1 SimGrid (3.34.1) not released (Target: fall 2023)
4 - Make the host_load plugin available from Python. See examples/python/plugin-host-load
6 ----------------------------------------------------------------------------
8 SimGrid (3.34) June 26. 2023
10 Save the planet, skip a release: 3.33 was due 6 months ago, so skip directly to 3.34.
13 - SimGrid now requires a compiler with C++17 support for public headers too.
14 Sibling projects should upgrade their FindSimGrid.cmake
15 - Remove the MSG API: its EOL was scheduled for 2020.
16 - Remove the Java bindings: they were limited to the MSG interface.
17 - On Windows, you now need to install WSL2 as the native builds are now disabled.
18 It was not really working anyway.
19 - Support for 32bits architecture is not tested anymore on our CI infrastructure.
20 It may break in the future, but we think that nobody's using SimGrid on 32 bits.
21 - Remove the surf module. It was replaced by the kernel/models module, and that
22 refactoring took almost 10 years to properly complete.
25 - Activity::set_remaining() is not public anymore. Use for example
26 Comm::set_payload_size() to change the size of the simulated data.
27 - New function: Engine::flatify_platform(), to get a fully detailed vision of the
29 - New Task abstraction: They are designed to represent dataflows, i.e, graphs of repeatable Activities.
30 See the examples under examples/cpp/task-* and the associated documentation.
31 - Full simDAG integration: Activity::start() actually starts only when all dependencies
32 are fulfilled. If it cannot be started right away, it will start as soon as it becomes
34 - Allow to set a concurrency limit on disks and hosts, as it was already the case for links.
35 - Rename Link::get_usage() to Link::get_load() for consistency with Host::
36 - Every signal now come with a static version that is invoked for every object of that class,
37 and an instance version that is invoked for this specific object only. For example,
38 s4u::Actor::on_suspend_cb() adds a callback that is invoked for the suspend of any actor while
39 s4u::Actor::on_this_suspend_cb() adds a callback for this specific actor only.
40 - Activity::on_suspended_cb() is renamed to Activity::on_suspend_cb(), and fired right before the suspend.
41 - Activity::on_resumed_cb() is renamed to Activity::on_resume_cb(), and fired right before the resume.
42 - Resource::on_state_change_cb() is renamed to Resource::on_onoff_cb() to distinguish from the
43 Activity::on_state_change_cb() that is related to the activity state machine, not on/off.
44 - Activity signals (veto, suspend, resume, completion) are now specialized by activity class.
45 That is, callbacks registered in Exec::on_suspend_cb will not be fired for Comms nor Ios.
48 - Battery: Enable the management of batteries on hosts.
49 See the examples under examples/cpp/battery-* and the documentation in the Plugins page.
50 - Photovoltaic: Enable the management of photovoltaic panels on hosts.
51 See the examples under examples/cpp/photovoltaic-* and the documentation in the Plugins page.
54 - optimize an internal data structure (use a set instead of a list for ongoing activities),
55 leading to a potentially big performance gain, in particular with many detached comms.
58 - New option smpi/barrier-collectives to add a barrier to some collectives
59 to detect dangerous code that /may/ work on some MPI implems.
60 - New function SMPI_app_instance_start() to easily start a MPI instance in your S4U simulation.
63 - Write the section of the manual about models, at least.
64 - WiFi: the total capacity of a link depends on the amount of flows on that link.
65 - Use the nonlinear callback feature of LMM to reflect this.
66 - Calibration values can be changed to match different MCS configurations
67 - See the example teshsuite/models/wifi_usage_decay/wifi_usage_decay.cpp
68 - See also "A Flow-Level Wi-Fi Model for Large Scale Network Simulation"
69 https://hal.archives-ouvertes.fr/hal-03777726
70 - Merge parameters network/bandwidth-factor and smpi/bw-factor that serve the same purpose.
71 - Same for the latency
72 - Rewrite the corresponding documentation.
73 - Allow to disable the TCP windowing modeling by setting network/TCP-gamma to 0.
74 - Finally kill the 'compound' host model. You can change the CPU or network model
75 with the default host model, as it should.
76 - Rename option "surf/precision" to "precision/timing" for clarity.
77 - Rename option "maxmin/precision" to "precision/work-amount" for clarity.
78 - New function: Engine::flatify_platform() to debug your platform.
81 - Implement pthread_join in MC mode.
82 - Implement semaphore functions in sthread.
83 - Add an intricated way to verify the access to non-reentrant data structures
84 It requires code annotation, as shown in examples/sthread/stdobject/stdobject.cpp
87 - Stateless model-checking is now usable on any system, including Mac OSX and ARM processors.
88 - The stateless aspects of the MC are now enabled by default in all SimGrid builds.
89 Liveness and stateful aspects are still controlled by the enabling_model-checking
91 - Introducing ODPOR and SDPOR reduction strategies
92 - Introducing guiding heuristics, trying to find bugs faster than DFS in reduced state space.
93 - Synchronize the MBI tests with upstream.
94 - Show the full actor backtraces when replaying a MC trace (with model-check/replay)
95 and the status of all actors on deadlocks in MC mode.
98 - simgrid::xbt::cmdline and simgrid::xbt::binary_name are gone.
99 Please use simgrid::s4u::Engine::get_cmdline() instead.
102 - New tutorial on simulating DAGs.
103 - New section in the user guide on the provided performance models.
104 - New section presenting some technical good practices for (potential) contributors.
105 - Add a section on errors and exceptions to the API documentation.
106 - Move the s4u examples to a section on their own to ease navigation.
108 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
109 (FG: issues on Framagit; GH: issues on GitHub)
110 - FG#18: Java bindings should be redone or removed
111 - FG!118: Wi-Fi callback mechanism
112 - FG!119: SMPI: add option to inject a barrier before every collective call
113 - GH#383: Segfault when adding a disk after load_platform(xml)
115 ----------------------------------------------------------------------------
117 SimGrid (3.32) October 3. 2022.
119 The Wiedervereinigung release. Germany was reunited 32 years ago.
122 - SimGrid now requires a compiler with C++17 support to compile the lib.
123 Our public headers still allow the user code to be compiled in C++14.
124 - Support graphviz v3 and ns-3 v3.36 (older versions are still supported).
125 - Tested with clang (v11, v13, v14 and v16), gcc (v7 to v13) and IntelCC v2022.2
129 - kill signal Comm::on_completion that was not working anyway.
130 - Expose signals Activity::on_suspend and Activity::on_resume
131 - New macro xbt_enforce(): similar to xbt_assert(), but throws an AssertionError
132 instead of calling abort().
133 - New: s4u::Exec::get_thread_count()
134 - Various cleanups around virtual machines:
135 - host_by_name() and friends now only return hosts. VMs are now excluded.
136 - It is now impossible to search a VM by name globally.
137 You can only search VM by name on a given PM, so either you know
138 the PM on which your VM runs and you can search by name, or you need
139 to manually iterate over all PMs to search this VM.
140 - The s4u::VirtualMachine constructor is now deprecated.
141 Please use s4u::Host::create_vm() instead.
142 - Rename s4u::VirtualMachine::on_creation() to on_vm_creation() to
143 avoid confusion with s4u:Host::on_creation() that is inherited.
144 Also s4u::VirtualMachine::on_destruction -> on_vm_destruction().
146 - One-sided communications (Comm::sendto) can now be detached,
147 and should now be more resilient to network and host faults.
150 - Added the following bindings / examples:
151 - Comm (now 100% covers the C++ interface):
152 - Comm.dst_data_size, Comm.mailbox, Comm.sender, Comm.start_time, Comm.finish_time
153 - Comm.state_str [examples: examples/python/comm-failure/, examples/python/comm-host2host/]
154 - Comm.remaining [examples: examples/python/comm-host2host/, examples/python/comm-suspend/]
155 - Comm.set_payload_size [example: examples/python/comm-host2host/]
156 - Comm.set_rate [example: examples/python/comm-throttling/]
157 - Comm.sendto, Comm.sendto_init, Comm.sendto_async [example: examples/python/comm-host2host/]
158 - Comm.start, Comm.suspend, Comm.resume [example: examples/python/comm-host2host/]
159 - Comm.test_any [example: examples/python/comm-testany/]
160 - Comm.wait_until [example: examples/python/comm-waituntil/]
162 - Engine.host_by_name [example: examples/python/comm-host2host/]
163 - Engine.mailbox_by_name_or_create [example: examples/python/comm-pingpong/]
165 - Mailbox: Mailbox.ready [example: examples/python/comm-ready/]
166 - Ptask [example: examples/python/exec-ptask/]:
167 - this_actor.exec_init
168 - this_actor.parallel_execute
171 - Added an AssertionError exception that may be thrown in case of error.
172 For instance, creating two hosts with the same name will now throw this exception
173 instead of killing the interpreter.
176 - Implement MPI_File_get_type_extent(), MPI_File_s/get_atomicity() and
177 MPI_File_get_byte_offset()
178 - Intercept getpid() calls to return the simulated ones.
179 - Fix various bugs in MPI IO.
181 Platform description & visualization:
182 - More robust sanity checks for platforms, to reject forbidden topologies with
183 a proper error message.
184 - New platform example: supernode.cpp and supernode.py.
185 The Python version generates a nice graphical representation of the platform.
186 - Bug fixes around fat-tree topologies.
187 - Allow to dump the platform topology as a CSV file representing the graph edges
188 with platform_graph_export_csv() (similar to the DOT export).
189 - Fix graphicator for "cluster" topologies (e.g. fat-tree, dragonfly).
192 - Fix a bug when using ptasks with multicores (FG!111).
195 - First bits of sthread, that intercepts pthread operations at runtime.
196 The intend is to use it together with simgrid-mc, but it is TBD.
197 - Sync MBI generators with upstream changes.
198 - Various cosmetics, small bug fixes and inner refactorings
200 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
201 (FG: issues on Framagit; GH: issues on GitHub)
202 - FG#105: "Variable penalty should not be negative!" with in-flight messages and bandwidth profiles
203 - FG#109: Application time reported by --cfg=smpi/display-timing:yes is wrong
204 - FG#110: Wait_any does not trigger new model solve when host events occur
205 - FG#111: Wrong execution time in rare cases when using multicore
206 - FG!98: Re-enable the tests for legacy stochastic profiles
207 - FG!109: Trigger new engine solve upon host events such as host on/off
208 - FG!116: SMPI/replay: Fix issue with recv of size =0
210 ----------------------------------------------------------------------------
212 SimGrid (3.31) March 22. 2022.
214 The ненасильство release. We stand against war.
216 Against the agression by a sick system that forces peoples to take arms against each other.
219 - Rework the internals, for simpler and modern code. This shall unlock many future improvements.
220 - You can now define plugins onto the DFS explorer (previously called SafetyChecker), using the
221 declared signals. See CommunicationDeterminism for an example.
222 - Support mutex, semaphore and barrier in DPOR reduction
223 - Seems to work on Arm64 architectures too.
224 - Display a nice error message when ptrace is not usable.
225 - New test suite, imported from the MPI Bugs Initiative (MBI). Not all MBI generators are integrated yet.
226 - Remove the ISP test suite: it's not free software, and it's superseeded by MBI.
229 - fix for FG#100 by ensuring small asynchronous messages never overtake larger
230 ones, conforming to the standard.
231 - replay: fix waitall behaviour to avoid forgetting requests and leaking
233 - tracing: ensure that we dump the TI traces continuously during execution and
234 not just at the end, reducing memory cost and performance hit.
235 - Update OpenMPI collectives selection logic to match current one (4.1.2)
236 - Add a coherence check for collective operation order and root/MPI_Op
237 coherence. Potentially costly so not activated unless smpi:pendantic is set
238 or -analyze is given.
241 - New signal: Engine::on_simulation_start_cb()
242 - Introduce a new execution mode with this_actor::thread_execute(). This simulate
243 the execution of a certain amount of flops by multiple threads ran by a host. Each
244 thread executes the same number of flops, given as argument. An example of this new
245 function can be found in examples/cpp/exec-threads.
246 - Reimplementation of barriers natively.
247 Previously, they were implemented on top of s4u::Mutex and s4u::ConditionVariable.
248 The new version should be faster (and can be used in the model-checker).
249 - Actor::get_restart_count(): Returns the number of reboots that this actor did.
252 - MSG_barrier_destroy now expects a non-const msg_barrier parameter.
254 New plugin: the Chaos Monkey (killing actors at any time)
255 - Along with the new simgrid-monkey script, it tests whether your simulation
256 resists resource failures at any possible timestamp in your simulation.
257 - It is mostly intended to test the SimGrid core in extreme conditions,
258 but some users may find it interesting too.
261 - New solver for parallel task: BMF.
262 - More realistic sharing of heterogeneous resources compared to the fair
263 bottleneck solver used by ptask_L07.
264 - Implement the BMF (Bottleneck max fairness) fairness.
265 - Improved resource sharing for parallel tasks with sub-flows (parallel
266 communications between same source and destination inside the ptask).
268 - "--cfg=host/model:ptask_L07 --cfg=host/solver:bmf": enable the ptask
269 model with BMF solver.
270 - "--cfg=bmf/max-iterations: <N>": maximum number of iterations performed
271 by BMF solver (default: 1000).
272 - "--cfg=bmf/precision: <N>": numerical precision used when computing
273 resource sharing (default: 1e-12).
274 - This model requires Eigen3 library. Make sure Eigen3 is installed to use BMF.
277 - Modifications of the Profile mechanism, with some impact on users
278 - Addition of a new (S4U) method to init profiles from generic functions to improve versatility
279 - Fix initial behaviour of state_profiles
280 - Modify periodicity to behave like a period, and not like a loop delay
283 - Drop xbt_dynar_shrink().
286 - Made the following bindings static (previously member functions):
287 - Actor: Actor.kill_all(), Actor.by_pid()
288 - Host: Host.by_name(), Host.current(), Host.on_creation_cb()
289 - Mailbox: Mailbox.by_name()
290 - Added the following bindings:
291 - this_actor.warning()
292 - Mailbox.put_init() [example: examples/python/comm-waitallfor/]
293 - Comm.detach() [example: examples/python/comm-waitallfor/]
294 - Comm.wait_for() [example: examples/python/comm-waitfor/]
295 - Comm.wait_any_for()
296 - Comm.wait_all_for() [example: examples/python/comm-waitallfor/]
297 - Mutex [example: examples/python/synchro-mutex/]
298 - Barrier [example: examples/python/synchro-barrier/]
299 - Semaphore [example: examples/python/synchro-semaphore/]
302 - Remove target "make uninstall" which was incomplete and no longer maintained.
304 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
305 (FG: issues on Framagit; GH: issues on GitHub)
306 - FG#57: Mc SimGrid should test whether ptrace is usable
307 - FG#87: Smpi scripts fail with spaces in paths
308 - FG#100: [SMPI] Order of the message matching is not guaranteed
309 - FG#101: LGPL 2.1 is deprecated license
310 - FG#104: "make uninstall" not up-to-date
311 - GH#151: Missing mutexes for DPOR.
313 ----------------------------------------------------------------------------
315 SimGrid (3.30) January 30. 2022.
317 The Sunday Bloody Sunday release.
319 Main user-visible changes:
320 - The SimDag API for the simulation of the scheduling of Directed Acyclic
321 Graphs has been dropped. It was marked as deprecated for a couple of years.
322 We finally complete the implementation of what has been called SimDag++
323 internally, i.e., porting the different features of SimDag on top of S4U.
324 The new way to simulate the execution of dependent activities directly by
325 maestro (without any other actor) is details in the examples/cpp/dag-* series
327 - The removal of SimDag led us to also remove the export to Jedule files that
328 was tightly coupled to SimDag. The instrumentation of DAG simulation is still
329 possible through the regular instrumentation API based on the Paje format.
330 - We also dropped the old and clumsy Lua bindings to create platforms in a
331 programmatic way. It can be done in C++ in a much cleaner way now, which
332 motivates this suppression.
335 - Introduce on_X_cb() functions for all signals, to attach a new
336 callback to the signal X. The signal variables are now hidden and
337 only these functions should be used.
338 Rationale: this enables the usual deprecation scheme where functions
339 remain for 4 releases if we need to modify the signals, while the
340 current code with the signal variables directly visible prevents any
342 - New function: Engine::run_until(date), to split the simulation.
343 - New signal: Activity::on_veto, to detect when an activity fails to start.
344 - Signal change: Comm::on_start(Comm&, bool) has been replaced by
345 Comm::on_send and Comm::recv. These two signals respectively correspond to
346 when the sending or receiving side of a Comm is ready. They are raised at
347 the same locations as the former Comm::on_start signal.
348 - New function: Engine::track_vetoed_activities() to interrupt run()
349 when an activity fails to start, and to keep track of such activities.
350 Please see the corresponding example for more info.
351 - New functions: s4u::Comm::{sendto_init, set_source, set_destination} to enable
352 the use of vetoers with direct host-to-host communications. Both source and
353 destination have to set for a comm to start. Each call to these setters check
354 if all vetoes are satisfied. When it is the case, the comm starts. A use case of
355 these functions is given in examples/cpp/dag-scheduling.
356 - New functions: {Exec, Io}::update_priority allow you to modify the priority of
357 these kinds of activities during their execution. Behavior is detailed in
358 examples/cpp/io-priority/.
361 - Dynamic costs for MPI operations: New API to allow users to dynamically
362 change injected costs for MPI_Recv, MPI_Send and MPI_Isend operations.
363 Alternative for smpi/or, smpi/os and smpi/ois configuration options.
364 - Fix some issues with the replay mechanism.
367 - Function xbt::Extendable::get_data() is now templated with the type of the
368 pointee. Untyped function is deprecated. Use get_data<void>() if you still
369 want to retrieve void*.
372 - New section: "SimGrid MPI calibration of a Grid5000 cluster"
373 presenting how to properly calibrate MPI communications in SimGrid.
374 - Complete and reword the platform section, which is now completed.
377 - Thread contexts are used by default with Python bindings. Other kinds of
378 contexts revealed unstable, specially starting with pybind11 v2.8.0.
380 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
381 (FG: issues on Framagit; GH: issues on GitHub)
382 - FG#95: Wrong computation time for multicore execution after pstate change
383 - FG#97: Wrong computation time for ptask+multicore+pstates
384 - FG#98: SMPI offline simulation is inconsistent with the online simulation
385 (deadlocks / message truncation)
386 - FG#99: Weird segfault when not sealing an host
388 ----------------------------------------------------------------------------
390 SimGrid (3.29) October 7. 2021
392 The "Ask a stupid question" release.
394 We wish that every user ask one question about SimGrid to celebrate.
395 On Mattermost, Stack Overflow or using the issues tracker.
397 New modeling features:
398 - Non-linear resource sharing, modeling resources whose performance heavily degrades with contention:
399 - The total capacity may be updated dynamically through a callback
400 and depends mainly on the number of concurrent flows.
401 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
403 - Dynamic factors: model variability in the speed of activities
404 - Each action can now have a factor that affects its progression.
405 This multiplicative factor is applied when updating the amount of work
406 remaining, thereby an activity with factor=0.5 only uses half of the
407 instantaneous power/bandwidth it is allocated and will appear twice
408 slower than what it actually consumes.
409 - This can be used to model a overhead (e.g., there is a 20 bytes
410 header in a 480 bytes TCP packet so the factor 0.9583) but the novelty
411 is this factor can now easily be adjusted depending on activity's and
412 resources characteristics.
413 - This existed for network (e.g., the effective bandwidth depends
414 on the message in SMPI piecewise-linear network model) but it is now
415 more general (the factor may depend on the source and destination and
416 thus account to different behaviors for intra-node communications and
417 extra-node communications) and is available for CPUs (e.g., if you
418 want to model an affinity as in the "Unrelated Machines" problem in
419 scheduling) and disks (e.g., if you want to model a stochastic
421 - For that, resources can be provided with a callback that computes
422 the activity factor when creating the action.
423 - Example: examples/cpp/exec-cpu-factors
424 - The same mechanism is also available for the latency, which
425 allows to easily introduce complex variability patterns.
428 - Added support to programmatic platform creation in Python.
429 Example: examples/python/clusters-multicpu
432 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
433 This can only be set through the API, not through XML files.
437 - Multiple fixes to ensure reproducibility of tracing
438 - scan/excan can now be replayed
439 - wait action now uses ranks and not pid, as the other ones.
440 - smpi/init and smpi/finalization-barrier are now valid for replays.
441 - exit() is now intercepted by SMPI to avoid premature shutdown of
442 simulation. First non 0 return codes is returned as simulation return
446 * New section "Release Notes" documenting recent and current developments.
447 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
448 * Improvements in API Reference for C++ and Python interfaces.
451 - Make wifi creation compatible with ns-3 version 3.34 too.
453 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
454 (FG: issues on Framagit; GH: issues on GitHub)
455 - FG#77: Search feature of doc is broken (update sphinx theme version)
456 - FG#78: Multiple fixes for SMPI replay:
457 - TI tracing of allotallv/w was outputting wrong values
458 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
460 ----------------------------------------------------------------------------
462 SimGrid (3.28) July 14. 2021
464 The Victoriadagarna Release.
467 - C++ platform interface: Users can now describe their platform directly in C++.
468 This provides greatly flexibility and performance improvement for complex
469 platforms. Main features:
470 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
471 "zones", instead of single hosts. This feature enables the description
472 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
473 - StarZone: new zone with a Star-like topology. The routes are defined
474 as a set of links used to communicate from node to everybody (node<->ALL).
475 - Split-Duplex links: auxiliary method to create split-duplex links in
476 the platform, easing its utilisation. It automatically creates both UP
477 and DOWN links (similarly as done in XML).
478 - Please refer to the documentation and the examples included:
479 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
480 - New plugin: Producer-Consumer with monitor. Just requires to include the
481 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
482 associated example (examples/cpp/plugin-prodcons).
485 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
486 s4u::Io::wait_any(), s4u::Io::wait_any_for().
487 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
488 parameter by reference, instead of a pointer.
489 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
490 Explicitly cancel the activity to get back to previous behavior.
491 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
494 - The default SMPI compiler flags are no more taken from the environment.
495 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
496 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
498 --cfg=smpi/finalization-barrier: which can be used to add
499 a barrier inside MPI_Finalize. This can help for some codes which cleanup
500 data attached to a process, but still used in other SMPI processes.
501 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
502 is active when set to false, to keep going after a small error
503 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
504 which may or may not be problematic in the end.
506 - fix behaviour, as maximum iteration count could be ignored
507 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
508 use sampling when the same kernel is called with a different set of parameters
509 which have an impact on the timing.
510 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
511 It should now work with smpi/auto-shared-malloc-thresh.
512 - Improve error handling and reporting in multiple places
513 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
514 validity checks, truncated messages are now an error, return errors
515 when explicitely deleted handles are reused, ...)
516 - RMA: multiple fixes and stability improvements.
517 - analysis (-analyze flag in smpirun):
518 - SMPI can now report buffer leaks as well as MPI handles leaks,
519 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
520 - if -trace-call-location is used when compiling, SMPI can report
521 origin of leaked handles/buffers
522 - group leaks by type/origin in output message if possible
523 - New implemented MPI calls: MPI_Comm_test_inter
526 - Changed internal implementation of bandwidth factors in network models.
527 Models affected: CM02, LV08 (default), SMPI, IB.
528 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
529 Bandwidth factors are applied to communications to describe that users
530 cannot use 100% of the available bandwidth. For example, the default network model,
531 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
532 behavior was implemented by limiting the bandwidth available in the LMM
533 system for this flow. This may give the false impression that there is
534 bandwidth available for other flows due to its underutilization, especially
535 for the dynamic bandwidth factors used in SMPI models.
536 To avoid this, we have modified the implementation so that each flow uses the
537 maximum physical bandwidth according to the LMM system.
538 However, the actual throughput of the flow seen by the user is defined by
539 the physical bandwidth multiplied by the bandwidth factor.
540 This change impacts on the simulation results for all network models on
541 which we have bandwidth factors configured.
542 *****************************************
543 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
544 *****************************************
545 This change may impact on the timing of your simulation results.
546 Take care when comparing simulations from different SimGrid's
547 versions. Sorry for the inconvenience.
548 - Dynamic network factors: users can configure a callback to define
549 the network factors dynamically. This API is available at
550 simgrid::kernel::resource::NetworkModelIntf.
551 - Users have access to complete information about the current communication
552 to decide which factor to apply. This includes: message size, source and
553 destination hosts, links and zones traversed.
554 - Dynamic factors for both latency and bandwidth.
555 - For more details, see the example in (examples/cpp/network-factors).
556 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
557 deprecated since version 3.24, are no longer supported. Instead, use
558 "wattage_off" and "wattage_per_state".
561 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
564 - New tutorial: Model-checking and formal assessment
565 - New sections: "Demystifying the routing" and "C++ platforms"
566 - Update and improve the part on visualization in MPI and Algo tutorials.
567 - Improve the section on routing: how to define it, how it's used internally
568 - Fix many issues, broken links and missing references in doxygen and Sphinx
571 - Lua platform files are deprecated. Their support will be dropped after v3.31.
574 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
576 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
577 (FG: issues on Framagit; GH: issues on GitHub)
578 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
579 - FG#64: Configuring smpi/IB-penalty-factors
580 - FG#67: Running computation concurrently with MPI_Iallreduce
581 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
582 - FG#71: Segmentation fault on invalid gw_src/gw_dst
583 - GH#322: Issue when an actor kills his host vm
585 ----------------------------------------------------------------------------
587 SimGrid (3.27) March 29. 2021
589 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
591 Main user-visible changes:
592 - All functions marked as deprecated for one year were removed.
593 - Removed all that was related to the "Storage" type in all APIs and bindings.
594 "Disks" have been introduced more than a year ago to replace them for the better.
595 It's time to finish this replacement.
596 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
597 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
598 avoiding "initializer element is not constant" errors with some codes. Please be
599 careful to recompile your MPI codes if switching to this new SimGrid version.
600 - Network model ns-3 now requires version ns-3.28 or higher.
601 - Removed some option aliases that were deprecated since v3.23 (June 2019)
602 - 'verbose-exit' is now 'debug/verbose-exit'
603 - 'simix/breakpoint' is now 'debug/breakpoint'
604 - 'clean-atexit' is now 'debug/clean-atexit'
607 - New functions: Comm::sendto_{init,async} to initiate a communication
608 on between two (possibly remote) hosts.
609 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
610 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
611 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
612 or Mailbox::get_async<void>() if you really want to play with void*.
613 - Unify the interface of Activity::on_{start/activity}
614 - New function: Comm::get_dst_data()
615 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
618 - ptask should now work better with multicore hosts (FG#37).
619 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
620 - Internal cleanups and refactoring around models, to prepare for programmatic
622 - Modernize our tests of the performance models
626 - New section: Configuring the logs from the command line.
627 - New section: Programming API of logs.
628 - New section: Getting realistic results
629 - New section: Examples of SMPI Usage
630 - New part: SimGrid's Internals, with a section on Design goals
631 - Move back to the classical Breathe tool chain, and fix many small glitches.
634 - Handle leak checking: SMPI will now display a message if MPI handles are still
635 present at the end of a simulation. Option smpi/list-leaks:n will display the n
636 first leaked pointers and their types.
637 - smpi/display-allocs option : Display Memory allocations analysis, with hints
638 about the largest allocations. This only works for allocations with malloc/calloc
639 and not compiled with SMPI_NO_OVERRIDE_MALLOC
640 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
641 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
642 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
643 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
644 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
645 - Replay now opens and parses the trace file only once for all ranks when possible.
647 - better handling of MPI_PROC_NULL
648 - MPI_Group_incl was broken in some cases
649 - PAPI tracing now fixed
650 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
651 - SMPI will now fail properly when receive size < sent size, instead of
652 just setting MPI_ERR_TRUNCATE and keep going.
653 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
654 - Remove useless build patches and split build/run patches on the page
655 - Build in parallel when possible
656 - Test on yum-based systems
657 - New nightly tested Proxy-apps:
658 - OpenMPI mpi-test-suite
659 - Coral MCB, LSMS, Qbox, CLOMP
660 - ECP/Mantevo Cloverleaf3D, Ember
661 - Codevault bhtree_mpi
664 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
665 - Start to overhaul the implementation to modernize it (still ongoing).
668 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
669 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
670 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
671 - Deprecate xbt_dynar_to_array().
672 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
674 C binding and interface:
675 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
676 - Constness requirement has been lowered for parameter argv of sg_actor_start()
677 and sg_actor_create() It's now (char* const*). The previous version is still
678 available as sg_actor_start_() and sg_actor_create_().
680 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
681 (FG: issues on Framagit; GH: issues on GitHub)
682 - FG#37: Parallel tasks are limited to 1 core per host
683 - FG#62: Running "smpirun -replay" on large networks
684 - FG!46: Fix a few potential memory leaks in SMPI colls
685 - GH#343: Wrong consumed energy when an exec timeouts
687 ----------------------------------------------------------------------------
689 SimGrid (3.26) December 15. 2020
691 The Release release (the French lockdown was eased today).
693 Important user-visible changes:
694 - SimGrid now requires a compiler with C++14 support.
695 Sibling projects should upgrade their FindSimGrid.cmake
696 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
697 several users had difficulties to understand issues when using high bandwidth or
698 small latency events. The new value was already the default for SMPI and
699 should not cause too much performance hit. It can change some simulation
700 timings if simulation was skipping a lot of small events (which was actually
701 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
702 - On some network models (L07, CM02, LV08), an implicit loopback link is created
703 for local communication. Default bandwidth was 500MBps and latency was 15µs,
704 which was rather bad. Default values have been changed to 10GBps and 0 latency.
705 Options network/loopback-lat and network/loopback-bw have been added to adapt
706 these values if needed. Network and routing models already providing separate
707 loopbacks (clusters and such) are not impacted by this change.
708 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
709 metadata that are added to each MPI message by MPI libraries (tag,
710 src/dest, and communicator information). This may change simulation times
711 for very small message sizes, but should be more realistic.
714 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
715 - New plugins: wifi_energy and link_load. See the associated examples.
716 - New examples: wifi networks, and communication suspend.
719 - LTO is now enabled for Intel/clang compilers.
720 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
721 Setting it to "auto" will use all available cores, while setting it to n will
722 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
723 - Remove obsolete runtime option 'contexts/parallel-threshold'.
724 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
725 has been kept for compatibility.
726 - Finally remove obsolete snake_case() aliases for runtime options.
727 - Further improve the documentation.
730 - the on_exit() of each actor is also executed when the simulation deadlocks.
731 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
732 An example is provided for s4u::Comm but it should work with Exec and Io.
735 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
736 https://framagit.org/simgrid/SMPI-proxy-apps
737 - MPI/IO: supports names starting with "./"
738 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
740 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
741 building shared libraries, this may cause issues, so environment variable
742 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
743 - most temporary files should now be created in /tmp dir (or equivalent).
744 If this one does not allow execution of code (noexec flag), this may cause issues.
745 Please use another tmp directory (using TMPDIR or equivalent system variable)
749 - A large refactoring is ongoing, but this should be transparent for users so far.
752 - Drop xbt_str_split().
754 C binding and interface:
755 - The return type of the sg_actor_on_exit() callbacks is now 'void'
756 (int value was previously ignored)
757 - Many C functions were renamed to follow a common naming style
758 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
759 names are kept until version 3.30.
760 - Many MSG tests were converted to the new S4U's interface in C, that
761 was extended for that.
763 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
764 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
765 - FG#41: Add sg_actor_create C interface
766 - FG#43: xbt::random needs some care
767 - FG#48: The Impossible Did Happen (yet again)
768 - FG#50: Suspending an actor executed at the current timestamp fails
769 - FG#52: Zero-seconds timeout : "That's in the past already"
770 - FG#53: Crash while using ns-3 network model
771 - FG#54: How to suspend a comm?
772 - FG!22: Stochastic Profiles
773 - FG!24: Documentation and fix for xbt/random
774 - FG!35: Add a modeling hint for parallel links in doc
775 - FG!36: [xbt/random] Read/Write the state of the RNG
776 - FG!42: Link cumulated load plugin
777 - FG!44: Test condition variable wait_until+notify_all
778 - GF#18137: Allow different stack sizes?
779 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
780 - GH#139: Allow pthread creation in SMPI
781 - GH#336: Packet-level simulation using SMPI?
782 - GH#345: Error in the LMM while migrating a VM
783 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
784 - GH#352: pip install / python setup.py install fail to find pybind11
785 - GH!337: Fix link_energy plugin for wifi platforms
786 - GH!339: Add Mailbox set_receiver method to python binding
787 - GH!344: Cast hugepages macros parameters to int64
789 ----------------------------------------------------------------------------
791 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
793 The Palindrom Day Release.
795 Important user-visible changes:
796 - Improve the Python usability (stability and documentation).
797 - A nasty synchronization bug was ironed out, see also below.
798 - Python's doc was organized as a separate tree, now integrated with C++.
799 - C bindings of S4U were not part of the doc.
800 - The C++ doc was also improved as methods are now split by theme.
801 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
802 - This is mandatory to use the Java bindings.
803 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
804 - SimDAG++: Automatic dependencies on S4U activities (experimental)
805 - Some features are already implemented but not all of them
806 - Cannot block an activity until it's scheduled on a resource
807 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
808 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
811 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
812 - Drop support for Viva/Triva (old visualization tools), hence removed
813 TRACE_get_node_types() and TRACE_get_edge_types() functions.
816 - Require pybind11 version 2.4.
817 - Greatly improve locking during context switches with Python.
818 - Declaring Python callbacks to this_actor::on_exit is now working.
819 - No more tests are failing currently, even on win32.
822 - Actor: Merge signals on_migration_start/end into on_host_change
823 - Actor: Rename migrate() into set_host()
824 - Disk: Allow users to get the read and write nominal bandwidth values
825 - Exec: Implement wait_for(timeout)
826 - Io: Implement wait_for(timeout)
829 - Parse errors now raise a simgrid::ParseError that you may want to catch.
832 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
833 been renamed to fix a spelling error. As usual, the previous definitions are
834 kept with a deprecation warning. However, to avoid runtime errors, it is now
835 forbidden to override these deprecated methods in a derived class. Please use
836 the new names immediately if you need to override them.
839 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
840 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
841 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
842 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
843 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
844 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
845 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
846 - New module: random, providing classical random numbers generators.
849 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
850 allocations' size above which they are considered "shared" by default (as if
851 they were performed through SMPI_SHARED_MALLOC macros).
852 Default = 0 = disabled feature.
853 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
854 This can cause some troubles if codes are already overriding these. If this
855 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
856 help, but will make this feature unusable.
857 - Input check is now performed much more consistently and outputs precise
858 warnings for each error found
859 - New supported calls : MPI_File_set_view, MPI_File_get_view
860 - MPI I/O now only supports disks, as storage is discontinued.
861 - MPI I/O allows opening files without specifying absolute path
862 - Note : SMPI configuration options are now initialized by smpi_init_options()
863 call, which should be called if SMPI is not being used through smpirun.
865 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
866 - FG#9: Python bindings crashing
867 - FG#39: Missing s4u::Comm::wait_any_for example
868 - FG#42: Add support for ThreadSanitizer (TSan)
869 - FG!19: Removing RngStream
870 - FG!20: A module for RNG calls
871 - FG!21: Choice between ad-hoc and standard distributions implementations
873 - FG!25: Fix link in Mutex doc
874 - FG!26: Fix links in SMPI interface doc
875 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
876 - GH#305: Unscheduled tasks are still excuted
877 - GH#313: smpirun: manual outdated w.r.t. --help
878 - GH#321: [S4U] Get task remaining work ratio
879 - GH#323: Crash when an actor turn off his physical host
880 - GH#335: Missing links on dragonfly example svg
881 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
882 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
884 ----------------------------------------------------------------------------
886 SimGrid (3.24) October 9. 2019
888 The Clean Disk Release.
891 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
892 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
893 thus becomes the default resource to execute I/O operations. Examples
894 and the file system plugin have been modified to use disks instead of
895 storage elements. Storage can still be used but is doomed to disappear.
896 Users are thus adviced to update their simulators accordingly.
897 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
898 for consistency with pthread_barrier_wait()
899 - Host::get_englobing_zone() returns the englobing netzone
900 - Actor::on_destruction is now called in the destructor
901 Actor::on_termination new signal called when the actor terminates
903 - Global signals are now part of the Engine:
904 - on_platform_creation: after config settings, before the XML parsing
905 - on_platform_created: right after the XML parsing
906 - on_time_advance: each time the clock advances
907 - on_simulation_end: after simulation, before cleanups
908 - on_deadlock: as the name implies.
910 - sg_{actor,host,link}_{data,data_set}() now all exist.
911 Use them to attach user data to the object and retrieve it.
914 - Introduce an experimental Wifi model. It sounds reasonable
915 according to the state of the art, but it still has to be properly
916 validated, at least against ns-3 if not against reality.
917 - Improved the usability of ns-3. Several bugs were ironed out.
918 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
919 It is now expressed as 'idle:epsilon:allCores' to properly model the
920 consumption of non-whole tasks on mono-core hosts. More info in the doc.
923 - convert a new set of functions to the S4U C interface and move the old MSG
924 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
927 - Fortran bindings for DVFS have been removed.
928 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
929 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
930 - SMPI can now be selected by cmake's find_module(MPI) with
931 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
932 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
933 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
934 - trace-call-location can be used with TI traces, and replayed, and is compatible
935 with smpi/comp-adjustment-file.
936 - sleep events are now correctly traced/replayed.
937 - Default for trace-call-location is now to use file names and not full paths.
938 To revert to previous behavior (in case of collision of filenames), option
939 "smpi/trace-call-use-absolute-path" can be set to yes.
942 - Use the included xxHash as an hash implem when C++14 is usable.
943 - Option model-checker/hash was removed. This is always activated now.
944 - New option smpi/buffering controls the MPI buffering in MC mode.
945 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
946 This is useful to check for MPI compliance.
949 - New section on plugins: how to define a new one, and existing ones.
952 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
953 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
956 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
957 and <mount> tags. The rationale is finish to decouple what should be handle
958 by SimGrid kernel (i.e., resources and activities) from plugins built in
959 user space (e.g., the file system plugin). This new tag currently accepts
960 only three attributes (id, read_bw, and write_bw). All the other information
961 that was declared with the storage related tags now has to be expressed as
962 properties. An example of platform using this new tag is available at
963 examples/platforms/hosts_with_disks.xml
966 - 'expect signal' can now accept more than one potential signal.
968 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
969 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
970 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
971 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
972 - FG#31: per-actor data
973 - FG#34: SG_BARRIER_SERIAL_THREAD?
974 - FG#35: model-checker does not like buster-produced binaries
975 - FG!13: MC: complete workaround in the error msg seen on modern systems
976 - FG!15: execute_flops now logs compute
977 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
978 - FG!17: ns-3: unblock the right number of communications + others issues
979 - FG!18: Improving the performance of the ns-3 bindings
980 - GH#189: Energy consumption of parallel tasks
981 - GH#219: Error in the throughput of TCP transfer
982 - GH!330: Minor polishing of English: Configuring_Simgrid
983 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
984 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
985 - GH!334: Rework energy plugin, again
987 ----------------------------------------------------------------------------
989 SimGrid (3.23.2) July 8. 2019
992 - Nicer introduction page.
993 - Migrate the "Deploy your application" page to the new doc.
994 - Move Java as a subtree of MSG.
997 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
1000 - Drop xbt_dynar_sort_strings().
1003 - Really fix FG#26: Turning off a link should raise NetworkFailureException
1004 - FG#27: Wrong exception thrown to wait_any when link is turned off
1005 - GH#328: Java: Canceling multiple tasks in a single vm/host
1007 ----------------------------------------------------------------------------
1009 SimGrid (3.23) June 25. 2019
1011 The Exotic Solstice Release.
1014 - SunOS and Haiku OS support. Because exotic platforms are fun.
1015 - Stop setting random seed with srand48() at initialization.
1016 - Use addr2line as a fallback for stacktraces when backtrace is not available.
1017 - Build option -Denable_documentation is now OFF by default.
1018 - Network model 'NS3' was renamed into 'ns-3'.
1021 - SimGrid can now hopefully be installed with pip.
1024 - wait_any can now be used for asynchronous executions too.
1027 - New log appenders: stdout and stderr. Use stdout for xbt_help.
1028 - Drop xbt_dict_dump.
1031 - 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
1032 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
1033 - MPI interface is now const correct for input parameters
1034 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
1035 - Fortran bindings for async collectives.
1036 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
1039 - Remove option 'model-check/record': Paths are recorded in any cases now.
1040 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
1041 always sparse. That's an immense gain in memory, and can even be
1042 faster because of cache effects. So there is no need to clutter the
1043 code to allow the user to go for the unefficient mode.
1046 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
1047 models proved to be more accurate than these old experiments.
1049 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
1050 - FG#1: Broken link in error messages
1051 - FG#2: missing installation documentation
1052 - FG#3: missing documentation in smpirun
1053 - FG#6: Python bindings not available on PyPI
1054 - FG#7: simple cmake call requires doxygen
1055 - FG#8: make python bindings an optional dependency
1056 - FG#10: Can not use MSG_process_set_data from SMPI any more
1057 - FG#11: Auto-restart actors forget their on_exit behavior
1058 - FG#12: -Denable_lto=OFF doesn't disable LTO
1059 - FG#13: Installs unstripped file 'bin/graphicator'
1060 - FG#14: Installs the empty directory 'doc/simgrid/html'
1061 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
1062 - FG#17: Dead link in doc (pls_ns3)
1063 - FG#20: 'tesh --help' should return 0
1064 - FG#21: Documentation link on http://simgrid.org/ broken
1065 - FG#22: Debian installation instruction are broken
1066 - FG#26: Turning off a link should raise NetworkFailureException exceptions
1067 - GH#133: Java: a process can run on a VM even if its host is off
1068 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
1069 - GH#326: Valgrind-detected error for join() when energy plugin is activated
1071 ----------------------------------------------------------------------------
1073 SimGrid (3.22) April 1. 2019
1075 The Easter Christmas Release (this one is somewhat late).
1078 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
1079 - Not complete yet: asynchronous activities (amongst others) are still missing
1080 - Still ongoing: the interface may change in the future. We need more
1081 testers! Please report any glitches.
1082 - No new project using Java should start now. Please switch to Python.
1085 - Some of the internal cleanups may lead to speed improvements:
1086 - The hard limitation on the amount of simulated actors with
1087 Java+Mac was removed. Now, the available memory is the only limit.
1088 - Our refcounting was tidyied, leading to 10% speedups in some cases.
1089 - We are still working on making our code robust to the actor kills
1090 and hosts' churn. Things are improving, but it's not perfect yet.
1091 - Replaced our own code to display a backtrace (that was forking addr2line)
1092 with the Boost.Stacktrace library.
1093 You won't see your backtraces without this optional dependency.
1094 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
1095 - Stop setting random seed with srand() at initialization.
1098 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
1099 XML file version remains 4.2 since old files are still compatible.
1102 - Process termination which was broken at version 3.21 has been repaired.
1103 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
1104 - Hide the examples into examples/deprecated. New users should use Python.
1107 - Drop MSG_process_create_from_stdfunc() from the API.
1108 This C++ function was a pimple in the C API, made necessary at some
1109 point by the Java bindings. This is fixed now.
1110 - Hide the examples into examples/deprecated. New users should use S4U.
1111 - MSG_process_create and MSG_process_attach now crash if the host used
1112 to run this process is off.
1113 - Fix the protype of MSG_process_on_exit()
1114 Now use (int,void*) callbacks instead of (void*,void*) ones.
1115 The implementation was ways too messy to actually work, I guess.
1118 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
1119 too often and provide more accurate timings.
1120 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
1121 - MPI_Alltoallw support
1122 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
1123 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
1124 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
1125 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
1127 - Basic implementation of generalized requests (SMPI doesn't
1128 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
1131 - Drop sg_cmdline. Please use xbt_cmdline instead.
1132 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
1133 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
1134 OsSemaphore is implemented in a portable way with C++11 threads.
1135 This should allow much more threads to be created at the same time,
1136 allowing Mac and Java users to simulate many more actors.
1137 - Implement the 'thread' factory with std::thread instead of xbt ones.
1138 It is not possible to set the stack size with threads anymore, but
1139 -fsplit-stack is the way to go nowadays when using threads.
1140 - Drop the xbt_os_thread_t module (now unused)
1141 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
1142 - Drop xbt_str_join_array().
1143 - Drop cunit, use Catch2 instead.
1146 - Many cleanups in the kernel::activity namespace. This was long
1147 overdue, and shall open the path to many future endeavors.
1150 - #132: Java: a process can not shut down its own host
1151 - #220: S4U: detached send causes exception if sender terminates soon after sending
1152 - #261: Document the parameters of parallel execution's constructor
1153 - #300: [s4u] BarrierPtr is missing
1154 - #314: SMPI args internal cleanup
1155 - #316: Fix a bug related to the CPU utilization of multi-core VM
1156 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
1157 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
1158 - #325: Turning off a host has different behavior on sleeping actors and computing actors
1160 ----------------------------------------------------------------------------
1162 SimGrid (3.21) October 3. 2018
1164 The Restarting Documentation (TRD) Release.
1167 - Convert the user manual to Sphinx for improved usability.
1168 Unstable version now hosted on https://simgrid.frama.io/simgrid
1169 Many glitches remain; stay tuned.
1170 - Restrict the Doxygen reference API to the installed header files.
1171 (documenting all internals this way is overkill + doxygen scales badly)
1172 - New tutorial on S4U (highly inspired from the old MSG one, but with
1173 a git repository to fork as a starting point, and a docker image)
1174 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
1175 - The Developper manual is still to be converted (not compiled until then)
1176 - Some parts are still missing in this conversion (such as the
1177 platform and deployment sections) while others were blindly converted
1178 and would need more love (such as the configuration flags).
1179 Things will certainly further improve in the future releases.
1182 - Fully reimplement auto_restart mechanism that was utterly broken
1183 (fix #22, #131 and #281 that were all very related)
1184 - Implement semaphores (fix #303 and #291)
1185 - When creating an actor from a function and its parameters,
1186 move-only parameters are not allowed anymore, as it would prevent
1187 the actor to be restartable if its parameters are consumed on run.
1188 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
1189 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
1190 IO operations that can be started, waited for, or canceled as a regular
1192 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
1196 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
1197 - New host variable: 'core_count'
1200 - Remove xbt_os_thread_specific features
1201 - Remove portability wrapper to condition variables
1202 - Remove xbt_os_thread_yield()
1205 - MPICH collective selector now mimics MPICH 3.3b
1206 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
1207 - MPI_Init(NULL,NULL) is now allowed
1208 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
1209 - Command-line arguments should now be handled properly in Fortran simulations
1210 - MPI Attributes and related callbacks should work in Fortran
1211 - Apps using C + Fortran should now work
1212 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
1213 - Support MPI_Error_String
1216 - Due to an internal bug, Msg.run() must now be your last line.
1217 We hope to fix it in a future release, and we are sorry for the inconvenience.
1220 - #22: Process autorestart seem to only work with CAS01 cpus
1221 - #56: Feature request: dynamic SMPI replay
1222 - #93: simgrid should not eat --help
1223 - #111: How to change the loopback link characteristics?
1224 - #116: Platform cluster radical
1225 - #117: Platform cluster/host order
1226 - #129: Replay_multiple on steroid
1227 - #131: Java : autostart does not work
1228 - #143: Setting a breakpoint at a given time
1229 - #235: Network energy models should be integrated
1230 - #264: Add ptask L07 resource tracing
1231 - #271: Dynamic and manual replay of SMPI traces
1232 - #279: Breakpoints option uses time deltas instead of absolute time
1233 - #281: Daemonized actor no longer daemonized after an auto-restart
1234 - #282: TIT convertor erase traces when using absolute path in the trace list file
1235 - #285: segfault when a s4u actor kills itself with msg process tracing activated
1236 - #286: Pajé traces are not exposing the number of cores
1237 - #287: Command-line parsing should stop when encountering '--'
1238 - #288: MPI_Init(NULL, NULL)
1239 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
1240 - #290: Method to check if a mailbox has a message ready to be consumed
1241 - #291: [s4u] semaphores do not seem available yet
1242 - #294: Alltoall collective implementation of mpich
1243 - #295: Starting a Host as OFF is not supported
1244 - #296: DTD too permissive
1245 - #299: Add a s4u barrier example
1247 ----------------------------------------------------------------------------
1249 SimGrid (3.20) Released June 24. 2018
1251 The proxy snake_case() release.
1254 - The whole API is now using snake_case() instead of mixing it with camelCase.
1255 Compatibility wrappers in place for one year.
1256 - We took the opportunity to increase the consistency of the API.
1257 Please report any remaining glitches.
1258 - New function simgrid::s4u::Host::get_actor_count:
1259 Returns the number of actors running on a specific host.
1262 - SMPI is now tested with ~45 proxy apps from various sources, with none or
1263 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
1264 - Replay: The replay file has been re-written in C++.
1265 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
1266 supported. They are stored in the trace and used when replayed.
1267 - Basic support of MPI_Cancel. Robustness not guaranteed.
1268 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
1269 (only for MPI_COMM_TYPE_SHARED).
1270 - New option: smpi/privatize-libs, to add external shared libs to be privatized
1271 by SMPI. They will be copied locally and loaded separately by each process.
1272 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
1273 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
1274 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
1275 with internal index optind with multiple processes. Only works if MPI_Init has
1276 already been called.
1277 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
1278 syntax without preprocessing tricks.
1281 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
1282 behavior tracing is no longer limited to MSG
1283 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
1284 this means that traces now need to be updated or re-obtained; both Irecv
1285 and Isend lines in your traces have now in total 5 mandatory fields:
1286 <rankid> <command> <to/from rankid> <tag> <size>
1287 To update your traces, it suffices to add a 0 for the tag here.
1288 - TIT now also supports waiting for a distinct request via MPI_Wait.
1289 Wait/Test now wait for a specific request, not just the last one that was
1290 issued. This unfortunately means another update, because we need to
1291 identify which request you want to wait for. We do this via the
1292 triplet (sender, receiver, tag), which needs to be added:
1293 <rankid> <command> <sender> <receiver> <tag>
1294 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
1295 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
1298 - The deprecation of MSG is ongoing (but this should not impact you).
1299 Many MSG functions are now simple wrappers on the C API of S4U. If
1300 you wish to convert your code to S4U, find the S4U counterparts of
1301 your MSG calls in src/msg/msg_legacy.cpp.
1302 - MSG can still be used, but won't evolve anymore.
1305 - Allow to run the Link energy plugin from the command line with
1306 --cfg=plugin:link_energy
1307 - Rename Energy plugin into host_energy
1308 - Rename Load plugin into host_load
1311 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
1312 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
1313 - Improve performance when handling timeouts of simix synchros.
1316 - Config: the C API is now deprecated (will be removed in 3.23), and
1317 the C++ API has been slightly improved.
1320 - Fix several build issues on OSX.
1321 - Move simgrid_config.h to simgrid/config.h (old header still working)
1324 - #143: Setting a breakpoint at a given time
1325 - #150: Inconsistent event names in SMPI replay
1326 - #258: daemonized actors hang after all non-daemonized actors have completed
1327 - #267: Linker error on unit_tmgr
1328 - #269: SMPI: tracing of MPI_Wait/all/any broken
1329 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
1331 ----------------------------------------------------------------------------
1333 SimGrid (3.19.1) Released March 22 2018
1335 The Fixed ABI Release.
1337 As you may know, we are currently refactoring SimGrid in deep.
1338 Upcoming SimGrid4 will be really different from SimGrid3: modular,
1339 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
1341 Our goal is to smooth this transition, with backward compatibility
1342 and automatic update paths, while still progressing toward SimGrid4.
1344 SimGrid remains open during works: The last pure SimGrid3 release was
1345 v3.12 while all subsequent versions are usable alpha versions of
1346 SimGrid4: Existing interfaces remain unchanged, but the new S4U
1347 interface is budding and the internals are deeply reorganized.
1349 Since 2015, we work hard to reduce the changes to public APIs. When
1350 we need to rename a public library symbol in S4U, we let your
1351 compiler issue an explicative warning when you use the deprecated
1352 function. These messages remain for four releases, i.e. for one full
1353 year, before turning into an error. Starting with v3.15, your can
1354 also adapt to API changes with the SIMGRID_VERSION macro, that is
1355 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
1357 Starting with this v3.19.1, our commitment to reduce the changes to
1358 the public interfaces is extended from the API to the ABI: a program
1359 using only MSG or SimDag and compiled against a given version of
1360 simgrid can probably be used with a later version of SimGrid without
1361 recompilation. We will do our best... but don't expect too much of
1362 it, that's a really difficult goal during such profund refactoring.
1364 The difference between v3.19 and v3.19.1 is that the former was
1365 accidentally breaking the ABI of MSG, while the later is restoring
1368 S4U and kernel APIs will still evolve until SimGrid4, with one-year
1369 deprecation warnings as currently. In fact, cleaning up these
1370 interfaces and converting them to snake_case() is one release goal of
1371 v3.20. But don't worry, we are working to smooth this upgrade path.
1373 In summary, new projects should start with S4U to benefit of the
1374 future, but old MSG projects should still be usable with no change.
1376 ----------------------------------------------------------------------------
1378 SimGrid (3.19) Released March 20 2018
1380 The Moscovitly-cold Spring release.
1382 SMPI user-visible improvements:
1383 - Support of MPI_Type_create_subarray
1384 - Unimplemented MPI calls now abort (instead of issuing a warning)
1385 Some harmless calls around errhandlers still only issue a warning.
1386 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
1387 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
1388 is now possible (mmap privatization was refactored to this end).
1389 Adding new SMPI instances should also be technically doable now.
1391 SMPI internal cleanups:
1392 - Removed index notion from SMPI -- no more "getPid() - 1"!
1393 In tracing files, actors are called rank-<PROCESS-ID> but these are
1394 NOT the real ranks (tracing several communicators is not supported yet).
1395 - Deprecate smpi_process_index() (will be removed in 3.22)
1396 - Better testing on classical proxy apps, through an external project:
1397 https://github.com/simgrid/SMPI-proxy-apps/
1400 - Use a graphical TOC to make it easier to find the documentation you need
1401 - Revamp the MSG tutorial
1404 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
1405 - Remove parameter of MSG_process_killall().
1406 Resetting the PID was bogus anyway (several actors could have the same PID).
1409 - Execution->setHost() can be called after start(), to migrate it.
1410 - Comm::test_any() is now implemented.
1411 - s4u::Actor now has onCreation() and onDestruction() signals.
1412 - Install some kernel header files for the users' plugins and more.
1413 Warning, their API is really not stable yet.
1414 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1415 include/simgrid/kernel/routing/*Zone.hpp
1418 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1420 Build System and other cleanups:
1421 - Remove unused run-time parameter "tracing/onelink-only".
1422 - SimGrid now works with intel compilers (MC still buggy)
1423 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1424 - tesh kills subprocesses on timeout (but not on windows)
1427 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1430 - #194: Feature request: simgrid::s4u::Comm::test_any()
1431 - #245: migrating an actor does not migrate its execution
1432 - #253: Feature Request: expose clusters as objects
1433 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1434 - #255: Tesh broken on Windows
1435 - #256: Modernize FindSimGrid.cmake
1436 - #257: Fix (ab)use of CMake install
1438 ----------------------------------------------------------------------------
1440 SimGrid (3.18) Released December 24 2017
1442 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1444 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1445 - Support for MSG should not be removed from SimGrid before 2020,
1446 but future evolutions will be limited to the integration of
1447 user-provided patches.
1450 - s4u::onDeadlock() signal, run before stopping the simulation.
1451 - s4u::Exec: asynchronous executions (abstraction of background
1453 - s4u::Host->getLoad() returns the achieved speed in flops/s
1456 - Link::name() is deprecated and will be removed in v3.21, use
1457 Link::getCname() instead.
1458 - Mailbox::getName() changed to return a std::string, use
1459 Mailbox::getCname() instead to get a char*.
1460 - Storage::getName() changed to return a std::string, use
1461 Storage::getCname() instead to get a char*.
1462 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1463 to have both allocation and deallocation of the map in user space.
1466 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1467 The amount of flops remaining to do is a vector, not a scalar.
1468 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1469 name implies on both sequential and parallel tasks.
1470 - Both changes fix GitHub's #223 using PR #237 as a basis.
1471 Thanks Michael Mercier.
1472 - Most examples were converted to S4U and hidden elsewhere at tests.
1473 Rationale: we still want MSG to work; we want newcomers to use S4U.
1476 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1477 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1480 - Switch to the faster dlopen privatization mechanism by default
1481 - Documentation improvments
1484 - Live migration is getting moved to a plugin. Dirty page tracking is
1485 the first part of this plugin. This imply that VM migration is now
1486 only possible if one this function is called:
1487 - C/MSG: MSG_vm_live_migration_plugin_init()
1488 - C/C++: sg_vm_live_migration_plugin_init()
1489 - Java: Msg.liveMigrationInit()
1490 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1494 - Define class simgrid::xbt::Path to manage file names.
1495 - Removed unused functions:
1496 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1497 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1498 xbt_graph_export_graphviz()
1499 - xbt/str.h: xbt_str_join()
1500 - Remove unused datatypes:
1501 - xbt/heap.h: use std::priority_queue or boost::heap instead
1502 - xbt/swag.h: use boost::intrusive::list instead
1505 - New link_energy plugin for the consumption of the links.
1506 - All of the operations on files and storage contents have been
1507 packaged into a plugin (src/plugins/file_system). The current
1508 public interface can be found in
1509 include/simgrid/plugins/file_system.h
1510 To use these functions you now have to initialize the plugin by
1511 calling MSG_storage_file_system_init() just after calling
1512 MSG_init() or sg_storage_file_system_init() just after creating
1516 - Remove the undocumented/untested tag <include>
1519 - Remove viva specific tracing as the tool is no longer maintained
1522 - #248: Kill of finished processes leads to segfault
1523 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1524 - #239: please implement signal s4u::onDeadlock()
1525 - #230: segfaults when exit() without run()
1526 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1527 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1528 - #222: Actor::kill() doesn't really kill and segfaults
1529 - #221: odd LMM warning when killing an actor
1530 - #120: Memory leak when the processes are forcefully killed
1532 ----------------------------------------------------------------------------
1534 SimGrid (3.17) Released October 8 2017
1536 The Drained Leaks release: (almost) no known leaks despite the tests.
1538 Even very long simulations will preserve your memory: our testsuite
1539 stresses SimGrid for over 45mn on fast machines for a coverage of
1540 over 80%, and there is only one single known leak, of about 4kb.
1544 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1545 handler may be given to free resources if the comm cannot be completed.
1546 - this_actor::send(mailbox) is now mailbox->put()
1547 - New: simgrid::s4u::Comm::wait_all()
1548 - New: Host.execute() for remote executions.
1551 - Unused option network/sender-gap is removed.
1554 - Deprecate MSG_task_isend_with_matching(): This unused feature
1555 really complicates our internals. Will be removed in v3.20.
1558 - Improved context termination. It is now possible to free resources, even
1559 when a process is forcibly killed.
1560 - ContextBoost: add support for Boost versions above 1.61.
1563 - Bring back run-time option --cfg=exception/cutpath to remove exception
1565 - Removed unused functions:
1566 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1567 xbt_str_rtrim(), xbt_str_trim().
1568 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1571 - Removed header files obsolete since SimGrid 3.12:
1572 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1574 - #3: SD_exit should be made optional
1575 - #120: Memory leak when the processes are forcefully killed
1576 - #159: Threading test regression in Actor refcounting
1577 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1578 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1579 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1580 - #191: VM migration and pstate
1581 - #192: Updating the energy consumptions of all hosts crashes with VMs
1582 - #195: All actors have PID=0 in the logs
1583 - #204: Sometimes segfault with thread contexts and mmap privatization
1584 - #222: Actor::kill() doesn't really kill and segfaults
1585 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1587 ----------------------------------------------------------------------------
1589 SimGrid (3.16) Released June 22. 2017.
1591 The Blooming Spring Release: developments are budding.
1594 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1595 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1596 - Java JAR file should be finally fully working on Mac OSX too.
1598 XML platforms: Switch to platform v4.1 format.
1599 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1600 - <zone> can be used as a synonym for the now deprecated <as>
1601 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1602 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1603 - <actor> can be used as a synonym for the now deprecated <process>
1604 - state_file and avail_file periodicity is now easier to express
1605 (check the documentation) the old behavior should still work.
1606 * Storage: not backward compatible, but it seems that nobody ever used it
1607 - Remove attribute 'content_type' of <storage_type>: was never used
1608 - Make attribute 'model' of <storage_type> optional: for future usage
1609 - Remove Bconnection model property: was never (in)validated. Replaced by
1610 the maximum of the read and write bandwidth as a resource constraint.
1613 - New and Backwards Compatibility break:
1614 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1615 When one wants to get the list of tasks whose states have changed during a
1616 simulation round, s/he has to allocate and free a dynar and use it as
1617 argument to this function. The former SD_simulate (double how_long)
1621 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1622 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1625 - New function to update the consumption of all hosts at once.
1626 - Fix the model for multi-core hosts, linear in the amount of busy
1627 cores with an abnormality for pIdle (see paper).
1628 BUG: Still not satisfactory for ptask on mono-cores.
1631 - The netzone are now available from the MSG API.
1632 The old names still work, but are now deprecated.
1635 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1636 - New option: smpi/keep-temps to not cleanup temp files
1637 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1638 mallocs mode is used, allows to change the size of the fake file used
1639 (default 1MB), to potentially limit the number of mappings for large runs.
1640 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1641 - Fortran ifort and flang compilers support
1642 - New RMA calls supported (experimental):
1643 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1644 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1645 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1646 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1647 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1648 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1651 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1652 - Behind the scene, ActivityImpl are now automatically refcounted.
1653 Removing this nasty bottleneck will greatly simplify our internals.
1656 - Replay: New function xbt_replay_action_get():
1657 Retrieve the function previously associated to an event type.
1658 - DROPPED FUNCTION: xbt_str_varsubst()
1659 - DROPPED MODULE: strbuff. We don't need it anymore.
1660 - DROPPED MODULE: matrix. We don't need it anymore.
1661 - DROPPED MODULE: lib. We don't need it anymore.
1663 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1665 SimGrid (3.15) Released March 22. 2017
1667 The Spring Release: continuous integration servers become green
1668 We fixed even the transient bugs on all target architectures:
1669 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1670 Windows, FreeBSD, NetBSD.
1672 - Rename NetCards to NetPoints.
1673 This was intended to help NS3 users, but that's not a netcard.
1674 That's a point in the routing algorithm, let's avoid wrong simplifications.
1677 - New: MSG_process_yield(). Stop and yield to other processes.
1678 - New: MSG_process_daemon(). Daemon processes are automatically killed
1679 when the last non-daemon process terminates
1680 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1681 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1682 to make room for the upcoming network energy plugin.
1683 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1686 - Ensure that an actor can kill itself with Process::exit()
1687 - Kill the obscure NativeException. Nobody want to survive the issues
1688 it denotes, so use JniException that is a RuntimeException (not to
1689 be caught explicitly).
1690 - Partial bug fix in initialization. SimGrid flags on command line were
1691 consumed at C level but stayed in the original Java String[] args.
1692 This could mess users' args[i] if SG flags were not put at the end of
1694 The SimGrid flags are now removed from the Java arguments. However,
1695 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1696 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1697 It might be if you have --log or --cfg flags in the command line.
1698 - Fix numerous memleaks all around the place. In particular, around VMs.
1702 - s4u::onPlatformCreated: right before the simulation starts
1703 - s4u::onSimulationEnd: right after the main simulation loop
1704 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1705 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1706 event from the availability_file changes the avail speed.
1707 - Links are now usable from s4u
1708 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1709 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1710 - New examples: The conversion of MSG examples to S4U has begun
1711 - Actors: create, kill, migration, and suspend
1712 - Applications: master-worker and token-ring
1713 - Action replay: communications and storage
1714 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1717 - Backwards Compatibility breaks
1718 - The SD_route_get_list and SD_route_get_size functions have been removed.
1719 They are replaced by the unique
1720 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1721 The route from src to dst is built in the links dynar whose size can
1722 be obtained with xbt_dynar_length.
1723 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1724 sg_host_route_bandwidth and sg_host_route_latency.
1725 Macros ensure the backwards compatibility, but you should fix your code
1728 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1729 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1730 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1731 We don't support 100% of it, but it should be enough. Contact us if not.
1732 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1733 naive and probably inaccurate)
1734 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1735 mapped onto a unique small file using some system magic.
1736 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1737 - Fortran: cleanups, fixes, support of user-added operations
1738 - RMA: MPI_Accumulate are applied in correct order.
1739 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1740 - Better support for MPI_IN_PLACE
1741 - Support for MPI_Win attrs and keyvals.
1742 - Support MPI_Comm_group_create, MPI_Type_size_x
1743 - MPI ops should be properly applied to all allowed types
1744 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1745 algorithm by a non failing one from mpich.
1748 - Kill the fifo data container: we don't use it anymore.
1751 - A <cluster> can now be created with different speed values to represent pstates.
1752 see examples/platforms/energy_cluster.xml for an example.
1754 SimGrid (3.14.159) Released December 28. 2016
1756 The previous release was only a crude approximation of a Pi release;
1757 this one should be a bit better.
1759 - Revert a last minute change that broke on Mac OSX.
1760 - Fix the changelog and NEWS files.
1761 - Improve the documentation, in particular of the routing module.
1762 - Integrate some patches flying around in Debian.
1764 SimGrid (3.14) Released December 24. 2016
1767 * The whole documentation was reworked and reorganized. There is
1768 still much room for improvement here, but we're on it.
1771 * We now need python3 for our testing infrastructure.
1772 * Model checking needs libevent
1775 * Now works on FreeBSD
1778 * Privatization now works on FreeBSD
1780 * Privatization is now activated by default in smpirun
1781 You can turn it off with -no-privatize if something goes wrong.
1783 * Call-location tracing for SMPI.
1784 You can add the exact location (filename / linenumber) of an MPI call to
1785 your trace files and slow down or speed up the simulation between two
1786 consecutive calls by using an adjustment file (see the documentation).
1788 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1789 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1790 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1791 values were ignored in some cases. The timings of these functions can now
1792 be significantly different.
1794 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1795 smpi/running-power is renamed to smpi/host-speed
1797 * smpi/grow-injected-times option to enable or disable multiplication of the
1798 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1799 make simulation less precise (but also much faster).
1801 * smpirun script should be (much) faster for large deployments.
1803 * SMPI tracing: fixed issue with poor matching of send/receives.
1805 * Replay: Fix broken waitall
1807 New functions and features
1808 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1810 Dropped / renamed functions and features
1811 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1812 - MSG_mailbox_is_empty() -> Mailbox::empty()
1813 - MSG_mailbox_front() -> Mailbox::front()
1814 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1815 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1816 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1817 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1818 * Don't pass the free_f parameter to property related functions:
1819 - MSG_host_set_property_value()
1820 - MSG_as_router_set_property_value()
1821 - MSG_storage_set_property_value()
1822 * VM properties. Since msg_vm_t are msg_host_t, just use
1823 MSG_host_get_property_value() and friends
1824 * VM I/O related things:
1825 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1826 - Unimplemented save/restore methods
1827 * MSG_as_router_get_property_value() was redundant with
1828 MSG_environment_as_get_property_value().
1829 - Removed MSG_as_router_*propert*() functions
1830 - Added MSG_environment_as_set_property_value() for consistency
1831 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1832 Well, they are still there for now, but deprecated with a warning.
1833 Please switch to xbt_dict_new_homogeneous() before this is removed
1835 * Task affinity. Its intended behavior (that was very badly tested
1836 and probably not really working) was deceiving what most users
1837 would have hoped here.
1838 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1839 * Option network/coordinates is now useless and should be dropped.
1842 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1845 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1846 * Added Dragonfly topology support
1849 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1850 EiBps, EBps, Eibps, Ebps for bandwidth.
1851 They may become useful to some lucky ones.
1854 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1855 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1856 * ex/async_waitAll: new example, on asynchronous communications
1859 * Memory usage should be decreased for simulations with a large number
1860 of processes. This also helps for SMPI.
1862 SimGrid (3.13) stable; urgency=low
1864 The Half Release, a.k.a. the Zealous Easter Trim.
1866 - We removed half of the lines, that were mostly experimental cruft.
1867 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1868 (not counting blanks and comments -- according to openhub.net)
1869 - The internals are now compiled in C++ (and will soon be clean C++)
1870 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1871 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1873 * Backwards Compatibility breaks
1874 - Removed Lua simulation bindings (switch to C or Java for that).
1875 Lua can still be used to describe platforms
1876 - Removed Java kernel plug-ins.
1877 Will be reintroduced after the ongoing major internals reorg.
1879 - the following functions were removed.
1880 They were too specific and should be reimplemented in a generic
1881 way, with filter function.
1882 - MSG_task_listen_from_host
1883 - MSG_mailbox_get_count_host_waiting_tasks
1884 - MSG_mailbox_put_with_timeout was removed.
1885 Please use MSG_task_send_with_timeout instead.
1887 - the SD_application_reinit function was removed. It has been a noop for a while.
1888 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1889 replaced by a more flexible mechanism.
1890 - The following functions thus do not exist anymore
1891 - SD_workstation_get_access_mode
1892 - SD_workstation_set_access_mode
1893 - SD_workstation_get_current_task
1894 - Basic estimation functions have been removed but can easily be replaced
1895 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1896 - SD_workstation_get_computation_time => amount / sg_host_speed()
1898 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1899 Use VM.getSpeed()*load/100 for the legacy behavior.
1901 - option enable_tracing was removed. It was not doing anything for a while.
1902 - In the ModelChecker:
1903 - the model-checker now ptraces the model-checked process which means
1904 you cannot use a debugger on the latter anymore (we might make this
1905 optional in the feature);
1906 - removed soft-dirty page tracking;
1907 - remove model-checked side snapshot management,
1908 MC_snapshot() and MC_compare_snapshot();
1909 - keep the MC_cut() function as a stub (it was not really working
1910 in the previous release).
1913 * All options are consistently in kebab-case. Old names are kept as alias.
1916 * Switch to platform v4 format.
1917 - Rename from 'power' to 'speed' the attributes describing the amount of
1918 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1919 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1920 - In <host> and <link>, attributes availability and state are gone.
1921 It was redundant with state and availability traces, and with peak values.
1922 - In <cluster>, attributes availability_file and state_file are gone.
1923 It was too complex and unused.
1924 - Kill <gpu>. Was not doing anything.
1925 - The DOCTYPE points to the right URL:
1926 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1927 (the file at this address now documents the changelog since its v1)
1928 - A warning is emitted for unit-less values (they are still accepted).
1929 - speed. Default: 'f' or 'flops'. Also defined:
1930 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1931 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1932 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1933 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1934 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1935 - latency. Default: 's' second. Also defined:
1936 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1938 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1939 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1942 * s4u::Host is now the preferred public interface to the Host features.
1943 sg_host_* functions are C bindings to the exact same behavior
1944 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1947 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1948 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1949 They were deprecated since a few years, and probably did not even compile anymore.
1952 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1953 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1954 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1955 encouraged to update their codes. A list of the modified functions can be found at the end of
1956 include/simgrid/simdag.h
1959 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1960 the lambda or closure passed as a parameter will run in kernel mode.
1961 All the callback functions should be rewritten to that interface at some point.
1964 * Reorganizing and cleaning the internals all around the place.
1967 * Remove old default barrier/bcast buggy algorithms (see #18407)
1968 * Various bug fixes to handle more codes
1969 * Remove the need for the --foreground option of smpirun (it is still
1970 accepted for backward compatibility).
1973 * Kill the setset data container: MC don't use it anymore.
1974 * Kill the queue data container: it made more sense with GRAS.
1975 * Kill the xbt_peer_t data type: it's useless without GRAS.
1976 * Kill rm_cb feature of config sets: it was never useful.
1977 * Kill graphxml parsing feature. It was not used.
1978 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1980 - xbt_dynar_sort_strings(), when the content is char*
1981 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1982 They throw exceptions on invalid input;
1983 * C++ support for declaring CLI flags (simgrid::config::Flag);
1984 * class for abstracting different signal backends (simgrid::xbt::signal).
1985 with no external dependencies (we need very simple signals).
1988 * refactoring and cleanup of the code;
1989 * ongoing process to cleanly separate the model-checking algorithms
1990 from the code model-checking support.
1992 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1994 SimGrid (3.12) stable; urgency=low
1996 The Facelift Release.
1999 * Require g++ v4.7 at least to not speak prehistorical C++.
2000 * Require Boost 1.48 (for signal2 component).
2001 * Java must be version 7 at least when activated.
2002 * Builds on Windows again (including Java bindings).
2003 * Tracing is now always enabled (no way to turn it out)
2004 * Remove GTNetS. It was not working anyway.
2005 * Various cleanups in the cmake scripts.
2006 * Move headers around to sort them out on installed systems:
2007 - instr/instr.h -> simgrid/instr.h
2008 - instr/jedule/* -> simgrid/jedule
2009 - simdag/datatypes.h was removed
2010 - simdag/simdag.h -> simgrid/simdag.h
2011 - msg/datatypes.h was removed
2012 - msg/msg.h -> simgrid/msg.h
2015 * Interface improvement:
2016 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
2017 - Sanitize the interface in MSG_task_ module:
2018 - Merge two functions that were close enough but misleading:
2019 set_compute_duration(t) -> set_flops_amount(t)
2020 get_remaining_computation(t) -> get_flops_amount(t)
2021 - set_data_size(t) -> set_bytes_amount(t)
2022 get_data_size(t) -> get_bytes_amount(t)
2023 - Massive cleanups in the functions related to the energy
2024 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
2025 - New: MSG_host_get_pstate()
2027 - msg/energy/onoff: switching hosts on and off
2029 * Interface improvement:
2030 - Rename Host.isAvail() to Host.isOn()
2031 - Rename Process.currentProcess() to Process.getCurrentProcess()
2032 - Rename Task.setDataSize() to Task.setBytesAmount()
2033 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
2035 - #18874: Actually allows the GC to reclaim tasks
2039 - SIMIX_process_throw: raises an exception in a remote process
2040 * Refactoring: Separate sub-modules
2041 - libsmx: the public interface, as libc in a real system
2042 - popping: the strange dance that converts a user request into a kernel handling
2043 - smx_context_*: the virtualization mechanisms that embed the user code
2044 - smx_*: the handling of each simcalls
2045 * Interface cleanup:
2046 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
2047 * Rename smx_action_t into smx_synchro_t, making explicit that these
2048 things are used to synchronize processes with their environment.
2049 For example, a communication is a sort of synchronization involving
2050 the communicating processes (that may block until the exchange) and
2051 the platform. The same can be said from computations, etc.
2053 - #18888: segfault when a process ends before its kill_time
2057 - 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.
2058 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
2059 - MPI_Type_set_name, MPI_Type_get_name
2060 - MPI_*_c2f and MPI_*_f2c functions
2061 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
2062 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
2063 - Activate a lot of new tests from the mpich 3 testsuite
2065 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
2066 - InfiniBand network model added: Based on the works of Jerome Vienne
2067 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
2068 - When smpi/display_timing is set, also display global simulation time and application times
2069 - Have smpirun, smpicc and friends display the SimGrid git hash version on --git-version
2070 * Collective communications
2071 - 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
2072 - 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/).
2073 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
2075 - Replay now uses algorithms from wanted collective selector
2076 - Replay can be used with SMP-aware algorithms
2077 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
2078 - 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.
2080 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
2081 - Correctly use loopback on fat-tree clusters
2082 - Asynchronous small messages shouldn't trigger deadlocks anymore
2083 * Energy/DVFS cleanup and improvement
2084 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
2085 - new: smpi_get_host_pstate()
2088 - "Full" network optimization flag was broken since Surf++
2089 - Better handling of precision flags in maxmin
2090 - Fix bug causing sometimes "Impossible" errors
2091 - Properly pass cluster properties to included hosts
2092 * Improvement of the Energy plugin.
2093 - Always update the consumption before returning that value
2094 - New property: watt_off to denote the dissipation when the host is off
2095 - New functions getWattMinAt and getWattMaxAt to retrieve the
2096 dissipation of pstates that we are not currently at.
2097 * Java: class NetworkLink renamed to Link
2098 * New function: simcall_process_get_kill_time()
2099 * Massive rename s/workstation/host/
2100 - That's intrusive, but that's good for the project consistency. Sorry.
2101 - Change config option "workstation/model" into "host/model"
2104 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
2105 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
2106 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
2107 - xbt_dirname and xbt_basename for non-POSIX systems
2109 * The model checker now runs as a separate process.
2110 * The model checker runs must now be launched with the new simgrid-mc program.
2111 * Record/Replay: the MC can display a textual representation of a path in the
2112 execution graph. It can then be replayed outside of the model checker.
2114 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2116 SimGrid (3.11) stable; urgency=low
2121 * Normalizing pointers addresses tool for better diff between logs
2124 * Add cloud examples using new VMs
2125 - examples/msg/cloud/two_tasks_vm.tesh
2126 - examples/msg/cloud/simple_vm.tesh
2127 - examples/java/cloud/cloud.tesh
2128 - examples/java/cloud/migration/migration.tesh
2129 * Add java surf examples:
2130 - examples/java/surfPlugin/surf_plugin.tesh
2131 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
2132 - examples/java/surfCpuModel/surf_cpu_model.tesh
2133 * Add SMPI+MSG example:
2134 - examples/smpi/smpi_msg_masterslave/
2139 - msg task destroy cancel test
2140 - msg_host on/off test
2141 * Move all tests in testsuite to teshsuite (adding tesh files)
2142 * Restructure teshsuites
2143 - one folder for each kind of test
2144 * Restructure AddTests.cmake
2146 - structure the order of tests (with sections)
2149 * Add virtual machine
2150 - creation of a VM on a PM
2151 - migration of a VM from a PM to another PM
2153 - MSG_process_join(msg_process_t process, double timeout)
2154 - msg_bar_t MSG_barrier_init(unsigned int count)
2155 - int MSG_barrier_wait(msg_bar_t barrier)
2156 - void MSG_barrier_destroy(msg_bar_t barrier)
2157 - msg_as_t MSG_environment_as_get_by_name(const char * name)
2158 * New option "msg/debug_multiple_use" to help debugging when a task is used
2161 - Improvements and finalization of MSG_storage, MSG_file APIs and their
2163 - Increase code coverage in test suites
2167 * Protect context stacks against stack overflow. The number of protected memory
2168 pages allocated on the top of each stack (1 by default) can be configured
2169 with the new command line option --cfg=contexts/guard_size:<value>.
2170 * Simcalls are now generated by a python script that
2171 - generates files included by SimGrid
2172 - checks that all the functions exist, and proposes prototypes
2174 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
2175 host_set_data, host_get_data
2177 - simcall_process_join(smx_process_t process, double timeout)
2178 * Fix bug where sleeping processing could not be suspended.
2181 * Translate surf models from C to C++
2182 - Generic classes for all models: Model, Resource, Action
2183 - A generic interface for each kind of model (CPU, Network, Storage
2184 Workstation, WorkstationVM)
2186 * Translate surf routings from C to C++
2187 * Add callbacks using sigc++ or boost::signals2
2188 - Add callback functions for resource creation/destruction
2189 - Add callback functions for action state change
2190 - Handle Energy as a plugin
2191 * Replace swag by boost::intrusive
2192 * Add new routing models for clusters. For documentation, see
2193 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
2194 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
2195 parameters for cluster tag
2196 - Fat trees, with topology="FAT_TREE" and
2197 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
2198 - see examples/platforms/torus_cluster.xml and
2199 examples/platforms/fat_tree_cluster.xml
2200 * More documentation
2203 * Hostfiles support host:nb_processes construct to deploy several processes on
2205 * Collective communication algorithms should not crash if used with
2206 improper number of nodes and report the error.
2207 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
2208 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
2209 MPI_Dims_create, MPI_Cart_sub are supported.
2210 * New interface to use SMPI programmatically (still depends on MSG for
2211 some parts, see examples/smpi/smpi_msg_masterslave):
2212 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
2216 * Global variables privatization in MPI executables is now performed at runtime
2217 with the option smpi/privatize_global_variables (default:no).
2218 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
2219 dynamic libraries loaded by the application are not privatized (static
2220 linking with these libraries is advised in this case)
2223 * Options defined in XML work correctly now.
2226 * New cmake option, enable_lib_in_jar, to control whether native libraries are
2227 copied into simgrid.jar or not (ON by default). Use this option if you want
2228 to reduce the size of the installed simgrid.jar, *and* the native libraries
2229 are kept installed elsewhere.
2230 * Surf binding with SWIG (code generated in maintainer mode only):
2231 - plugin to handle callbacks
2232 - CPU model only for the moment
2235 * Supernovae build mode is definitively removed. It was used to improve
2236 inlining and inter-module optimizations. It is nowadays superseded by
2237 link-time optimizations commonly available in compilers.
2238 * Update ns-3 find lib. Bindings for ns-3 should work again now.
2239 * Add boost dependency for surf++
2240 * Add new macro for tests
2241 - ADD_TESH(name <tesh_args>)
2242 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
2246 - xbt_bar_t XBT_barrier_init(unsigned int count)
2247 - int XBT_barrier_wait(xbt_bar_t barrier)
2248 - void XBT_barrier_destroy(xbt_bar_t barrier)
2249 * Make the xbt_os_time module public
2251 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2253 SimGrid (3.10) stable; urgency=low
2255 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
2258 * Reintegrate Java to the main archive as desynchronizing these
2259 package is not acceptable anymore (Java is now considered stable)
2260 * Add explicit synchronization facilities through semaphores
2261 * Bug fix: Task.setDataSize() only changed the C world, not the value
2262 cached in the Java world
2265 * Dramatically change the way files are handled. API and internals changed, but
2266 this part of MSG was not considered as production grade either.
2267 * Add explicit synchronization facilities through semaphores
2268 * Add a new function MSG_host_get_process_list()
2269 * Preliminary DVFS support (see examples/msg/energy/ for details)
2272 * SMPI is now included directly in the libsimgrid as the windows
2273 linker doesn't force us on splitting it anymore.
2274 * Improvements of the SMPI replay tool:
2275 - Most of the collective communications are now rooted in the same process as
2276 in the original application.
2277 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
2278 used until now). Multiple data types can now be used in a trace.
2279 - The replay tool now supports traces produce either by TAU or a modified
2281 - Bug Fix: the compute part of the reduce action is now taken into account.
2282 - Gatherv collective is now supported
2283 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
2284 of smpirun outputs time independent traces for the current run. One file
2285 is created per process. If too many processes are simulated, this behavior
2286 can be changed to one file for all processes by using the
2287 tracing/smpi/format/ti_one_file flag
2288 * smpirun generates the host file if needed (with given host count and platform)
2289 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
2290 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
2291 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
2292 --cfg=smpi/coll_selector:(mpich/ompi)
2293 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
2295 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
2296 happening inside a collective SMPI call.
2297 * Fix the behavior of complex data types handling.
2298 * Make MPI_Wtime another synchronization point to take computations into
2300 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
2301 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
2302 * Add all missing Fortran bindings, SMPI should work with Fortran 90
2303 (no privatization of global variables yet)
2304 * Preliminary DVFS support (see examples/smpi/energy/ for details)
2307 * Verification of liveness properties is now available for SMPI applications
2308 (in addition to MSG applications)
2309 * Bugged examples using SMPI in examples/smpi/mc/
2310 * Add --cfg=model-check/visited option. Allows the verification of infinite
2311 programs. Detection of loops in the execution thanks to the system state
2312 comparison and reduction of the state space to explore. Can be combined with
2313 DPOR for safety properties.
2316 * Allow to change SimGrid configuration (see --help) within the code
2317 thanks to SD_config() as it can be done in MSG.
2318 * Add a new function SD_task_set_amount() upon user request.
2321 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
2322 * Remove rule based routing (no more PCRE dependency)
2323 * Add a limiter_link option to cluster tag, to specify a maximum reachable
2324 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
2325 * Add a loopback_bw and loopback_lat options to cluster tag.
2326 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
2327 routing. Now peers are encapsulated in an AS and have their own private
2328 router but this is transparent.
2331 * Our own implementation of getline is renamed xbt_getline, and gets
2332 used even if the OS provide a getline(). This should reduce the
2333 configuration complexity by using the same code on all platforms.
2334 * New type: xbt_cfg_elm_boolean.
2335 * Allow to use yes/no for boolean configuration options in the command line.
2336 * Allow to disable SimGrid cleanups at exit from command line option.
2337 There are situations where one may want a simulation to end with an exit.
2338 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
2339 annoying when scripting around the simulator. Adding a
2340 --cfg=clean_atexit:no allows to circumvent this issue.
2343 * Lots of memory leaks were corrected in this release.
2344 There is no known memory leaks anymore, in all of our 600+ tests.
2345 * New command line option --version, to get SimGrid version information.
2346 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
2348 * Supernovae builds are deprecated, and expected to be removed in the next
2351 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2353 SimGrid (3.9) stable; urgency=low
2355 The Grasgory release: GRAS is really dead now.
2357 * Complete overhaul of the internal host structures scheme.
2360 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
2361 was considered as experimental and badly maintained since then.
2362 * Keeping it was thus a trap to our potential users, that could take
2363 it instead of MSG or SMPI by mistake despite is pity state.
2364 * GRAS seems to have very few users (if any), and no one volunteered
2365 to maintain it further. It also induces a lot of XBT code (for
2366 portability sake), that must be maintained too.
2367 * For all these reasons, we killed GRAS. If someone wants to revive it
2368 in the future, don't cry, our git history still remembers of GRAS.
2371 * Major overhaul. Merge our documentation again as time proved that
2372 splitting it was really not helping our users.
2373 * Further improve the developer documentation to help newcomers
2374 hacking on SimGrid itself. The user documentation (and in
2375 particular, the beginner documentation) is still in a sorry state.
2378 * Now works on Windows too!
2379 * Much more extensive test suite, from MPICH
2382 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
2383 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
2384 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
2385 * Bug fix: let task be scheduled when the last dependency to be solved is
2386 a control dependency.
2387 * Remove SD_load_environment_script function.
2388 Use the C sg_platf function if you want to declare a platform
2392 * New function: MSG_process_get_number()
2393 * Old function documented: MSG_config()
2394 * Remove MSG_load_platform_script function
2395 Use the C sg_platf function if you want to declare a platform
2399 * Change the default value of the TCP_gamma constant (maximal size of TCP
2400 congestion window) to a more realistic 4MiB value. If you notice changes in
2401 your simulation results, you can fall back to the previous 20k tiny window
2402 by adding --cfg=network/TCP_gamma:20000 on command line.
2403 * (Hopefully) fix a bug wrt periodic availability and state traces
2404 * Bug fix: use default values at start when first event in availability/state
2405 trace is not at time 0.
2408 * remove the "new_" part of function name sg_platf_new_trace_connect
2409 (resulting in sg_platf_trace_connect), since it does not create
2413 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2414 use case where it's really mandatory, and maintaining it was a pain in
2416 * New: xbt_fifo_search(), search an item with a user-provided
2417 comparison function instead of dumb pointer comparison.
2420 * Fix the lua deployment:
2421 Use `simgrid.init_application()` before deployment instead of
2422 `simgrid.msg_register_application()` after.
2425 * Transfer the tracing files into the corresponding modules.
2427 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2429 SimGrid (3.8.1) stable; urgency=low
2431 The "we are told that some people want to also *install* the simgrid
2434 * Add missing file "tesh.1" to the archive.
2436 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2438 SimGrid (3.8) stable; urgency=low
2440 The Psssshiiiit release: SimGrid jumps into the Cloud.
2443 * Add an experimental interface to manipulate VMs. They are mainly
2444 process groups with very few intrinsic semantic, but they should
2445 allow you to build the semantic you want easily.
2446 * New function: MSG_host_set_property_value()
2447 * New function: MSG_process_on_exit(). To clean memory in all cases.
2448 * Bug fixes that made the host (and link) failures unusable.
2449 * Add a way to auto-restart process when the host in which they are
2450 executing comes back (ON_FAILURE="RESTART" on deployment file,
2451 MSG_process_auto_restart_set).
2452 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2453 please stop using the old ones, they are DEPRECATED.
2455 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2456 Please use MSG_init() instead. (reducing the amount of entry
2457 points in the library helps us).
2458 * Make it impossible to link against the wrong version of the lib
2459 * Deprecate MSG_clean(). No need to call it anymore.
2460 * Function MSG_get_host_number() is not deprecated anymore.
2463 * Split the doc into a user guide and a reference guide.
2464 * Start a developper guide to help people hacking on SimGrid.
2467 * Enable tracing by default. This modules rocks you should use it.
2468 * Remove option custom_flags. Now use environment variables CFLAGS
2470 * Use default cmake things to detect lua instead of home grown ones.
2471 * New option "enable_mallocators" to disable mallocators, for debugging
2472 purpose ("on" by default).
2475 * Bug fixes around the resource failures: don't let the processes
2476 survive the host they are running onto.
2477 * Add an interface to auto-restart processes when the host in which they are
2478 executing comes back.
2479 * Ensures that SIMIX_clean is called automatically. It's not part of
2480 the public interface anymore (bindings should be updated).
2483 * Bug fix for when SD_Simulate is called with a positive value: be careful
2484 when comparing doubles. Sometimes they are different for non significant
2486 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2487 parallel task whose initial work is distributed among host according
2488 to the Amdahl's law. Such tasks are created with a parameter alpha
2489 that corresponds to the non-parallelizable part of the computation.
2490 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2491 between two sets of workstations assuming a 1D block distribution (each
2492 workstation owns a similar share of data) on both sides.
2494 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2495 Data redistribution will be automatically scheduled once parent and child
2496 are both scheduled. The filling of computation_amount and
2497 communication_amount structures is now done seamlessly thanks to the chosen
2499 * New function SD_workstation_dump to display various information
2500 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2501 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2502 bandwidth on the route onto which the task is scheduled and the amount of
2504 To divide the nominal bandwidth by 2, the rate then has to be:
2505 rate = bandwidth/(2*amount)
2506 * Compute tasks that have failed can now be rescheduled and executed again
2507 (from their beginning)
2508 * Increasing source code coverage (src/simdag is now covered at 95.8%
2512 * Re-implement time-independent trace replay using SMPI (at the
2513 smpi_smp_* level) instead of MSG. This should replace
2514 examples/msg/actions/actions.c
2515 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2516 hindexed and structs)
2517 * Implement the exchange of non-contiguous data.
2518 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2519 * Correct behavior of smpi/sender_gap and set its default value to 0
2520 * Add option to asynchronously send small messages to allow better
2521 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2522 specifies the size in bytes under which messages will be asynchronously sent.
2523 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2524 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2527 * New C interface to define a platform: XML is now optional.
2528 For more info, please check include/simgrid/platf.h.
2529 * New interface to define random platforms from the C:
2530 For more info, please check include/simgrid/platf_generator.h and
2531 examples/msg/masterslave/masterslave_platfgen.c
2532 * Export a sg_cmdline dynar containing all the arguments we got from
2536 * Two new tracing options for adding comments to trace file so you
2537 can track your experiments (see --help-tracing for details).
2538 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2539 * Adding the SimGrid version that generated the trace file as a comment.
2540 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2541 * Fix to avoid key clashes on Paje links
2542 * Other minor fixes related to the Paje specification
2545 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2546 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2547 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2548 * Greatly improve the robustness of mmalloc to user errors (such as
2549 using an area after freeing it, or freeing it twice)
2551 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2553 SimGrid-java (3.8.1) stable; urgency=low
2555 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2556 Warning: all methods are visible, but only some of them are
2557 implemented so far. Check the source (src/jtrace.c)
2558 for further information.
2559 * New module: org.simgrid.msg.File (SimGrid File management functions)
2560 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2561 * Change the meaning of Process.restart: now restart the process from
2562 the beginning, like MSG_process_restart in C.
2563 * Add Process.setAutoRestart: handling of process restart when failed
2565 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2566 you to retrieve the properties of the processes/hosts
2567 * Deprecate Msg.clean(): you can just forget about it now.
2568 * New function Process.getCount(), that only works when compiling
2569 with the not yet released version 3.9 of the C library.
2571 * New context factory based on Coroutines. It mandates a modified JVM
2572 but then, the simulations run about five times faster, and there is
2573 no limit to the amount of processes (beside of the available memory).
2575 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2577 SimGrid (3.7.1) stable; urgency=low
2580 * Restore the prototype of MSG_process_create_with_environment() to
2581 the pre-3.7 situation by removing the kill_time argument.
2582 * Add a MSG_process_set_kill_time() function instead.
2585 * Fix weird behaviors when dealing with parallel tasks.
2588 * Simgrid is now built as a dll.
2589 * Simgrid-java now works on Windows.
2590 * Simgrid-Java is now included into Windows package.
2593 * First pre-build package for MacOSX.
2596 * Fix compilation when using MSG_USE_DEPRECATED.
2597 * Fix some compilation issues on Macs and Windows.
2598 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2599 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2601 We discovered that the Lua console is broken, but we are missing the
2602 manpower to fix it right now. The problem existed in 3.7 too, so we
2603 are not blocking the release for that. Sorry if you depended on this
2604 feature, any help would be really welcome.
2606 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2609 SimGrid-java (3.7.1) stable; urgency=low
2611 The "Java aint got to be bloated and slow" release
2614 * Various internal cleanups and performance improvement
2615 Simulations are expected to run up to twice faster or so
2616 * Make Process.kill(process) an instance method, not a static one
2617 * User processes are not java.lang.Thread subclasses.
2618 This breaks the compatibility (sorry), but previous API was
2619 brain-dead, making it impossible to have non-trivial
2620 initializations in the process constructor.
2621 * Require a full constructor per Process sub-class.
2622 Kinda breaks the compatibility (sorry), but this allows a much more
2623 efficient way to launch the processes at simulation startup.
2624 * Do not embeed our version of semaphores, java 1.5 can be considered
2625 as sufficiently prevalent for us to not dupplicate its features.
2629 * Add examples for almost every part of the API
2630 We spotted and fixed a lot of bugs in the process
2632 * New module: asynchronous communication API
2633 * New function: Process.sleep()
2634 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2635 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2636 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2638 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2640 SimGrid (3.7) stable; urgency=low
2642 The "spring cleanups (before the next Big Project kicks in)" release.
2645 * We can specify the SMPI latency/bandwidth factor with command line
2646 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2647 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2648 You can also use the "config tag" from platform file by adding this line
2649 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2650 (see "example/platforms/tag_config.xml" to use "config tag").
2651 Note that the command line supersedes the platform file configuration.
2652 * Change the correction factors used in LMM model, according to
2653 the latest experiments described in INRIA RR-7821.
2654 Accuracy should be improved this way.
2655 * Use the partial invalidation optimization by default for the
2656 network too. Should produce the exact same results, only faster.
2657 * Major cleanup in surf to merge models and split some optimization
2658 mechanisms from the core of the models. As a result you can now
2659 specify which model to use (e.g., --cfg=network/model:LV08
2660 --cfg=cpu/model:Cas01) and which optimization mode to use
2661 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2662 Incompatible combinations should err at initialization. See
2663 --help-models for the list of all models and optimization modes.
2664 * The CLM03 workstation model was dropped for simplicity because it
2665 used the deprecated CM02 network model. Use default instead.
2666 * Rename the TCP_gamma configuration option to network/TCP_gamma
2667 * Rename the coordinates configuration option to
2668 network/coordinates, and document it
2669 * Use now crosstraffic keyword instead of the terribly misleading
2670 fullduplex keyword. It is activated by default now in the current
2671 default model, use --cfg=network/crosstraffic:0 to turn it off.
2672 * Ongoing refactoring the model parsing to make XML files optional
2673 See include/simgrid/platf.h for details (still to be completed)
2676 * Major overhaul of the documentation. Almost instructive now :/
2677 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2678 functions and friends. This interface was considered as
2679 deprecated since over 2 years, it's time to inform our users that it is.
2680 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2681 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2682 use these (crappy) functions in your code.
2683 These functions will be removed soon. Stop using them now.
2684 * Deprecate MSG_get_host_{table,number}
2685 Implement MSG_hosts_as_dynar() instead.
2686 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2687 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2690 * Stabilize the parallel execution mode of user contexts
2691 * Introduce configuration variables to control parallel execution:
2692 - contexts/synchro: Synchronization mode to use when running
2693 contexts in parallel (either futex, posix or busy_wait)
2694 - contexts/parallel_threshold: Minimal number of user contexts
2695 that must be part of a scheduling round to switch to parallel
2696 execution mode (raw contexts only)
2697 * Fix bugs that prevented to use suspend/resume along with
2698 synchronization structures.
2699 * Fix bugs in process termination that lead to invalid memory access
2700 in very specific conditions.
2703 * Introduce a parallel mode for the models (controlled by surf/nthreads
2704 configuration item). In our tests, running the models in parallel
2705 never lead to any speedups because they are so fast that the gain
2706 of computing each model in parallel does not amortizes the
2707 synchronization costs, even when ultra fast futexes are used.
2708 This is released anyway because YMMV.
2711 * Performance boost by using a swag internally to compute the set of
2712 tasks that are finished and should constitute the return value of
2716 * Enable it by default now that it is considered rather stable.
2719 * Documentation of the tracing functions.
2720 * Performance gains when tracing categorized/uncategorized resource
2721 utilization by avoiding calls to get route when updating resource
2722 variables. LMM constraints are being used instead.
2723 * API changed to set task categories. Use MSG_task_set_category instead
2724 of TRACE_msg_set_task_category, and SD_task_set_category instead
2725 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2726 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2727 * Examples for link user variables added (see at examples/msg/tracing/)
2728 * Deprecated function TRACE_msg_set_process_category completely removed
2729 * Trace header updated according to the latest Paje file format
2730 * Tracing network lazy updates, no longer obligate users to use full updates
2731 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2732 * Experimental: let user code declare/set/push/pop application states for hosts
2733 * API changed to allow the manual creation of graph configuration files
2734 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2737 * Improve the API of Lua MSG bindings, using the Lua spirit.
2738 * Each simulated process now lives in its own Lua world (globals are
2739 automatically duplicated). It helps writing simulators. Will allow
2740 to run Splay programs within SimGrid in the future.
2741 * Add a Chord example in Lua, equivalent to the MSG one.
2744 * Start the implementation of a solution to express temporal
2745 properties, not only local assertions. This is still an
2746 experimental work in progress, stay clear from it to be safe.
2750 - Add new runtime parameters --help-logs and --help-log-categories
2751 to display information about supported logging parameters and
2753 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2754 don't exists anymore.
2755 * Mallocators: allow value NULL for the reset function.
2757 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2758 create homogeneous dictionaries, where all the elements share the
2759 same free function. Non homogeneous dictionaries will be
2760 deprecated in the next release.
2761 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2762 - Multi-level dictionaries are deprecated.
2764 - new function xbt_dynar_search_or_negative() that is useful when
2765 you have less than 2 million elements in your dynar and don't
2766 want of the extra complexity of catching exceptions when the
2767 element is not found.
2769 - Make xbt_os_thread module (for thread portability) public.
2770 Documentation is still to come, sorry.
2772 - Cleanups and simplifications to make it maintainable again.
2773 - Exotic features (such as memalign and valloc) were removed.
2774 - The metadata were extended and improved so that the
2775 model-checker becomes able to explore and inspect the heaps.
2776 - This may induce a performance drop when enable_model-checking is
2777 ON in cmake (even if it's not used in the simulation), but it is
2778 necessary at this point to get MC working.
2780 Turn model-checking OFF if simulation performance matters to you.
2781 Not enabling it at runtime is not enough, disable it in cmake.
2783 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2786 SimGrid (3.6.2) stable; urgency=low
2788 The "Not coding new stuff allows to polish old things" release.
2791 * New bindings to the NS3 packet level simulator (experimental)
2792 * Use the raw (efficient) execution contexts instead of the sysv
2793 (portable) ones when possible.
2794 * libpcre is now mandatory in any cases since not using it led to
2795 severe performance loss and possibly other issues
2796 * Update the XML platforms:
2797 - G5K: include the latest machine in Nancy
2798 - GridPP and LCG: new platforms
2799 * Documentation was partially updated, at least (more to come)
2801 Bug fixes, cosmetics and small improvements
2802 * Free terminated processes before the end of the simulation to avoid
2803 exhausting the memory of users having very dynamic amount of
2805 * Bug fix and cosmetics about canceling non-running tasks
2806 * Bug fix about the dot loader's issues when using libcgraph
2809 * Create an installer for windows with nsis (amd64 and win32)
2810 - Add an hello world project to illustrate SimGrid project creation.
2811 - Embed libpcre into the SimGrid installer to avoid
2812 its compilation burden
2813 * The raw execution contexts should work on Apple now
2814 * Port to Windows 64 bits
2815 - Sysv contexts now have an implementation for this arch
2816 - GRAS communication features now support this arch
2817 * Drop support for borland compiler on windows
2818 - this code was not maintained, and we kinda depend on gcc nowadays
2819 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2820 * Fix portability issue on unstable ubuntu: linker became picky on
2823 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2825 SimGrid (3.6.1) stable; urgency=low
2827 The "Oops, we broke Macs too" release
2830 * Fixed contexts detection so that raw ones are used when possible
2831 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2832 leads to a strange error, with user code segfaulting sometimes when
2833 the generated code is not perfectly aligned (which is not
2834 controllable from the user side, depends on the amount of code)
2837 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2839 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2841 SimGrid-java (3.6) unstable; urgency=low
2844 * Split of every thing from SimGrid v3.5 into a separate package.
2846 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2848 SimGrid (3.6) stable; urgency=medium
2850 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2853 * Bindings now constitute their own package, separated from the main one.
2854 Rationale: reduce our maintenance nightmare by reducing the module coupling
2855 They will soon be released on their own on gforge.
2857 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2858 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2860 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2861 * It's not quite deprecated for now because we have no replacement,
2862 but it may soon become the case.
2865 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2866 * Fortran: New user-level cache variable to store the rank of the running
2867 process. This improves performance by an order of magnitude.
2868 * C: New coccinelle script to automatically locate and modify global and
2869 local static variables.
2870 * Improved SMPI network model with a sender-side gap to account for multiple
2874 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2875 only say if a communication is finished, no matter whether it succeeded or
2876 failed. You can call MSG_comm_get_status() to know the status of a finished
2878 * New function MSG_task_dsend() to send a task and detach it. When a
2879 communication is detached, you are never notified of its success or failure
2880 and the memory is released automatically once it is finished. This function
2881 is useful when you don't care about the end nor the success of a
2883 * Change the prototypes of action replay. Sorry for inconvenience,
2884 but this is really more efficient this way (and to adapt your code,
2885 you just have to fix the initialization, that shouldn't be too long)
2886 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2887 ever managed to do anything useful with it.
2888 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2889 takes a dynar of communications. It returns immediately and gives the
2890 index of a finished communication (if any).
2891 * New example: a basic implementation of the Chord P2P algorithm.
2894 * New model for multi-core CPUs. You can now use the core attribute to
2895 precise the number of cores of a host. This is a basic model. Every
2896 process running on the host receives at most the power provided in
2897 the DTD (throughput<=power). Total throughput of process cannot exceed
2899 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2900 router linked by an up-link and a down-link (possibly asymmetrical).
2901 This kind of pattern allows to easily build last-mile model style platforms.
2902 Aggregating such patterns in a rule-based AS is thus the technique of
2903 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2904 * New model for Vivaldi routing. We transformed the Vivaldi network model
2905 into a Vivaldi routing model (based on the rule-based model). This allows to
2906 combine Vivaldi based latencies with last-mile platforms.
2909 * Added a check for NaN of IEEE754 infinite in the double entries of
2911 * Introduce a new context factory "raw", highly inspirated from the
2912 ucontext factory, but using manually crafted functions in assembly to
2913 do the work in an efficient manner.
2914 * Allow to change the used context factory at run time, not only at
2915 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2916 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2917 contexts at runtime (only with raw contexts or ucontexts).
2918 * Completely rewrote this module to allow parallel execution of user
2919 processes. Use --cfg=contexts/nthreads:N to execute user processes
2920 with N parallel threads (the default is 1, meaning no parallelism).
2921 * Allow to decide dynamically between sequential and parallel modes.
2922 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2923 processes in parallel only when their number is greater than or equal to P
2925 * Added a check for NaN of IEEE754 infinite in the double entries of
2929 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2930 won't output the state of processes when interrupted with Ctrl-C
2931 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2932 NULL-terminated array. This may solve backward compatibility issues
2933 due to the change to return type of SD_simulate. See also:
2934 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2935 * Add new macros with variable number of arguments.
2936 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2937 - in xbt/asserts.h: xbt_assert
2938 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2939 - in xbt/ex.h: THROWF and RETHROWF.
2940 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2942 * Change xbt_die() to accept a format string with arguments, just like printf.
2943 * New data structure: xbt_lib_t, like a dict but more general and with better
2947 * New configuration options
2948 Options triva/categorized and triva/uncategorized can be used to generate
2949 graph configuration files for Triva visualization tool.
2950 * Configuration option tracing/platform is renamed to tracing/categorized
2951 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2952 * New TRACE_user_link_variable interface:
2953 User provides the name of the link and the tracing variable to attach to it
2954 * the declaration of tracing categories must be done after the environment
2956 * simpler tracing interface, just one way to declare categories
2957 TRACE_category or TRACE_category_with_color, it is up to you
2958 * links in the trace file are again identified by their names
2959 * trace contains the full platform hierarchy exactly as declared using the ASes
2960 * Options tracing/msg/[task|process]:1 groups the process by hosts
2961 for both cases, tasks and processes must have names that are unique during the simulation
2962 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2963 * The experimental option tracing/msg/volume is deprecated
2964 its functionality may be reincorporated if needed
2966 The tracing generates a trace file with unordered timestamped events,
2967 because of the way the core simulator (surf) works. A script available
2968 at the tools directory (fix-paje-trace.sh) can be used to put the events
2969 in order. We have changed the tracing so it can generate ordered timestamped
2970 events in the final trace, but depending on the simulator (and how much time
2971 is simulated) that can lead to a huge memory utilization. It is deactivated
2972 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2974 Build Infrastructure
2975 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2976 - We are trying hard to keep the API stable, but it may happen that
2977 some things change (we're a research project after all, not a
2978 nuclear plant operating system). If such things should happen, you
2979 could rely on that macro to adapt.
2980 - current value: 30600 for 3.06.00, aka 3.6
2981 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2982 a number that can be compared with SIMGRID_VERSION.
2983 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2984 assertions and verbose/debug logging events are disabled at compile time.
2986 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2988 SimGrid (3.5) stable; urgency=medium
2991 * New feature to any SimGrid-based simulator: Model-Checking
2992 Check SIN#1 for more details.
2995 * New Model SMPI (three-interval linear regression for correction factors)
2996 See RR-7426, available at http://hal.inria.fr/inria-00527150
2997 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2998 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2999 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
3000 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
3001 * See also src/smpi/README
3006 - Tracing API changes: TRACE_start and TRACE_end should not be called
3007 by user-code. They are automatically called by simulators created
3008 with SimDAG, MSG and SMPI if the toolkit is compiled with
3009 tracing_enabled=ON. Categories declaration and utilization remain the
3010 same for MSG and SimDag.
3011 - A function was added to the tracing API to declare categories with
3013 - TRACE_category_with_color (char *category, char *color)
3014 where color must be in the following format
3015 "%f %f %f", red, green, blue
3016 and red, green, blue are float values in the interval [0, 1]
3017 - User can specify NULL as color parameter, or continue calling
3018 TRACE_category (cat)
3019 On that case, the tracing system will define random colors
3020 - The following command-line options are supported:
3021 --cfg=tracing/filename:msg.trace
3022 --cfg=tracing:1 (activate tracing, needed to use others)
3023 --cfg=tracing/platform:1 (categorized resource use)
3024 --cfg=tracing/uncategorized:1 (uncategorized resource use)
3025 --cfg=tracing/msg/task:1 (task creation)
3026 --cfg=tracing/msg/process:1 (process creation, migration)
3027 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
3028 --cfg=tracing/smpi:1 (SMPI interface tracing)
3029 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
3030 - examples of examples/msg/tracing updated
3032 - DAXLoader and DOTLoader functions can generate tasks with categories
3033 - A new function to attribute a category to SD tasks:
3034 TRACE_sd_set_task_category (SD_task_t task, char *category)
3035 * Tracing the MPI interface implemented by SMPI
3036 - Collective operations are traced with states
3037 - Point-to-Point operations are traced with states/links
3038 - Tracing activated by a parameter "-trace filename" passed
3039 to smpirun during execution (considering that simgrid
3040 is compiled with tracing enabled)
3041 - To run the simulation with gdb, the simulator
3042 accepts --cfg=tracing/smpi:1 to trace SMPI
3043 - tesh files to check if smpi tracing is ok
3044 - See examples/smpi/NAS/DT-trace
3045 * GTNetS tracing re-worked
3046 - adaptation to the tracing system of GTNets to cope
3047 with modifications regarding the fullduplex mode
3048 - new tesh files to check if gtnets tracing is ok
3051 * Asynchronous communications through the functions:
3052 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
3053 * New function: MSG_load_platform_script()
3054 to make possible using a lua script instead of XML files to set up platforms
3055 * New function: MSG_set_function
3056 to associate functions to processes, used when bypassing the parser
3057 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
3059 Platforms: Add some more examples in examples/platforms
3060 * Grid'5000: see www.grid5000.fr
3061 * *_30000_hosts.xml: various huge files [mainly scalability testing]
3064 * Change the XML format. This is a very important modification. SimGrid
3065 3.5 introduces a new hierarchical format based on the notion of
3066 Autonomous Systems. Compatibility with old format is ensured
3067 through the perl script provided in the install bin directory
3068 bin/simgrid_update_xml.
3069 It is now possible to build platforms with specific routing
3070 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
3071 connect several platforms together. We will try to provide soon
3072 set of realistic platforms exploiting these properties (have a
3073 look at examples/platforms/ for the moment).
3074 * Take the opportunity of the XML format change to be a good XML citizen:
3075 rename link:ctn to link_ctn and similar changes (also dealt with by
3077 * Add a new routing scheme (rule-based) using regular expressions. It
3078 enables to have an extremely low memory footprint when the
3079 underlying routing is simple and can be compactly described. You
3080 need to have libpcre4-dev (perl regular expressions) installed if
3081 you want to use this routing scheme.
3082 * Revive the cluster TAG and allow to easily and efficiently (both in
3083 term of memory and speed) connect clusters together. Have a look
3084 at teshsuite/simdag/platforms/ to see how this can be done. With
3085 this tag, you can create clusters with thousands of tasks at no
3086 cost (have a look at examples/platforms/).
3087 Note: clusters are implemented as ASes, so there is no need for an
3088 enclosing AS tag if you have only one cluster in your platform.
3089 * Add new generic functions in the public interface that allows the user
3090 to call SURF 'create_resource' methods from your code (same
3091 functionality as the XML bypass mechanism but with a much lighter
3093 * Add a new model (enabled through command line --cfg=network/model:SMPI)
3094 that uses a piecewise linear approximation to produce better
3095 results when exchanging small messages.
3096 * Add a new parameter to handle correctly full duplex link and account
3097 for interferences between uplink and downlink communications
3098 (activate with --cfg=fullduplex:1).
3101 * Rename the SD_READY (all dependencies are satisfied and task is
3102 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
3103 dependencies are satisfied) state.
3104 This prevents a confusion between the notion of "ready to schedule"
3105 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
3106 simulated" (SD_RUNNABLE) used by the simulation kernel.
3107 * Change the way a task is considered as ready. Instead of removing
3108 dependencies when a task is done, a counter is decreased. This way,
3109 it is always possible to reach ancestors thanks to the
3110 SD_taks_get_parents function (even after the end of the simulation.)
3111 * Change the return type of SD_Simulate from (SD_task_t*) into
3112 xbt_dynar_t. This function was in handling a dynar internally and
3113 converted it into a NULL terminated array for historical reasons.
3114 * New function SD_dotload(char*) to load a DAG described in dot
3115 format. This loader and the corresponding examples require the
3116 installation of the graphviz library.
3117 * Fix a bug in the management of tasks of size 0 in the surf network
3118 models. This problem was only visible with SIMDAG and you should
3119 thus disregard results produced with earlier versions if you
3120 relied on this feature (some tasks were blocked because of this).
3121 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
3122 even though your applications were doing only point-to-point
3123 communications and sequential computations. Now you can really use any
3124 model you want (of course, if you create real parallel tasks, which are
3125 not implemented in most models beside ptaskL07, this will abort).
3126 * Add an example that schedules a DAX on an heterogeneous platform
3127 using a Min-Min strategy.
3128 * New function SD_workstation_get_current_task() that returns the kind
3129 of task currently running on a workstation in the sequential access
3131 * Raise some warnings when unexecuted tasks remains at the end of the
3132 simulation. This is usually caused by cycles in the DAG.
3135 * New function: SIMIX_process_set_function() called by MSG_set_function
3136 * Change the underlying waiting queue in semaphores so that a process
3137 can wait on several of them simultaneously (as in waitany).
3138 * Fix the way to handle tokens in semaphores so that all access patterns
3139 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
3140 * kill the dirty pimple SIMIX_message_sizes_output()
3141 Please use (proper) visualization instead
3144 * New data container: setset (set of sets of elements)
3145 * New module: mmalloc (mapped malloc, allowing to have several
3146 independent segments of malloc)
3147 * New function: xbt_dict_cursor_set_data()
3148 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
3149 * New function: xbt_dynar_is_empty()
3150 * New function: xbt_fifo_get_last_item()
3151 * Fix xbt_dynar_shrink(): use the right element size.
3152 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
3153 uninitialized areas during expand.
3154 * Fix semaphores: previous implementation was severely broken.
3155 * Use library init/fini functions for our initialization.
3156 - you can use logs and other feature as soon as you want in your
3157 code (even before the xbt_init / MSG_init)
3158 - xbt_exit is now a no-op and produce a warning when used.
3161 * Port GRAS to new SIMIX mechanisms. This allows gras users to
3162 benefit from the latest improvement to the simulation kernel.
3163 * Kill measurement sockets for now. If you rely on them, sorry. This
3164 release is not for you. This feature will be reintroduced in the
3165 future, but we cannot delay the release any further.
3166 * New function: gras_msgtype_get_name().
3167 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
3168 * Fix (at last) the pmm example: it should not randomly fail anymore.
3170 Build chain: bug fixes and overall polishing
3171 * Cmake is now stable enough. Hence, we killed the autotools.
3172 * Port to windows ( TM :)
3173 * Fix the 'make install' target.
3174 No need to use 'make install-simgrid' anymore
3175 * Introduce a 'make dist' target compiling a *source* archive
3176 'make package' compiles a binary archive
3177 * Compile java files only on need
3178 * Add --cd and --setenv command line options to tesh
3179 * Out of source builds are not fully supported yet, but we are close
3180 * Enable supernovae and optimization flags by default for our users
3183 * Add layer to set up environment directly from lua, without XML.
3184 * The effect of gras_stub_generator can be achieved through
3185 lua too (check examples/gras/console/ping_generator.lua)
3187 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
3189 SimGrid (3.4.1) stable; urgency=low
3191 The "Polishing easter eggs is probably a good idea" release.
3192 This is a bug fixes release only.
3196 * Fix a bug preventing the tasks from begin garbage collected.
3199 * Fix a bug occurring when a host involved in a communication fails.
3200 This was not detected properly by the other peer involved in the
3201 communication. Now, it's reported as a network error.
3204 * Warn the user about loop dependencies in data flow of DAX files
3205 * Obey the control-flow dependencies of DAX files
3208 * Add option "enable_smpi" allowing to not compile SMPI.
3209 Probably useful for the (Mac) users experiencing a build error here
3210 * Improve the detection of lua5.1 and ruby1.8
3212 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
3214 SimGrid (3.4) stable; urgency=low
3216 The "Easter in Cargese" release. Also known as (major changes):
3218 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
3219 ~> bindings were greatly improved
3220 ~> new tracing infrastructure for better visualization introduced
3222 * the "Welcome to configury modernity" release.
3223 ~> we switched from autotools to cmake, and improved our cdash
3226 A more detailled list of changes follow (full detail in svn log).
3228 Java Bindings: Various Cleanups
3229 * (install java-gcj-compat-dev on debian-like to use them)
3230 * Remove put/get: no need to export deprecated interface in Java
3231 Use send/receive instead.
3232 * Cleanup the examples and add a README per directory
3233 * Remove example autoDestination (that's the only way to go now)
3234 * Remove example explicitDestination (was a plain copy of basic)
3235 * Make JniException a runtime exception, so that there is no need to
3236 declare the fact that you may encounter such a beast. I guess that
3237 nobody will ever want to survive such error.
3238 * Create specific errors for each MSG case of failure:
3239 host failure, transfer failure, timeout, task cancelled
3240 * Cleanup the exceptions that may get thrown by each function
3241 * Other internal cleanups in Java bindings. Performance still bad :/
3242 Ruby and Lua Bindings: create them
3243 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
3244 * That's new and great, you should try them out.
3245 Same functionalities than Java bindings, only even less polished
3247 * Kill the useless "rate" argument of SD_task_get_execution_time()
3248 Everyone used to provide -1 as a value, it was not used, and the
3249 semantic of a possible use wasn't even clear.
3250 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
3251 either as comm costs or compute costs to mean that there is no
3252 such thing for that specific task.
3253 * Add a SD_task_set_name() function
3254 * Fix SD_task_unschedule() on typed tasks
3255 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
3256 * In DAX loader, accept useless 'level' attributes to <job> since
3257 LIGO DAGs have them (seem to be to ease graphical representation).
3259 * Add an example masterslave_mailbox.c using send/receive and not
3260 the deprecated put/get interface.
3261 * Kill the MSG_paje_output() function. It's a noop since 2 years.
3262 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
3264 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
3265 (declare MSG_USE_DEPRECATED to still have the old name)
3266 * Add a MSG_task_set_data() function
3267 * About trace replay (see examples/msg/actions):
3269 - Allow to work with splitted trace files for each process
3270 Give the specific trace file as argument of each process,
3271 and call MSG_action_trace_run(NULL)
3272 You can still have one merged file for all processes.
3273 - Fix implementation of collective operations
3274 * Allow task_execute() on 0-sized tasks (closes #10063)
3276 * This is the first release of SimGrid where SMPI is not considered
3277 beta anymore (even if some corners should still be improved)
3278 * Port over the new SIMIX_network submodule (internal refactoring)
3279 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
3280 * Implement more missing elements of the standard:
3282 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
3284 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
3285 MPI_Scan() MPI_Get_processor_name()
3286 - Added implementation of missing case for Alltoall (warning: it's
3287 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
3288 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
3289 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
3290 SMPI_MPI_Allgatherv()
3291 * Bug fixes include:
3292 - MPI_Waitsome() was broken
3293 - Allow relative includes in smpicc
3294 - Command line cfg argument 'reference_speed' was ignored...
3295 - Some functions did not properly lead to auto-benching of user code
3296 - smpicc passes -O2 by default (just like openmpi one)
3298 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
3299 * Bug fixes about timeouts during communications
3300 * add SIMIX_message_sizes_output() as a pimple to write to file the
3301 amount of messages per size. Use gnuplot to get histogram.
3302 Pimple because that's the only user-visible function of simix,
3303 defined directly in xbt.h (irk, sorry)
3305 - Add a SIMIX_sem_get_capacity() function
3306 - Fix interactions with processe resume/suspende
3307 - release_forever() was stupidly broken
3308 - Fix SIMIX_display_process_status() for processes in a semaphore
3309 - Make SIMIX_sem_block_onto() user-visible
3310 * Refactoring context stuff:
3311 - Use pseudo-OOP for better modularity
3312 - reimplement SIMIX_process_kill() without process_schedule() so
3313 that the latter can take as invariant that it is called from
3315 - Merge context_start into context_new for sake of simplicity
3317 * Add a Vivaldi network model, coded live during SUD'10 ;)
3318 * Rename configuration variables to start a hierarchy:
3319 o cpu_model -> cpu/model
3320 o network_model -> network/model
3321 o workstation_model -> workstation/model
3322 * New configuration variables:
3323 o network/bandwidth_factor: correction to bandwidth
3324 o network/latency_factor: correction to latency
3325 o netwotk/weight_S: correction to the weight of competing streams
3326 * Add a long description to the models, that users can see with such
3327 argument on the command line: --cfg=cpu/model:help
3328 * --help-models display the long description of all known models
3330 * config: add the ability to set a default value after registration
3331 Does not override any previously set value (e.g. from cmd line)
3332 * dict: allow to have integer key and data.
3333 When so, you need to use the following functions
3334 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
3335 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
3336 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
3337 In contrary to regular dicts, the key is not malloced before copy.
3338 Mixing scalar and regular elements in the same dict is not tested
3340 * Allow to use xbt_dynar_shrink() to expend the dynar instead
3341 Tracing for Visualization:
3342 * SimGrid is now instrumented in order to generate a trace file for
3343 visualization analysis: to use it, need to compile SimGrid with the
3344 "tracing" option enabled, and instrument the program using SimGrid with
3345 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
3346 (among other functions).
3347 * The instrumentation only traces the platform utilization for now
3348 * Documentation to use the tracing functions and how to analyze the
3349 traces with the Triva tool is written.
3350 * More information about: SimGrid FAQ (in the section Tracing Simulations
3353 * We moved to cmake as default build system. Autotools support will
3354 be dropped soon. Check the FAQ for more info about how to use it.
3355 * Greatly improved our cdash/ctest interactions
3356 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
3357 * Added memory checking tests with valgrind; lot of memleak fixing.
3358 This may be the first release of SimGrid with so few memory issues
3359 * Added code coverage tests.
3360 Our coverage is still improvable, but at least we see it on cdash.
3362 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
3364 SimGrid (3.3.4) stable; urgency=low
3366 The "Desktop Grid needs love too" release (also called Xmas release).
3368 Models improvements:
3369 * Major speedup in the maxmin system solving by using lazy evaluation
3370 Instead of solving completely the maxmin system at each iteration,
3371 only invalidate (and recompute) the modified parts.
3372 This new feature is enabled in default models but you can try to
3373 turn it on with "--cfg:maxmin-selective-update=1" for other models.
3374 * Cas01 IMproved as default CPU model
3375 This CPU model is the same Cas01 model, but it uses the
3376 maxmin-selective-update flag and a heap structure to manage
3377 actions on SURF kernel.
3378 It reduces the complexity to find the next action to finish and,
3379 consequently, it's faster than the old Cas01.
3380 This is the new default CPU model (Cas01).
3381 * Rename the old Cas01 model to Cas01_fullupdate
3382 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
3383 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
3384 * CpuTI (CPU Trace Integration)
3385 A new CPU model whose objective is simulate faster when using
3386 availability trace files.
3387 Instead of using a full featured, over engineered maxmin system for
3388 CPU modeling, this model does the pre-integration of traces files
3389 to calculate the amount of CPU power available, and so, executes
3390 faster than the old CPU models.
3391 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
3392 * Use LV08 as default network model since it gives better accuracy
3393 for small messages and shouldn't change things for big ones.
3394 Use --cfg=network_model:CM02 to get the previous behavior.
3397 ******************************************
3398 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
3399 ******************************************
3400 * The new CPU model may changes simulations!
3401 The point is that events occurring at the exact same timestamp
3402 are not scheduled in the same order with the old and new
3403 version. This may be enough to completely change the execution
3404 of simulations in some cases.
3405 * The new network model will change simulations!
3406 This new model is more realistic than the previous one, so you
3407 should consider redoing your old experiments with this model.
3408 Sorry for the inconvenience.
3411 * Introduce the supernovae compilation mode
3412 When compiled that way, the whole SimGrid (or almost) is put in a
3413 single compilation unit and compiled in one shoot.
3414 This is to help gcc which has difficulties to inline stuff from one
3416 The speedup seem to be above 15%, althrough more tests are needed on
3417 amd64 to confirm that gain.
3420 * Port of MSG's mailbox on top of SIMIX network
3421 The put/get mechanism was greatly simplified on the way.
3424 * New SIMIX network module. Provides:
3425 - Mailbox: rendez-vous mechanism to find with who you want to speak
3426 - Synchronous send/recv: easier and hopefully faster since the
3427 logic is handled in the maestro process directly now
3428 - Asynchronous send/recv: you dreamt of it? It's here now
3429 Too bad that nobody cared enough to propagate the change to MSG.
3430 * Add semaphores as SIMIX synchronization mechanism.
3433 * new function SD_daxload(char*) to load a DAX file
3434 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3435 * Introduce typed tasks. Specify its kind and cost at creation.
3436 At scheduling, just give where it should be placed, and the cost
3437 for each involved resource is automatically computed.
3438 Existing constructors so far (more to come of course):
3439 - SD_task_create_comm_e2e() for end-to-end communication
3440 - SD_task_create_comp_seq() for sequential computation
3441 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3442 * new function SD_task_dump() for debugging display
3443 * new function SD_task_dotty(task,FILE*) writing to file the info
3444 about the task in dotty format
3445 * SD_task_dependency_exists() can now cope with having one of its
3446 arguments NULL. If so, it tests whether the other argument has any
3448 * Add getters on list of preceding/following tasks:
3449 SD_task_get_parents(task) and SD_task_get_children(task)
3450 * Add getters on amount of workstations and list:
3451 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3452 * Add getter on task kind: SD_task_get_kind(task)
3453 * Update the start_time and finish_time of tasks on completion/failure
3454 * Bugfix: Remove task from state swags when destroyed
3457 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3460 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3461 dynar with one separate thread per value of the dynar.
3462 * Change the prototype of xbt_thread_create(), sorry.
3463 Added a boolean parameter indicating whether we want to join this
3464 thread (used in SG only for now)
3465 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3468 * GTNetS wrappers should now be usable again (and betterly tested too)
3469 * Fix a major regression from 3.2 where the timeout provided to
3470 MSG_task_put_with_timeout() was used as absolute time before which
3471 the comm should be done.
3472 * Start to fix the <cluster> tag.
3473 - Internal links should be good now (beside of the loopback, which
3474 use the private link instead)
3475 - paths to the external world is still rather broken
3476 - the <route:multi> tag is just broken. Actually that's brain-dead.
3477 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3478 to make it less stupid
3479 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3480 * Fix a source-level compatibility glitch from 3.2: after defining
3481 MSG_USE_DEPRECATED, you can use the old name
3482 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3483 * Allow to compile from the SVN with automake 1.11
3484 * Fix some problems when using the "start_time" tag in deployment XMLs.
3485 * Fix #8569: XBT/synchro.h has redundant declarations
3486 * Fix #8563: MSG return values and exceptions
3487 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3488 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3489 Thanks to Robson Peixoto.
3490 * Fix a vicious bug in dictionaries inducing that some elements were
3491 not freed on xbt_dict_free()
3493 Portability report of this version:
3494 * Main portability targets:
3495 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3496 on (amd64/i386/ia64)
3497 - mac leopard on i386
3498 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3499 but nothing critical.
3500 * Other platforms: windows, AIX and others were not tested for this release
3502 Timing report of this version:
3503 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3504 scenario parameters). From 8h to a few seconds in desktop grid settings.
3505 * Supernovae brings about 25% speedup on i386.
3507 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3509 SimGrid (3.3.3) stable; urgency=low
3511 The "Need for Speed" release.
3513 The timings done to validate the 3.3.2 were faulty.
3514 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3516 The problem was a conversion from a manually handled vector to
3517 xbt_dynar_t on the critical path.
3518 xbt_dynar_foreach calls functions, inducing stack management crap.
3520 We inlined these functions and xbt_dynar_foreach is now breath taking.
3521 We also inlined xbt_swag_belong on the way.
3523 Here are some approximate speedup measurements (on master/slaves
3524 simulations lasting between 10s and 20s each):
3525 3.3.1 -> 3.3.2: about same performance
3526 3.3.2 -> 3.3.3: 40% speedup
3527 3.3.1 -> 3.3.3: 40% speedup
3528 3.3.1 with inline patch -> 3.3.3: 30% speedup
3530 Our reading is that the refactoring which occurred in 3.3.2 made us
3531 suffer much more from the xbt_dynar_foreach low performance, but
3532 once we solved this, this refactoring proved to be very performance
3533 effective. From the 40% speedup, somehow, 10% are due to the
3534 inlining and 30% to the refactoring.
3536 That's a pitty that gcc cannot inline functions placed in other files
3537 alone. We have to choose between:
3538 - break the encapsulation (by putting private data structures and
3539 accessors in headers files to help gcc)
3540 - live with low performance
3541 - switch to a decent compiler such as icc (not quite possible).
3543 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3545 SimGrid (3.3.2) stable; urgency=low
3547 The "Simplicity does not preceed complexity, but follows it" release.
3549 The main contributors of this release were (lexical order):
3550 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3553 * Extract the routing logic into its own object.
3554 (was dupplicated in network.c and workstation_LV07.c;
3555 Allows to implement other ways of storing that info)
3556 => kill now useless network_card concept
3557 - Use dynar to represent routes (instead of void** + int*)
3558 - kill link_set (use surf_network_model->resource_set instead)
3559 - Add a command-line option to choose the routing schema to use
3560 - Add three new models:
3561 * Floyd (shortest path computed at initialization)
3562 * Dijikstra (shortest path recomputed all the time)
3563 * Cached Dijikstra (shortest path computed on need)
3564 All these models where contributed by Silas De Munck, and are
3565 described in his ICCS09 paper.
3567 * Simplify model declaration
3568 (less redirections, less function to write when defining a model)
3569 - Factorize stuff between models:
3572 surf_model_resource_set(model)
3573 surf_model_resource_by_name(model, name)
3574 - Unify the types of models in s_surf_model_t (using an union)
3575 - Embeed fields of common_public directly into s_surf_model_t
3576 - Rename model methods:
3577 action_free ~> action_unref
3578 action_change_state ~> action_state_set
3579 action_get_state ~> action_state_get
3580 - Change model methods into functions:
3581 (model)->common_public->action_use ~> surf_action_ref
3583 * Implement a generic resource; use it as ancestor to specific ones
3584 (allows to kill duplicated code in models)
3585 Drawback: timer command don't need no name nor properties;
3586 workstation_CLM03 don't need no properties
3587 (but I guess we can live with those few bytes wasted)
3589 * Improve the action object model
3590 - implement a constructor avoiding dupplicated code about field
3591 initialization in generic_action part.
3593 * Kill the SDP model: it has an external dependency, is deprecated
3594 in flavor of modern lmm models, and didn't compile since a while
3597 * Relocation of the context module from XBT to SIMIX.
3598 (the context were decoupled from the simix processes, duplicating a lot of code)
3599 => a lot of code was factorized
3600 - less overhead is introduced during scheduling
3601 - simpler API for the context factory
3602 - the logic for process creation,destruction and manipulation was simplified
3603 * Simplification of the s_smx_process_t data structure.
3604 => accesing the simix level data associated to a process is faster now,
3605 and the code is a lot more readable.
3608 * Implement some more MPI primitives:
3609 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3610 -implementation: Bcast: flat or 2-ary tree (default),
3611 Barrier: 4-ary tree,
3613 Allreduce: Reduce then Bcast
3614 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3615 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3616 Alltoallv: flat tree, like ompi
3618 * Add support for optimized collectives (Bcast is now binomial by default)
3619 * Port smpirun and smpicc to OS X
3622 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3623 Shout out if you used it.
3626 * Display the list of still queued messages in SG mode when existing
3630 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3631 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3632 * Add xbt_str_from_file(FILE*)
3633 * Add xbt_dict_get_key achieving a linear reverse search
3634 * Remove the context module
3636 Portability report of this version:
3637 * Main portability targets:
3638 - Linux(debian)/x86/context
3639 - Linux(debian)/x86/pthreads
3640 - Linux(debian)/amd64/context
3641 - Linux(debian)/amd64/pthreads
3642 On these, we still have the eratic breakages of gras/pmm and
3643 amok/saturate_sg reported in previous version. We still think
3644 that the tests are the cause of the fault, not the tested code.
3646 - Mac OSX Leopard/x86/context
3647 Still false negative in tesh autotesting.
3648 Smpi still fails, but this time because readlink does not accept -f
3649 Everything seems to work properly beside of that.
3652 - AIX version 5.3 (only tested contexts this time)
3653 Smpi still fails there because mktemp is not installed.
3654 Everything seems to work properly beside of that.
3656 I managed to compile it for the first time, but several breakages.
3657 Won't delay the release for this exotic platform.
3659 * Windows: it's still lagging behind. If you want to help, please
3662 Timing report of this version:
3663 This version seem to be more than 5% faster than 3.3.1 (on linux
3664 64bits with contexts). The gain is less than expected, we are
3665 investigating this for next release.
3667 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3669 SimGrid (3.3.1) stable; urgency=low
3672 * Implement a --cfg-help to show existing configuration variables
3673 * Build chain do not require doxygen in maintainer mode
3676 * fix a bug on struct sizeof computation, which prevented the
3677 exchange of arrays of structs in some conditions
3678 - added a regression test about this in datadesc_usage
3679 * Allow the exchange of 0-long dynamic vectors.
3680 - for that, use -1 as indicator of dynamic size instead of 0
3681 - This implied to change any size from unsigned long to long,
3682 reducing a bit communication abilities, but I guess that with
3683 64bits being quite common, this is more than enough.
3684 - This also induce a protocol change, thus bumping network protocol
3685 version from 0 to 1 (if we have external users, we have to get
3686 clean on that point too ;)
3687 - added two regression tests about this in datadesc_usage
3688 * Be more verbose when propagating local exceptions
3689 This helps debugging.
3690 * Display the status of simulated processes when receiving SIGINT in
3694 * Allow to control the simulation from a trace file.
3695 New functions MSG_action_register() and MSG_action_trace_run()
3696 The first one allows to associate a function execution to each
3697 kind of action while the second one parses a trace file and
3698 triggers the corresponding actions within the system.
3699 For now, only a toy example is provided in examples/msg/actions
3700 * Add an example of process migration in examples/msg/migration
3701 * Fix a bug in task exchange which broke MSG_task_get_sender()
3702 Add a teshsuite regression test for that.
3703 [Bug: if MSG_task_get_sender() is called after sender exit,
3705 * Fix a bug which prevented suspend/resume to work properly
3706 * Display the status of simulated processes when receiving SIGINT
3707 This fixes a regression of v3.3. due to the introduction of SIMIX
3708 * Bug fixing in failure management:
3709 - trace could not start by a failure at time 0
3710 - failure during communications were not working
3713 * Add SIMIX_process_set_name() to change the name of the current
3714 process in the log messages.
3715 * Store smx_hosts in a dict since we only retrieve them by name
3716 * Move the configuration infrastructure to surf
3719 * Move the configuration infrastructure to surf
3722 * Massive internal cleanups:
3723 - Store internal structures on processes instead of hosts (allows
3724 to have more than one process per host, in addition of being more
3726 - Cleanup the initialization/finalization process
3727 - Kill a whole bunch of unneeded synchronization:
3728 processes run in exclusive manner within the simulator
3729 - Move queues from global tables to process data fields
3731 - now accept -platform and -hostfile arguments
3732 - Pass the right rank value to processes according to the hostfile
3733 * Compile the examples by default, and use them as regression tests
3734 * Implement MPI_Wtime()
3735 * Change the reference speed to a command line option
3738 * TCP_gamma can now be specified as command line option using
3739 --cfg=TCP_gamma:10000000.0
3740 * Change the --surf-path cmd line option into --cfg=path:
3743 * Also include strbuff from xbt.h public header
3744 * xbt_ex_display(): do not free the exception after displaying
3745 This allows to do more with the given exception afterward.
3746 Users should call xbt_ex_free() themselves.
3750 Portability report of this version:
3751 * Main portability targets:
3752 - Linux(debian)/x86/context
3753 - Linux(debian)/x86/pthreads
3754 - Linux(debian)/amd64/context
3755 - Linux(debian)/amd64/pthreads
3756 These targets fail about 1/10 of times on gras/pmm, but we believe
3757 that this is because of the test, not because of SimGrid.
3758 amok/saturate_sg fails even more rarely, and the test may not be
3761 - Mac OSX Leopard/x86/context
3762 The test suite still spits tons of errors because some obscure
3763 force prevents us from removing the temporary directories
3764 arguing that they still contain some metadata I've never heard of.
3765 Smpi fails because seq is not installed.
3766 Everything seems to work properly beside of that.
3769 - AIX version 5.3 (both contexts and pthread)
3770 Smpi still fails there because mktemp is not installed.
3771 XML inclusions seems rosty on AIX.
3773 * Windows: it's still lagging behind. If you want to help, please
3776 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3778 SimGrid (3.3) stable; urgency=high
3782 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3783 [Malek Cherier & Mt]
3785 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3788 Until now, GRAS were using MSG as an interface to SURF. It was
3789 quite difficult because both interface have several differences
3790 (MSG channels vs GRAS sockets were the most notable point).
3792 This also opens the gate to SMPI (which should occur soon) and speed
3793 up simulations by to 40% (even if it were not the main goal).
3795 **************************************
3796 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3797 **************************************
3798 The point is that events occurring at the exact same timestamp are
3799 not scheduled in the same order with the old and new version. This
3800 may be enough to completely change the execution of simulations in
3801 some cases. Sorry for the inconvenience.
3803 * Cleanup and upgrade the XML format to push further scalability
3804 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3806 * Improve the testing infrastructure with tesh. Now a very large part of
3807 the code is tested not only by being run but also by checking that the
3808 output match an expected output [Mt].
3810 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3811 is really less memory-demanding, which should allow you to use
3812 larger files in SimGrid [AL].
3814 * Inform valgrind about our contexts, so that it becomes usable
3815 with the default (and more effecient) version of SimGrid
3816 [contributed by Sékou Diakite, many thanks]
3819 * Introduce a listener thread in charge of receiving incoming
3820 messages from the network. It allows to overlap communication and
3821 computation but most notably, it removes some stupid deadlocks due
3822 to the fact that so far, a process could not send and receive at
3823 the same time. This made most non trivial communication schema
3825 * Convert the PIDs from long int to int to match the MSG ones (and
3826 linux ones too) [Mt]
3827 * New function: gras_agent_spawn() to launch a new process on
3828 current host. Only working in simulation for now. [Mt]
3829 * New function: gras_os_hostport() returning a constant form (ie,
3830 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3833 * Make the backtrace of exceptions more human readable [Mt]
3834 * New module: xbt/str [Mt]
3835 a ton of string utility functions (split, join, printf to a newly
3836 allocated buffer, trim, etc)
3837 * New module: xbt/hash [Mt]
3838 SHA1 hashing algorithm (more to come if needed)
3839 * New module: xbt/synchro [Mt]
3840 synchronization tools (mutex and conditions) working the same way
3841 in simulation and in real life (mainly useful for GRAS, but not
3843 * New module: xbt/queue [Mt]
3844 classical producer/consumer synchronization scheme
3845 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3846 (using the classical functions will get serialized) [Mt]
3847 * Make dictionary internal table dynamic. No need to specify its size
3848 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3850 * Make sure the log channels are organized as a tree under windows
3851 (because of ANSI C compatibility issue, any channel were child of
3852 root directly) [Mt].
3855 * Cleaned many thing in surf and fixed a few bugs [AL].
3856 * Add a nice command line configuration mechanism to compose models [AL].
3857 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3858 the previous one (KCCFLN05). It relies on something that looks like
3859 a max-min sharing mechanism but cannot be written as such. A new solver
3860 was thus designed [AL].
3861 * Add a new solver to lmm. Based on Lagrange optimization and
3862 gradient-based descent, it enables to efficiently maximise systems s.a
3864 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3866 This solver enables to propose two new network models for TCP Reno and
3867 TCP Vegas based on Low's work. These models still need to be fully
3868 tested though [Pedro Velho].
3871 * Bug fix in SD_simulate. Now the time bound given as argument is
3873 * Use the new parallel task model (ptask_L07) as default.
3874 * Use the SURF command line configuration mechanism.
3875 * 0-size tasks (for synchronization) should now work.
3877 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3879 SimGrid (3.2) stable; urgency=high
3883 We still experience issues on this platform, but we believe that at
3884 least MSG is usable.
3886 GRAS API BREAKAGE (for simplification purpose, sorry):
3887 * the gras_msgtype_by_name is not used anymore. Instead of
3888 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3889 you can write (and must)
3890 gras_msg_send(toserver, "request", &request);
3891 - If you still want to pass a gras_msgtype_t to the function (to cache
3892 the type and avoid the lookup time), use the gras_msg_send_() variant.
3893 - Impacted functions:
3894 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3895 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3896 * The callbacks are now expected to return 0 when everything went well
3897 (just like the main() function)
3899 GRAS new features and improvements:
3900 * New module mechanism where user code can use per process globals [Mt]
3901 This is similar to gras_userdata_*() functions, but for libraries. It
3902 factorize some code developped over and over in the examples and AMOK.
3903 It has still to be documented and used (only amok/peermanagement is
3905 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3906 loss when they were small enough to fit into the buffer and sent quickly
3907 enough so that they can all get received in one shoot.
3908 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3909 if not found. Use the *_or_null() variant for the old semantic.
3910 * In gras_msg_handle, do not discard messages without callback.
3911 They are probably messages to be explicitly awaited later (ie, proofs of
3912 mis-synchronization in userland since they are sent before being awaited)
3913 No big deal usually.
3914 * gras_socket_meas_send/recv: semantic changed!
3915 The numerical arguments used to be (1) the total amount of data to send
3916 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3917 messages. This was need for the fool willing to send more than MAXINT
3918 bytes on quite fat pipes.
3921 * Do really rename the hostmanagement module to peermanagement. [Mt]
3922 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3923 but this is rather new and this was documented in the module
3924 documentation (poor excuses, I admit)
3925 * Bandwidth measurement semantic changed! This follows the changes to
3926 gras_socket_meas_send/recv explained above.
3929 * A sequential mode has been added to the workstations. When a workstation
3930 is in sequential mode, it can execute only one task, and the other tasks
3931 are waiting in a FIFO. [Christophe Thiery]
3934 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3935 model for SIMDAG. [Christophe Thiery]
3936 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3937 equal to 0, which caused some bad behaviors in
3938 saturated_constraint_set_update. I now use a threshold mechanism like in
3942 * When running manually src/testall, you select specific units [Mt]
3943 testall is the result of our cunit mechanism, and should replace all
3944 the scripty thingy around since bash don't run easily on billware.
3946 * A mallocator system has been added. [Christophe Thiery]
3947 Mallocators allow you to recycle your unused objects instead of freeing them
3948 and allocating new ones.
3950 Documentation update:
3951 * FAQ reworking + New FAQs:
3952 - "Valgrind spits tons of errors!" [Mt]
3953 - "How to repport bugs" [Mt]
3954 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3955 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3956 - Communication time measurement within MSG [AL]
3957 - I experience weird communication times when I change the latency [AL]
3958 * GRAS tutorial [Mt]
3960 - an introduction to the framework and to the used communication model
3961 - an initiatic tour introducing the most proheminent features:
3963 . Lesson 0: Installing GRAS
3964 . Lesson 1: Setting up your own project
3965 o Part 2: Message passing
3966 . Lesson 2: Exchanging simple messages
3967 . Lesson 3: Passing arguments to the processes (in SG)
3968 . Lesson 4: Attaching callbacks to messages
3969 . Lesson 5: Using globals in processes
3970 . Lesson 6: Logging information properly
3971 . Lesson 7: Using internal timers
3972 . Lesson 8: Handling errors through exceptions
3973 . Lesson 9: Exchanging simple data
3974 . Lesson 10: Remote Procedure Calling (RPC)
3975 . Lesson 11: Explicitly waiting for messages
3976 . Recapping of message passing features in GRAS
3977 - A HOWTO section containing:
3978 o HOWTO design a GRAS application
3979 More are due, of course. They will come latter. In the meanwhile, you can
3980 check the examples which are still here.
3982 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3984 SimGrid (3.1) stable; urgency=high
3988 There was a stack corruption somewhere, visible only when optimizing
3989 with these versions. [Vince]
3992 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3993 module that enabled to play with Directed Acyclic Graphs. It is built
3994 directly on top of SURF and provides an API rather close to the old
3995 SG. Some old codes using SG are currently under rewrite to check that
3996 all needful functions are provided. [Christophe Thiery]
3999 * Complete rewrite of the KCCFLN05 workstation model. It is now an
4000 extension of the classical CLM03 model that gracefully handles
4001 failures. This is now the default model for MSG and GRAS. It doesn't
4002 handle parallel tasks yet though. [AL]
4003 * Bug fix: Weights were not correctly set in the network part.
4004 WARNING: This may have resulted in incorrect results with simulations
4005 where there are more than one flow on a given link. [AL]
4008 * After a (long ?) discussion on simgrid-devel, we have decided that the
4009 convention we had on units was stupid. That is why it has been decided
4010 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
4011 WARNING: This means that all previous platform files will not work as
4012 such with this version! A warning is issued to ask users to update
4014 A conversion script can be found in the contrib module of the CVS, under
4015 the name contrib/platform_generation/surfxml_update.pl [MQ]
4018 * Bug fix: Processes were started in reverse order, wrt deployment file.
4019 WARNING: if your code relies on this bug, please fix it. [AL]
4020 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
4021 being executed on two different locations. [AL]
4022 * Bug fix: Failures are now better supported thanks to Derrick's tests
4023 (there was many failure situations I hadn't thought of and that weren't
4024 correctly handled). [AL]
4025 * New function: MSG_host_is_avail indicates you whether a given m_host_t
4029 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
4030 Exception occurring on server-side are propagated back to client (!).
4032 API CHANGE: the callback changed their prototype. Change:
4033 int my_handler(gras_socket_t expeditor, void *payload_data) {
4035 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
4036 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
4038 * New! function: gras_msg_handleall to deal with all messages arriving
4039 within a given period.
4040 * New! function: gras_socket_server_range to get a server socket in a
4041 range of port numbers (ease to avoid port number conflicts) [MQ]
4042 * New! gras processes display their backtrace when they get a SIGUSR1
4043 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
4044 Sweet to debug RL processes [MQ]
4048 - Do not force experiment sizes to be expressed in kb, or it becomes
4049 impossible to measure the latency this way (needs one byte-long tests)
4050 WARNING: this changes the amok_bw_* function semantic. [MQ]
4051 - Implements the link saturation stuff. [MQ]
4052 * Peer management module:
4053 New! module factorizing code that we wrote over and over [MQ].
4056 * New module: cunit (my jUnit implementation in ansi C) [MQ]
4057 - Test units are placed directly into the library code, they get extracted
4058 automatically and placed into the src/testall binary.
4059 - Convert most of the XBT tests to this system.
4060 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
4061 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
4062 live in the same C code. This required to change a little bit the API
4063 of surfxml parsing but shouldn't be an issue for end-users. [AL]
4064 * New module: sparse graph structure with basic algorithms (this is work
4065 in progress and the API is not considered to be frozen yet). [AL]
4066 * Display more information on backtraces: source line & function names are
4067 now displayed just like valgrind does (rely on addr2line tool) [MQ]
4068 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
4069 * Reworked a little bit some #include statements to load only required
4070 headers. Some user code that relied on SimGrid to include stdlib or
4071 stdio may need to include it by themselves. [AL]
4072 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
4074 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
4075 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
4076 but I guess that almost nobody used those parts. [MQ]
4078 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
4080 SimGrid (3.0.1) stable; urgency=low
4083 * Unfortunately, I had missed 5 misnamed functions:
4084 xbt_fifo_item_t xbt_fifo_newitem(void);
4085 void xbt_fifo_freeitem(xbt_fifo_item_t);
4086 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
4087 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
4088 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
4089 They're now deprecated. Please use their new versions:
4090 xbt_fifo_item_t xbt_fifo_new_item(void);
4091 void xbt_fifo_free_item(xbt_fifo_item_t);
4092 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
4093 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
4094 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
4096 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
4097 * Documentation: xbt_log module unmercifully reworked [MQ]
4098 * Bugfix: there was a problem with the ending of contexts with
4099 the pthread backend. It caused some weird deadlock or behavior
4100 depending on the pthread implementation. [AL]
4101 * Bugfix: get the exceptions raised in the simulator repport where
4102 and why they come from when they are not caught in time [AL, MQ]
4105 * Bugfix: Do repport the error when two non-connected hosts try to
4106 exchange data (Thanks to Flavien for stumbling into this one) [AL]
4109 * Add additionnal checkings on communications. Assert that two
4110 communicating hosts are connected by a set of links... [AL]
4113 * Add additionnal checkings on channel values in communication [AL]
4114 * New: MSG_task_get_source to see on which host a task was generated [HC]
4115 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
4116 the number of tasks waiting to be received on channel and sent
4118 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
4119 waits for the first task coming from a given host.. [AL]
4121 GRAS new functionnalities: [MQ]
4122 * Enhance the parsing macro to allow the size of multidimentional objects
4123 to be given thru annotations.
4124 * New example (and documentation): Matrix Multiplication a la RPC
4125 (as when I was young!) and fix a bunch of bugs found on the way.
4127 GRAS performance improvements: [MQ]
4129 * Reduce the amount of cbps creation/destruction by making it static to
4130 datadesc_send/recv() and using a (newly created) cbps_reset (based on
4133 * Change libdata to a set so that we can search for stuff by ID (and thus
4134 reduce the insane amount of dict lookups)
4137 * Actually implement gras_datadesc_copy() so that we don't have to mimick
4138 RL communication on top of SG since it's so uneffective.
4139 It may also be used for inter-thread communication in RL, one day.
4140 * Use gras_datadesc_copy() to exchange messages on top of SG
4142 - improve message exchange performance on top of SG
4143 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
4144 * Don't exchange on the network the size of the used part of buffer,
4145 instead, specify the possible buffer size to read().
4147 - reduces the amount of read/write calls (one pair per exchange)
4148 - reduces the amount of exchanged data (the size)
4149 - allows to retrieve all arrived data on receiver side, if we don't need
4150 it right now (subsequent read will peek the buffer)
4151 - allows the receiver to proceed with the beginning of the stream before
4152 everything is arrived
4153 - make it possible to build an iov transport (using readv/writev)
4155 - take care of the data with non-stable storage (like stacked data),
4157 * If possible, TCP send uses vector I/O (when writev() is here)
4158 - Don't use it for receive since we send data sizes and data on the
4159 same stream, so we wouldn't be able to chain large amount of chunks
4160 before having to flush the stuff to read the size.
4161 * Rework the transport plugin mechanism to simplify it and reduce the
4162 amount of pointer dereferencement when searching for the right function
4165 * I guess that now, we do almost as few system calls as possible while
4166 doing as few data copy as possible.
4168 To improve it further, we could try to send all the sizes first and then
4169 all the data (to use iov on receiving size), but it's only a partial
4170 solution: when you have 2 dimensional data, the sizes of the second
4171 dimension is data of the first dimension, so you need 3 streams.
4173 I'm not sure the potential performance gains justify the coding burden.
4175 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
4177 SimGrid (3.00) stable; urgency=high
4180 * New! Give the possibility to hijack the surf parser and thus bypass
4181 MSG_create_environment and MSG_launch_application. Have a look at
4182 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
4184 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
4186 SimGrid (2.96) unstable; urgency=low
4191 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
4192 This deprecates the xbt_error_t mechanisms.
4193 It modifies (simplifies) all XBT and GRAS API.
4194 MSG API keeps unchanged (exceptions raised by XBT are caught from
4195 within MSG and masked with existing error handling facilities)
4198 * New! Add a FATPIPE model. [AL]
4199 * New! Add a parallel task model. [AL]
4200 * New! Add automatically a loopback interface (in the default
4201 network model) if none was precised.
4204 * Bugfix: MSG_process_resume now works with the current running process.
4206 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
4207 * Modification of MSG_task_get_compute_duration. Once a task has been
4208 processed, the value returned by this function is now equal to 0. [AL]
4209 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
4210 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
4211 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
4212 of a m_task. For now, MSG_task_cancel only works with computation
4214 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
4215 of the processor (in Mflop/s) regardless of the current load on the
4217 * API Change: use proper naming convention for MSG_getClock and
4218 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
4220 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
4221 This function changes the priority of a computation task. This priority
4222 doesn't affect the transfer rate. A priority of 2 will make a task
4223 receive two times more cpu power than the other ones. This function
4224 has been added to suit the needs of Nguyen The Loc and hasn't been that
4225 much tested yet. So if it fails, please report it and send me your code.
4227 * API Change: removed all functions and types that were marked "deprecated"
4228 since many months. Renamed MSG_global_init_args to MSG_global_init.
4230 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
4232 SimGrid (2.95) unstable; urgency=low
4235 * Steal some nice code to GNU pth to fix context detection and usage [AL]
4236 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
4237 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
4240 * Bugfix: Allow absolute paths for platform description files [MQ]
4241 * Bugfix: do free the variables after use. Leads to drastic performance
4243 * Implement max_duration (ie, timeouts) on resources [AL]
4246 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
4248 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
4249 the needed support (timeouts) [AL]
4252 * Implement measurement sockets. You can now get the bandwidth between two
4253 hosts thanks to AMOK (see below). [MQ]
4254 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
4255 dynar over the network (yeah) [MQ]
4256 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
4257 * simple/static token-ring example. [Alexandre Colucci and MQ]
4258 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
4259 in place before [MQ]
4262 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
4263 type parsing. "short" and "long" modifiers were also ignored; other
4264 modifier (such as reference level) are still ignored. [MQ]
4265 * Embeed the buffer size within the buffer itself on SG. [MQ]
4266 That way, send() are atomic and cannot get intermixed anymore (at least
4267 the ones which are less than 100k; bigger messages still have the issue)
4268 * Array size pushed by the field, not by the field type (or each
4269 and every long int will push stuff to the cbps) [MQ]
4270 * use select() to sleep since it allows to portably sleep less than one
4273 GRAS (minor cleanups)
4274 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
4277 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
4278 * type category 'ignored' killed as it was never used and were difficult
4280 * whether a type can cycle or not is now a flag, leaving room for more
4281 flags (as "ignored", if we feel the need one day ;)
4282 * Rename raw sockets to measurement sockets since "raw" has another
4283 meaning in networking community.
4286 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
4287 and offers features not belonging to GRAS but that most applications
4288 need. One day, it may be a set of plugins loadable at runtime.
4289 * New module: bandwidth
4290 bandwidth measurement between arbitrary nodes running this module. [MQ]
4292 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
4294 SimGrid (2.94) unstable; urgency=low
4296 The first beta release of SimGrid 3 !
4300 * Update the main page and the FAQ. Adding references to gforge.
4303 * Add a gras_os_getpid function.
4306 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
4307 * Extend the logs so that they also print PID, hostname, date, ... if
4309 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
4310 and kill the old version which were in testsuite/
4311 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
4315 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
4316 for users. Instead, each programming environment (GRAS, MSG,...) use it
4317 when needed to provide such a feature to users.
4318 Don't shortcut the mechanism or you will also shortcut the virtualization
4319 you need on the simulator.
4323 * Cleanups in configury with regard to compile optimization/warning flags.
4324 Also add -fno-loop-optimize to any powerpc since it's the optimization
4325 killing gcc (< 3.4.0).
4326 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
4327 needed by MSG examples complications
4328 * Borrow configury beautifications from PHP
4331 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
4332 warning (thanks loris for stumbling into this one).
4333 * Bugfix: stop loading private headers (gras_config.h) from the public
4337 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
4338 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
4339 RL would imply a DNS resolver, which is *hard* to do in a portable way
4340 (and therefore delayed).
4341 * Implement a real timer mechanism and use it in timing macros. This allows
4342 to avoid rounding errors and get a 0.000005 sec precision in timing
4343 macros. While I was at it, various cleanups:
4344 - allow to declare more than one timed section per file (fix a stupid bug)
4345 - move some private declaration to the right place
4346 - merge conditional execution and timing macros into emulation module
4347 - document the module
4348 - make sure the module cleanups its mess on gras_exit
4349 * Documentation improvements:
4350 - (new) how to compile applications using GRAS
4351 - (new) emulation support (timing macros)
4353 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
4355 SimGrid (2.93) unstable; urgency=low
4357 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
4360 - Use Paje properly where used. Still to be sanitized properly.
4361 - Portability fix: Add an implementation of the contexts using pthread
4365 - Add xbt_procname(): returns the name of the current process.
4366 Use it to show the current process's name in all logging.
4368 - fix detection of older flex version and the reaction, since we do
4369 depend on modern ones (we use lex_destroy)
4370 - Better separation of SG and RL in the libs: remove all simulation code
4371 from libgras. As a result, this lib is now only 200k when stripped.
4372 Some of the xbt modules may also be duplicated (two sets and such) and
4373 should be cleaned/killed before SG3.
4374 - Insist on using xlC on AIX because of weird problems involving gcc there.
4375 - Cleanup the make remote stuff. This is now done by scripts
4376 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
4377 mainly for our private use, but we're working on changing them to user
4380 - Bugfix: flush the socket on close only if there is some *output*.
4381 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
4383 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
4386 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
4388 SimGrid (2.92) unstable; urgency=low
4390 Alpha 3 on the path to SimGrid 3
4394 - New! First try of benchmarking macros.
4395 - New! First try so that gras_stub_generator generate deployment and
4396 remote compilation helpers.
4398 - Bugfix: Initialization fix in msg_test.
4402 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
4404 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
4406 - API Change: gras_os_sleep to take the amount of seconds as a double.
4407 Accepting an int was error prone since it was the only location where
4408 seconds were coded as such. It leaded to damn rounding errors.
4409 - Bugfix: Hard to belive that timers ever worked before this.
4411 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4413 SimGrid (2.91) unstable; urgency=low
4415 Alpha 2 on the path to SimGrid 3
4419 - Bug fix in the lmm_solver.
4421 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4422 through the function MSG_paje_output.
4423 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4424 - It is possible to bound the rate of a communication in MSG with
4425 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4426 to write it in the changelog).
4427 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4432 - Various cleanups to the autotools stuff
4433 - Begin to move Gras examples to examples/gras/
4434 - Let make distcheck work again (yeah!)
4436 - documentation overhauled using doxygen.
4437 gtk-doc-tools is dead in SimGrid now.
4438 - Automatically extract all existing logging categories, and add the list
4439 to the documentation (long standing one, to say the less)
4441 - Cleanup the known architecture table. Reorder the entries to group what
4442 should be, and use a more consistent naming scheme.
4443 (some of the test dataset are still to be regenerated)
4444 - New! Allow library to register globals on each process just as userdata
4446 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4447 lookup time (for now).
4448 Use it in msg and trp.
4449 This cleans a lot the internals and helps enforcing privacy of the
4450 headers between the gras components.
4451 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4452 - Bugfix: gras_os_time was delirious in RL.
4453 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4455 - Reenable GRAS now that it works.
4457 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4459 SimGrid (2.90) unstable; urgency=low
4461 Alpha 1 on the path to SimGrid 3
4463 * It is a long time since the last release of SimGrid. I'm sorry about
4464 that but as I had told you, I was rewriting a lot of things. I apologize
4465 to those who had been reporting bugs to me and that I had not answered.
4466 If your bug is still in the new version, please tell me. Here is a
4467 summary of the main changes.
4469 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4470 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4471 portability, speed, and a lot more but you'll figure it out later.
4472 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4473 I wanted to release the new MSG as soon as possible and I have
4474 broken GRAS, which is the reason why, for now, only MSG is fully
4475 functional. A laconic description of these projects is available
4476 in the documentation.
4478 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4479 kernel optimized for our needs (called SURF but only the developers
4480 should use it). Hence, MSG is now roughly 30 times faster and I think
4481 that by rewriting a little bit MSG, I could event speed it up a little
4482 bit more. Beside the gain in speed, it is also much easier to encode a
4483 new platform model with SURF than it was with SG. More to come...
4485 * REVOLUTION 3: I have tried to change a little as possible the API of
4486 MSG but a few things really had to disappear. The main differences
4487 with the previous version are:
4488 1) no more m_links_t and the corresponding functions. Platforms are
4489 directly read from an XML description and cannot be hard-coded
4490 anymore. The same format is used for application deployment
4491 description. The new format is described in the documentation.
4492 Have a look in tools/platform_generation. There is a tiny script
4493 that converts from the old platform format to the new one. Concerning
4494 the application deployment format, parsing the old one is tricky.
4495 I think most of you should however be able to convert your files. If
4496 it is really an issue, I can write a C code that does the conversion.
4498 2) the toolbox tbx does not exist anymore. We now have a library
4499 with much more data-structures but without the hash-tables (we have
4500 dictionaries that are much faster).
4502 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4504 *****************************************************************************
4505 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4506 * don't feel like converting the oldies. (Mt) *
4507 *****************************************************************************
4510 Version 2.90: "the long awaited one"
4511 - Finished rewriting and debugging MSG. Rewrote the documentation.
4512 - disable GRAS for now since it needs to be ported to the newest SG
4515 - Finish the port to windows (using mingw32 for cross-compile)
4518 - Main loop and datastructures of SURF. A cpu resource object is
4519 functional. Surf can thus be used to create cpu's with variable
4520 performance on which you can execute some actions.
4522 2004-11-15 Martin Quinson
4523 - Port to ARM. Simply added the alignment and size descriptions. Should
4524 work, but the ARM machines are so slow that I didn't had the opportunity
4525 to 'make check' over there yet.
4527 2004-11-15 Arnaud Legrand
4528 - Trace manager now written. It uses a heap structure and is therefore
4529 expected to be efficient. It may however be speeded up (particularly
4530 when many events occur at the same date) by using red and black
4531 trees. One day maybe...
4532 - Max-min linear system solver written. It uses a sparse matrix
4533 structure taking advantage of its expected use. Most operations are
4534 O(1) and free/calloc are called as few as possible. The computation of
4535 the minimum could however be improved by using a red and black tree
4538 2004-11-03 Arnaud Legrand
4539 - Rename every gras_* function that was in xbt/ to its xbt_
4541 - Add a heap and a doubly-linked list to xbt
4542 - Added a dichotomy to the dictionaries. make check works as well before
4543 so I assume that the patch is correct. I do not know however if things
4544 run effectively faster than before now. :)
4546 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4547 SimGrid, and the version number is bumped to 2.x
4549 2004-10-29 Martin Quinson
4550 - Introduction of the remote errors.
4551 They are the result of a RMI/RPC on the remote machine.
4552 ErrCodes being scalar values, you can't get the host on which those
4553 errors did happen. Extending the error mechanism as in Gnome is possible.
4554 No idea yet whether it is a good idea.
4556 2004-10-28 Martin Quinson
4557 - Interface revolution: the Starred Structure Eradication.
4558 I used to do typedef struct {} toto_t; and then handle *toto_t.
4559 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4560 - ???_t is a valid type (builded with typedef)
4561 - s_toto_t is a structure (access to fields with .)
4562 - s_toto is a structure needing 'struct' keyword to be used
4563 - e_toto_t is an enum
4564 - toto_t is an 'object' (struct*)
4566 typedef struct s_toto {} s_toto_t, *toto_t;
4567 typedef enum {} e_toto_t;
4568 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4569 s_toto_t) is private.
4571 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4572 so that it can set the variable to NULL. It was so for dicts and sets,
4573 it changed for dynars.
4575 - Fix a bunch of memleaks in dict_remove
4576 - Fix a bug in sg/server_socket opening: it failed all the time.
4578 2004-10-07 Martin Quinson
4579 - Speed up dynar lookup operation a bit.
4581 gras_dynar_get is dead.
4583 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4584 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4585 (which gives you the address of the stored data).
4587 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4588 copy of the data using an affectation to do the job and not a memcpy.
4590 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4591 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4593 It does not really speedup the dynar test because they are
4594 setting elements all the time (and look them seldom). But the dict does
4595 far more lookup than setting.
4597 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4599 2004-10-05 Martin Quinson
4600 - Allow to (en/dis)able the cycle detection at run time.
4602 Whether we should check for cycle or not is now a property of each
4603 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4604 datadesc_cycle_unset allow to remove this property when previously set.
4606 Note that the cycle detection is off by default since it impacts the
4607 performance. Watch the data you feed GRAS with ;)
4609 This property is hereditary. Any element embedded in a structure having it
4610 set have it set for the time of this data exchange.
4612 You should set it both on sender and receiver side. If you don't set it on
4613 sender side, it will enter an endless loop. If you forget on receiver
4614 side, the cycles won't be recreated after communication.
4616 - Header reorganization.
4617 Kill gras_private.h, each submodule must load the headers it needs.
4619 2004-10-04 Martin Quinson
4620 - Interface revolution: do not try to survive to malloc failure.
4622 Now, gras_malloc and friends call gras_abort() on failure.
4623 As a conclusion, malloc_error is not a valid error anymore, and all
4624 functions for which it was the only gras_error_t return value are
4625 changed. They now return void, or there result directly.
4626 This simplify the API a lot.
4628 2004-09-29 Martin Quinson
4629 - Re-enable raw sockets.
4630 Created by gras_socket_{client,server}_ext;
4631 Used with gras_raw_{send,recv}
4634 It should allow to kill the last bits of gras first version soon.
4636 This is not completely satisfactory yet (duplicate code with
4637 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4640 - Simplify transport plugin (internal) interface by not passing any
4641 argument to _server and _client, but embedding them in the socket
4644 2004-09-28 Martin Quinson
4645 - Finish the port to AIX.
4646 autoconf was my problem (segfault within the malloc replacement
4647 function. No idea why)
4649 2004-09-16 Martin Quinson
4650 - Fix some size_t madness on 64bit architectures.
4652 2004-09-08 Martin Quinson
4653 - Reduce the number of system headers loaded, overload some more system
4654 calls (such as malloc to cast the result of the system one, and work
4656 - Fix and reintroduce the config support
4658 2004-09-07 Martin Quinson
4659 - Source code reorganization to allow Arnaud to surf all over there.
4660 - Allow to document the logging categories.
4661 - Remove all uppercase from logging categories and useless cleanup in names.
4663 2004-08-18 Martin Quinson
4664 Version 0.6.2 (protocol not changed; API changed)
4665 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4666 gras_error_t), and NULL when not found. Functions expecting a msgtype
4667 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4668 hopefully usefull message.
4669 - Portability to prehistoric sparcs again
4671 2004-08-17 Martin Quinson
4672 Version 0.6.1 (protocol not changed; ABI not changed)
4673 - prealloc some buffers to speed things up
4675 2004-08-11 Martin Quinson
4676 Version 0.6 (protocol not changed; ABI expended)
4677 - The parsing macro can deal with the references, provided that you add
4678 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4680 2004-08-09 Martin Quinson
4681 Version 0.5 (protocol not changed; ABI changed)
4682 - Allow to off turn the cycle detection code in data exchange at
4683 compilation time. It should be at run time, but I'm short of time (and
4684 the config stuff is still broken). That way, we keep dict out of the
4685 critical path, which is good because the performance is poor:
4686 - search not dichotomial yet
4687 - dynar give no way to access their content and memcpy everytime
4688 - In composed data description (struct, ref and so on), stop foolness of
4689 keeping the subtype's ID, but store the type itself. This keeps sets out
4690 of the critical path, which is good since they rely on dynar and
4691 dictionaries. The only loose of that is that we cannot detect the
4692 redeclaration of a structure/union with another content (but I'm not sure
4693 the code detected well this error before anyway). We still can detect
4694 the redefinition discrepancy for the other types.
4695 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4696 it breaks the code because of type-punning used all over the place).
4697 This breaks on all non-gcc architectures (for now).
4699 All those changes (plus the buffer of last time) allow me to gain 2 order
4700 of magnitude on cruel tests consisting of 800000 array of integers on two
4701 level of a hierarchical structure (200 secondes -> 4 secondes)
4704 - the selector of reference must now return the type it points to, not
4705 the ID of this type.
4707 2004-08-06 Martin Quinson
4708 Version 0.4 (protocol changed; ABI not changed)
4709 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4710 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4711 - Implement a buffer transport, and use it by default (it relies on tcp in
4712 real life and on sg in simulation).
4713 That's a bit hackish since I had a new field to the structure to store
4714 its data without interfering with the subtype ones. Inheritance
4715 is tricky in C. And that's a kind of reverse inheritance with one class
4716 derivating two classes. Or maybe a game with java interfaces. Anyway,
4717 that's damn hard in C (at least).
4718 Moreover, I got tired while trying to ensure plugin separation and
4719 genericity in SG mode. MSG wants me to do weird things, so let's go for
4720 cruel hacks (temporarily of course ;).
4721 See comment in transport_private.h:71
4722 - do not include all the _interface headers in private but in the files
4723 which really need them (to cut the compilation time when they are
4726 2004-07-26 Martin Quinson
4727 Version 0.3 (protocol not changed; ABI changed)
4728 - Major overhault of the datadesc interface to simplify it:
4729 - shorted the function names:
4730 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4731 - add a trivial way to push/pop integers into the cbps without malloc.
4732 This allows to make really generic sub_type description, which simply
4733 pop their size of the stack.
4734 - add a function gras_datadesc_ref_pop_arr() which does what users want
4735 most of the time: Declare a dynamic array (which pops its size of the
4736 stack) and declare a reference to it. Poor name, but anyway.
4737 - kill the post-send callback, add a post-receive one
4739 2004-07-23 Martin Quinson
4740 Version 0.2 (protocol changed; ABI changed)
4741 - add some testing for cpbs in the test cases, and fix some more bugs.
4742 This invalidate again the little64 data file, since I cannot regenerate
4744 - remove an awfull optimization in the logging stuff, allowing me to:
4745 - understand it again
4746 - learn gcc how to check that the argument match the provided format
4747 - fix all errors revealed by gcc after that
4748 - internal keys of dict are not \0 terminated. Deal with it properly in
4749 loggings instead of segfaulting when the user want to see the logs :-/
4751 2004-07-22 Martin Quinson
4752 - Fix some stupid bug preventing cbps (callback postit) from working
4754 2004-07-21 Martin Quinson
4755 - Some documentation cleanups
4756 - remove the useless last argument of msgtype_declare
4757 - rename the Virtu functions to fit into the 'os' namespace
4758 - move headers src/include -> src/include/gras/ and stop fooling with
4759 gras -> . symbolic link
4760 - make distcheck is now successful
4762 2004-07-19 Martin Quinson
4764 - Build shared library also
4765 - Install html doc to the right location
4766 - stop removing maintainer files in make clean
4767 - build tests only on make check
4769 2004-07-13 Martin Quinson
4771 - No major issue in previous version => change versioning schema
4772 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4773 corresponding dataset.
4775 2004-07-11 Martin Quinson
4777 - Get it working with any kind of structure (we can compute the padding
4778 bytes remotely for all the architectures I have access to)
4779 - Implement the structure parsing macro (still not quite robust/complete)
4780 - Improvement to the remote testing toysuite
4782 2004-07-10 Martin Quinson
4783 [autoconf mechanism]
4784 - get ride of a bunch of deprecated macros
4785 - actually run the test for two-compliment, not only compile it :-/
4786 - test whether the structures get packed (and bail out if yes. Damn.
4787 Alignment is a serious matter)
4788 - test whether the structures get compacted (but respecting the alignment
4789 constraints of each types)
4790 - test whether the array fields of structures can straddle alignment boundaries
4792 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4793 (took me 2 hours to find that bug, looking at the wrong place)
4794 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4795 before sending/receiving and are used to compute the offsets of fields
4796 - Given that padding size depend even on compiler options, keep track of
4797 alignment and aligned_size only for the current architecture. Not a big
4798 deal since we send structure fields one after the other (seems
4800 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4801 program, let it work on linux/gcc/little32. portability todo.
4803 2004-07-08 Martin Quinson
4804 - import and improve remote compilation support from FAST
4805 - make sure make check works on half a dozen of machines out there
4807 2004-07-07 Martin Quinson
4808 Let's say it's version 0.0.3 ;)
4809 - Implement conversions (yuhu!)
4810 - Let it work on solaris (beside conversion, of course)
4811 - Stupid me, using rand() to generate the conversion datatests in not wise.
4813 2004-07-06 Martin Quinson
4814 - Let make dist work, since I'm gonna need it to compile on remote hosts
4815 - Let Tests/datadesc_usage write the architecture on which the file was
4816 generated as first byte.
4817 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4818 architecture descriptions.
4819 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4820 on those architectures.
4821 - Optimization: send/recv array of scalar in one shoot
4823 2004-07-05 Martin Quinson
4824 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4826 - Plug a whole bunch of memleaks
4827 - each process now have to call gras_{init,exit}. One day, their log
4828 settings will be separated
4829 - Continue the code factorisation between SG, RL and common in Transport.
4831 2004-07-04 Martin Quinson
4833 - Redistribution between SG and RL.
4834 We wanna have to accept in SG, so move accepted related parts of RL in
4835 the common part. (more precisely, the dynar of all known sockets is no
4836 more a static in transport.c, but part of the process_data)
4838 [gras_stub_generator]
4839 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4841 2004-07-03 Martin Quinson
4842 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4843 them all up in one shot)
4845 - Fix the ugly case of reference to dynamic array.
4846 - New (semi-public) function gras_datadesc_size to allow the messaging
4847 layer to malloc the needed space for the buffer.
4849 - gras_socket_close now expect the socket to close (and not its address to
4850 put NULL in it after it). This is because the socket passed to handlers
4851 is one of their argument (=> not writable).
4853 - propagate the interface cleanup from last week in datadesc, ie remove a
4854 superfluous level of indirection. User pass adress of variable
4855 containing data (both when sending and receiving), and not of a variable
4856 being a pointer to the data. Let's say that I like it better ;)
4857 The price for that is constructs like "int msg=*(int*)payload" in
4858 handlers, but it's a fine price, IMHO.
4860 - Let it work in RL (yuhu)
4862 2004-06-21 Martin Quinson
4864 - porting SG plugin and SG select to new standards (works almost).
4865 - plug memleaks and fix bugs around.
4868 - cleanup the prototype of data recv and force users to specify when they
4869 want to handle references to objects. Test case working even for cycles.
4870 - plug memleaks. Valgrind is perfectly ok with this.
4872 2004-06-12 Martin Quinson
4874 - cleanup the separation between plugin and main code in plugin creation
4876 2004-06-11 Martin Quinson
4878 - Reput hook for raw sockets, needed for BW experiments
4879 - kill a few lines of dead code
4880 [Data description] Interface cleanup
4881 - gras_datadesc_by_name returns the searched type or NULL.
4882 That way, no variable is needed to use a type desc once, which makes
4884 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4885 two parameters were strings (field name, type name), leading to
4887 [Dicos] Interface cleanup
4888 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4889 This is consistant with the dynar API.
4891 2004-04-21 Martin Quinson
4893 - Porting to new standards.
4895 - interface cleanup.
4896 There is no bag anymore, no need to take extra provision to mask the
4897 pointers behind "ID".
4898 Better splitup of functions between files create/exchange/convert.
4899 This is still a bit artificial since convert and receive are so
4900 interleaved, but anyway.
4902 - add a queued message list to procdata (the ones not matching criteria
4904 - factorize some more code between SG and RL wrt procdata
4906 - use gras_exit in example to track memleaks
4907 - get rid of gs_example now that GS is properly integrated into gras
4908 - update run_test to integrate the latest tests (datadesc)
4910 - rename WARNINGn macros to WARNn since it prooved error-prone
4912 2004-04-19 Martin Quinson
4914 - register init/exit functions within gras module mechanism
4915 - send/receive function.
4916 Convertion is not implemented, but short-cutted if not needed.
4917 struct/array elements are sent one by one (instead of block-wise), but
4918 nobody really cares (yet). Get a prototype before optimizing.
4919 - tests (using a file socket) for DD send/receive on:
4920 - base types: int, float
4921 - array: fixed size, string (ie ref to dynamic string)
4922 - structure: homogeneous, heterogeneous
4923 - chained list, graph with cycle
4924 Believe it or not, valgrind is not too unhappy with the results. The
4925 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4929 - Bugfix when using a filename explicitly (instead of '-')
4931 2004-04-09 Martin Quinson
4933 - factorize more code between RL and SG in socket creation
4934 - Complete the implementation and tests of:
4936 o file (only in RL, and mainly for debugging)
4938 I lost 3 days to design a portable address resolver, and then decided
4939 that the prototype mainly have to run on my box.
4940 Addressing portability too early may be like optimizing too early :-/
4942 - use gras_init in the Tests instead of the crappy parse_log_opt
4943 (the latter function is removed)
4944 [Conditional execution]
4945 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4946 [Code reorganisation]
4947 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4948 Build examples against the RL library, since there is no way to disable
4949 its creation for now.
4951 For information, the beginning of coding on GRAS was back in june
4952 2003. I guess that every line has been rewritten at least twice since