complexity, but follows it" release.
3401 The main contributors of this release were (lexical order):
3402 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3405 * Extract the routing logic into its own object.
3406 (was dupplicated in network.c and workstation_LV07.c;
3407 Allows to implement other ways of storing that info)
3408 => kill now useless network_card concept
3409 - Use dynar to represent routes (instead of void** + int*)
3410 - kill link_set (use surf_network_model->resource_set instead)
3411 - Add a command-line option to choose the routing schema to use
3412 - Add three new models:
3413 * Floyd (shortest path computed at initialization)
3414 * Dijikstra (shortest path recomputed all the time)
3415 * Cached Dijikstra (shortest path computed on need)
3416 All these models where contributed by Silas De Munck, and are
3417 described in his ICCS09 paper.
3419 * Simplify model declaration
3420 (less redirections, less function to write when defining a model)
3421 - Factorize stuff between models:
3424 surf_model_resource_set(model)
3425 surf_model_resource_by_name(model, name)
3426 - Unify the types of models in s_surf_model_t (using an union)
3427 - Embeed fields of common_public directly into s_surf_model_t
3428 - Rename model methods:
3429 action_free ~> action_unref
3430 action_change_state ~> action_state_set
3431 action_get_state ~> action_state_get
3432 - Change model methods into functions:
3433 (model)->common_public->action_use ~> surf_action_ref
3435 * Implement a generic resource; use it as ancestor to specific ones
3436 (allows to kill duplicated code in models)
3437 Drawback: timer command don't need no name nor properties;
3438 workstation_CLM03 don't need no properties
3439 (but I guess we can live with those few bytes wasted)
3441 * Improve the action object model
3442 - implement a constructor avoiding dupplicated code about field
3443 initialization in generic_action part.
3445 * Kill the SDP model: it has an external dependency, is deprecated
3446 in flavor of modern lmm models, and didn't compile since a while
3449 * Relocation of the context module from XBT to SIMIX.
3450 (the context were decoupled from the simix processes, duplicating a lot of code)
3451 => a lot of code was factorized
3452 - less overhead is introduced during scheduling
3453 - simpler API for the context factory
3454 - the logic for process creation,destruction and manipulation was simplified
3455 * Simplification of the s_smx_process_t data structure.
3456 => accesing the simix level data associated to a process is faster now,
3457 and the code is a lot more readable.
3460 * Implement some more MPI primitives:
3461 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3462 -implementation: Bcast: flat or 2-ary tree (default),
3463 Barrier: 4-ary tree,
3465 Allreduce: Reduce then Bcast
3466 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3467 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3468 Alltoallv: flat tree, like ompi
3470 * Add support for optimized collectives (Bcast is now binomial by default)
3471 * Port smpirun and smpicc to OS X
3474 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3475 Shout out if you used it.
3478 * Display the list of still queued messages in SG mode when existing
3482 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3483 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3484 * Add xbt_str_from_file(FILE*)
3485 * Add xbt_dict_get_key achieving a linear reverse search
3486 * Remove the context module
3488 Portability report of this version:
3489 * Main portability targets:
3490 - Linux(debian)/x86/context
3491 - Linux(debian)/x86/pthreads
3492 - Linux(debian)/amd64/context
3493 - Linux(debian)/amd64/pthreads
3494 On these, we still have the eratic breakages of gras/pmm and
3495 amok/saturate_sg reported in previous version. We still think
3496 that the tests are the cause of the fault, not the tested code.
3498 - Mac OSX Leopard/x86/context
3499 Still false negative in tesh autotesting.
3500 Smpi still fails, but this time because readlink does not accept -f
3501 Everything seems to work properly beside of that.
3504 - AIX version 5.3 (only tested contexts this time)
3505 Smpi still fails there because mktemp is not installed.
3506 Everything seems to work properly beside of that.
3508 I managed to compile it for the first time, but several breakages.
3509 Won't delay the release for this exotic platform.
3511 * Windows: it's still lagging behind. If you want to help, please
3514 Timing report of this version:
3515 This version seem to be more than 5% faster than 3.3.1 (on linux
3516 64bits with contextes). The gain is less than expected, we are
3517 investigating this for next release.
3519 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3521 SimGrid (3.3.1) stable; urgency=low
3524 * Implement a --cfg-help to show existing configuration variables
3525 * Build chain do not require doxygen in maintainer mode
3528 * fix a bug on struct sizeof computation, which prevented the
3529 exchange of arrays of structs in some conditions
3530 - added a regression test about this in datadesc_usage
3531 * Allow the exchange of 0-long dynamic vectors.
3532 - for that, use -1 as indicator of dynamic size instead of 0
3533 - This implied to change any size from unsigned long to long,
3534 reducing a bit communication abilities, but I guess that with
3535 64bits being quite common, this is more than enough.
3536 - This also induce a protocol change, thus bumping network protocol
3537 version from 0 to 1 (if we have external users, we have to get
3538 clean on that point too ;)
3539 - added two regression tests about this in datadesc_usage
3540 * Be more verbose when propagating local exceptions
3541 This helps debugging.
3542 * Display the status of simulated processes when receiving SIGINT in
3546 * Allow to control the simulation from a trace file.
3547 New functions MSG_action_register() and MSG_action_trace_run()
3548 The first one allows to associate a function execution to each
3549 kind of action while the second one parses a trace file and
3550 triggers the corresponding actions within the system.
3551 For now, only a toy example is provided in examples/msg/actions
3552 * Add an example of process migration in examples/msg/migration
3553 * Fix a bug in task exchange which broke MSG_task_get_sender()
3554 Add a teshsuite regression test for that.
3555 [Bug: if MSG_task_get_sender() is called after sender exit,
3557 * Fix a bug which prevented suspend/resume to work properly
3558 * Display the status of simulated processes when receiving SIGINT
3559 This fixes a regression of v3.3. due to the introduction of SIMIX
3560 * Bug fixing in failure management:
3561 - trace could not start by a failure at time 0
3562 - failure during communications were not working
3565 * Add SIMIX_process_set_name() to change the name of the current
3566 process in the log messages.
3567 * Store smx_hosts in a dict since we only retrieve them by name
3568 * Move the configuration infrastructure to surf
3571 * Move the configuration infrastructure to surf
3574 * Massive internal cleanups:
3575 - Store internal structures on processes instead of hosts (allows
3576 to have more than one process per host, in addition of being more
3578 - Cleanup the initialization/finalization process
3579 - Kill a whole bunch of unneeded synchronization:
3580 processes run in exclusive manner within the simulator
3581 - Move queues from global tables to process data fields
3583 - now accept -platform and -hostfile arguments
3584 - Pass the right rank value to processes according to the hostfile
3585 * Compile the examples by default, and use them as regression tests
3586 * Implement MPI_Wtime()
3587 * Change the reference speed to a command line option
3590 * TCP_gamma can now be specified as command line option using
3591 --cfg=TCP_gamma:10000000.0
3592 * Change the --surf-path cmd line option into --cfg=path:
3595 * Also include strbuff from xbt.h public header
3596 * xbt_ex_display(): do not free the exception after displaying
3597 This allows to do more with the given exception afterward.
3598 Users should call xbt_ex_free() themselves.
3602 Portability report of this version:
3603 * Main portability targets:
3604 - Linux(debian)/x86/context
3605 - Linux(debian)/x86/pthreads
3606 - Linux(debian)/amd64/context
3607 - Linux(debian)/amd64/pthreads
3608 These targets fail about 1/10 of times on gras/pmm, but we believe
3609 that this is because of the test, not because of simgrid.
3610 amok/saturate_sg fails even more rarely, and the test may not be
3613 - Mac OSX Leopard/x86/context
3614 The test suite still spits tons of errors because some obscure
3615 force prevents us from removing the temporary directories
3616 arguing that they still contain some metadata I've never heard of.
3617 Smpi fails because seq is not installed.
3618 Everything seems to work properly beside of that.
3621 - AIX version 5.3 (both contexts and pthread)
3622 Smpi still fails there because mktemp is not installed.
3623 XML inclusions seems rosty on AIX.
3625 * Windows: it's still lagging behind. If you want to help, please
3628 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3630 SimGrid (3.3) stable; urgency=high
3634 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3635 [Malek Cherier & Mt]
3637 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3640 Until now, GRAS were using MSG as an interface to SURF. It was
3641 quite difficult because both interface have several differences
3642 (MSG channels vs GRAS sockets were the most notable point).
3644 This also opens the gate to SMPI (which should occur soon) and speed
3645 up simulations by to 40% (even if it were not the main goal).
3647 **************************************
3648 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3649 **************************************
3650 The point is that events occurring at the exact same timestamp are
3651 not scheduled in the same order with the old and new version. This
3652 may be enough to completely change the execution of simulations in
3653 some cases. Sorry for the inconvenience.
3655 * Cleanup and upgrade the XML format to push further scalability
3656 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3658 * Improve the testing infrastructure with tesh. Now a very large part of
3659 the code is tested not only by being run but also by checking that the
3660 output match an expected output [Mt].
3662 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3663 is really less memory-demanding, which should allow you to use
3664 larger files in SimGrid [AL].
3666 * Inform valgrind about our contextes, so that it becomes usable
3667 with the default (and more effecient) version of SimGrid
3668 [contributed by Sékou Diakite, many thanks]
3671 * Introduce a listener thread in charge of receiving incoming
3672 messages from the network. It allows to overlap communication and
3673 computation but most notably, it removes some stupid deadlocks due
3674 to the fact that so far, a process could not send and receive at
3675 the same time. This made most non trivial communication schema
3677 * Convert the PIDs from long int to int to match the MSG ones (and
3678 linux ones too) [Mt]
3679 * New function: gras_agent_spawn() to launch a new process on
3680 current host. Only working in simulation for now. [Mt]
3681 * New function: gras_os_hostport() returning a constant form (ie,
3682 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3685 * Make the backtrace of exceptions more human readable [Mt]
3686 * New module: xbt/str [Mt]
3687 a ton of string utility functions (split, join, printf to a newly
3688 allocated buffer, trim, etc)
3689 * New module: xbt/hash [Mt]
3690 SHA1 hashing algorithm (more to come if needed)
3691 * New module: xbt/synchro [Mt]
3692 synchronization tools (mutex and conditions) working the same way
3693 in simulation and in real life (mainly useful for GRAS, but not
3695 * New module: xbt/queue [Mt]
3696 classical producer/consumer synchronization scheme
3697 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3698 (using the classical functions will get serialized) [Mt]
3699 * Make dictionary internal table dynamic. No need to specify its size
3700 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3702 * Make sure the log channels are organized as a tree under windows
3703 (because of ANSI C compatibility issue, any channel were child of
3704 root directly) [Mt].
3707 * Cleaned many thing in surf and fixed a few bugs [AL].
3708 * Add a nice command line configuration mechanism to compose models [AL].
3709 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3710 the previous one (KCCFLN05). It relies on something that looks like
3711 a max-min sharing mechanism but cannot be written as such. A new solver
3712 was thus designed [AL].
3713 * Add a new solver to lmm. Based on Lagrange optimization and
3714 gradient-based descent, it enables to efficiently maximise systems s.a
3716 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3718 This solver enables to propose two new network models for TCP Reno and
3719 TCP Vegas based on Low's work. These models still need to be fully
3720 tested though [Pedro Velho].
3723 * Bug fix in SD_simulate. Now the time bound given as argument is
3725 * Use the new parallel task model (ptask_L07) as default.
3726 * Use the SURF command line configuration mechanism.
3727 * 0-size tasks (for synchronization) should now work.
3729 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3731 SimGrid (3.2) stable; urgency=high
3735 We still experience issues on this platform, but we believe that at
3736 least MSG is usable.
3738 GRAS API BREAKAGE (for simplification purpose, sorry):
3739 * the gras_msgtype_by_name is not used anymore. Instead of
3740 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3741 you can write (and must)
3742 gras_msg_send(toserver, "request", &request);
3743 - If you still want to pass a gras_msgtype_t to the function (to cache
3744 the type and avoid the lookup time), use the gras_msg_send_() variant.
3745 - Impacted functions:
3746 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3747 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3748 * The callbacks are now expected to return 0 when everything went well
3749 (just like the main() function)
3751 GRAS new features and improvements:
3752 * New module mechanism where user code can use per process globals [Mt]
3753 This is similar to gras_userdata_*() functions, but for libraries. It
3754 factorize some code developped over and over in the examples and AMOK.
3755 It has still to be documented and used (only amok/peermanagement is
3757 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3758 loss when they were small enough to fit into the buffer and sent quickly
3759 enough so that they can all get received in one shoot.
3760 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3761 if not found. Use the *_or_null() variant for the old semantic.
3762 * In gras_msg_handle, do not discard messages without callback.
3763 They are probably messages to be explicitly awaited later (ie, proofs of
3764 mis-synchronization in userland since they are sent before being awaited)
3765 No big deal usually.
3766 * gras_socket_meas_send/recv: semantic changed!
3767 The numerical arguments used to be (1) the total amount of data to send
3768 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3769 messages. This was need for the fool willing to send more than MAXINT
3770 bytes on quite fat pipes.
3773 * Do really rename the hostmanagement module to peermanagement. [Mt]
3774 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3775 but this is rather new and this was documented in the module
3776 documentation (poor excuses, I admit)
3777 * Bandwidth measurement semantic changed! This follows the changes to
3778 gras_socket_meas_send/recv explained above.
3781 * A sequential mode has been added to the workstations. When a workstation
3782 is in sequential mode, it can execute only one task, and the other tasks
3783 are waiting in a FIFO. [Christophe Thiery]
3786 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3787 model for SIMDAG. [Christophe Thiery]
3788 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3789 equal to 0, which caused some bad behaviors in
3790 saturated_constraint_set_update. I now use a threshold mechanism like in
3794 * When running manually src/testall, you select specific units [Mt]
3795 testall is the result of our cunit mechanism, and should replace all
3796 the scripty thingy around since bash don't run easily on billware.
3798 * A mallocator system has been added. [Christophe Thiery]
3799 Mallocators allow you to recycle your unused objects instead of freeing them
3800 and allocating new ones.
3802 Documentation update:
3803 * FAQ reworking + New FAQs:
3804 - "Valgrind spits tons of errors!" [Mt]
3805 - "How to repport bugs" [Mt]
3806 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3807 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3808 - Communication time measurement within MSG [AL]
3809 - I experience weird communication times when I change the latency [AL]
3810 * GRAS tutorial [Mt]
3812 - an introduction to the framework and to the used communication model
3813 - an initiatic tour introducing the most proheminent features:
3815 . Lesson 0: Installing GRAS
3816 . Lesson 1: Setting up your own project
3817 o Part 2: Message passing
3818 . Lesson 2: Exchanging simple messages
3819 . Lesson 3: Passing arguments to the processes (in SG)
3820 . Lesson 4: Attaching callbacks to messages
3821 . Lesson 5: Using globals in processes
3822 . Lesson 6: Logging information properly
3823 . Lesson 7: Using internal timers
3824 . Lesson 8: Handling errors through exceptions
3825 . Lesson 9: Exchanging simple data
3826 . Lesson 10: Remote Procedure Calling (RPC)
3827 . Lesson 11: Explicitly waiting for messages
3828 . Recapping of message passing features in GRAS
3829 - A HOWTO section containing:
3830 o HOWTO design a GRAS application
3831 More are due, of course. They will come latter. In the meanwhile, you can
3832 check the examples which are still here.
3834 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3836 SimGrid (3.1) stable; urgency=high
3840 There was a stack corruption somewhere, visible only when optimizing
3841 with these versions. [Vince]
3844 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3845 module that enabled to play with Directed Acyclic Graphs. It is built
3846 directly on top of SURF and provides an API rather close to the old
3847 SG. Some old codes using SG are currently under rewrite to check that
3848 all needful functions are provided. [Christophe Thiery]
3851 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3852 extension of the classical CLM03 model that gracefully handles
3853 failures. This is now the default model for MSG and GRAS. It doesn't
3854 handle parallel tasks yet though. [AL]
3855 * Bug fix: Weights were not correctly set in the network part.
3856 WARNING: This may have resulted in incorrect results with simulations
3857 where there are more than one flow on a given link. [AL]
3860 * After a (long ?) discussion on simgrid-devel, we have decided that the
3861 convention we had on units was stupid. That is why it has been decided
3862 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3863 WARNING: This means that all previous platform files will not work as
3864 such with this version! A warning is issued to ask users to update
3866 A conversion script can be found in the contrib module of the CVS, under
3867 the name contrib/platform_generation/surfxml_update.pl [MQ]
3870 * Bug fix: Processes were started in reverse order, wrt deployment file.
3871 WARNING: if your code relies on this bug, please fix it. [AL]
3872 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3873 being executed on two different locations. [AL]
3874 * Bug fix: Failures are now better supported thanks to Derrick's tests
3875 (there was many failure situations I hadn't thought of and that weren't
3876 correctly handled). [AL]
3877 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3881 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3882 Exception occurring on server-side are propagated back to client (!).
3884 API CHANGE: the callback changed their prototype. Change:
3885 int my_handler(gras_socket_t expeditor, void *payload_data) {
3887 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3888 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3890 * New! function: gras_msg_handleall to deal with all messages arriving
3891 within a given period.
3892 * New! function: gras_socket_server_range to get a server socket in a
3893 range of port numbers (ease to avoid port number conflicts) [MQ]
3894 * New! gras processes display their backtrace when they get a SIGUSR1
3895 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3896 Sweet to debug RL processes [MQ]
3900 - Do not force experiment sizes to be expressed in kb, or it becomes
3901 impossible to measure the latency this way (needs one byte-long tests)
3902 WARNING: this changes the amok_bw_* function semantic. [MQ]
3903 - Implements the link saturation stuff. [MQ]
3904 * Peer management module:
3905 New! module factorizing code that we wrote over and over [MQ].
3908 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3909 - Test units are placed directly into the library code, they get extracted
3910 automatically and placed into the src/testall binary.
3911 - Convert most of the XBT tests to this system.
3912 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3913 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3914 live in the same C code. This required to change a little bit the API
3915 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3916 * New module: sparse graph structure with basic algorithms (this is work
3917 in progress and the API is not considered to be frozen yet). [AL]
3918 * Display more information on backtraces: source line & function names are
3919 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3920 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3921 * Reworked a little bit some #include statements to load only required
3922 headers. Some user code that relied on SimGrid to include stdlib or
3923 stdio may need to include it by themselves. [AL]
3924 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3926 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3927 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3928 but I guess that almost nobody used those parts. [MQ]
3930 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3932 SimGrid (3.0.1) stable; urgency=low
3935 * Unfortunately, I had missed 5 misnamed functions:
3936 xbt_fifo_item_t xbt_fifo_newitem(void);
3937 void xbt_fifo_freeitem(xbt_fifo_item_t);
3938 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3939 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3940 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3941 They're now deprecated. Please use their new versions:
3942 xbt_fifo_item_t xbt_fifo_new_item(void);
3943 void xbt_fifo_free_item(xbt_fifo_item_t);
3944 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3945 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3946 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3948 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3949 * Documentation: xbt_log module unmercifully reworked [MQ]
3950 * Bugfix: there was a problem with the ending of contexts with
3951 the pthread backend. It caused some weird deadlock or behavior
3952 depending on the pthread implementation. [AL]
3953 * Bugfix: get the exceptions raised in the simulator repport where
3954 and why they come from when they are not caught in time [AL, MQ]
3957 * Bugfix: Do repport the error when two non-connected hosts try to
3958 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3961 * Add additionnal checkings on communications. Assert that two
3962 communicating hosts are connected by a set of links... [AL]
3965 * Add additionnal checkings on channel values in communication [AL]
3966 * New: MSG_task_get_source to see on which host a task was generated [HC]
3967 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3968 the number of tasks waiting to be received on channel and sent
3970 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3971 waits for the first task coming from a given host.. [AL]
3973 GRAS new functionnalities: [MQ]
3974 * Enhance the parsing macro to allow the size of multidimentional objects
3975 to be given thru annotations.
3976 * New example (and documentation): Matrix Multiplication a la RPC
3977 (as when I was young!) and fix a bunch of bugs found on the way.
3979 GRAS performance improvements: [MQ]
3981 * Reduce the amount of cbps creation/destruction by making it static to
3982 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3985 * Change libdata to a set so that we can search for stuff by ID (and thus
3986 reduce the insane amount of dict lookups)
3989 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3990 RL communication on top of SG since it's so uneffective.
3991 It may also be used for inter-thread communication in RL, one day.
3992 * Use gras_datadesc_copy() to exchange messages on top of SG
3994 - improve message exchange performance on top of SG
3995 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3996 * Don't exchange on the network the size of the used part of buffer,
3997 instead, specify the possible buffer size to read().
3999 - reduces the amount of read/write calls (one pair per exchange)
4000 - reduces the amount of exchanged data (the size)
4001 - allows to retrieve all arrived data on receiver side, if we don't need
4002 it right now (subsequent read will peek the buffer)
4003 - allows the receiver to proceed with the beginning of the stream before
4004 everything is arrived
4005 - make it possible to build an iov transport (using readv/writev)
4007 - take care of the data with non-stable storage (like stacked data),
4009 * If possible, TCP send uses vector I/O (when writev() is here)
4010 - Don't use it for receive since we send data sizes and data on the
4011 same stream, so we wouldn't be able to chain large amount of chunks
4012 before having to flush the stuff to read the size.
4013 * Rework the transport plugin mechanism to simplify it and reduce the
4014 amount of pointer dereferencement when searching for the right function
4017 * I guess that now, we do almost as few system calls as possible while
4018 doing as few data copy as possible.
4020 To improve it further, we could try to send all the sizes first and then
4021 all the data (to use iov on receiving size), but it's only a partial
4022 solution: when you have 2 dimensional data, the sizes of the second
4023 dimension is data of the first dimension, so you need 3 streams.
4025 I'm not sure the potential performance gains justify the coding burden.
4027 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
4029 SimGrid (3.00) stable; urgency=high
4032 * New! Give the possibility to hijack the surf parser and thus bypass
4033 MSG_create_environment and MSG_launch_application. Have a look at
4034 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
4036 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
4038 SimGrid (2.96) unstable; urgency=low
4043 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
4044 This deprecates the xbt_error_t mechanisms.
4045 It modifies (simplifies) all XBT and GRAS API.
4046 MSG API keeps unchanged (exceptions raised by XBT are caught from
4047 within MSG and masked with existing error handling facilities)
4050 * New! Add a FATPIPE model. [AL]
4051 * New! Add a parallel task model. [AL]
4052 * New! Add automatically a loopback interface (in the default
4053 network model) if none was precised.
4056 * Bugfix: MSG_process_resume now works with the current running process.
4058 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
4059 * Modification of MSG_task_get_compute_duration. Once a task has been
4060 processed, the value returned by this function is now equal to 0. [AL]
4061 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
4062 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
4063 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
4064 of a m_task. For now, MSG_task_cancel only works with computation
4066 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
4067 of the processor (in Mflop/s) regardless of the current load on the
4069 * API Change: use proper naming convention for MSG_getClock and
4070 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
4072 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
4073 This function changes the priority of a computation task. This priority
4074 doesn't affect the transfer rate. A priority of 2 will make a task
4075 receive two times more cpu power than the other ones. This function
4076 has been added to suit the needs of Nguyen The Loc and hasn't been that
4077 much tested yet. So if it fails, please report it and send me your code.
4079 * API Change: removed all functions and types that were marked "deprecated"
4080 since many months. Renamed MSG_global_init_args to MSG_global_init.
4082 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
4084 SimGrid (2.95) unstable; urgency=low
4087 * Steal some nice code to GNU pth to fix context detection and usage [AL]
4088 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
4089 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
4092 * Bugfix: Allow absolute paths for platform description files [MQ]
4093 * Bugfix: do free the variables after use. Leads to drastic performance
4095 * Implement max_duration (ie, timeouts) on resources [AL]
4098 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
4100 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
4101 the needed support (timeouts) [AL]
4104 * Implement measurement sockets. You can now get the bandwidth between two
4105 hosts thanks to AMOK (see below). [MQ]
4106 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
4107 dynar over the network (yeah) [MQ]
4108 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
4109 * simple/static token-ring example. [Alexandre Colucci and MQ]
4110 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
4111 in place before [MQ]
4114 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
4115 type parsing. "short" and "long" modifiers were also ignored; other
4116 modifier (such as reference level) are still ignored. [MQ]
4117 * Embeed the buffer size within the buffer itself on SG. [MQ]
4118 That way, send() are atomic and cannot get intermixed anymore (at least
4119 the ones which are less than 100k; bigger messages still have the issue)
4120 * Array size pushed by the field, not by the field type (or each
4121 and every long int will push stuff to the cbps) [MQ]
4122 * use select() to sleep since it allows to portably sleep less than one
4125 GRAS (minor cleanups)
4126 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
4129 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
4130 * type category 'ignored' killed as it was never used and were difficult
4132 * whether a type can cycle or not is now a flag, leaving room for more
4133 flags (as "ignored", if we feel the need one day ;)
4134 * Rename raw sockets to measurement sockets since "raw" has another
4135 meaning in networking community.
4138 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
4139 and offers features not belonging to GRAS but that most applications
4140 need. One day, it may be a set of plugins loadable at runtime.
4141 * New module: bandwidth
4142 bandwidth measurement between arbitrary nodes running this module. [MQ]
4144 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
4146 SimGrid (2.94) unstable; urgency=low
4148 The first beta release of SimGrid 3 !
4152 * Update the main page and the FAQ. Adding references to gforge.
4155 * Add a gras_os_getpid function.
4158 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
4159 * Extend the logs so that they also print PID, hostname, date, ... if
4161 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
4162 and kill the old version which were in testsuite/
4163 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
4167 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
4168 for users. Instead, each programming environment (GRAS, MSG,...) use it
4169 when needed to provide such a feature to users.
4170 Don't shortcut the mechanism or you will also shortcut the virtualization
4171 you need on the simulator.
4175 * Cleanups in configury with regard to compile optimization/warning flags.
4176 Also add -fno-loop-optimize to any powerpc since it's the optimization
4177 killing gcc (< 3.4.0).
4178 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
4179 needed by MSG examples complications
4180 * Borrow configury beautifications from PHP
4183 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
4184 warning (thanks loris for stumbling into this one).
4185 * Bugfix: stop loading private headers (gras_config.h) from the public
4189 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
4190 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
4191 RL would imply a DNS resolver, which is *hard* to do in a portable way
4192 (and therefore delayed).
4193 * Implement a real timer mechanism and use it in timing macros. This allows
4194 to avoid rounding errors and get a 0.000005 sec precision in timing
4195 macros. While I was at it, various cleanups:
4196 - allow to declare more than one timed section per file (fix a stupid bug)
4197 - move some private declaration to the right place
4198 - merge conditional execution and timing macros into emulation module
4199 - document the module
4200 - make sure the module cleanups its mess on gras_exit
4201 * Documentation improvements:
4202 - (new) how to compile applications using GRAS
4203 - (new) emulation support (timing macros)
4205 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
4207 SimGrid (2.93) unstable; urgency=low
4209 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
4212 - Use Paje properly where used. Still to be sanitized properly.
4213 - Portability fix: Add an implementation of the contexts using pthread
4217 - Add xbt_procname(): returns the name of the current process.
4218 Use it to show the current process's name in all logging.
4220 - fix detection of older flex version and the reaction, since we do
4221 depend on modern ones (we use lex_destroy)
4222 - Better separation of SG and RL in the libs: remove all simulation code
4223 from libgras. As a result, this lib is now only 200k when stripped.
4224 Some of the xbt modules may also be duplicated (two sets and such) and
4225 should be cleaned/killed before SG3.
4226 - Insist on using xlC on AIX because of weird problems involving gcc there.
4227 - Cleanup the make remote stuff. This is now done by scripts
4228 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
4229 mainly for our private use, but we're working on changing them to user
4232 - Bugfix: flush the socket on close only if there is some *output*.
4233 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
4235 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
4238 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
4240 SimGrid (2.92) unstable; urgency=low
4242 Alpha 3 on the path to SimGrid 3
4246 - New! First try of benchmarking macros.
4247 - New! First try so that gras_stub_generator generate deployment and
4248 remote compilation helpers.
4250 - Bugfix: Initialization fix in msg_test.
4254 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
4256 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
4258 - API Change: gras_os_sleep to take the amount of seconds as a double.
4259 Accepting an int was error prone since it was the only location where
4260 seconds were coded as such. It leaded to damn rounding errors.
4261 - Bugfix: Hard to belive that timers ever worked before this.
4263 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4265 SimGrid (2.91) unstable; urgency=low
4267 Alpha 2 on the path to SimGrid 3
4271 - Bug fix in the lmm_solver.
4273 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4274 through the function MSG_paje_output.
4275 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4276 - It is possible to bound the rate of a communication in MSG with
4277 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4278 to write it in the changelog).
4279 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4284 - Various cleanups to the autotools stuff
4285 - Begin to move Gras examples to examples/gras/
4286 - Let make distcheck work again (yeah!)
4288 - documentation overhauled using doxygen.
4289 gtk-doc-tools is dead in SimGrid now.
4290 - Automatically extract all existing logging categories, and add the list
4291 to the documentation (long standing one, to say the less)
4293 - Cleanup the known architecture table. Reorder the entries to group what
4294 should be, and use a more consistent naming scheme.
4295 (some of the test dataset are still to be regenerated)
4296 - New! Allow library to register globals on each process just as userdata
4298 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4299 lookup time (for now).
4300 Use it in msg and trp.
4301 This cleans a lot the internals and helps enforcing privacy of the
4302 headers between the gras components.
4303 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4304 - Bugfix: gras_os_time was delirious in RL.
4305 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4307 - Reenable GRAS now that it works.
4309 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4311 SimGrid (2.90) unstable; urgency=low
4313 Alpha 1 on the path to SimGrid 3
4315 * It is a long time since the last release of SimGrid. I'm sorry about
4316 that but as I had told you, I was rewriting a lot of things. I apologize
4317 to those who had been reporting bugs to me and that I had not answered.
4318 If your bug is still in the new version, please tell me. Here is a
4319 summary of the main changes.
4321 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4322 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4323 portability, speed, and a lot more but you'll figure it out later.
4324 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4325 I wanted to release the new MSG as soon as possible and I have
4326 broken GRAS, which is the reason why, for now, only MSG is fully
4327 functional. A laconic description of these projects is available
4328 in the documentation.
4330 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4331 kernel optimized for our needs (called SURF but only the developers
4332 should use it). Hence, MSG is now roughly 30 times faster and I think
4333 that by rewriting a little bit MSG, I could event speed it up a little
4334 bit more. Beside the gain in speed, it is also much easier to encode a
4335 new platform model with SURF than it was with SG. More to come...
4337 * REVOLUTION 3: I have tried to change a little as possible the API of
4338 MSG but a few things really had to disappear. The main differences
4339 with the previous version are:
4340 1) no more m_links_t and the corresponding functions. Platforms are
4341 directly read from an XML description and cannot be hard-coded
4342 anymore. The same format is used for application deployment
4343 description. The new format is described in the documentation.
4344 Have a look in tools/platform_generation. There is a tiny script
4345 that converts from the old platform format to the new one. Concerning
4346 the application deployment format, parsing the old one is tricky.
4347 I think most of you should however be able to convert your files. If
4348 it is really an issue, I can write a C code that does the conversion.
4350 2) the toolbox tbx does not exist anymore. We now have a library
4351 with much more data-structures but without the hash-tables (we have
4352 dictionaries that are much faster).
4354 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4356 *****************************************************************************
4357 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4358 * don't feel like converting the oldies. (Mt) *
4359 *****************************************************************************
4362 Version 2.90: "the long awaited one"
4363 - Finished rewriting and debugging MSG. Rewrote the documentation.
4364 - disable GRAS for now since it needs to be ported to the newest SG
4367 - Finish the port to windows (using mingw32 for cross-compile)
4370 - Main loop and datastructures of SURF. A cpu resource object is
4371 functional. Surf can thus be used to create cpu's with variable
4372 performance on which you can execute some actions.
4374 2004-11-15 Martin Quinson
4375 - Port to ARM. Simply added the alignment and size descriptions. Should
4376 work, but the ARM machines are so slow that I didn't had the opportunity
4377 to 'make check' over there yet.
4379 2004-11-15 Arnaud Legrand
4380 - Trace manager now written. It uses a heap structure and is therefore
4381 expected to be efficient. It may however be speeded up (particularly
4382 when many events occur at the same date) by using red and black
4383 trees. One day maybe...
4384 - Max-min linear system solver written. It uses a sparse matrix
4385 structure taking advantage of its expected use. Most operations are
4386 O(1) and free/calloc are called as few as possible. The computation of
4387 the minimum could however be improved by using a red and black tree
4390 2004-11-03 Arnaud Legrand
4391 - Rename every gras_* function that was in xbt/ to its xbt_
4393 - Add a heap and a doubly-linked list to xbt
4394 - Added a dichotomy to the dictionaries. make check works as well before
4395 so I assume that the patch is correct. I do not know however if things
4396 run effectively faster than before now. :)
4398 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4399 SimGrid, and the version number is bumped to 2.x
4401 2004-10-29 Martin Quinson
4402 - Introduction of the remote errors.
4403 They are the result of a RMI/RPC on the remote machine.
4404 ErrCodes being scalar values, you can't get the host on which those
4405 errors did happen. Extending the error mechanism as in Gnome is possible.
4406 No idea yet whether it is a good idea.
4408 2004-10-28 Martin Quinson
4409 - Interface revolution: the Starred Structure Eradication.
4410 I used to do typedef struct {} toto_t; and then handle *toto_t.
4411 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4412 - ???_t is a valid type (builded with typedef)
4413 - s_toto_t is a structure (access to fields with .)
4414 - s_toto is a structure needing 'struct' keyword to be used
4415 - e_toto_t is an enum
4416 - toto_t is an 'object' (struct*)
4418 typedef struct s_toto {} s_toto_t, *toto_t;
4419 typedef enum {} e_toto_t;
4420 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4421 s_toto_t) is private.
4423 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4424 so that it can set the variable to NULL. It was so for dicts and sets,
4425 it changed for dynars.
4427 - Fix a bunch of memleaks in dict_remove
4428 - Fix a bug in sg/server_socket opening: it failed all the time.
4430 2004-10-07 Martin Quinson
4431 - Speed up dynar lookup operation a bit.
4433 gras_dynar_get is dead.
4435 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4436 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4437 (which gives you the address of the stored data).
4439 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4440 copy of the data using an affectation to do the job and not a memcpy.
4442 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4443 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4445 It does not really speedup the dynar test because they are
4446 setting elements all the time (and look them seldom). But the dict does
4447 far more lookup than setting.
4449 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4451 2004-10-05 Martin Quinson
4452 - Allow to (en/dis)able the cycle detection at run time.
4454 Whether we should check for cycle or not is now a property of each
4455 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4456 datadesc_cycle_unset allow to remove this property when previously set.
4458 Note that the cycle detection is off by default since it impacts the
4459 performance. Watch the data you feed GRAS with ;)
4461 This property is hereditary. Any element embedded in a structure having it
4462 set have it set for the time of this data exchange.
4464 You should set it both on sender and receiver side. If you don't set it on
4465 sender side, it will enter an endless loop. If you forget on receiver
4466 side, the cycles won't be recreated after communication.
4468 - Header reorganization.
4469 Kill gras_private.h, each submodule must load the headers it needs.
4471 2004-10-04 Martin Quinson
4472 - Interface revolution: do not try to survive to malloc failure.
4474 Now, gras_malloc and friends call gras_abort() on failure.
4475 As a conclusion, malloc_error is not a valid error anymore, and all
4476 functions for which it was the only gras_error_t return value are
4477 changed. They now return void, or there result directly.
4478 This simplify the API a lot.
4480 2004-09-29 Martin Quinson
4481 - Re-enable raw sockets.
4482 Created by gras_socket_{client,server}_ext;
4483 Used with gras_raw_{send,recv}
4486 It should allow to kill the last bits of gras first version soon.
4488 This is not completely satisfactory yet (duplicate code with
4489 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4492 - Simplify transport plugin (internal) interface by not passing any
4493 argument to _server and _client, but embedding them in the socket
4496 2004-09-28 Martin Quinson
4497 - Finish the port to AIX.
4498 autoconf was my problem (segfault within the malloc replacement
4499 function. No idea why)
4501 2004-09-16 Martin Quinson
4502 - Fix some size_t madness on 64bit architectures.
4504 2004-09-08 Martin Quinson
4505 - Reduce the number of system headers loaded, overload some more system
4506 calls (such as malloc to cast the result of the system one, and work
4508 - Fix and reintroduce the config support
4510 2004-09-07 Martin Quinson
4511 - Source code reorganization to allow Arnaud to surf all over there.
4512 - Allow to document the logging categories.
4513 - Remove all uppercase from logging categories and useless cleanup in names.
4515 2004-08-18 Martin Quinson
4516 Version 0.6.2 (protocol not changed; API changed)
4517 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4518 gras_error_t), and NULL when not found. Functions expecting a msgtype
4519 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4520 hopefully usefull message.
4521 - Portability to prehistoric sparcs again
4523 2004-08-17 Martin Quinson
4524 Version 0.6.1 (protocol not changed; ABI not changed)
4525 - prealloc some buffers to speed things up
4527 2004-08-11 Martin Quinson
4528 Version 0.6 (protocol not changed; ABI expended)
4529 - The parsing macro can deal with the references, provided that you add
4530 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4532 2004-08-09 Martin Quinson
4533 Version 0.5 (protocol not changed; ABI changed)
4534 - Allow to off turn the cycle detection code in data exchange at
4535 compilation time. It should be at run time, but I'm short of time (and
4536 the config stuff is still broken). That way, we keep dict out of the
4537 critical path, which is good because the performance is poor:
4538 - search not dichotomial yet
4539 - dynar give no way to access their content and memcpy everytime
4540 - In composed data description (struct, ref and so on), stop foolness of
4541 keeping the subtype's ID, but store the type itself. This keeps sets out
4542 of the critical path, which is good since they rely on dynar and
4543 dictionaries. The only loose of that is that we cannot detect the
4544 redeclaration of a structure/union with another content (but I'm not sure
4545 the code detected well this error before anyway). We still can detect
4546 the redefinition discrepancy for the other types.
4547 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4548 it breaks the code because of type-punning used all over the place).
4549 This breaks on all non-gcc architectures (for now).
4551 All those changes (plus the buffer of last time) allow me to gain 2 order
4552 of magnitude on cruel tests consisting of 800000 array of integers on two
4553 level of a hierarchical structure (200 secondes -> 4 secondes)
4556 - the selector of reference must now return the type it points to, not
4557 the ID of this type.
4559 2004-08-06 Martin Quinson
4560 Version 0.4 (protocol changed; ABI not changed)
4561 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4562 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4563 - Implement a buffer transport, and use it by default (it relies on tcp in
4564 real life and on sg in simulation).
4565 That's a bit hackish since I had a new field to the structure to store
4566 its data without interfering with the subtype ones. Inheritance
4567 is tricky in C. And that's a kind of reverse inheritance with one class
4568 derivating two classes. Or maybe a game with java interfaces. Anyway,
4569 that's damn hard in C (at least).
4570 Moreover, I got tired while trying to ensure plugin separation and
4571 genericity in SG mode. MSG wants me to do weird things, so let's go for
4572 cruel hacks (temporarily of course ;).
4573 See comment in transport_private.h:71
4574 - do not include all the _interface headers in private but in the files
4575 which really need them (to cut the compilation time when they are
4578 2004-07-26 Martin Quinson
4579 Version 0.3 (protocol not changed; ABI changed)
4580 - Major overhault of the datadesc interface to simplify it:
4581 - shorted the function names:
4582 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4583 - add a trivial way to push/pop integers into the cbps without malloc.
4584 This allows to make really generic sub_type description, which simply
4585 pop their size of the stack.
4586 - add a function gras_datadesc_ref_pop_arr() which does what users want
4587 most of the time: Declare a dynamic array (which pops its size of the
4588 stack) and declare a reference to it. Poor name, but anyway.
4589 - kill the post-send callback, add a post-receive one
4591 2004-07-23 Martin Quinson
4592 Version 0.2 (protocol changed; ABI changed)
4593 - add some testing for cpbs in the test cases, and fix some more bugs.
4594 This invalidate again the little64 data file, since I cannot regenerate
4596 - remove an awfull optimization in the logging stuff, allowing me to:
4597 - understand it again
4598 - learn gcc how to check that the argument match the provided format
4599 - fix all errors revealed by gcc after that
4600 - internal keys of dict are not \0 terminated. Deal with it properly in
4601 loggings instead of segfaulting when the user want to see the logs :-/
4603 2004-07-22 Martin Quinson
4604 - Fix some stupid bug preventing cbps (callback postit) from working
4606 2004-07-21 Martin Quinson
4607 - Some documentation cleanups
4608 - remove the useless last argument of msgtype_declare
4609 - rename the Virtu functions to fit into the 'os' namespace
4610 - move headers src/include -> src/include/gras/ and stop fooling with
4611 gras -> . symbolic link
4612 - make distcheck is now successful
4614 2004-07-19 Martin Quinson
4616 - Build shared library also
4617 - Install html doc to the right location
4618 - stop removing maintainer files in make clean
4619 - build tests only on make check
4621 2004-07-13 Martin Quinson
4623 - No major issue in previous version => change versioning schema
4624 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4625 corresponding dataset.
4627 2004-07-11 Martin Quinson
4629 - Get it working with any kind of structure (we can compute the padding
4630 bytes remotely for all the architectures I have access to)
4631 - Implement the structure parsing macro (still not quite robust/complete)
4632 - Improvement to the remote testing toysuite
4634 2004-07-10 Martin Quinson
4635 [autoconf mechanism]
4636 - get ride of a bunch of deprecated macros
4637 - actually run the test for two-compliment, not only compile it :-/
4638 - test whether the structures get packed (and bail out if yes. Damn.
4639 Alignment is a serious matter)
4640 - test whether the structures get compacted (but respecting the alignment
4641 constraints of each types)
4642 - test whether the array fields of structures can straddle alignment boundaries
4644 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4645 (took me 2 hours to find that bug, looking at the wrong place)
4646 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4647 before sending/receiving and are used to compute the offsets of fields
4648 - Given that padding size depend even on compiler options, keep track of
4649 alignment and aligned_size only for the current architecture. Not a big
4650 deal since we send structure fields one after the other (seems
4652 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4653 program, let it work on linux/gcc/little32. portability todo.
4655 2004-07-08 Martin Quinson
4656 - import and improve remote compilation support from FAST
4657 - make sure make check works on half a dozen of machines out there
4659 2004-07-07 Martin Quinson
4660 Let's say it's version 0.0.3 ;)
4661 - Implement conversions (yuhu!)
4662 - Let it work on solaris (beside conversion, of course)
4663 - Stupid me, using rand() to generate the conversion datatests in not wise.
4665 2004-07-06 Martin Quinson
4666 - Let make dist work, since I'm gonna need it to compile on remote hosts
4667 - Let Tests/datadesc_usage write the architecture on which the file was
4668 generated as first byte.
4669 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4670 architecture descriptions.
4671 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4672 on those architectures.
4673 - Optimization: send/recv array of scalar in one shoot
4675 2004-07-05 Martin Quinson
4676 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4678 - Plug a whole bunch of memleaks
4679 - each process now have to call gras_{init,exit}. One day, their log
4680 settings will be separated
4681 - Continue the code factorisation between SG, RL and common in Transport.
4683 2004-07-04 Martin Quinson
4685 - Redistribution between SG and RL.
4686 We wanna have to accept in SG, so move accepted related parts of RL in
4687 the common part. (more precisely, the dynar of all known sockets is no
4688 more a static in transport.c, but part of the process_data)
4690 [gras_stub_generator]
4691 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4693 2004-07-03 Martin Quinson
4694 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4695 them all up in one shot)
4697 - Fix the ugly case of reference to dynamic array.
4698 - New (semi-public) function gras_datadesc_size to allow the messaging
4699 layer to malloc the needed space for the buffer.
4701 - gras_socket_close now expect the socket to close (and not its address to
4702 put NULL in it after it). This is because the socket passed to handlers
4703 is one of their argument (=> not writable).
4705 - propagate the interface cleanup from last week in datadesc, ie remove a
4706 superfluous level of indirection. User pass adress of variable
4707 containing data (both when sending and receiving), and not of a variable
4708 being a pointer to the data. Let's say that I like it better ;)
4709 The price for that is constructs like "int msg=*(int*)payload" in
4710 handlers, but it's a fine price, IMHO.
4712 - Let it work in RL (yuhu)
4714 2004-06-21 Martin Quinson
4716 - porting SG plugin and SG select to new standards (works almost).
4717 - plug memleaks and fix bugs around.
4720 - cleanup the prototype of data recv and force users to specify when they
4721 want to handle references to objects. Test case working even for cycles.
4722 - plug memleaks. Valgrind is perfectly ok with this.
4724 2004-06-12 Martin Quinson
4726 - cleanup the separation between plugin and main code in plugin creation
4728 2004-06-11 Martin Quinson
4730 - Reput hook for raw sockets, needed for BW experiments
4731 - kill a few lines of dead code
4732 [Data description] Interface cleanup
4733 - gras_datadesc_by_name returns the searched type or NULL.
4734 That way, no variable is needed to use a type desc once, which makes
4736 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4737 two parameters were strings (field name, type name), leading to
4739 [Dicos] Interface cleanup
4740 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4741 This is consistant with the dynar API.
4743 2004-04-21 Martin Quinson
4745 - Porting to new standards.
4747 - interface cleanup.
4748 There is no bag anymore, no need to take extra provision to mask the
4749 pointers behind "ID".
4750 Better splitup of functions between files create/exchange/convert.
4751 This is still a bit artificial since convert and receive are so
4752 interleaved, but anyway.
4754 - add a queued message list to procdata (the ones not matching criteria
4756 - factorize some more code between SG and RL wrt procdata
4758 - use gras_exit in example to track memleaks
4759 - get rid of gs_example now that GS is properly integrated into gras
4760 - update run_test to integrate the latest tests (datadesc)
4762 - rename WARNINGn macros to WARNn since it prooved error-prone
4764 2004-04-19 Martin Quinson
4766 - register init/exit functions within gras module mechanism
4767 - send/receive function.
4768 Convertion is not implemented, but short-cutted if not needed.
4769 struct/array elements are sent one by one (instead of block-wise), but
4770 nobody really cares (yet). Get a prototype before optimizing.
4771 - tests (using a file socket) for DD send/receive on:
4772 - base types: int, float
4773 - array: fixed size, string (ie ref to dynamic string)
4774 - structure: homogeneous, heterogeneous
4775 - chained list, graph with cycle
4776 Believe it or not, valgrind is not too unhappy with the results. The
4777 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4781 - Bugfix when using a filename explicitly (instead of '-')
4783 2004-04-09 Martin Quinson
4785 - factorize more code between RL and SG in socket creation
4786 - Complete the implementation and tests of:
4788 o file (only in RL, and mainly for debugging)
4790 I lost 3 days to design a portable address resolver, and then decided
4791 that the prototype mainly have to run on my box.
4792 Addressing portability too early may be like optimizing too early :-/
4794 - use gras_init in the Tests instead of the crappy parse_log_opt
4795 (the latter function is removed)
4796 [Conditional execution]
4797 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4798 [Code reorganisation]
4799 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4800 Build examples against the RL library, since there is no way to disable
4801 its creation for now.
4803 For information, the beginning of coding on GRAS was back in june
4804 2003. I guess that every line has been rewritten at least twice since