1 SimGrid (3.32.1) not released yet (target december 22)
4 - Remove the MSG API: its EOL was scheduled for 2020.
5 - Remove the Java bindings: they were limited to the MSG interface.
6 - On Windows, you now need to install WSL2 as the native builds are now disabled.
7 It was not really working anyway.
10 - Activity::set_remaining() is not public anymore. Use for example
11 Comm::set_payload_size() to change the size of the simulated data.
14 - optimize an internal datastructure, leading to a potentially big
15 performance gain (in particular with many detached comms)
18 - New option smpi/barrier-collectives to add a barrier to some collectives
19 to detect dangerous code that /may/ work on some MPI implems.
22 - Write the section of the manual about models, at least.
23 - WiFi: the total capacity of a link depends on the amout of flows on that link.
24 - Use the nonlinear callback feature of LMM to reflect this.
25 - Calibration values can be changed to match different MCS configurations
26 - See the example teshsuite/models/wifi_usage_decay/wifi_usage_decay.cpp
27 - See also "A Flow-Level Wi-Fi Model for Large Scale Network Simulation"
28 https://hal.archives-ouvertes.fr/hal-03777726
29 - Merge parameters network/bandwidth-factor and smpi/bw-factor that serve the same purpose.
30 - Same for the latency
31 - Rewrite the corresponding documentation.
32 - Allow to disable the TCP windowing modeling by setting network/TCP-gamma to 0.
35 - Implement pthread_join in MC mode.
38 - New section in the user guide on the provided performance models.
39 - New section presenting some technical good practices for (potential) contributors.
41 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
42 (FG: issues on Framagit; GH: issues on GitHub)
43 - FG#18: Java bindings should be redone or removed
44 - FG!118: Wi-Fi callback mechanism
45 - FG!119: SMPI: add option to inject a barrier before every collective call, to allow...
47 ----------------------------------------------------------------------------
49 SimGrid (3.32) October 3. 2022.
51 The Wiedervereinigung release. Germany was reunited 32 years ago.
54 - SimGrid now requires a compiler with C++17 support to compile the lib.
55 Our public headers still allow the user code to be compiled in C++14.
56 - Support graphviz v3 and ns-3 v3.36 (older versions are still supported).
57 - Tested with clang (v11, v13, v14 and v16), gcc (v7 to v13) and IntelCC v2022.2
61 - kill signal Comm::on_completion that was not working anyway.
62 - Expose signals Activity::on_suspend and Activity::on_resume
63 - New macro xbt_enforce(): similar to xbt_assert(), but throws an AssertionError
64 instead of calling abort().
65 - New: s4u::Exec::get_thread_count()
66 - Various cleanups around virtual machines:
67 - host_by_name() and friends now only return hosts. VMs are now excluded.
68 - It is now impossible to search a VM by name globally.
69 You can only search VM by name on a given PM, so either you know
70 the PM on which your VM runs and you can search by name, or you need
71 to manually iterate over all PMs to search this VM.
72 - The s4u::VirtualMachine constructor is now deprecated.
73 Please use s4u::Host::create_vm() instead.
74 - Rename s4u::VirtualMachine::on_creation() to on_vm_creation() to
75 avoid confusion with s4u:Host::on_creation() that is inherited.
76 Also s4u::VirtualMachine::on_destruction -> on_vm_destruction().
78 - One-sided communications (Comm::sendto) can now be detached,
79 and should now be more resilient to network and host faults.
82 - Added the following bindings / examples:
83 - Comm (now 100% covers the C++ interface):
84 - Comm.dst_data_size, Comm.mailbox, Comm.sender, Comm.start_time, Comm.finish_time
85 - Comm.state_str [examples: examples/python/comm-failure/, examples/python/comm-host2host/]
86 - Comm.remaining [examples: examples/python/comm-host2host/, examples/python/comm-suspend/]
87 - Comm.set_payload_size [example: examples/python/comm-host2host/]
88 - Comm.set_rate [example: examples/python/comm-throttling/]
89 - Comm.sendto, Comm.sendto_init, Comm.sendto_async [example: examples/python/comm-host2host/]
90 - Comm.start, Comm.suspend, Comm.resume [example: examples/python/comm-host2host/]
91 - Comm.test_any [example: examples/python/comm-testany/]
92 - Comm.wait_until [example: examples/python/comm-waituntil/]
94 - Engine.host_by_name [example: examples/python/comm-host2host/]
95 - Engine.mailbox_by_name_or_create [example: examples/python/comm-pingpong/]
97 - Mailbox: Mailbox.ready [example: examples/python/comm-ready/]
98 - Ptask [example: examples/python/exec-ptask/]:
99 - this_actor.exec_init
100 - this_actor.parallel_execute
103 - Added an AssertionError exception that may be thrown in case of error.
104 For instance, creating two hosts with the same name will now throw this exception
105 instead of killing the interpreter.
108 - Implement MPI_File_get_type_extent(), MPI_File_s/get_atomicity() and
109 MPI_File_get_byte_offset()
110 - Intercept getpid() calls to return the simulated ones.
111 - Fix various bugs in MPI IO.
113 Platform description & visualization:
114 - More robust sanity checks for platforms, to reject forbidden topologies with
115 a proper error message.
116 - New platform example: supernode.cpp and supernode.py.
117 The Python version generates a nice graphical representation of the platform.
118 - Bug fixes around fat-tree topologies.
119 - Allow to dump the platform topology as a CSV file representing the graph edges
120 with platform_graph_export_csv() (similar to the DOT export).
121 - Fix graphicator for "cluster" topologies (e.g. fat-tree, dragonfly).
124 - Fix a bug when using ptasks with multicores (FG!111).
127 - First bits of sthread, that intercepts pthread operations at runtime.
128 The intend is to use it together with simgrid-mc, but it is TBD.
129 - Sync MBI generators with upstream changes.
130 - Various cosmetics, small bug fixes and inner refactorings
132 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
133 (FG: issues on Framagit; GH: issues on GitHub)
134 - FG#105: "Variable penalty should not be negative!" with in-flight messages and bandwidth profiles
135 - FG#109: Application time reported by --cfg=smpi/display-timing:yes is wrong
136 - FG#110: Wait_any does not trigger new model solve when host events occur
137 - FG#111: Wrong execution time in rare cases when using multicore
138 - FG!98: Re-enable the tests for legacy stochastic profiles
139 - FG!109: Trigger new engine solve upon host events such as host on/off
140 - FG!116: SMPI/replay: Fix issue with recv of size =0
142 ----------------------------------------------------------------------------
144 SimGrid (3.31) March 22. 2022.
146 The ненасильство release. We stand against war.
148 Against the agression by a sick system that forces peoples to take arms against each other.
151 - Rework the internals, for simpler and modern code. This shall unlock many future improvements.
152 - You can now define plugins onto the DFS explorer (previously called SafetyChecker), using the
153 declared signals. See CommunicationDeterminism for an example.
154 - Support mutex, semaphore and barrier in DPOR reduction
155 - Seems to work on Arm64 architectures too.
156 - Display a nice error message when ptrace is not usable.
157 - New test suite, imported from the MPI Bugs Initiative (MBI). Not all MBI generators are integrated yet.
158 - Remove the ISP test suite: it's not free software, and it's superseeded by MBI.
161 - fix for FG#100 by ensuring small asynchronous messages never overtake larger
162 ones, conforming to the standard.
163 - replay: fix waitall behaviour to avoid forgetting requests and leaking
165 - tracing: ensure that we dump the TI traces continuously during execution and
166 not just at the end, reducing memory cost and performance hit.
167 - Update OpenMPI collectives selection logic to match current one (4.1.2)
168 - Add a coherence check for collective operation order and root/MPI_Op
169 coherence. Potentially costly so not activated unless smpi:pendantic is set
170 or -analyze is given.
173 - New signal: Engine::on_simulation_start_cb()
174 - Introduce a new execution mode with this_actor::thread_execute(). This simulate
175 the execution of a certain amount of flops by multiple threads ran by a host. Each
176 thread executes the same number of flops, given as argument. An example of this new
177 function can be found in examples/cpp/exec-threads.
178 - Reimplementation of barriers natively.
179 Previously, they were implemented on top of s4u::Mutex and s4u::ConditionVariable.
180 The new version should be faster (and can be used in the model-checker).
181 - Actor::get_restart_count(): Returns the number of reboots that this actor did.
184 - MSG_barrier_destroy now expects a non-const msg_barrier parameter.
186 New plugin: the Chaos Monkey (killing actors at any time)
187 - Along with the new simgrid-monkey script, it tests whether your simulation
188 resists resource failures at any possible timestamp in your simulation.
189 - It is mostly intended to test the simgrid core in extreme conditions,
190 but some users may find it interesting too.
193 - New solver for parallel task: BMF.
194 - More realistic sharing of heterogeneous resources compared to the fair
195 bottleneck solver used by ptask_L07.
196 - Implement the BMF (Bottleneck max fairness) fairness.
197 - Improved resource sharing for parallel tasks with sub-flows (parallel
198 communications between same source and destination inside the ptask).
200 - "--cfg=host/model:ptask_L07 --cfg=host/solver:bmf": enable the ptask
201 model with BMF solver.
202 - "--cfg=bmf/max-iterations: <N>": maximum number of iterations performed
203 by BMF solver (default: 1000).
204 - "--cfg=bmf/precision: <N>": numerical precision used when computing
205 resource sharing (default: 1e-12).
206 - This model requires Eigen3 library. Make sure Eigen3 is installed to use BMF.
209 - Modifications of the Profile mechanism, with some impact on users
210 - Addition of a new (S4U) method to init profiles from generic functions to improve versatility
211 - Fix initial behaviour of state_profiles
212 - Modify periodicity to behave like a period, and not like a loop delay
215 - Drop xbt_dynar_shrink().
218 - Made the following bindings static (previously member functions):
219 - Actor: Actor.kill_all(), Actor.by_pid()
220 - Host: Host.by_name(), Host.current(), Host.on_creation_cb()
221 - Mailbox: Mailbox.by_name()
222 - Added the following bindings:
223 - this_actor.warning()
224 - Mailbox.put_init() [example: examples/python/comm-waitallfor/]
225 - Comm.detach() [example: examples/python/comm-waitallfor/]
226 - Comm.wait_for() [example: examples/python/comm-waitfor/]
227 - Comm.wait_any_for()
228 - Comm.wait_all_for() [example: examples/python/comm-waitallfor/]
229 - Mutex [example: examples/python/synchro-mutex/]
230 - Barrier [example: examples/python/synchro-barrier/]
231 - Semaphore [example: examples/python/synchro-semaphore/]
234 - Remove target "make uninstall" which was incomplete and no longer maintained.
236 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
237 (FG: issues on Framagit; GH: issues on GitHub)
238 - FG#57: Mc SimGrid should test whether ptrace is usable
239 - FG#87: Smpi scripts fail with spaces in paths
240 - FG#100: [SMPI] Order of the message matching is not guaranteed
241 - FG#101: LGPL 2.1 is deprecated license
242 - FG#104: "make uninstall" not up-to-date
243 - GH#151: Missing mutexes for DPOR.
245 ----------------------------------------------------------------------------
247 SimGrid (3.30) January 30. 2022.
249 The Sunday Bloody Sunday release.
251 Main user-visible changes:
252 - The SimDag API for the simulation of the scheduling of Directed Acyclic
253 Graphs has been dropped. It was marked as deprecated for a couple of years.
254 We finally complete the implementation of what has been called SimDag++
255 internally, i.e., porting the different features of SimDag on top of S4U.
256 The new way to simulate the execution of dependent activities directly by
257 maestro (without any other actor) is details in the examples/cpp/dag-* series
259 - The removal of SimDag led us to also remove the export to Jedule files that
260 was tightly coupled to SimDag. The instrumentation of DAG simulation is still
261 possible through the regular instrumentation API based on the Paje format.
262 - We also dropped the old and clumsy Lua bindings to create platforms in a
263 programmatic way. It can be done in C++ in a much cleaner way now, which
264 motivates this suppression.
267 - Introduce on_X_cb() functions for all signals, to attach a new
268 callback to the signal X. The signal variables are now hidden and
269 only these functions should be used.
270 Rationale: this enables the usual deprecation scheme where functions
271 remain for 4 releases if we need to modify the signals, while the
272 current code with the signal variables directly visible prevents any
274 - New function: Engine::run_until(date), to split the simulation.
275 - New signal: Activity::on_veto, to detect when an activity fails to start.
276 - Signal change: Comm::on_start(Comm&, bool) has been replaced by
277 Comm::on_send and Comm::recv. These two signals respectively correspond to
278 when the sending or receiving side of a Comm is ready. They are raised at
279 the same locations as the former Comm::on_start signal.
280 - New function: Engine::track_vetoed_activities() to interrupt run()
281 when an activity fails to start, and to keep track of such activities.
282 Please see the corresponding example for more info.
283 - New functions: s4u::Comm::{sendto_init, set_source, set_destination} to enable
284 the use of vetoers with direct host-to-host communications. Both source and
285 destination have to set for a comm to start. Each call to these setters check
286 if all vetoes are satisfied. When it is the case, the comm starts. A use case of
287 these functions is given in examples/cpp/dag-scheduling.
288 - New functions: {Exec, Io}::update_priority allow you to modify the priority of
289 these kinds of activities during their execution. Behavior is detailed in
290 examples/cpp/io-priority/.
293 - Dynamic costs for MPI operations: New API to allow users to dynamically
294 change injected costs for MPI_Recv, MPI_Send and MPI_Isend operations.
295 Alternative for smpi/or, smpi/os and smpi/ois configuration options.
296 - Fix some issues with the replay mechanism.
299 - Function xbt::Extendable::get_data() is now templated with the type of the
300 pointee. Untyped function is deprecated. Use get_data<void>() if you still
301 want to retrieve void*.
304 - New section: "SimGrid MPI calibration of a Grid5000 cluster"
305 presenting how to properly calibrate MPI communications in SimGrid.
306 - Complete and reword the platform section, which is now completed.
309 - Thread contexts are used by default with Python bindings. Other kinds of
310 contexts revealed unstable, specially starting with pybind11 v2.8.0.
312 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
313 (FG: issues on Framagit; GH: issues on GitHub)
314 - FG#95: Wrong computation time for multicore execution after pstate change
315 - FG#97: Wrong computation time for ptask+multicore+pstates
316 - FG#98: SMPI offline simulation is inconsistent with the online simulation
317 (deadlocks / message truncation)
318 - FG#99: Weird segfault when not sealing an host
320 ----------------------------------------------------------------------------
322 SimGrid (3.29) October 7. 2021
324 The "Ask a stupid question" release.
326 We wish that every user ask one question about SimGrid to celebrate.
327 On Mattermost, Stack Overflow or using the issues tracker.
329 New modeling features:
330 - Non-linear resource sharing, modeling resources whose performance heavily degrades with contention:
331 - The total capacity may be updated dynamically through a callback
332 and depends mainly on the number of concurrent flows.
333 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
335 - Dynamic factors: model variability in the speed of activities
336 - Each action can now have a factor that affects its progression.
337 This multiplicative factor is applied when updating the amount of work
338 remaining, thereby an activity with factor=0.5 only uses half of the
339 instantaneous power/bandwidth it is allocated and will appear twice
340 slower than what it actually consumes.
341 - This can be used to model a overhead (e.g., there is a 20 bytes
342 header in a 480 bytes TCP packet so the factor 0.9583) but the novelty
343 is this factor can now easily be adjusted depending on activity's and
344 resources characteristics.
345 - This existed for network (e.g., the effective bandwidth depends
346 on the message in SMPI piecewise-linear network model) but it is now
347 more general (the factor may depend on the source and destination and
348 thus account to different behaviors for intra-node communications and
349 extra-node communications) and is available for CPUs (e.g., if you
350 want to model an affinity as in the "Unrelated Machines" problem in
351 scheduling) and disks (e.g., if you want to model a stochastic
353 - For that, resources can be provided with a callback that computes
354 the activity factor when creating the action.
355 - Example: examples/cpp/exec-cpu-factors
356 - The same mechanism is also available for the latency, which
357 allows to easily introduce complex variability patterns.
360 - Added support to programmatic platform creation in Python.
361 Example: examples/python/clusters-multicpu
364 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
365 This can only be set through the API, not through XML files.
369 - Multiple fixes to ensure reproducibility of tracing
370 - scan/excan can now be replayed
371 - wait action now uses ranks and not pid, as the other ones.
372 - smpi/init and smpi/finalization-barrier are now valid for replays.
373 - exit() is now intercepted by SMPI to avoid premature shutdown of
374 simulation. First non 0 return codes is returned as simulation return
378 * New section "Release Notes" documenting recent and current developments.
379 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
380 * Improvements in API Reference for C++ and Python interfaces.
383 - Make wifi creation compatible with ns-3 version 3.34 too.
385 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
386 (FG: issues on Framagit; GH: issues on GitHub)
387 - FG#77: Search feature of doc is broken (update sphinx theme version)
388 - FG#78: Multiple fixes for SMPI replay:
389 - TI tracing of allotallv/w was outputting wrong values
390 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
392 ----------------------------------------------------------------------------
394 SimGrid (3.28) July 14. 2021
396 The Victoriadagarna Release.
399 - C++ platform interface: Users can now describe their platform directly in C++.
400 This provides greatly flexibility and performance improvement for complex
401 platforms. Main features:
402 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
403 "zones", instead of single hosts. This feature enables the description
404 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
405 - StarZone: new zone with a Star-like topology. The routes are defined
406 as a set of links used to communicate from node to everybody (node<->ALL).
407 - Split-Duplex links: auxiliary method to create split-duplex links in
408 the platform, easing its utilisation. It automatically creates both UP
409 and DOWN links (similarly as done in XML).
410 - Please refer to the documentation and the examples included:
411 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
412 - New plugin: Producer-Consumer with monitor. Just requires to include the
413 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
414 associated example (examples/cpp/plugin-prodcons).
417 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
418 s4u::Io::wait_any(), s4u::Io::wait_any_for().
419 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
420 parameter by reference, instead of a pointer.
421 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
422 Explicitly cancel the activity to get back to previous behavior.
423 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
426 - The default SMPI compiler flags are no more taken from the environment.
427 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
428 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
430 --cfg=smpi/finalization-barrier: which can be used to add
431 a barrier inside MPI_Finalize. This can help for some codes which cleanup
432 data attached to a process, but still used in other SMPI processes.
433 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
434 is active when set to false, to keep going after a small error
435 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
436 which may or may not be problematic in the end.
438 - fix behaviour, as maximum iteration count could be ignored
439 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
440 use sampling when the same kernel is called with a different set of parameters
441 which have an impact on the timing.
442 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
443 It should now work with smpi/auto-shared-malloc-thresh.
444 - Improve error handling and reporting in multiple places
445 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
446 validity checks, truncated messages are now an error, return errors
447 when explicitely deleted handles are reused, ...)
448 - RMA: multiple fixes and stability improvements.
449 - analysis (-analyze flag in smpirun):
450 - SMPI can now report buffer leaks as well as MPI handles leaks,
451 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
452 - if -trace-call-location is used when compiling, SMPI can report
453 origin of leaked handles/buffers
454 - group leaks by type/origin in output message if possible
455 - New implemented MPI calls: MPI_Comm_test_inter
458 - Changed internal implementation of bandwidth factors in network models.
459 Models affected: CM02, LV08 (default), SMPI, IB.
460 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
461 Bandwidth factors are applied to communications to describe that users
462 cannot use 100% of the available bandwidth. For example, the default network model,
463 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
464 behavior was implemented by limiting the bandwidth available in the LMM
465 system for this flow. This may give the false impression that there is
466 bandwidth available for other flows due to its underutilization, especially
467 for the dynamic bandwidth factors used in SMPI models.
468 To avoid this, we have modified the implementation so that each flow uses the
469 maximum physical bandwidth according to the LMM system.
470 However, the actual throughput of the flow seen by the user is defined by
471 the physical bandwidth multiplied by the bandwidth factor.
472 This change impacts on the simulation results for all network models on
473 which we have bandwidth factors configured.
474 *****************************************
475 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
476 *****************************************
477 This change may impact on the timing of your simulation results.
478 Take care when comparing simulations from different SimGrid's
479 versions. Sorry for the inconvenience.
480 - Dynamic network factors: users can configure a callback to define
481 the network factors dynamically. This API is available at
482 simgrid::kernel::resource::NetworkModelIntf.
483 - Users have access to complete information about the current communication
484 to decide which factor to apply. This includes: message size, source and
485 destination hosts, links and zones traversed.
486 - Dynamic factors for both latency and bandwidth.
487 - For more details, see the example in (examples/cpp/network-factors).
488 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
489 deprecated since version 3.24, are no longer supported. Instead, use
490 "wattage_off" and "wattage_per_state".
493 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
496 - New tutorial: Model-checking and formal assessment
497 - New sections: "Demystifying the routing" and "C++ platforms"
498 - Update and improve the part on visualization in MPI and Algo tutorials.
499 - Improve the section on routing: how to define it, how it's used internally
500 - Fix many issues, broken links and missing references in doxygen and Sphinx
503 - Lua platform files are deprecated. Their support will be dropped after v3.31.
506 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
508 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
509 (FG: issues on Framagit; GH: issues on GitHub)
510 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
511 - FG#64: Configuring smpi/IB-penalty-factors
512 - FG#67: Running computation concurrently with MPI_Iallreduce
513 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
514 - FG#71: Segmentation fault on invalid gw_src/gw_dst
515 - GH#322: Issue when an actor kills his host vm
517 ----------------------------------------------------------------------------
519 SimGrid (3.27) March 29. 2021
521 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
523 Main user-visible changes:
524 - All functions marked as deprecated for one year were removed.
525 - Removed all that was related to the "Storage" type in all APIs and bindings.
526 "Disks" have been introduced more than a year ago to replace them for the better.
527 It's time to finish this replacement.
528 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
529 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
530 avoiding "initializer element is not constant" errors with some codes. Please be
531 careful to recompile your MPI codes if switching to this new SimGrid version.
532 - Network model ns-3 now requires version ns-3.28 or higher.
533 - Removed some option aliases that were deprecated since v3.23 (June 2019)
534 - 'verbose-exit' is now 'debug/verbose-exit'
535 - 'simix/breakpoint' is now 'debug/breakpoint'
536 - 'clean-atexit' is now 'debug/clean-atexit'
539 - New functions: Comm::sendto_{init,async} to initiate a communication
540 on between two (possibly remote) hosts.
541 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
542 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
543 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
544 or Mailbox::get_async<void>() if you really want to play with void*.
545 - Unify the interface of Activity::on_{start/activity}
546 - New function: Comm::get_dst_data()
547 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
550 - ptask should now work better with multicore hosts (FG#37).
551 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
552 - Internal cleanups and refactoring around models, to prepare for programmatic
554 - Modernize our tests of the performance models
558 - New section: Configuring the logs from the command line.
559 - New section: Programming API of logs.
560 - New section: Getting realistic results
561 - New section: Examples of SMPI Usage
562 - New part: SimGrid's Internals, with a section on Design goals
563 - Move back to the classical Breathe tool chain, and fix many small glitches.
566 - Handle leak checking: SMPI will now display a message if MPI handles are still
567 present at the end of a simulation. Option smpi/list-leaks:n will display the n
568 first leaked pointers and their types.
569 - smpi/display-allocs option : Display Memory allocations analysis, with hints
570 about the largest allocations. This only works for allocations with malloc/calloc
571 and not compiled with SMPI_NO_OVERRIDE_MALLOC
572 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
573 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
574 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
575 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
576 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
577 - Replay now opens and parses the trace file only once for all ranks when possible.
579 - better handling of MPI_PROC_NULL
580 - MPI_Group_incl was broken in some cases
581 - PAPI tracing now fixed
582 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
583 - SMPI will now fail properly when receive size < sent size, instead of
584 just setting MPI_ERR_TRUNCATE and keep going.
585 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
586 - Remove useless build patches and split build/run patches on the page
587 - Build in parallel when possible
588 - Test on yum-based systems
589 - New nightly tested Proxy-apps:
590 - OpenMPI mpi-test-suite
591 - Coral MCB, LSMS, Qbox, CLOMP
592 - ECP/Mantevo Cloverleaf3D, Ember
593 - Codevault bhtree_mpi
596 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
597 - Start to overhaul the implementation to modernize it (still ongoing).
600 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
601 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
602 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
603 - Deprecate xbt_dynar_to_array().
604 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
606 C binding and interface:
607 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
608 - Constness requirement has been lowered for parameter argv of sg_actor_start()
609 and sg_actor_create() It's now (char* const*). The previous version is still
610 available as sg_actor_start_() and sg_actor_create_().
612 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
613 (FG: issues on Framagit; GH: issues on GitHub)
614 - FG#37: Parallel tasks are limited to 1 core per host
615 - FG#62: Running "smpirun -replay" on large networks
616 - FG!46: Fix a few potential memory leaks in SMPI colls
617 - GH#343: Wrong consumed energy when an exec timeouts
619 ----------------------------------------------------------------------------
621 SimGrid (3.26) December 15. 2020
623 The Release release (the French lockdown was eased today).
625 Important user-visible changes:
626 - SimGrid now requires a compiler with C++14 support.
627 Sibling projects should upgrade their FindSimgrid.cmake
628 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
629 several users had difficulties to understand issues when using high bandwidth or
630 small latency events. The new value was already the default for SMPI and
631 should not cause too much performance hit. It can change some simulation
632 timings if simulation was skipping a lot of small events (which was actually
633 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
634 - On some network models (L07, CM02, LV08), an implicit loopback link is created
635 for local communication. Default bandwidth was 500MBps and latency was 15µs,
636 which was rather bad. Default values have been changed to 10GBps and 0 latency.
637 Options network/loopback-lat and network/loopback-bw have been added to adapt
638 these values if needed. Network and routing models already providing separate
639 loopbacks (clusters and such) are not impacted by this change.
640 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
641 metadata that are added to each MPI message by MPI libraries (tag,
642 src/dest, and communicator information). This may change simulation times
643 for very small message sizes, but should be more realistic.
646 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
647 - New plugins: wifi_energy and link_load. See the associated examples.
648 - New examples: wifi networks, and communication suspend.
651 - LTO is now enabled for Intel/clang compilers.
652 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
653 Setting it to "auto" will use all available cores, while setting it to n will
654 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
655 - Remove obsolete runtime option 'contexts/parallel-threshold'.
656 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
657 has been kept for compatibility.
658 - Finally remove obsolete snake_case() aliases for runtime options.
659 - Further improve the documentation.
662 - the on_exit() of each actor is also executed when the simulation deadlocks.
663 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
664 An example is provided for s4u::Comm but it should work with Exec and Io.
667 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
668 https://framagit.org/simgrid/SMPI-proxy-apps
669 - MPI/IO: supports names starting with "./"
670 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
672 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
673 building shared libraries, this may cause issues, so environment variable
674 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
675 - most temporary files should now be created in /tmp dir (or equivalent).
676 If this one does not allow execution of code (noexec flag), this may cause issues.
677 Please use another tmp directory (using TMPDIR or equivalent system variable)
681 - A large refactoring is ongoing, but this should be transparent for users so far.
684 - Drop xbt_str_split().
686 C binding and interface:
687 - The return type of the sg_actor_on_exit() callbacks is now 'void'
688 (int value was previously ignored)
689 - Many C functions were renamed to follow a common naming style
690 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
691 names are kept until version 3.30.
692 - Many MSG tests were converted to the new S4U's interface in C, that
693 was extended for that.
695 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
696 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
697 - FG#41: Add sg_actor_create C interface
698 - FG#43: xbt::random needs some care
699 - FG#48: The Impossible Did Happen (yet again)
700 - FG#50: Suspending an actor executed at the current timestamp fails
701 - FG#52: Zero-seconds timeout : "That's in the past already"
702 - FG#53: Crash while using ns-3 network model
703 - FG#54: How to suspend a comm?
704 - FG!22: Stochastic Profiles
705 - FG!24: Documentation and fix for xbt/random
706 - FG!35: Add a modeling hint for parallel links in doc
707 - FG!36: [xbt/random] Read/Write the state of the RNG
708 - FG!42: Link cumulated load plugin
709 - FG!44: Test condition variable wait_until+notify_all
710 - GF#18137: Allow different stack sizes?
711 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
712 - GH#139: Allow pthread creation in SMPI
713 - GH#336: Packet-level simulation using SMPI?
714 - GH#345: Error in the LMM while migrating a VM
715 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
716 - GH#352: pip install / python setup.py install fail to find pybind11
717 - GH!337: Fix link_energy plugin for wifi platforms
718 - GH!339: Add Mailbox set_receiver method to python binding
719 - GH!344: Cast hugepages macros parameters to int64
721 ----------------------------------------------------------------------------
723 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
725 The Palindrom Day Release.
727 Important user-visible changes:
728 - Improve the Python usability (stability and documentation).
729 - A nasty synchronization bug was ironed out, see also below.
730 - Python's doc was organized as a separate tree, now integrated with C++.
731 - C bindings of S4U were not part of the doc.
732 - The C++ doc was also improved as methods are now split by theme.
733 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
734 - This is mandatory to use the Java bindings.
735 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
736 - SimDAG++: Automatic dependencies on S4U activities (experimental)
737 - Some features are already implemented but not all of them
738 - Cannot block an activity until it's scheduled on a resource
739 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
740 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
743 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
744 - Drop support for Viva/Triva (old visualization tools), hence removed
745 TRACE_get_node_types() and TRACE_get_edge_types() functions.
748 - Require pybind11 version 2.4.
749 - Greatly improve locking during context switches with Python.
750 - Declaring Python callbacks to this_actor::on_exit is now working.
751 - No more tests are failing currently, even on win32.
754 - Actor: Merge signals on_migration_start/end into on_host_change
755 - Actor: Rename migrate() into set_host()
756 - Disk: Allow users to get the read and write nominal bandwidth values
757 - Exec: Implement wait_for(timeout)
758 - Io: Implement wait_for(timeout)
761 - Parse errors now raise a simgrid::ParseError that you may want to catch.
764 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
765 been renamed to fix a spelling error. As usual, the previous definitions are
766 kept with a deprecation warning. However, to avoid runtime errors, it is now
767 forbidden to override these deprecated methods in a derived class. Please use
768 the new names immediately if you need to override them.
771 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
772 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
773 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
774 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
775 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
776 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
777 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
778 - New module: random, providing classical random numbers generators.
781 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
782 allocations' size above which they are considered "shared" by default (as if
783 they were performed through SMPI_SHARED_MALLOC macros).
784 Default = 0 = disabled feature.
785 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
786 This can cause some troubles if codes are already overriding these. If this
787 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
788 help, but will make this feature unusable.
789 - Input check is now performed much more consistently and outputs precise
790 warnings for each error found
791 - New supported calls : MPI_File_set_view, MPI_File_get_view
792 - MPI I/O now only supports disks, as storage is discontinued.
793 - MPI I/O allows opening files without specifying absolute path
794 - Note : SMPI configuration options are now initialized by smpi_init_options()
795 call, which should be called if SMPI is not being used through smpirun.
797 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
798 - FG#9: Python bindings crashing
799 - FG#39: Missing s4u::Comm::wait_any_for example
800 - FG#42: Add support for ThreadSanitizer (TSan)
801 - FG!19: Removing RngStream
802 - FG!20: A module for RNG calls
803 - FG!21: Choice between ad-hoc and standard distributions implementations
805 - FG!25: Fix link in Mutex doc
806 - FG!26: Fix links in SMPI interface doc
807 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
808 - GH#305: Unscheduled tasks are still excuted
809 - GH#313: smpirun: manual outdated w.r.t. --help
810 - GH#321: [S4U] Get task remaining work ratio
811 - GH#323: Crash when an actor turn off his physical host
812 - GH#335: Missing links on dragonfly example svg
813 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
814 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
816 ----------------------------------------------------------------------------
818 SimGrid (3.24) October 9. 2019
820 The Clean Disk Release.
823 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
824 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
825 thus becomes the default resource to execute I/O operations. Examples
826 and the file system plugin have been modified to use disks instead of
827 storage elements. Storage can still be used but is doomed to disappear.
828 Users are thus adviced to update their simulators accordingly.
829 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
830 for consistency with pthread_barrier_wait()
831 - Host::get_englobing_zone() returns the englobing netzone
832 - Actor::on_destruction is now called in the destructor
833 Actor::on_termination new signal called when the actor terminates
835 - Global signals are now part of the Engine:
836 - on_platform_creation: after config settings, before the XML parsing
837 - on_platform_created: right after the XML parsing
838 - on_time_advance: each time the clock advances
839 - on_simulation_end: after simulation, before cleanups
840 - on_deadlock: as the name implies.
842 - sg_{actor,host,link}_{data,data_set}() now all exist.
843 Use them to attach user data to the object and retrieve it.
846 - Introduce an experimental Wifi model. It sounds reasonable
847 according to the state of the art, but it still has to be properly
848 validated, at least against ns-3 if not against reality.
849 - Improved the usability of ns-3. Several bugs were ironed out.
850 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
851 It is now expressed as 'idle:epsilon:allCores' to properly model the
852 consumption of non-whole tasks on mono-core hosts. More info in the doc.
855 - convert a new set of functions to the S4U C interface and move the old MSG
856 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
859 - Fortran bindings for DVFS have been removed.
860 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
861 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
862 - SMPI can now be selected by cmake's find_module(MPI) with
863 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
864 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
865 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
866 - trace-call-location can be used with TI traces, and replayed, and is compatible
867 with smpi/comp-adjustment-file.
868 - sleep events are now correctly traced/replayed.
869 - Default for trace-call-location is now to use file names and not full paths.
870 To revert to previous behavior (in case of collision of filenames), option
871 "smpi/trace-call-use-absolute-path" can be set to yes.
874 - Use the included xxHash as an hash implem when C++14 is usable.
875 - Option model-checker/hash was removed. This is always activated now.
876 - New option smpi/buffering controls the MPI buffering in MC mode.
877 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
878 This is useful to check for MPI compliance.
881 - New section on plugins: how to define a new one, and existing ones.
884 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
885 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
888 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
889 and <mount> tags. The rationale is finish to decouple what should be handle
890 by SimGrid kernel (i.e., resources and activities) from plugins built in
891 user space (e.g., the file system plugin). This new tag currently accepts
892 only three attributes (id, read_bw, and write_bw). All the other information
893 that was declared with the storage related tags now has to be expressed as
894 properties. An example of platform using this new tag is available at
895 examples/platforms/hosts_with_disks.xml
898 - 'expect signal' can now accept more than one potential signal.
900 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
901 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
902 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
903 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
904 - FG#31: per-actor data
905 - FG#34: SG_BARRIER_SERIAL_THREAD?
906 - FG#35: model-checker does not like buster-produced binaries
907 - FG!13: MC: complete workaround in the error msg seen on modern systems
908 - FG!15: execute_flops now logs compute
909 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
910 - FG!17: ns-3: unblock the right number of communications + others issues
911 - FG!18: Improving the performance of the ns-3 bindings
912 - GH#189: Energy consumption of parallel tasks
913 - GH#219: Error in the throughput of TCP transfer
914 - GH!330: Minor polishing of English: Configuring_Simgrid
915 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
916 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
917 - GH!334: Rework energy plugin, again
919 ----------------------------------------------------------------------------
921 SimGrid (3.23.2) July 8. 2019
924 - Nicer introduction page.
925 - Migrate the "Deploy your application" page to the new doc.
926 - Move Java as a subtree of MSG.
929 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
932 - Drop xbt_dynar_sort_strings().
935 - Really fix FG#26: Turning off a link should raise NetworkFailureException
936 - FG#27: Wrong exception thrown to wait_any when link is turned off
937 - GH#328: Java: Canceling multiple tasks in a single vm/host
939 ----------------------------------------------------------------------------
941 SimGrid (3.23) June 25. 2019
943 The Exotic Solstice Release.
946 - SunOS and Haiku OS support. Because exotic platforms are fun.
947 - Stop setting random seed with srand48() at initialization.
948 - Use addr2line as a fallback for stacktraces when backtrace is not available.
949 - Build option -Denable_documentation is now OFF by default.
950 - Network model 'NS3' was renamed into 'ns-3'.
953 - Simgrid can now hopefully be installed with pip.
956 - wait_any can now be used for asynchronous executions too.
959 - New log appenders: stdout and stderr. Use stdout for xbt_help.
960 - Drop xbt_dict_dump.
963 - 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
964 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
965 - MPI interface is now const correct for input parameters
966 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
967 - Fortran bindings for async collectives.
968 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
971 - Remove option 'model-check/record': Paths are recorded in any cases now.
972 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
973 always sparse. That's an immense gain in memory, and can even be
974 faster because of cache effects. So there is no need to clutter the
975 code to allow the user to go for the unefficient mode.
978 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
979 models proved to be more accurate than these old experiments.
981 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
982 - FG#1: Broken link in error messages
983 - FG#2: missing installation documentation
984 - FG#3: missing documentation in smpirun
985 - FG#6: Python bindings not available on PyPI
986 - FG#7: simple cmake call requires doxygen
987 - FG#8: make python bindings an optional dependency
988 - FG#10: Can not use MSG_process_set_data from SMPI any more
989 - FG#11: Auto-restart actors forget their on_exit behavior
990 - FG#12: -Denable_lto=OFF doesn't disable LTO
991 - FG#13: Installs unstripped file 'bin/graphicator'
992 - FG#14: Installs the empty directory 'doc/simgrid/html'
993 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
994 - FG#17: Dead link in doc (pls_ns3)
995 - FG#20: 'tesh --help' should return 0
996 - FG#21: Documentation link on http://simgrid.org/ broken
997 - FG#22: Debian installation instruction are broken
998 - FG#26: Turning off a link should raise NetworkFailureException exceptions
999 - GH#133: Java: a process can run on a VM even if its host is off
1000 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
1001 - GH#326: Valgrind-detected error for join() when energy plugin is activated
1003 ----------------------------------------------------------------------------
1005 SimGrid (3.22) April 1. 2019
1007 The Easter Christmas Release (this one is somewhat late).
1010 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
1011 - Not complete yet: asynchronous activities (amongst others) are still missing
1012 - Still ongoing: the interface may change in the future. We need more
1013 testers! Please report any glitches.
1014 - No new project using Java should start now. Please switch to Python.
1017 - Some of the internal cleanups may lead to speed improvements:
1018 - The hard limitation on the amount of simulated actors with
1019 Java+Mac was removed. Now, the available memory is the only limit.
1020 - Our refcounting was tidyied, leading to 10% speedups in some cases.
1021 - We are still working on making our code robust to the actor kills
1022 and hosts' churn. Things are improving, but it's not perfect yet.
1023 - Replaced our own code to display a backtrace (that was forking addr2line)
1024 with the Boost.Stacktrace library.
1025 You won't see your backtraces without this optional dependency.
1026 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
1027 - Stop setting random seed with srand() at initialization.
1030 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
1031 XML file version remains 4.2 since old files are still compatible.
1034 - Process termination which was broken at version 3.21 has been repaired.
1035 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
1036 - Hide the examples into examples/deprecated. New users should use Python.
1039 - Drop MSG_process_create_from_stdfunc() from the API.
1040 This C++ function was a pimple in the C API, made necessary at some
1041 point by the Java bindings. This is fixed now.
1042 - Hide the examples into examples/deprecated. New users should use S4U.
1043 - MSG_process_create and MSG_process_attach now crash if the host used
1044 to run this process is off.
1045 - Fix the protype of MSG_process_on_exit()
1046 Now use (int,void*) callbacks instead of (void*,void*) ones.
1047 The implementation was ways too messy to actually work, I guess.
1050 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
1051 too often and provide more accurate timings.
1052 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
1053 - MPI_Alltoallw support
1054 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
1055 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
1056 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
1057 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
1059 - Basic implementation of generalized requests (SMPI doesn't
1060 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
1063 - Drop sg_cmdline. Please use xbt_cmdline instead.
1064 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
1065 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
1066 OsSemaphore is implemented in a portable way with C++11 threads.
1067 This should allow much more threads to be created at the same time,
1068 allowing Mac and Java users to simulate many more actors.
1069 - Implement the 'thread' factory with std::thread instead of xbt ones.
1070 It is not possible to set the stack size with threads anymore, but
1071 -fsplit-stack is the way to go nowadays when using threads.
1072 - Drop the xbt_os_thread_t module (now unused)
1073 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
1074 - Drop xbt_str_join_array().
1075 - Drop cunit, use Catch2 instead.
1078 - Many cleanups in the kernel::activity namespace. This was long
1079 overdue, and shall open the path to many future endeavors.
1082 - #132: Java: a process can not shut down its own host
1083 - #220: S4U: detached send causes exception if sender terminates soon after sending
1084 - #261: Document the parameters of parallel execution's constructor
1085 - #300: [s4u] BarrierPtr is missing
1086 - #314: SMPI args internal cleanup
1087 - #316: Fix a bug related to the CPU utilization of multi-core VM
1088 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
1089 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
1090 - #325: Turning off a host has different behavior on sleeping actors and computing actors
1092 ----------------------------------------------------------------------------
1094 SimGrid (3.21) October 3. 2018
1096 The Restarting Documentation (TRD) Release.
1099 - Convert the user manual to Sphinx for improved usability.
1100 Unstable version now hosted on https://simgrid.frama.io/simgrid
1101 Many glitches remain; stay tuned.
1102 - Restrict the Doxygen reference API to the installed header files.
1103 (documenting all internals this way is overkill + doxygen scales badly)
1104 - New tutorial on S4U (highly inspired from the old MSG one, but with
1105 a git repository to fork as a starting point, and a docker image)
1106 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
1107 - The Developper manual is still to be converted (not compiled until then)
1108 - Some parts are still missing in this conversion (such as the
1109 platform and deployment sections) while others were blindly converted
1110 and would need more love (such as the configuration flags).
1111 Things will certainly further improve in the future releases.
1114 - Fully reimplement auto_restart mechanism that was utterly broken
1115 (fix #22, #131 and #281 that were all very related)
1116 - Implement semaphores (fix #303 and #291)
1117 - When creating an actor from a function and its parameters,
1118 move-only parameters are not allowed anymore, as it would prevent
1119 the actor to be restartable if its parameters are consumed on run.
1120 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
1121 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
1122 IO operations that can be started, waited for, or canceled as a regular
1124 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
1128 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
1129 - New host variable: 'core_count'
1132 - Remove xbt_os_thread_specific features
1133 - Remove portability wrapper to condition variables
1134 - Remove xbt_os_thread_yield()
1137 - MPICH collective selector now mimics MPICH 3.3b
1138 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
1139 - MPI_Init(NULL,NULL) is now allowed
1140 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
1141 - Command-line arguments should now be handled properly in Fortran simulations
1142 - MPI Attributes and related callbacks should work in Fortran
1143 - Apps using C + Fortran should now work
1144 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
1145 - Support MPI_Error_String
1148 - Due to an internal bug, Msg.run() must now be your last line.
1149 We hope to fix it in a future release, and we are sorry for the inconvenience.
1152 - #22: Process autorestart seem to only work with CAS01 cpus
1153 - #56: Feature request: dynamic SMPI replay
1154 - #93: simgrid should not eat --help
1155 - #111: How to change the loopback link characteristics?
1156 - #116: Platform cluster radical
1157 - #117: Platform cluster/host order
1158 - #129: Replay_multiple on steroid
1159 - #131: Java : autostart does not work
1160 - #143: Setting a breakpoint at a given time
1161 - #235: Network energy models should be integrated
1162 - #264: Add ptask L07 resource tracing
1163 - #271: Dynamic and manual replay of SMPI traces
1164 - #279: Breakpoints option uses time deltas instead of absolute time
1165 - #281: Daemonized actor no longer daemonized after an auto-restart
1166 - #282: TIT convertor erase traces when using absolute path in the trace list file
1167 - #285: segfault when a s4u actor kills itself with msg process tracing activated
1168 - #286: Pajé traces are not exposing the number of cores
1169 - #287: Command-line parsing should stop when encountering '--'
1170 - #288: MPI_Init(NULL, NULL)
1171 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
1172 - #290: Method to check if a mailbox has a message ready to be consumed
1173 - #291: [s4u] semaphores do not seem available yet
1174 - #294: Alltoall collective implementation of mpich
1175 - #295: Starting a Host as OFF is not supported
1176 - #296: DTD too permissive
1177 - #299: Add a s4u barrier example
1179 ----------------------------------------------------------------------------
1181 SimGrid (3.20) Released June 24. 2018
1183 The proxy snake_case() release.
1186 - The whole API is now using snake_case() instead of mixing it with camelCase.
1187 Compatibility wrappers in place for one year.
1188 - We took the opportunity to increase the consistency of the API.
1189 Please report any remaining glitches.
1190 - New function simgrid::s4u::Host::get_actor_count:
1191 Returns the number of actors running on a specific host.
1194 - SMPI is now tested with ~45 proxy apps from various sources, with none or
1195 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
1196 - Replay: The replay file has been re-written in C++.
1197 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
1198 supported. They are stored in the trace and used when replayed.
1199 - Basic support of MPI_Cancel. Robustness not guaranteed.
1200 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
1201 (only for MPI_COMM_TYPE_SHARED).
1202 - New option: smpi/privatize-libs, to add external shared libs to be privatized
1203 by SMPI. They will be copied locally and loaded separately by each process.
1204 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
1205 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
1206 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
1207 with internal index optind with multiple processes. Only works if MPI_Init has
1208 already been called.
1209 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
1210 syntax without preprocessing tricks.
1213 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
1214 behavior tracing is no longer limited to MSG
1215 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
1216 this means that traces now need to be updated or re-obtained; both Irecv
1217 and Isend lines in your traces have now in total 5 mandatory fields:
1218 <rankid> <command> <to/from rankid> <tag> <size>
1219 To update your traces, it suffices to add a 0 for the tag here.
1220 - TIT now also supports waiting for a distinct request via MPI_Wait.
1221 Wait/Test now wait for a specific request, not just the last one that was
1222 issued. This unfortunately means another update, because we need to
1223 identify which request you want to wait for. We do this via the
1224 triplet (sender, receiver, tag), which needs to be added:
1225 <rankid> <command> <sender> <receiver> <tag>
1226 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
1227 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
1230 - The deprecation of MSG is ongoing (but this should not impact you).
1231 Many MSG functions are now simple wrappers on the C API of S4U. If
1232 you wish to convert your code to S4U, find the S4U counterparts of
1233 your MSG calls in src/msg/msg_legacy.cpp.
1234 - MSG can still be used, but won't evolve anymore.
1237 - Allow to run the Link energy plugin from the command line with
1238 --cfg=plugin:link_energy
1239 - Rename Energy plugin into host_energy
1240 - Rename Load plugin into host_load
1243 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
1244 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
1245 - Improve performance when handling timeouts of simix synchros.
1248 - Config: the C API is now deprecated (will be removed in 3.23), and
1249 the C++ API has been slightly improved.
1252 - Fix several build issues on OSX.
1253 - Move simgrid_config.h to simgrid/config.h (old header still working)
1256 - #143: Setting a breakpoint at a given time
1257 - #150: Inconsistent event names in SMPI replay
1258 - #258: daemonized actors hang after all non-daemonized actors have completed
1259 - #267: Linker error on unit_tmgr
1260 - #269: SMPI: tracing of MPI_Wait/all/any broken
1261 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
1263 ----------------------------------------------------------------------------
1265 SimGrid (3.19.1) Released March 22 2018
1267 The Fixed ABI Release.
1269 As you may know, we are currently refactoring SimGrid in deep.
1270 Upcoming SimGrid4 will be really different from SimGrid3: modular,
1271 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
1273 Our goal is to smooth this transition, with backward compatibility
1274 and automatic update paths, while still progressing toward SimGrid4.
1276 SimGrid remains open during works: The last pure SimGrid3 release was
1277 v3.12 while all subsequent versions are usable alpha versions of
1278 SimGrid4: Existing interfaces remain unchanged, but the new S4U
1279 interface is budding and the internals are deeply reorganized.
1281 Since 2015, we work hard to reduce the changes to public APIs. When
1282 we need to rename a public library symbol in S4U, we let your
1283 compiler issue an explicative warning when you use the deprecated
1284 function. These messages remain for four releases, i.e. for one full
1285 year, before turning into an error. Starting with v3.15, your can
1286 also adapt to API changes with the SIMGRID_VERSION macro, that is
1287 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
1289 Starting with this v3.19.1, our commitment to reduce the changes to
1290 the public interfaces is extended from the API to the ABI: a program
1291 using only MSG or SimDag and compiled against a given version of
1292 simgrid can probably be used with a later version of SimGrid without
1293 recompilation. We will do our best... but don't expect too much of
1294 it, that's a really difficult goal during such profund refactoring.
1296 The difference between v3.19 and v3.19.1 is that the former was
1297 accidentally breaking the ABI of MSG, while the later is restoring
1300 S4U and kernel APIs will still evolve until SimGrid4, with one-year
1301 deprecation warnings as currently. In fact, cleaning up these
1302 interfaces and converting them to snake_case() is one release goal of
1303 v3.20. But don't worry, we are working to smooth this upgrade path.
1305 In summary, new projects should start with S4U to benefit of the
1306 future, but old MSG projects should still be usable with no change.
1308 ----------------------------------------------------------------------------
1310 SimGrid (3.19) Released March 20 2018
1312 The Moscovitly-cold Spring release.
1314 SMPI user-visible improvements:
1315 - Support of MPI_Type_create_subarray
1316 - Unimplemented MPI calls now abort (instead of issuing a warning)
1317 Some harmless calls around errhandlers still only issue a warning.
1318 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
1319 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
1320 is now possible (mmap privatization was refactored to this end).
1321 Adding new SMPI instances should also be technically doable now.
1323 SMPI internal cleanups:
1324 - Removed index notion from SMPI -- no more "getPid() - 1"!
1325 In tracing files, actors are called rank-<PROCESS-ID> but these are
1326 NOT the real ranks (tracing several communicators is not supported yet).
1327 - Deprecate smpi_process_index() (will be removed in 3.22)
1328 - Better testing on classical proxy apps, through an external project:
1329 https://github.com/simgrid/SMPI-proxy-apps/
1332 - Use a graphical TOC to make it easier to find the documentation you need
1333 - Revamp the MSG tutorial
1336 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
1337 - Remove parameter of MSG_process_killall().
1338 Resetting the PID was bogus anyway (several actors could have the same PID).
1341 - Execution->setHost() can be called after start(), to migrate it.
1342 - Comm::test_any() is now implemented.
1343 - s4u::Actor now has onCreation() and onDestruction() signals.
1344 - Install some kernel header files for the users' plugins and more.
1345 Warning, their API is really not stable yet.
1346 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1347 include/simgrid/kernel/routing/*Zone.hpp
1350 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1352 Build System and other cleanups:
1353 - Remove unused run-time parameter "tracing/onelink-only".
1354 - SimGrid now works with intel compilers (MC still buggy)
1355 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1356 - tesh kills subprocesses on timeout (but not on windows)
1359 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1362 - #194: Feature request: simgrid::s4u::Comm::test_any()
1363 - #245: migrating an actor does not migrate its execution
1364 - #253: Feature Request: expose clusters as objects
1365 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1366 - #255: Tesh broken on Windows
1367 - #256: Modernize FindSimGrid.cmake
1368 - #257: Fix (ab)use of CMake install
1370 ----------------------------------------------------------------------------
1372 SimGrid (3.18) Released December 24 2017
1374 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1376 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1377 - Support for MSG should not be removed from SimGrid before 2020,
1378 but future evolutions will be limited to the integration of
1379 user-provided patches.
1382 - s4u::onDeadlock() signal, run before stopping the simulation.
1383 - s4u::Exec: asynchronous executions (abstraction of background
1385 - s4u::Host->getLoad() returns the achieved speed in flops/s
1388 - Link::name() is deprecated and will be removed in v3.21, use
1389 Link::getCname() instead.
1390 - Mailbox::getName() changed to return a std::string, use
1391 Mailbox::getCname() instead to get a char*.
1392 - Storage::getName() changed to return a std::string, use
1393 Storage::getCname() instead to get a char*.
1394 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1395 to have both allocation and deallocation of the map in user space.
1398 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1399 The amount of flops remaining to do is a vector, not a scalar.
1400 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1401 name implies on both sequential and parallel tasks.
1402 - Both changes fix GitHub's #223 using PR #237 as a basis.
1403 Thanks Michael Mercier.
1404 - Most examples were converted to S4U and hidden elsewhere at tests.
1405 Rationale: we still want MSG to work; we want newcomers to use S4U.
1408 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1409 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1412 - Switch to the faster dlopen privatization mechanism by default
1413 - Documentation improvments
1416 - Live migration is getting moved to a plugin. Dirty page tracking is
1417 the first part of this plugin. This imply that VM migration is now
1418 only possible if one this function is called:
1419 - C/MSG: MSG_vm_live_migration_plugin_init()
1420 - C/C++: sg_vm_live_migration_plugin_init()
1421 - Java: Msg.liveMigrationInit()
1422 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1426 - Define class simgrid::xbt::Path to manage file names.
1427 - Removed unused functions:
1428 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1429 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1430 xbt_graph_export_graphviz()
1431 - xbt/str.h: xbt_str_join()
1432 - Remove unused datatypes:
1433 - xbt/heap.h: use std::priority_queue or boost::heap instead
1434 - xbt/swag.h: use boost::intrusive::list instead
1437 - New link_energy plugin for the consumption of the links.
1438 - All of the operations on files and storage contents have been
1439 packaged into a plugin (src/plugins/file_system). The current
1440 public interface can be found in
1441 include/simgrid/plugins/file_system.h
1442 To use these functions you now have to initialize the plugin by
1443 calling MSG_storage_file_system_init() just after calling
1444 MSG_init() or sg_storage_file_system_init() just after creating
1448 - Remove the undocumented/untested tag <include>
1451 - Remove viva specific tracing as the tool is no longer maintained
1454 - #248: Kill of finished processes leads to segfault
1455 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1456 - #239: please implement signal s4u::onDeadlock()
1457 - #230: segfaults when exit() without run()
1458 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1459 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1460 - #222: Actor::kill() doesn't really kill and segfaults
1461 - #221: odd LMM warning when killing an actor
1462 - #120: Memory leak when the processes are forcefully killed
1464 ----------------------------------------------------------------------------
1466 SimGrid (3.17) Released October 8 2017
1468 The Drained Leaks release: (almost) no known leaks despite the tests.
1470 Even very long simulations will preserve your memory: our testsuite
1471 stresses SimGrid for over 45mn on fast machines for a coverage of
1472 over 80%, and there is only one single known leak, of about 4kb.
1476 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1477 handler may be given to free resources if the comm cannot be completed.
1478 - this_actor::send(mailbox) is now mailbox->put()
1479 - New: simgrid::s4u::Comm::wait_all()
1480 - New: Host.execute() for remote executions.
1483 - Unused option network/sender-gap is removed.
1486 - Deprecate MSG_task_isend_with_matching(): This unused feature
1487 really complicates our internals. Will be removed in v3.20.
1490 - Improved context termination. It is now possible to free resources, even
1491 when a process is forcibly killed.
1492 - ContextBoost: add support for Boost versions above 1.61.
1495 - Bring back run-time option --cfg=exception/cutpath to remove exception
1497 - Removed unused functions:
1498 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1499 xbt_str_rtrim(), xbt_str_trim().
1500 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1503 - Removed header files obsolete since SimGrid 3.12:
1504 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1506 - #3: SD_exit should be made optional
1507 - #120: Memory leak when the processes are forcefully killed
1508 - #159: Threading test regression in Actor refcounting
1509 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1510 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1511 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1512 - #191: VM migration and pstate
1513 - #192: Updating the energy consumptions of all hosts crashes with VMs
1514 - #195: All actors have PID=0 in the logs
1515 - #204: Sometimes segfault with thread contexts and mmap privatization
1516 - #222: Actor::kill() doesn't really kill and segfaults
1517 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1519 ----------------------------------------------------------------------------
1521 SimGrid (3.16) Released June 22. 2017.
1523 The Blooming Spring Release: developments are budding.
1526 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1527 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1528 - Java JAR file should be finally fully working on Mac OSX too.
1530 XML platforms: Switch to platform v4.1 format.
1531 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1532 - <zone> can be used as a synonym for the now deprecated <as>
1533 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1534 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1535 - <actor> can be used as a synonym for the now deprecated <process>
1536 - state_file and avail_file periodicity is now easier to express
1537 (check the documentation) the old behavior should still work.
1538 * Storage: not backward compatible, but it seems that nobody ever used it
1539 - Remove attribute 'content_type' of <storage_type>: was never used
1540 - Make attribute 'model' of <storage_type> optional: for future usage
1541 - Remove Bconnection model property: was never (in)validated. Replaced by
1542 the maximum of the read and write bandwidth as a resource constraint.
1545 - New and Backwards Compatibility break:
1546 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1547 When one wants to get the list of tasks whose states have changed during a
1548 simulation round, s/he has to allocate and free a dynar and use it as
1549 argument to this function. The former SD_simulate (double how_long)
1553 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1554 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1557 - New function to update the consumption of all hosts at once.
1558 - Fix the model for multi-core hosts, linear in the amount of busy
1559 cores with an abnormality for pIdle (see paper).
1560 BUG: Still not satisfactory for ptask on mono-cores.
1563 - The netzone are now available from the MSG API.
1564 The old names still work, but are now deprecated.
1567 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1568 - New option: smpi/keep-temps to not cleanup temp files
1569 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1570 mallocs mode is used, allows to change the size of the fake file used
1571 (default 1MB), to potentially limit the number of mappings for large runs.
1572 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1573 - Fortran ifort and flang compilers support
1574 - New RMA calls supported (experimental):
1575 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1576 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1577 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1578 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1579 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1580 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1583 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1584 - Behind the scene, ActivityImpl are now automatically refcounted.
1585 Removing this nasty bottleneck will greatly simplify our internals.
1588 - Replay: New function xbt_replay_action_get():
1589 Retrieve the function previously associated to an event type.
1590 - DROPPED FUNCTION: xbt_str_varsubst()
1591 - DROPPED MODULE: strbuff. We don't need it anymore.
1592 - DROPPED MODULE: matrix. We don't need it anymore.
1593 - DROPPED MODULE: lib. We don't need it anymore.
1595 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1597 SimGrid (3.15) Released March 22. 2017
1599 The Spring Release: continuous integration servers become green
1600 We fixed even the transient bugs on all target architectures:
1601 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1602 Windows, FreeBSD, NetBSD.
1604 - Rename NetCards to NetPoints.
1605 This was intended to help NS3 users, but that's not a netcard.
1606 That's a point in the routing algorithm, let's avoid wrong simplifications.
1609 - New: MSG_process_yield(). Stop and yield to other processes.
1610 - New: MSG_process_daemon(). Daemon processes are automatically killed
1611 when the last non-daemon process terminates
1612 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1613 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1614 to make room for the upcoming network energy plugin.
1615 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1618 - Ensure that an actor can kill itself with Process::exit()
1619 - Kill the obscure NativeException. Nobody want to survive the issues
1620 it denotes, so use JniException that is a RuntimeException (not to
1621 be caught explicitly).
1622 - Partial bug fix in initialization. SimGrid flags on command line were
1623 consumed at C level but stayed in the original Java String[] args.
1624 This could mess users' args[i] if SG flags were not put at the end of
1626 The SimGrid flags are now removed from the Java arguments. However,
1627 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1628 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1629 It might be if you have --log or --cfg flags in the command line.
1630 - Fix numerous memleaks all around the place. In particular, around VMs.
1634 - s4u::onPlatformCreated: right before the simulation starts
1635 - s4u::onSimulationEnd: right after the main simulation loop
1636 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1637 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1638 event from the availability_file changes the avail speed.
1639 - Links are now usable from s4u
1640 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1641 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1642 - New examples: The conversion of MSG examples to S4U has begun
1643 - Actors: create, kill, migration, and suspend
1644 - Applications: master-worker and token-ring
1645 - Action replay: communications and storage
1646 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1649 - Backwards Compatibility breaks
1650 - The SD_route_get_list and SD_route_get_size functions have been removed.
1651 They are replaced by the unique
1652 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1653 The route from src to dst is built in the links dynar whose size can
1654 be obtained with xbt_dynar_length.
1655 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1656 sg_host_route_bandwidth and sg_host_route_latency.
1657 Macros ensure the backwards compatibility, but you should fix your code
1660 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1661 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1662 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1663 We don't support 100% of it, but it should be enough. Contact us if not.
1664 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1665 naive and probably inaccurate)
1666 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1667 mapped onto a unique small file using some system magic.
1668 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1669 - Fortran: cleanups, fixes, support of user-added operations
1670 - RMA: MPI_Accumulate are applied in correct order.
1671 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1672 - Better support for MPI_IN_PLACE
1673 - Support for MPI_Win attrs and keyvals.
1674 - Support MPI_Comm_group_create, MPI_Type_size_x
1675 - MPI ops should be properly applied to all allowed types
1676 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1677 algorithm by a non failing one from mpich.
1680 - Kill the fifo data container: we don't use it anymore.
1683 - A <cluster> can now be created with different speed values to represent pstates.
1684 see examples/platforms/energy_cluster.xml for an example.
1686 SimGrid (3.14.159) Released December 28. 2016
1688 The previous release was only a crude approximation of a Pi release;
1689 this one should be a bit better.
1691 - Revert a last minute change that broke on Mac OSX.
1692 - Fix the changelog and NEWS files.
1693 - Improve the documentation, in particular of the routing module.
1694 - Integrate some patches flying around in Debian.
1696 SimGrid (3.14) Released December 24. 2016
1699 * The whole documentation was reworked and reorganized. There is
1700 still much room for improvement here, but we're on it.
1703 * We now need python3 for our testing infrastructure.
1704 * Model checking needs libevent
1707 * Now works on FreeBSD
1710 * Privatization now works on FreeBSD
1712 * Privatization is now activated by default in smpirun
1713 You can turn it off with -no-privatize if something goes wrong.
1715 * Call-location tracing for SMPI.
1716 You can add the exact location (filename / linenumber) of an MPI call to
1717 your trace files and slow down or speed up the simulation between two
1718 consecutive calls by using an adjustment file (see the documentation).
1720 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1721 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1722 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1723 values were ignored in some cases. The timings of these functions can now
1724 be significantly different.
1726 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1727 smpi/running-power is renamed to smpi/host-speed
1729 * smpi/grow-injected-times option to enable or disable multiplication of the
1730 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1731 make simulation less precise (but also much faster).
1733 * smpirun script should be (much) faster for large deployments.
1735 * SMPI tracing: fixed issue with poor matching of send/receives.
1737 * Replay: Fix broken waitall
1739 New functions and features
1740 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1742 Dropped / renamed functions and features
1743 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1744 - MSG_mailbox_is_empty() -> Mailbox::empty()
1745 - MSG_mailbox_front() -> Mailbox::front()
1746 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1747 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1748 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1749 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1750 * Don't pass the free_f parameter to property related functions:
1751 - MSG_host_set_property_value()
1752 - MSG_as_router_set_property_value()
1753 - MSG_storage_set_property_value()
1754 * VM properties. Since msg_vm_t are msg_host_t, just use
1755 MSG_host_get_property_value() and friends
1756 * VM I/O related things:
1757 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1758 - Unimplemented save/restore methods
1759 * MSG_as_router_get_property_value() was redundant with
1760 MSG_environment_as_get_property_value().
1761 - Removed MSG_as_router_*propert*() functions
1762 - Added MSG_environment_as_set_property_value() for consistency
1763 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1764 Well, they are still there for now, but deprecated with a warning.
1765 Please switch to xbt_dict_new_homogeneous() before this is removed
1767 * Task affinity. Its intended behavior (that was very badly tested
1768 and probably not really working) was deceiving what most users
1769 would have hoped here.
1770 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1771 * Option network/coordinates is now useless and should be dropped.
1774 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1777 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1778 * Added Dragonfly topology support
1781 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1782 EiBps, EBps, Eibps, Ebps for bandwidth.
1783 They may become useful to some lucky ones.
1786 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1787 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1788 * ex/async_waitAll: new example, on asynchronous communications
1791 * Memory usage should be decreased for simulations with a large number
1792 of processes. This also helps for SMPI.
1794 SimGrid (3.13) stable; urgency=low
1796 The Half Release, a.k.a. the Zealous Easter Trim.
1798 - We removed half of the lines, that were mostly experimental cruft.
1799 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1800 (not counting blanks and comments -- according to openhub.net)
1801 - The internals are now compiled in C++ (and will soon be clean C++)
1802 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1803 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1805 * Backwards Compatibility breaks
1806 - Removed Lua simulation bindings (switch to C or Java for that).
1807 Lua can still be used to describe platforms
1808 - Removed Java kernel plug-ins.
1809 Will be reintroduced after the ongoing major internals reorg.
1811 - the following functions were removed.
1812 They were too specific and should be reimplemented in a generic
1813 way, with filter function.
1814 - MSG_task_listen_from_host
1815 - MSG_mailbox_get_count_host_waiting_tasks
1816 - MSG_mailbox_put_with_timeout was removed.
1817 Please use MSG_task_send_with_timeout instead.
1819 - the SD_application_reinit function was removed. It has been a noop for a while.
1820 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1821 replaced by a more flexible mechanism.
1822 - The following functions thus do not exist anymore
1823 - SD_workstation_get_access_mode
1824 - SD_workstation_set_access_mode
1825 - SD_workstation_get_current_task
1826 - Basic estimation functions have been removed but can easily be replaced
1827 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1828 - SD_workstation_get_computation_time => amount / sg_host_speed()
1830 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1831 Use VM.getSpeed()*load/100 for the legacy behavior.
1833 - option enable_tracing was removed. It was not doing anything for a while.
1834 - In the ModelChecker:
1835 - the model-checker now ptraces the model-checked process which means
1836 you cannot use a debugger on the latter anymore (we might make this
1837 optional in the feature);
1838 - removed soft-dirty page tracking;
1839 - remove model-checked side snapshot management,
1840 MC_snapshot() and MC_compare_snapshot();
1841 - keep the MC_cut() function as a stub (it was not really working
1842 in the previous release).
1845 * All options are consistently in kebab-case. Old names are kept as alias.
1848 * Switch to platform v4 format.
1849 - Rename from 'power' to 'speed' the attributes describing the amount of
1850 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1851 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1852 - In <host> and <link>, attributes availability and state are gone.
1853 It was redundant with state and availability traces, and with peak values.
1854 - In <cluster>, attributes availability_file and state_file are gone.
1855 It was too complex and unused.
1856 - Kill <gpu>. Was not doing anything.
1857 - The DOCTYPE points to the right URL:
1858 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1859 (the file at this address now documents the changelog since its v1)
1860 - A warning is emitted for unit-less values (they are still accepted).
1861 - speed. Default: 'f' or 'flops'. Also defined:
1862 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1863 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1864 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1865 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1866 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1867 - latency. Default: 's' second. Also defined:
1868 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1870 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1871 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1874 * s4u::Host is now the preferred public interface to the Host features.
1875 sg_host_* functions are C bindings to the exact same behavior
1876 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1879 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1880 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1881 They were deprecated since a few years, and probably did not even compile anymore.
1884 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1885 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1886 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1887 encouraged to update their codes. A list of the modified functions can be found at the end of
1888 include/simgrid/simdag.h
1891 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1892 the lambda or closure passed as a parameter will run in kernel mode.
1893 All the callback functions should be rewritten to that interface at some point.
1896 * Reorganizing and cleaning the internals all around the place.
1899 * Remove old default barrier/bcast buggy algorithms (see #18407)
1900 * Various bug fixes to handle more codes
1901 * Remove the need for the --foreground option of smpirun (it is still
1902 accepted for backward compatibility).
1905 * Kill the setset data container: MC don't use it anymore.
1906 * Kill the queue data container: it made more sense with GRAS.
1907 * Kill the xbt_peer_t data type: it's useless without GRAS.
1908 * Kill rm_cb feature of config sets: it was never useful.
1909 * Kill graphxml parsing feature. It was not used.
1910 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1912 - xbt_dynar_sort_strings(), when the content is char*
1913 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1914 They throw exceptions on invalid input;
1915 * C++ support for declaring CLI flags (simgrid::config::Flag);
1916 * class for abstracting different signal backends (simgrid::xbt::signal).
1917 with no external dependencies (we need very simple signals).
1920 * refactoring and cleanup of the code;
1921 * ongoing process to cleanly separate the model-checking algorithms
1922 from the code model-checking support.
1924 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1926 SimGrid (3.12) stable; urgency=low
1928 The Facelift Release.
1931 * Require g++ v4.7 at least to not speak prehistorical C++.
1932 * Require Boost 1.48 (for signal2 component).
1933 * Java must be version 7 at least when activated.
1934 * Builds on Windows again (including Java bindings).
1935 * Tracing is now always enabled (no way to turn it out)
1936 * Remove GTNetS. It was not working anyway.
1937 * Various cleanups in the cmake scripts.
1938 * Move headers around to sort them out on installed systems:
1939 - instr/instr.h -> simgrid/instr.h
1940 - instr/jedule/* -> simgrid/jedule
1941 - simdag/datatypes.h was removed
1942 - simdag/simdag.h -> simgrid/simdag.h
1943 - msg/datatypes.h was removed
1944 - msg/msg.h -> simgrid/msg.h
1947 * Interface improvement:
1948 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1949 - Sanitize the interface in MSG_task_ module:
1950 - Merge two functions that were close enough but misleading:
1951 set_compute_duration(t) -> set_flops_amount(t)
1952 get_remaining_computation(t) -> get_flops_amount(t)
1953 - set_data_size(t) -> set_bytes_amount(t)
1954 get_data_size(t) -> get_bytes_amount(t)
1955 - Massive cleanups in the functions related to the energy
1956 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1957 - New: MSG_host_get_pstate()
1959 - msg/energy/onoff: switching hosts on and off
1961 * Interface improvement:
1962 - Rename Host.isAvail() to Host.isOn()
1963 - Rename Process.currentProcess() to Process.getCurrentProcess()
1964 - Rename Task.setDataSize() to Task.setBytesAmount()
1965 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1967 - #18874: Actually allows the GC to reclaim tasks
1971 - SIMIX_process_throw: raises an exception in a remote process
1972 * Refactoring: Separate sub-modules
1973 - libsmx: the public interface, as libc in a real system
1974 - popping: the strange dance that converts a user request into a kernel handling
1975 - smx_context_*: the virtualization mechanisms that embed the user code
1976 - smx_*: the handling of each simcalls
1977 * Interface cleanup:
1978 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1979 * Rename smx_action_t into smx_synchro_t, making explicit that these
1980 things are used to synchronize processes with their environment.
1981 For example, a communication is a sort of synchronization involving
1982 the communicating processes (that may block until the exchange) and
1983 the platform. The same can be said from computations, etc.
1985 - #18888: segfault when a process ends before its kill_time
1989 - 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.
1990 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1991 - MPI_Type_set_name, MPI_Type_get_name
1992 - MPI_*_c2f and MPI_*_f2c functions
1993 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1994 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1995 - Activate a lot of new tests from the mpich 3 testsuite
1997 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1998 - InfiniBand network model added: Based on the works of Jerome Vienne
1999 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
2000 - When smpi/display_timing is set, also display global simulation time and application times
2001 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
2002 * Collective communications
2003 - 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
2004 - 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/).
2005 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
2007 - Replay now uses algorithms from wanted collective selector
2008 - Replay can be used with SMP-aware algorithms
2009 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
2010 - 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.
2012 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
2013 - Correctly use loopback on fat-tree clusters
2014 - Asynchronous small messages shouldn't trigger deadlocks anymore
2015 * Energy/DVFS cleanup and improvement
2016 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
2017 - new: smpi_get_host_pstate()
2020 - "Full" network optimization flag was broken since Surf++
2021 - Better handling of precision flags in maxmin
2022 - Fix bug causing sometimes "Impossible" errors
2023 - Properly pass cluster properties to included hosts
2024 * Improvement of the Energy plugin.
2025 - Always update the consumption before returning that value
2026 - New property: watt_off to denote the dissipation when the host is off
2027 - New functions getWattMinAt and getWattMaxAt to retrieve the
2028 dissipation of pstates that we are not currently at.
2029 * Java: class NetworkLink renamed to Link
2030 * New function: simcall_process_get_kill_time()
2031 * Massive rename s/workstation/host/
2032 - That's intrusive, but that's good for the project consistency. Sorry.
2033 - Change config option "workstation/model" into "host/model"
2036 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
2037 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
2038 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
2039 - xbt_dirname and xbt_basename for non-POSIX systems
2041 * The model checker now runs as a separate process.
2042 * The model checker runs must now be launched with the new simgrid-mc program.
2043 * Record/Replay: the MC can display a textual representation of a path in the
2044 execution graph. It can then be replayed outside of the model checker.
2046 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2048 SimGrid (3.11) stable; urgency=low
2053 * Normalizing pointers addresses tool for better diff between logs
2056 * Add cloud examples using new VMs
2057 - examples/msg/cloud/two_tasks_vm.tesh
2058 - examples/msg/cloud/simple_vm.tesh
2059 - examples/java/cloud/cloud.tesh
2060 - examples/java/cloud/migration/migration.tesh
2061 * Add java surf examples:
2062 - examples/java/surfPlugin/surf_plugin.tesh
2063 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
2064 - examples/java/surfCpuModel/surf_cpu_model.tesh
2065 * Add SMPI+MSG example:
2066 - examples/smpi/smpi_msg_masterslave/
2071 - msg task destroy cancel test
2072 - msg_host on/off test
2073 * Move all tests in testsuite to teshsuite (adding tesh files)
2074 * Restructure teshsuites
2075 - one folder for each kind of test
2076 * Restructure AddTests.cmake
2078 - structure the order of tests (with sections)
2081 * Add virtual machine
2082 - creation of a VM on a PM
2083 - migration of a VM from a PM to another PM
2085 - MSG_process_join(msg_process_t process, double timeout)
2086 - msg_bar_t MSG_barrier_init(unsigned int count)
2087 - int MSG_barrier_wait(msg_bar_t barrier)
2088 - void MSG_barrier_destroy(msg_bar_t barrier)
2089 - msg_as_t MSG_environment_as_get_by_name(const char * name)
2090 * New option "msg/debug_multiple_use" to help debugging when a task is used
2093 - Improvements and finalization of MSG_storage, MSG_file APIs and their
2095 - Increase code coverage in test suites
2099 * Protect context stacks against stack overflow. The number of protected memory
2100 pages allocated on the top of each stack (1 by default) can be configured
2101 with the new command line option --cfg=contexts/guard_size:<value>.
2102 * Simcalls are now generated by a python script that
2103 - generates files included by SimGrid
2104 - checks that all the functions exist, and proposes prototypes
2106 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
2107 host_set_data, host_get_data
2109 - simcall_process_join(smx_process_t process, double timeout)
2110 * Fix bug where sleeping processing could not be suspended.
2113 * Translate surf models from C to C++
2114 - Generic classes for all models: Model, Resource, Action
2115 - A generic interface for each kind of model (CPU, Network, Storage
2116 Workstation, WorkstationVM)
2118 * Translate surf routings from C to C++
2119 * Add callbacks using sigc++ or boost::signals2
2120 - Add callback functions for resource creation/destruction
2121 - Add callback functions for action state change
2122 - Handle Energy as a plugin
2123 * Replace swag by boost::intrusive
2124 * Add new routing models for clusters. For documentation, see
2125 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
2126 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
2127 parameters for cluster tag
2128 - Fat trees, with topology="FAT_TREE" and
2129 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
2130 - see examples/platforms/torus_cluster.xml and
2131 examples/platforms/fat_tree_cluster.xml
2132 * More documentation
2135 * Hostfiles support host:nb_processes construct to deploy several processes on
2137 * Collective communication algorithms should not crash if used with
2138 improper number of nodes and report the error.
2139 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
2140 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
2141 MPI_Dims_create, MPI_Cart_sub are supported.
2142 * New interface to use SMPI programmatically (still depends on MSG for
2143 some parts, see examples/smpi/smpi_msg_masterslave):
2144 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
2148 * Global variables privatization in MPI executables is now performed at runtime
2149 with the option smpi/privatize_global_variables (default:no).
2150 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
2151 dynamic libraries loaded by the application are not privatized (static
2152 linking with these libraries is advised in this case)
2155 * Options defined in XML work correctly now.
2158 * New cmake option, enable_lib_in_jar, to control whether native libraries are
2159 copied into simgrid.jar or not (ON by default). Use this option if you want
2160 to reduce the size of the installed simgrid.jar, *and* the native libraries
2161 are kept installed elsewhere.
2162 * Surf binding with SWIG (code generated in maintainer mode only):
2163 - plugin to handle callbacks
2164 - CPU model only for the moment
2167 * Supernovae build mode is definitively removed. It was used to improve
2168 inlining and inter-module optimizations. It is nowadays superseded by
2169 link-time optimizations commonly available in compilers.
2170 * Update ns-3 find lib. Bindings for ns-3 should work again now.
2171 * Add boost dependency for surf++
2172 * Add new macro for tests
2173 - ADD_TESH(name <tesh_args>)
2174 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
2178 - xbt_bar_t XBT_barrier_init(unsigned int count)
2179 - int XBT_barrier_wait(xbt_bar_t barrier)
2180 - void XBT_barrier_destroy(xbt_bar_t barrier)
2181 * Make the xbt_os_time module public
2183 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2185 SimGrid (3.10) stable; urgency=low
2187 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
2190 * Reintegrate Java to the main archive as desynchronizing these
2191 package is not acceptable anymore (Java is now considered stable)
2192 * Add explicit synchronization facilities through semaphores
2193 * Bug fix: Task.setDataSize() only changed the C world, not the value
2194 cached in the Java world
2197 * Dramatically change the way files are handled. API and internals changed, but
2198 this part of MSG was not considered as production grade either.
2199 * Add explicit synchronization facilities through semaphores
2200 * Add a new function MSG_host_get_process_list()
2201 * Preliminary DVFS support (see examples/msg/energy/ for details)
2204 * SMPI is now included directly in the libsimgrid as the windows
2205 linker doesn't force us on splitting it anymore.
2206 * Improvements of the SMPI replay tool:
2207 - Most of the collective communications are now rooted in the same process as
2208 in the original application.
2209 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
2210 used until now). Multiple data types can now be used in a trace.
2211 - The replay tool now supports traces produce either by TAU or a modified
2213 - Bug Fix: the compute part of the reduce action is now taken into account.
2214 - Gatherv collective is now supported
2215 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
2216 of smpirun outputs time independent traces for the current run. One file
2217 is created per process. If too many processes are simulated, this behavior
2218 can be changed to one file for all processes by using the
2219 tracing/smpi/format/ti_one_file flag
2220 * smpirun generates the host file if needed (with given host count and platform)
2221 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
2222 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
2223 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
2224 --cfg=smpi/coll_selector:(mpich/ompi)
2225 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
2227 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
2228 happening inside a collective SMPI call.
2229 * Fix the behavior of complex data types handling.
2230 * Make MPI_Wtime another synchronization point to take computations into
2232 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
2233 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
2234 * Add all missing Fortran bindings, SMPI should work with Fortran 90
2235 (no privatization of global variables yet)
2236 * Preliminary DVFS support (see examples/smpi/energy/ for details)
2239 * Verification of liveness properties is now available for SMPI applications
2240 (in addition to MSG applications)
2241 * Bugged examples using SMPI in examples/smpi/mc/
2242 * Add --cfg=model-check/visited option. Allows the verification of infinite
2243 programs. Detection of loops in the execution thanks to the system state
2244 comparison and reduction of the state space to explore. Can be combined with
2245 DPOR for safety properties.
2248 * Allow to change SimGrid configuration (see --help) within the code
2249 thanks to SD_config() as it can be done in MSG.
2250 * Add a new function SD_task_set_amount() upon user request.
2253 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
2254 * Remove rule based routing (no more PCRE dependency)
2255 * Add a limiter_link option to cluster tag, to specify a maximum reachable
2256 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
2257 * Add a loopback_bw and loopback_lat options to cluster tag.
2258 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
2259 routing. Now peers are encapsulated in an AS and have their own private
2260 router but this is transparent.
2263 * Our own implementation of getline is renamed xbt_getline, and gets
2264 used even if the OS provide a getline(). This should reduce the
2265 configuration complexity by using the same code on all platforms.
2266 * New type: xbt_cfg_elm_boolean.
2267 * Allow to use yes/no for boolean configuration options in the command line.
2268 * Allow to disable SimGrid cleanups at exit from command line option.
2269 There are situations where one may want a simulation to end with an exit.
2270 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
2271 annoying when scripting around the simulator. Adding a
2272 --cfg=clean_atexit:no allows to circumvent this issue.
2275 * Lots of memory leaks were corrected in this release.
2276 There is no known memory leaks anymore, in all of our 600+ tests.
2277 * New command line option --version, to get SimGrid version information.
2278 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
2280 * Supernovae builds are deprecated, and expected to be removed in the next
2283 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2285 SimGrid (3.9) stable; urgency=low
2287 The Grasgory release: GRAS is really dead now.
2289 * Complete overhaul of the internal host structures scheme.
2292 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
2293 was considered as experimental and badly maintained since then.
2294 * Keeping it was thus a trap to our potential users, that could take
2295 it instead of MSG or SMPI by mistake despite is pity state.
2296 * GRAS seems to have very few users (if any), and no one volunteered
2297 to maintain it further. It also induces a lot of XBT code (for
2298 portability sake), that must be maintained too.
2299 * For all these reasons, we killed GRAS. If someone wants to revive it
2300 in the future, don't cry, our git history still remembers of GRAS.
2303 * Major overhaul. Merge our documentation again as time proved that
2304 splitting it was really not helping our users.
2305 * Further improve the developer documentation to help newcomers
2306 hacking on SimGrid itself. The user documentation (and in
2307 particular, the beginner documentation) is still in a sorry state.
2310 * Now works on Windows too!
2311 * Much more extensive test suite, from MPICH
2314 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
2315 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
2316 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
2317 * Bug fix: let task be scheduled when the last dependency to be solved is
2318 a control dependency.
2319 * Remove SD_load_environment_script function.
2320 Use the C sg_platf function if you want to declare a platform
2324 * New function: MSG_process_get_number()
2325 * Old function documented: MSG_config()
2326 * Remove MSG_load_platform_script function
2327 Use the C sg_platf function if you want to declare a platform
2331 * Change the default value of the TCP_gamma constant (maximal size of TCP
2332 congestion window) to a more realistic 4MiB value. If you notice changes in
2333 your simulation results, you can fall back to the previous 20k tiny window
2334 by adding --cfg=network/TCP_gamma:20000 on command line.
2335 * (Hopefully) fix a bug wrt periodic availability and state traces
2336 * Bug fix: use default values at start when first event in availability/state
2337 trace is not at time 0.
2340 * remove the "new_" part of function name sg_platf_new_trace_connect
2341 (resulting in sg_platf_trace_connect), since it does not create
2345 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2346 use case where it's really mandatory, and maintaining it was a pain in
2348 * New: xbt_fifo_search(), search an item with a user-provided
2349 comparison function instead of dumb pointer comparison.
2352 * Fix the lua deployment:
2353 Use `simgrid.init_application()` before deployment instead of
2354 `simgrid.msg_register_application()` after.
2357 * Transfer the tracing files into the corresponding modules.
2359 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2361 SimGrid (3.8.1) stable; urgency=low
2363 The "we are told that some people want to also *install* the simgrid
2366 * Add missing file "tesh.1" to the archive.
2368 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2370 SimGrid (3.8) stable; urgency=low
2372 The Psssshiiiit release: SimGrid jumps into the Cloud.
2375 * Add an experimental interface to manipulate VMs. They are mainly
2376 process groups with very few intrinsic semantic, but they should
2377 allow you to build the semantic you want easily.
2378 * New function: MSG_host_set_property_value()
2379 * New function: MSG_process_on_exit(). To clean memory in all cases.
2380 * Bug fixes that made the host (and link) failures unusable.
2381 * Add a way to auto-restart process when the host in which they are
2382 executing comes back (ON_FAILURE="RESTART" on deployment file,
2383 MSG_process_auto_restart_set).
2384 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2385 please stop using the old ones, they are DEPRECATED.
2387 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2388 Please use MSG_init() instead. (reducing the amount of entry
2389 points in the library helps us).
2390 * Make it impossible to link against the wrong version of the lib
2391 * Deprecate MSG_clean(). No need to call it anymore.
2392 * Function MSG_get_host_number() is not deprecated anymore.
2395 * Split the doc into a user guide and a reference guide.
2396 * Start a developper guide to help people hacking on SimGrid.
2399 * Enable tracing by default. This modules rocks you should use it.
2400 * Remove option custom_flags. Now use environment variables CFLAGS
2402 * Use default cmake things to detect lua instead of home grown ones.
2403 * New option "enable_mallocators" to disable mallocators, for debugging
2404 purpose ("on" by default).
2407 * Bug fixes around the resource failures: don't let the processes
2408 survive the host they are running onto.
2409 * Add an interface to auto-restart processes when the host in which they are
2410 executing comes back.
2411 * Ensures that SIMIX_clean is called automatically. It's not part of
2412 the public interface anymore (bindings should be updated).
2415 * Bug fix for when SD_Simulate is called with a positive value: be careful
2416 when comparing doubles. Sometimes they are different for non significant
2418 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2419 parallel task whose initial work is distributed among host according
2420 to the Amdahl's law. Such tasks are created with a parameter alpha
2421 that corresponds to the non-parallelizable part of the computation.
2422 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2423 between two sets of workstations assuming a 1D block distribution (each
2424 workstation owns a similar share of data) on both sides.
2426 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2427 Data redistribution will be automatically scheduled once parent and child
2428 are both scheduled. The filling of computation_amount and
2429 communication_amount structures is now done seamlessly thanks to the chosen
2431 * New function SD_workstation_dump to display various information
2432 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2433 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2434 bandwidth on the route onto which the task is scheduled and the amount of
2436 To divide the nominal bandwidth by 2, the rate then has to be:
2437 rate = bandwidth/(2*amount)
2438 * Compute tasks that have failed can now be rescheduled and executed again
2439 (from their beginning)
2440 * Increasing source code coverage (src/simdag is now covered at 95.8%
2444 * Re-implement time-independent trace replay using SMPI (at the
2445 smpi_smp_* level) instead of MSG. This should replace
2446 examples/msg/actions/actions.c
2447 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2448 hindexed and structs)
2449 * Implement the exchange of non-contiguous data.
2450 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2451 * Correct behavior of smpi/sender_gap and set its default value to 0
2452 * Add option to asynchronously send small messages to allow better
2453 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2454 specifies the size in bytes under which messages will be asynchronously sent.
2455 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2456 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2459 * New C interface to define a platform: XML is now optional.
2460 For more info, please check include/simgrid/platf.h.
2461 * New interface to define random platforms from the C:
2462 For more info, please check include/simgrid/platf_generator.h and
2463 examples/msg/masterslave/masterslave_platfgen.c
2464 * Export a sg_cmdline dynar containing all the arguments we got from
2468 * Two new tracing options for adding comments to trace file so you
2469 can track your experiments (see --help-tracing for details).
2470 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2471 * Adding the SimGrid version that generated the trace file as a comment.
2472 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2473 * Fix to avoid key clashes on Paje links
2474 * Other minor fixes related to the Paje specification
2477 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2478 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2479 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2480 * Greatly improve the robustness of mmalloc to user errors (such as
2481 using an area after freeing it, or freeing it twice)
2483 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2485 SimGrid-java (3.8.1) stable; urgency=low
2487 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2488 Warning: all methods are visible, but only some of them are
2489 implemented so far. Check the source (src/jtrace.c)
2490 for further information.
2491 * New module: org.simgrid.msg.File (SimGrid File management functions)
2492 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2493 * Change the meaning of Process.restart: now restart the process from
2494 the beginning, like MSG_process_restart in C.
2495 * Add Process.setAutoRestart: handling of process restart when failed
2497 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2498 you to retrieve the properties of the processes/hosts
2499 * Deprecate Msg.clean(): you can just forget about it now.
2500 * New function Process.getCount(), that only works when compiling
2501 with the not yet released version 3.9 of the C library.
2503 * New context factory based on Coroutines. It mandates a modified JVM
2504 but then, the simulations run about five times faster, and there is
2505 no limit to the amount of processes (beside of the available memory).
2507 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2509 SimGrid (3.7.1) stable; urgency=low
2512 * Restore the prototype of MSG_process_create_with_environment() to
2513 the pre-3.7 situation by removing the kill_time argument.
2514 * Add a MSG_process_set_kill_time() function instead.
2517 * Fix weird behaviors when dealing with parallel tasks.
2520 * Simgrid is now built as a dll.
2521 * Simgrid-java now works on Windows.
2522 * Simgrid-Java is now included into Windows package.
2525 * First pre-build package for MacOSX.
2528 * Fix compilation when using MSG_USE_DEPRECATED.
2529 * Fix some compilation issues on Macs and Windows.
2530 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2531 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2533 We discovered that the Lua console is broken, but we are missing the
2534 manpower to fix it right now. The problem existed in 3.7 too, so we
2535 are not blocking the release for that. Sorry if you depended on this
2536 feature, any help would be really welcome.
2538 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2541 SimGrid-java (3.7.1) stable; urgency=low
2543 The "Java aint got to be bloated and slow" release
2546 * Various internal cleanups and performance improvement
2547 Simulations are expected to run up to twice faster or so
2548 * Make Process.kill(process) an instance method, not a static one
2549 * User processes are not java.lang.Thread subclasses.
2550 This breaks the compatibility (sorry), but previous API was
2551 brain-dead, making it impossible to have non-trivial
2552 initializations in the process constructor.
2553 * Require a full constructor per Process sub-class.
2554 Kinda breaks the compatibility (sorry), but this allows a much more
2555 efficient way to launch the processes at simulation startup.
2556 * Do not embeed our version of semaphores, java 1.5 can be considered
2557 as sufficiently prevalent for us to not dupplicate its features.
2561 * Add examples for almost every part of the API
2562 We spotted and fixed a lot of bugs in the process
2564 * New module: asynchronous communication API
2565 * New function: Process.sleep()
2566 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2567 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2568 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2570 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2572 SimGrid (3.7) stable; urgency=low
2574 The "spring cleanups (before the next Big Project kicks in)" release.
2577 * We can specify the SMPI latency/bandwidth factor with command line
2578 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2579 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2580 You can also use the "config tag" from platform file by adding this line
2581 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2582 (see "example/platforms/tag_config.xml" to use "config tag").
2583 Note that the command line supersedes the platform file configuration.
2584 * Change the correction factors used in LMM model, according to
2585 the latest experiments described in INRIA RR-7821.
2586 Accuracy should be improved this way.
2587 * Use the partial invalidation optimization by default for the
2588 network too. Should produce the exact same results, only faster.
2589 * Major cleanup in surf to merge models and split some optimization
2590 mechanisms from the core of the models. As a result you can now
2591 specify which model to use (e.g., --cfg=network/model:LV08
2592 --cfg=cpu/model:Cas01) and which optimization mode to use
2593 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2594 Incompatible combinations should err at initialization. See
2595 --help-models for the list of all models and optimization modes.
2596 * The CLM03 workstation model was dropped for simplicity because it
2597 used the deprecated CM02 network model. Use default instead.
2598 * Rename the TCP_gamma configuration option to network/TCP_gamma
2599 * Rename the coordinates configuration option to
2600 network/coordinates, and document it
2601 * Use now crosstraffic keyword instead of the terribly misleading
2602 fullduplex keyword. It is activated by default now in the current
2603 default model, use --cfg=network/crosstraffic:0 to turn it off.
2604 * Ongoing refactoring the model parsing to make XML files optional
2605 See include/simgrid/platf.h for details (still to be completed)
2608 * Major overhaul of the documentation. Almost instructive now :/
2609 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2610 functions and friends. This interface was considered as
2611 deprecated since over 2 years, it's time to inform our users that it is.
2612 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2613 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2614 use these (crappy) functions in your code.
2615 These functions will be removed soon. Stop using them now.
2616 * Deprecate MSG_get_host_{table,number}
2617 Implement MSG_hosts_as_dynar() instead.
2618 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2619 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2622 * Stabilize the parallel execution mode of user contexts
2623 * Introduce configuration variables to control parallel execution:
2624 - contexts/synchro: Synchronization mode to use when running
2625 contexts in parallel (either futex, posix or busy_wait)
2626 - contexts/parallel_threshold: Minimal number of user contexts
2627 that must be part of a scheduling round to switch to parallel
2628 execution mode (raw contexts only)
2629 * Fix bugs that prevented to use suspend/resume along with
2630 synchronization structures.
2631 * Fix bugs in process termination that lead to invalid memory access
2632 in very specific conditions.
2635 * Introduce a parallel mode for the models (controlled by surf/nthreads
2636 configuration item). In our tests, running the models in parallel
2637 never lead to any speedups because they are so fast that the gain
2638 of computing each model in parallel does not amortizes the
2639 synchronization costs, even when ultra fast futexes are used.
2640 This is released anyway because YMMV.
2643 * Performance boost by using a swag internally to compute the set of
2644 tasks that are finished and should constitute the return value of
2648 * Enable it by default now that it is considered rather stable.
2651 * Documentation of the tracing functions.
2652 * Performance gains when tracing categorized/uncategorized resource
2653 utilization by avoiding calls to get route when updating resource
2654 variables. LMM constraints are being used instead.
2655 * API changed to set task categories. Use MSG_task_set_category instead
2656 of TRACE_msg_set_task_category, and SD_task_set_category instead
2657 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2658 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2659 * Examples for link user variables added (see at examples/msg/tracing/)
2660 * Deprecated function TRACE_msg_set_process_category completely removed
2661 * Trace header updated according to the latest Paje file format
2662 * Tracing network lazy updates, no longer obligate users to use full updates
2663 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2664 * Experimental: let user code declare/set/push/pop application states for hosts
2665 * API changed to allow the manual creation of graph configuration files
2666 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2669 * Improve the API of Lua MSG bindings, using the Lua spirit.
2670 * Each simulated process now lives in its own Lua world (globals are
2671 automatically duplicated). It helps writing simulators. Will allow
2672 to run Splay programs within SimGrid in the future.
2673 * Add a Chord example in Lua, equivalent to the MSG one.
2676 * Start the implementation of a solution to express temporal
2677 properties, not only local assertions. This is still an
2678 experimental work in progress, stay clear from it to be safe.
2682 - Add new runtime parameters --help-logs and --help-log-categories
2683 to display information about supported logging parameters and
2685 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2686 don't exists anymore.
2687 * Mallocators: allow value NULL for the reset function.
2689 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2690 create homogeneous dictionaries, where all the elements share the
2691 same free function. Non homogeneous dictionaries will be
2692 deprecated in the next release.
2693 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2694 - Multi-level dictionaries are deprecated.
2696 - new function xbt_dynar_search_or_negative() that is useful when
2697 you have less than 2 million elements in your dynar and don't
2698 want of the extra complexity of catching exceptions when the
2699 element is not found.
2701 - Make xbt_os_thread module (for thread portability) public.
2702 Documentation is still to come, sorry.
2704 - Cleanups and simplifications to make it maintainable again.
2705 - Exotic features (such as memalign and valloc) were removed.
2706 - The metadata were extended and improved so that the
2707 model-checker becomes able to explore and inspect the heaps.
2708 - This may induce a performance drop when enable_model-checking is
2709 ON in cmake (even if it's not used in the simulation), but it is
2710 necessary at this point to get MC working.
2712 Turn model-checking OFF if simulation performance matters to you.
2713 Not enabling it at runtime is not enough, disable it in cmake.
2715 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2718 SimGrid (3.6.2) stable; urgency=low
2720 The "Not coding new stuff allows to polish old things" release.
2723 * New bindings to the NS3 packet level simulator (experimental)
2724 * Use the raw (efficient) execution contexts instead of the sysv
2725 (portable) ones when possible.
2726 * libpcre is now mandatory in any cases since not using it led to
2727 severe performance loss and possibly other issues
2728 * Update the XML platforms:
2729 - G5K: include the latest machine in Nancy
2730 - GridPP and LCG: new platforms
2731 * Documentation was partially updated, at least (more to come)
2733 Bug fixes, cosmetics and small improvements
2734 * Free terminated processes before the end of the simulation to avoid
2735 exhausting the memory of users having very dynamic amount of
2737 * Bug fix and cosmetics about canceling non-running tasks
2738 * Bug fix about the dot loader's issues when using libcgraph
2741 * Create an installer for windows with nsis (amd64 and win32)
2742 - Add an hello world project to illustrate simgrid project creation.
2743 - Embed libpcre into the Simgrid installer to avoid
2744 its compilation burden
2745 * The raw execution contexts should work on Apple now
2746 * Port to Windows 64 bits
2747 - Sysv contexts now have an implementation for this arch
2748 - GRAS communication features now support this arch
2749 * Drop support for borland compiler on windows
2750 - this code was not maintained, and we kinda depend on gcc nowadays
2751 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2752 * Fix portability issue on unstable ubuntu: linker became picky on
2755 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2757 SimGrid (3.6.1) stable; urgency=low
2759 The "Oops, we broke Macs too" release
2762 * Fixed contexts detection so that raw ones are used when possible
2763 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2764 leads to a strange error, with user code segfaulting sometimes when
2765 the generated code is not perfectly aligned (which is not
2766 controllable from the user side, depends on the amount of code)
2769 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2771 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2773 SimGrid-java (3.6) unstable; urgency=low
2776 * Split of every thing from simgrid v3.5 into a separate package.
2778 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2780 SimGrid (3.6) stable; urgency=medium
2782 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2785 * Bindings now constitute their own package, separated from the main one.
2786 Rationale: reduce our maintenance nightmare by reducing the module coupling
2787 They will soon be released on their own on gforge.
2789 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2790 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2792 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2793 * It's not quite deprecated for now because we have no replacement,
2794 but it may soon become the case.
2797 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2798 * Fortran: New user-level cache variable to store the rank of the running
2799 process. This improves performance by an order of magnitude.
2800 * C: New coccinelle script to automatically locate and modify global and
2801 local static variables.
2802 * Improved SMPI network model with a sender-side gap to account for multiple
2806 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2807 only say if a communication is finished, no matter whether it succeeded or
2808 failed. You can call MSG_comm_get_status() to know the status of a finished
2810 * New function MSG_task_dsend() to send a task and detach it. When a
2811 communication is detached, you are never notified of its success or failure
2812 and the memory is released automatically once it is finished. This function
2813 is useful when you don't care about the end nor the success of a
2815 * Change the prototypes of action replay. Sorry for inconvenience,
2816 but this is really more efficient this way (and to adapt your code,
2817 you just have to fix the initialization, that shouldn't be too long)
2818 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2819 ever managed to do anything useful with it.
2820 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2821 takes a dynar of communications. It returns immediately and gives the
2822 index of a finished communication (if any).
2823 * New example: a basic implementation of the Chord P2P algorithm.
2826 * New model for multi-core CPUs. You can now use the core attribute to
2827 precise the number of cores of a host. This is a basic model. Every
2828 process running on the host receives at most the power provided in
2829 the DTD (throughput<=power). Total throughput of process cannot exceed
2831 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2832 router linked by an up-link and a down-link (possibly asymmetrical).
2833 This kind of pattern allows to easily build last-mile model style platforms.
2834 Aggregating such patterns in a rule-based AS is thus the technique of
2835 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2836 * New model for Vivaldi routing. We transformed the Vivaldi network model
2837 into a Vivaldi routing model (based on the rule-based model). This allows to
2838 combine Vivaldi based latencies with last-mile platforms.
2841 * Added a check for NaN of IEEE754 infinite in the double entries of
2843 * Introduce a new context factory "raw", highly inspirated from the
2844 ucontext factory, but using manually crafted functions in assembly to
2845 do the work in an efficient manner.
2846 * Allow to change the used context factory at run time, not only at
2847 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2848 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2849 contexts at runtime (only with raw contexts or ucontexts).
2850 * Completely rewrote this module to allow parallel execution of user
2851 processes. Use --cfg=contexts/nthreads:N to execute user processes
2852 with N parallel threads (the default is 1, meaning no parallelism).
2853 * Allow to decide dynamically between sequential and parallel modes.
2854 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2855 processes in parallel only when their number is greater than or equal to P
2857 * Added a check for NaN of IEEE754 infinite in the double entries of
2861 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2862 won't output the state of processes when interrupted with Ctrl-C
2863 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2864 NULL-terminated array. This may solve backward compatibility issues
2865 due to the change to return type of SD_simulate. See also:
2866 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2867 * Add new macros with variable number of arguments.
2868 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2869 - in xbt/asserts.h: xbt_assert
2870 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2871 - in xbt/ex.h: THROWF and RETHROWF.
2872 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2874 * Change xbt_die() to accept a format string with arguments, just like printf.
2875 * New data structure: xbt_lib_t, like a dict but more general and with better
2879 * New configuration options
2880 Options triva/categorized and triva/uncategorized can be used to generate
2881 graph configuration files for Triva visualization tool.
2882 * Configuration option tracing/platform is renamed to tracing/categorized
2883 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2884 * New TRACE_user_link_variable interface:
2885 User provides the name of the link and the tracing variable to attach to it
2886 * the declaration of tracing categories must be done after the environment
2888 * simpler tracing interface, just one way to declare categories
2889 TRACE_category or TRACE_category_with_color, it is up to you
2890 * links in the trace file are again identified by their names
2891 * trace contains the full platform hierarchy exactly as declared using the ASes
2892 * Options tracing/msg/[task|process]:1 groups the process by hosts
2893 for both cases, tasks and processes must have names that are unique during the simulation
2894 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2895 * The experimental option tracing/msg/volume is deprecated
2896 its functionality may be reincorporated if needed
2898 The tracing generates a trace file with unordered timestamped events,
2899 because of the way the core simulator (surf) works. A script available
2900 at the tools directory (fix-paje-trace.sh) can be used to put the events
2901 in order. We have changed the tracing so it can generate ordered timestamped
2902 events in the final trace, but depending on the simulator (and how much time
2903 is simulated) that can lead to a huge memory utilization. It is deactivated
2904 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2906 Build Infrastructure
2907 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2908 - We are trying hard to keep the API stable, but it may happen that
2909 some things change (we're a research project after all, not a
2910 nuclear plant operating system). If such things should happen, you
2911 could rely on that macro to adapt.
2912 - current value: 30600 for 3.06.00, aka 3.6
2913 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2914 a number that can be compared with SIMGRID_VERSION.
2915 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2916 assertions and verbose/debug logging events are disabled at compile time.
2918 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2920 SimGrid (3.5) stable; urgency=medium
2923 * New feature to any SimGrid-based simulator: Model-Checking
2924 Check SIN#1 for more details.
2927 * New Model SMPI (three-interval linear regression for correction factors)
2928 See RR-7426, available at http://hal.inria.fr/inria-00527150
2929 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2930 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2931 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2932 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2933 * See also src/smpi/README
2938 - Tracing API changes: TRACE_start and TRACE_end should not be called
2939 by user-code. They are automatically called by simulators created
2940 with SimDAG, MSG and SMPI if the toolkit is compiled with
2941 tracing_enabled=ON. Categories declaration and utilization remain the
2942 same for MSG and SimDag.
2943 - A function was added to the tracing API to declare categories with
2945 - TRACE_category_with_color (char *category, char *color)
2946 where color must be in the following format
2947 "%f %f %f", red, green, blue
2948 and red, green, blue are float values in the interval [0, 1]
2949 - User can specify NULL as color parameter, or continue calling
2950 TRACE_category (cat)
2951 On that case, the tracing system will define random colors
2952 - The following command-line options are supported:
2953 --cfg=tracing/filename:msg.trace
2954 --cfg=tracing:1 (activate tracing, needed to use others)
2955 --cfg=tracing/platform:1 (categorized resource use)
2956 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2957 --cfg=tracing/msg/task:1 (task creation)
2958 --cfg=tracing/msg/process:1 (process creation, migration)
2959 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2960 --cfg=tracing/smpi:1 (SMPI interface tracing)
2961 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2962 - examples of examples/msg/tracing updated
2964 - DAXLoader and DOTLoader functions can generate tasks with categories
2965 - A new function to attribute a category to SD tasks:
2966 TRACE_sd_set_task_category (SD_task_t task, char *category)
2967 * Tracing the MPI interface implemented by SMPI
2968 - Collective operations are traced with states
2969 - Point-to-Point operations are traced with states/links
2970 - Tracing activated by a parameter "-trace filename" passed
2971 to smpirun during execution (considering that simgrid
2972 is compiled with tracing enabled)
2973 - To run the simulation with gdb, the simulator
2974 accepts --cfg=tracing/smpi:1 to trace SMPI
2975 - tesh files to check if smpi tracing is ok
2976 - See examples/smpi/NAS/DT-trace
2977 * GTNetS tracing re-worked
2978 - adaptation to the tracing system of GTNets to cope
2979 with modifications regarding the fullduplex mode
2980 - new tesh files to check if gtnets tracing is ok
2983 * Asynchronous communications through the functions:
2984 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2985 * New function: MSG_load_platform_script()
2986 to make possible using a lua script instead of XML files to set up platforms
2987 * New function: MSG_set_function
2988 to associate functions to processes, used when bypassing the parser
2989 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2991 Platforms: Add some more examples in examples/platforms
2992 * Grid'5000: see www.grid5000.fr
2993 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2996 * Change the XML format. This is a very important modification. SimGrid
2997 3.5 introduces a new hierarchical format based on the notion of
2998 Autonomous Systems. Compatibility with old format is ensured
2999 through the perl script provided in the install bin directory
3000 bin/simgrid_update_xml.
3001 It is now possible to build platforms with specific routing
3002 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
3003 connect several platforms together. We will try to provide soon
3004 set of realistic platforms exploiting these properties (have a
3005 look at examples/platforms/ for the moment).
3006 * Take the opportunity of the XML format change to be a good XML citizen:
3007 rename link:ctn to link_ctn and similar changes (also dealt with by
3009 * Add a new routing scheme (rule-based) using regular expressions. It
3010 enables to have an extremely low memory footprint when the
3011 underlying routing is simple and can be compactly described. You
3012 need to have libpcre4-dev (perl regular expressions) installed if
3013 you want to use this routing scheme.
3014 * Revive the cluster TAG and allow to easily and efficiently (both in
3015 term of memory and speed) connect clusters together. Have a look
3016 at teshsuite/simdag/platforms/ to see how this can be done. With
3017 this tag, you can create clusters with thousands of tasks at no
3018 cost (have a look at examples/platforms/).
3019 Note: clusters are implemented as ASes, so there is no need for an
3020 enclosing AS tag if you have only one cluster in your platform.
3021 * Add new generic functions in the public interface that allows the user
3022 to call SURF 'create_resource' methods from your code (same
3023 functionality as the XML bypass mechanism but with a much lighter
3025 * Add a new model (enabled through command line --cfg=network/model:SMPI)
3026 that uses a piecewise linear approximation to produce better
3027 results when exchanging small messages.
3028 * Add a new parameter to handle correctly full duplex link and account
3029 for interferences between uplink and downlink communications
3030 (activate with --cfg=fullduplex:1).
3033 * Rename the SD_READY (all dependencies are satisfied and task is
3034 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
3035 dependencies are satisfied) state.
3036 This prevents a confusion between the notion of "ready to schedule"
3037 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
3038 simulated" (SD_RUNNABLE) used by the simulation kernel.
3039 * Change the way a task is considered as ready. Instead of removing
3040 dependencies when a task is done, a counter is decreased. This way,
3041 it is always possible to reach ancestors thanks to the
3042 SD_taks_get_parents function (even after the end of the simulation.)
3043 * Change the return type of SD_Simulate from (SD_task_t*) into
3044 xbt_dynar_t. This function was in handling a dynar internally and
3045 converted it into a NULL terminated array for historical reasons.
3046 * New function SD_dotload(char*) to load a DAG described in dot
3047 format. This loader and the corresponding examples require the
3048 installation of the graphviz library.
3049 * Fix a bug in the management of tasks of size 0 in the surf network
3050 models. This problem was only visible with SIMDAG and you should
3051 thus disregard results produced with earlier versions if you
3052 relied on this feature (some tasks were blocked because of this).
3053 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
3054 even though your applications were doing only point-to-point
3055 communications and sequential computations. Now you can really use any
3056 model you want (of course, if you create real parallel tasks, which are
3057 not implemented in most models beside ptaskL07, this will abort).
3058 * Add an example that schedules a DAX on an heterogeneous platform
3059 using a Min-Min strategy.
3060 * New function SD_workstation_get_current_task() that returns the kind
3061 of task currently running on a workstation in the sequential access
3063 * Raise some warnings when unexecuted tasks remains at the end of the
3064 simulation. This is usually caused by cycles in the DAG.
3067 * New function: SIMIX_process_set_function() called by MSG_set_function
3068 * Change the underlying waiting queue in semaphores so that a process
3069 can wait on several of them simultaneously (as in waitany).
3070 * Fix the way to handle tokens in semaphores so that all access patterns
3071 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
3072 * kill the dirty pimple SIMIX_message_sizes_output()
3073 Please use (proper) visualization instead
3076 * New data container: setset (set of sets of elements)
3077 * New module: mmalloc (mapped malloc, allowing to have several
3078 independent segments of malloc)
3079 * New function: xbt_dict_cursor_set_data()
3080 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
3081 * New function: xbt_dynar_is_empty()
3082 * New function: xbt_fifo_get_last_item()
3083 * Fix xbt_dynar_shrink(): use the right element size.
3084 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
3085 uninitialized areas during expand.
3086 * Fix semaphores: previous implementation was severely broken.
3087 * Use library init/fini functions for our initialization.
3088 - you can use logs and other feature as soon as you want in your
3089 code (even before the xbt_init / MSG_init)
3090 - xbt_exit is now a no-op and produce a warning when used.
3093 * Port GRAS to new SIMIX mechanisms. This allows gras users to
3094 benefit from the latest improvement to the simulation kernel.
3095 * Kill measurement sockets for now. If you rely on them, sorry. This
3096 release is not for you. This feature will be reintroduced in the
3097 future, but we cannot delay the release any further.
3098 * New function: gras_msgtype_get_name().
3099 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
3100 * Fix (at last) the pmm example: it should not randomly fail anymore.
3102 Build chain: bug fixes and overall polishing
3103 * Cmake is now stable enough. Hence, we killed the autotools.
3104 * Port to windows ( TM :)
3105 * Fix the 'make install' target.
3106 No need to use 'make install-simgrid' anymore
3107 * Introduce a 'make dist' target compiling a *source* archive
3108 'make package' compiles a binary archive
3109 * Compile java files only on need
3110 * Add --cd and --setenv command line options to tesh
3111 * Out of source builds are not fully supported yet, but we are close
3112 * Enable supernovae and optimization flags by default for our users
3115 * Add layer to set up environment directly from lua, without XML.
3116 * The effect of gras_stub_generator can be achieved through
3117 lua too (check examples/gras/console/ping_generator.lua)
3119 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
3121 SimGrid (3.4.1) stable; urgency=low
3123 The "Polishing easter eggs is probably a good idea" release.
3124 This is a bug fixes release only.
3128 * Fix a bug preventing the tasks from begin garbage collected.
3131 * Fix a bug occurring when a host involved in a communication fails.
3132 This was not detected properly by the other peer involved in the
3133 communication. Now, it's reported as a network error.
3136 * Warn the user about loop dependencies in data flow of DAX files
3137 * Obey the control-flow dependencies of DAX files
3140 * Add option "enable_smpi" allowing to not compile SMPI.
3141 Probably useful for the (Mac) users experiencing a build error here
3142 * Improve the detection of lua5.1 and ruby1.8
3144 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
3146 SimGrid (3.4) stable; urgency=low
3148 The "Easter in Cargese" release. Also known as (major changes):
3150 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
3151 ~> bindings were greatly improved
3152 ~> new tracing infrastructure for better visualization introduced
3154 * the "Welcome to configury modernity" release.
3155 ~> we switched from autotools to cmake, and improved our cdash
3158 A more detailled list of changes follow (full detail in svn log).
3160 Java Bindings: Various Cleanups
3161 * (install java-gcj-compat-dev on debian-like to use them)
3162 * Remove put/get: no need to export deprecated interface in Java
3163 Use send/receive instead.
3164 * Cleanup the examples and add a README per directory
3165 * Remove example autoDestination (that's the only way to go now)
3166 * Remove example explicitDestination (was a plain copy of basic)
3167 * Make JniException a runtime exception, so that there is no need to
3168 declare the fact that you may encounter such a beast. I guess that
3169 nobody will ever want to survive such error.
3170 * Create specific errors for each MSG case of failure:
3171 host failure, transfer failure, timeout, task cancelled
3172 * Cleanup the exceptions that may get thrown by each function
3173 * Other internal cleanups in Java bindings. Performance still bad :/
3174 Ruby and Lua Bindings: create them
3175 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
3176 * That's new and great, you should try them out.
3177 Same functionalities than Java bindings, only even less polished
3179 * Kill the useless "rate" argument of SD_task_get_execution_time()
3180 Everyone used to provide -1 as a value, it was not used, and the
3181 semantic of a possible use wasn't even clear.
3182 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
3183 either as comm costs or compute costs to mean that there is no
3184 such thing for that specific task.
3185 * Add a SD_task_set_name() function
3186 * Fix SD_task_unschedule() on typed tasks
3187 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
3188 * In DAX loader, accept useless 'level' attributes to <job> since
3189 LIGO DAGs have them (seem to be to ease graphical representation).
3191 * Add an example masterslave_mailbox.c using send/receive and not
3192 the deprecated put/get interface.
3193 * Kill the MSG_paje_output() function. It's a noop since 2 years.
3194 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
3196 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
3197 (declare MSG_USE_DEPRECATED to still have the old name)
3198 * Add a MSG_task_set_data() function
3199 * About trace replay (see examples/msg/actions):
3201 - Allow to work with splitted trace files for each process
3202 Give the specific trace file as argument of each process,
3203 and call MSG_action_trace_run(NULL)
3204 You can still have one merged file for all processes.
3205 - Fix implementation of collective operations
3206 * Allow task_execute() on 0-sized tasks (closes #10063)
3208 * This is the first release of SimGrid where SMPI is not considered
3209 beta anymore (even if some corners should still be improved)
3210 * Port over the new SIMIX_network submodule (internal refactoring)
3211 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
3212 * Implement more missing elements of the standard:
3214 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
3216 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
3217 MPI_Scan() MPI_Get_processor_name()
3218 - Added implementation of missing case for Alltoall (warning: it's
3219 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
3220 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
3221 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
3222 SMPI_MPI_Allgatherv()
3223 * Bug fixes include:
3224 - MPI_Waitsome() was broken
3225 - Allow relative includes in smpicc
3226 - Command line cfg argument 'reference_speed' was ignored...
3227 - Some functions did not properly lead to auto-benching of user code
3228 - smpicc passes -O2 by default (just like openmpi one)
3230 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
3231 * Bug fixes about timeouts during communications
3232 * add SIMIX_message_sizes_output() as a pimple to write to file the
3233 amount of messages per size. Use gnuplot to get histogram.
3234 Pimple because that's the only user-visible function of simix,
3235 defined directly in xbt.h (irk, sorry)
3237 - Add a SIMIX_sem_get_capacity() function
3238 - Fix interactions with processe resume/suspende
3239 - release_forever() was stupidly broken
3240 - Fix SIMIX_display_process_status() for processes in a semaphore
3241 - Make SIMIX_sem_block_onto() user-visible
3242 * Refactoring context stuff:
3243 - Use pseudo-OOP for better modularity
3244 - reimplement SIMIX_process_kill() without process_schedule() so
3245 that the latter can take as invariant that it is called from
3247 - Merge context_start into context_new for sake of simplicity
3249 * Add a Vivaldi network model, coded live during SUD'10 ;)
3250 * Rename configuration variables to start a hierarchy:
3251 o cpu_model -> cpu/model
3252 o network_model -> network/model
3253 o workstation_model -> workstation/model
3254 * New configuration variables:
3255 o network/bandwidth_factor: correction to bandwidth
3256 o network/latency_factor: correction to latency
3257 o netwotk/weight_S: correction to the weight of competing streams
3258 * Add a long description to the models, that users can see with such
3259 argument on the command line: --cfg=cpu/model:help
3260 * --help-models display the long description of all known models
3262 * config: add the ability to set a default value after registration
3263 Does not override any previously set value (e.g. from cmd line)
3264 * dict: allow to have integer key and data.
3265 When so, you need to use the following functions
3266 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
3267 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
3268 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
3269 In contrary to regular dicts, the key is not malloced before copy.
3270 Mixing scalar and regular elements in the same dict is not tested
3272 * Allow to use xbt_dynar_shrink() to expend the dynar instead
3273 Tracing for Visualization:
3274 * SimGrid is now instrumented in order to generate a trace file for
3275 visualization analysis: to use it, need to compile SimGrid with the
3276 "tracing" option enabled, and instrument the program using SimGrid with
3277 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
3278 (among other functions).
3279 * The instrumentation only traces the platform utilization for now
3280 * Documentation to use the tracing functions and how to analyze the
3281 traces with the Triva tool is written.
3282 * More information about: SimGrid FAQ (in the section Tracing Simulations
3285 * We moved to cmake as default build system. Autotools support will
3286 be dropped soon. Check the FAQ for more info about how to use it.
3287 * Greatly improved our cdash/ctest interactions
3288 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
3289 * Added memory checking tests with valgrind; lot of memleak fixing.
3290 This may be the first release of simgrid with so few memory issues
3291 * Added code coverage tests.
3292 Our coverage is still improvable, but at least we see it on cdash.
3294 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
3296 SimGrid (3.3.4) stable; urgency=low
3298 The "Desktop Grid needs love too" release (also called Xmas release).
3300 Models improvements:
3301 * Major speedup in the maxmin system solving by using lazy evaluation
3302 Instead of solving completely the maxmin system at each iteration,
3303 only invalidate (and recompute) the modified parts.
3304 This new feature is enabled in default models but you can try to
3305 turn it on with "--cfg:maxmin-selective-update=1" for other models.
3306 * Cas01 IMproved as default CPU model
3307 This CPU model is the same Cas01 model, but it uses the
3308 maxmin-selective-update flag and a heap structure to manage
3309 actions on SURF kernel.
3310 It reduces the complexity to find the next action to finish and,
3311 consequently, it's faster than the old Cas01.
3312 This is the new default CPU model (Cas01).
3313 * Rename the old Cas01 model to Cas01_fullupdate
3314 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
3315 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
3316 * CpuTI (CPU Trace Integration)
3317 A new CPU model whose objective is simulate faster when using
3318 availability trace files.
3319 Instead of using a full featured, over engineered maxmin system for
3320 CPU modeling, this model does the pre-integration of traces files
3321 to calculate the amount of CPU power available, and so, executes
3322 faster than the old CPU models.
3323 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
3324 * Use LV08 as default network model since it gives better accuracy
3325 for small messages and shouldn't change things for big ones.
3326 Use --cfg=network_model:CM02 to get the previous behavior.
3329 ******************************************
3330 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
3331 ******************************************
3332 * The new CPU model may changes simulations!
3333 The point is that events occurring at the exact same timestamp
3334 are not scheduled in the same order with the old and new
3335 version. This may be enough to completely change the execution
3336 of simulations in some cases.
3337 * The new network model will change simulations!
3338 This new model is more realistic than the previous one, so you
3339 should consider redoing your old experiments with this model.
3340 Sorry for the inconvenience.
3343 * Introduce the supernovae compilation mode
3344 When compiled that way, the whole SimGrid (or almost) is put in a
3345 single compilation unit and compiled in one shoot.
3346 This is to help gcc which has difficulties to inline stuff from one
3348 The speedup seem to be above 15%, althrough more tests are needed on
3349 amd64 to confirm that gain.
3352 * Port of MSG's mailbox on top of SIMIX network
3353 The put/get mechanism was greatly simplified on the way.
3356 * New SIMIX network module. Provides:
3357 - Mailbox: rendez-vous mechanism to find with who you want to speak
3358 - Synchronous send/recv: easier and hopefully faster since the
3359 logic is handled in the maestro process directly now
3360 - Asynchronous send/recv: you dreamt of it? It's here now
3361 Too bad that nobody cared enough to propagate the change to MSG.
3362 * Add semaphores as SIMIX synchronization mechanism.
3365 * new function SD_daxload(char*) to load a DAX file
3366 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3367 * Introduce typed tasks. Specify its kind and cost at creation.
3368 At scheduling, just give where it should be placed, and the cost
3369 for each involved resource is automatically computed.
3370 Existing constructors so far (more to come of course):
3371 - SD_task_create_comm_e2e() for end-to-end communication
3372 - SD_task_create_comp_seq() for sequential computation
3373 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3374 * new function SD_task_dump() for debugging display
3375 * new function SD_task_dotty(task,FILE*) writing to file the info
3376 about the task in dotty format
3377 * SD_task_dependency_exists() can now cope with having one of its
3378 arguments NULL. If so, it tests whether the other argument has any
3380 * Add getters on list of preceding/following tasks:
3381 SD_task_get_parents(task) and SD_task_get_children(task)
3382 * Add getters on amount of workstations and list:
3383 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3384 * Add getter on task kind: SD_task_get_kind(task)
3385 * Update the start_time and finish_time of tasks on completion/failure
3386 * Bugfix: Remove task from state swags when destroyed
3389 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3392 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3393 dynar with one separate thread per value of the dynar.
3394 * Change the prototype of xbt_thread_create(), sorry.
3395 Added a boolean parameter indicating whether we want to join this
3396 thread (used in SG only for now)
3397 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3400 * GTNetS wrappers should now be usable again (and betterly tested too)
3401 * Fix a major regression from 3.2 where the timeout provided to
3402 MSG_task_put_with_timeout() was used as absolute time before which
3403 the comm should be done.
3404 * Start to fix the <cluster> tag.
3405 - Internal links should be good now (beside of the loopback, which
3406 use the private link instead)
3407 - paths to the external world is still rather broken
3408 - the <route:multi> tag is just broken. Actually that's brain-dead.
3409 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3410 to make it less stupid
3411 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3412 * Fix a source-level compatibility glitch from 3.2: after defining
3413 MSG_USE_DEPRECATED, you can use the old name
3414 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3415 * Allow to compile from the SVN with automake 1.11
3416 * Fix some problems when using the "start_time" tag in deployment XMLs.
3417 * Fix #8569: XBT/synchro.h has redundant declarations
3418 * Fix #8563: MSG return values and exceptions
3419 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3420 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3421 Thanks to Robson Peixoto.
3422 * Fix a vicious bug in dictionaries inducing that some elements were
3423 not freed on xbt_dict_free()
3425 Portability report of this version:
3426 * Main portability targets:
3427 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3428 on (amd64/i386/ia64)
3429 - mac leopard on i386
3430 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3431 but nothing critical.
3432 * Other platforms: windows, AIX and others were not tested for this release
3434 Timing report of this version:
3435 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3436 scenario parameters). From 8h to a few seconds in desktop grid settings.
3437 * Supernovae brings about 25% speedup on i386.
3439 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3441 SimGrid (3.3.3) stable; urgency=low
3443 The "Need for Speed" release.
3445 The timings done to validate the 3.3.2 were faulty.
3446 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3448 The problem was a conversion from a manually handled vector to
3449 xbt_dynar_t on the critical path.
3450 xbt_dynar_foreach calls functions, inducing stack management crap.
3452 We inlined these functions and xbt_dynar_foreach is now breath taking.
3453 We also inlined xbt_swag_belong on the way.
3455 Here are some approximate speedup measurements (on master/slaves
3456 simulations lasting between 10s and 20s each):
3457 3.3.1 -> 3.3.2: about same performance
3458 3.3.2 -> 3.3.3: 40% speedup
3459 3.3.1 -> 3.3.3: 40% speedup
3460 3.3.1 with inline patch -> 3.3.3: 30% speedup
3462 Our reading is that the refactoring which occurred in 3.3.2 made us
3463 suffer much more from the xbt_dynar_foreach low performance, but
3464 once we solved this, this refactoring proved to be very performance
3465 effective. From the 40% speedup, somehow, 10% are due to the
3466 inlining and 30% to the refactoring.
3468 That's a pitty that gcc cannot inline functions placed in other files
3469 alone. We have to choose between:
3470 - break the encapsulation (by putting private data structures and
3471 accessors in headers files to help gcc)
3472 - live with low performance
3473 - switch to a decent compiler such as icc (not quite possible).
3475 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3477 SimGrid (3.3.2) stable; urgency=low
3479 The "Simplicity does not preceed complexity, but follows it" release.
3481 The main contributors of this release were (lexical order):
3482 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3485 * Extract the routing logic into its own object.
3486 (was dupplicated in network.c and workstation_LV07.c;
3487 Allows to implement other ways of storing that info)
3488 => kill now useless network_card concept
3489 - Use dynar to represent routes (instead of void** + int*)
3490 - kill link_set (use surf_network_model->resource_set instead)
3491 - Add a command-line option to choose the routing schema to use
3492 - Add three new models:
3493 * Floyd (shortest path computed at initialization)
3494 * Dijikstra (shortest path recomputed all the time)
3495 * Cached Dijikstra (shortest path computed on need)
3496 All these models where contributed by Silas De Munck, and are
3497 described in his ICCS09 paper.
3499 * Simplify model declaration
3500 (less redirections, less function to write when defining a model)
3501 - Factorize stuff between models:
3504 surf_model_resource_set(model)
3505 surf_model_resource_by_name(model, name)
3506 - Unify the types of models in s_surf_model_t (using an union)
3507 - Embeed fields of common_public directly into s_surf_model_t
3508 - Rename model methods:
3509 action_free ~> action_unref
3510 action_change_state ~> action_state_set
3511 action_get_state ~> action_state_get
3512 - Change model methods into functions:
3513 (model)->common_public->action_use ~> surf_action_ref
3515 * Implement a generic resource; use it as ancestor to specific ones
3516 (allows to kill duplicated code in models)
3517 Drawback: timer command don't need no name nor properties;
3518 workstation_CLM03 don't need no properties
3519 (but I guess we can live with those few bytes wasted)
3521 * Improve the action object model
3522 - implement a constructor avoiding dupplicated code about field
3523 initialization in generic_action part.
3525 * Kill the SDP model: it has an external dependency, is deprecated
3526 in flavor of modern lmm models, and didn't compile since a while
3529 * Relocation of the context module from XBT to SIMIX.
3530 (the context were decoupled from the simix processes, duplicating a lot of code)
3531 => a lot of code was factorized
3532 - less overhead is introduced during scheduling
3533 - simpler API for the context factory
3534 - the logic for process creation,destruction and manipulation was simplified
3535 * Simplification of the s_smx_process_t data structure.
3536 => accesing the simix level data associated to a process is faster now,
3537 and the code is a lot more readable.
3540 * Implement some more MPI primitives:
3541 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3542 -implementation: Bcast: flat or 2-ary tree (default),
3543 Barrier: 4-ary tree,
3545 Allreduce: Reduce then Bcast
3546 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3547 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3548 Alltoallv: flat tree, like ompi
3550 * Add support for optimized collectives (Bcast is now binomial by default)
3551 * Port smpirun and smpicc to OS X
3554 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3555 Shout out if you used it.
3558 * Display the list of still queued messages in SG mode when existing
3562 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3563 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3564 * Add xbt_str_from_file(FILE*)
3565 * Add xbt_dict_get_key achieving a linear reverse search
3566 * Remove the context module
3568 Portability report of this version:
3569 * Main portability targets:
3570 - Linux(debian)/x86/context
3571 - Linux(debian)/x86/pthreads
3572 - Linux(debian)/amd64/context
3573 - Linux(debian)/amd64/pthreads
3574 On these, we still have the eratic breakages of gras/pmm and
3575 amok/saturate_sg reported in previous version. We still think
3576 that the tests are the cause of the fault, not the tested code.
3578 - Mac OSX Leopard/x86/context
3579 Still false negative in tesh autotesting.
3580 Smpi still fails, but this time because readlink does not accept -f
3581 Everything seems to work properly beside of that.
3584 - AIX version 5.3 (only tested contexts this time)
3585 Smpi still fails there because mktemp is not installed.
3586 Everything seems to work properly beside of that.
3588 I managed to compile it for the first time, but several breakages.
3589 Won't delay the release for this exotic platform.
3591 * Windows: it's still lagging behind. If you want to help, please
3594 Timing report of this version:
3595 This version seem to be more than 5% faster than 3.3.1 (on linux
3596 64bits with contextes). The gain is less than expected, we are
3597 investigating this for next release.
3599 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3601 SimGrid (3.3.1) stable; urgency=low
3604 * Implement a --cfg-help to show existing configuration variables
3605 * Build chain do not require doxygen in maintainer mode
3608 * fix a bug on struct sizeof computation, which prevented the
3609 exchange of arrays of structs in some conditions
3610 - added a regression test about this in datadesc_usage
3611 * Allow the exchange of 0-long dynamic vectors.
3612 - for that, use -1 as indicator of dynamic size instead of 0
3613 - This implied to change any size from unsigned long to long,
3614 reducing a bit communication abilities, but I guess that with
3615 64bits being quite common, this is more than enough.
3616 - This also induce a protocol change, thus bumping network protocol
3617 version from 0 to 1 (if we have external users, we have to get
3618 clean on that point too ;)
3619 - added two regression tests about this in datadesc_usage
3620 * Be more verbose when propagating local exceptions
3621 This helps debugging.
3622 * Display the status of simulated processes when receiving SIGINT in
3626 * Allow to control the simulation from a trace file.
3627 New functions MSG_action_register() and MSG_action_trace_run()
3628 The first one allows to associate a function execution to each
3629 kind of action while the second one parses a trace file and
3630 triggers the corresponding actions within the system.
3631 For now, only a toy example is provided in examples/msg/actions
3632 * Add an example of process migration in examples/msg/migration
3633 * Fix a bug in task exchange which broke MSG_task_get_sender()
3634 Add a teshsuite regression test for that.
3635 [Bug: if MSG_task_get_sender() is called after sender exit,
3637 * Fix a bug which prevented suspend/resume to work properly
3638 * Display the status of simulated processes when receiving SIGINT
3639 This fixes a regression of v3.3. due to the introduction of SIMIX
3640 * Bug fixing in failure management:
3641 - trace could not start by a failure at time 0
3642 - failure during communications were not working
3645 * Add SIMIX_process_set_name() to change the name of the current
3646 process in the log messages.
3647 * Store smx_hosts in a dict since we only retrieve them by name
3648 * Move the configuration infrastructure to surf
3651 * Move the configuration infrastructure to surf
3654 * Massive internal cleanups:
3655 - Store internal structures on processes instead of hosts (allows
3656 to have more than one process per host, in addition of being more
3658 - Cleanup the initialization/finalization process
3659 - Kill a whole bunch of unneeded synchronization:
3660 processes run in exclusive manner within the simulator
3661 - Move queues from global tables to process data fields
3663 - now accept -platform and -hostfile arguments
3664 - Pass the right rank value to processes according to the hostfile
3665 * Compile the examples by default, and use them as regression tests
3666 * Implement MPI_Wtime()
3667 * Change the reference speed to a command line option
3670 * TCP_gamma can now be specified as command line option using
3671 --cfg=TCP_gamma:10000000.0
3672 * Change the --surf-path cmd line option into --cfg=path:
3675 * Also include strbuff from xbt.h public header
3676 * xbt_ex_display(): do not free the exception after displaying
3677 This allows to do more with the given exception afterward.
3678 Users should call xbt_ex_free() themselves.
3682 Portability report of this version:
3683 * Main portability targets:
3684 - Linux(debian)/x86/context
3685 - Linux(debian)/x86/pthreads
3686 - Linux(debian)/amd64/context
3687 - Linux(debian)/amd64/pthreads
3688 These targets fail about 1/10 of times on gras/pmm, but we believe
3689 that this is because of the test, not because of simgrid.
3690 amok/saturate_sg fails even more rarely, and the test may not be
3693 - Mac OSX Leopard/x86/context
3694 The test suite still spits tons of errors because some obscure
3695 force prevents us from removing the temporary directories
3696 arguing that they still contain some metadata I've never heard of.
3697 Smpi fails because seq is not installed.
3698 Everything seems to work properly beside of that.
3701 - AIX version 5.3 (both contexts and pthread)
3702 Smpi still fails there because mktemp is not installed.
3703 XML inclusions seems rosty on AIX.
3705 * Windows: it's still lagging behind. If you want to help, please
3708 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3710 SimGrid (3.3) stable; urgency=high
3714 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3715 [Malek Cherier & Mt]
3717 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3720 Until now, GRAS were using MSG as an interface to SURF. It was
3721 quite difficult because both interface have several differences
3722 (MSG channels vs GRAS sockets were the most notable point).
3724 This also opens the gate to SMPI (which should occur soon) and speed
3725 up simulations by to 40% (even if it were not the main goal).
3727 **************************************
3728 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3729 **************************************
3730 The point is that events occurring at the exact same timestamp are
3731 not scheduled in the same order with the old and new version. This
3732 may be enough to completely change the execution of simulations in
3733 some cases. Sorry for the inconvenience.
3735 * Cleanup and upgrade the XML format to push further scalability
3736 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3738 * Improve the testing infrastructure with tesh. Now a very large part of
3739 the code is tested not only by being run but also by checking that the
3740 output match an expected output [Mt].
3742 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3743 is really less memory-demanding, which should allow you to use
3744 larger files in SimGrid [AL].
3746 * Inform valgrind about our contextes, so that it becomes usable
3747 with the default (and more effecient) version of SimGrid
3748 [contributed by Sékou Diakite, many thanks]
3751 * Introduce a listener thread in charge of receiving incoming
3752 messages from the network. It allows to overlap communication and
3753 computation but most notably, it removes some stupid deadlocks due
3754 to the fact that so far, a process could not send and receive at
3755 the same time. This made most non trivial communication schema
3757 * Convert the PIDs from long int to int to match the MSG ones (and
3758 linux ones too) [Mt]
3759 * New function: gras_agent_spawn() to launch a new process on
3760 current host. Only working in simulation for now. [Mt]
3761 * New function: gras_os_hostport() returning a constant form (ie,
3762 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3765 * Make the backtrace of exceptions more human readable [Mt]
3766 * New module: xbt/str [Mt]
3767 a ton of string utility functions (split, join, printf to a newly
3768 allocated buffer, trim, etc)
3769 * New module: xbt/hash [Mt]
3770 SHA1 hashing algorithm (more to come if needed)
3771 * New module: xbt/synchro [Mt]
3772 synchronization tools (mutex and conditions) working the same way
3773 in simulation and in real life (mainly useful for GRAS, but not
3775 * New module: xbt/queue [Mt]
3776 classical producer/consumer synchronization scheme
3777 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3778 (using the classical functions will get serialized) [Mt]
3779 * Make dictionary internal table dynamic. No need to specify its size
3780 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3782 * Make sure the log channels are organized as a tree under windows
3783 (because of ANSI C compatibility issue, any channel were child of
3784 root directly) [Mt].
3787 * Cleaned many thing in surf and fixed a few bugs [AL].
3788 * Add a nice command line configuration mechanism to compose models [AL].
3789 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3790 the previous one (KCCFLN05). It relies on something that looks like
3791 a max-min sharing mechanism but cannot be written as such. A new solver
3792 was thus designed [AL].
3793 * Add a new solver to lmm. Based on Lagrange optimization and
3794 gradient-based descent, it enables to efficiently maximise systems s.a
3796 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3798 This solver enables to propose two new network models for TCP Reno and
3799 TCP Vegas based on Low's work. These models still need to be fully
3800 tested though [Pedro Velho].
3803 * Bug fix in SD_simulate. Now the time bound given as argument is
3805 * Use the new parallel task model (ptask_L07) as default.
3806 * Use the SURF command line configuration mechanism.
3807 * 0-size tasks (for synchronization) should now work.
3809 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3811 SimGrid (3.2) stable; urgency=high
3815 We still experience issues on this platform, but we believe that at
3816 least MSG is usable.
3818 GRAS API BREAKAGE (for simplification purpose, sorry):
3819 * the gras_msgtype_by_name is not used anymore. Instead of
3820 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3821 you can write (and must)
3822 gras_msg_send(toserver, "request", &request);
3823 - If you still want to pass a gras_msgtype_t to the function (to cache
3824 the type and avoid the lookup time), use the gras_msg_send_() variant.
3825 - Impacted functions:
3826 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3827 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3828 * The callbacks are now expected to return 0 when everything went well
3829 (just like the main() function)
3831 GRAS new features and improvements:
3832 * New module mechanism where user code can use per process globals [Mt]
3833 This is similar to gras_userdata_*() functions, but for libraries. It
3834 factorize some code developped over and over in the examples and AMOK.
3835 It has still to be documented and used (only amok/peermanagement is
3837 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3838 loss when they were small enough to fit into the buffer and sent quickly
3839 enough so that they can all get received in one shoot.
3840 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3841 if not found. Use the *_or_null() variant for the old semantic.
3842 * In gras_msg_handle, do not discard messages without callback.
3843 They are probably messages to be explicitly awaited later (ie, proofs of
3844 mis-synchronization in userland since they are sent before being awaited)
3845 No big deal usually.
3846 * gras_socket_meas_send/recv: semantic changed!
3847 The numerical arguments used to be (1) the total amount of data to send
3848 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3849 messages. This was need for the fool willing to send more than MAXINT
3850 bytes on quite fat pipes.
3853 * Do really rename the hostmanagement module to peermanagement. [Mt]
3854 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3855 but this is rather new and this was documented in the module
3856 documentation (poor excuses, I admit)
3857 * Bandwidth measurement semantic changed! This follows the changes to
3858 gras_socket_meas_send/recv explained above.
3861 * A sequential mode has been added to the workstations. When a workstation
3862 is in sequential mode, it can execute only one task, and the other tasks
3863 are waiting in a FIFO. [Christophe Thiery]
3866 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3867 model for SIMDAG. [Christophe Thiery]
3868 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3869 equal to 0, which caused some bad behaviors in
3870 saturated_constraint_set_update. I now use a threshold mechanism like in
3874 * When running manually src/testall, you select specific units [Mt]
3875 testall is the result of our cunit mechanism, and should replace all
3876 the scripty thingy around since bash don't run easily on billware.
3878 * A mallocator system has been added. [Christophe Thiery]
3879 Mallocators allow you to recycle your unused objects instead of freeing them
3880 and allocating new ones.
3882 Documentation update:
3883 * FAQ reworking + New FAQs:
3884 - "Valgrind spits tons of errors!" [Mt]
3885 - "How to repport bugs" [Mt]
3886 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3887 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3888 - Communication time measurement within MSG [AL]
3889 - I experience weird communication times when I change the latency [AL]
3890 * GRAS tutorial [Mt]
3892 - an introduction to the framework and to the used communication model
3893 - an initiatic tour introducing the most proheminent features:
3895 . Lesson 0: Installing GRAS
3896 . Lesson 1: Setting up your own project
3897 o Part 2: Message passing
3898 . Lesson 2: Exchanging simple messages
3899 . Lesson 3: Passing arguments to the processes (in SG)
3900 . Lesson 4: Attaching callbacks to messages
3901 . Lesson 5: Using globals in processes
3902 . Lesson 6: Logging information properly
3903 . Lesson 7: Using internal timers
3904 . Lesson 8: Handling errors through exceptions
3905 . Lesson 9: Exchanging simple data
3906 . Lesson 10: Remote Procedure Calling (RPC)
3907 . Lesson 11: Explicitly waiting for messages
3908 . Recapping of message passing features in GRAS
3909 - A HOWTO section containing:
3910 o HOWTO design a GRAS application
3911 More are due, of course. They will come latter. In the meanwhile, you can
3912 check the examples which are still here.
3914 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3916 SimGrid (3.1) stable; urgency=high
3920 There was a stack corruption somewhere, visible only when optimizing
3921 with these versions. [Vince]
3924 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3925 module that enabled to play with Directed Acyclic Graphs. It is built
3926 directly on top of SURF and provides an API rather close to the old
3927 SG. Some old codes using SG are currently under rewrite to check that
3928 all needful functions are provided. [Christophe Thiery]
3931 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3932 extension of the classical CLM03 model that gracefully handles
3933 failures. This is now the default model for MSG and GRAS. It doesn't
3934 handle parallel tasks yet though. [AL]
3935 * Bug fix: Weights were not correctly set in the network part.
3936 WARNING: This may have resulted in incorrect results with simulations
3937 where there are more than one flow on a given link. [AL]
3940 * After a (long ?) discussion on simgrid-devel, we have decided that the
3941 convention we had on units was stupid. That is why it has been decided
3942 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3943 WARNING: This means that all previous platform files will not work as
3944 such with this version! A warning is issued to ask users to update
3946 A conversion script can be found in the contrib module of the CVS, under
3947 the name contrib/platform_generation/surfxml_update.pl [MQ]
3950 * Bug fix: Processes were started in reverse order, wrt deployment file.
3951 WARNING: if your code relies on this bug, please fix it. [AL]
3952 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3953 being executed on two different locations. [AL]
3954 * Bug fix: Failures are now better supported thanks to Derrick's tests
3955 (there was many failure situations I hadn't thought of and that weren't
3956 correctly handled). [AL]
3957 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3961 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3962 Exception occurring on server-side are propagated back to client (!).
3964 API CHANGE: the callback changed their prototype. Change:
3965 int my_handler(gras_socket_t expeditor, void *payload_data) {
3967 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3968 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3970 * New! function: gras_msg_handleall to deal with all messages arriving
3971 within a given period.
3972 * New! function: gras_socket_server_range to get a server socket in a
3973 range of port numbers (ease to avoid port number conflicts) [MQ]
3974 * New! gras processes display their backtrace when they get a SIGUSR1
3975 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3976 Sweet to debug RL processes [MQ]
3980 - Do not force experiment sizes to be expressed in kb, or it becomes
3981 impossible to measure the latency this way (needs one byte-long tests)
3982 WARNING: this changes the amok_bw_* function semantic. [MQ]
3983 - Implements the link saturation stuff. [MQ]
3984 * Peer management module:
3985 New! module factorizing code that we wrote over and over [MQ].
3988 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3989 - Test units are placed directly into the library code, they get extracted
3990 automatically and placed into the src/testall binary.
3991 - Convert most of the XBT tests to this system.
3992 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3993 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3994 live in the same C code. This required to change a little bit the API
3995 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3996 * New module: sparse graph structure with basic algorithms (this is work
3997 in progress and the API is not considered to be frozen yet). [AL]
3998 * Display more information on backtraces: source line & function names are
3999 now displayed just like valgrind does (rely on addr2line tool) [MQ]
4000 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
4001 * Reworked a little bit some #include statements to load only required
4002 headers. Some user code that relied on SimGrid to include stdlib or
4003 stdio may need to include it by themselves. [AL]
4004 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
4006 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
4007 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
4008 but I guess that almost nobody used those parts. [MQ]
4010 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
4012 SimGrid (3.0.1) stable; urgency=low
4015 * Unfortunately, I had missed 5 misnamed functions:
4016 xbt_fifo_item_t xbt_fifo_newitem(void);
4017 void xbt_fifo_freeitem(xbt_fifo_item_t);
4018 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
4019 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
4020 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
4021 They're now deprecated. Please use their new versions:
4022 xbt_fifo_item_t xbt_fifo_new_item(void);
4023 void xbt_fifo_free_item(xbt_fifo_item_t);
4024 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
4025 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
4026 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
4028 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
4029 * Documentation: xbt_log module unmercifully reworked [MQ]
4030 * Bugfix: there was a problem with the ending of contexts with
4031 the pthread backend. It caused some weird deadlock or behavior
4032 depending on the pthread implementation. [AL]
4033 * Bugfix: get the exceptions raised in the simulator repport where
4034 and why they come from when they are not caught in time [AL, MQ]
4037 * Bugfix: Do repport the error when two non-connected hosts try to
4038 exchange data (Thanks to Flavien for stumbling into this one) [AL]
4041 * Add additionnal checkings on communications. Assert that two
4042 communicating hosts are connected by a set of links... [AL]
4045 * Add additionnal checkings on channel values in communication [AL]
4046 * New: MSG_task_get_source to see on which host a task was generated [HC]
4047 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
4048 the number of tasks waiting to be received on channel and sent
4050 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
4051 waits for the first task coming from a given host.. [AL]
4053 GRAS new functionnalities: [MQ]
4054 * Enhance the parsing macro to allow the size of multidimentional objects
4055 to be given thru annotations.
4056 * New example (and documentation): Matrix Multiplication a la RPC
4057 (as when I was young!) and fix a bunch of bugs found on the way.
4059 GRAS performance improvements: [MQ]
4061 * Reduce the amount of cbps creation/destruction by making it static to
4062 datadesc_send/recv() and using a (newly created) cbps_reset (based on
4065 * Change libdata to a set so that we can search for stuff by ID (and thus
4066 reduce the insane amount of dict lookups)
4069 * Actually implement gras_datadesc_copy() so that we don't have to mimick
4070 RL communication on top of SG since it's so uneffective.
4071 It may also be used for inter-thread communication in RL, one day.
4072 * Use gras_datadesc_copy() to exchange messages on top of SG
4074 - improve message exchange performance on top of SG
4075 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
4076 * Don't exchange on the network the size of the used part of buffer,
4077 instead, specify the possible buffer size to read().
4079 - reduces the amount of read/write calls (one pair per exchange)
4080 - reduces the amount of exchanged data (the size)
4081 - allows to retrieve all arrived data on receiver side, if we don't need
4082 it right now (subsequent read will peek the buffer)
4083 - allows the receiver to proceed with the beginning of the stream before
4084 everything is arrived
4085 - make it possible to build an iov transport (using readv/writev)
4087 - take care of the data with non-stable storage (like stacked data),
4089 * If possible, TCP send uses vector I/O (when writev() is here)
4090 - Don't use it for receive since we send data sizes and data on the
4091 same stream, so we wouldn't be able to chain large amount of chunks
4092 before having to flush the stuff to read the size.
4093 * Rework the transport plugin mechanism to simplify it and reduce the
4094 amount of pointer dereferencement when searching for the right function
4097 * I guess that now, we do almost as few system calls as possible while
4098 doing as few data copy as possible.
4100 To improve it further, we could try to send all the sizes first and then
4101 all the data (to use iov on receiving size), but it's only a partial
4102 solution: when you have 2 dimensional data, the sizes of the second
4103 dimension is data of the first dimension, so you need 3 streams.
4105 I'm not sure the potential performance gains justify the coding burden.
4107 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
4109 SimGrid (3.00) stable; urgency=high
4112 * New! Give the possibility to hijack the surf parser and thus bypass
4113 MSG_create_environment and MSG_launch_application. Have a look at
4114 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
4116 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
4118 SimGrid (2.96) unstable; urgency=low
4123 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
4124 This deprecates the xbt_error_t mechanisms.
4125 It modifies (simplifies) all XBT and GRAS API.
4126 MSG API keeps unchanged (exceptions raised by XBT are caught from
4127 within MSG and masked with existing error handling facilities)
4130 * New! Add a FATPIPE model. [AL]
4131 * New! Add a parallel task model. [AL]
4132 * New! Add automatically a loopback interface (in the default
4133 network model) if none was precised.
4136 * Bugfix: MSG_process_resume now works with the current running process.
4138 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
4139 * Modification of MSG_task_get_compute_duration. Once a task has been
4140 processed, the value returned by this function is now equal to 0. [AL]
4141 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
4142 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
4143 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
4144 of a m_task. For now, MSG_task_cancel only works with computation
4146 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
4147 of the processor (in Mflop/s) regardless of the current load on the
4149 * API Change: use proper naming convention for MSG_getClock and
4150 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
4152 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
4153 This function changes the priority of a computation task. This priority
4154 doesn't affect the transfer rate. A priority of 2 will make a task
4155 receive two times more cpu power than the other ones. This function
4156 has been added to suit the needs of Nguyen The Loc and hasn't been that
4157 much tested yet. So if it fails, please report it and send me your code.
4159 * API Change: removed all functions and types that were marked "deprecated"
4160 since many months. Renamed MSG_global_init_args to MSG_global_init.
4162 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
4164 SimGrid (2.95) unstable; urgency=low
4167 * Steal some nice code to GNU pth to fix context detection and usage [AL]
4168 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
4169 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
4172 * Bugfix: Allow absolute paths for platform description files [MQ]
4173 * Bugfix: do free the variables after use. Leads to drastic performance
4175 * Implement max_duration (ie, timeouts) on resources [AL]
4178 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
4180 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
4181 the needed support (timeouts) [AL]
4184 * Implement measurement sockets. You can now get the bandwidth between two
4185 hosts thanks to AMOK (see below). [MQ]
4186 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
4187 dynar over the network (yeah) [MQ]
4188 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
4189 * simple/static token-ring example. [Alexandre Colucci and MQ]
4190 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
4191 in place before [MQ]
4194 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
4195 type parsing. "short" and "long" modifiers were also ignored; other
4196 modifier (such as reference level) are still ignored. [MQ]
4197 * Embeed the buffer size within the buffer itself on SG. [MQ]
4198 That way, send() are atomic and cannot get intermixed anymore (at least
4199 the ones which are less than 100k; bigger messages still have the issue)
4200 * Array size pushed by the field, not by the field type (or each
4201 and every long int will push stuff to the cbps) [MQ]
4202 * use select() to sleep since it allows to portably sleep less than one
4205 GRAS (minor cleanups)
4206 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
4209 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
4210 * type category 'ignored' killed as it was never used and were difficult
4212 * whether a type can cycle or not is now a flag, leaving room for more
4213 flags (as "ignored", if we feel the need one day ;)
4214 * Rename raw sockets to measurement sockets since "raw" has another
4215 meaning in networking community.
4218 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
4219 and offers features not belonging to GRAS but that most applications
4220 need. One day, it may be a set of plugins loadable at runtime.
4221 * New module: bandwidth
4222 bandwidth measurement between arbitrary nodes running this module. [MQ]
4224 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
4226 SimGrid (2.94) unstable; urgency=low
4228 The first beta release of SimGrid 3 !
4232 * Update the main page and the FAQ. Adding references to gforge.
4235 * Add a gras_os_getpid function.
4238 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
4239 * Extend the logs so that they also print PID, hostname, date, ... if
4241 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
4242 and kill the old version which were in testsuite/
4243 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
4247 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
4248 for users. Instead, each programming environment (GRAS, MSG,...) use it
4249 when needed to provide such a feature to users.
4250 Don't shortcut the mechanism or you will also shortcut the virtualization
4251 you need on the simulator.
4255 * Cleanups in configury with regard to compile optimization/warning flags.
4256 Also add -fno-loop-optimize to any powerpc since it's the optimization
4257 killing gcc (< 3.4.0).
4258 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
4259 needed by MSG examples complications
4260 * Borrow configury beautifications from PHP
4263 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
4264 warning (thanks loris for stumbling into this one).
4265 * Bugfix: stop loading private headers (gras_config.h) from the public
4269 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
4270 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
4271 RL would imply a DNS resolver, which is *hard* to do in a portable way
4272 (and therefore delayed).
4273 * Implement a real timer mechanism and use it in timing macros. This allows
4274 to avoid rounding errors and get a 0.000005 sec precision in timing
4275 macros. While I was at it, various cleanups:
4276 - allow to declare more than one timed section per file (fix a stupid bug)
4277 - move some private declaration to the right place
4278 - merge conditional execution and timing macros into emulation module
4279 - document the module
4280 - make sure the module cleanups its mess on gras_exit
4281 * Documentation improvements:
4282 - (new) how to compile applications using GRAS
4283 - (new) emulation support (timing macros)
4285 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
4287 SimGrid (2.93) unstable; urgency=low
4289 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
4292 - Use Paje properly where used. Still to be sanitized properly.
4293 - Portability fix: Add an implementation of the contexts using pthread
4297 - Add xbt_procname(): returns the name of the current process.
4298 Use it to show the current process's name in all logging.
4300 - fix detection of older flex version and the reaction, since we do
4301 depend on modern ones (we use lex_destroy)
4302 - Better separation of SG and RL in the libs: remove all simulation code
4303 from libgras. As a result, this lib is now only 200k when stripped.
4304 Some of the xbt modules may also be duplicated (two sets and such) and
4305 should be cleaned/killed before SG3.
4306 - Insist on using xlC on AIX because of weird problems involving gcc there.
4307 - Cleanup the make remote stuff. This is now done by scripts
4308 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
4309 mainly for our private use, but we're working on changing them to user
4312 - Bugfix: flush the socket on close only if there is some *output*.
4313 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
4315 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
4318 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
4320 SimGrid (2.92) unstable; urgency=low
4322 Alpha 3 on the path to SimGrid 3
4326 - New! First try of benchmarking macros.
4327 - New! First try so that gras_stub_generator generate deployment and
4328 remote compilation helpers.
4330 - Bugfix: Initialization fix in msg_test.
4334 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
4336 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
4338 - API Change: gras_os_sleep to take the amount of seconds as a double.
4339 Accepting an int was error prone since it was the only location where
4340 seconds were coded as such. It leaded to damn rounding errors.
4341 - Bugfix: Hard to belive that timers ever worked before this.
4343 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4345 SimGrid (2.91) unstable; urgency=low
4347 Alpha 2 on the path to SimGrid 3
4351 - Bug fix in the lmm_solver.
4353 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4354 through the function MSG_paje_output.
4355 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4356 - It is possible to bound the rate of a communication in MSG with
4357 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4358 to write it in the changelog).
4359 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4364 - Various cleanups to the autotools stuff
4365 - Begin to move Gras examples to examples/gras/
4366 - Let make distcheck work again (yeah!)
4368 - documentation overhauled using doxygen.
4369 gtk-doc-tools is dead in SimGrid now.
4370 - Automatically extract all existing logging categories, and add the list
4371 to the documentation (long standing one, to say the less)
4373 - Cleanup the known architecture table. Reorder the entries to group what
4374 should be, and use a more consistent naming scheme.
4375 (some of the test dataset are still to be regenerated)
4376 - New! Allow library to register globals on each process just as userdata
4378 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4379 lookup time (for now).
4380 Use it in msg and trp.
4381 This cleans a lot the internals and helps enforcing privacy of the
4382 headers between the gras components.
4383 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4384 - Bugfix: gras_os_time was delirious in RL.
4385 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4387 - Reenable GRAS now that it works.
4389 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4391 SimGrid (2.90) unstable; urgency=low
4393 Alpha 1 on the path to SimGrid 3
4395 * It is a long time since the last release of SimGrid. I'm sorry about
4396 that but as I had told you, I was rewriting a lot of things. I apologize
4397 to those who had been reporting bugs to me and that I had not answered.
4398 If your bug is still in the new version, please tell me. Here is a
4399 summary of the main changes.
4401 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4402 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4403 portability, speed, and a lot more but you'll figure it out later.
4404 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4405 I wanted to release the new MSG as soon as possible and I have
4406 broken GRAS, which is the reason why, for now, only MSG is fully
4407 functional. A laconic description of these projects is available
4408 in the documentation.
4410 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4411 kernel optimized for our needs (called SURF but only the developers
4412 should use it). Hence, MSG is now roughly 30 times faster and I think
4413 that by rewriting a little bit MSG, I could event speed it up a little
4414 bit more. Beside the gain in speed, it is also much easier to encode a
4415 new platform model with SURF than it was with SG. More to come...
4417 * REVOLUTION 3: I have tried to change a little as possible the API of
4418 MSG but a few things really had to disappear. The main differences
4419 with the previous version are:
4420 1) no more m_links_t and the corresponding functions. Platforms are
4421 directly read from an XML description and cannot be hard-coded
4422 anymore. The same format is used for application deployment
4423 description. The new format is described in the documentation.
4424 Have a look in tools/platform_generation. There is a tiny script
4425 that converts from the old platform format to the new one. Concerning
4426 the application deployment format, parsing the old one is tricky.
4427 I think most of you should however be able to convert your files. If
4428 it is really an issue, I can write a C code that does the conversion.
4430 2) the toolbox tbx does not exist anymore. We now have a library
4431 with much more data-structures but without the hash-tables (we have
4432 dictionaries that are much faster).
4434 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4436 *****************************************************************************
4437 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4438 * don't feel like converting the oldies. (Mt) *
4439 *****************************************************************************
4442 Version 2.90: "the long awaited one"
4443 - Finished rewriting and debugging MSG. Rewrote the documentation.
4444 - disable GRAS for now since it needs to be ported to the newest SG
4447 - Finish the port to windows (using mingw32 for cross-compile)
4450 - Main loop and datastructures of SURF. A cpu resource object is
4451 functional. Surf can thus be used to create cpu's with variable
4452 performance on which you can execute some actions.
4454 2004-11-15 Martin Quinson
4455 - Port to ARM. Simply added the alignment and size descriptions. Should
4456 work, but the ARM machines are so slow that I didn't had the opportunity
4457 to 'make check' over there yet.
4459 2004-11-15 Arnaud Legrand
4460 - Trace manager now written. It uses a heap structure and is therefore
4461 expected to be efficient. It may however be speeded up (particularly
4462 when many events occur at the same date) by using red and black
4463 trees. One day maybe...
4464 - Max-min linear system solver written. It uses a sparse matrix
4465 structure taking advantage of its expected use. Most operations are
4466 O(1) and free/calloc are called as few as possible. The computation of
4467 the minimum could however be improved by using a red and black tree
4470 2004-11-03 Arnaud Legrand
4471 - Rename every gras_* function that was in xbt/ to its xbt_
4473 - Add a heap and a doubly-linked list to xbt
4474 - Added a dichotomy to the dictionaries. make check works as well before
4475 so I assume that the patch is correct. I do not know however if things
4476 run effectively faster than before now. :)
4478 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4479 SimGrid, and the version number is bumped to 2.x
4481 2004-10-29 Martin Quinson
4482 - Introduction of the remote errors.
4483 They are the result of a RMI/RPC on the remote machine.
4484 ErrCodes being scalar values, you can't get the host on which those
4485 errors did happen. Extending the error mechanism as in Gnome is possible.
4486 No idea yet whether it is a good idea.
4488 2004-10-28 Martin Quinson
4489 - Interface revolution: the Starred Structure Eradication.
4490 I used to do typedef struct {} toto_t; and then handle *toto_t.
4491 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4492 - ???_t is a valid type (builded with typedef)
4493 - s_toto_t is a structure (access to fields with .)
4494 - s_toto is a structure needing 'struct' keyword to be used
4495 - e_toto_t is an enum
4496 - toto_t is an 'object' (struct*)
4498 typedef struct s_toto {} s_toto_t, *toto_t;
4499 typedef enum {} e_toto_t;
4500 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4501 s_toto_t) is private.
4503 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4504 so that it can set the variable to NULL. It was so for dicts and sets,
4505 it changed for dynars.
4507 - Fix a bunch of memleaks in dict_remove
4508 - Fix a bug in sg/server_socket opening: it failed all the time.
4510 2004-10-07 Martin Quinson
4511 - Speed up dynar lookup operation a bit.
4513 gras_dynar_get is dead.
4515 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4516 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4517 (which gives you the address of the stored data).
4519 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4520 copy of the data using an affectation to do the job and not a memcpy.
4522 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4523 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4525 It does not really speedup the dynar test because they are
4526 setting elements all the time (and look them seldom). But the dict does
4527 far more lookup than setting.
4529 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4531 2004-10-05 Martin Quinson
4532 - Allow to (en/dis)able the cycle detection at run time.
4534 Whether we should check for cycle or not is now a property of each
4535 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4536 datadesc_cycle_unset allow to remove this property when previously set.
4538 Note that the cycle detection is off by default since it impacts the
4539 performance. Watch the data you feed GRAS with ;)
4541 This property is hereditary. Any element embedded in a structure having it
4542 set have it set for the time of this data exchange.
4544 You should set it both on sender and receiver side. If you don't set it on
4545 sender side, it will enter an endless loop. If you forget on receiver
4546 side, the cycles won't be recreated after communication.
4548 - Header reorganization.
4549 Kill gras_private.h, each submodule must load the headers it needs.
4551 2004-10-04 Martin Quinson
4552 - Interface revolution: do not try to survive to malloc failure.
4554 Now, gras_malloc and friends call gras_abort() on failure.
4555 As a conclusion, malloc_error is not a valid error anymore, and all
4556 functions for which it was the only gras_error_t return value are
4557 changed. They now return void, or there result directly.
4558 This simplify the API a lot.
4560 2004-09-29 Martin Quinson
4561 - Re-enable raw sockets.
4562 Created by gras_socket_{client,server}_ext;
4563 Used with gras_raw_{send,recv}
4566 It should allow to kill the last bits of gras first version soon.
4568 This is not completely satisfactory yet (duplicate code with
4569 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4572 - Simplify transport plugin (internal) interface by not passing any
4573 argument to _server and _client, but embedding them in the socket
4576 2004-09-28 Martin Quinson
4577 - Finish the port to AIX.
4578 autoconf was my problem (segfault within the malloc replacement
4579 function. No idea why)
4581 2004-09-16 Martin Quinson
4582 - Fix some size_t madness on 64bit architectures.
4584 2004-09-08 Martin Quinson
4585 - Reduce the number of system headers loaded, overload some more system
4586 calls (such as malloc to cast the result of the system one, and work
4588 - Fix and reintroduce the config support
4590 2004-09-07 Martin Quinson
4591 - Source code reorganization to allow Arnaud to surf all over there.
4592 - Allow to document the logging categories.
4593 - Remove all uppercase from logging categories and useless cleanup in names.
4595 2004-08-18 Martin Quinson
4596 Version 0.6.2 (protocol not changed; API changed)
4597 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4598 gras_error_t), and NULL when not found. Functions expecting a msgtype
4599 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4600 hopefully usefull message.
4601 - Portability to prehistoric sparcs again
4603 2004-08-17 Martin Quinson
4604 Version 0.6.1 (protocol not changed; ABI not changed)
4605 - prealloc some buffers to speed things up
4607 2004-08-11 Martin Quinson
4608 Version 0.6 (protocol not changed; ABI expended)
4609 - The parsing macro can deal with the references, provided that you add
4610 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4612 2004-08-09 Martin Quinson
4613 Version 0.5 (protocol not changed; ABI changed)
4614 - Allow to off turn the cycle detection code in data exchange at
4615 compilation time. It should be at run time, but I'm short of time (and
4616 the config stuff is still broken). That way, we keep dict out of the
4617 critical path, which is good because the performance is poor:
4618 - search not dichotomial yet
4619 - dynar give no way to access their content and memcpy everytime
4620 - In composed data description (struct, ref and so on), stop foolness of
4621 keeping the subtype's ID, but store the type itself. This keeps sets out
4622 of the critical path, which is good since they rely on dynar and
4623 dictionaries. The only loose of that is that we cannot detect the
4624 redeclaration of a structure/union with another content (but I'm not sure
4625 the code detected well this error before anyway). We still can detect
4626 the redefinition discrepancy for the other types.
4627 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4628 it breaks the code because of type-punning used all over the place).
4629 This breaks on all non-gcc architectures (for now).
4631 All those changes (plus the buffer of last time) allow me to gain 2 order
4632 of magnitude on cruel tests consisting of 800000 array of integers on two
4633 level of a hierarchical structure (200 secondes -> 4 secondes)
4636 - the selector of reference must now return the type it points to, not
4637 the ID of this type.
4639 2004-08-06 Martin Quinson
4640 Version 0.4 (protocol changed; ABI not changed)
4641 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4642 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4643 - Implement a buffer transport, and use it by default (it relies on tcp in
4644 real life and on sg in simulation).
4645 That's a bit hackish since I had a new field to the structure to store
4646 its data without interfering with the subtype ones. Inheritance
4647 is tricky in C. And that's a kind of reverse inheritance with one class
4648 derivating two classes. Or maybe a game with java interfaces. Anyway,
4649 that's damn hard in C (at least).
4650 Moreover, I got tired while trying to ensure plugin separation and
4651 genericity in SG mode. MSG wants me to do weird things, so let's go for
4652 cruel hacks (temporarily of course ;).
4653 See comment in transport_private.h:71
4654 - do not include all the _interface headers in private but in the files
4655 which really need them (to cut the compilation time when they are
4658 2004-07-26 Martin Quinson
4659 Version 0.3 (protocol not changed; ABI changed)
4660 - Major overhault of the datadesc interface to simplify it:
4661 - shorted the function names:
4662 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4663 - add a trivial way to push/pop integers into the cbps without malloc.
4664 This allows to make really generic sub_type description, which simply
4665 pop their size of the stack.
4666 - add a function gras_datadesc_ref_pop_arr() which does what users want
4667 most of the time: Declare a dynamic array (which pops its size of the
4668 stack) and declare a reference to it. Poor name, but anyway.
4669 - kill the post-send callback, add a post-receive one
4671 2004-07-23 Martin Quinson
4672 Version 0.2 (protocol changed; ABI changed)
4673 - add some testing for cpbs in the test cases, and fix some more bugs.
4674 This invalidate again the little64 data file, since I cannot regenerate
4676 - remove an awfull optimization in the logging stuff, allowing me to:
4677 - understand it again
4678 - learn gcc how to check that the argument match the provided format
4679 - fix all errors revealed by gcc after that
4680 - internal keys of dict are not \0 terminated. Deal with it properly in
4681 loggings instead of segfaulting when the user want to see the logs :-/
4683 2004-07-22 Martin Quinson
4684 - Fix some stupid bug preventing cbps (callback postit) from working
4686 2004-07-21 Martin Quinson
4687 - Some documentation cleanups
4688 - remove the useless last argument of msgtype_declare
4689 - rename the Virtu functions to fit into the 'os' namespace
4690 - move headers src/include -> src/include/gras/ and stop fooling with
4691 gras -> . symbolic link
4692 - make distcheck is now successful
4694 2004-07-19 Martin Quinson
4696 - Build shared library also
4697 - Install html doc to the right location
4698 - stop removing maintainer files in make clean
4699 - build tests only on make check
4701 2004-07-13 Martin Quinson
4703 - No major issue in previous version => change versioning schema
4704 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4705 corresponding dataset.
4707 2004-07-11 Martin Quinson
4709 - Get it working with any kind of structure (we can compute the padding
4710 bytes remotely for all the architectures I have access to)
4711 - Implement the structure parsing macro (still not quite robust/complete)
4712 - Improvement to the remote testing toysuite
4714 2004-07-10 Martin Quinson
4715 [autoconf mechanism]
4716 - get ride of a bunch of deprecated macros
4717 - actually run the test for two-compliment, not only compile it :-/
4718 - test whether the structures get packed (and bail out if yes. Damn.
4719 Alignment is a serious matter)
4720 - test whether the structures get compacted (but respecting the alignment
4721 constraints of each types)
4722 - test whether the array fields of structures can straddle alignment boundaries
4724 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4725 (took me 2 hours to find that bug, looking at the wrong place)
4726 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4727 before sending/receiving and are used to compute the offsets of fields
4728 - Given that padding size depend even on compiler options, keep track of
4729 alignment and aligned_size only for the current architecture. Not a big
4730 deal since we send structure fields one after the other (seems
4732 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4733 program, let it work on linux/gcc/little32. portability todo.
4735 2004-07-08 Martin Quinson
4736 - import and improve remote compilation support from FAST
4737 - make sure make check works on half a dozen of machines out there
4739 2004-07-07 Martin Quinson
4740 Let's say it's version 0.0.3 ;)
4741 - Implement conversions (yuhu!)
4742 - Let it work on solaris (beside conversion, of course)
4743 - Stupid me, using rand() to generate the conversion datatests in not wise.
4745 2004-07-06 Martin Quinson
4746 - Let make dist work, since I'm gonna need it to compile on remote hosts
4747 - Let Tests/datadesc_usage write the architecture on which the file was
4748 generated as first byte.
4749 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4750 architecture descriptions.
4751 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4752 on those architectures.
4753 - Optimization: send/recv array of scalar in one shoot
4755 2004-07-05 Martin Quinson
4756 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4758 - Plug a whole bunch of memleaks
4759 - each process now have to call gras_{init,exit}. One day, their log
4760 settings will be separated
4761 - Continue the code factorisation between SG, RL and common in Transport.
4763 2004-07-04 Martin Quinson
4765 - Redistribution between SG and RL.
4766 We wanna have to accept in SG, so move accepted related parts of RL in
4767 the common part. (more precisely, the dynar of all known sockets is no
4768 more a static in transport.c, but part of the process_data)
4770 [gras_stub_generator]
4771 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4773 2004-07-03 Martin Quinson
4774 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4775 them all up in one shot)
4777 - Fix the ugly case of reference to dynamic array.
4778 - New (semi-public) function gras_datadesc_size to allow the messaging
4779 layer to malloc the needed space for the buffer.
4781 - gras_socket_close now expect the socket to close (and not its address to
4782 put NULL in it after it). This is because the socket passed to handlers
4783 is one of their argument (=> not writable).
4785 - propagate the interface cleanup from last week in datadesc, ie remove a
4786 superfluous level of indirection. User pass adress of variable
4787 containing data (both when sending and receiving), and not of a variable
4788 being a pointer to the data. Let's say that I like it better ;)
4789 The price for that is constructs like "int msg=*(int*)payload" in
4790 handlers, but it's a fine price, IMHO.
4792 - Let it work in RL (yuhu)
4794 2004-06-21 Martin Quinson
4796 - porting SG plugin and SG select to new standards (works almost).
4797 - plug memleaks and fix bugs around.
4800 - cleanup the prototype of data recv and force users to specify when they
4801 want to handle references to objects. Test case working even for cycles.
4802 - plug memleaks. Valgrind is perfectly ok with this.
4804 2004-06-12 Martin Quinson
4806 - cleanup the separation between plugin and main code in plugin creation
4808 2004-06-11 Martin Quinson
4810 - Reput hook for raw sockets, needed for BW experiments
4811 - kill a few lines of dead code
4812 [Data description] Interface cleanup
4813 - gras_datadesc_by_name returns the searched type or NULL.
4814 That way, no variable is needed to use a type desc once, which makes
4816 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4817 two parameters were strings (field name, type name), leading to
4819 [Dicos] Interface cleanup
4820 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4821 This is consistant with the dynar API.
4823 2004-04-21 Martin Quinson
4825 - Porting to new standards.
4827 - interface cleanup.
4828 There is no bag anymore, no need to take extra provision to mask the
4829 pointers behind "ID".
4830 Better splitup of functions between files create/exchange/convert.
4831 This is still a bit artificial since convert and receive are so
4832 interleaved, but anyway.
4834 - add a queued message list to procdata (the ones not matching criteria
4836 - factorize some more code between SG and RL wrt procdata
4838 - use gras_exit in example to track memleaks
4839 - get rid of gs_example now that GS is properly integrated into gras
4840 - update run_test to integrate the latest tests (datadesc)
4842 - rename WARNINGn macros to WARNn since it prooved error-prone
4844 2004-04-19 Martin Quinson
4846 - register init/exit functions within gras module mechanism
4847 - send/receive function.
4848 Convertion is not implemented, but short-cutted if not needed.
4849 struct/array elements are sent one by one (instead of block-wise), but
4850 nobody really cares (yet). Get a prototype before optimizing.
4851 - tests (using a file socket) for DD send/receive on:
4852 - base types: int, float
4853 - array: fixed size, string (ie ref to dynamic string)
4854 - structure: homogeneous, heterogeneous
4855 - chained list, graph with cycle
4856 Believe it or not, valgrind is not too unhappy with the results. The
4857 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4861 - Bugfix when using a filename explicitly (instead of '-')
4863 2004-04-09 Martin Quinson
4865 - factorize more code between RL and SG in socket creation
4866 - Complete the implementation and tests of:
4868 o file (only in RL, and mainly for debugging)
4870 I lost 3 days to design a portable address resolver, and then decided
4871 that the prototype mainly have to run on my box.
4872 Addressing portability too early may be like optimizing too early :-/
4874 - use gras_init in the Tests instead of the crappy parse_log_opt
4875 (the latter function is removed)
4876 [Conditional execution]
4877 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4878 [Code reorganisation]
4879 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4880 Build examples against the RL library, since there is no way to disable
4881 its creation for now.
4883 For information, the beginning of coding on GRAS was back in june
4884 2003. I guess that every line has been rewritten at least twice since