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.
8 - Support for 32bits architecture is not tested anymore on our CI infrastructure.
9 It may break in the future, but we think that nobody's using SimGrid on 32 bits.
10 - Remove the surf module. It was replaced by the kernel/models module, and that
11 refactoring took almost 10 years to properly complete.
14 - Activity::set_remaining() is not public anymore. Use for example
15 Comm::set_payload_size() to change the size of the simulated data.
16 - New function: Engine::flatify_platform(), to get a fully detailed vision of the
18 - Full simDAG integration: Activity::start() actually starts only when all dependencies
19 are fullfiled. If it cannot be started right away, it will start as soon as it becomes
21 - Allow to set a concurrency limit on disks and hosts, as it was already the case for links.
22 - Rename Link::get_usage() to Link::get_load() for consistency with Host::
25 - Operation: They are designed to represent workflows, i.e, graphs of repeatable Activities.
26 See the examples under examples/cpp/operation-* and the documentation in the Plugins page.
27 - Battery: Enable the management of batteries on hosts.
28 See the examples under examples/cpp/battery-* and the documentation in the Plugins page.
31 - optimize an internal datastructure (use a set instead of a list for ongoing activities),
32 leading to a potentially big performance gain, in particular with many detached comms.
35 - New option smpi/barrier-collectives to add a barrier to some collectives
36 to detect dangerous code that /may/ work on some MPI implems.
37 - New function SMPI_app_instance_start() to easily start a MPI instance in your S4U simulation.
40 - Write the section of the manual about models, at least.
41 - WiFi: the total capacity of a link depends on the amout of flows on that link.
42 - Use the nonlinear callback feature of LMM to reflect this.
43 - Calibration values can be changed to match different MCS configurations
44 - See the example teshsuite/models/wifi_usage_decay/wifi_usage_decay.cpp
45 - See also "A Flow-Level Wi-Fi Model for Large Scale Network Simulation"
46 https://hal.archives-ouvertes.fr/hal-03777726
47 - Merge parameters network/bandwidth-factor and smpi/bw-factor that serve the same purpose.
48 - Same for the latency
49 - Rewrite the corresponding documentation.
50 - Allow to disable the TCP windowing modeling by setting network/TCP-gamma to 0.
51 - Finally kill the 'compound' host model. You can change the CPU or network model
52 with the default host model, as it should.
53 - Rename option "surf/precision" to "precision/timing" for clarity.
54 - Rename option "maxmin/precision" to "precision/work-amount" for clarity.
55 - New function: Engine::flatify_platform() to debug your platform.
58 - Implement pthread_join in MC mode.
59 - Implement semaphore functions in sthread.
60 - Add an intricated way to verify the access to non-reentrant data structures
61 It requires code annotation, as shown in examples/sthread/stdobject/stdobject.cpp
64 - Synchronize the MBI tests with upstream.
65 - Show the full actor bactraces when replaying a MC trace (with model-check/replay)
66 and the status of all actors on deadlocks in MC mode.
67 - The safety/stateless aspects of the MC are now enabled by default in all simgrid builds.
68 Liveness and stateful aspects are still controled by the enabling_model-checking
70 - Stateless model-checking is now usable on any system, including Mac OSX and ARM processors.
73 - simgrid::xbt::cmdline and simgrid::xbt::binary_name are gone.
74 Please use simgrid::s4u::Engine::get_cmdline() instead.
77 - New section in the user guide on the provided performance models.
78 - New section presenting some technical good practices for (potential) contributors.
79 - Add a section on errors and exceptions to the API documentation.
80 - Move the s4u examples to a section on their own to ease navigation.
82 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
83 (FG: issues on Framagit; GH: issues on GitHub)
84 - FG#18: Java bindings should be redone or removed
85 - FG!118: Wi-Fi callback mechanism
86 - FG!119: SMPI: add option to inject a barrier before every collective call
87 - GH#383: Segfault when adding a disk after load_platform(xml)
89 ----------------------------------------------------------------------------
91 SimGrid (3.32) October 3. 2022.
93 The Wiedervereinigung release. Germany was reunited 32 years ago.
96 - SimGrid now requires a compiler with C++17 support to compile the lib.
97 Our public headers still allow the user code to be compiled in C++14.
98 - Support graphviz v3 and ns-3 v3.36 (older versions are still supported).
99 - Tested with clang (v11, v13, v14 and v16), gcc (v7 to v13) and IntelCC v2022.2
103 - kill signal Comm::on_completion that was not working anyway.
104 - Expose signals Activity::on_suspend and Activity::on_resume
105 - New macro xbt_enforce(): similar to xbt_assert(), but throws an AssertionError
106 instead of calling abort().
107 - New: s4u::Exec::get_thread_count()
108 - Various cleanups around virtual machines:
109 - host_by_name() and friends now only return hosts. VMs are now excluded.
110 - It is now impossible to search a VM by name globally.
111 You can only search VM by name on a given PM, so either you know
112 the PM on which your VM runs and you can search by name, or you need
113 to manually iterate over all PMs to search this VM.
114 - The s4u::VirtualMachine constructor is now deprecated.
115 Please use s4u::Host::create_vm() instead.
116 - Rename s4u::VirtualMachine::on_creation() to on_vm_creation() to
117 avoid confusion with s4u:Host::on_creation() that is inherited.
118 Also s4u::VirtualMachine::on_destruction -> on_vm_destruction().
120 - One-sided communications (Comm::sendto) can now be detached,
121 and should now be more resilient to network and host faults.
124 - Added the following bindings / examples:
125 - Comm (now 100% covers the C++ interface):
126 - Comm.dst_data_size, Comm.mailbox, Comm.sender, Comm.start_time, Comm.finish_time
127 - Comm.state_str [examples: examples/python/comm-failure/, examples/python/comm-host2host/]
128 - Comm.remaining [examples: examples/python/comm-host2host/, examples/python/comm-suspend/]
129 - Comm.set_payload_size [example: examples/python/comm-host2host/]
130 - Comm.set_rate [example: examples/python/comm-throttling/]
131 - Comm.sendto, Comm.sendto_init, Comm.sendto_async [example: examples/python/comm-host2host/]
132 - Comm.start, Comm.suspend, Comm.resume [example: examples/python/comm-host2host/]
133 - Comm.test_any [example: examples/python/comm-testany/]
134 - Comm.wait_until [example: examples/python/comm-waituntil/]
136 - Engine.host_by_name [example: examples/python/comm-host2host/]
137 - Engine.mailbox_by_name_or_create [example: examples/python/comm-pingpong/]
139 - Mailbox: Mailbox.ready [example: examples/python/comm-ready/]
140 - Ptask [example: examples/python/exec-ptask/]:
141 - this_actor.exec_init
142 - this_actor.parallel_execute
145 - Added an AssertionError exception that may be thrown in case of error.
146 For instance, creating two hosts with the same name will now throw this exception
147 instead of killing the interpreter.
150 - Implement MPI_File_get_type_extent(), MPI_File_s/get_atomicity() and
151 MPI_File_get_byte_offset()
152 - Intercept getpid() calls to return the simulated ones.
153 - Fix various bugs in MPI IO.
155 Platform description & visualization:
156 - More robust sanity checks for platforms, to reject forbidden topologies with
157 a proper error message.
158 - New platform example: supernode.cpp and supernode.py.
159 The Python version generates a nice graphical representation of the platform.
160 - Bug fixes around fat-tree topologies.
161 - Allow to dump the platform topology as a CSV file representing the graph edges
162 with platform_graph_export_csv() (similar to the DOT export).
163 - Fix graphicator for "cluster" topologies (e.g. fat-tree, dragonfly).
166 - Fix a bug when using ptasks with multicores (FG!111).
169 - First bits of sthread, that intercepts pthread operations at runtime.
170 The intend is to use it together with simgrid-mc, but it is TBD.
171 - Sync MBI generators with upstream changes.
172 - Various cosmetics, small bug fixes and inner refactorings
174 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
175 (FG: issues on Framagit; GH: issues on GitHub)
176 - FG#105: "Variable penalty should not be negative!" with in-flight messages and bandwidth profiles
177 - FG#109: Application time reported by --cfg=smpi/display-timing:yes is wrong
178 - FG#110: Wait_any does not trigger new model solve when host events occur
179 - FG#111: Wrong execution time in rare cases when using multicore
180 - FG!98: Re-enable the tests for legacy stochastic profiles
181 - FG!109: Trigger new engine solve upon host events such as host on/off
182 - FG!116: SMPI/replay: Fix issue with recv of size =0
184 ----------------------------------------------------------------------------
186 SimGrid (3.31) March 22. 2022.
188 The ненасильство release. We stand against war.
190 Against the agression by a sick system that forces peoples to take arms against each other.
193 - Rework the internals, for simpler and modern code. This shall unlock many future improvements.
194 - You can now define plugins onto the DFS explorer (previously called SafetyChecker), using the
195 declared signals. See CommunicationDeterminism for an example.
196 - Support mutex, semaphore and barrier in DPOR reduction
197 - Seems to work on Arm64 architectures too.
198 - Display a nice error message when ptrace is not usable.
199 - New test suite, imported from the MPI Bugs Initiative (MBI). Not all MBI generators are integrated yet.
200 - Remove the ISP test suite: it's not free software, and it's superseeded by MBI.
203 - fix for FG#100 by ensuring small asynchronous messages never overtake larger
204 ones, conforming to the standard.
205 - replay: fix waitall behaviour to avoid forgetting requests and leaking
207 - tracing: ensure that we dump the TI traces continuously during execution and
208 not just at the end, reducing memory cost and performance hit.
209 - Update OpenMPI collectives selection logic to match current one (4.1.2)
210 - Add a coherence check for collective operation order and root/MPI_Op
211 coherence. Potentially costly so not activated unless smpi:pendantic is set
212 or -analyze is given.
215 - New signal: Engine::on_simulation_start_cb()
216 - Introduce a new execution mode with this_actor::thread_execute(). This simulate
217 the execution of a certain amount of flops by multiple threads ran by a host. Each
218 thread executes the same number of flops, given as argument. An example of this new
219 function can be found in examples/cpp/exec-threads.
220 - Reimplementation of barriers natively.
221 Previously, they were implemented on top of s4u::Mutex and s4u::ConditionVariable.
222 The new version should be faster (and can be used in the model-checker).
223 - Actor::get_restart_count(): Returns the number of reboots that this actor did.
226 - MSG_barrier_destroy now expects a non-const msg_barrier parameter.
228 New plugin: the Chaos Monkey (killing actors at any time)
229 - Along with the new simgrid-monkey script, it tests whether your simulation
230 resists resource failures at any possible timestamp in your simulation.
231 - It is mostly intended to test the simgrid core in extreme conditions,
232 but some users may find it interesting too.
235 - New solver for parallel task: BMF.
236 - More realistic sharing of heterogeneous resources compared to the fair
237 bottleneck solver used by ptask_L07.
238 - Implement the BMF (Bottleneck max fairness) fairness.
239 - Improved resource sharing for parallel tasks with sub-flows (parallel
240 communications between same source and destination inside the ptask).
242 - "--cfg=host/model:ptask_L07 --cfg=host/solver:bmf": enable the ptask
243 model with BMF solver.
244 - "--cfg=bmf/max-iterations: <N>": maximum number of iterations performed
245 by BMF solver (default: 1000).
246 - "--cfg=bmf/precision: <N>": numerical precision used when computing
247 resource sharing (default: 1e-12).
248 - This model requires Eigen3 library. Make sure Eigen3 is installed to use BMF.
251 - Modifications of the Profile mechanism, with some impact on users
252 - Addition of a new (S4U) method to init profiles from generic functions to improve versatility
253 - Fix initial behaviour of state_profiles
254 - Modify periodicity to behave like a period, and not like a loop delay
257 - Drop xbt_dynar_shrink().
260 - Made the following bindings static (previously member functions):
261 - Actor: Actor.kill_all(), Actor.by_pid()
262 - Host: Host.by_name(), Host.current(), Host.on_creation_cb()
263 - Mailbox: Mailbox.by_name()
264 - Added the following bindings:
265 - this_actor.warning()
266 - Mailbox.put_init() [example: examples/python/comm-waitallfor/]
267 - Comm.detach() [example: examples/python/comm-waitallfor/]
268 - Comm.wait_for() [example: examples/python/comm-waitfor/]
269 - Comm.wait_any_for()
270 - Comm.wait_all_for() [example: examples/python/comm-waitallfor/]
271 - Mutex [example: examples/python/synchro-mutex/]
272 - Barrier [example: examples/python/synchro-barrier/]
273 - Semaphore [example: examples/python/synchro-semaphore/]
276 - Remove target "make uninstall" which was incomplete and no longer maintained.
278 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
279 (FG: issues on Framagit; GH: issues on GitHub)
280 - FG#57: Mc SimGrid should test whether ptrace is usable
281 - FG#87: Smpi scripts fail with spaces in paths
282 - FG#100: [SMPI] Order of the message matching is not guaranteed
283 - FG#101: LGPL 2.1 is deprecated license
284 - FG#104: "make uninstall" not up-to-date
285 - GH#151: Missing mutexes for DPOR.
287 ----------------------------------------------------------------------------
289 SimGrid (3.30) January 30. 2022.
291 The Sunday Bloody Sunday release.
293 Main user-visible changes:
294 - The SimDag API for the simulation of the scheduling of Directed Acyclic
295 Graphs has been dropped. It was marked as deprecated for a couple of years.
296 We finally complete the implementation of what has been called SimDag++
297 internally, i.e., porting the different features of SimDag on top of S4U.
298 The new way to simulate the execution of dependent activities directly by
299 maestro (without any other actor) is details in the examples/cpp/dag-* series
301 - The removal of SimDag led us to also remove the export to Jedule files that
302 was tightly coupled to SimDag. The instrumentation of DAG simulation is still
303 possible through the regular instrumentation API based on the Paje format.
304 - We also dropped the old and clumsy Lua bindings to create platforms in a
305 programmatic way. It can be done in C++ in a much cleaner way now, which
306 motivates this suppression.
309 - Introduce on_X_cb() functions for all signals, to attach a new
310 callback to the signal X. The signal variables are now hidden and
311 only these functions should be used.
312 Rationale: this enables the usual deprecation scheme where functions
313 remain for 4 releases if we need to modify the signals, while the
314 current code with the signal variables directly visible prevents any
316 - New function: Engine::run_until(date), to split the simulation.
317 - New signal: Activity::on_veto, to detect when an activity fails to start.
318 - Signal change: Comm::on_start(Comm&, bool) has been replaced by
319 Comm::on_send and Comm::recv. These two signals respectively correspond to
320 when the sending or receiving side of a Comm is ready. They are raised at
321 the same locations as the former Comm::on_start signal.
322 - New function: Engine::track_vetoed_activities() to interrupt run()
323 when an activity fails to start, and to keep track of such activities.
324 Please see the corresponding example for more info.
325 - New functions: s4u::Comm::{sendto_init, set_source, set_destination} to enable
326 the use of vetoers with direct host-to-host communications. Both source and
327 destination have to set for a comm to start. Each call to these setters check
328 if all vetoes are satisfied. When it is the case, the comm starts. A use case of
329 these functions is given in examples/cpp/dag-scheduling.
330 - New functions: {Exec, Io}::update_priority allow you to modify the priority of
331 these kinds of activities during their execution. Behavior is detailed in
332 examples/cpp/io-priority/.
335 - Dynamic costs for MPI operations: New API to allow users to dynamically
336 change injected costs for MPI_Recv, MPI_Send and MPI_Isend operations.
337 Alternative for smpi/or, smpi/os and smpi/ois configuration options.
338 - Fix some issues with the replay mechanism.
341 - Function xbt::Extendable::get_data() is now templated with the type of the
342 pointee. Untyped function is deprecated. Use get_data<void>() if you still
343 want to retrieve void*.
346 - New section: "SimGrid MPI calibration of a Grid5000 cluster"
347 presenting how to properly calibrate MPI communications in SimGrid.
348 - Complete and reword the platform section, which is now completed.
351 - Thread contexts are used by default with Python bindings. Other kinds of
352 contexts revealed unstable, specially starting with pybind11 v2.8.0.
354 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
355 (FG: issues on Framagit; GH: issues on GitHub)
356 - FG#95: Wrong computation time for multicore execution after pstate change
357 - FG#97: Wrong computation time for ptask+multicore+pstates
358 - FG#98: SMPI offline simulation is inconsistent with the online simulation
359 (deadlocks / message truncation)
360 - FG#99: Weird segfault when not sealing an host
362 ----------------------------------------------------------------------------
364 SimGrid (3.29) October 7. 2021
366 The "Ask a stupid question" release.
368 We wish that every user ask one question about SimGrid to celebrate.
369 On Mattermost, Stack Overflow or using the issues tracker.
371 New modeling features:
372 - Non-linear resource sharing, modeling resources whose performance heavily degrades with contention:
373 - The total capacity may be updated dynamically through a callback
374 and depends mainly on the number of concurrent flows.
375 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
377 - Dynamic factors: model variability in the speed of activities
378 - Each action can now have a factor that affects its progression.
379 This multiplicative factor is applied when updating the amount of work
380 remaining, thereby an activity with factor=0.5 only uses half of the
381 instantaneous power/bandwidth it is allocated and will appear twice
382 slower than what it actually consumes.
383 - This can be used to model a overhead (e.g., there is a 20 bytes
384 header in a 480 bytes TCP packet so the factor 0.9583) but the novelty
385 is this factor can now easily be adjusted depending on activity's and
386 resources characteristics.
387 - This existed for network (e.g., the effective bandwidth depends
388 on the message in SMPI piecewise-linear network model) but it is now
389 more general (the factor may depend on the source and destination and
390 thus account to different behaviors for intra-node communications and
391 extra-node communications) and is available for CPUs (e.g., if you
392 want to model an affinity as in the "Unrelated Machines" problem in
393 scheduling) and disks (e.g., if you want to model a stochastic
395 - For that, resources can be provided with a callback that computes
396 the activity factor when creating the action.
397 - Example: examples/cpp/exec-cpu-factors
398 - The same mechanism is also available for the latency, which
399 allows to easily introduce complex variability patterns.
402 - Added support to programmatic platform creation in Python.
403 Example: examples/python/clusters-multicpu
406 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
407 This can only be set through the API, not through XML files.
411 - Multiple fixes to ensure reproducibility of tracing
412 - scan/excan can now be replayed
413 - wait action now uses ranks and not pid, as the other ones.
414 - smpi/init and smpi/finalization-barrier are now valid for replays.
415 - exit() is now intercepted by SMPI to avoid premature shutdown of
416 simulation. First non 0 return codes is returned as simulation return
420 * New section "Release Notes" documenting recent and current developments.
421 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
422 * Improvements in API Reference for C++ and Python interfaces.
425 - Make wifi creation compatible with ns-3 version 3.34 too.
427 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
428 (FG: issues on Framagit; GH: issues on GitHub)
429 - FG#77: Search feature of doc is broken (update sphinx theme version)
430 - FG#78: Multiple fixes for SMPI replay:
431 - TI tracing of allotallv/w was outputting wrong values
432 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
434 ----------------------------------------------------------------------------
436 SimGrid (3.28) July 14. 2021
438 The Victoriadagarna Release.
441 - C++ platform interface: Users can now describe their platform directly in C++.
442 This provides greatly flexibility and performance improvement for complex
443 platforms. Main features:
444 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
445 "zones", instead of single hosts. This feature enables the description
446 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
447 - StarZone: new zone with a Star-like topology. The routes are defined
448 as a set of links used to communicate from node to everybody (node<->ALL).
449 - Split-Duplex links: auxiliary method to create split-duplex links in
450 the platform, easing its utilisation. It automatically creates both UP
451 and DOWN links (similarly as done in XML).
452 - Please refer to the documentation and the examples included:
453 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
454 - New plugin: Producer-Consumer with monitor. Just requires to include the
455 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
456 associated example (examples/cpp/plugin-prodcons).
459 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
460 s4u::Io::wait_any(), s4u::Io::wait_any_for().
461 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
462 parameter by reference, instead of a pointer.
463 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
464 Explicitly cancel the activity to get back to previous behavior.
465 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
468 - The default SMPI compiler flags are no more taken from the environment.
469 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
470 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
472 --cfg=smpi/finalization-barrier: which can be used to add
473 a barrier inside MPI_Finalize. This can help for some codes which cleanup
474 data attached to a process, but still used in other SMPI processes.
475 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
476 is active when set to false, to keep going after a small error
477 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
478 which may or may not be problematic in the end.
480 - fix behaviour, as maximum iteration count could be ignored
481 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
482 use sampling when the same kernel is called with a different set of parameters
483 which have an impact on the timing.
484 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
485 It should now work with smpi/auto-shared-malloc-thresh.
486 - Improve error handling and reporting in multiple places
487 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
488 validity checks, truncated messages are now an error, return errors
489 when explicitely deleted handles are reused, ...)
490 - RMA: multiple fixes and stability improvements.
491 - analysis (-analyze flag in smpirun):
492 - SMPI can now report buffer leaks as well as MPI handles leaks,
493 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
494 - if -trace-call-location is used when compiling, SMPI can report
495 origin of leaked handles/buffers
496 - group leaks by type/origin in output message if possible
497 - New implemented MPI calls: MPI_Comm_test_inter
500 - Changed internal implementation of bandwidth factors in network models.
501 Models affected: CM02, LV08 (default), SMPI, IB.
502 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
503 Bandwidth factors are applied to communications to describe that users
504 cannot use 100% of the available bandwidth. For example, the default network model,
505 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
506 behavior was implemented by limiting the bandwidth available in the LMM
507 system for this flow. This may give the false impression that there is
508 bandwidth available for other flows due to its underutilization, especially
509 for the dynamic bandwidth factors used in SMPI models.
510 To avoid this, we have modified the implementation so that each flow uses the
511 maximum physical bandwidth according to the LMM system.
512 However, the actual throughput of the flow seen by the user is defined by
513 the physical bandwidth multiplied by the bandwidth factor.
514 This change impacts on the simulation results for all network models on
515 which we have bandwidth factors configured.
516 *****************************************
517 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
518 *****************************************
519 This change may impact on the timing of your simulation results.
520 Take care when comparing simulations from different SimGrid's
521 versions. Sorry for the inconvenience.
522 - Dynamic network factors: users can configure a callback to define
523 the network factors dynamically. This API is available at
524 simgrid::kernel::resource::NetworkModelIntf.
525 - Users have access to complete information about the current communication
526 to decide which factor to apply. This includes: message size, source and
527 destination hosts, links and zones traversed.
528 - Dynamic factors for both latency and bandwidth.
529 - For more details, see the example in (examples/cpp/network-factors).
530 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
531 deprecated since version 3.24, are no longer supported. Instead, use
532 "wattage_off" and "wattage_per_state".
535 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
538 - New tutorial: Model-checking and formal assessment
539 - New sections: "Demystifying the routing" and "C++ platforms"
540 - Update and improve the part on visualization in MPI and Algo tutorials.
541 - Improve the section on routing: how to define it, how it's used internally
542 - Fix many issues, broken links and missing references in doxygen and Sphinx
545 - Lua platform files are deprecated. Their support will be dropped after v3.31.
548 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
550 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
551 (FG: issues on Framagit; GH: issues on GitHub)
552 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
553 - FG#64: Configuring smpi/IB-penalty-factors
554 - FG#67: Running computation concurrently with MPI_Iallreduce
555 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
556 - FG#71: Segmentation fault on invalid gw_src/gw_dst
557 - GH#322: Issue when an actor kills his host vm
559 ----------------------------------------------------------------------------
561 SimGrid (3.27) March 29. 2021
563 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
565 Main user-visible changes:
566 - All functions marked as deprecated for one year were removed.
567 - Removed all that was related to the "Storage" type in all APIs and bindings.
568 "Disks" have been introduced more than a year ago to replace them for the better.
569 It's time to finish this replacement.
570 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
571 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
572 avoiding "initializer element is not constant" errors with some codes. Please be
573 careful to recompile your MPI codes if switching to this new SimGrid version.
574 - Network model ns-3 now requires version ns-3.28 or higher.
575 - Removed some option aliases that were deprecated since v3.23 (June 2019)
576 - 'verbose-exit' is now 'debug/verbose-exit'
577 - 'simix/breakpoint' is now 'debug/breakpoint'
578 - 'clean-atexit' is now 'debug/clean-atexit'
581 - New functions: Comm::sendto_{init,async} to initiate a communication
582 on between two (possibly remote) hosts.
583 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
584 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
585 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
586 or Mailbox::get_async<void>() if you really want to play with void*.
587 - Unify the interface of Activity::on_{start/activity}
588 - New function: Comm::get_dst_data()
589 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
592 - ptask should now work better with multicore hosts (FG#37).
593 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
594 - Internal cleanups and refactoring around models, to prepare for programmatic
596 - Modernize our tests of the performance models
600 - New section: Configuring the logs from the command line.
601 - New section: Programming API of logs.
602 - New section: Getting realistic results
603 - New section: Examples of SMPI Usage
604 - New part: SimGrid's Internals, with a section on Design goals
605 - Move back to the classical Breathe tool chain, and fix many small glitches.
608 - Handle leak checking: SMPI will now display a message if MPI handles are still
609 present at the end of a simulation. Option smpi/list-leaks:n will display the n
610 first leaked pointers and their types.
611 - smpi/display-allocs option : Display Memory allocations analysis, with hints
612 about the largest allocations. This only works for allocations with malloc/calloc
613 and not compiled with SMPI_NO_OVERRIDE_MALLOC
614 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
615 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
616 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
617 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
618 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
619 - Replay now opens and parses the trace file only once for all ranks when possible.
621 - better handling of MPI_PROC_NULL
622 - MPI_Group_incl was broken in some cases
623 - PAPI tracing now fixed
624 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
625 - SMPI will now fail properly when receive size < sent size, instead of
626 just setting MPI_ERR_TRUNCATE and keep going.
627 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
628 - Remove useless build patches and split build/run patches on the page
629 - Build in parallel when possible
630 - Test on yum-based systems
631 - New nightly tested Proxy-apps:
632 - OpenMPI mpi-test-suite
633 - Coral MCB, LSMS, Qbox, CLOMP
634 - ECP/Mantevo Cloverleaf3D, Ember
635 - Codevault bhtree_mpi
638 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
639 - Start to overhaul the implementation to modernize it (still ongoing).
642 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
643 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
644 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
645 - Deprecate xbt_dynar_to_array().
646 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
648 C binding and interface:
649 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
650 - Constness requirement has been lowered for parameter argv of sg_actor_start()
651 and sg_actor_create() It's now (char* const*). The previous version is still
652 available as sg_actor_start_() and sg_actor_create_().
654 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
655 (FG: issues on Framagit; GH: issues on GitHub)
656 - FG#37: Parallel tasks are limited to 1 core per host
657 - FG#62: Running "smpirun -replay" on large networks
658 - FG!46: Fix a few potential memory leaks in SMPI colls
659 - GH#343: Wrong consumed energy when an exec timeouts
661 ----------------------------------------------------------------------------
663 SimGrid (3.26) December 15. 2020
665 The Release release (the French lockdown was eased today).
667 Important user-visible changes:
668 - SimGrid now requires a compiler with C++14 support.
669 Sibling projects should upgrade their FindSimgrid.cmake
670 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
671 several users had difficulties to understand issues when using high bandwidth or
672 small latency events. The new value was already the default for SMPI and
673 should not cause too much performance hit. It can change some simulation
674 timings if simulation was skipping a lot of small events (which was actually
675 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
676 - On some network models (L07, CM02, LV08), an implicit loopback link is created
677 for local communication. Default bandwidth was 500MBps and latency was 15µs,
678 which was rather bad. Default values have been changed to 10GBps and 0 latency.
679 Options network/loopback-lat and network/loopback-bw have been added to adapt
680 these values if needed. Network and routing models already providing separate
681 loopbacks (clusters and such) are not impacted by this change.
682 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
683 metadata that are added to each MPI message by MPI libraries (tag,
684 src/dest, and communicator information). This may change simulation times
685 for very small message sizes, but should be more realistic.
688 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
689 - New plugins: wifi_energy and link_load. See the associated examples.
690 - New examples: wifi networks, and communication suspend.
693 - LTO is now enabled for Intel/clang compilers.
694 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
695 Setting it to "auto" will use all available cores, while setting it to n will
696 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
697 - Remove obsolete runtime option 'contexts/parallel-threshold'.
698 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
699 has been kept for compatibility.
700 - Finally remove obsolete snake_case() aliases for runtime options.
701 - Further improve the documentation.
704 - the on_exit() of each actor is also executed when the simulation deadlocks.
705 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
706 An example is provided for s4u::Comm but it should work with Exec and Io.
709 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
710 https://framagit.org/simgrid/SMPI-proxy-apps
711 - MPI/IO: supports names starting with "./"
712 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
714 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
715 building shared libraries, this may cause issues, so environment variable
716 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
717 - most temporary files should now be created in /tmp dir (or equivalent).
718 If this one does not allow execution of code (noexec flag), this may cause issues.
719 Please use another tmp directory (using TMPDIR or equivalent system variable)
723 - A large refactoring is ongoing, but this should be transparent for users so far.
726 - Drop xbt_str_split().
728 C binding and interface:
729 - The return type of the sg_actor_on_exit() callbacks is now 'void'
730 (int value was previously ignored)
731 - Many C functions were renamed to follow a common naming style
732 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
733 names are kept until version 3.30.
734 - Many MSG tests were converted to the new S4U's interface in C, that
735 was extended for that.
737 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
738 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
739 - FG#41: Add sg_actor_create C interface
740 - FG#43: xbt::random needs some care
741 - FG#48: The Impossible Did Happen (yet again)
742 - FG#50: Suspending an actor executed at the current timestamp fails
743 - FG#52: Zero-seconds timeout : "That's in the past already"
744 - FG#53: Crash while using ns-3 network model
745 - FG#54: How to suspend a comm?
746 - FG!22: Stochastic Profiles
747 - FG!24: Documentation and fix for xbt/random
748 - FG!35: Add a modeling hint for parallel links in doc
749 - FG!36: [xbt/random] Read/Write the state of the RNG
750 - FG!42: Link cumulated load plugin
751 - FG!44: Test condition variable wait_until+notify_all
752 - GF#18137: Allow different stack sizes?
753 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
754 - GH#139: Allow pthread creation in SMPI
755 - GH#336: Packet-level simulation using SMPI?
756 - GH#345: Error in the LMM while migrating a VM
757 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
758 - GH#352: pip install / python setup.py install fail to find pybind11
759 - GH!337: Fix link_energy plugin for wifi platforms
760 - GH!339: Add Mailbox set_receiver method to python binding
761 - GH!344: Cast hugepages macros parameters to int64
763 ----------------------------------------------------------------------------
765 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
767 The Palindrom Day Release.
769 Important user-visible changes:
770 - Improve the Python usability (stability and documentation).
771 - A nasty synchronization bug was ironed out, see also below.
772 - Python's doc was organized as a separate tree, now integrated with C++.
773 - C bindings of S4U were not part of the doc.
774 - The C++ doc was also improved as methods are now split by theme.
775 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
776 - This is mandatory to use the Java bindings.
777 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
778 - SimDAG++: Automatic dependencies on S4U activities (experimental)
779 - Some features are already implemented but not all of them
780 - Cannot block an activity until it's scheduled on a resource
781 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
782 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
785 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
786 - Drop support for Viva/Triva (old visualization tools), hence removed
787 TRACE_get_node_types() and TRACE_get_edge_types() functions.
790 - Require pybind11 version 2.4.
791 - Greatly improve locking during context switches with Python.
792 - Declaring Python callbacks to this_actor::on_exit is now working.
793 - No more tests are failing currently, even on win32.
796 - Actor: Merge signals on_migration_start/end into on_host_change
797 - Actor: Rename migrate() into set_host()
798 - Disk: Allow users to get the read and write nominal bandwidth values
799 - Exec: Implement wait_for(timeout)
800 - Io: Implement wait_for(timeout)
803 - Parse errors now raise a simgrid::ParseError that you may want to catch.
806 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
807 been renamed to fix a spelling error. As usual, the previous definitions are
808 kept with a deprecation warning. However, to avoid runtime errors, it is now
809 forbidden to override these deprecated methods in a derived class. Please use
810 the new names immediately if you need to override them.
813 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
814 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
815 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
816 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
817 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
818 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
819 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
820 - New module: random, providing classical random numbers generators.
823 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
824 allocations' size above which they are considered "shared" by default (as if
825 they were performed through SMPI_SHARED_MALLOC macros).
826 Default = 0 = disabled feature.
827 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
828 This can cause some troubles if codes are already overriding these. If this
829 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
830 help, but will make this feature unusable.
831 - Input check is now performed much more consistently and outputs precise
832 warnings for each error found
833 - New supported calls : MPI_File_set_view, MPI_File_get_view
834 - MPI I/O now only supports disks, as storage is discontinued.
835 - MPI I/O allows opening files without specifying absolute path
836 - Note : SMPI configuration options are now initialized by smpi_init_options()
837 call, which should be called if SMPI is not being used through smpirun.
839 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
840 - FG#9: Python bindings crashing
841 - FG#39: Missing s4u::Comm::wait_any_for example
842 - FG#42: Add support for ThreadSanitizer (TSan)
843 - FG!19: Removing RngStream
844 - FG!20: A module for RNG calls
845 - FG!21: Choice between ad-hoc and standard distributions implementations
847 - FG!25: Fix link in Mutex doc
848 - FG!26: Fix links in SMPI interface doc
849 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
850 - GH#305: Unscheduled tasks are still excuted
851 - GH#313: smpirun: manual outdated w.r.t. --help
852 - GH#321: [S4U] Get task remaining work ratio
853 - GH#323: Crash when an actor turn off his physical host
854 - GH#335: Missing links on dragonfly example svg
855 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
856 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
858 ----------------------------------------------------------------------------
860 SimGrid (3.24) October 9. 2019
862 The Clean Disk Release.
865 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
866 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
867 thus becomes the default resource to execute I/O operations. Examples
868 and the file system plugin have been modified to use disks instead of
869 storage elements. Storage can still be used but is doomed to disappear.
870 Users are thus adviced to update their simulators accordingly.
871 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
872 for consistency with pthread_barrier_wait()
873 - Host::get_englobing_zone() returns the englobing netzone
874 - Actor::on_destruction is now called in the destructor
875 Actor::on_termination new signal called when the actor terminates
877 - Global signals are now part of the Engine:
878 - on_platform_creation: after config settings, before the XML parsing
879 - on_platform_created: right after the XML parsing
880 - on_time_advance: each time the clock advances
881 - on_simulation_end: after simulation, before cleanups
882 - on_deadlock: as the name implies.
884 - sg_{actor,host,link}_{data,data_set}() now all exist.
885 Use them to attach user data to the object and retrieve it.
888 - Introduce an experimental Wifi model. It sounds reasonable
889 according to the state of the art, but it still has to be properly
890 validated, at least against ns-3 if not against reality.
891 - Improved the usability of ns-3. Several bugs were ironed out.
892 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
893 It is now expressed as 'idle:epsilon:allCores' to properly model the
894 consumption of non-whole tasks on mono-core hosts. More info in the doc.
897 - convert a new set of functions to the S4U C interface and move the old MSG
898 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
901 - Fortran bindings for DVFS have been removed.
902 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
903 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
904 - SMPI can now be selected by cmake's find_module(MPI) with
905 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
906 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
907 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
908 - trace-call-location can be used with TI traces, and replayed, and is compatible
909 with smpi/comp-adjustment-file.
910 - sleep events are now correctly traced/replayed.
911 - Default for trace-call-location is now to use file names and not full paths.
912 To revert to previous behavior (in case of collision of filenames), option
913 "smpi/trace-call-use-absolute-path" can be set to yes.
916 - Use the included xxHash as an hash implem when C++14 is usable.
917 - Option model-checker/hash was removed. This is always activated now.
918 - New option smpi/buffering controls the MPI buffering in MC mode.
919 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
920 This is useful to check for MPI compliance.
923 - New section on plugins: how to define a new one, and existing ones.
926 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
927 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
930 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
931 and <mount> tags. The rationale is finish to decouple what should be handle
932 by SimGrid kernel (i.e., resources and activities) from plugins built in
933 user space (e.g., the file system plugin). This new tag currently accepts
934 only three attributes (id, read_bw, and write_bw). All the other information
935 that was declared with the storage related tags now has to be expressed as
936 properties. An example of platform using this new tag is available at
937 examples/platforms/hosts_with_disks.xml
940 - 'expect signal' can now accept more than one potential signal.
942 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
943 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
944 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
945 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
946 - FG#31: per-actor data
947 - FG#34: SG_BARRIER_SERIAL_THREAD?
948 - FG#35: model-checker does not like buster-produced binaries
949 - FG!13: MC: complete workaround in the error msg seen on modern systems
950 - FG!15: execute_flops now logs compute
951 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
952 - FG!17: ns-3: unblock the right number of communications + others issues
953 - FG!18: Improving the performance of the ns-3 bindings
954 - GH#189: Energy consumption of parallel tasks
955 - GH#219: Error in the throughput of TCP transfer
956 - GH!330: Minor polishing of English: Configuring_Simgrid
957 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
958 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
959 - GH!334: Rework energy plugin, again
961 ----------------------------------------------------------------------------
963 SimGrid (3.23.2) July 8. 2019
966 - Nicer introduction page.
967 - Migrate the "Deploy your application" page to the new doc.
968 - Move Java as a subtree of MSG.
971 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
974 - Drop xbt_dynar_sort_strings().
977 - Really fix FG#26: Turning off a link should raise NetworkFailureException
978 - FG#27: Wrong exception thrown to wait_any when link is turned off
979 - GH#328: Java: Canceling multiple tasks in a single vm/host
981 ----------------------------------------------------------------------------
983 SimGrid (3.23) June 25. 2019
985 The Exotic Solstice Release.
988 - SunOS and Haiku OS support. Because exotic platforms are fun.
989 - Stop setting random seed with srand48() at initialization.
990 - Use addr2line as a fallback for stacktraces when backtrace is not available.
991 - Build option -Denable_documentation is now OFF by default.
992 - Network model 'NS3' was renamed into 'ns-3'.
995 - Simgrid can now hopefully be installed with pip.
998 - wait_any can now be used for asynchronous executions too.
1001 - New log appenders: stdout and stderr. Use stdout for xbt_help.
1002 - Drop xbt_dict_dump.
1005 - 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
1006 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
1007 - MPI interface is now const correct for input parameters
1008 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
1009 - Fortran bindings for async collectives.
1010 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
1013 - Remove option 'model-check/record': Paths are recorded in any cases now.
1014 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
1015 always sparse. That's an immense gain in memory, and can even be
1016 faster because of cache effects. So there is no need to clutter the
1017 code to allow the user to go for the unefficient mode.
1020 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
1021 models proved to be more accurate than these old experiments.
1023 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
1024 - FG#1: Broken link in error messages
1025 - FG#2: missing installation documentation
1026 - FG#3: missing documentation in smpirun
1027 - FG#6: Python bindings not available on PyPI
1028 - FG#7: simple cmake call requires doxygen
1029 - FG#8: make python bindings an optional dependency
1030 - FG#10: Can not use MSG_process_set_data from SMPI any more
1031 - FG#11: Auto-restart actors forget their on_exit behavior
1032 - FG#12: -Denable_lto=OFF doesn't disable LTO
1033 - FG#13: Installs unstripped file 'bin/graphicator'
1034 - FG#14: Installs the empty directory 'doc/simgrid/html'
1035 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
1036 - FG#17: Dead link in doc (pls_ns3)
1037 - FG#20: 'tesh --help' should return 0
1038 - FG#21: Documentation link on http://simgrid.org/ broken
1039 - FG#22: Debian installation instruction are broken
1040 - FG#26: Turning off a link should raise NetworkFailureException exceptions
1041 - GH#133: Java: a process can run on a VM even if its host is off
1042 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
1043 - GH#326: Valgrind-detected error for join() when energy plugin is activated
1045 ----------------------------------------------------------------------------
1047 SimGrid (3.22) April 1. 2019
1049 The Easter Christmas Release (this one is somewhat late).
1052 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
1053 - Not complete yet: asynchronous activities (amongst others) are still missing
1054 - Still ongoing: the interface may change in the future. We need more
1055 testers! Please report any glitches.
1056 - No new project using Java should start now. Please switch to Python.
1059 - Some of the internal cleanups may lead to speed improvements:
1060 - The hard limitation on the amount of simulated actors with
1061 Java+Mac was removed. Now, the available memory is the only limit.
1062 - Our refcounting was tidyied, leading to 10% speedups in some cases.
1063 - We are still working on making our code robust to the actor kills
1064 and hosts' churn. Things are improving, but it's not perfect yet.
1065 - Replaced our own code to display a backtrace (that was forking addr2line)
1066 with the Boost.Stacktrace library.
1067 You won't see your backtraces without this optional dependency.
1068 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
1069 - Stop setting random seed with srand() at initialization.
1072 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
1073 XML file version remains 4.2 since old files are still compatible.
1076 - Process termination which was broken at version 3.21 has been repaired.
1077 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
1078 - Hide the examples into examples/deprecated. New users should use Python.
1081 - Drop MSG_process_create_from_stdfunc() from the API.
1082 This C++ function was a pimple in the C API, made necessary at some
1083 point by the Java bindings. This is fixed now.
1084 - Hide the examples into examples/deprecated. New users should use S4U.
1085 - MSG_process_create and MSG_process_attach now crash if the host used
1086 to run this process is off.
1087 - Fix the protype of MSG_process_on_exit()
1088 Now use (int,void*) callbacks instead of (void*,void*) ones.
1089 The implementation was ways too messy to actually work, I guess.
1092 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
1093 too often and provide more accurate timings.
1094 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
1095 - MPI_Alltoallw support
1096 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
1097 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
1098 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
1099 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
1101 - Basic implementation of generalized requests (SMPI doesn't
1102 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
1105 - Drop sg_cmdline. Please use xbt_cmdline instead.
1106 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
1107 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
1108 OsSemaphore is implemented in a portable way with C++11 threads.
1109 This should allow much more threads to be created at the same time,
1110 allowing Mac and Java users to simulate many more actors.
1111 - Implement the 'thread' factory with std::thread instead of xbt ones.
1112 It is not possible to set the stack size with threads anymore, but
1113 -fsplit-stack is the way to go nowadays when using threads.
1114 - Drop the xbt_os_thread_t module (now unused)
1115 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
1116 - Drop xbt_str_join_array().
1117 - Drop cunit, use Catch2 instead.
1120 - Many cleanups in the kernel::activity namespace. This was long
1121 overdue, and shall open the path to many future endeavors.
1124 - #132: Java: a process can not shut down its own host
1125 - #220: S4U: detached send causes exception if sender terminates soon after sending
1126 - #261: Document the parameters of parallel execution's constructor
1127 - #300: [s4u] BarrierPtr is missing
1128 - #314: SMPI args internal cleanup
1129 - #316: Fix a bug related to the CPU utilization of multi-core VM
1130 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
1131 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
1132 - #325: Turning off a host has different behavior on sleeping actors and computing actors
1134 ----------------------------------------------------------------------------
1136 SimGrid (3.21) October 3. 2018
1138 The Restarting Documentation (TRD) Release.
1141 - Convert the user manual to Sphinx for improved usability.
1142 Unstable version now hosted on https://simgrid.frama.io/simgrid
1143 Many glitches remain; stay tuned.
1144 - Restrict the Doxygen reference API to the installed header files.
1145 (documenting all internals this way is overkill + doxygen scales badly)
1146 - New tutorial on S4U (highly inspired from the old MSG one, but with
1147 a git repository to fork as a starting point, and a docker image)
1148 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
1149 - The Developper manual is still to be converted (not compiled until then)
1150 - Some parts are still missing in this conversion (such as the
1151 platform and deployment sections) while others were blindly converted
1152 and would need more love (such as the configuration flags).
1153 Things will certainly further improve in the future releases.
1156 - Fully reimplement auto_restart mechanism that was utterly broken
1157 (fix #22, #131 and #281 that were all very related)
1158 - Implement semaphores (fix #303 and #291)
1159 - When creating an actor from a function and its parameters,
1160 move-only parameters are not allowed anymore, as it would prevent
1161 the actor to be restartable if its parameters are consumed on run.
1162 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
1163 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
1164 IO operations that can be started, waited for, or canceled as a regular
1166 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
1170 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
1171 - New host variable: 'core_count'
1174 - Remove xbt_os_thread_specific features
1175 - Remove portability wrapper to condition variables
1176 - Remove xbt_os_thread_yield()
1179 - MPICH collective selector now mimics MPICH 3.3b
1180 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
1181 - MPI_Init(NULL,NULL) is now allowed
1182 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
1183 - Command-line arguments should now be handled properly in Fortran simulations
1184 - MPI Attributes and related callbacks should work in Fortran
1185 - Apps using C + Fortran should now work
1186 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
1187 - Support MPI_Error_String
1190 - Due to an internal bug, Msg.run() must now be your last line.
1191 We hope to fix it in a future release, and we are sorry for the inconvenience.
1194 - #22: Process autorestart seem to only work with CAS01 cpus
1195 - #56: Feature request: dynamic SMPI replay
1196 - #93: simgrid should not eat --help
1197 - #111: How to change the loopback link characteristics?
1198 - #116: Platform cluster radical
1199 - #117: Platform cluster/host order
1200 - #129: Replay_multiple on steroid
1201 - #131: Java : autostart does not work
1202 - #143: Setting a breakpoint at a given time
1203 - #235: Network energy models should be integrated
1204 - #264: Add ptask L07 resource tracing
1205 - #271: Dynamic and manual replay of SMPI traces
1206 - #279: Breakpoints option uses time deltas instead of absolute time
1207 - #281: Daemonized actor no longer daemonized after an auto-restart
1208 - #282: TIT convertor erase traces when using absolute path in the trace list file
1209 - #285: segfault when a s4u actor kills itself with msg process tracing activated
1210 - #286: Pajé traces are not exposing the number of cores
1211 - #287: Command-line parsing should stop when encountering '--'
1212 - #288: MPI_Init(NULL, NULL)
1213 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
1214 - #290: Method to check if a mailbox has a message ready to be consumed
1215 - #291: [s4u] semaphores do not seem available yet
1216 - #294: Alltoall collective implementation of mpich
1217 - #295: Starting a Host as OFF is not supported
1218 - #296: DTD too permissive
1219 - #299: Add a s4u barrier example
1221 ----------------------------------------------------------------------------
1223 SimGrid (3.20) Released June 24. 2018
1225 The proxy snake_case() release.
1228 - The whole API is now using snake_case() instead of mixing it with camelCase.
1229 Compatibility wrappers in place for one year.
1230 - We took the opportunity to increase the consistency of the API.
1231 Please report any remaining glitches.
1232 - New function simgrid::s4u::Host::get_actor_count:
1233 Returns the number of actors running on a specific host.
1236 - SMPI is now tested with ~45 proxy apps from various sources, with none or
1237 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
1238 - Replay: The replay file has been re-written in C++.
1239 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
1240 supported. They are stored in the trace and used when replayed.
1241 - Basic support of MPI_Cancel. Robustness not guaranteed.
1242 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
1243 (only for MPI_COMM_TYPE_SHARED).
1244 - New option: smpi/privatize-libs, to add external shared libs to be privatized
1245 by SMPI. They will be copied locally and loaded separately by each process.
1246 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
1247 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
1248 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
1249 with internal index optind with multiple processes. Only works if MPI_Init has
1250 already been called.
1251 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
1252 syntax without preprocessing tricks.
1255 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
1256 behavior tracing is no longer limited to MSG
1257 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
1258 this means that traces now need to be updated or re-obtained; both Irecv
1259 and Isend lines in your traces have now in total 5 mandatory fields:
1260 <rankid> <command> <to/from rankid> <tag> <size>
1261 To update your traces, it suffices to add a 0 for the tag here.
1262 - TIT now also supports waiting for a distinct request via MPI_Wait.
1263 Wait/Test now wait for a specific request, not just the last one that was
1264 issued. This unfortunately means another update, because we need to
1265 identify which request you want to wait for. We do this via the
1266 triplet (sender, receiver, tag), which needs to be added:
1267 <rankid> <command> <sender> <receiver> <tag>
1268 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
1269 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
1272 - The deprecation of MSG is ongoing (but this should not impact you).
1273 Many MSG functions are now simple wrappers on the C API of S4U. If
1274 you wish to convert your code to S4U, find the S4U counterparts of
1275 your MSG calls in src/msg/msg_legacy.cpp.
1276 - MSG can still be used, but won't evolve anymore.
1279 - Allow to run the Link energy plugin from the command line with
1280 --cfg=plugin:link_energy
1281 - Rename Energy plugin into host_energy
1282 - Rename Load plugin into host_load
1285 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
1286 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
1287 - Improve performance when handling timeouts of simix synchros.
1290 - Config: the C API is now deprecated (will be removed in 3.23), and
1291 the C++ API has been slightly improved.
1294 - Fix several build issues on OSX.
1295 - Move simgrid_config.h to simgrid/config.h (old header still working)
1298 - #143: Setting a breakpoint at a given time
1299 - #150: Inconsistent event names in SMPI replay
1300 - #258: daemonized actors hang after all non-daemonized actors have completed
1301 - #267: Linker error on unit_tmgr
1302 - #269: SMPI: tracing of MPI_Wait/all/any broken
1303 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
1305 ----------------------------------------------------------------------------
1307 SimGrid (3.19.1) Released March 22 2018
1309 The Fixed ABI Release.
1311 As you may know, we are currently refactoring SimGrid in deep.
1312 Upcoming SimGrid4 will be really different from SimGrid3: modular,
1313 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
1315 Our goal is to smooth this transition, with backward compatibility
1316 and automatic update paths, while still progressing toward SimGrid4.
1318 SimGrid remains open during works: The last pure SimGrid3 release was
1319 v3.12 while all subsequent versions are usable alpha versions of
1320 SimGrid4: Existing interfaces remain unchanged, but the new S4U
1321 interface is budding and the internals are deeply reorganized.
1323 Since 2015, we work hard to reduce the changes to public APIs. When
1324 we need to rename a public library symbol in S4U, we let your
1325 compiler issue an explicative warning when you use the deprecated
1326 function. These messages remain for four releases, i.e. for one full
1327 year, before turning into an error. Starting with v3.15, your can
1328 also adapt to API changes with the SIMGRID_VERSION macro, that is
1329 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
1331 Starting with this v3.19.1, our commitment to reduce the changes to
1332 the public interfaces is extended from the API to the ABI: a program
1333 using only MSG or SimDag and compiled against a given version of
1334 simgrid can probably be used with a later version of SimGrid without
1335 recompilation. We will do our best... but don't expect too much of
1336 it, that's a really difficult goal during such profund refactoring.
1338 The difference between v3.19 and v3.19.1 is that the former was
1339 accidentally breaking the ABI of MSG, while the later is restoring
1342 S4U and kernel APIs will still evolve until SimGrid4, with one-year
1343 deprecation warnings as currently. In fact, cleaning up these
1344 interfaces and converting them to snake_case() is one release goal of
1345 v3.20. But don't worry, we are working to smooth this upgrade path.
1347 In summary, new projects should start with S4U to benefit of the
1348 future, but old MSG projects should still be usable with no change.
1350 ----------------------------------------------------------------------------
1352 SimGrid (3.19) Released March 20 2018
1354 The Moscovitly-cold Spring release.
1356 SMPI user-visible improvements:
1357 - Support of MPI_Type_create_subarray
1358 - Unimplemented MPI calls now abort (instead of issuing a warning)
1359 Some harmless calls around errhandlers still only issue a warning.
1360 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
1361 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
1362 is now possible (mmap privatization was refactored to this end).
1363 Adding new SMPI instances should also be technically doable now.
1365 SMPI internal cleanups:
1366 - Removed index notion from SMPI -- no more "getPid() - 1"!
1367 In tracing files, actors are called rank-<PROCESS-ID> but these are
1368 NOT the real ranks (tracing several communicators is not supported yet).
1369 - Deprecate smpi_process_index() (will be removed in 3.22)
1370 - Better testing on classical proxy apps, through an external project:
1371 https://github.com/simgrid/SMPI-proxy-apps/
1374 - Use a graphical TOC to make it easier to find the documentation you need
1375 - Revamp the MSG tutorial
1378 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
1379 - Remove parameter of MSG_process_killall().
1380 Resetting the PID was bogus anyway (several actors could have the same PID).
1383 - Execution->setHost() can be called after start(), to migrate it.
1384 - Comm::test_any() is now implemented.
1385 - s4u::Actor now has onCreation() and onDestruction() signals.
1386 - Install some kernel header files for the users' plugins and more.
1387 Warning, their API is really not stable yet.
1388 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1389 include/simgrid/kernel/routing/*Zone.hpp
1392 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1394 Build System and other cleanups:
1395 - Remove unused run-time parameter "tracing/onelink-only".
1396 - SimGrid now works with intel compilers (MC still buggy)
1397 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1398 - tesh kills subprocesses on timeout (but not on windows)
1401 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1404 - #194: Feature request: simgrid::s4u::Comm::test_any()
1405 - #245: migrating an actor does not migrate its execution
1406 - #253: Feature Request: expose clusters as objects
1407 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1408 - #255: Tesh broken on Windows
1409 - #256: Modernize FindSimGrid.cmake
1410 - #257: Fix (ab)use of CMake install
1412 ----------------------------------------------------------------------------
1414 SimGrid (3.18) Released December 24 2017
1416 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1418 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1419 - Support for MSG should not be removed from SimGrid before 2020,
1420 but future evolutions will be limited to the integration of
1421 user-provided patches.
1424 - s4u::onDeadlock() signal, run before stopping the simulation.
1425 - s4u::Exec: asynchronous executions (abstraction of background
1427 - s4u::Host->getLoad() returns the achieved speed in flops/s
1430 - Link::name() is deprecated and will be removed in v3.21, use
1431 Link::getCname() instead.
1432 - Mailbox::getName() changed to return a std::string, use
1433 Mailbox::getCname() instead to get a char*.
1434 - Storage::getName() changed to return a std::string, use
1435 Storage::getCname() instead to get a char*.
1436 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1437 to have both allocation and deallocation of the map in user space.
1440 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1441 The amount of flops remaining to do is a vector, not a scalar.
1442 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1443 name implies on both sequential and parallel tasks.
1444 - Both changes fix GitHub's #223 using PR #237 as a basis.
1445 Thanks Michael Mercier.
1446 - Most examples were converted to S4U and hidden elsewhere at tests.
1447 Rationale: we still want MSG to work; we want newcomers to use S4U.
1450 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1451 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1454 - Switch to the faster dlopen privatization mechanism by default
1455 - Documentation improvments
1458 - Live migration is getting moved to a plugin. Dirty page tracking is
1459 the first part of this plugin. This imply that VM migration is now
1460 only possible if one this function is called:
1461 - C/MSG: MSG_vm_live_migration_plugin_init()
1462 - C/C++: sg_vm_live_migration_plugin_init()
1463 - Java: Msg.liveMigrationInit()
1464 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1468 - Define class simgrid::xbt::Path to manage file names.
1469 - Removed unused functions:
1470 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1471 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1472 xbt_graph_export_graphviz()
1473 - xbt/str.h: xbt_str_join()
1474 - Remove unused datatypes:
1475 - xbt/heap.h: use std::priority_queue or boost::heap instead
1476 - xbt/swag.h: use boost::intrusive::list instead
1479 - New link_energy plugin for the consumption of the links.
1480 - All of the operations on files and storage contents have been
1481 packaged into a plugin (src/plugins/file_system). The current
1482 public interface can be found in
1483 include/simgrid/plugins/file_system.h
1484 To use these functions you now have to initialize the plugin by
1485 calling MSG_storage_file_system_init() just after calling
1486 MSG_init() or sg_storage_file_system_init() just after creating
1490 - Remove the undocumented/untested tag <include>
1493 - Remove viva specific tracing as the tool is no longer maintained
1496 - #248: Kill of finished processes leads to segfault
1497 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1498 - #239: please implement signal s4u::onDeadlock()
1499 - #230: segfaults when exit() without run()
1500 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1501 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1502 - #222: Actor::kill() doesn't really kill and segfaults
1503 - #221: odd LMM warning when killing an actor
1504 - #120: Memory leak when the processes are forcefully killed
1506 ----------------------------------------------------------------------------
1508 SimGrid (3.17) Released October 8 2017
1510 The Drained Leaks release: (almost) no known leaks despite the tests.
1512 Even very long simulations will preserve your memory: our testsuite
1513 stresses SimGrid for over 45mn on fast machines for a coverage of
1514 over 80%, and there is only one single known leak, of about 4kb.
1518 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1519 handler may be given to free resources if the comm cannot be completed.
1520 - this_actor::send(mailbox) is now mailbox->put()
1521 - New: simgrid::s4u::Comm::wait_all()
1522 - New: Host.execute() for remote executions.
1525 - Unused option network/sender-gap is removed.
1528 - Deprecate MSG_task_isend_with_matching(): This unused feature
1529 really complicates our internals. Will be removed in v3.20.
1532 - Improved context termination. It is now possible to free resources, even
1533 when a process is forcibly killed.
1534 - ContextBoost: add support for Boost versions above 1.61.
1537 - Bring back run-time option --cfg=exception/cutpath to remove exception
1539 - Removed unused functions:
1540 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1541 xbt_str_rtrim(), xbt_str_trim().
1542 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1545 - Removed header files obsolete since SimGrid 3.12:
1546 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1548 - #3: SD_exit should be made optional
1549 - #120: Memory leak when the processes are forcefully killed
1550 - #159: Threading test regression in Actor refcounting
1551 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1552 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1553 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1554 - #191: VM migration and pstate
1555 - #192: Updating the energy consumptions of all hosts crashes with VMs
1556 - #195: All actors have PID=0 in the logs
1557 - #204: Sometimes segfault with thread contexts and mmap privatization
1558 - #222: Actor::kill() doesn't really kill and segfaults
1559 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1561 ----------------------------------------------------------------------------
1563 SimGrid (3.16) Released June 22. 2017.
1565 The Blooming Spring Release: developments are budding.
1568 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1569 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1570 - Java JAR file should be finally fully working on Mac OSX too.
1572 XML platforms: Switch to platform v4.1 format.
1573 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1574 - <zone> can be used as a synonym for the now deprecated <as>
1575 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1576 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1577 - <actor> can be used as a synonym for the now deprecated <process>
1578 - state_file and avail_file periodicity is now easier to express
1579 (check the documentation) the old behavior should still work.
1580 * Storage: not backward compatible, but it seems that nobody ever used it
1581 - Remove attribute 'content_type' of <storage_type>: was never used
1582 - Make attribute 'model' of <storage_type> optional: for future usage
1583 - Remove Bconnection model property: was never (in)validated. Replaced by
1584 the maximum of the read and write bandwidth as a resource constraint.
1587 - New and Backwards Compatibility break:
1588 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1589 When one wants to get the list of tasks whose states have changed during a
1590 simulation round, s/he has to allocate and free a dynar and use it as
1591 argument to this function. The former SD_simulate (double how_long)
1595 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1596 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1599 - New function to update the consumption of all hosts at once.
1600 - Fix the model for multi-core hosts, linear in the amount of busy
1601 cores with an abnormality for pIdle (see paper).
1602 BUG: Still not satisfactory for ptask on mono-cores.
1605 - The netzone are now available from the MSG API.
1606 The old names still work, but are now deprecated.
1609 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1610 - New option: smpi/keep-temps to not cleanup temp files
1611 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1612 mallocs mode is used, allows to change the size of the fake file used
1613 (default 1MB), to potentially limit the number of mappings for large runs.
1614 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1615 - Fortran ifort and flang compilers support
1616 - New RMA calls supported (experimental):
1617 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1618 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1619 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1620 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1621 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1622 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1625 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1626 - Behind the scene, ActivityImpl are now automatically refcounted.
1627 Removing this nasty bottleneck will greatly simplify our internals.
1630 - Replay: New function xbt_replay_action_get():
1631 Retrieve the function previously associated to an event type.
1632 - DROPPED FUNCTION: xbt_str_varsubst()
1633 - DROPPED MODULE: strbuff. We don't need it anymore.
1634 - DROPPED MODULE: matrix. We don't need it anymore.
1635 - DROPPED MODULE: lib. We don't need it anymore.
1637 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1639 SimGrid (3.15) Released March 22. 2017
1641 The Spring Release: continuous integration servers become green
1642 We fixed even the transient bugs on all target architectures:
1643 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1644 Windows, FreeBSD, NetBSD.
1646 - Rename NetCards to NetPoints.
1647 This was intended to help NS3 users, but that's not a netcard.
1648 That's a point in the routing algorithm, let's avoid wrong simplifications.
1651 - New: MSG_process_yield(). Stop and yield to other processes.
1652 - New: MSG_process_daemon(). Daemon processes are automatically killed
1653 when the last non-daemon process terminates
1654 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1655 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1656 to make room for the upcoming network energy plugin.
1657 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1660 - Ensure that an actor can kill itself with Process::exit()
1661 - Kill the obscure NativeException. Nobody want to survive the issues
1662 it denotes, so use JniException that is a RuntimeException (not to
1663 be caught explicitly).
1664 - Partial bug fix in initialization. SimGrid flags on command line were
1665 consumed at C level but stayed in the original Java String[] args.
1666 This could mess users' args[i] if SG flags were not put at the end of
1668 The SimGrid flags are now removed from the Java arguments. However,
1669 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1670 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1671 It might be if you have --log or --cfg flags in the command line.
1672 - Fix numerous memleaks all around the place. In particular, around VMs.
1676 - s4u::onPlatformCreated: right before the simulation starts
1677 - s4u::onSimulationEnd: right after the main simulation loop
1678 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1679 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1680 event from the availability_file changes the avail speed.
1681 - Links are now usable from s4u
1682 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1683 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1684 - New examples: The conversion of MSG examples to S4U has begun
1685 - Actors: create, kill, migration, and suspend
1686 - Applications: master-worker and token-ring
1687 - Action replay: communications and storage
1688 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1691 - Backwards Compatibility breaks
1692 - The SD_route_get_list and SD_route_get_size functions have been removed.
1693 They are replaced by the unique
1694 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1695 The route from src to dst is built in the links dynar whose size can
1696 be obtained with xbt_dynar_length.
1697 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1698 sg_host_route_bandwidth and sg_host_route_latency.
1699 Macros ensure the backwards compatibility, but you should fix your code
1702 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1703 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1704 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1705 We don't support 100% of it, but it should be enough. Contact us if not.
1706 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1707 naive and probably inaccurate)
1708 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1709 mapped onto a unique small file using some system magic.
1710 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1711 - Fortran: cleanups, fixes, support of user-added operations
1712 - RMA: MPI_Accumulate are applied in correct order.
1713 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1714 - Better support for MPI_IN_PLACE
1715 - Support for MPI_Win attrs and keyvals.
1716 - Support MPI_Comm_group_create, MPI_Type_size_x
1717 - MPI ops should be properly applied to all allowed types
1718 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1719 algorithm by a non failing one from mpich.
1722 - Kill the fifo data container: we don't use it anymore.
1725 - A <cluster> can now be created with different speed values to represent pstates.
1726 see examples/platforms/energy_cluster.xml for an example.
1728 SimGrid (3.14.159) Released December 28. 2016
1730 The previous release was only a crude approximation of a Pi release;
1731 this one should be a bit better.
1733 - Revert a last minute change that broke on Mac OSX.
1734 - Fix the changelog and NEWS files.
1735 - Improve the documentation, in particular of the routing module.
1736 - Integrate some patches flying around in Debian.
1738 SimGrid (3.14) Released December 24. 2016
1741 * The whole documentation was reworked and reorganized. There is
1742 still much room for improvement here, but we're on it.
1745 * We now need python3 for our testing infrastructure.
1746 * Model checking needs libevent
1749 * Now works on FreeBSD
1752 * Privatization now works on FreeBSD
1754 * Privatization is now activated by default in smpirun
1755 You can turn it off with -no-privatize if something goes wrong.
1757 * Call-location tracing for SMPI.
1758 You can add the exact location (filename / linenumber) of an MPI call to
1759 your trace files and slow down or speed up the simulation between two
1760 consecutive calls by using an adjustment file (see the documentation).
1762 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1763 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1764 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1765 values were ignored in some cases. The timings of these functions can now
1766 be significantly different.
1768 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1769 smpi/running-power is renamed to smpi/host-speed
1771 * smpi/grow-injected-times option to enable or disable multiplication of the
1772 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1773 make simulation less precise (but also much faster).
1775 * smpirun script should be (much) faster for large deployments.
1777 * SMPI tracing: fixed issue with poor matching of send/receives.
1779 * Replay: Fix broken waitall
1781 New functions and features
1782 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1784 Dropped / renamed functions and features
1785 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1786 - MSG_mailbox_is_empty() -> Mailbox::empty()
1787 - MSG_mailbox_front() -> Mailbox::front()
1788 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1789 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1790 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1791 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1792 * Don't pass the free_f parameter to property related functions:
1793 - MSG_host_set_property_value()
1794 - MSG_as_router_set_property_value()
1795 - MSG_storage_set_property_value()
1796 * VM properties. Since msg_vm_t are msg_host_t, just use
1797 MSG_host_get_property_value() and friends
1798 * VM I/O related things:
1799 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1800 - Unimplemented save/restore methods
1801 * MSG_as_router_get_property_value() was redundant with
1802 MSG_environment_as_get_property_value().
1803 - Removed MSG_as_router_*propert*() functions
1804 - Added MSG_environment_as_set_property_value() for consistency
1805 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1806 Well, they are still there for now, but deprecated with a warning.
1807 Please switch to xbt_dict_new_homogeneous() before this is removed
1809 * Task affinity. Its intended behavior (that was very badly tested
1810 and probably not really working) was deceiving what most users
1811 would have hoped here.
1812 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1813 * Option network/coordinates is now useless and should be dropped.
1816 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1819 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1820 * Added Dragonfly topology support
1823 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1824 EiBps, EBps, Eibps, Ebps for bandwidth.
1825 They may become useful to some lucky ones.
1828 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1829 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1830 * ex/async_waitAll: new example, on asynchronous communications
1833 * Memory usage should be decreased for simulations with a large number
1834 of processes. This also helps for SMPI.
1836 SimGrid (3.13) stable; urgency=low
1838 The Half Release, a.k.a. the Zealous Easter Trim.
1840 - We removed half of the lines, that were mostly experimental cruft.
1841 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1842 (not counting blanks and comments -- according to openhub.net)
1843 - The internals are now compiled in C++ (and will soon be clean C++)
1844 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1845 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1847 * Backwards Compatibility breaks
1848 - Removed Lua simulation bindings (switch to C or Java for that).
1849 Lua can still be used to describe platforms
1850 - Removed Java kernel plug-ins.
1851 Will be reintroduced after the ongoing major internals reorg.
1853 - the following functions were removed.
1854 They were too specific and should be reimplemented in a generic
1855 way, with filter function.
1856 - MSG_task_listen_from_host
1857 - MSG_mailbox_get_count_host_waiting_tasks
1858 - MSG_mailbox_put_with_timeout was removed.
1859 Please use MSG_task_send_with_timeout instead.
1861 - the SD_application_reinit function was removed. It has been a noop for a while.
1862 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1863 replaced by a more flexible mechanism.
1864 - The following functions thus do not exist anymore
1865 - SD_workstation_get_access_mode
1866 - SD_workstation_set_access_mode
1867 - SD_workstation_get_current_task
1868 - Basic estimation functions have been removed but can easily be replaced
1869 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1870 - SD_workstation_get_computation_time => amount / sg_host_speed()
1872 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1873 Use VM.getSpeed()*load/100 for the legacy behavior.
1875 - option enable_tracing was removed. It was not doing anything for a while.
1876 - In the ModelChecker:
1877 - the model-checker now ptraces the model-checked process which means
1878 you cannot use a debugger on the latter anymore (we might make this
1879 optional in the feature);
1880 - removed soft-dirty page tracking;
1881 - remove model-checked side snapshot management,
1882 MC_snapshot() and MC_compare_snapshot();
1883 - keep the MC_cut() function as a stub (it was not really working
1884 in the previous release).
1887 * All options are consistently in kebab-case. Old names are kept as alias.
1890 * Switch to platform v4 format.
1891 - Rename from 'power' to 'speed' the attributes describing the amount of
1892 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1893 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1894 - In <host> and <link>, attributes availability and state are gone.
1895 It was redundant with state and availability traces, and with peak values.
1896 - In <cluster>, attributes availability_file and state_file are gone.
1897 It was too complex and unused.
1898 - Kill <gpu>. Was not doing anything.
1899 - The DOCTYPE points to the right URL:
1900 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1901 (the file at this address now documents the changelog since its v1)
1902 - A warning is emitted for unit-less values (they are still accepted).
1903 - speed. Default: 'f' or 'flops'. Also defined:
1904 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1905 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1906 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1907 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1908 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1909 - latency. Default: 's' second. Also defined:
1910 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1912 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1913 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1916 * s4u::Host is now the preferred public interface to the Host features.
1917 sg_host_* functions are C bindings to the exact same behavior
1918 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1921 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1922 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1923 They were deprecated since a few years, and probably did not even compile anymore.
1926 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1927 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1928 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1929 encouraged to update their codes. A list of the modified functions can be found at the end of
1930 include/simgrid/simdag.h
1933 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1934 the lambda or closure passed as a parameter will run in kernel mode.
1935 All the callback functions should be rewritten to that interface at some point.
1938 * Reorganizing and cleaning the internals all around the place.
1941 * Remove old default barrier/bcast buggy algorithms (see #18407)
1942 * Various bug fixes to handle more codes
1943 * Remove the need for the --foreground option of smpirun (it is still
1944 accepted for backward compatibility).
1947 * Kill the setset data container: MC don't use it anymore.
1948 * Kill the queue data container: it made more sense with GRAS.
1949 * Kill the xbt_peer_t data type: it's useless without GRAS.
1950 * Kill rm_cb feature of config sets: it was never useful.
1951 * Kill graphxml parsing feature. It was not used.
1952 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1954 - xbt_dynar_sort_strings(), when the content is char*
1955 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1956 They throw exceptions on invalid input;
1957 * C++ support for declaring CLI flags (simgrid::config::Flag);
1958 * class for abstracting different signal backends (simgrid::xbt::signal).
1959 with no external dependencies (we need very simple signals).
1962 * refactoring and cleanup of the code;
1963 * ongoing process to cleanly separate the model-checking algorithms
1964 from the code model-checking support.
1966 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1968 SimGrid (3.12) stable; urgency=low
1970 The Facelift Release.
1973 * Require g++ v4.7 at least to not speak prehistorical C++.
1974 * Require Boost 1.48 (for signal2 component).
1975 * Java must be version 7 at least when activated.
1976 * Builds on Windows again (including Java bindings).
1977 * Tracing is now always enabled (no way to turn it out)
1978 * Remove GTNetS. It was not working anyway.
1979 * Various cleanups in the cmake scripts.
1980 * Move headers around to sort them out on installed systems:
1981 - instr/instr.h -> simgrid/instr.h
1982 - instr/jedule/* -> simgrid/jedule
1983 - simdag/datatypes.h was removed
1984 - simdag/simdag.h -> simgrid/simdag.h
1985 - msg/datatypes.h was removed
1986 - msg/msg.h -> simgrid/msg.h
1989 * Interface improvement:
1990 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1991 - Sanitize the interface in MSG_task_ module:
1992 - Merge two functions that were close enough but misleading:
1993 set_compute_duration(t) -> set_flops_amount(t)
1994 get_remaining_computation(t) -> get_flops_amount(t)
1995 - set_data_size(t) -> set_bytes_amount(t)
1996 get_data_size(t) -> get_bytes_amount(t)
1997 - Massive cleanups in the functions related to the energy
1998 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1999 - New: MSG_host_get_pstate()
2001 - msg/energy/onoff: switching hosts on and off
2003 * Interface improvement:
2004 - Rename Host.isAvail() to Host.isOn()
2005 - Rename Process.currentProcess() to Process.getCurrentProcess()
2006 - Rename Task.setDataSize() to Task.setBytesAmount()
2007 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
2009 - #18874: Actually allows the GC to reclaim tasks
2013 - SIMIX_process_throw: raises an exception in a remote process
2014 * Refactoring: Separate sub-modules
2015 - libsmx: the public interface, as libc in a real system
2016 - popping: the strange dance that converts a user request into a kernel handling
2017 - smx_context_*: the virtualization mechanisms that embed the user code
2018 - smx_*: the handling of each simcalls
2019 * Interface cleanup:
2020 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
2021 * Rename smx_action_t into smx_synchro_t, making explicit that these
2022 things are used to synchronize processes with their environment.
2023 For example, a communication is a sort of synchronization involving
2024 the communicating processes (that may block until the exchange) and
2025 the platform. The same can be said from computations, etc.
2027 - #18888: segfault when a process ends before its kill_time
2031 - 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.
2032 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
2033 - MPI_Type_set_name, MPI_Type_get_name
2034 - MPI_*_c2f and MPI_*_f2c functions
2035 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
2036 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
2037 - Activate a lot of new tests from the mpich 3 testsuite
2039 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
2040 - InfiniBand network model added: Based on the works of Jerome Vienne
2041 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
2042 - When smpi/display_timing is set, also display global simulation time and application times
2043 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
2044 * Collective communications
2045 - 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
2046 - 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/).
2047 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
2049 - Replay now uses algorithms from wanted collective selector
2050 - Replay can be used with SMP-aware algorithms
2051 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
2052 - 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.
2054 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
2055 - Correctly use loopback on fat-tree clusters
2056 - Asynchronous small messages shouldn't trigger deadlocks anymore
2057 * Energy/DVFS cleanup and improvement
2058 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
2059 - new: smpi_get_host_pstate()
2062 - "Full" network optimization flag was broken since Surf++
2063 - Better handling of precision flags in maxmin
2064 - Fix bug causing sometimes "Impossible" errors
2065 - Properly pass cluster properties to included hosts
2066 * Improvement of the Energy plugin.
2067 - Always update the consumption before returning that value
2068 - New property: watt_off to denote the dissipation when the host is off
2069 - New functions getWattMinAt and getWattMaxAt to retrieve the
2070 dissipation of pstates that we are not currently at.
2071 * Java: class NetworkLink renamed to Link
2072 * New function: simcall_process_get_kill_time()
2073 * Massive rename s/workstation/host/
2074 - That's intrusive, but that's good for the project consistency. Sorry.
2075 - Change config option "workstation/model" into "host/model"
2078 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
2079 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
2080 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
2081 - xbt_dirname and xbt_basename for non-POSIX systems
2083 * The model checker now runs as a separate process.
2084 * The model checker runs must now be launched with the new simgrid-mc program.
2085 * Record/Replay: the MC can display a textual representation of a path in the
2086 execution graph. It can then be replayed outside of the model checker.
2088 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2090 SimGrid (3.11) stable; urgency=low
2095 * Normalizing pointers addresses tool for better diff between logs
2098 * Add cloud examples using new VMs
2099 - examples/msg/cloud/two_tasks_vm.tesh
2100 - examples/msg/cloud/simple_vm.tesh
2101 - examples/java/cloud/cloud.tesh
2102 - examples/java/cloud/migration/migration.tesh
2103 * Add java surf examples:
2104 - examples/java/surfPlugin/surf_plugin.tesh
2105 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
2106 - examples/java/surfCpuModel/surf_cpu_model.tesh
2107 * Add SMPI+MSG example:
2108 - examples/smpi/smpi_msg_masterslave/
2113 - msg task destroy cancel test
2114 - msg_host on/off test
2115 * Move all tests in testsuite to teshsuite (adding tesh files)
2116 * Restructure teshsuites
2117 - one folder for each kind of test
2118 * Restructure AddTests.cmake
2120 - structure the order of tests (with sections)
2123 * Add virtual machine
2124 - creation of a VM on a PM
2125 - migration of a VM from a PM to another PM
2127 - MSG_process_join(msg_process_t process, double timeout)
2128 - msg_bar_t MSG_barrier_init(unsigned int count)
2129 - int MSG_barrier_wait(msg_bar_t barrier)
2130 - void MSG_barrier_destroy(msg_bar_t barrier)
2131 - msg_as_t MSG_environment_as_get_by_name(const char * name)
2132 * New option "msg/debug_multiple_use" to help debugging when a task is used
2135 - Improvements and finalization of MSG_storage, MSG_file APIs and their
2137 - Increase code coverage in test suites
2141 * Protect context stacks against stack overflow. The number of protected memory
2142 pages allocated on the top of each stack (1 by default) can be configured
2143 with the new command line option --cfg=contexts/guard_size:<value>.
2144 * Simcalls are now generated by a python script that
2145 - generates files included by SimGrid
2146 - checks that all the functions exist, and proposes prototypes
2148 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
2149 host_set_data, host_get_data
2151 - simcall_process_join(smx_process_t process, double timeout)
2152 * Fix bug where sleeping processing could not be suspended.
2155 * Translate surf models from C to C++
2156 - Generic classes for all models: Model, Resource, Action
2157 - A generic interface for each kind of model (CPU, Network, Storage
2158 Workstation, WorkstationVM)
2160 * Translate surf routings from C to C++
2161 * Add callbacks using sigc++ or boost::signals2
2162 - Add callback functions for resource creation/destruction
2163 - Add callback functions for action state change
2164 - Handle Energy as a plugin
2165 * Replace swag by boost::intrusive
2166 * Add new routing models for clusters. For documentation, see
2167 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
2168 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
2169 parameters for cluster tag
2170 - Fat trees, with topology="FAT_TREE" and
2171 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
2172 - see examples/platforms/torus_cluster.xml and
2173 examples/platforms/fat_tree_cluster.xml
2174 * More documentation
2177 * Hostfiles support host:nb_processes construct to deploy several processes on
2179 * Collective communication algorithms should not crash if used with
2180 improper number of nodes and report the error.
2181 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
2182 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
2183 MPI_Dims_create, MPI_Cart_sub are supported.
2184 * New interface to use SMPI programmatically (still depends on MSG for
2185 some parts, see examples/smpi/smpi_msg_masterslave):
2186 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
2190 * Global variables privatization in MPI executables is now performed at runtime
2191 with the option smpi/privatize_global_variables (default:no).
2192 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
2193 dynamic libraries loaded by the application are not privatized (static
2194 linking with these libraries is advised in this case)
2197 * Options defined in XML work correctly now.
2200 * New cmake option, enable_lib_in_jar, to control whether native libraries are
2201 copied into simgrid.jar or not (ON by default). Use this option if you want
2202 to reduce the size of the installed simgrid.jar, *and* the native libraries
2203 are kept installed elsewhere.
2204 * Surf binding with SWIG (code generated in maintainer mode only):
2205 - plugin to handle callbacks
2206 - CPU model only for the moment
2209 * Supernovae build mode is definitively removed. It was used to improve
2210 inlining and inter-module optimizations. It is nowadays superseded by
2211 link-time optimizations commonly available in compilers.
2212 * Update ns-3 find lib. Bindings for ns-3 should work again now.
2213 * Add boost dependency for surf++
2214 * Add new macro for tests
2215 - ADD_TESH(name <tesh_args>)
2216 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
2220 - xbt_bar_t XBT_barrier_init(unsigned int count)
2221 - int XBT_barrier_wait(xbt_bar_t barrier)
2222 - void XBT_barrier_destroy(xbt_bar_t barrier)
2223 * Make the xbt_os_time module public
2225 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2227 SimGrid (3.10) stable; urgency=low
2229 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
2232 * Reintegrate Java to the main archive as desynchronizing these
2233 package is not acceptable anymore (Java is now considered stable)
2234 * Add explicit synchronization facilities through semaphores
2235 * Bug fix: Task.setDataSize() only changed the C world, not the value
2236 cached in the Java world
2239 * Dramatically change the way files are handled. API and internals changed, but
2240 this part of MSG was not considered as production grade either.
2241 * Add explicit synchronization facilities through semaphores
2242 * Add a new function MSG_host_get_process_list()
2243 * Preliminary DVFS support (see examples/msg/energy/ for details)
2246 * SMPI is now included directly in the libsimgrid as the windows
2247 linker doesn't force us on splitting it anymore.
2248 * Improvements of the SMPI replay tool:
2249 - Most of the collective communications are now rooted in the same process as
2250 in the original application.
2251 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
2252 used until now). Multiple data types can now be used in a trace.
2253 - The replay tool now supports traces produce either by TAU or a modified
2255 - Bug Fix: the compute part of the reduce action is now taken into account.
2256 - Gatherv collective is now supported
2257 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
2258 of smpirun outputs time independent traces for the current run. One file
2259 is created per process. If too many processes are simulated, this behavior
2260 can be changed to one file for all processes by using the
2261 tracing/smpi/format/ti_one_file flag
2262 * smpirun generates the host file if needed (with given host count and platform)
2263 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
2264 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
2265 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
2266 --cfg=smpi/coll_selector:(mpich/ompi)
2267 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
2269 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
2270 happening inside a collective SMPI call.
2271 * Fix the behavior of complex data types handling.
2272 * Make MPI_Wtime another synchronization point to take computations into
2274 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
2275 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
2276 * Add all missing Fortran bindings, SMPI should work with Fortran 90
2277 (no privatization of global variables yet)
2278 * Preliminary DVFS support (see examples/smpi/energy/ for details)
2281 * Verification of liveness properties is now available for SMPI applications
2282 (in addition to MSG applications)
2283 * Bugged examples using SMPI in examples/smpi/mc/
2284 * Add --cfg=model-check/visited option. Allows the verification of infinite
2285 programs. Detection of loops in the execution thanks to the system state
2286 comparison and reduction of the state space to explore. Can be combined with
2287 DPOR for safety properties.
2290 * Allow to change SimGrid configuration (see --help) within the code
2291 thanks to SD_config() as it can be done in MSG.
2292 * Add a new function SD_task_set_amount() upon user request.
2295 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
2296 * Remove rule based routing (no more PCRE dependency)
2297 * Add a limiter_link option to cluster tag, to specify a maximum reachable
2298 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
2299 * Add a loopback_bw and loopback_lat options to cluster tag.
2300 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
2301 routing. Now peers are encapsulated in an AS and have their own private
2302 router but this is transparent.
2305 * Our own implementation of getline is renamed xbt_getline, and gets
2306 used even if the OS provide a getline(). This should reduce the
2307 configuration complexity by using the same code on all platforms.
2308 * New type: xbt_cfg_elm_boolean.
2309 * Allow to use yes/no for boolean configuration options in the command line.
2310 * Allow to disable SimGrid cleanups at exit from command line option.
2311 There are situations where one may want a simulation to end with an exit.
2312 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
2313 annoying when scripting around the simulator. Adding a
2314 --cfg=clean_atexit:no allows to circumvent this issue.
2317 * Lots of memory leaks were corrected in this release.
2318 There is no known memory leaks anymore, in all of our 600+ tests.
2319 * New command line option --version, to get SimGrid version information.
2320 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
2322 * Supernovae builds are deprecated, and expected to be removed in the next
2325 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2327 SimGrid (3.9) stable; urgency=low
2329 The Grasgory release: GRAS is really dead now.
2331 * Complete overhaul of the internal host structures scheme.
2334 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
2335 was considered as experimental and badly maintained since then.
2336 * Keeping it was thus a trap to our potential users, that could take
2337 it instead of MSG or SMPI by mistake despite is pity state.
2338 * GRAS seems to have very few users (if any), and no one volunteered
2339 to maintain it further. It also induces a lot of XBT code (for
2340 portability sake), that must be maintained too.
2341 * For all these reasons, we killed GRAS. If someone wants to revive it
2342 in the future, don't cry, our git history still remembers of GRAS.
2345 * Major overhaul. Merge our documentation again as time proved that
2346 splitting it was really not helping our users.
2347 * Further improve the developer documentation to help newcomers
2348 hacking on SimGrid itself. The user documentation (and in
2349 particular, the beginner documentation) is still in a sorry state.
2352 * Now works on Windows too!
2353 * Much more extensive test suite, from MPICH
2356 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
2357 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
2358 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
2359 * Bug fix: let task be scheduled when the last dependency to be solved is
2360 a control dependency.
2361 * Remove SD_load_environment_script function.
2362 Use the C sg_platf function if you want to declare a platform
2366 * New function: MSG_process_get_number()
2367 * Old function documented: MSG_config()
2368 * Remove MSG_load_platform_script function
2369 Use the C sg_platf function if you want to declare a platform
2373 * Change the default value of the TCP_gamma constant (maximal size of TCP
2374 congestion window) to a more realistic 4MiB value. If you notice changes in
2375 your simulation results, you can fall back to the previous 20k tiny window
2376 by adding --cfg=network/TCP_gamma:20000 on command line.
2377 * (Hopefully) fix a bug wrt periodic availability and state traces
2378 * Bug fix: use default values at start when first event in availability/state
2379 trace is not at time 0.
2382 * remove the "new_" part of function name sg_platf_new_trace_connect
2383 (resulting in sg_platf_trace_connect), since it does not create
2387 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2388 use case where it's really mandatory, and maintaining it was a pain in
2390 * New: xbt_fifo_search(), search an item with a user-provided
2391 comparison function instead of dumb pointer comparison.
2394 * Fix the lua deployment:
2395 Use `simgrid.init_application()` before deployment instead of
2396 `simgrid.msg_register_application()` after.
2399 * Transfer the tracing files into the corresponding modules.
2401 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2403 SimGrid (3.8.1) stable; urgency=low
2405 The "we are told that some people want to also *install* the simgrid
2408 * Add missing file "tesh.1" to the archive.
2410 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2412 SimGrid (3.8) stable; urgency=low
2414 The Psssshiiiit release: SimGrid jumps into the Cloud.
2417 * Add an experimental interface to manipulate VMs. They are mainly
2418 process groups with very few intrinsic semantic, but they should
2419 allow you to build the semantic you want easily.
2420 * New function: MSG_host_set_property_value()
2421 * New function: MSG_process_on_exit(). To clean memory in all cases.
2422 * Bug fixes that made the host (and link) failures unusable.
2423 * Add a way to auto-restart process when the host in which they are
2424 executing comes back (ON_FAILURE="RESTART" on deployment file,
2425 MSG_process_auto_restart_set).
2426 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2427 please stop using the old ones, they are DEPRECATED.
2429 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2430 Please use MSG_init() instead. (reducing the amount of entry
2431 points in the library helps us).
2432 * Make it impossible to link against the wrong version of the lib
2433 * Deprecate MSG_clean(). No need to call it anymore.
2434 * Function MSG_get_host_number() is not deprecated anymore.
2437 * Split the doc into a user guide and a reference guide.
2438 * Start a developper guide to help people hacking on SimGrid.
2441 * Enable tracing by default. This modules rocks you should use it.
2442 * Remove option custom_flags. Now use environment variables CFLAGS
2444 * Use default cmake things to detect lua instead of home grown ones.
2445 * New option "enable_mallocators" to disable mallocators, for debugging
2446 purpose ("on" by default).
2449 * Bug fixes around the resource failures: don't let the processes
2450 survive the host they are running onto.
2451 * Add an interface to auto-restart processes when the host in which they are
2452 executing comes back.
2453 * Ensures that SIMIX_clean is called automatically. It's not part of
2454 the public interface anymore (bindings should be updated).
2457 * Bug fix for when SD_Simulate is called with a positive value: be careful
2458 when comparing doubles. Sometimes they are different for non significant
2460 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2461 parallel task whose initial work is distributed among host according
2462 to the Amdahl's law. Such tasks are created with a parameter alpha
2463 that corresponds to the non-parallelizable part of the computation.
2464 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2465 between two sets of workstations assuming a 1D block distribution (each
2466 workstation owns a similar share of data) on both sides.
2468 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2469 Data redistribution will be automatically scheduled once parent and child
2470 are both scheduled. The filling of computation_amount and
2471 communication_amount structures is now done seamlessly thanks to the chosen
2473 * New function SD_workstation_dump to display various information
2474 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2475 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2476 bandwidth on the route onto which the task is scheduled and the amount of
2478 To divide the nominal bandwidth by 2, the rate then has to be:
2479 rate = bandwidth/(2*amount)
2480 * Compute tasks that have failed can now be rescheduled and executed again
2481 (from their beginning)
2482 * Increasing source code coverage (src/simdag is now covered at 95.8%
2486 * Re-implement time-independent trace replay using SMPI (at the
2487 smpi_smp_* level) instead of MSG. This should replace
2488 examples/msg/actions/actions.c
2489 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2490 hindexed and structs)
2491 * Implement the exchange of non-contiguous data.
2492 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2493 * Correct behavior of smpi/sender_gap and set its default value to 0
2494 * Add option to asynchronously send small messages to allow better
2495 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2496 specifies the size in bytes under which messages will be asynchronously sent.
2497 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2498 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2501 * New C interface to define a platform: XML is now optional.
2502 For more info, please check include/simgrid/platf.h.
2503 * New interface to define random platforms from the C:
2504 For more info, please check include/simgrid/platf_generator.h and
2505 examples/msg/masterslave/masterslave_platfgen.c
2506 * Export a sg_cmdline dynar containing all the arguments we got from
2510 * Two new tracing options for adding comments to trace file so you
2511 can track your experiments (see --help-tracing for details).
2512 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2513 * Adding the SimGrid version that generated the trace file as a comment.
2514 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2515 * Fix to avoid key clashes on Paje links
2516 * Other minor fixes related to the Paje specification
2519 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2520 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2521 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2522 * Greatly improve the robustness of mmalloc to user errors (such as
2523 using an area after freeing it, or freeing it twice)
2525 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2527 SimGrid-java (3.8.1) stable; urgency=low
2529 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2530 Warning: all methods are visible, but only some of them are
2531 implemented so far. Check the source (src/jtrace.c)
2532 for further information.
2533 * New module: org.simgrid.msg.File (SimGrid File management functions)
2534 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2535 * Change the meaning of Process.restart: now restart the process from
2536 the beginning, like MSG_process_restart in C.
2537 * Add Process.setAutoRestart: handling of process restart when failed
2539 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2540 you to retrieve the properties of the processes/hosts
2541 * Deprecate Msg.clean(): you can just forget about it now.
2542 * New function Process.getCount(), that only works when compiling
2543 with the not yet released version 3.9 of the C library.
2545 * New context factory based on Coroutines. It mandates a modified JVM
2546 but then, the simulations run about five times faster, and there is
2547 no limit to the amount of processes (beside of the available memory).
2549 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2551 SimGrid (3.7.1) stable; urgency=low
2554 * Restore the prototype of MSG_process_create_with_environment() to
2555 the pre-3.7 situation by removing the kill_time argument.
2556 * Add a MSG_process_set_kill_time() function instead.
2559 * Fix weird behaviors when dealing with parallel tasks.
2562 * Simgrid is now built as a dll.
2563 * Simgrid-java now works on Windows.
2564 * Simgrid-Java is now included into Windows package.
2567 * First pre-build package for MacOSX.
2570 * Fix compilation when using MSG_USE_DEPRECATED.
2571 * Fix some compilation issues on Macs and Windows.
2572 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2573 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2575 We discovered that the Lua console is broken, but we are missing the
2576 manpower to fix it right now. The problem existed in 3.7 too, so we
2577 are not blocking the release for that. Sorry if you depended on this
2578 feature, any help would be really welcome.
2580 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2583 SimGrid-java (3.7.1) stable; urgency=low
2585 The "Java aint got to be bloated and slow" release
2588 * Various internal cleanups and performance improvement
2589 Simulations are expected to run up to twice faster or so
2590 * Make Process.kill(process) an instance method, not a static one
2591 * User processes are not java.lang.Thread subclasses.
2592 This breaks the compatibility (sorry), but previous API was
2593 brain-dead, making it impossible to have non-trivial
2594 initializations in the process constructor.
2595 * Require a full constructor per Process sub-class.
2596 Kinda breaks the compatibility (sorry), but this allows a much more
2597 efficient way to launch the processes at simulation startup.
2598 * Do not embeed our version of semaphores, java 1.5 can be considered
2599 as sufficiently prevalent for us to not dupplicate its features.
2603 * Add examples for almost every part of the API
2604 We spotted and fixed a lot of bugs in the process
2606 * New module: asynchronous communication API
2607 * New function: Process.sleep()
2608 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2609 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2610 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2612 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2614 SimGrid (3.7) stable; urgency=low
2616 The "spring cleanups (before the next Big Project kicks in)" release.
2619 * We can specify the SMPI latency/bandwidth factor with command line
2620 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2621 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2622 You can also use the "config tag" from platform file by adding this line
2623 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2624 (see "example/platforms/tag_config.xml" to use "config tag").
2625 Note that the command line supersedes the platform file configuration.
2626 * Change the correction factors used in LMM model, according to
2627 the latest experiments described in INRIA RR-7821.
2628 Accuracy should be improved this way.
2629 * Use the partial invalidation optimization by default for the
2630 network too. Should produce the exact same results, only faster.
2631 * Major cleanup in surf to merge models and split some optimization
2632 mechanisms from the core of the models. As a result you can now
2633 specify which model to use (e.g., --cfg=network/model:LV08
2634 --cfg=cpu/model:Cas01) and which optimization mode to use
2635 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2636 Incompatible combinations should err at initialization. See
2637 --help-models for the list of all models and optimization modes.
2638 * The CLM03 workstation model was dropped for simplicity because it
2639 used the deprecated CM02 network model. Use default instead.
2640 * Rename the TCP_gamma configuration option to network/TCP_gamma
2641 * Rename the coordinates configuration option to
2642 network/coordinates, and document it
2643 * Use now crosstraffic keyword instead of the terribly misleading
2644 fullduplex keyword. It is activated by default now in the current
2645 default model, use --cfg=network/crosstraffic:0 to turn it off.
2646 * Ongoing refactoring the model parsing to make XML files optional
2647 See include/simgrid/platf.h for details (still to be completed)
2650 * Major overhaul of the documentation. Almost instructive now :/
2651 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2652 functions and friends. This interface was considered as
2653 deprecated since over 2 years, it's time to inform our users that it is.
2654 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2655 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2656 use these (crappy) functions in your code.
2657 These functions will be removed soon. Stop using them now.
2658 * Deprecate MSG_get_host_{table,number}
2659 Implement MSG_hosts_as_dynar() instead.
2660 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2661 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2664 * Stabilize the parallel execution mode of user contexts
2665 * Introduce configuration variables to control parallel execution:
2666 - contexts/synchro: Synchronization mode to use when running
2667 contexts in parallel (either futex, posix or busy_wait)
2668 - contexts/parallel_threshold: Minimal number of user contexts
2669 that must be part of a scheduling round to switch to parallel
2670 execution mode (raw contexts only)
2671 * Fix bugs that prevented to use suspend/resume along with
2672 synchronization structures.
2673 * Fix bugs in process termination that lead to invalid memory access
2674 in very specific conditions.
2677 * Introduce a parallel mode for the models (controlled by surf/nthreads
2678 configuration item). In our tests, running the models in parallel
2679 never lead to any speedups because they are so fast that the gain
2680 of computing each model in parallel does not amortizes the
2681 synchronization costs, even when ultra fast futexes are used.
2682 This is released anyway because YMMV.
2685 * Performance boost by using a swag internally to compute the set of
2686 tasks that are finished and should constitute the return value of
2690 * Enable it by default now that it is considered rather stable.
2693 * Documentation of the tracing functions.
2694 * Performance gains when tracing categorized/uncategorized resource
2695 utilization by avoiding calls to get route when updating resource
2696 variables. LMM constraints are being used instead.
2697 * API changed to set task categories. Use MSG_task_set_category instead
2698 of TRACE_msg_set_task_category, and SD_task_set_category instead
2699 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2700 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2701 * Examples for link user variables added (see at examples/msg/tracing/)
2702 * Deprecated function TRACE_msg_set_process_category completely removed
2703 * Trace header updated according to the latest Paje file format
2704 * Tracing network lazy updates, no longer obligate users to use full updates
2705 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2706 * Experimental: let user code declare/set/push/pop application states for hosts
2707 * API changed to allow the manual creation of graph configuration files
2708 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2711 * Improve the API of Lua MSG bindings, using the Lua spirit.
2712 * Each simulated process now lives in its own Lua world (globals are
2713 automatically duplicated). It helps writing simulators. Will allow
2714 to run Splay programs within SimGrid in the future.
2715 * Add a Chord example in Lua, equivalent to the MSG one.
2718 * Start the implementation of a solution to express temporal
2719 properties, not only local assertions. This is still an
2720 experimental work in progress, stay clear from it to be safe.
2724 - Add new runtime parameters --help-logs and --help-log-categories
2725 to display information about supported logging parameters and
2727 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2728 don't exists anymore.
2729 * Mallocators: allow value NULL for the reset function.
2731 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2732 create homogeneous dictionaries, where all the elements share the
2733 same free function. Non homogeneous dictionaries will be
2734 deprecated in the next release.
2735 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2736 - Multi-level dictionaries are deprecated.
2738 - new function xbt_dynar_search_or_negative() that is useful when
2739 you have less than 2 million elements in your dynar and don't
2740 want of the extra complexity of catching exceptions when the
2741 element is not found.
2743 - Make xbt_os_thread module (for thread portability) public.
2744 Documentation is still to come, sorry.
2746 - Cleanups and simplifications to make it maintainable again.
2747 - Exotic features (such as memalign and valloc) were removed.
2748 - The metadata were extended and improved so that the
2749 model-checker becomes able to explore and inspect the heaps.
2750 - This may induce a performance drop when enable_model-checking is
2751 ON in cmake (even if it's not used in the simulation), but it is
2752 necessary at this point to get MC working.
2754 Turn model-checking OFF if simulation performance matters to you.
2755 Not enabling it at runtime is not enough, disable it in cmake.
2757 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2760 SimGrid (3.6.2) stable; urgency=low
2762 The "Not coding new stuff allows to polish old things" release.
2765 * New bindings to the NS3 packet level simulator (experimental)
2766 * Use the raw (efficient) execution contexts instead of the sysv
2767 (portable) ones when possible.
2768 * libpcre is now mandatory in any cases since not using it led to
2769 severe performance loss and possibly other issues
2770 * Update the XML platforms:
2771 - G5K: include the latest machine in Nancy
2772 - GridPP and LCG: new platforms
2773 * Documentation was partially updated, at least (more to come)
2775 Bug fixes, cosmetics and small improvements
2776 * Free terminated processes before the end of the simulation to avoid
2777 exhausting the memory of users having very dynamic amount of
2779 * Bug fix and cosmetics about canceling non-running tasks
2780 * Bug fix about the dot loader's issues when using libcgraph
2783 * Create an installer for windows with nsis (amd64 and win32)
2784 - Add an hello world project to illustrate simgrid project creation.
2785 - Embed libpcre into the Simgrid installer to avoid
2786 its compilation burden
2787 * The raw execution contexts should work on Apple now
2788 * Port to Windows 64 bits
2789 - Sysv contexts now have an implementation for this arch
2790 - GRAS communication features now support this arch
2791 * Drop support for borland compiler on windows
2792 - this code was not maintained, and we kinda depend on gcc nowadays
2793 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2794 * Fix portability issue on unstable ubuntu: linker became picky on
2797 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2799 SimGrid (3.6.1) stable; urgency=low
2801 The "Oops, we broke Macs too" release
2804 * Fixed contexts detection so that raw ones are used when possible
2805 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2806 leads to a strange error, with user code segfaulting sometimes when
2807 the generated code is not perfectly aligned (which is not
2808 controllable from the user side, depends on the amount of code)
2811 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2813 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2815 SimGrid-java (3.6) unstable; urgency=low
2818 * Split of every thing from simgrid v3.5 into a separate package.
2820 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2822 SimGrid (3.6) stable; urgency=medium
2824 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2827 * Bindings now constitute their own package, separated from the main one.
2828 Rationale: reduce our maintenance nightmare by reducing the module coupling
2829 They will soon be released on their own on gforge.
2831 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2832 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2834 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2835 * It's not quite deprecated for now because we have no replacement,
2836 but it may soon become the case.
2839 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2840 * Fortran: New user-level cache variable to store the rank of the running
2841 process. This improves performance by an order of magnitude.
2842 * C: New coccinelle script to automatically locate and modify global and
2843 local static variables.
2844 * Improved SMPI network model with a sender-side gap to account for multiple
2848 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2849 only say if a communication is finished, no matter whether it succeeded or
2850 failed. You can call MSG_comm_get_status() to know the status of a finished
2852 * New function MSG_task_dsend() to send a task and detach it. When a
2853 communication is detached, you are never notified of its success or failure
2854 and the memory is released automatically once it is finished. This function
2855 is useful when you don't care about the end nor the success of a
2857 * Change the prototypes of action replay. Sorry for inconvenience,
2858 but this is really more efficient this way (and to adapt your code,
2859 you just have to fix the initialization, that shouldn't be too long)
2860 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2861 ever managed to do anything useful with it.
2862 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2863 takes a dynar of communications. It returns immediately and gives the
2864 index of a finished communication (if any).
2865 * New example: a basic implementation of the Chord P2P algorithm.
2868 * New model for multi-core CPUs. You can now use the core attribute to
2869 precise the number of cores of a host. This is a basic model. Every
2870 process running on the host receives at most the power provided in
2871 the DTD (throughput<=power). Total throughput of process cannot exceed
2873 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2874 router linked by an up-link and a down-link (possibly asymmetrical).
2875 This kind of pattern allows to easily build last-mile model style platforms.
2876 Aggregating such patterns in a rule-based AS is thus the technique of
2877 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2878 * New model for Vivaldi routing. We transformed the Vivaldi network model
2879 into a Vivaldi routing model (based on the rule-based model). This allows to
2880 combine Vivaldi based latencies with last-mile platforms.
2883 * Added a check for NaN of IEEE754 infinite in the double entries of
2885 * Introduce a new context factory "raw", highly inspirated from the
2886 ucontext factory, but using manually crafted functions in assembly to
2887 do the work in an efficient manner.
2888 * Allow to change the used context factory at run time, not only at
2889 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2890 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2891 contexts at runtime (only with raw contexts or ucontexts).
2892 * Completely rewrote this module to allow parallel execution of user
2893 processes. Use --cfg=contexts/nthreads:N to execute user processes
2894 with N parallel threads (the default is 1, meaning no parallelism).
2895 * Allow to decide dynamically between sequential and parallel modes.
2896 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2897 processes in parallel only when their number is greater than or equal to P
2899 * Added a check for NaN of IEEE754 infinite in the double entries of
2903 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2904 won't output the state of processes when interrupted with Ctrl-C
2905 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2906 NULL-terminated array. This may solve backward compatibility issues
2907 due to the change to return type of SD_simulate. See also:
2908 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2909 * Add new macros with variable number of arguments.
2910 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2911 - in xbt/asserts.h: xbt_assert
2912 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2913 - in xbt/ex.h: THROWF and RETHROWF.
2914 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2916 * Change xbt_die() to accept a format string with arguments, just like printf.
2917 * New data structure: xbt_lib_t, like a dict but more general and with better
2921 * New configuration options
2922 Options triva/categorized and triva/uncategorized can be used to generate
2923 graph configuration files for Triva visualization tool.
2924 * Configuration option tracing/platform is renamed to tracing/categorized
2925 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2926 * New TRACE_user_link_variable interface:
2927 User provides the name of the link and the tracing variable to attach to it
2928 * the declaration of tracing categories must be done after the environment
2930 * simpler tracing interface, just one way to declare categories
2931 TRACE_category or TRACE_category_with_color, it is up to you
2932 * links in the trace file are again identified by their names
2933 * trace contains the full platform hierarchy exactly as declared using the ASes
2934 * Options tracing/msg/[task|process]:1 groups the process by hosts
2935 for both cases, tasks and processes must have names that are unique during the simulation
2936 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2937 * The experimental option tracing/msg/volume is deprecated
2938 its functionality may be reincorporated if needed
2940 The tracing generates a trace file with unordered timestamped events,
2941 because of the way the core simulator (surf) works. A script available
2942 at the tools directory (fix-paje-trace.sh) can be used to put the events
2943 in order. We have changed the tracing so it can generate ordered timestamped
2944 events in the final trace, but depending on the simulator (and how much time
2945 is simulated) that can lead to a huge memory utilization. It is deactivated
2946 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2948 Build Infrastructure
2949 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2950 - We are trying hard to keep the API stable, but it may happen that
2951 some things change (we're a research project after all, not a
2952 nuclear plant operating system). If such things should happen, you
2953 could rely on that macro to adapt.
2954 - current value: 30600 for 3.06.00, aka 3.6
2955 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2956 a number that can be compared with SIMGRID_VERSION.
2957 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2958 assertions and verbose/debug logging events are disabled at compile time.
2960 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2962 SimGrid (3.5) stable; urgency=medium
2965 * New feature to any SimGrid-based simulator: Model-Checking
2966 Check SIN#1 for more details.
2969 * New Model SMPI (three-interval linear regression for correction factors)
2970 See RR-7426, available at http://hal.inria.fr/inria-00527150
2971 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2972 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2973 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2974 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2975 * See also src/smpi/README
2980 - Tracing API changes: TRACE_start and TRACE_end should not be called
2981 by user-code. They are automatically called by simulators created
2982 with SimDAG, MSG and SMPI if the toolkit is compiled with
2983 tracing_enabled=ON. Categories declaration and utilization remain the
2984 same for MSG and SimDag.
2985 - A function was added to the tracing API to declare categories with
2987 - TRACE_category_with_color (char *category, char *color)
2988 where color must be in the following format
2989 "%f %f %f", red, green, blue
2990 and red, green, blue are float values in the interval [0, 1]
2991 - User can specify NULL as color parameter, or continue calling
2992 TRACE_category (cat)
2993 On that case, the tracing system will define random colors
2994 - The following command-line options are supported:
2995 --cfg=tracing/filename:msg.trace
2996 --cfg=tracing:1 (activate tracing, needed to use others)
2997 --cfg=tracing/platform:1 (categorized resource use)
2998 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2999 --cfg=tracing/msg/task:1 (task creation)
3000 --cfg=tracing/msg/process:1 (process creation, migration)
3001 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
3002 --cfg=tracing/smpi:1 (SMPI interface tracing)
3003 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
3004 - examples of examples/msg/tracing updated
3006 - DAXLoader and DOTLoader functions can generate tasks with categories
3007 - A new function to attribute a category to SD tasks:
3008 TRACE_sd_set_task_category (SD_task_t task, char *category)
3009 * Tracing the MPI interface implemented by SMPI
3010 - Collective operations are traced with states
3011 - Point-to-Point operations are traced with states/links
3012 - Tracing activated by a parameter "-trace filename" passed
3013 to smpirun during execution (considering that simgrid
3014 is compiled with tracing enabled)
3015 - To run the simulation with gdb, the simulator
3016 accepts --cfg=tracing/smpi:1 to trace SMPI
3017 - tesh files to check if smpi tracing is ok
3018 - See examples/smpi/NAS/DT-trace
3019 * GTNetS tracing re-worked
3020 - adaptation to the tracing system of GTNets to cope
3021 with modifications regarding the fullduplex mode
3022 - new tesh files to check if gtnets tracing is ok
3025 * Asynchronous communications through the functions:
3026 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
3027 * New function: MSG_load_platform_script()
3028 to make possible using a lua script instead of XML files to set up platforms
3029 * New function: MSG_set_function
3030 to associate functions to processes, used when bypassing the parser
3031 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
3033 Platforms: Add some more examples in examples/platforms
3034 * Grid'5000: see www.grid5000.fr
3035 * *_30000_hosts.xml: various huge files [mainly scalability testing]
3038 * Change the XML format. This is a very important modification. SimGrid
3039 3.5 introduces a new hierarchical format based on the notion of
3040 Autonomous Systems. Compatibility with old format is ensured
3041 through the perl script provided in the install bin directory
3042 bin/simgrid_update_xml.
3043 It is now possible to build platforms with specific routing
3044 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
3045 connect several platforms together. We will try to provide soon
3046 set of realistic platforms exploiting these properties (have a
3047 look at examples/platforms/ for the moment).
3048 * Take the opportunity of the XML format change to be a good XML citizen:
3049 rename link:ctn to link_ctn and similar changes (also dealt with by
3051 * Add a new routing scheme (rule-based) using regular expressions. It
3052 enables to have an extremely low memory footprint when the
3053 underlying routing is simple and can be compactly described. You
3054 need to have libpcre4-dev (perl regular expressions) installed if
3055 you want to use this routing scheme.
3056 * Revive the cluster TAG and allow to easily and efficiently (both in
3057 term of memory and speed) connect clusters together. Have a look
3058 at teshsuite/simdag/platforms/ to see how this can be done. With
3059 this tag, you can create clusters with thousands of tasks at no
3060 cost (have a look at examples/platforms/).
3061 Note: clusters are implemented as ASes, so there is no need for an
3062 enclosing AS tag if you have only one cluster in your platform.
3063 * Add new generic functions in the public interface that allows the user
3064 to call SURF 'create_resource' methods from your code (same
3065 functionality as the XML bypass mechanism but with a much lighter
3067 * Add a new model (enabled through command line --cfg=network/model:SMPI)
3068 that uses a piecewise linear approximation to produce better
3069 results when exchanging small messages.
3070 * Add a new parameter to handle correctly full duplex link and account
3071 for interferences between uplink and downlink communications
3072 (activate with --cfg=fullduplex:1).
3075 * Rename the SD_READY (all dependencies are satisfied and task is
3076 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
3077 dependencies are satisfied) state.
3078 This prevents a confusion between the notion of "ready to schedule"
3079 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
3080 simulated" (SD_RUNNABLE) used by the simulation kernel.
3081 * Change the way a task is considered as ready. Instead of removing
3082 dependencies when a task is done, a counter is decreased. This way,
3083 it is always possible to reach ancestors thanks to the
3084 SD_taks_get_parents function (even after the end of the simulation.)
3085 * Change the return type of SD_Simulate from (SD_task_t*) into
3086 xbt_dynar_t. This function was in handling a dynar internally and
3087 converted it into a NULL terminated array for historical reasons.
3088 * New function SD_dotload(char*) to load a DAG described in dot
3089 format. This loader and the corresponding examples require the
3090 installation of the graphviz library.
3091 * Fix a bug in the management of tasks of size 0 in the surf network
3092 models. This problem was only visible with SIMDAG and you should
3093 thus disregard results produced with earlier versions if you
3094 relied on this feature (some tasks were blocked because of this).
3095 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
3096 even though your applications were doing only point-to-point
3097 communications and sequential computations. Now you can really use any
3098 model you want (of course, if you create real parallel tasks, which are
3099 not implemented in most models beside ptaskL07, this will abort).
3100 * Add an example that schedules a DAX on an heterogeneous platform
3101 using a Min-Min strategy.
3102 * New function SD_workstation_get_current_task() that returns the kind
3103 of task currently running on a workstation in the sequential access
3105 * Raise some warnings when unexecuted tasks remains at the end of the
3106 simulation. This is usually caused by cycles in the DAG.
3109 * New function: SIMIX_process_set_function() called by MSG_set_function
3110 * Change the underlying waiting queue in semaphores so that a process
3111 can wait on several of them simultaneously (as in waitany).
3112 * Fix the way to handle tokens in semaphores so that all access patterns
3113 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
3114 * kill the dirty pimple SIMIX_message_sizes_output()
3115 Please use (proper) visualization instead
3118 * New data container: setset (set of sets of elements)
3119 * New module: mmalloc (mapped malloc, allowing to have several
3120 independent segments of malloc)
3121 * New function: xbt_dict_cursor_set_data()
3122 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
3123 * New function: xbt_dynar_is_empty()
3124 * New function: xbt_fifo_get_last_item()
3125 * Fix xbt_dynar_shrink(): use the right element size.
3126 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
3127 uninitialized areas during expand.
3128 * Fix semaphores: previous implementation was severely broken.
3129 * Use library init/fini functions for our initialization.
3130 - you can use logs and other feature as soon as you want in your
3131 code (even before the xbt_init / MSG_init)
3132 - xbt_exit is now a no-op and produce a warning when used.
3135 * Port GRAS to new SIMIX mechanisms. This allows gras users to
3136 benefit from the latest improvement to the simulation kernel.
3137 * Kill measurement sockets for now. If you rely on them, sorry. This
3138 release is not for you. This feature will be reintroduced in the
3139 future, but we cannot delay the release any further.
3140 * New function: gras_msgtype_get_name().
3141 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
3142 * Fix (at last) the pmm example: it should not randomly fail anymore.
3144 Build chain: bug fixes and overall polishing
3145 * Cmake is now stable enough. Hence, we killed the autotools.
3146 * Port to windows ( TM :)
3147 * Fix the 'make install' target.
3148 No need to use 'make install-simgrid' anymore
3149 * Introduce a 'make dist' target compiling a *source* archive
3150 'make package' compiles a binary archive
3151 * Compile java files only on need
3152 * Add --cd and --setenv command line options to tesh
3153 * Out of source builds are not fully supported yet, but we are close
3154 * Enable supernovae and optimization flags by default for our users
3157 * Add layer to set up environment directly from lua, without XML.
3158 * The effect of gras_stub_generator can be achieved through
3159 lua too (check examples/gras/console/ping_generator.lua)
3161 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
3163 SimGrid (3.4.1) stable; urgency=low
3165 The "Polishing easter eggs is probably a good idea" release.
3166 This is a bug fixes release only.
3170 * Fix a bug preventing the tasks from begin garbage collected.
3173 * Fix a bug occurring when a host involved in a communication fails.
3174 This was not detected properly by the other peer involved in the
3175 communication. Now, it's reported as a network error.
3178 * Warn the user about loop dependencies in data flow of DAX files
3179 * Obey the control-flow dependencies of DAX files
3182 * Add option "enable_smpi" allowing to not compile SMPI.
3183 Probably useful for the (Mac) users experiencing a build error here
3184 * Improve the detection of lua5.1 and ruby1.8
3186 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
3188 SimGrid (3.4) stable; urgency=low
3190 The "Easter in Cargese" release. Also known as (major changes):
3192 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
3193 ~> bindings were greatly improved
3194 ~> new tracing infrastructure for better visualization introduced
3196 * the "Welcome to configury modernity" release.
3197 ~> we switched from autotools to cmake, and improved our cdash
3200 A more detailled list of changes follow (full detail in svn log).
3202 Java Bindings: Various Cleanups
3203 * (install java-gcj-compat-dev on debian-like to use them)
3204 * Remove put/get: no need to export deprecated interface in Java
3205 Use send/receive instead.
3206 * Cleanup the examples and add a README per directory
3207 * Remove example autoDestination (that's the only way to go now)
3208 * Remove example explicitDestination (was a plain copy of basic)
3209 * Make JniException a runtime exception, so that there is no need to
3210 declare the fact that you may encounter such a beast. I guess that
3211 nobody will ever want to survive such error.
3212 * Create specific errors for each MSG case of failure:
3213 host failure, transfer failure, timeout, task cancelled
3214 * Cleanup the exceptions that may get thrown by each function
3215 * Other internal cleanups in Java bindings. Performance still bad :/
3216 Ruby and Lua Bindings: create them
3217 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
3218 * That's new and great, you should try them out.
3219 Same functionalities than Java bindings, only even less polished
3221 * Kill the useless "rate" argument of SD_task_get_execution_time()
3222 Everyone used to provide -1 as a value, it was not used, and the
3223 semantic of a possible use wasn't even clear.
3224 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
3225 either as comm costs or compute costs to mean that there is no
3226 such thing for that specific task.
3227 * Add a SD_task_set_name() function
3228 * Fix SD_task_unschedule() on typed tasks
3229 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
3230 * In DAX loader, accept useless 'level' attributes to <job> since
3231 LIGO DAGs have them (seem to be to ease graphical representation).
3233 * Add an example masterslave_mailbox.c using send/receive and not
3234 the deprecated put/get interface.
3235 * Kill the MSG_paje_output() function. It's a noop since 2 years.
3236 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
3238 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
3239 (declare MSG_USE_DEPRECATED to still have the old name)
3240 * Add a MSG_task_set_data() function
3241 * About trace replay (see examples/msg/actions):
3243 - Allow to work with splitted trace files for each process
3244 Give the specific trace file as argument of each process,
3245 and call MSG_action_trace_run(NULL)
3246 You can still have one merged file for all processes.
3247 - Fix implementation of collective operations
3248 * Allow task_execute() on 0-sized tasks (closes #10063)
3250 * This is the first release of SimGrid where SMPI is not considered
3251 beta anymore (even if some corners should still be improved)
3252 * Port over the new SIMIX_network submodule (internal refactoring)
3253 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
3254 * Implement more missing elements of the standard:
3256 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
3258 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
3259 MPI_Scan() MPI_Get_processor_name()
3260 - Added implementation of missing case for Alltoall (warning: it's
3261 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
3262 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
3263 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
3264 SMPI_MPI_Allgatherv()
3265 * Bug fixes include:
3266 - MPI_Waitsome() was broken
3267 - Allow relative includes in smpicc
3268 - Command line cfg argument 'reference_speed' was ignored...
3269 - Some functions did not properly lead to auto-benching of user code
3270 - smpicc passes -O2 by default (just like openmpi one)
3272 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
3273 * Bug fixes about timeouts during communications
3274 * add SIMIX_message_sizes_output() as a pimple to write to file the
3275 amount of messages per size. Use gnuplot to get histogram.
3276 Pimple because that's the only user-visible function of simix,
3277 defined directly in xbt.h (irk, sorry)
3279 - Add a SIMIX_sem_get_capacity() function
3280 - Fix interactions with processe resume/suspende
3281 - release_forever() was stupidly broken
3282 - Fix SIMIX_display_process_status() for processes in a semaphore
3283 - Make SIMIX_sem_block_onto() user-visible
3284 * Refactoring context stuff:
3285 - Use pseudo-OOP for better modularity
3286 - reimplement SIMIX_process_kill() without process_schedule() so
3287 that the latter can take as invariant that it is called from
3289 - Merge context_start into context_new for sake of simplicity
3291 * Add a Vivaldi network model, coded live during SUD'10 ;)
3292 * Rename configuration variables to start a hierarchy:
3293 o cpu_model -> cpu/model
3294 o network_model -> network/model
3295 o workstation_model -> workstation/model
3296 * New configuration variables:
3297 o network/bandwidth_factor: correction to bandwidth
3298 o network/latency_factor: correction to latency
3299 o netwotk/weight_S: correction to the weight of competing streams
3300 * Add a long description to the models, that users can see with such
3301 argument on the command line: --cfg=cpu/model:help
3302 * --help-models display the long description of all known models
3304 * config: add the ability to set a default value after registration
3305 Does not override any previously set value (e.g. from cmd line)
3306 * dict: allow to have integer key and data.
3307 When so, you need to use the following functions
3308 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
3309 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
3310 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
3311 In contrary to regular dicts, the key is not malloced before copy.
3312 Mixing scalar and regular elements in the same dict is not tested
3314 * Allow to use xbt_dynar_shrink() to expend the dynar instead
3315 Tracing for Visualization:
3316 * SimGrid is now instrumented in order to generate a trace file for
3317 visualization analysis: to use it, need to compile SimGrid with the
3318 "tracing" option enabled, and instrument the program using SimGrid with
3319 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
3320 (among other functions).
3321 * The instrumentation only traces the platform utilization for now
3322 * Documentation to use the tracing functions and how to analyze the
3323 traces with the Triva tool is written.
3324 * More information about: SimGrid FAQ (in the section Tracing Simulations
3327 * We moved to cmake as default build system. Autotools support will
3328 be dropped soon. Check the FAQ for more info about how to use it.
3329 * Greatly improved our cdash/ctest interactions
3330 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
3331 * Added memory checking tests with valgrind; lot of memleak fixing.
3332 This may be the first release of simgrid with so few memory issues
3333 * Added code coverage tests.
3334 Our coverage is still improvable, but at least we see it on cdash.
3336 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
3338 SimGrid (3.3.4) stable; urgency=low
3340 The "Desktop Grid needs love too" release (also called Xmas release).
3342 Models improvements:
3343 * Major speedup in the maxmin system solving by using lazy evaluation
3344 Instead of solving completely the maxmin system at each iteration,
3345 only invalidate (and recompute) the modified parts.
3346 This new feature is enabled in default models but you can try to
3347 turn it on with "--cfg:maxmin-selective-update=1" for other models.
3348 * Cas01 IMproved as default CPU model
3349 This CPU model is the same Cas01 model, but it uses the
3350 maxmin-selective-update flag and a heap structure to manage
3351 actions on SURF kernel.
3352 It reduces the complexity to find the next action to finish and,
3353 consequently, it's faster than the old Cas01.
3354 This is the new default CPU model (Cas01).
3355 * Rename the old Cas01 model to Cas01_fullupdate
3356 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
3357 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
3358 * CpuTI (CPU Trace Integration)
3359 A new CPU model whose objective is simulate faster when using
3360 availability trace files.
3361 Instead of using a full featured, over engineered maxmin system for
3362 CPU modeling, this model does the pre-integration of traces files
3363 to calculate the amount of CPU power available, and so, executes
3364 faster than the old CPU models.
3365 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
3366 * Use LV08 as default network model since it gives better accuracy
3367 for small messages and shouldn't change things for big ones.
3368 Use --cfg=network_model:CM02 to get the previous behavior.
3371 ******************************************
3372 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
3373 ******************************************
3374 * The new CPU model may changes simulations!
3375 The point is that events occurring at the exact same timestamp
3376 are not scheduled in the same order with the old and new
3377 version. This may be enough to completely change the execution
3378 of simulations in some cases.
3379 * The new network model will change simulations!
3380 This new model is more realistic than the previous one, so you
3381 should consider redoing your old experiments with this model.
3382 Sorry for the inconvenience.
3385 * Introduce the supernovae compilation mode
3386 When compiled that way, the whole SimGrid (or almost) is put in a
3387 single compilation unit and compiled in one shoot.
3388 This is to help gcc which has difficulties to inline stuff from one
3390 The speedup seem to be above 15%, althrough more tests are needed on
3391 amd64 to confirm that gain.
3394 * Port of MSG's mailbox on top of SIMIX network
3395 The put/get mechanism was greatly simplified on the way.
3398 * New SIMIX network module. Provides:
3399 - Mailbox: rendez-vous mechanism to find with who you want to speak
3400 - Synchronous send/recv: easier and hopefully faster since the
3401 logic is handled in the maestro process directly now
3402 - Asynchronous send/recv: you dreamt of it? It's here now
3403 Too bad that nobody cared enough to propagate the change to MSG.
3404 * Add semaphores as SIMIX synchronization mechanism.
3407 * new function SD_daxload(char*) to load a DAX file
3408 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3409 * Introduce typed tasks. Specify its kind and cost at creation.
3410 At scheduling, just give where it should be placed, and the cost
3411 for each involved resource is automatically computed.
3412 Existing constructors so far (more to come of course):
3413 - SD_task_create_comm_e2e() for end-to-end communication
3414 - SD_task_create_comp_seq() for sequential computation
3415 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3416 * new function SD_task_dump() for debugging display
3417 * new function SD_task_dotty(task,FILE*) writing to file the info
3418 about the task in dotty format
3419 * SD_task_dependency_exists() can now cope with having one of its
3420 arguments NULL. If so, it tests whether the other argument has any
3422 * Add getters on list of preceding/following tasks:
3423 SD_task_get_parents(task) and SD_task_get_children(task)
3424 * Add getters on amount of workstations and list:
3425 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3426 * Add getter on task kind: SD_task_get_kind(task)
3427 * Update the start_time and finish_time of tasks on completion/failure
3428 * Bugfix: Remove task from state swags when destroyed
3431 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3434 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3435 dynar with one separate thread per value of the dynar.
3436 * Change the prototype of xbt_thread_create(), sorry.
3437 Added a boolean parameter indicating whether we want to join this
3438 thread (used in SG only for now)
3439 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3442 * GTNetS wrappers should now be usable again (and betterly tested too)
3443 * Fix a major regression from 3.2 where the timeout provided to
3444 MSG_task_put_with_timeout() was used as absolute time before which
3445 the comm should be done.
3446 * Start to fix the <cluster> tag.
3447 - Internal links should be good now (beside of the loopback, which
3448 use the private link instead)
3449 - paths to the external world is still rather broken
3450 - the <route:multi> tag is just broken. Actually that's brain-dead.
3451 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3452 to make it less stupid
3453 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3454 * Fix a source-level compatibility glitch from 3.2: after defining
3455 MSG_USE_DEPRECATED, you can use the old name
3456 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3457 * Allow to compile from the SVN with automake 1.11
3458 * Fix some problems when using the "start_time" tag in deployment XMLs.
3459 * Fix #8569: XBT/synchro.h has redundant declarations
3460 * Fix #8563: MSG return values and exceptions
3461 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3462 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3463 Thanks to Robson Peixoto.
3464 * Fix a vicious bug in dictionaries inducing that some elements were
3465 not freed on xbt_dict_free()
3467 Portability report of this version:
3468 * Main portability targets:
3469 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3470 on (amd64/i386/ia64)
3471 - mac leopard on i386
3472 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3473 but nothing critical.
3474 * Other platforms: windows, AIX and others were not tested for this release
3476 Timing report of this version:
3477 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3478 scenario parameters). From 8h to a few seconds in desktop grid settings.
3479 * Supernovae brings about 25% speedup on i386.
3481 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3483 SimGrid (3.3.3) stable; urgency=low
3485 The "Need for Speed" release.
3487 The timings done to validate the 3.3.2 were faulty.
3488 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3490 The problem was a conversion from a manually handled vector to
3491 xbt_dynar_t on the critical path.
3492 xbt_dynar_foreach calls functions, inducing stack management crap.
3494 We inlined these functions and xbt_dynar_foreach is now breath taking.
3495 We also inlined xbt_swag_belong on the way.
3497 Here are some approximate speedup measurements (on master/slaves
3498 simulations lasting between 10s and 20s each):
3499 3.3.1 -> 3.3.2: about same performance
3500 3.3.2 -> 3.3.3: 40% speedup
3501 3.3.1 -> 3.3.3: 40% speedup
3502 3.3.1 with inline patch -> 3.3.3: 30% speedup
3504 Our reading is that the refactoring which occurred in 3.3.2 made us
3505 suffer much more from the xbt_dynar_foreach low performance, but
3506 once we solved this, this refactoring proved to be very performance
3507 effective. From the 40% speedup, somehow, 10% are due to the
3508 inlining and 30% to the refactoring.
3510 That's a pitty that gcc cannot inline functions placed in other files
3511 alone. We have to choose between:
3512 - break the encapsulation (by putting private data structures and
3513 accessors in headers files to help gcc)
3514 - live with low performance
3515 - switch to a decent compiler such as icc (not quite possible).
3517 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3519 SimGrid (3.3.2) stable; urgency=low
3521 The "Simplicity does not preceed complexity, but follows it" release.
3523 The main contributors of this release were (lexical order):
3524 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3527 * Extract the routing logic into its own object.
3528 (was dupplicated in network.c and workstation_LV07.c;
3529 Allows to implement other ways of storing that info)
3530 => kill now useless network_card concept
3531 - Use dynar to represent routes (instead of void** + int*)
3532 - kill link_set (use surf_network_model->resource_set instead)
3533 - Add a command-line option to choose the routing schema to use
3534 - Add three new models:
3535 * Floyd (shortest path computed at initialization)
3536 * Dijikstra (shortest path recomputed all the time)
3537 * Cached Dijikstra (shortest path computed on need)
3538 All these models where contributed by Silas De Munck, and are
3539 described in his ICCS09 paper.
3541 * Simplify model declaration
3542 (less redirections, less function to write when defining a model)
3543 - Factorize stuff between models:
3546 surf_model_resource_set(model)
3547 surf_model_resource_by_name(model, name)
3548 - Unify the types of models in s_surf_model_t (using an union)
3549 - Embeed fields of common_public directly into s_surf_model_t
3550 - Rename model methods:
3551 action_free ~> action_unref
3552 action_change_state ~> action_state_set
3553 action_get_state ~> action_state_get
3554 - Change model methods into functions:
3555 (model)->common_public->action_use ~> surf_action_ref
3557 * Implement a generic resource; use it as ancestor to specific ones
3558 (allows to kill duplicated code in models)
3559 Drawback: timer command don't need no name nor properties;
3560 workstation_CLM03 don't need no properties
3561 (but I guess we can live with those few bytes wasted)
3563 * Improve the action object model
3564 - implement a constructor avoiding dupplicated code about field
3565 initialization in generic_action part.
3567 * Kill the SDP model: it has an external dependency, is deprecated
3568 in flavor of modern lmm models, and didn't compile since a while
3571 * Relocation of the context module from XBT to SIMIX.
3572 (the context were decoupled from the simix processes, duplicating a lot of code)
3573 => a lot of code was factorized
3574 - less overhead is introduced during scheduling
3575 - simpler API for the context factory
3576 - the logic for process creation,destruction and manipulation was simplified
3577 * Simplification of the s_smx_process_t data structure.
3578 => accesing the simix level data associated to a process is faster now,
3579 and the code is a lot more readable.
3582 * Implement some more MPI primitives:
3583 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3584 -implementation: Bcast: flat or 2-ary tree (default),
3585 Barrier: 4-ary tree,
3587 Allreduce: Reduce then Bcast
3588 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3589 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3590 Alltoallv: flat tree, like ompi
3592 * Add support for optimized collectives (Bcast is now binomial by default)
3593 * Port smpirun and smpicc to OS X
3596 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3597 Shout out if you used it.
3600 * Display the list of still queued messages in SG mode when existing
3604 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3605 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3606 * Add xbt_str_from_file(FILE*)
3607 * Add xbt_dict_get_key achieving a linear reverse search
3608 * Remove the context module
3610 Portability report of this version:
3611 * Main portability targets:
3612 - Linux(debian)/x86/context
3613 - Linux(debian)/x86/pthreads
3614 - Linux(debian)/amd64/context
3615 - Linux(debian)/amd64/pthreads
3616 On these, we still have the eratic breakages of gras/pmm and
3617 amok/saturate_sg reported in previous version. We still think
3618 that the tests are the cause of the fault, not the tested code.
3620 - Mac OSX Leopard/x86/context
3621 Still false negative in tesh autotesting.
3622 Smpi still fails, but this time because readlink does not accept -f
3623 Everything seems to work properly beside of that.
3626 - AIX version 5.3 (only tested contexts this time)
3627 Smpi still fails there because mktemp is not installed.
3628 Everything seems to work properly beside of that.
3630 I managed to compile it for the first time, but several breakages.
3631 Won't delay the release for this exotic platform.
3633 * Windows: it's still lagging behind. If you want to help, please
3636 Timing report of this version:
3637 This version seem to be more than 5% faster than 3.3.1 (on linux
3638 64bits with contextes). The gain is less than expected, we are
3639 investigating this for next release.
3641 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3643 SimGrid (3.3.1) stable; urgency=low
3646 * Implement a --cfg-help to show existing configuration variables
3647 * Build chain do not require doxygen in maintainer mode
3650 * fix a bug on struct sizeof computation, which prevented the
3651 exchange of arrays of structs in some conditions
3652 - added a regression test about this in datadesc_usage
3653 * Allow the exchange of 0-long dynamic vectors.
3654 - for that, use -1 as indicator of dynamic size instead of 0
3655 - This implied to change any size from unsigned long to long,
3656 reducing a bit communication abilities, but I guess that with
3657 64bits being quite common, this is more than enough.
3658 - This also induce a protocol change, thus bumping network protocol
3659 version from 0 to 1 (if we have external users, we have to get
3660 clean on that point too ;)
3661 - added two regression tests about this in datadesc_usage
3662 * Be more verbose when propagating local exceptions
3663 This helps debugging.
3664 * Display the status of simulated processes when receiving SIGINT in
3668 * Allow to control the simulation from a trace file.
3669 New functions MSG_action_register() and MSG_action_trace_run()
3670 The first one allows to associate a function execution to each
3671 kind of action while the second one parses a trace file and
3672 triggers the corresponding actions within the system.
3673 For now, only a toy example is provided in examples/msg/actions
3674 * Add an example of process migration in examples/msg/migration
3675 * Fix a bug in task exchange which broke MSG_task_get_sender()
3676 Add a teshsuite regression test for that.
3677 [Bug: if MSG_task_get_sender() is called after sender exit,
3679 * Fix a bug which prevented suspend/resume to work properly
3680 * Display the status of simulated processes when receiving SIGINT
3681 This fixes a regression of v3.3. due to the introduction of SIMIX
3682 * Bug fixing in failure management:
3683 - trace could not start by a failure at time 0
3684 - failure during communications were not working
3687 * Add SIMIX_process_set_name() to change the name of the current
3688 process in the log messages.
3689 * Store smx_hosts in a dict since we only retrieve them by name
3690 * Move the configuration infrastructure to surf
3693 * Move the configuration infrastructure to surf
3696 * Massive internal cleanups:
3697 - Store internal structures on processes instead of hosts (allows
3698 to have more than one process per host, in addition of being more
3700 - Cleanup the initialization/finalization process
3701 - Kill a whole bunch of unneeded synchronization:
3702 processes run in exclusive manner within the simulator
3703 - Move queues from global tables to process data fields
3705 - now accept -platform and -hostfile arguments
3706 - Pass the right rank value to processes according to the hostfile
3707 * Compile the examples by default, and use them as regression tests
3708 * Implement MPI_Wtime()
3709 * Change the reference speed to a command line option
3712 * TCP_gamma can now be specified as command line option using
3713 --cfg=TCP_gamma:10000000.0
3714 * Change the --surf-path cmd line option into --cfg=path:
3717 * Also include strbuff from xbt.h public header
3718 * xbt_ex_display(): do not free the exception after displaying
3719 This allows to do more with the given exception afterward.
3720 Users should call xbt_ex_free() themselves.
3724 Portability report of this version:
3725 * Main portability targets:
3726 - Linux(debian)/x86/context
3727 - Linux(debian)/x86/pthreads
3728 - Linux(debian)/amd64/context
3729 - Linux(debian)/amd64/pthreads
3730 These targets fail about 1/10 of times on gras/pmm, but we believe
3731 that this is because of the test, not because of simgrid.
3732 amok/saturate_sg fails even more rarely, and the test may not be
3735 - Mac OSX Leopard/x86/context
3736 The test suite still spits tons of errors because some obscure
3737 force prevents us from removing the temporary directories
3738 arguing that they still contain some metadata I've never heard of.
3739 Smpi fails because seq is not installed.
3740 Everything seems to work properly beside of that.
3743 - AIX version 5.3 (both contexts and pthread)
3744 Smpi still fails there because mktemp is not installed.
3745 XML inclusions seems rosty on AIX.
3747 * Windows: it's still lagging behind. If you want to help, please
3750 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3752 SimGrid (3.3) stable; urgency=high
3756 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3757 [Malek Cherier & Mt]
3759 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3762 Until now, GRAS were using MSG as an interface to SURF. It was
3763 quite difficult because both interface have several differences
3764 (MSG channels vs GRAS sockets were the most notable point).
3766 This also opens the gate to SMPI (which should occur soon) and speed
3767 up simulations by to 40% (even if it were not the main goal).
3769 **************************************
3770 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3771 **************************************
3772 The point is that events occurring at the exact same timestamp are
3773 not scheduled in the same order with the old and new version. This
3774 may be enough to completely change the execution of simulations in
3775 some cases. Sorry for the inconvenience.
3777 * Cleanup and upgrade the XML format to push further scalability
3778 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3780 * Improve the testing infrastructure with tesh. Now a very large part of
3781 the code is tested not only by being run but also by checking that the
3782 output match an expected output [Mt].
3784 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3785 is really less memory-demanding, which should allow you to use
3786 larger files in SimGrid [AL].
3788 * Inform valgrind about our contextes, so that it becomes usable
3789 with the default (and more effecient) version of SimGrid
3790 [contributed by Sékou Diakite, many thanks]
3793 * Introduce a listener thread in charge of receiving incoming
3794 messages from the network. It allows to overlap communication and
3795 computation but most notably, it removes some stupid deadlocks due
3796 to the fact that so far, a process could not send and receive at
3797 the same time. This made most non trivial communication schema
3799 * Convert the PIDs from long int to int to match the MSG ones (and
3800 linux ones too) [Mt]
3801 * New function: gras_agent_spawn() to launch a new process on
3802 current host. Only working in simulation for now. [Mt]
3803 * New function: gras_os_hostport() returning a constant form (ie,
3804 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3807 * Make the backtrace of exceptions more human readable [Mt]
3808 * New module: xbt/str [Mt]
3809 a ton of string utility functions (split, join, printf to a newly
3810 allocated buffer, trim, etc)
3811 * New module: xbt/hash [Mt]
3812 SHA1 hashing algorithm (more to come if needed)
3813 * New module: xbt/synchro [Mt]
3814 synchronization tools (mutex and conditions) working the same way
3815 in simulation and in real life (mainly useful for GRAS, but not
3817 * New module: xbt/queue [Mt]
3818 classical producer/consumer synchronization scheme
3819 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3820 (using the classical functions will get serialized) [Mt]
3821 * Make dictionary internal table dynamic. No need to specify its size
3822 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3824 * Make sure the log channels are organized as a tree under windows
3825 (because of ANSI C compatibility issue, any channel were child of
3826 root directly) [Mt].
3829 * Cleaned many thing in surf and fixed a few bugs [AL].
3830 * Add a nice command line configuration mechanism to compose models [AL].
3831 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3832 the previous one (KCCFLN05). It relies on something that looks like
3833 a max-min sharing mechanism but cannot be written as such. A new solver
3834 was thus designed [AL].
3835 * Add a new solver to lmm. Based on Lagrange optimization and
3836 gradient-based descent, it enables to efficiently maximise systems s.a
3838 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3840 This solver enables to propose two new network models for TCP Reno and
3841 TCP Vegas based on Low's work. These models still need to be fully
3842 tested though [Pedro Velho].
3845 * Bug fix in SD_simulate. Now the time bound given as argument is
3847 * Use the new parallel task model (ptask_L07) as default.
3848 * Use the SURF command line configuration mechanism.
3849 * 0-size tasks (for synchronization) should now work.
3851 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3853 SimGrid (3.2) stable; urgency=high
3857 We still experience issues on this platform, but we believe that at
3858 least MSG is usable.
3860 GRAS API BREAKAGE (for simplification purpose, sorry):
3861 * the gras_msgtype_by_name is not used anymore. Instead of
3862 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3863 you can write (and must)
3864 gras_msg_send(toserver, "request", &request);
3865 - If you still want to pass a gras_msgtype_t to the function (to cache
3866 the type and avoid the lookup time), use the gras_msg_send_() variant.
3867 - Impacted functions:
3868 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3869 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3870 * The callbacks are now expected to return 0 when everything went well
3871 (just like the main() function)
3873 GRAS new features and improvements:
3874 * New module mechanism where user code can use per process globals [Mt]
3875 This is similar to gras_userdata_*() functions, but for libraries. It
3876 factorize some code developped over and over in the examples and AMOK.
3877 It has still to be documented and used (only amok/peermanagement is
3879 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3880 loss when they were small enough to fit into the buffer and sent quickly
3881 enough so that they can all get received in one shoot.
3882 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3883 if not found. Use the *_or_null() variant for the old semantic.
3884 * In gras_msg_handle, do not discard messages without callback.
3885 They are probably messages to be explicitly awaited later (ie, proofs of
3886 mis-synchronization in userland since they are sent before being awaited)
3887 No big deal usually.
3888 * gras_socket_meas_send/recv: semantic changed!
3889 The numerical arguments used to be (1) the total amount of data to send
3890 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3891 messages. This was need for the fool willing to send more than MAXINT
3892 bytes on quite fat pipes.
3895 * Do really rename the hostmanagement module to peermanagement. [Mt]
3896 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3897 but this is rather new and this was documented in the module
3898 documentation (poor excuses, I admit)
3899 * Bandwidth measurement semantic changed! This follows the changes to
3900 gras_socket_meas_send/recv explained above.
3903 * A sequential mode has been added to the workstations. When a workstation
3904 is in sequential mode, it can execute only one task, and the other tasks
3905 are waiting in a FIFO. [Christophe Thiery]
3908 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3909 model for SIMDAG. [Christophe Thiery]
3910 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3911 equal to 0, which caused some bad behaviors in
3912 saturated_constraint_set_update. I now use a threshold mechanism like in
3916 * When running manually src/testall, you select specific units [Mt]
3917 testall is the result of our cunit mechanism, and should replace all
3918 the scripty thingy around since bash don't run easily on billware.
3920 * A mallocator system has been added. [Christophe Thiery]
3921 Mallocators allow you to recycle your unused objects instead of freeing them
3922 and allocating new ones.
3924 Documentation update:
3925 * FAQ reworking + New FAQs:
3926 - "Valgrind spits tons of errors!" [Mt]
3927 - "How to repport bugs" [Mt]
3928 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3929 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3930 - Communication time measurement within MSG [AL]
3931 - I experience weird communication times when I change the latency [AL]
3932 * GRAS tutorial [Mt]
3934 - an introduction to the framework and to the used communication model
3935 - an initiatic tour introducing the most proheminent features:
3937 . Lesson 0: Installing GRAS
3938 . Lesson 1: Setting up your own project
3939 o Part 2: Message passing
3940 . Lesson 2: Exchanging simple messages
3941 . Lesson 3: Passing arguments to the processes (in SG)
3942 . Lesson 4: Attaching callbacks to messages
3943 . Lesson 5: Using globals in processes
3944 . Lesson 6: Logging information properly
3945 . Lesson 7: Using internal timers
3946 . Lesson 8: Handling errors through exceptions
3947 . Lesson 9: Exchanging simple data
3948 . Lesson 10: Remote Procedure Calling (RPC)
3949 . Lesson 11: Explicitly waiting for messages
3950 . Recapping of message passing features in GRAS
3951 - A HOWTO section containing:
3952 o HOWTO design a GRAS application
3953 More are due, of course. They will come latter. In the meanwhile, you can
3954 check the examples which are still here.
3956 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3958 SimGrid (3.1) stable; urgency=high
3962 There was a stack corruption somewhere, visible only when optimizing
3963 with these versions. [Vince]
3966 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3967 module that enabled to play with Directed Acyclic Graphs. It is built
3968 directly on top of SURF and provides an API rather close to the old
3969 SG. Some old codes using SG are currently under rewrite to check that
3970 all needful functions are provided. [Christophe Thiery]
3973 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3974 extension of the classical CLM03 model that gracefully handles
3975 failures. This is now the default model for MSG and GRAS. It doesn't
3976 handle parallel tasks yet though. [AL]
3977 * Bug fix: Weights were not correctly set in the network part.
3978 WARNING: This may have resulted in incorrect results with simulations
3979 where there are more than one flow on a given link. [AL]
3982 * After a (long ?) discussion on simgrid-devel, we have decided that the
3983 convention we had on units was stupid. That is why it has been decided
3984 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3985 WARNING: This means that all previous platform files will not work as
3986 such with this version! A warning is issued to ask users to update
3988 A conversion script can be found in the contrib module of the CVS, under
3989 the name contrib/platform_generation/surfxml_update.pl [MQ]
3992 * Bug fix: Processes were started in reverse order, wrt deployment file.
3993 WARNING: if your code relies on this bug, please fix it. [AL]
3994 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3995 being executed on two different locations. [AL]
3996 * Bug fix: Failures are now better supported thanks to Derrick's tests
3997 (there was many failure situations I hadn't thought of and that weren't
3998 correctly handled). [AL]
3999 * New function: MSG_host_is_avail indicates you whether a given m_host_t
4003 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
4004 Exception occurring on server-side are propagated back to client (!).
4006 API CHANGE: the callback changed their prototype. Change:
4007 int my_handler(gras_socket_t expeditor, void *payload_data) {
4009 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
4010 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
4012 * New! function: gras_msg_handleall to deal with all messages arriving
4013 within a given period.
4014 * New! function: gras_socket_server_range to get a server socket in a
4015 range of port numbers (ease to avoid port number conflicts) [MQ]
4016 * New! gras processes display their backtrace when they get a SIGUSR1
4017 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
4018 Sweet to debug RL processes [MQ]
4022 - Do not force experiment sizes to be expressed in kb, or it becomes
4023 impossible to measure the latency this way (needs one byte-long tests)
4024 WARNING: this changes the amok_bw_* function semantic. [MQ]
4025 - Implements the link saturation stuff. [MQ]
4026 * Peer management module:
4027 New! module factorizing code that we wrote over and over [MQ].
4030 * New module: cunit (my jUnit implementation in ansi C) [MQ]
4031 - Test units are placed directly into the library code, they get extracted
4032 automatically and placed into the src/testall binary.
4033 - Convert most of the XBT tests to this system.
4034 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
4035 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
4036 live in the same C code. This required to change a little bit the API
4037 of surfxml parsing but shouldn't be an issue for end-users. [AL]
4038 * New module: sparse graph structure with basic algorithms (this is work
4039 in progress and the API is not considered to be frozen yet). [AL]
4040 * Display more information on backtraces: source line & function names are
4041 now displayed just like valgrind does (rely on addr2line tool) [MQ]
4042 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
4043 * Reworked a little bit some #include statements to load only required
4044 headers. Some user code that relied on SimGrid to include stdlib or
4045 stdio may need to include it by themselves. [AL]
4046 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
4048 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
4049 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
4050 but I guess that almost nobody used those parts. [MQ]
4052 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
4054 SimGrid (3.0.1) stable; urgency=low
4057 * Unfortunately, I had missed 5 misnamed functions:
4058 xbt_fifo_item_t xbt_fifo_newitem(void);
4059 void xbt_fifo_freeitem(xbt_fifo_item_t);
4060 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
4061 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
4062 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
4063 They're now deprecated. Please use their new versions:
4064 xbt_fifo_item_t xbt_fifo_new_item(void);
4065 void xbt_fifo_free_item(xbt_fifo_item_t);
4066 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
4067 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
4068 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
4070 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
4071 * Documentation: xbt_log module unmercifully reworked [MQ]
4072 * Bugfix: there was a problem with the ending of contexts with
4073 the pthread backend. It caused some weird deadlock or behavior
4074 depending on the pthread implementation. [AL]
4075 * Bugfix: get the exceptions raised in the simulator repport where
4076 and why they come from when they are not caught in time [AL, MQ]
4079 * Bugfix: Do repport the error when two non-connected hosts try to
4080 exchange data (Thanks to Flavien for stumbling into this one) [AL]
4083 * Add additionnal checkings on communications. Assert that two
4084 communicating hosts are connected by a set of links... [AL]
4087 * Add additionnal checkings on channel values in communication [AL]
4088 * New: MSG_task_get_source to see on which host a task was generated [HC]
4089 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
4090 the number of tasks waiting to be received on channel and sent
4092 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
4093 waits for the first task coming from a given host.. [AL]
4095 GRAS new functionnalities: [MQ]
4096 * Enhance the parsing macro to allow the size of multidimentional objects
4097 to be given thru annotations.
4098 * New example (and documentation): Matrix Multiplication a la RPC
4099 (as when I was young!) and fix a bunch of bugs found on the way.
4101 GRAS performance improvements: [MQ]
4103 * Reduce the amount of cbps creation/destruction by making it static to
4104 datadesc_send/recv() and using a (newly created) cbps_reset (based on
4107 * Change libdata to a set so that we can search for stuff by ID (and thus
4108 reduce the insane amount of dict lookups)
4111 * Actually implement gras_datadesc_copy() so that we don't have to mimick
4112 RL communication on top of SG since it's so uneffective.
4113 It may also be used for inter-thread communication in RL, one day.
4114 * Use gras_datadesc_copy() to exchange messages on top of SG
4116 - improve message exchange performance on top of SG
4117 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
4118 * Don't exchange on the network the size of the used part of buffer,
4119 instead, specify the possible buffer size to read().
4121 - reduces the amount of read/write calls (one pair per exchange)
4122 - reduces the amount of exchanged data (the size)
4123 - allows to retrieve all arrived data on receiver side, if we don't need
4124 it right now (subsequent read will peek the buffer)
4125 - allows the receiver to proceed with the beginning of the stream before
4126 everything is arrived
4127 - make it possible to build an iov transport (using readv/writev)
4129 - take care of the data with non-stable storage (like stacked data),
4131 * If possible, TCP send uses vector I/O (when writev() is here)
4132 - Don't use it for receive since we send data sizes and data on the
4133 same stream, so we wouldn't be able to chain large amount of chunks
4134 before having to flush the stuff to read the size.
4135 * Rework the transport plugin mechanism to simplify it and reduce the
4136 amount of pointer dereferencement when searching for the right function
4139 * I guess that now, we do almost as few system calls as possible while
4140 doing as few data copy as possible.
4142 To improve it further, we could try to send all the sizes first and then
4143 all the data (to use iov on receiving size), but it's only a partial
4144 solution: when you have 2 dimensional data, the sizes of the second
4145 dimension is data of the first dimension, so you need 3 streams.
4147 I'm not sure the potential performance gains justify the coding burden.
4149 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
4151 SimGrid (3.00) stable; urgency=high
4154 * New! Give the possibility to hijack the surf parser and thus bypass
4155 MSG_create_environment and MSG_launch_application. Have a look at
4156 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
4158 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
4160 SimGrid (2.96) unstable; urgency=low
4165 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
4166 This deprecates the xbt_error_t mechanisms.
4167 It modifies (simplifies) all XBT and GRAS API.
4168 MSG API keeps unchanged (exceptions raised by XBT are caught from
4169 within MSG and masked with existing error handling facilities)
4172 * New! Add a FATPIPE model. [AL]
4173 * New! Add a parallel task model. [AL]
4174 * New! Add automatically a loopback interface (in the default
4175 network model) if none was precised.
4178 * Bugfix: MSG_process_resume now works with the current running process.
4180 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
4181 * Modification of MSG_task_get_compute_duration. Once a task has been
4182 processed, the value returned by this function is now equal to 0. [AL]
4183 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
4184 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
4185 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
4186 of a m_task. For now, MSG_task_cancel only works with computation
4188 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
4189 of the processor (in Mflop/s) regardless of the current load on the
4191 * API Change: use proper naming convention for MSG_getClock and
4192 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
4194 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
4195 This function changes the priority of a computation task. This priority
4196 doesn't affect the transfer rate. A priority of 2 will make a task
4197 receive two times more cpu power than the other ones. This function
4198 has been added to suit the needs of Nguyen The Loc and hasn't been that
4199 much tested yet. So if it fails, please report it and send me your code.
4201 * API Change: removed all functions and types that were marked "deprecated"
4202 since many months. Renamed MSG_global_init_args to MSG_global_init.
4204 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
4206 SimGrid (2.95) unstable; urgency=low
4209 * Steal some nice code to GNU pth to fix context detection and usage [AL]
4210 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
4211 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
4214 * Bugfix: Allow absolute paths for platform description files [MQ]
4215 * Bugfix: do free the variables after use. Leads to drastic performance
4217 * Implement max_duration (ie, timeouts) on resources [AL]
4220 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
4222 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
4223 the needed support (timeouts) [AL]
4226 * Implement measurement sockets. You can now get the bandwidth between two
4227 hosts thanks to AMOK (see below). [MQ]
4228 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
4229 dynar over the network (yeah) [MQ]
4230 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
4231 * simple/static token-ring example. [Alexandre Colucci and MQ]
4232 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
4233 in place before [MQ]
4236 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
4237 type parsing. "short" and "long" modifiers were also ignored; other
4238 modifier (such as reference level) are still ignored. [MQ]
4239 * Embeed the buffer size within the buffer itself on SG. [MQ]
4240 That way, send() are atomic and cannot get intermixed anymore (at least
4241 the ones which are less than 100k; bigger messages still have the issue)
4242 * Array size pushed by the field, not by the field type (or each
4243 and every long int will push stuff to the cbps) [MQ]
4244 * use select() to sleep since it allows to portably sleep less than one
4247 GRAS (minor cleanups)
4248 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
4251 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
4252 * type category 'ignored' killed as it was never used and were difficult
4254 * whether a type can cycle or not is now a flag, leaving room for more
4255 flags (as "ignored", if we feel the need one day ;)
4256 * Rename raw sockets to measurement sockets since "raw" has another
4257 meaning in networking community.
4260 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
4261 and offers features not belonging to GRAS but that most applications
4262 need. One day, it may be a set of plugins loadable at runtime.
4263 * New module: bandwidth
4264 bandwidth measurement between arbitrary nodes running this module. [MQ]
4266 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
4268 SimGrid (2.94) unstable; urgency=low
4270 The first beta release of SimGrid 3 !
4274 * Update the main page and the FAQ. Adding references to gforge.
4277 * Add a gras_os_getpid function.
4280 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
4281 * Extend the logs so that they also print PID, hostname, date, ... if
4283 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
4284 and kill the old version which were in testsuite/
4285 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
4289 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
4290 for users. Instead, each programming environment (GRAS, MSG,...) use it
4291 when needed to provide such a feature to users.
4292 Don't shortcut the mechanism or you will also shortcut the virtualization
4293 you need on the simulator.
4297 * Cleanups in configury with regard to compile optimization/warning flags.
4298 Also add -fno-loop-optimize to any powerpc since it's the optimization
4299 killing gcc (< 3.4.0).
4300 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
4301 needed by MSG examples complications
4302 * Borrow configury beautifications from PHP
4305 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
4306 warning (thanks loris for stumbling into this one).
4307 * Bugfix: stop loading private headers (gras_config.h) from the public
4311 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
4312 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
4313 RL would imply a DNS resolver, which is *hard* to do in a portable way
4314 (and therefore delayed).
4315 * Implement a real timer mechanism and use it in timing macros. This allows
4316 to avoid rounding errors and get a 0.000005 sec precision in timing
4317 macros. While I was at it, various cleanups:
4318 - allow to declare more than one timed section per file (fix a stupid bug)
4319 - move some private declaration to the right place
4320 - merge conditional execution and timing macros into emulation module
4321 - document the module
4322 - make sure the module cleanups its mess on gras_exit
4323 * Documentation improvements:
4324 - (new) how to compile applications using GRAS
4325 - (new) emulation support (timing macros)
4327 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
4329 SimGrid (2.93) unstable; urgency=low
4331 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
4334 - Use Paje properly where used. Still to be sanitized properly.
4335 - Portability fix: Add an implementation of the contexts using pthread
4339 - Add xbt_procname(): returns the name of the current process.
4340 Use it to show the current process's name in all logging.
4342 - fix detection of older flex version and the reaction, since we do
4343 depend on modern ones (we use lex_destroy)
4344 - Better separation of SG and RL in the libs: remove all simulation code
4345 from libgras. As a result, this lib is now only 200k when stripped.
4346 Some of the xbt modules may also be duplicated (two sets and such) and
4347 should be cleaned/killed before SG3.
4348 - Insist on using xlC on AIX because of weird problems involving gcc there.
4349 - Cleanup the make remote stuff. This is now done by scripts
4350 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
4351 mainly for our private use, but we're working on changing them to user
4354 - Bugfix: flush the socket on close only if there is some *output*.
4355 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
4357 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
4360 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
4362 SimGrid (2.92) unstable; urgency=low
4364 Alpha 3 on the path to SimGrid 3
4368 - New! First try of benchmarking macros.
4369 - New! First try so that gras_stub_generator generate deployment and
4370 remote compilation helpers.
4372 - Bugfix: Initialization fix in msg_test.
4376 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
4378 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
4380 - API Change: gras_os_sleep to take the amount of seconds as a double.
4381 Accepting an int was error prone since it was the only location where
4382 seconds were coded as such. It leaded to damn rounding errors.
4383 - Bugfix: Hard to belive that timers ever worked before this.
4385 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4387 SimGrid (2.91) unstable; urgency=low
4389 Alpha 2 on the path to SimGrid 3
4393 - Bug fix in the lmm_solver.
4395 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4396 through the function MSG_paje_output.
4397 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4398 - It is possible to bound the rate of a communication in MSG with
4399 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4400 to write it in the changelog).
4401 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4406 - Various cleanups to the autotools stuff
4407 - Begin to move Gras examples to examples/gras/
4408 - Let make distcheck work again (yeah!)
4410 - documentation overhauled using doxygen.
4411 gtk-doc-tools is dead in SimGrid now.
4412 - Automatically extract all existing logging categories, and add the list
4413 to the documentation (long standing one, to say the less)
4415 - Cleanup the known architecture table. Reorder the entries to group what
4416 should be, and use a more consistent naming scheme.
4417 (some of the test dataset are still to be regenerated)
4418 - New! Allow library to register globals on each process just as userdata
4420 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4421 lookup time (for now).
4422 Use it in msg and trp.
4423 This cleans a lot the internals and helps enforcing privacy of the
4424 headers between the gras components.
4425 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4426 - Bugfix: gras_os_time was delirious in RL.
4427 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4429 - Reenable GRAS now that it works.
4431 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4433 SimGrid (2.90) unstable; urgency=low
4435 Alpha 1 on the path to SimGrid 3
4437 * It is a long time since the last release of SimGrid. I'm sorry about
4438 that but as I had told you, I was rewriting a lot of things. I apologize
4439 to those who had been reporting bugs to me and that I had not answered.
4440 If your bug is still in the new version, please tell me. Here is a
4441 summary of the main changes.
4443 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4444 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4445 portability, speed, and a lot more but you'll figure it out later.
4446 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4447 I wanted to release the new MSG as soon as possible and I have
4448 broken GRAS, which is the reason why, for now, only MSG is fully
4449 functional. A laconic description of these projects is available
4450 in the documentation.
4452 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4453 kernel optimized for our needs (called SURF but only the developers
4454 should use it). Hence, MSG is now roughly 30 times faster and I think
4455 that by rewriting a little bit MSG, I could event speed it up a little
4456 bit more. Beside the gain in speed, it is also much easier to encode a
4457 new platform model with SURF than it was with SG. More to come...
4459 * REVOLUTION 3: I have tried to change a little as possible the API of
4460 MSG but a few things really had to disappear. The main differences
4461 with the previous version are:
4462 1) no more m_links_t and the corresponding functions. Platforms are
4463 directly read from an XML description and cannot be hard-coded
4464 anymore. The same format is used for application deployment
4465 description. The new format is described in the documentation.
4466 Have a look in tools/platform_generation. There is a tiny script
4467 that converts from the old platform format to the new one. Concerning
4468 the application deployment format, parsing the old one is tricky.
4469 I think most of you should however be able to convert your files. If
4470 it is really an issue, I can write a C code that does the conversion.
4472 2) the toolbox tbx does not exist anymore. We now have a library
4473 with much more data-structures but without the hash-tables (we have
4474 dictionaries that are much faster).
4476 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4478 *****************************************************************************
4479 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4480 * don't feel like converting the oldies. (Mt) *
4481 *****************************************************************************
4484 Version 2.90: "the long awaited one"
4485 - Finished rewriting and debugging MSG. Rewrote the documentation.
4486 - disable GRAS for now since it needs to be ported to the newest SG
4489 - Finish the port to windows (using mingw32 for cross-compile)
4492 - Main loop and datastructures of SURF. A cpu resource object is
4493 functional. Surf can thus be used to create cpu's with variable
4494 performance on which you can execute some actions.
4496 2004-11-15 Martin Quinson
4497 - Port to ARM. Simply added the alignment and size descriptions. Should
4498 work, but the ARM machines are so slow that I didn't had the opportunity
4499 to 'make check' over there yet.
4501 2004-11-15 Arnaud Legrand
4502 - Trace manager now written. It uses a heap structure and is therefore
4503 expected to be efficient. It may however be speeded up (particularly
4504 when many events occur at the same date) by using red and black
4505 trees. One day maybe...
4506 - Max-min linear system solver written. It uses a sparse matrix
4507 structure taking advantage of its expected use. Most operations are
4508 O(1) and free/calloc are called as few as possible. The computation of
4509 the minimum could however be improved by using a red and black tree
4512 2004-11-03 Arnaud Legrand
4513 - Rename every gras_* function that was in xbt/ to its xbt_
4515 - Add a heap and a doubly-linked list to xbt
4516 - Added a dichotomy to the dictionaries. make check works as well before
4517 so I assume that the patch is correct. I do not know however if things
4518 run effectively faster than before now. :)
4520 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4521 SimGrid, and the version number is bumped to 2.x
4523 2004-10-29 Martin Quinson
4524 - Introduction of the remote errors.
4525 They are the result of a RMI/RPC on the remote machine.
4526 ErrCodes being scalar values, you can't get the host on which those
4527 errors did happen. Extending the error mechanism as in Gnome is possible.
4528 No idea yet whether it is a good idea.
4530 2004-10-28 Martin Quinson
4531 - Interface revolution: the Starred Structure Eradication.
4532 I used to do typedef struct {} toto_t; and then handle *toto_t.
4533 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4534 - ???_t is a valid type (builded with typedef)
4535 - s_toto_t is a structure (access to fields with .)
4536 - s_toto is a structure needing 'struct' keyword to be used
4537 - e_toto_t is an enum
4538 - toto_t is an 'object' (struct*)
4540 typedef struct s_toto {} s_toto_t, *toto_t;
4541 typedef enum {} e_toto_t;
4542 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4543 s_toto_t) is private.
4545 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4546 so that it can set the variable to NULL. It was so for dicts and sets,
4547 it changed for dynars.
4549 - Fix a bunch of memleaks in dict_remove
4550 - Fix a bug in sg/server_socket opening: it failed all the time.
4552 2004-10-07 Martin Quinson
4553 - Speed up dynar lookup operation a bit.
4555 gras_dynar_get is dead.
4557 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4558 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4559 (which gives you the address of the stored data).
4561 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4562 copy of the data using an affectation to do the job and not a memcpy.
4564 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4565 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4567 It does not really speedup the dynar test because they are
4568 setting elements all the time (and look them seldom). But the dict does
4569 far more lookup than setting.
4571 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4573 2004-10-05 Martin Quinson
4574 - Allow to (en/dis)able the cycle detection at run time.
4576 Whether we should check for cycle or not is now a property of each
4577 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4578 datadesc_cycle_unset allow to remove this property when previously set.
4580 Note that the cycle detection is off by default since it impacts the
4581 performance. Watch the data you feed GRAS with ;)
4583 This property is hereditary. Any element embedded in a structure having it
4584 set have it set for the time of this data exchange.
4586 You should set it both on sender and receiver side. If you don't set it on
4587 sender side, it will enter an endless loop. If you forget on receiver
4588 side, the cycles won't be recreated after communication.
4590 - Header reorganization.
4591 Kill gras_private.h, each submodule must load the headers it needs.
4593 2004-10-04 Martin Quinson
4594 - Interface revolution: do not try to survive to malloc failure.
4596 Now, gras_malloc and friends call gras_abort() on failure.
4597 As a conclusion, malloc_error is not a valid error anymore, and all
4598 functions for which it was the only gras_error_t return value are
4599 changed. They now return void, or there result directly.
4600 This simplify the API a lot.
4602 2004-09-29 Martin Quinson
4603 - Re-enable raw sockets.
4604 Created by gras_socket_{client,server}_ext;
4605 Used with gras_raw_{send,recv}
4608 It should allow to kill the last bits of gras first version soon.
4610 This is not completely satisfactory yet (duplicate code with
4611 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4614 - Simplify transport plugin (internal) interface by not passing any
4615 argument to _server and _client, but embedding them in the socket
4618 2004-09-28 Martin Quinson
4619 - Finish the port to AIX.
4620 autoconf was my problem (segfault within the malloc replacement
4621 function. No idea why)
4623 2004-09-16 Martin Quinson
4624 - Fix some size_t madness on 64bit architectures.
4626 2004-09-08 Martin Quinson
4627 - Reduce the number of system headers loaded, overload some more system
4628 calls (such as malloc to cast the result of the system one, and work
4630 - Fix and reintroduce the config support
4632 2004-09-07 Martin Quinson
4633 - Source code reorganization to allow Arnaud to surf all over there.
4634 - Allow to document the logging categories.
4635 - Remove all uppercase from logging categories and useless cleanup in names.
4637 2004-08-18 Martin Quinson
4638 Version 0.6.2 (protocol not changed; API changed)
4639 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4640 gras_error_t), and NULL when not found. Functions expecting a msgtype
4641 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4642 hopefully usefull message.
4643 - Portability to prehistoric sparcs again
4645 2004-08-17 Martin Quinson
4646 Version 0.6.1 (protocol not changed; ABI not changed)
4647 - prealloc some buffers to speed things up
4649 2004-08-11 Martin Quinson
4650 Version 0.6 (protocol not changed; ABI expended)
4651 - The parsing macro can deal with the references, provided that you add
4652 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4654 2004-08-09 Martin Quinson
4655 Version 0.5 (protocol not changed; ABI changed)
4656 - Allow to off turn the cycle detection code in data exchange at
4657 compilation time. It should be at run time, but I'm short of time (and
4658 the config stuff is still broken). That way, we keep dict out of the
4659 critical path, which is good because the performance is poor:
4660 - search not dichotomial yet
4661 - dynar give no way to access their content and memcpy everytime
4662 - In composed data description (struct, ref and so on), stop foolness of
4663 keeping the subtype's ID, but store the type itself. This keeps sets out
4664 of the critical path, which is good since they rely on dynar and
4665 dictionaries. The only loose of that is that we cannot detect the
4666 redeclaration of a structure/union with another content (but I'm not sure
4667 the code detected well this error before anyway). We still can detect
4668 the redefinition discrepancy for the other types.
4669 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4670 it breaks the code because of type-punning used all over the place).
4671 This breaks on all non-gcc architectures (for now).
4673 All those changes (plus the buffer of last time) allow me to gain 2 order
4674 of magnitude on cruel tests consisting of 800000 array of integers on two
4675 level of a hierarchical structure (200 secondes -> 4 secondes)
4678 - the selector of reference must now return the type it points to, not
4679 the ID of this type.
4681 2004-08-06 Martin Quinson
4682 Version 0.4 (protocol changed; ABI not changed)
4683 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4684 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4685 - Implement a buffer transport, and use it by default (it relies on tcp in
4686 real life and on sg in simulation).
4687 That's a bit hackish since I had a new field to the structure to store
4688 its data without interfering with the subtype ones. Inheritance
4689 is tricky in C. And that's a kind of reverse inheritance with one class
4690 derivating two classes. Or maybe a game with java interfaces. Anyway,
4691 that's damn hard in C (at least).
4692 Moreover, I got tired while trying to ensure plugin separation and
4693 genericity in SG mode. MSG wants me to do weird things, so let's go for
4694 cruel hacks (temporarily of course ;).
4695 See comment in transport_private.h:71
4696 - do not include all the _interface headers in private but in the files
4697 which really need them (to cut the compilation time when they are
4700 2004-07-26 Martin Quinson
4701 Version 0.3 (protocol not changed; ABI changed)
4702 - Major overhault of the datadesc interface to simplify it:
4703 - shorted the function names:
4704 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4705 - add a trivial way to push/pop integers into the cbps without malloc.
4706 This allows to make really generic sub_type description, which simply
4707 pop their size of the stack.
4708 - add a function gras_datadesc_ref_pop_arr() which does what users want
4709 most of the time: Declare a dynamic array (which pops its size of the
4710 stack) and declare a reference to it. Poor name, but anyway.
4711 - kill the post-send callback, add a post-receive one
4713 2004-07-23 Martin Quinson
4714 Version 0.2 (protocol changed; ABI changed)
4715 - add some testing for cpbs in the test cases, and fix some more bugs.
4716 This invalidate again the little64 data file, since I cannot regenerate
4718 - remove an awfull optimization in the logging stuff, allowing me to:
4719 - understand it again
4720 - learn gcc how to check that the argument match the provided format
4721 - fix all errors revealed by gcc after that
4722 - internal keys of dict are not \0 terminated. Deal with it properly in
4723 loggings instead of segfaulting when the user want to see the logs :-/
4725 2004-07-22 Martin Quinson
4726 - Fix some stupid bug preventing cbps (callback postit) from working
4728 2004-07-21 Martin Quinson
4729 - Some documentation cleanups
4730 - remove the useless last argument of msgtype_declare
4731 - rename the Virtu functions to fit into the 'os' namespace
4732 - move headers src/include -> src/include/gras/ and stop fooling with
4733 gras -> . symbolic link
4734 - make distcheck is now successful
4736 2004-07-19 Martin Quinson
4738 - Build shared library also
4739 - Install html doc to the right location
4740 - stop removing maintainer files in make clean
4741 - build tests only on make check
4743 2004-07-13 Martin Quinson
4745 - No major issue in previous version => change versioning schema
4746 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4747 corresponding dataset.
4749 2004-07-11 Martin Quinson
4751 - Get it working with any kind of structure (we can compute the padding
4752 bytes remotely for all the architectures I have access to)
4753 - Implement the structure parsing macro (still not quite robust/complete)
4754 - Improvement to the remote testing toysuite
4756 2004-07-10 Martin Quinson
4757 [autoconf mechanism]
4758 - get ride of a bunch of deprecated macros
4759 - actually run the test for two-compliment, not only compile it :-/
4760 - test whether the structures get packed (and bail out if yes. Damn.
4761 Alignment is a serious matter)
4762 - test whether the structures get compacted (but respecting the alignment
4763 constraints of each types)
4764 - test whether the array fields of structures can straddle alignment boundaries
4766 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4767 (took me 2 hours to find that bug, looking at the wrong place)
4768 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4769 before sending/receiving and are used to compute the offsets of fields
4770 - Given that padding size depend even on compiler options, keep track of
4771 alignment and aligned_size only for the current architecture. Not a big
4772 deal since we send structure fields one after the other (seems
4774 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4775 program, let it work on linux/gcc/little32. portability todo.
4777 2004-07-08 Martin Quinson
4778 - import and improve remote compilation support from FAST
4779 - make sure make check works on half a dozen of machines out there
4781 2004-07-07 Martin Quinson
4782 Let's say it's version 0.0.3 ;)
4783 - Implement conversions (yuhu!)
4784 - Let it work on solaris (beside conversion, of course)
4785 - Stupid me, using rand() to generate the conversion datatests in not wise.
4787 2004-07-06 Martin Quinson
4788 - Let make dist work, since I'm gonna need it to compile on remote hosts
4789 - Let Tests/datadesc_usage write the architecture on which the file was
4790 generated as first byte.
4791 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4792 architecture descriptions.
4793 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4794 on those architectures.
4795 - Optimization: send/recv array of scalar in one shoot
4797 2004-07-05 Martin Quinson
4798 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4800 - Plug a whole bunch of memleaks
4801 - each process now have to call gras_{init,exit}. One day, their log
4802 settings will be separated
4803 - Continue the code factorisation between SG, RL and common in Transport.
4805 2004-07-04 Martin Quinson
4807 - Redistribution between SG and RL.
4808 We wanna have to accept in SG, so move accepted related parts of RL in
4809 the common part. (more precisely, the dynar of all known sockets is no
4810 more a static in transport.c, but part of the process_data)
4812 [gras_stub_generator]
4813 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4815 2004-07-03 Martin Quinson
4816 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4817 them all up in one shot)
4819 - Fix the ugly case of reference to dynamic array.
4820 - New (semi-public) function gras_datadesc_size to allow the messaging
4821 layer to malloc the needed space for the buffer.
4823 - gras_socket_close now expect the socket to close (and not its address to
4824 put NULL in it after it). This is because the socket passed to handlers
4825 is one of their argument (=> not writable).
4827 - propagate the interface cleanup from last week in datadesc, ie remove a
4828 superfluous level of indirection. User pass adress of variable
4829 containing data (both when sending and receiving), and not of a variable
4830 being a pointer to the data. Let's say that I like it better ;)
4831 The price for that is constructs like "int msg=*(int*)payload" in
4832 handlers, but it's a fine price, IMHO.
4834 - Let it work in RL (yuhu)
4836 2004-06-21 Martin Quinson
4838 - porting SG plugin and SG select to new standards (works almost).
4839 - plug memleaks and fix bugs around.
4842 - cleanup the prototype of data recv and force users to specify when they
4843 want to handle references to objects. Test case working even for cycles.
4844 - plug memleaks. Valgrind is perfectly ok with this.
4846 2004-06-12 Martin Quinson
4848 - cleanup the separation between plugin and main code in plugin creation
4850 2004-06-11 Martin Quinson
4852 - Reput hook for raw sockets, needed for BW experiments
4853 - kill a few lines of dead code
4854 [Data description] Interface cleanup
4855 - gras_datadesc_by_name returns the searched type or NULL.
4856 That way, no variable is needed to use a type desc once, which makes
4858 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4859 two parameters were strings (field name, type name), leading to
4861 [Dicos] Interface cleanup
4862 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4863 This is consistant with the dynar API.
4865 2004-04-21 Martin Quinson
4867 - Porting to new standards.
4869 - interface cleanup.
4870 There is no bag anymore, no need to take extra provision to mask the
4871 pointers behind "ID".
4872 Better splitup of functions between files create/exchange/convert.
4873 This is still a bit artificial since convert and receive are so
4874 interleaved, but anyway.
4876 - add a queued message list to procdata (the ones not matching criteria
4878 - factorize some more code between SG and RL wrt procdata
4880 - use gras_exit in example to track memleaks
4881 - get rid of gs_example now that GS is properly integrated into gras
4882 - update run_test to integrate the latest tests (datadesc)
4884 - rename WARNINGn macros to WARNn since it prooved error-prone
4886 2004-04-19 Martin Quinson
4888 - register init/exit functions within gras module mechanism
4889 - send/receive function.
4890 Convertion is not implemented, but short-cutted if not needed.
4891 struct/array elements are sent one by one (instead of block-wise), but
4892 nobody really cares (yet). Get a prototype before optimizing.
4893 - tests (using a file socket) for DD send/receive on:
4894 - base types: int, float
4895 - array: fixed size, string (ie ref to dynamic string)
4896 - structure: homogeneous, heterogeneous
4897 - chained list, graph with cycle
4898 Believe it or not, valgrind is not too unhappy with the results. The
4899 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4903 - Bugfix when using a filename explicitly (instead of '-')
4905 2004-04-09 Martin Quinson
4907 - factorize more code between RL and SG in socket creation
4908 - Complete the implementation and tests of:
4910 o file (only in RL, and mainly for debugging)
4912 I lost 3 days to design a portable address resolver, and then decided
4913 that the prototype mainly have to run on my box.
4914 Addressing portability too early may be like optimizing too early :-/
4916 - use gras_init in the Tests instead of the crappy parse_log_opt
4917 (the latter function is removed)
4918 [Conditional execution]
4919 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4920 [Code reorganisation]
4921 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4922 Build examples against the RL library, since there is no way to disable
4923 its creation for now.
4925 For information, the beginning of coding on GRAS was back in june
4926 2003. I guess that every line has been rewritten at least twice since