1 SimGrid (3.35.1) not released (target: Feb 24)
4 ----------------------------------------------------------------------------
6 SimGrid (3.35) November 23. 2023
8 The "Thanks Giving up stateful model-checking" release. Stateless model checking remains.
11 - New class ActivitySet to ease wait_any()/test_any()/wait_all()
12 - Deprecate {Comm,Io,Exec}::{wait_any,wait_all,test_any} and friends
13 - Simplify a bit the declaration of multi-zoned platforms from C++
14 - New function NetZone::add_route(host1, host2, links) when you don't need gateways
15 - Also add a variant with s4u::Link, when you don't want to specify the directions
17 - Zone's gateways can now be controlled directly.
18 - Add NetZone::add_route(zone1, zone2, links) specifying the route between zones
19 - Introduce a Mailbox::get_async() with no payload parameter. You can use the new
20 Comm::get_payload() once the communication is over to retrieve the payload.
21 - Implement recursive mutexes. Simply pass true to the constructor to get one.
22 - Simplify the expression of horizontal scaling of Tasks.
23 - Each Task now consists of a dispatcher, a collector and one or more instances.
24 - The parallelism degree of each of these can be set.
25 - Several examples have been added or modified accordingly.
26 - Update s4u::create_DAG_from_json() to support wfformat 1.4.
27 - Introduce a new MessageQueue abstraction and associated Mess simulated object.
28 The behavior of a MessageQueue is similar to that of a Mailbox, but intended for
29 control messages that do not incur any simulated cost. Information is automagically
30 transported over thin air between producer and consumer. See examples/cpp/mess-wait
31 - New function: Mutex::get_owner()
34 - New: Add a JBOD (just a bunch of disks) concept. It's a sort of host with many disks.
35 - Revamp the battery plugin: rewrite completely the API, for a better usability.
36 The examples were updated accordingly.
37 The battery can now act as a simple connector (see battery-connector example).
38 - Revamp of the photovoltaic plugin: now called SolarPanel and complete rewrite of the API
39 - Add chiller plugin: enable the management of chillers consuming electrical energy
40 to compensate heat generated by hosts.
41 - Add a battery-chiller-solar example combining several plugins to evaluate the amount
42 of brown energy (from the electrical grid) and green energy (from the solar panel)
43 during a given computation.
46 - New SMPI_app_instance_join(): wait for the completion of a started MPI instance
47 - MPI_UNIVERSE_SIZE now initialized to the total amount of hosts in the platform
48 - Memory usage due to SMPI for non-MPI actors greatly reduced.
49 - New implemented calls: MPI_Isendrecv, MPI_Isendrecv_replace
52 - Allow to use on valgrind-observed or gdb-observed processes.
53 - Install sthread on user's disk.
54 - Implement recursive pthreads.
55 - Implement pthread_barrier and pthread_cond (but conditional are not supported by the MC yet).
56 - Add some McMini codes to test sthread further (controlled with enable_testsuite_McMini).
59 - Remove stateful model-checking. This was not used, not really working, and very hard to fix.
60 Liveness properties cannot be verified anymore.
61 - More informative backtraces on assertion failure.
62 - Fix dependency bugs for mutex and other transitions
63 - Fix some reversible_race definitions, and also the rest of ODPOR.
66 - Make the host_load plugin available from Python. See examples/python/plugin-host-load
67 - Mailbox::get_async() does not return a pair anymore. Use comm.get_payload() instead.
68 - Comm::waitall/waitany/testany() are gone. Please use ActivitySet() instead.
69 - Comm::waitallfor() is gone too. Its semantic was unclear on timeout anyway.
70 - Io::waitany() and waitanyfor() are gone. Please use ActivitySet() instead.
71 - Do not export the values of enums. So you need to write e.g. SharingPolicy.LINEAR
72 while it should have been possible to write LINEAR alone before. This is the advised
73 behavior for modern C++ code.
76 - Introduce sg_activity_set_t and deprecate wait_all/wait_any/test_any for
80 - optimize an internal data structure (replace boost::circular_buffer_space_optimized by
81 std::deque to store pending and unmatched Comms in Mailboxes). It is actually a revert
82 to what was used a few years back. The boost structure had a lower memory footprint than
83 deques, but it appeared that their "space_optimized" character was generating a huge lot
84 of refcount changes on the stored Comms.
87 - Fix errors with ns-3 v3.36+
88 - Many other small bug fixes, in particular in MC and sthread.
90 ----------------------------------------------------------------------------
92 SimGrid (3.34) June 26. 2023
94 Save the planet, skip a release: 3.33 was due 6 months ago, so skip directly to 3.34.
97 - SimGrid now requires a compiler with C++17 support for public headers too.
98 Sibling projects should upgrade their FindSimGrid.cmake
99 - Remove the MSG API: its EOL was scheduled for 2020.
100 - Remove the Java bindings: they were limited to the MSG interface.
101 - On Windows, you now need to install WSL2 as the native builds are now disabled.
102 It was not really working anyway.
103 - Support for 32bits architecture is not tested anymore on our CI infrastructure.
104 It may break in the future, but we think that nobody's using SimGrid on 32 bits.
105 - Remove the surf module. It was replaced by the kernel/models module, and that
106 refactoring took almost 10 years to properly complete.
109 - Activity::set_remaining() is not public anymore. Use for example
110 Comm::set_payload_size() to change the size of the simulated data.
111 - New function: Engine::flatify_platform(), to get a fully detailed vision of the
113 - New Task abstraction: They are designed to represent dataflows, i.e, graphs of repeatable Activities.
114 See the examples under examples/cpp/task-* and the associated documentation.
115 - Full simDAG integration: Activity::start() actually starts only when all dependencies
116 are fulfilled. If it cannot be started right away, it will start as soon as it becomes
118 - Allow to set a concurrency limit on disks and hosts, as it was already the case for links.
119 - Rename Link::get_usage() to Link::get_load() for consistency with Host::
120 - Every signal now come with a static version that is invoked for every object of that class,
121 and an instance version that is invoked for this specific object only. For example,
122 s4u::Actor::on_suspend_cb() adds a callback that is invoked for the suspend of any actor while
123 s4u::Actor::on_this_suspend_cb() adds a callback for this specific actor only.
124 - Activity::on_suspended_cb() is renamed to Activity::on_suspend_cb(), and fired right before the suspend.
125 - Activity::on_resumed_cb() is renamed to Activity::on_resume_cb(), and fired right before the resume.
126 - Resource::on_state_change_cb() is renamed to Resource::on_onoff_cb() to distinguish from the
127 Activity::on_state_change_cb() that is related to the activity state machine, not on/off.
128 - Activity signals (veto, suspend, resume, completion) are now specialized by activity class.
129 That is, callbacks registered in Exec::on_suspend_cb will not be fired for Comms nor Ios.
132 - Battery: Enable the management of batteries on hosts.
133 See the examples under examples/cpp/battery-* and the documentation in the Plugins page.
134 - Photovoltaic: Enable the management of photovoltaic panels on hosts.
135 See the examples under examples/cpp/photovoltaic-* and the documentation in the Plugins page.
138 - optimize an internal data structure (use a set instead of a list for ongoing activities),
139 leading to a potentially big performance gain, in particular with many detached comms.
142 - New option smpi/barrier-collectives to add a barrier to some collectives
143 to detect dangerous code that /may/ work on some MPI implems.
144 - New function SMPI_app_instance_start() to easily start a MPI instance in your S4U simulation.
147 - Write the section of the manual about models, at least.
148 - WiFi: the total capacity of a link depends on the amount of flows on that link.
149 - Use the nonlinear callback feature of LMM to reflect this.
150 - Calibration values can be changed to match different MCS configurations
151 - See the example teshsuite/models/wifi_usage_decay/wifi_usage_decay.cpp
152 - See also "A Flow-Level Wi-Fi Model for Large Scale Network Simulation"
153 https://hal.archives-ouvertes.fr/hal-03777726
154 - Merge parameters network/bandwidth-factor and smpi/bw-factor that serve the same purpose.
155 - Same for the latency
156 - Rewrite the corresponding documentation.
157 - Allow to disable the TCP windowing modeling by setting network/TCP-gamma to 0.
158 - Finally kill the 'compound' host model. You can change the CPU or network model
159 with the default host model, as it should.
160 - Rename option "surf/precision" to "precision/timing" for clarity.
161 - Rename option "maxmin/precision" to "precision/work-amount" for clarity.
162 - New function: Engine::flatify_platform() to debug your platform.
165 - Implement pthread_join in MC mode.
166 - Implement semaphore functions in sthread.
167 - Add an intricated way to verify the access to non-reentrant data structures
168 It requires code annotation, as shown in examples/sthread/stdobject/stdobject.cpp
171 - Stateless model-checking is now usable on any system, including Mac OSX and ARM processors.
172 - The stateless aspects of the MC are now enabled by default in all SimGrid builds.
173 Liveness and stateful aspects are still controlled by the enabling_model-checking
174 configuration option.
175 - Introducing ODPOR and SDPOR reduction strategies
176 - Introducing guiding heuristics, trying to find bugs faster than DFS in reduced state space.
177 - Synchronize the MBI tests with upstream.
178 - Show the full actor backtraces when replaying a MC trace (with model-check/replay)
179 and the status of all actors on deadlocks in MC mode.
182 - simgrid::xbt::cmdline and simgrid::xbt::binary_name are gone.
183 Please use simgrid::s4u::Engine::get_cmdline() instead.
186 - New tutorial on simulating DAGs.
187 - New section in the user guide on the provided performance models.
188 - New section presenting some technical good practices for (potential) contributors.
189 - Add a section on errors and exceptions to the API documentation.
190 - Move the s4u examples to a section on their own to ease navigation.
192 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
193 (FG: issues on Framagit; GH: issues on GitHub)
194 - FG#18: Java bindings should be redone or removed
195 - FG!118: Wi-Fi callback mechanism
196 - FG!119: SMPI: add option to inject a barrier before every collective call
197 - GH#383: Segfault when adding a disk after load_platform(xml)
199 ----------------------------------------------------------------------------
201 SimGrid (3.32) October 3. 2022.
203 The Wiedervereinigung release. Germany was reunited 32 years ago.
206 - SimGrid now requires a compiler with C++17 support to compile the lib.
207 Our public headers still allow the user code to be compiled in C++14.
208 - Support graphviz v3 and ns-3 v3.36 (older versions are still supported).
209 - Tested with clang (v11, v13, v14 and v16), gcc (v7 to v13) and IntelCC v2022.2
213 - kill signal Comm::on_completion that was not working anyway.
214 - Expose signals Activity::on_suspend and Activity::on_resume
215 - New macro xbt_enforce(): similar to xbt_assert(), but throws an AssertionError
216 instead of calling abort().
217 - New: s4u::Exec::get_thread_count()
218 - Various cleanups around virtual machines:
219 - host_by_name() and friends now only return hosts. VMs are now excluded.
220 - It is now impossible to search a VM by name globally.
221 You can only search VM by name on a given PM, so either you know
222 the PM on which your VM runs and you can search by name, or you need
223 to manually iterate over all PMs to search this VM.
224 - The s4u::VirtualMachine constructor is now deprecated.
225 Please use s4u::Host::create_vm() instead.
226 - Rename s4u::VirtualMachine::on_creation() to on_vm_creation() to
227 avoid confusion with s4u:Host::on_creation() that is inherited.
228 Also s4u::VirtualMachine::on_destruction -> on_vm_destruction().
230 - One-sided communications (Comm::sendto) can now be detached,
231 and should now be more resilient to network and host faults.
234 - Added the following bindings / examples:
235 - Comm (now 100% covers the C++ interface):
236 - Comm.dst_data_size, Comm.mailbox, Comm.sender, Comm.start_time, Comm.finish_time
237 - Comm.state_str [examples: examples/python/comm-failure/, examples/python/comm-host2host/]
238 - Comm.remaining [examples: examples/python/comm-host2host/, examples/python/comm-suspend/]
239 - Comm.set_payload_size [example: examples/python/comm-host2host/]
240 - Comm.set_rate [example: examples/python/comm-throttling/]
241 - Comm.sendto, Comm.sendto_init, Comm.sendto_async [example: examples/python/comm-host2host/]
242 - Comm.start, Comm.suspend, Comm.resume [example: examples/python/comm-host2host/]
243 - Comm.test_any [example: examples/python/comm-testany/]
244 - Comm.wait_until [example: examples/python/comm-waituntil/]
246 - Engine.host_by_name [example: examples/python/comm-host2host/]
247 - Engine.mailbox_by_name_or_create [example: examples/python/comm-pingpong/]
249 - Mailbox: Mailbox.ready [example: examples/python/comm-ready/]
250 - Ptask [example: examples/python/exec-ptask/]:
251 - this_actor.exec_init
252 - this_actor.parallel_execute
255 - Added an AssertionError exception that may be thrown in case of error.
256 For instance, creating two hosts with the same name will now throw this exception
257 instead of killing the interpreter.
260 - Implement MPI_File_get_type_extent(), MPI_File_s/get_atomicity() and
261 MPI_File_get_byte_offset()
262 - Intercept getpid() calls to return the simulated ones.
263 - Fix various bugs in MPI IO.
265 Platform description & visualization:
266 - More robust sanity checks for platforms, to reject forbidden topologies with
267 a proper error message.
268 - New platform example: supernode.cpp and supernode.py.
269 The Python version generates a nice graphical representation of the platform.
270 - Bug fixes around fat-tree topologies.
271 - Allow to dump the platform topology as a CSV file representing the graph edges
272 with platform_graph_export_csv() (similar to the DOT export).
273 - Fix graphicator for "cluster" topologies (e.g. fat-tree, dragonfly).
276 - Fix a bug when using ptasks with multicores (FG!111).
279 - First bits of sthread, that intercepts pthread operations at runtime.
280 The intend is to use it together with simgrid-mc, but it is TBD.
281 - Sync MBI generators with upstream changes.
282 - Various cosmetics, small bug fixes and inner refactorings
284 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
285 (FG: issues on Framagit; GH: issues on GitHub)
286 - FG#105: "Variable penalty should not be negative!" with in-flight messages and bandwidth profiles
287 - FG#109: Application time reported by --cfg=smpi/display-timing:yes is wrong
288 - FG#110: Wait_any does not trigger new model solve when host events occur
289 - FG#111: Wrong execution time in rare cases when using multicore
290 - FG!98: Re-enable the tests for legacy stochastic profiles
291 - FG!109: Trigger new engine solve upon host events such as host on/off
292 - FG!116: SMPI/replay: Fix issue with recv of size =0
294 ----------------------------------------------------------------------------
296 SimGrid (3.31) March 22. 2022.
298 The ненасильство release. We stand against war.
300 Against the agression by a sick system that forces peoples to take arms against each other.
303 - Rework the internals, for simpler and modern code. This shall unlock many future improvements.
304 - You can now define plugins onto the DFS explorer (previously called SafetyChecker), using the
305 declared signals. See CommunicationDeterminism for an example.
306 - Support mutex, semaphore and barrier in DPOR reduction
307 - Seems to work on Arm64 architectures too.
308 - Display a nice error message when ptrace is not usable.
309 - New test suite, imported from the MPI Bugs Initiative (MBI). Not all MBI generators are integrated yet.
310 - Remove the ISP test suite: it's not free software, and it's superseeded by MBI.
313 - fix for FG#100 by ensuring small asynchronous messages never overtake larger
314 ones, conforming to the standard.
315 - replay: fix waitall behaviour to avoid forgetting requests and leaking
317 - tracing: ensure that we dump the TI traces continuously during execution and
318 not just at the end, reducing memory cost and performance hit.
319 - Update OpenMPI collectives selection logic to match current one (4.1.2)
320 - Add a coherence check for collective operation order and root/MPI_Op
321 coherence. Potentially costly so not activated unless smpi:pendantic is set
322 or -analyze is given.
325 - New signal: Engine::on_simulation_start_cb()
326 - Introduce a new execution mode with this_actor::thread_execute(). This simulate
327 the execution of a certain amount of flops by multiple threads ran by a host. Each
328 thread executes the same number of flops, given as argument. An example of this new
329 function can be found in examples/cpp/exec-threads.
330 - Reimplementation of barriers natively.
331 Previously, they were implemented on top of s4u::Mutex and s4u::ConditionVariable.
332 The new version should be faster (and can be used in the model-checker).
333 - Actor::get_restart_count(): Returns the number of reboots that this actor did.
336 - MSG_barrier_destroy now expects a non-const msg_barrier parameter.
338 New plugin: the Chaos Monkey (killing actors at any time)
339 - Along with the new simgrid-monkey script, it tests whether your simulation
340 resists resource failures at any possible timestamp in your simulation.
341 - It is mostly intended to test the SimGrid core in extreme conditions,
342 but some users may find it interesting too.
345 - New solver for parallel task: BMF.
346 - More realistic sharing of heterogeneous resources compared to the fair
347 bottleneck solver used by ptask_L07.
348 - Implement the BMF (Bottleneck max fairness) fairness.
349 - Improved resource sharing for parallel tasks with sub-flows (parallel
350 communications between same source and destination inside the ptask).
352 - "--cfg=host/model:ptask_L07 --cfg=host/solver:bmf": enable the ptask
353 model with BMF solver.
354 - "--cfg=bmf/max-iterations: <N>": maximum number of iterations performed
355 by BMF solver (default: 1000).
356 - "--cfg=bmf/precision: <N>": numerical precision used when computing
357 resource sharing (default: 1e-12).
358 - This model requires Eigen3 library. Make sure Eigen3 is installed to use BMF.
361 - Modifications of the Profile mechanism, with some impact on users
362 - Addition of a new (S4U) method to init profiles from generic functions to improve versatility
363 - Fix initial behaviour of state_profiles
364 - Modify periodicity to behave like a period, and not like a loop delay
367 - Drop xbt_dynar_shrink().
370 - Made the following bindings static (previously member functions):
371 - Actor: Actor.kill_all(), Actor.by_pid()
372 - Host: Host.by_name(), Host.current(), Host.on_creation_cb()
373 - Mailbox: Mailbox.by_name()
374 - Added the following bindings:
375 - this_actor.warning()
376 - Mailbox.put_init() [example: examples/python/comm-waitallfor/]
377 - Comm.detach() [example: examples/python/comm-waitallfor/]
378 - Comm.wait_for() [example: examples/python/comm-waitfor/]
379 - Comm.wait_any_for()
380 - Comm.wait_all_for() [example: examples/python/comm-waitallfor/]
381 - Mutex [example: examples/python/synchro-mutex/]
382 - Barrier [example: examples/python/synchro-barrier/]
383 - Semaphore [example: examples/python/synchro-semaphore/]
386 - Remove target "make uninstall" which was incomplete and no longer maintained.
388 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
389 (FG: issues on Framagit; GH: issues on GitHub)
390 - FG#57: Mc SimGrid should test whether ptrace is usable
391 - FG#87: Smpi scripts fail with spaces in paths
392 - FG#100: [SMPI] Order of the message matching is not guaranteed
393 - FG#101: LGPL 2.1 is deprecated license
394 - FG#104: "make uninstall" not up-to-date
395 - GH#151: Missing mutexes for DPOR.
397 ----------------------------------------------------------------------------
399 SimGrid (3.30) January 30. 2022.
401 The Sunday Bloody Sunday release.
403 Main user-visible changes:
404 - The SimDag API for the simulation of the scheduling of Directed Acyclic
405 Graphs has been dropped. It was marked as deprecated for a couple of years.
406 We finally complete the implementation of what has been called SimDag++
407 internally, i.e., porting the different features of SimDag on top of S4U.
408 The new way to simulate the execution of dependent activities directly by
409 maestro (without any other actor) is details in the examples/cpp/dag-* series
411 - The removal of SimDag led us to also remove the export to Jedule files that
412 was tightly coupled to SimDag. The instrumentation of DAG simulation is still
413 possible through the regular instrumentation API based on the Paje format.
414 - We also dropped the old and clumsy Lua bindings to create platforms in a
415 programmatic way. It can be done in C++ in a much cleaner way now, which
416 motivates this suppression.
419 - Introduce on_X_cb() functions for all signals, to attach a new
420 callback to the signal X. The signal variables are now hidden and
421 only these functions should be used.
422 Rationale: this enables the usual deprecation scheme where functions
423 remain for 4 releases if we need to modify the signals, while the
424 current code with the signal variables directly visible prevents any
426 - New function: Engine::run_until(date), to split the simulation.
427 - New signal: Activity::on_veto, to detect when an activity fails to start.
428 - Signal change: Comm::on_start(Comm&, bool) has been replaced by
429 Comm::on_send and Comm::recv. These two signals respectively correspond to
430 when the sending or receiving side of a Comm is ready. They are raised at
431 the same locations as the former Comm::on_start signal.
432 - New function: Engine::track_vetoed_activities() to interrupt run()
433 when an activity fails to start, and to keep track of such activities.
434 Please see the corresponding example for more info.
435 - New functions: s4u::Comm::{sendto_init, set_source, set_destination} to enable
436 the use of vetoers with direct host-to-host communications. Both source and
437 destination have to set for a comm to start. Each call to these setters check
438 if all vetoes are satisfied. When it is the case, the comm starts. A use case of
439 these functions is given in examples/cpp/dag-scheduling.
440 - New functions: {Exec, Io}::update_priority allow you to modify the priority of
441 these kinds of activities during their execution. Behavior is detailed in
442 examples/cpp/io-priority/.
445 - Dynamic costs for MPI operations: New API to allow users to dynamically
446 change injected costs for MPI_Recv, MPI_Send and MPI_Isend operations.
447 Alternative for smpi/or, smpi/os and smpi/ois configuration options.
448 - Fix some issues with the replay mechanism.
451 - Function xbt::Extendable::get_data() is now templated with the type of the
452 pointee. Untyped function is deprecated. Use get_data<void>() if you still
453 want to retrieve void*.
456 - New section: "SimGrid MPI calibration of a Grid5000 cluster"
457 presenting how to properly calibrate MPI communications in SimGrid.
458 - Complete and reword the platform section, which is now completed.
461 - Thread contexts are used by default with Python bindings. Other kinds of
462 contexts revealed unstable, specially starting with pybind11 v2.8.0.
464 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
465 (FG: issues on Framagit; GH: issues on GitHub)
466 - FG#95: Wrong computation time for multicore execution after pstate change
467 - FG#97: Wrong computation time for ptask+multicore+pstates
468 - FG#98: SMPI offline simulation is inconsistent with the online simulation
469 (deadlocks / message truncation)
470 - FG#99: Weird segfault when not sealing an host
472 ----------------------------------------------------------------------------
474 SimGrid (3.29) October 7. 2021
476 The "Ask a stupid question" release.
478 We wish that every user ask one question about SimGrid to celebrate.
479 On Mattermost, Stack Overflow or using the issues tracker.
481 New modeling features:
482 - Non-linear resource sharing, modeling resources whose performance heavily degrades with contention:
483 - The total capacity may be updated dynamically through a callback
484 and depends mainly on the number of concurrent flows.
485 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
487 - Dynamic factors: model variability in the speed of activities
488 - Each action can now have a factor that affects its progression.
489 This multiplicative factor is applied when updating the amount of work
490 remaining, thereby an activity with factor=0.5 only uses half of the
491 instantaneous power/bandwidth it is allocated and will appear twice
492 slower than what it actually consumes.
493 - This can be used to model a overhead (e.g., there is a 20 bytes
494 header in a 480 bytes TCP packet so the factor 0.9583) but the novelty
495 is this factor can now easily be adjusted depending on activity's and
496 resources characteristics.
497 - This existed for network (e.g., the effective bandwidth depends
498 on the message in SMPI piecewise-linear network model) but it is now
499 more general (the factor may depend on the source and destination and
500 thus account to different behaviors for intra-node communications and
501 extra-node communications) and is available for CPUs (e.g., if you
502 want to model an affinity as in the "Unrelated Machines" problem in
503 scheduling) and disks (e.g., if you want to model a stochastic
505 - For that, resources can be provided with a callback that computes
506 the activity factor when creating the action.
507 - Example: examples/cpp/exec-cpu-factors
508 - The same mechanism is also available for the latency, which
509 allows to easily introduce complex variability patterns.
512 - Added support to programmatic platform creation in Python.
513 Example: examples/python/clusters-multicpu
516 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
517 This can only be set through the API, not through XML files.
521 - Multiple fixes to ensure reproducibility of tracing
522 - scan/excan can now be replayed
523 - wait action now uses ranks and not pid, as the other ones.
524 - smpi/init and smpi/finalization-barrier are now valid for replays.
525 - exit() is now intercepted by SMPI to avoid premature shutdown of
526 simulation. First non 0 return codes is returned as simulation return
530 * New section "Release Notes" documenting recent and current developments.
531 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
532 * Improvements in API Reference for C++ and Python interfaces.
535 - Make wifi creation compatible with ns-3 version 3.34 too.
537 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
538 (FG: issues on Framagit; GH: issues on GitHub)
539 - FG#77: Search feature of doc is broken (update sphinx theme version)
540 - FG#78: Multiple fixes for SMPI replay:
541 - TI tracing of allotallv/w was outputting wrong values
542 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
544 ----------------------------------------------------------------------------
546 SimGrid (3.28) July 14. 2021
548 The Victoriadagarna Release.
551 - C++ platform interface: Users can now describe their platform directly in C++.
552 This provides greatly flexibility and performance improvement for complex
553 platforms. Main features:
554 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
555 "zones", instead of single hosts. This feature enables the description
556 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
557 - StarZone: new zone with a Star-like topology. The routes are defined
558 as a set of links used to communicate from node to everybody (node<->ALL).
559 - Split-Duplex links: auxiliary method to create split-duplex links in
560 the platform, easing its utilisation. It automatically creates both UP
561 and DOWN links (similarly as done in XML).
562 - Please refer to the documentation and the examples included:
563 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
564 - New plugin: Producer-Consumer with monitor. Just requires to include the
565 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
566 associated example (examples/cpp/plugin-prodcons).
569 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
570 s4u::Io::wait_any(), s4u::Io::wait_any_for().
571 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
572 parameter by reference, instead of a pointer.
573 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
574 Explicitly cancel the activity to get back to previous behavior.
575 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
578 - The default SMPI compiler flags are no more taken from the environment.
579 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
580 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
582 --cfg=smpi/finalization-barrier: which can be used to add
583 a barrier inside MPI_Finalize. This can help for some codes which cleanup
584 data attached to a process, but still used in other SMPI processes.
585 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
586 is active when set to false, to keep going after a small error
587 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
588 which may or may not be problematic in the end.
590 - fix behaviour, as maximum iteration count could be ignored
591 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
592 use sampling when the same kernel is called with a different set of parameters
593 which have an impact on the timing.
594 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
595 It should now work with smpi/auto-shared-malloc-thresh.
596 - Improve error handling and reporting in multiple places
597 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
598 validity checks, truncated messages are now an error, return errors
599 when explicitely deleted handles are reused, ...)
600 - RMA: multiple fixes and stability improvements.
601 - analysis (-analyze flag in smpirun):
602 - SMPI can now report buffer leaks as well as MPI handles leaks,
603 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
604 - if -trace-call-location is used when compiling, SMPI can report
605 origin of leaked handles/buffers
606 - group leaks by type/origin in output message if possible
607 - New implemented MPI calls: MPI_Comm_test_inter
610 - Changed internal implementation of bandwidth factors in network models.
611 Models affected: CM02, LV08 (default), SMPI, IB.
612 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
613 Bandwidth factors are applied to communications to describe that users
614 cannot use 100% of the available bandwidth. For example, the default network model,
615 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
616 behavior was implemented by limiting the bandwidth available in the LMM
617 system for this flow. This may give the false impression that there is
618 bandwidth available for other flows due to its underutilization, especially
619 for the dynamic bandwidth factors used in SMPI models.
620 To avoid this, we have modified the implementation so that each flow uses the
621 maximum physical bandwidth according to the LMM system.
622 However, the actual throughput of the flow seen by the user is defined by
623 the physical bandwidth multiplied by the bandwidth factor.
624 This change impacts on the simulation results for all network models on
625 which we have bandwidth factors configured.
626 *****************************************
627 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
628 *****************************************
629 This change may impact on the timing of your simulation results.
630 Take care when comparing simulations from different SimGrid's
631 versions. Sorry for the inconvenience.
632 - Dynamic network factors: users can configure a callback to define
633 the network factors dynamically. This API is available at
634 simgrid::kernel::resource::NetworkModelIntf.
635 - Users have access to complete information about the current communication
636 to decide which factor to apply. This includes: message size, source and
637 destination hosts, links and zones traversed.
638 - Dynamic factors for both latency and bandwidth.
639 - For more details, see the example in (examples/cpp/network-factors).
640 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
641 deprecated since version 3.24, are no longer supported. Instead, use
642 "wattage_off" and "wattage_per_state".
645 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
648 - New tutorial: Model-checking and formal assessment
649 - New sections: "Demystifying the routing" and "C++ platforms"
650 - Update and improve the part on visualization in MPI and Algo tutorials.
651 - Improve the section on routing: how to define it, how it's used internally
652 - Fix many issues, broken links and missing references in doxygen and Sphinx
655 - Lua platform files are deprecated. Their support will be dropped after v3.31.
658 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
660 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
661 (FG: issues on Framagit; GH: issues on GitHub)
662 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
663 - FG#64: Configuring smpi/IB-penalty-factors
664 - FG#67: Running computation concurrently with MPI_Iallreduce
665 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
666 - FG#71: Segmentation fault on invalid gw_src/gw_dst
667 - GH#322: Issue when an actor kills his host vm
669 ----------------------------------------------------------------------------
671 SimGrid (3.27) March 29. 2021
673 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
675 Main user-visible changes:
676 - All functions marked as deprecated for one year were removed.
677 - Removed all that was related to the "Storage" type in all APIs and bindings.
678 "Disks" have been introduced more than a year ago to replace them for the better.
679 It's time to finish this replacement.
680 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
681 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
682 avoiding "initializer element is not constant" errors with some codes. Please be
683 careful to recompile your MPI codes if switching to this new SimGrid version.
684 - Network model ns-3 now requires version ns-3.28 or higher.
685 - Removed some option aliases that were deprecated since v3.23 (June 2019)
686 - 'verbose-exit' is now 'debug/verbose-exit'
687 - 'simix/breakpoint' is now 'debug/breakpoint'
688 - 'clean-atexit' is now 'debug/clean-atexit'
691 - New functions: Comm::sendto_{init,async} to initiate a communication
692 on between two (possibly remote) hosts.
693 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
694 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
695 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
696 or Mailbox::get_async<void>() if you really want to play with void*.
697 - Unify the interface of Activity::on_{start/activity}
698 - New function: Comm::get_dst_data()
699 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
702 - ptask should now work better with multicore hosts (FG#37).
703 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
704 - Internal cleanups and refactoring around models, to prepare for programmatic
706 - Modernize our tests of the performance models
710 - New section: Configuring the logs from the command line.
711 - New section: Programming API of logs.
712 - New section: Getting realistic results
713 - New section: Examples of SMPI Usage
714 - New part: SimGrid's Internals, with a section on Design goals
715 - Move back to the classical Breathe tool chain, and fix many small glitches.
718 - Handle leak checking: SMPI will now display a message if MPI handles are still
719 present at the end of a simulation. Option smpi/list-leaks:n will display the n
720 first leaked pointers and their types.
721 - smpi/display-allocs option : Display Memory allocations analysis, with hints
722 about the largest allocations. This only works for allocations with malloc/calloc
723 and not compiled with SMPI_NO_OVERRIDE_MALLOC
724 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
725 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
726 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
727 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
728 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
729 - Replay now opens and parses the trace file only once for all ranks when possible.
731 - better handling of MPI_PROC_NULL
732 - MPI_Group_incl was broken in some cases
733 - PAPI tracing now fixed
734 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
735 - SMPI will now fail properly when receive size < sent size, instead of
736 just setting MPI_ERR_TRUNCATE and keep going.
737 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
738 - Remove useless build patches and split build/run patches on the page
739 - Build in parallel when possible
740 - Test on yum-based systems
741 - New nightly tested Proxy-apps:
742 - OpenMPI mpi-test-suite
743 - Coral MCB, LSMS, Qbox, CLOMP
744 - ECP/Mantevo Cloverleaf3D, Ember
745 - Codevault bhtree_mpi
748 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
749 - Start to overhaul the implementation to modernize it (still ongoing).
752 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
753 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
754 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
755 - Deprecate xbt_dynar_to_array().
756 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
758 C binding and interface:
759 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
760 - Constness requirement has been lowered for parameter argv of sg_actor_start()
761 and sg_actor_create() It's now (char* const*). The previous version is still
762 available as sg_actor_start_() and sg_actor_create_().
764 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
765 (FG: issues on Framagit; GH: issues on GitHub)
766 - FG#37: Parallel tasks are limited to 1 core per host
767 - FG#62: Running "smpirun -replay" on large networks
768 - FG!46: Fix a few potential memory leaks in SMPI colls
769 - GH#343: Wrong consumed energy when an exec timeouts
771 ----------------------------------------------------------------------------
773 SimGrid (3.26) December 15. 2020
775 The Release release (the French lockdown was eased today).
777 Important user-visible changes:
778 - SimGrid now requires a compiler with C++14 support.
779 Sibling projects should upgrade their FindSimGrid.cmake
780 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
781 several users had difficulties to understand issues when using high bandwidth or
782 small latency events. The new value was already the default for SMPI and
783 should not cause too much performance hit. It can change some simulation
784 timings if simulation was skipping a lot of small events (which was actually
785 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
786 - On some network models (L07, CM02, LV08), an implicit loopback link is created
787 for local communication. Default bandwidth was 500MBps and latency was 15µs,
788 which was rather bad. Default values have been changed to 10GBps and 0 latency.
789 Options network/loopback-lat and network/loopback-bw have been added to adapt
790 these values if needed. Network and routing models already providing separate
791 loopbacks (clusters and such) are not impacted by this change.
792 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
793 metadata that are added to each MPI message by MPI libraries (tag,
794 src/dest, and communicator information). This may change simulation times
795 for very small message sizes, but should be more realistic.
798 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
799 - New plugins: wifi_energy and link_load. See the associated examples.
800 - New examples: wifi networks, and communication suspend.
803 - LTO is now enabled for Intel/clang compilers.
804 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
805 Setting it to "auto" will use all available cores, while setting it to n will
806 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
807 - Remove obsolete runtime option 'contexts/parallel-threshold'.
808 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
809 has been kept for compatibility.
810 - Finally remove obsolete snake_case() aliases for runtime options.
811 - Further improve the documentation.
814 - the on_exit() of each actor is also executed when the simulation deadlocks.
815 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
816 An example is provided for s4u::Comm but it should work with Exec and Io.
819 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
820 https://framagit.org/simgrid/SMPI-proxy-apps
821 - MPI/IO: supports names starting with "./"
822 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
824 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
825 building shared libraries, this may cause issues, so environment variable
826 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
827 - most temporary files should now be created in /tmp dir (or equivalent).
828 If this one does not allow execution of code (noexec flag), this may cause issues.
829 Please use another tmp directory (using TMPDIR or equivalent system variable)
833 - A large refactoring is ongoing, but this should be transparent for users so far.
836 - Drop xbt_str_split().
838 C binding and interface:
839 - The return type of the sg_actor_on_exit() callbacks is now 'void'
840 (int value was previously ignored)
841 - Many C functions were renamed to follow a common naming style
842 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
843 names are kept until version 3.30.
844 - Many MSG tests were converted to the new S4U's interface in C, that
845 was extended for that.
847 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
848 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
849 - FG#41: Add sg_actor_create C interface
850 - FG#43: xbt::random needs some care
851 - FG#48: The Impossible Did Happen (yet again)
852 - FG#50: Suspending an actor executed at the current timestamp fails
853 - FG#52: Zero-seconds timeout : "That's in the past already"
854 - FG#53: Crash while using ns-3 network model
855 - FG#54: How to suspend a comm?
856 - FG!22: Stochastic Profiles
857 - FG!24: Documentation and fix for xbt/random
858 - FG!35: Add a modeling hint for parallel links in doc
859 - FG!36: [xbt/random] Read/Write the state of the RNG
860 - FG!42: Link cumulated load plugin
861 - FG!44: Test condition variable wait_until+notify_all
862 - GF#18137: Allow different stack sizes?
863 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
864 - GH#139: Allow pthread creation in SMPI
865 - GH#336: Packet-level simulation using SMPI?
866 - GH#345: Error in the LMM while migrating a VM
867 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
868 - GH#352: pip install / python setup.py install fail to find pybind11
869 - GH!337: Fix link_energy plugin for wifi platforms
870 - GH!339: Add Mailbox set_receiver method to python binding
871 - GH!344: Cast hugepages macros parameters to int64
873 ----------------------------------------------------------------------------
875 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
877 The Palindrom Day Release.
879 Important user-visible changes:
880 - Improve the Python usability (stability and documentation).
881 - A nasty synchronization bug was ironed out, see also below.
882 - Python's doc was organized as a separate tree, now integrated with C++.
883 - C bindings of S4U were not part of the doc.
884 - The C++ doc was also improved as methods are now split by theme.
885 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
886 - This is mandatory to use the Java bindings.
887 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
888 - SimDAG++: Automatic dependencies on S4U activities (experimental)
889 - Some features are already implemented but not all of them
890 - Cannot block an activity until it's scheduled on a resource
891 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
892 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
895 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
896 - Drop support for Viva/Triva (old visualization tools), hence removed
897 TRACE_get_node_types() and TRACE_get_edge_types() functions.
900 - Require pybind11 version 2.4.
901 - Greatly improve locking during context switches with Python.
902 - Declaring Python callbacks to this_actor::on_exit is now working.
903 - No more tests are failing currently, even on win32.
906 - Actor: Merge signals on_migration_start/end into on_host_change
907 - Actor: Rename migrate() into set_host()
908 - Disk: Allow users to get the read and write nominal bandwidth values
909 - Exec: Implement wait_for(timeout)
910 - Io: Implement wait_for(timeout)
913 - Parse errors now raise a simgrid::ParseError that you may want to catch.
916 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
917 been renamed to fix a spelling error. As usual, the previous definitions are
918 kept with a deprecation warning. However, to avoid runtime errors, it is now
919 forbidden to override these deprecated methods in a derived class. Please use
920 the new names immediately if you need to override them.
923 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
924 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
925 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
926 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
927 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
928 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
929 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
930 - New module: random, providing classical random numbers generators.
933 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
934 allocations' size above which they are considered "shared" by default (as if
935 they were performed through SMPI_SHARED_MALLOC macros).
936 Default = 0 = disabled feature.
937 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
938 This can cause some troubles if codes are already overriding these. If this
939 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
940 help, but will make this feature unusable.
941 - Input check is now performed much more consistently and outputs precise
942 warnings for each error found
943 - New supported calls : MPI_File_set_view, MPI_File_get_view
944 - MPI I/O now only supports disks, as storage is discontinued.
945 - MPI I/O allows opening files without specifying absolute path
946 - Note : SMPI configuration options are now initialized by smpi_init_options()
947 call, which should be called if SMPI is not being used through smpirun.
949 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
950 - FG#9: Python bindings crashing
951 - FG#39: Missing s4u::Comm::wait_any_for example
952 - FG#42: Add support for ThreadSanitizer (TSan)
953 - FG!19: Removing RngStream
954 - FG!20: A module for RNG calls
955 - FG!21: Choice between ad-hoc and standard distributions implementations
957 - FG!25: Fix link in Mutex doc
958 - FG!26: Fix links in SMPI interface doc
959 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
960 - GH#305: Unscheduled tasks are still excuted
961 - GH#313: smpirun: manual outdated w.r.t. --help
962 - GH#321: [S4U] Get task remaining work ratio
963 - GH#323: Crash when an actor turn off his physical host
964 - GH#335: Missing links on dragonfly example svg
965 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
966 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
968 ----------------------------------------------------------------------------
970 SimGrid (3.24) October 9. 2019
972 The Clean Disk Release.
975 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
976 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
977 thus becomes the default resource to execute I/O operations. Examples
978 and the file system plugin have been modified to use disks instead of
979 storage elements. Storage can still be used but is doomed to disappear.
980 Users are thus adviced to update their simulators accordingly.
981 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
982 for consistency with pthread_barrier_wait()
983 - Host::get_englobing_zone() returns the englobing netzone
984 - Actor::on_destruction is now called in the destructor
985 Actor::on_termination new signal called when the actor terminates
987 - Global signals are now part of the Engine:
988 - on_platform_creation: after config settings, before the XML parsing
989 - on_platform_created: right after the XML parsing
990 - on_time_advance: each time the clock advances
991 - on_simulation_end: after simulation, before cleanups
992 - on_deadlock: as the name implies.
994 - sg_{actor,host,link}_{data,data_set}() now all exist.
995 Use them to attach user data to the object and retrieve it.
998 - Introduce an experimental Wifi model. It sounds reasonable
999 according to the state of the art, but it still has to be properly
1000 validated, at least against ns-3 if not against reality.
1001 - Improved the usability of ns-3. Several bugs were ironed out.
1002 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
1003 It is now expressed as 'idle:epsilon:allCores' to properly model the
1004 consumption of non-whole tasks on mono-core hosts. More info in the doc.
1007 - convert a new set of functions to the S4U C interface and move the old MSG
1008 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
1011 - Fortran bindings for DVFS have been removed.
1012 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
1013 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
1014 - SMPI can now be selected by cmake's find_module(MPI) with
1015 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
1016 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
1017 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
1018 - trace-call-location can be used with TI traces, and replayed, and is compatible
1019 with smpi/comp-adjustment-file.
1020 - sleep events are now correctly traced/replayed.
1021 - Default for trace-call-location is now to use file names and not full paths.
1022 To revert to previous behavior (in case of collision of filenames), option
1023 "smpi/trace-call-use-absolute-path" can be set to yes.
1026 - Use the included xxHash as an hash implem when C++14 is usable.
1027 - Option model-checker/hash was removed. This is always activated now.
1028 - New option smpi/buffering controls the MPI buffering in MC mode.
1029 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
1030 This is useful to check for MPI compliance.
1033 - New section on plugins: how to define a new one, and existing ones.
1036 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
1037 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
1040 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
1041 and <mount> tags. The rationale is finish to decouple what should be handle
1042 by SimGrid kernel (i.e., resources and activities) from plugins built in
1043 user space (e.g., the file system plugin). This new tag currently accepts
1044 only three attributes (id, read_bw, and write_bw). All the other information
1045 that was declared with the storage related tags now has to be expressed as
1046 properties. An example of platform using this new tag is available at
1047 examples/platforms/hosts_with_disks.xml
1050 - 'expect signal' can now accept more than one potential signal.
1052 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
1053 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
1054 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
1055 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
1056 - FG#31: per-actor data
1057 - FG#34: SG_BARRIER_SERIAL_THREAD?
1058 - FG#35: model-checker does not like buster-produced binaries
1059 - FG!13: MC: complete workaround in the error msg seen on modern systems
1060 - FG!15: execute_flops now logs compute
1061 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
1062 - FG!17: ns-3: unblock the right number of communications + others issues
1063 - FG!18: Improving the performance of the ns-3 bindings
1064 - GH#189: Energy consumption of parallel tasks
1065 - GH#219: Error in the throughput of TCP transfer
1066 - GH!330: Minor polishing of English: Configuring_Simgrid
1067 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
1068 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
1069 - GH!334: Rework energy plugin, again
1071 ----------------------------------------------------------------------------
1073 SimGrid (3.23.2) July 8. 2019
1076 - Nicer introduction page.
1077 - Migrate the "Deploy your application" page to the new doc.
1078 - Move Java as a subtree of MSG.
1081 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
1084 - Drop xbt_dynar_sort_strings().
1087 - Really fix FG#26: Turning off a link should raise NetworkFailureException
1088 - FG#27: Wrong exception thrown to wait_any when link is turned off
1089 - GH#328: Java: Canceling multiple tasks in a single vm/host
1091 ----------------------------------------------------------------------------
1093 SimGrid (3.23) June 25. 2019
1095 The Exotic Solstice Release.
1098 - SunOS and Haiku OS support. Because exotic platforms are fun.
1099 - Stop setting random seed with srand48() at initialization.
1100 - Use addr2line as a fallback for stacktraces when backtrace is not available.
1101 - Build option -Denable_documentation is now OFF by default.
1102 - Network model 'NS3' was renamed into 'ns-3'.
1105 - SimGrid can now hopefully be installed with pip.
1108 - wait_any can now be used for asynchronous executions too.
1111 - New log appenders: stdout and stderr. Use stdout for xbt_help.
1112 - Drop xbt_dict_dump.
1115 - 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
1116 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
1117 - MPI interface is now const correct for input parameters
1118 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
1119 - Fortran bindings for async collectives.
1120 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
1123 - Remove option 'model-check/record': Paths are recorded in any cases now.
1124 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
1125 always sparse. That's an immense gain in memory, and can even be
1126 faster because of cache effects. So there is no need to clutter the
1127 code to allow the user to go for the unefficient mode.
1130 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
1131 models proved to be more accurate than these old experiments.
1133 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
1134 - FG#1: Broken link in error messages
1135 - FG#2: missing installation documentation
1136 - FG#3: missing documentation in smpirun
1137 - FG#6: Python bindings not available on PyPI
1138 - FG#7: simple cmake call requires doxygen
1139 - FG#8: make python bindings an optional dependency
1140 - FG#10: Can not use MSG_process_set_data from SMPI any more
1141 - FG#11: Auto-restart actors forget their on_exit behavior
1142 - FG#12: -Denable_lto=OFF doesn't disable LTO
1143 - FG#13: Installs unstripped file 'bin/graphicator'
1144 - FG#14: Installs the empty directory 'doc/simgrid/html'
1145 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
1146 - FG#17: Dead link in doc (pls_ns3)
1147 - FG#20: 'tesh --help' should return 0
1148 - FG#21: Documentation link on http://simgrid.org/ broken
1149 - FG#22: Debian installation instruction are broken
1150 - FG#26: Turning off a link should raise NetworkFailureException exceptions
1151 - GH#133: Java: a process can run on a VM even if its host is off
1152 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
1153 - GH#326: Valgrind-detected error for join() when energy plugin is activated
1155 ----------------------------------------------------------------------------
1157 SimGrid (3.22) April 1. 2019
1159 The Easter Christmas Release (this one is somewhat late).
1162 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
1163 - Not complete yet: asynchronous activities (amongst others) are still missing
1164 - Still ongoing: the interface may change in the future. We need more
1165 testers! Please report any glitches.
1166 - No new project using Java should start now. Please switch to Python.
1169 - Some of the internal cleanups may lead to speed improvements:
1170 - The hard limitation on the amount of simulated actors with
1171 Java+Mac was removed. Now, the available memory is the only limit.
1172 - Our refcounting was tidyied, leading to 10% speedups in some cases.
1173 - We are still working on making our code robust to the actor kills
1174 and hosts' churn. Things are improving, but it's not perfect yet.
1175 - Replaced our own code to display a backtrace (that was forking addr2line)
1176 with the Boost.Stacktrace library.
1177 You won't see your backtraces without this optional dependency.
1178 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
1179 - Stop setting random seed with srand() at initialization.
1182 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
1183 XML file version remains 4.2 since old files are still compatible.
1186 - Process termination which was broken at version 3.21 has been repaired.
1187 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
1188 - Hide the examples into examples/deprecated. New users should use Python.
1191 - Drop MSG_process_create_from_stdfunc() from the API.
1192 This C++ function was a pimple in the C API, made necessary at some
1193 point by the Java bindings. This is fixed now.
1194 - Hide the examples into examples/deprecated. New users should use S4U.
1195 - MSG_process_create and MSG_process_attach now crash if the host used
1196 to run this process is off.
1197 - Fix the protype of MSG_process_on_exit()
1198 Now use (int,void*) callbacks instead of (void*,void*) ones.
1199 The implementation was ways too messy to actually work, I guess.
1202 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
1203 too often and provide more accurate timings.
1204 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
1205 - MPI_Alltoallw support
1206 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
1207 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
1208 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
1209 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
1211 - Basic implementation of generalized requests (SMPI doesn't
1212 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
1215 - Drop sg_cmdline. Please use xbt_cmdline instead.
1216 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
1217 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
1218 OsSemaphore is implemented in a portable way with C++11 threads.
1219 This should allow much more threads to be created at the same time,
1220 allowing Mac and Java users to simulate many more actors.
1221 - Implement the 'thread' factory with std::thread instead of xbt ones.
1222 It is not possible to set the stack size with threads anymore, but
1223 -fsplit-stack is the way to go nowadays when using threads.
1224 - Drop the xbt_os_thread_t module (now unused)
1225 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
1226 - Drop xbt_str_join_array().
1227 - Drop cunit, use Catch2 instead.
1230 - Many cleanups in the kernel::activity namespace. This was long
1231 overdue, and shall open the path to many future endeavors.
1234 - #132: Java: a process can not shut down its own host
1235 - #220: S4U: detached send causes exception if sender terminates soon after sending
1236 - #261: Document the parameters of parallel execution's constructor
1237 - #300: [s4u] BarrierPtr is missing
1238 - #314: SMPI args internal cleanup
1239 - #316: Fix a bug related to the CPU utilization of multi-core VM
1240 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
1241 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
1242 - #325: Turning off a host has different behavior on sleeping actors and computing actors
1244 ----------------------------------------------------------------------------
1246 SimGrid (3.21) October 3. 2018
1248 The Restarting Documentation (TRD) Release.
1251 - Convert the user manual to Sphinx for improved usability.
1252 Unstable version now hosted on https://simgrid.frama.io/simgrid
1253 Many glitches remain; stay tuned.
1254 - Restrict the Doxygen reference API to the installed header files.
1255 (documenting all internals this way is overkill + doxygen scales badly)
1256 - New tutorial on S4U (highly inspired from the old MSG one, but with
1257 a git repository to fork as a starting point, and a docker image)
1258 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
1259 - The Developper manual is still to be converted (not compiled until then)
1260 - Some parts are still missing in this conversion (such as the
1261 platform and deployment sections) while others were blindly converted
1262 and would need more love (such as the configuration flags).
1263 Things will certainly further improve in the future releases.
1266 - Fully reimplement auto_restart mechanism that was utterly broken
1267 (fix #22, #131 and #281 that were all very related)
1268 - Implement semaphores (fix #303 and #291)
1269 - When creating an actor from a function and its parameters,
1270 move-only parameters are not allowed anymore, as it would prevent
1271 the actor to be restartable if its parameters are consumed on run.
1272 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
1273 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
1274 IO operations that can be started, waited for, or canceled as a regular
1276 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
1280 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
1281 - New host variable: 'core_count'
1284 - Remove xbt_os_thread_specific features
1285 - Remove portability wrapper to condition variables
1286 - Remove xbt_os_thread_yield()
1289 - MPICH collective selector now mimics MPICH 3.3b
1290 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
1291 - MPI_Init(NULL,NULL) is now allowed
1292 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
1293 - Command-line arguments should now be handled properly in Fortran simulations
1294 - MPI Attributes and related callbacks should work in Fortran
1295 - Apps using C + Fortran should now work
1296 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
1297 - Support MPI_Error_String
1300 - Due to an internal bug, Msg.run() must now be your last line.
1301 We hope to fix it in a future release, and we are sorry for the inconvenience.
1304 - #22: Process autorestart seem to only work with CAS01 cpus
1305 - #56: Feature request: dynamic SMPI replay
1306 - #93: simgrid should not eat --help
1307 - #111: How to change the loopback link characteristics?
1308 - #116: Platform cluster radical
1309 - #117: Platform cluster/host order
1310 - #129: Replay_multiple on steroid
1311 - #131: Java : autostart does not work
1312 - #143: Setting a breakpoint at a given time
1313 - #235: Network energy models should be integrated
1314 - #264: Add ptask L07 resource tracing
1315 - #271: Dynamic and manual replay of SMPI traces
1316 - #279: Breakpoints option uses time deltas instead of absolute time
1317 - #281: Daemonized actor no longer daemonized after an auto-restart
1318 - #282: TIT convertor erase traces when using absolute path in the trace list file
1319 - #285: segfault when a s4u actor kills itself with msg process tracing activated
1320 - #286: Pajé traces are not exposing the number of cores
1321 - #287: Command-line parsing should stop when encountering '--'
1322 - #288: MPI_Init(NULL, NULL)
1323 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
1324 - #290: Method to check if a mailbox has a message ready to be consumed
1325 - #291: [s4u] semaphores do not seem available yet
1326 - #294: Alltoall collective implementation of mpich
1327 - #295: Starting a Host as OFF is not supported
1328 - #296: DTD too permissive
1329 - #299: Add a s4u barrier example
1331 ----------------------------------------------------------------------------
1333 SimGrid (3.20) Released June 24. 2018
1335 The proxy snake_case() release.
1338 - The whole API is now using snake_case() instead of mixing it with camelCase.
1339 Compatibility wrappers in place for one year.
1340 - We took the opportunity to increase the consistency of the API.
1341 Please report any remaining glitches.
1342 - New function simgrid::s4u::Host::get_actor_count:
1343 Returns the number of actors running on a specific host.
1346 - SMPI is now tested with ~45 proxy apps from various sources, with none or
1347 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
1348 - Replay: The replay file has been re-written in C++.
1349 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
1350 supported. They are stored in the trace and used when replayed.
1351 - Basic support of MPI_Cancel. Robustness not guaranteed.
1352 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
1353 (only for MPI_COMM_TYPE_SHARED).
1354 - New option: smpi/privatize-libs, to add external shared libs to be privatized
1355 by SMPI. They will be copied locally and loaded separately by each process.
1356 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
1357 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
1358 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
1359 with internal index optind with multiple processes. Only works if MPI_Init has
1360 already been called.
1361 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
1362 syntax without preprocessing tricks.
1365 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
1366 behavior tracing is no longer limited to MSG
1367 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
1368 this means that traces now need to be updated or re-obtained; both Irecv
1369 and Isend lines in your traces have now in total 5 mandatory fields:
1370 <rankid> <command> <to/from rankid> <tag> <size>
1371 To update your traces, it suffices to add a 0 for the tag here.
1372 - TIT now also supports waiting for a distinct request via MPI_Wait.
1373 Wait/Test now wait for a specific request, not just the last one that was
1374 issued. This unfortunately means another update, because we need to
1375 identify which request you want to wait for. We do this via the
1376 triplet (sender, receiver, tag), which needs to be added:
1377 <rankid> <command> <sender> <receiver> <tag>
1378 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
1379 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
1382 - The deprecation of MSG is ongoing (but this should not impact you).
1383 Many MSG functions are now simple wrappers on the C API of S4U. If
1384 you wish to convert your code to S4U, find the S4U counterparts of
1385 your MSG calls in src/msg/msg_legacy.cpp.
1386 - MSG can still be used, but won't evolve anymore.
1389 - Allow to run the Link energy plugin from the command line with
1390 --cfg=plugin:link_energy
1391 - Rename Energy plugin into host_energy
1392 - Rename Load plugin into host_load
1395 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
1396 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
1397 - Improve performance when handling timeouts of simix synchros.
1400 - Config: the C API is now deprecated (will be removed in 3.23), and
1401 the C++ API has been slightly improved.
1404 - Fix several build issues on OSX.
1405 - Move simgrid_config.h to simgrid/config.h (old header still working)
1408 - #143: Setting a breakpoint at a given time
1409 - #150: Inconsistent event names in SMPI replay
1410 - #258: daemonized actors hang after all non-daemonized actors have completed
1411 - #267: Linker error on unit_tmgr
1412 - #269: SMPI: tracing of MPI_Wait/all/any broken
1413 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
1415 ----------------------------------------------------------------------------
1417 SimGrid (3.19.1) Released March 22 2018
1419 The Fixed ABI Release.
1421 As you may know, we are currently refactoring SimGrid in deep.
1422 Upcoming SimGrid4 will be really different from SimGrid3: modular,
1423 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
1425 Our goal is to smooth this transition, with backward compatibility
1426 and automatic update paths, while still progressing toward SimGrid4.
1428 SimGrid remains open during works: The last pure SimGrid3 release was
1429 v3.12 while all subsequent versions are usable alpha versions of
1430 SimGrid4: Existing interfaces remain unchanged, but the new S4U
1431 interface is budding and the internals are deeply reorganized.
1433 Since 2015, we work hard to reduce the changes to public APIs. When
1434 we need to rename a public library symbol in S4U, we let your
1435 compiler issue an explicative warning when you use the deprecated
1436 function. These messages remain for four releases, i.e. for one full
1437 year, before turning into an error. Starting with v3.15, your can
1438 also adapt to API changes with the SIMGRID_VERSION macro, that is
1439 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
1441 Starting with this v3.19.1, our commitment to reduce the changes to
1442 the public interfaces is extended from the API to the ABI: a program
1443 using only MSG or SimDag and compiled against a given version of
1444 simgrid can probably be used with a later version of SimGrid without
1445 recompilation. We will do our best... but don't expect too much of
1446 it, that's a really difficult goal during such profund refactoring.
1448 The difference between v3.19 and v3.19.1 is that the former was
1449 accidentally breaking the ABI of MSG, while the later is restoring
1452 S4U and kernel APIs will still evolve until SimGrid4, with one-year
1453 deprecation warnings as currently. In fact, cleaning up these
1454 interfaces and converting them to snake_case() is one release goal of
1455 v3.20. But don't worry, we are working to smooth this upgrade path.
1457 In summary, new projects should start with S4U to benefit of the
1458 future, but old MSG projects should still be usable with no change.
1460 ----------------------------------------------------------------------------
1462 SimGrid (3.19) Released March 20 2018
1464 The Moscovitly-cold Spring release.
1466 SMPI user-visible improvements:
1467 - Support of MPI_Type_create_subarray
1468 - Unimplemented MPI calls now abort (instead of issuing a warning)
1469 Some harmless calls around errhandlers still only issue a warning.
1470 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
1471 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
1472 is now possible (mmap privatization was refactored to this end).
1473 Adding new SMPI instances should also be technically doable now.
1475 SMPI internal cleanups:
1476 - Removed index notion from SMPI -- no more "getPid() - 1"!
1477 In tracing files, actors are called rank-<PROCESS-ID> but these are
1478 NOT the real ranks (tracing several communicators is not supported yet).
1479 - Deprecate smpi_process_index() (will be removed in 3.22)
1480 - Better testing on classical proxy apps, through an external project:
1481 https://github.com/simgrid/SMPI-proxy-apps/
1484 - Use a graphical TOC to make it easier to find the documentation you need
1485 - Revamp the MSG tutorial
1488 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
1489 - Remove parameter of MSG_process_killall().
1490 Resetting the PID was bogus anyway (several actors could have the same PID).
1493 - Execution->setHost() can be called after start(), to migrate it.
1494 - Comm::test_any() is now implemented.
1495 - s4u::Actor now has onCreation() and onDestruction() signals.
1496 - Install some kernel header files for the users' plugins and more.
1497 Warning, their API is really not stable yet.
1498 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1499 include/simgrid/kernel/routing/*Zone.hpp
1502 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1504 Build System and other cleanups:
1505 - Remove unused run-time parameter "tracing/onelink-only".
1506 - SimGrid now works with intel compilers (MC still buggy)
1507 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1508 - tesh kills subprocesses on timeout (but not on windows)
1511 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1514 - #194: Feature request: simgrid::s4u::Comm::test_any()
1515 - #245: migrating an actor does not migrate its execution
1516 - #253: Feature Request: expose clusters as objects
1517 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1518 - #255: Tesh broken on Windows
1519 - #256: Modernize FindSimGrid.cmake
1520 - #257: Fix (ab)use of CMake install
1522 ----------------------------------------------------------------------------
1524 SimGrid (3.18) Released December 24 2017
1526 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1528 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1529 - Support for MSG should not be removed from SimGrid before 2020,
1530 but future evolutions will be limited to the integration of
1531 user-provided patches.
1534 - s4u::onDeadlock() signal, run before stopping the simulation.
1535 - s4u::Exec: asynchronous executions (abstraction of background
1537 - s4u::Host->getLoad() returns the achieved speed in flops/s
1540 - Link::name() is deprecated and will be removed in v3.21, use
1541 Link::getCname() instead.
1542 - Mailbox::getName() changed to return a std::string, use
1543 Mailbox::getCname() instead to get a char*.
1544 - Storage::getName() changed to return a std::string, use
1545 Storage::getCname() instead to get a char*.
1546 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1547 to have both allocation and deallocation of the map in user space.
1550 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1551 The amount of flops remaining to do is a vector, not a scalar.
1552 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1553 name implies on both sequential and parallel tasks.
1554 - Both changes fix GitHub's #223 using PR #237 as a basis.
1555 Thanks Michael Mercier.
1556 - Most examples were converted to S4U and hidden elsewhere at tests.
1557 Rationale: we still want MSG to work; we want newcomers to use S4U.
1560 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1561 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1564 - Switch to the faster dlopen privatization mechanism by default
1565 - Documentation improvments
1568 - Live migration is getting moved to a plugin. Dirty page tracking is
1569 the first part of this plugin. This imply that VM migration is now
1570 only possible if one this function is called:
1571 - C/MSG: MSG_vm_live_migration_plugin_init()
1572 - C/C++: sg_vm_live_migration_plugin_init()
1573 - Java: Msg.liveMigrationInit()
1574 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1578 - Define class simgrid::xbt::Path to manage file names.
1579 - Removed unused functions:
1580 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1581 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1582 xbt_graph_export_graphviz()
1583 - xbt/str.h: xbt_str_join()
1584 - Remove unused datatypes:
1585 - xbt/heap.h: use std::priority_queue or boost::heap instead
1586 - xbt/swag.h: use boost::intrusive::list instead
1589 - New link_energy plugin for the consumption of the links.
1590 - All of the operations on files and storage contents have been
1591 packaged into a plugin (src/plugins/file_system). The current
1592 public interface can be found in
1593 include/simgrid/plugins/file_system.h
1594 To use these functions you now have to initialize the plugin by
1595 calling MSG_storage_file_system_init() just after calling
1596 MSG_init() or sg_storage_file_system_init() just after creating
1600 - Remove the undocumented/untested tag <include>
1603 - Remove viva specific tracing as the tool is no longer maintained
1606 - #248: Kill of finished processes leads to segfault
1607 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1608 - #239: please implement signal s4u::onDeadlock()
1609 - #230: segfaults when exit() without run()
1610 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1611 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1612 - #222: Actor::kill() doesn't really kill and segfaults
1613 - #221: odd LMM warning when killing an actor
1614 - #120: Memory leak when the processes are forcefully killed
1616 ----------------------------------------------------------------------------
1618 SimGrid (3.17) Released October 8 2017
1620 The Drained Leaks release: (almost) no known leaks despite the tests.
1622 Even very long simulations will preserve your memory: our testsuite
1623 stresses SimGrid for over 45mn on fast machines for a coverage of
1624 over 80%, and there is only one single known leak, of about 4kb.
1628 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1629 handler may be given to free resources if the comm cannot be completed.
1630 - this_actor::send(mailbox) is now mailbox->put()
1631 - New: simgrid::s4u::Comm::wait_all()
1632 - New: Host.execute() for remote executions.
1635 - Unused option network/sender-gap is removed.
1638 - Deprecate MSG_task_isend_with_matching(): This unused feature
1639 really complicates our internals. Will be removed in v3.20.
1642 - Improved context termination. It is now possible to free resources, even
1643 when a process is forcibly killed.
1644 - ContextBoost: add support for Boost versions above 1.61.
1647 - Bring back run-time option --cfg=exception/cutpath to remove exception
1649 - Removed unused functions:
1650 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1651 xbt_str_rtrim(), xbt_str_trim().
1652 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1655 - Removed header files obsolete since SimGrid 3.12:
1656 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1658 - #3: SD_exit should be made optional
1659 - #120: Memory leak when the processes are forcefully killed
1660 - #159: Threading test regression in Actor refcounting
1661 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1662 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1663 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1664 - #191: VM migration and pstate
1665 - #192: Updating the energy consumptions of all hosts crashes with VMs
1666 - #195: All actors have PID=0 in the logs
1667 - #204: Sometimes segfault with thread contexts and mmap privatization
1668 - #222: Actor::kill() doesn't really kill and segfaults
1669 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1671 ----------------------------------------------------------------------------
1673 SimGrid (3.16) Released June 22. 2017.
1675 The Blooming Spring Release: developments are budding.
1678 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1679 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1680 - Java JAR file should be finally fully working on Mac OSX too.
1682 XML platforms: Switch to platform v4.1 format.
1683 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1684 - <zone> can be used as a synonym for the now deprecated <as>
1685 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1686 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1687 - <actor> can be used as a synonym for the now deprecated <process>
1688 - state_file and avail_file periodicity is now easier to express
1689 (check the documentation) the old behavior should still work.
1690 * Storage: not backward compatible, but it seems that nobody ever used it
1691 - Remove attribute 'content_type' of <storage_type>: was never used
1692 - Make attribute 'model' of <storage_type> optional: for future usage
1693 - Remove Bconnection model property: was never (in)validated. Replaced by
1694 the maximum of the read and write bandwidth as a resource constraint.
1697 - New and Backwards Compatibility break:
1698 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1699 When one wants to get the list of tasks whose states have changed during a
1700 simulation round, s/he has to allocate and free a dynar and use it as
1701 argument to this function. The former SD_simulate (double how_long)
1705 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1706 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1709 - New function to update the consumption of all hosts at once.
1710 - Fix the model for multi-core hosts, linear in the amount of busy
1711 cores with an abnormality for pIdle (see paper).
1712 BUG: Still not satisfactory for ptask on mono-cores.
1715 - The netzone are now available from the MSG API.
1716 The old names still work, but are now deprecated.
1719 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1720 - New option: smpi/keep-temps to not cleanup temp files
1721 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1722 mallocs mode is used, allows to change the size of the fake file used
1723 (default 1MB), to potentially limit the number of mappings for large runs.
1724 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1725 - Fortran ifort and flang compilers support
1726 - New RMA calls supported (experimental):
1727 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1728 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1729 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1730 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1731 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1732 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1735 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1736 - Behind the scene, ActivityImpl are now automatically refcounted.
1737 Removing this nasty bottleneck will greatly simplify our internals.
1740 - Replay: New function xbt_replay_action_get():
1741 Retrieve the function previously associated to an event type.
1742 - DROPPED FUNCTION: xbt_str_varsubst()
1743 - DROPPED MODULE: strbuff. We don't need it anymore.
1744 - DROPPED MODULE: matrix. We don't need it anymore.
1745 - DROPPED MODULE: lib. We don't need it anymore.
1747 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1749 SimGrid (3.15) Released March 22. 2017
1751 The Spring Release: continuous integration servers become green
1752 We fixed even the transient bugs on all target architectures:
1753 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1754 Windows, FreeBSD, NetBSD.
1756 - Rename NetCards to NetPoints.
1757 This was intended to help NS3 users, but that's not a netcard.
1758 That's a point in the routing algorithm, let's avoid wrong simplifications.
1761 - New: MSG_process_yield(). Stop and yield to other processes.
1762 - New: MSG_process_daemon(). Daemon processes are automatically killed
1763 when the last non-daemon process terminates
1764 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1765 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1766 to make room for the upcoming network energy plugin.
1767 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1770 - Ensure that an actor can kill itself with Process::exit()
1771 - Kill the obscure NativeException. Nobody want to survive the issues
1772 it denotes, so use JniException that is a RuntimeException (not to
1773 be caught explicitly).
1774 - Partial bug fix in initialization. SimGrid flags on command line were
1775 consumed at C level but stayed in the original Java String[] args.
1776 This could mess users' args[i] if SG flags were not put at the end of
1778 The SimGrid flags are now removed from the Java arguments. However,
1779 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1780 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1781 It might be if you have --log or --cfg flags in the command line.
1782 - Fix numerous memleaks all around the place. In particular, around VMs.
1786 - s4u::onPlatformCreated: right before the simulation starts
1787 - s4u::onSimulationEnd: right after the main simulation loop
1788 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1789 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1790 event from the availability_file changes the avail speed.
1791 - Links are now usable from s4u
1792 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1793 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1794 - New examples: The conversion of MSG examples to S4U has begun
1795 - Actors: create, kill, migration, and suspend
1796 - Applications: master-worker and token-ring
1797 - Action replay: communications and storage
1798 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1801 - Backwards Compatibility breaks
1802 - The SD_route_get_list and SD_route_get_size functions have been removed.
1803 They are replaced by the unique
1804 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1805 The route from src to dst is built in the links dynar whose size can
1806 be obtained with xbt_dynar_length.
1807 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1808 sg_host_route_bandwidth and sg_host_route_latency.
1809 Macros ensure the backwards compatibility, but you should fix your code
1812 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1813 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1814 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1815 We don't support 100% of it, but it should be enough. Contact us if not.
1816 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1817 naive and probably inaccurate)
1818 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1819 mapped onto a unique small file using some system magic.
1820 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1821 - Fortran: cleanups, fixes, support of user-added operations
1822 - RMA: MPI_Accumulate are applied in correct order.
1823 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1824 - Better support for MPI_IN_PLACE
1825 - Support for MPI_Win attrs and keyvals.
1826 - Support MPI_Comm_group_create, MPI_Type_size_x
1827 - MPI ops should be properly applied to all allowed types
1828 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1829 algorithm by a non failing one from mpich.
1832 - Kill the fifo data container: we don't use it anymore.
1835 - A <cluster> can now be created with different speed values to represent pstates.
1836 see examples/platforms/energy_cluster.xml for an example.
1838 SimGrid (3.14.159) Released December 28. 2016
1840 The previous release was only a crude approximation of a Pi release;
1841 this one should be a bit better.
1843 - Revert a last minute change that broke on Mac OSX.
1844 - Fix the changelog and NEWS files.
1845 - Improve the documentation, in particular of the routing module.
1846 - Integrate some patches flying around in Debian.
1848 SimGrid (3.14) Released December 24. 2016
1851 * The whole documentation was reworked and reorganized. There is
1852 still much room for improvement here, but we're on it.
1855 * We now need python3 for our testing infrastructure.
1856 * Model checking needs libevent
1859 * Now works on FreeBSD
1862 * Privatization now works on FreeBSD
1864 * Privatization is now activated by default in smpirun
1865 You can turn it off with -no-privatize if something goes wrong.
1867 * Call-location tracing for SMPI.
1868 You can add the exact location (filename / linenumber) of an MPI call to
1869 your trace files and slow down or speed up the simulation between two
1870 consecutive calls by using an adjustment file (see the documentation).
1872 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1873 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1874 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1875 values were ignored in some cases. The timings of these functions can now
1876 be significantly different.
1878 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1879 smpi/running-power is renamed to smpi/host-speed
1881 * smpi/grow-injected-times option to enable or disable multiplication of the
1882 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1883 make simulation less precise (but also much faster).
1885 * smpirun script should be (much) faster for large deployments.
1887 * SMPI tracing: fixed issue with poor matching of send/receives.
1889 * Replay: Fix broken waitall
1891 New functions and features
1892 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1894 Dropped / renamed functions and features
1895 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1896 - MSG_mailbox_is_empty() -> Mailbox::empty()
1897 - MSG_mailbox_front() -> Mailbox::front()
1898 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1899 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1900 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1901 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1902 * Don't pass the free_f parameter to property related functions:
1903 - MSG_host_set_property_value()
1904 - MSG_as_router_set_property_value()
1905 - MSG_storage_set_property_value()
1906 * VM properties. Since msg_vm_t are msg_host_t, just use
1907 MSG_host_get_property_value() and friends
1908 * VM I/O related things:
1909 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1910 - Unimplemented save/restore methods
1911 * MSG_as_router_get_property_value() was redundant with
1912 MSG_environment_as_get_property_value().
1913 - Removed MSG_as_router_*propert*() functions
1914 - Added MSG_environment_as_set_property_value() for consistency
1915 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1916 Well, they are still there for now, but deprecated with a warning.
1917 Please switch to xbt_dict_new_homogeneous() before this is removed
1919 * Task affinity. Its intended behavior (that was very badly tested
1920 and probably not really working) was deceiving what most users
1921 would have hoped here.
1922 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1923 * Option network/coordinates is now useless and should be dropped.
1926 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1929 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1930 * Added Dragonfly topology support
1933 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1934 EiBps, EBps, Eibps, Ebps for bandwidth.
1935 They may become useful to some lucky ones.
1938 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1939 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1940 * ex/async_waitAll: new example, on asynchronous communications
1943 * Memory usage should be decreased for simulations with a large number
1944 of processes. This also helps for SMPI.
1946 SimGrid (3.13) stable; urgency=low
1948 The Half Release, a.k.a. the Zealous Easter Trim.
1950 - We removed half of the lines, that were mostly experimental cruft.
1951 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1952 (not counting blanks and comments -- according to openhub.net)
1953 - The internals are now compiled in C++ (and will soon be clean C++)
1954 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1955 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1957 * Backwards Compatibility breaks
1958 - Removed Lua simulation bindings (switch to C or Java for that).
1959 Lua can still be used to describe platforms
1960 - Removed Java kernel plug-ins.
1961 Will be reintroduced after the ongoing major internals reorg.
1963 - the following functions were removed.
1964 They were too specific and should be reimplemented in a generic
1965 way, with filter function.
1966 - MSG_task_listen_from_host
1967 - MSG_mailbox_get_count_host_waiting_tasks
1968 - MSG_mailbox_put_with_timeout was removed.
1969 Please use MSG_task_send_with_timeout instead.
1971 - the SD_application_reinit function was removed. It has been a noop for a while.
1972 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1973 replaced by a more flexible mechanism.
1974 - The following functions thus do not exist anymore
1975 - SD_workstation_get_access_mode
1976 - SD_workstation_set_access_mode
1977 - SD_workstation_get_current_task
1978 - Basic estimation functions have been removed but can easily be replaced
1979 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1980 - SD_workstation_get_computation_time => amount / sg_host_speed()
1982 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1983 Use VM.getSpeed()*load/100 for the legacy behavior.
1985 - option enable_tracing was removed. It was not doing anything for a while.
1986 - In the ModelChecker:
1987 - the model-checker now ptraces the model-checked process which means
1988 you cannot use a debugger on the latter anymore (we might make this
1989 optional in the feature);
1990 - removed soft-dirty page tracking;
1991 - remove model-checked side snapshot management,
1992 MC_snapshot() and MC_compare_snapshot();
1993 - keep the MC_cut() function as a stub (it was not really working
1994 in the previous release).
1997 * All options are consistently in kebab-case. Old names are kept as alias.
2000 * Switch to platform v4 format.
2001 - Rename from 'power' to 'speed' the attributes describing the amount of
2002 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
2003 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
2004 - In <host> and <link>, attributes availability and state are gone.
2005 It was redundant with state and availability traces, and with peak values.
2006 - In <cluster>, attributes availability_file and state_file are gone.
2007 It was too complex and unused.
2008 - Kill <gpu>. Was not doing anything.
2009 - The DOCTYPE points to the right URL:
2010 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
2011 (the file at this address now documents the changelog since its v1)
2012 - A warning is emitted for unit-less values (they are still accepted).
2013 - speed. Default: 'f' or 'flops'. Also defined:
2014 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
2015 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
2016 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
2017 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
2018 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
2019 - latency. Default: 's' second. Also defined:
2020 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
2022 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
2023 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
2026 * s4u::Host is now the preferred public interface to the Host features.
2027 sg_host_* functions are C bindings to the exact same behavior
2028 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
2031 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
2032 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
2033 They were deprecated since a few years, and probably did not even compile anymore.
2036 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
2037 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
2038 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
2039 encouraged to update their codes. A list of the modified functions can be found at the end of
2040 include/simgrid/simdag.h
2043 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
2044 the lambda or closure passed as a parameter will run in kernel mode.
2045 All the callback functions should be rewritten to that interface at some point.
2048 * Reorganizing and cleaning the internals all around the place.
2051 * Remove old default barrier/bcast buggy algorithms (see #18407)
2052 * Various bug fixes to handle more codes
2053 * Remove the need for the --foreground option of smpirun (it is still
2054 accepted for backward compatibility).
2057 * Kill the setset data container: MC don't use it anymore.
2058 * Kill the queue data container: it made more sense with GRAS.
2059 * Kill the xbt_peer_t data type: it's useless without GRAS.
2060 * Kill rm_cb feature of config sets: it was never useful.
2061 * Kill graphxml parsing feature. It was not used.
2062 * Kill the deprecated code protected by XBT_USE_DEPRECATED
2064 - xbt_dynar_sort_strings(), when the content is char*
2065 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
2066 They throw exceptions on invalid input;
2067 * C++ support for declaring CLI flags (simgrid::config::Flag);
2068 * class for abstracting different signal backends (simgrid::xbt::signal).
2069 with no external dependencies (we need very simple signals).
2072 * refactoring and cleanup of the code;
2073 * ongoing process to cleanly separate the model-checking algorithms
2074 from the code model-checking support.
2076 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2078 SimGrid (3.12) stable; urgency=low
2080 The Facelift Release.
2083 * Require g++ v4.7 at least to not speak prehistorical C++.
2084 * Require Boost 1.48 (for signal2 component).
2085 * Java must be version 7 at least when activated.
2086 * Builds on Windows again (including Java bindings).
2087 * Tracing is now always enabled (no way to turn it out)
2088 * Remove GTNetS. It was not working anyway.
2089 * Various cleanups in the cmake scripts.
2090 * Move headers around to sort them out on installed systems:
2091 - instr/instr.h -> simgrid/instr.h
2092 - instr/jedule/* -> simgrid/jedule
2093 - simdag/datatypes.h was removed
2094 - simdag/simdag.h -> simgrid/simdag.h
2095 - msg/datatypes.h was removed
2096 - msg/msg.h -> simgrid/msg.h
2099 * Interface improvement:
2100 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
2101 - Sanitize the interface in MSG_task_ module:
2102 - Merge two functions that were close enough but misleading:
2103 set_compute_duration(t) -> set_flops_amount(t)
2104 get_remaining_computation(t) -> get_flops_amount(t)
2105 - set_data_size(t) -> set_bytes_amount(t)
2106 get_data_size(t) -> get_bytes_amount(t)
2107 - Massive cleanups in the functions related to the energy
2108 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
2109 - New: MSG_host_get_pstate()
2111 - msg/energy/onoff: switching hosts on and off
2113 * Interface improvement:
2114 - Rename Host.isAvail() to Host.isOn()
2115 - Rename Process.currentProcess() to Process.getCurrentProcess()
2116 - Rename Task.setDataSize() to Task.setBytesAmount()
2117 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
2119 - #18874: Actually allows the GC to reclaim tasks
2123 - SIMIX_process_throw: raises an exception in a remote process
2124 * Refactoring: Separate sub-modules
2125 - libsmx: the public interface, as libc in a real system
2126 - popping: the strange dance that converts a user request into a kernel handling
2127 - smx_context_*: the virtualization mechanisms that embed the user code
2128 - smx_*: the handling of each simcalls
2129 * Interface cleanup:
2130 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
2131 * Rename smx_action_t into smx_synchro_t, making explicit that these
2132 things are used to synchronize processes with their environment.
2133 For example, a communication is a sort of synchronization involving
2134 the communicating processes (that may block until the exchange) and
2135 the platform. The same can be said from computations, etc.
2137 - #18888: segfault when a process ends before its kill_time
2141 - 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.
2142 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
2143 - MPI_Type_set_name, MPI_Type_get_name
2144 - MPI_*_c2f and MPI_*_f2c functions
2145 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
2146 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
2147 - Activate a lot of new tests from the mpich 3 testsuite
2149 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
2150 - InfiniBand network model added: Based on the works of Jerome Vienne
2151 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
2152 - When smpi/display_timing is set, also display global simulation time and application times
2153 - Have smpirun, smpicc and friends display the SimGrid git hash version on --git-version
2154 * Collective communications
2155 - 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
2156 - 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/).
2157 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
2159 - Replay now uses algorithms from wanted collective selector
2160 - Replay can be used with SMP-aware algorithms
2161 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
2162 - 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.
2164 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
2165 - Correctly use loopback on fat-tree clusters
2166 - Asynchronous small messages shouldn't trigger deadlocks anymore
2167 * Energy/DVFS cleanup and improvement
2168 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
2169 - new: smpi_get_host_pstate()
2172 - "Full" network optimization flag was broken since Surf++
2173 - Better handling of precision flags in maxmin
2174 - Fix bug causing sometimes "Impossible" errors
2175 - Properly pass cluster properties to included hosts
2176 * Improvement of the Energy plugin.
2177 - Always update the consumption before returning that value
2178 - New property: watt_off to denote the dissipation when the host is off
2179 - New functions getWattMinAt and getWattMaxAt to retrieve the
2180 dissipation of pstates that we are not currently at.
2181 * Java: class NetworkLink renamed to Link
2182 * New function: simcall_process_get_kill_time()
2183 * Massive rename s/workstation/host/
2184 - That's intrusive, but that's good for the project consistency. Sorry.
2185 - Change config option "workstation/model" into "host/model"
2188 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
2189 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
2190 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
2191 - xbt_dirname and xbt_basename for non-POSIX systems
2193 * The model checker now runs as a separate process.
2194 * The model checker runs must now be launched with the new simgrid-mc program.
2195 * Record/Replay: the MC can display a textual representation of a path in the
2196 execution graph. It can then be replayed outside of the model checker.
2198 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2200 SimGrid (3.11) stable; urgency=low
2205 * Normalizing pointers addresses tool for better diff between logs
2208 * Add cloud examples using new VMs
2209 - examples/msg/cloud/two_tasks_vm.tesh
2210 - examples/msg/cloud/simple_vm.tesh
2211 - examples/java/cloud/cloud.tesh
2212 - examples/java/cloud/migration/migration.tesh
2213 * Add java surf examples:
2214 - examples/java/surfPlugin/surf_plugin.tesh
2215 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
2216 - examples/java/surfCpuModel/surf_cpu_model.tesh
2217 * Add SMPI+MSG example:
2218 - examples/smpi/smpi_msg_masterslave/
2223 - msg task destroy cancel test
2224 - msg_host on/off test
2225 * Move all tests in testsuite to teshsuite (adding tesh files)
2226 * Restructure teshsuites
2227 - one folder for each kind of test
2228 * Restructure AddTests.cmake
2230 - structure the order of tests (with sections)
2233 * Add virtual machine
2234 - creation of a VM on a PM
2235 - migration of a VM from a PM to another PM
2237 - MSG_process_join(msg_process_t process, double timeout)
2238 - msg_bar_t MSG_barrier_init(unsigned int count)
2239 - int MSG_barrier_wait(msg_bar_t barrier)
2240 - void MSG_barrier_destroy(msg_bar_t barrier)
2241 - msg_as_t MSG_environment_as_get_by_name(const char * name)
2242 * New option "msg/debug_multiple_use" to help debugging when a task is used
2245 - Improvements and finalization of MSG_storage, MSG_file APIs and their
2247 - Increase code coverage in test suites
2251 * Protect context stacks against stack overflow. The number of protected memory
2252 pages allocated on the top of each stack (1 by default) can be configured
2253 with the new command line option --cfg=contexts/guard_size:<value>.
2254 * Simcalls are now generated by a python script that
2255 - generates files included by SimGrid
2256 - checks that all the functions exist, and proposes prototypes
2258 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
2259 host_set_data, host_get_data
2261 - simcall_process_join(smx_process_t process, double timeout)
2262 * Fix bug where sleeping processing could not be suspended.
2265 * Translate surf models from C to C++
2266 - Generic classes for all models: Model, Resource, Action
2267 - A generic interface for each kind of model (CPU, Network, Storage
2268 Workstation, WorkstationVM)
2270 * Translate surf routings from C to C++
2271 * Add callbacks using sigc++ or boost::signals2
2272 - Add callback functions for resource creation/destruction
2273 - Add callback functions for action state change
2274 - Handle Energy as a plugin
2275 * Replace swag by boost::intrusive
2276 * Add new routing models for clusters. For documentation, see
2277 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
2278 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
2279 parameters for cluster tag
2280 - Fat trees, with topology="FAT_TREE" and
2281 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
2282 - see examples/platforms/torus_cluster.xml and
2283 examples/platforms/fat_tree_cluster.xml
2284 * More documentation
2287 * Hostfiles support host:nb_processes construct to deploy several processes on
2289 * Collective communication algorithms should not crash if used with
2290 improper number of nodes and report the error.
2291 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
2292 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
2293 MPI_Dims_create, MPI_Cart_sub are supported.
2294 * New interface to use SMPI programmatically (still depends on MSG for
2295 some parts, see examples/smpi/smpi_msg_masterslave):
2296 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
2300 * Global variables privatization in MPI executables is now performed at runtime
2301 with the option smpi/privatize_global_variables (default:no).
2302 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
2303 dynamic libraries loaded by the application are not privatized (static
2304 linking with these libraries is advised in this case)
2307 * Options defined in XML work correctly now.
2310 * New cmake option, enable_lib_in_jar, to control whether native libraries are
2311 copied into simgrid.jar or not (ON by default). Use this option if you want
2312 to reduce the size of the installed simgrid.jar, *and* the native libraries
2313 are kept installed elsewhere.
2314 * Surf binding with SWIG (code generated in maintainer mode only):
2315 - plugin to handle callbacks
2316 - CPU model only for the moment
2319 * Supernovae build mode is definitively removed. It was used to improve
2320 inlining and inter-module optimizations. It is nowadays superseded by
2321 link-time optimizations commonly available in compilers.
2322 * Update ns-3 find lib. Bindings for ns-3 should work again now.
2323 * Add boost dependency for surf++
2324 * Add new macro for tests
2325 - ADD_TESH(name <tesh_args>)
2326 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
2330 - xbt_bar_t XBT_barrier_init(unsigned int count)
2331 - int XBT_barrier_wait(xbt_bar_t barrier)
2332 - void XBT_barrier_destroy(xbt_bar_t barrier)
2333 * Make the xbt_os_time module public
2335 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2337 SimGrid (3.10) stable; urgency=low
2339 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
2342 * Reintegrate Java to the main archive as desynchronizing these
2343 package is not acceptable anymore (Java is now considered stable)
2344 * Add explicit synchronization facilities through semaphores
2345 * Bug fix: Task.setDataSize() only changed the C world, not the value
2346 cached in the Java world
2349 * Dramatically change the way files are handled. API and internals changed, but
2350 this part of MSG was not considered as production grade either.
2351 * Add explicit synchronization facilities through semaphores
2352 * Add a new function MSG_host_get_process_list()
2353 * Preliminary DVFS support (see examples/msg/energy/ for details)
2356 * SMPI is now included directly in the libsimgrid as the windows
2357 linker doesn't force us on splitting it anymore.
2358 * Improvements of the SMPI replay tool:
2359 - Most of the collective communications are now rooted in the same process as
2360 in the original application.
2361 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
2362 used until now). Multiple data types can now be used in a trace.
2363 - The replay tool now supports traces produce either by TAU or a modified
2365 - Bug Fix: the compute part of the reduce action is now taken into account.
2366 - Gatherv collective is now supported
2367 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
2368 of smpirun outputs time independent traces for the current run. One file
2369 is created per process. If too many processes are simulated, this behavior
2370 can be changed to one file for all processes by using the
2371 tracing/smpi/format/ti_one_file flag
2372 * smpirun generates the host file if needed (with given host count and platform)
2373 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
2374 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
2375 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
2376 --cfg=smpi/coll_selector:(mpich/ompi)
2377 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
2379 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
2380 happening inside a collective SMPI call.
2381 * Fix the behavior of complex data types handling.
2382 * Make MPI_Wtime another synchronization point to take computations into
2384 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
2385 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
2386 * Add all missing Fortran bindings, SMPI should work with Fortran 90
2387 (no privatization of global variables yet)
2388 * Preliminary DVFS support (see examples/smpi/energy/ for details)
2391 * Verification of liveness properties is now available for SMPI applications
2392 (in addition to MSG applications)
2393 * Bugged examples using SMPI in examples/smpi/mc/
2394 * Add --cfg=model-check/visited option. Allows the verification of infinite
2395 programs. Detection of loops in the execution thanks to the system state
2396 comparison and reduction of the state space to explore. Can be combined with
2397 DPOR for safety properties.
2400 * Allow to change SimGrid configuration (see --help) within the code
2401 thanks to SD_config() as it can be done in MSG.
2402 * Add a new function SD_task_set_amount() upon user request.
2405 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
2406 * Remove rule based routing (no more PCRE dependency)
2407 * Add a limiter_link option to cluster tag, to specify a maximum reachable
2408 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
2409 * Add a loopback_bw and loopback_lat options to cluster tag.
2410 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
2411 routing. Now peers are encapsulated in an AS and have their own private
2412 router but this is transparent.
2415 * Our own implementation of getline is renamed xbt_getline, and gets
2416 used even if the OS provide a getline(). This should reduce the
2417 configuration complexity by using the same code on all platforms.
2418 * New type: xbt_cfg_elm_boolean.
2419 * Allow to use yes/no for boolean configuration options in the command line.
2420 * Allow to disable SimGrid cleanups at exit from command line option.
2421 There are situations where one may want a simulation to end with an exit.
2422 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
2423 annoying when scripting around the simulator. Adding a
2424 --cfg=clean_atexit:no allows to circumvent this issue.
2427 * Lots of memory leaks were corrected in this release.
2428 There is no known memory leaks anymore, in all of our 600+ tests.
2429 * New command line option --version, to get SimGrid version information.
2430 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
2432 * Supernovae builds are deprecated, and expected to be removed in the next
2435 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2437 SimGrid (3.9) stable; urgency=low
2439 The Grasgory release: GRAS is really dead now.
2441 * Complete overhaul of the internal host structures scheme.
2444 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
2445 was considered as experimental and badly maintained since then.
2446 * Keeping it was thus a trap to our potential users, that could take
2447 it instead of MSG or SMPI by mistake despite is pity state.
2448 * GRAS seems to have very few users (if any), and no one volunteered
2449 to maintain it further. It also induces a lot of XBT code (for
2450 portability sake), that must be maintained too.
2451 * For all these reasons, we killed GRAS. If someone wants to revive it
2452 in the future, don't cry, our git history still remembers of GRAS.
2455 * Major overhaul. Merge our documentation again as time proved that
2456 splitting it was really not helping our users.
2457 * Further improve the developer documentation to help newcomers
2458 hacking on SimGrid itself. The user documentation (and in
2459 particular, the beginner documentation) is still in a sorry state.
2462 * Now works on Windows too!
2463 * Much more extensive test suite, from MPICH
2466 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
2467 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
2468 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
2469 * Bug fix: let task be scheduled when the last dependency to be solved is
2470 a control dependency.
2471 * Remove SD_load_environment_script function.
2472 Use the C sg_platf function if you want to declare a platform
2476 * New function: MSG_process_get_number()
2477 * Old function documented: MSG_config()
2478 * Remove MSG_load_platform_script function
2479 Use the C sg_platf function if you want to declare a platform
2483 * Change the default value of the TCP_gamma constant (maximal size of TCP
2484 congestion window) to a more realistic 4MiB value. If you notice changes in
2485 your simulation results, you can fall back to the previous 20k tiny window
2486 by adding --cfg=network/TCP_gamma:20000 on command line.
2487 * (Hopefully) fix a bug wrt periodic availability and state traces
2488 * Bug fix: use default values at start when first event in availability/state
2489 trace is not at time 0.
2492 * remove the "new_" part of function name sg_platf_new_trace_connect
2493 (resulting in sg_platf_trace_connect), since it does not create
2497 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2498 use case where it's really mandatory, and maintaining it was a pain in
2500 * New: xbt_fifo_search(), search an item with a user-provided
2501 comparison function instead of dumb pointer comparison.
2504 * Fix the lua deployment:
2505 Use `simgrid.init_application()` before deployment instead of
2506 `simgrid.msg_register_application()` after.
2509 * Transfer the tracing files into the corresponding modules.
2511 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2513 SimGrid (3.8.1) stable; urgency=low
2515 The "we are told that some people want to also *install* the simgrid
2518 * Add missing file "tesh.1" to the archive.
2520 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2522 SimGrid (3.8) stable; urgency=low
2524 The Psssshiiiit release: SimGrid jumps into the Cloud.
2527 * Add an experimental interface to manipulate VMs. They are mainly
2528 process groups with very few intrinsic semantic, but they should
2529 allow you to build the semantic you want easily.
2530 * New function: MSG_host_set_property_value()
2531 * New function: MSG_process_on_exit(). To clean memory in all cases.
2532 * Bug fixes that made the host (and link) failures unusable.
2533 * Add a way to auto-restart process when the host in which they are
2534 executing comes back (ON_FAILURE="RESTART" on deployment file,
2535 MSG_process_auto_restart_set).
2536 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2537 please stop using the old ones, they are DEPRECATED.
2539 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2540 Please use MSG_init() instead. (reducing the amount of entry
2541 points in the library helps us).
2542 * Make it impossible to link against the wrong version of the lib
2543 * Deprecate MSG_clean(). No need to call it anymore.
2544 * Function MSG_get_host_number() is not deprecated anymore.
2547 * Split the doc into a user guide and a reference guide.
2548 * Start a developper guide to help people hacking on SimGrid.
2551 * Enable tracing by default. This modules rocks you should use it.
2552 * Remove option custom_flags. Now use environment variables CFLAGS
2554 * Use default cmake things to detect lua instead of home grown ones.
2555 * New option "enable_mallocators" to disable mallocators, for debugging
2556 purpose ("on" by default).
2559 * Bug fixes around the resource failures: don't let the processes
2560 survive the host they are running onto.
2561 * Add an interface to auto-restart processes when the host in which they are
2562 executing comes back.
2563 * Ensures that SIMIX_clean is called automatically. It's not part of
2564 the public interface anymore (bindings should be updated).
2567 * Bug fix for when SD_Simulate is called with a positive value: be careful
2568 when comparing doubles. Sometimes they are different for non significant
2570 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2571 parallel task whose initial work is distributed among host according
2572 to the Amdahl's law. Such tasks are created with a parameter alpha
2573 that corresponds to the non-parallelizable part of the computation.
2574 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2575 between two sets of workstations assuming a 1D block distribution (each
2576 workstation owns a similar share of data) on both sides.
2578 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2579 Data redistribution will be automatically scheduled once parent and child
2580 are both scheduled. The filling of computation_amount and
2581 communication_amount structures is now done seamlessly thanks to the chosen
2583 * New function SD_workstation_dump to display various information
2584 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2585 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2586 bandwidth on the route onto which the task is scheduled and the amount of
2588 To divide the nominal bandwidth by 2, the rate then has to be:
2589 rate = bandwidth/(2*amount)
2590 * Compute tasks that have failed can now be rescheduled and executed again
2591 (from their beginning)
2592 * Increasing source code coverage (src/simdag is now covered at 95.8%
2596 * Re-implement time-independent trace replay using SMPI (at the
2597 smpi_smp_* level) instead of MSG. This should replace
2598 examples/msg/actions/actions.c
2599 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2600 hindexed and structs)
2601 * Implement the exchange of non-contiguous data.
2602 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2603 * Correct behavior of smpi/sender_gap and set its default value to 0
2604 * Add option to asynchronously send small messages to allow better
2605 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2606 specifies the size in bytes under which messages will be asynchronously sent.
2607 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2608 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2611 * New C interface to define a platform: XML is now optional.
2612 For more info, please check include/simgrid/platf.h.
2613 * New interface to define random platforms from the C:
2614 For more info, please check include/simgrid/platf_generator.h and
2615 examples/msg/masterslave/masterslave_platfgen.c
2616 * Export a sg_cmdline dynar containing all the arguments we got from
2620 * Two new tracing options for adding comments to trace file so you
2621 can track your experiments (see --help-tracing for details).
2622 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2623 * Adding the SimGrid version that generated the trace file as a comment.
2624 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2625 * Fix to avoid key clashes on Paje links
2626 * Other minor fixes related to the Paje specification
2629 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2630 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2631 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2632 * Greatly improve the robustness of mmalloc to user errors (such as
2633 using an area after freeing it, or freeing it twice)
2635 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2637 SimGrid-java (3.8.1) stable; urgency=low
2639 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2640 Warning: all methods are visible, but only some of them are
2641 implemented so far. Check the source (src/jtrace.c)
2642 for further information.
2643 * New module: org.simgrid.msg.File (SimGrid File management functions)
2644 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2645 * Change the meaning of Process.restart: now restart the process from
2646 the beginning, like MSG_process_restart in C.
2647 * Add Process.setAutoRestart: handling of process restart when failed
2649 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2650 you to retrieve the properties of the processes/hosts
2651 * Deprecate Msg.clean(): you can just forget about it now.
2652 * New function Process.getCount(), that only works when compiling
2653 with the not yet released version 3.9 of the C library.
2655 * New context factory based on Coroutines. It mandates a modified JVM
2656 but then, the simulations run about five times faster, and there is
2657 no limit to the amount of processes (beside of the available memory).
2659 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2661 SimGrid (3.7.1) stable; urgency=low
2664 * Restore the prototype of MSG_process_create_with_environment() to
2665 the pre-3.7 situation by removing the kill_time argument.
2666 * Add a MSG_process_set_kill_time() function instead.
2669 * Fix weird behaviors when dealing with parallel tasks.
2672 * Simgrid is now built as a dll.
2673 * Simgrid-java now works on Windows.
2674 * Simgrid-Java is now included into Windows package.
2677 * First pre-build package for MacOSX.
2680 * Fix compilation when using MSG_USE_DEPRECATED.
2681 * Fix some compilation issues on Macs and Windows.
2682 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2683 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2685 We discovered that the Lua console is broken, but we are missing the
2686 manpower to fix it right now. The problem existed in 3.7 too, so we
2687 are not blocking the release for that. Sorry if you depended on this
2688 feature, any help would be really welcome.
2690 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2693 SimGrid-java (3.7.1) stable; urgency=low
2695 The "Java aint got to be bloated and slow" release
2698 * Various internal cleanups and performance improvement
2699 Simulations are expected to run up to twice faster or so
2700 * Make Process.kill(process) an instance method, not a static one
2701 * User processes are not java.lang.Thread subclasses.
2702 This breaks the compatibility (sorry), but previous API was
2703 brain-dead, making it impossible to have non-trivial
2704 initializations in the process constructor.
2705 * Require a full constructor per Process sub-class.
2706 Kinda breaks the compatibility (sorry), but this allows a much more
2707 efficient way to launch the processes at simulation startup.
2708 * Do not embeed our version of semaphores, java 1.5 can be considered
2709 as sufficiently prevalent for us to not dupplicate its features.
2713 * Add examples for almost every part of the API
2714 We spotted and fixed a lot of bugs in the process
2716 * New module: asynchronous communication API
2717 * New function: Process.sleep()
2718 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2719 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2720 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2722 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2724 SimGrid (3.7) stable; urgency=low
2726 The "spring cleanups (before the next Big Project kicks in)" release.
2729 * We can specify the SMPI latency/bandwidth factor with command line
2730 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2731 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2732 You can also use the "config tag" from platform file by adding this line
2733 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2734 (see "example/platforms/tag_config.xml" to use "config tag").
2735 Note that the command line supersedes the platform file configuration.
2736 * Change the correction factors used in LMM model, according to
2737 the latest experiments described in INRIA RR-7821.
2738 Accuracy should be improved this way.
2739 * Use the partial invalidation optimization by default for the
2740 network too. Should produce the exact same results, only faster.
2741 * Major cleanup in surf to merge models and split some optimization
2742 mechanisms from the core of the models. As a result you can now
2743 specify which model to use (e.g., --cfg=network/model:LV08
2744 --cfg=cpu/model:Cas01) and which optimization mode to use
2745 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2746 Incompatible combinations should err at initialization. See
2747 --help-models for the list of all models and optimization modes.
2748 * The CLM03 workstation model was dropped for simplicity because it
2749 used the deprecated CM02 network model. Use default instead.
2750 * Rename the TCP_gamma configuration option to network/TCP_gamma
2751 * Rename the coordinates configuration option to
2752 network/coordinates, and document it
2753 * Use now crosstraffic keyword instead of the terribly misleading
2754 fullduplex keyword. It is activated by default now in the current
2755 default model, use --cfg=network/crosstraffic:0 to turn it off.
2756 * Ongoing refactoring the model parsing to make XML files optional
2757 See include/simgrid/platf.h for details (still to be completed)
2760 * Major overhaul of the documentation. Almost instructive now :/
2761 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2762 functions and friends. This interface was considered as
2763 deprecated since over 2 years, it's time to inform our users that it is.
2764 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2765 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2766 use these (crappy) functions in your code.
2767 These functions will be removed soon. Stop using them now.
2768 * Deprecate MSG_get_host_{table,number}
2769 Implement MSG_hosts_as_dynar() instead.
2770 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2771 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2774 * Stabilize the parallel execution mode of user contexts
2775 * Introduce configuration variables to control parallel execution:
2776 - contexts/synchro: Synchronization mode to use when running
2777 contexts in parallel (either futex, posix or busy_wait)
2778 - contexts/parallel_threshold: Minimal number of user contexts
2779 that must be part of a scheduling round to switch to parallel
2780 execution mode (raw contexts only)
2781 * Fix bugs that prevented to use suspend/resume along with
2782 synchronization structures.
2783 * Fix bugs in process termination that lead to invalid memory access
2784 in very specific conditions.
2787 * Introduce a parallel mode for the models (controlled by surf/nthreads
2788 configuration item). In our tests, running the models in parallel
2789 never lead to any speedups because they are so fast that the gain
2790 of computing each model in parallel does not amortizes the
2791 synchronization costs, even when ultra fast futexes are used.
2792 This is released anyway because YMMV.
2795 * Performance boost by using a swag internally to compute the set of
2796 tasks that are finished and should constitute the return value of
2800 * Enable it by default now that it is considered rather stable.
2803 * Documentation of the tracing functions.
2804 * Performance gains when tracing categorized/uncategorized resource
2805 utilization by avoiding calls to get route when updating resource
2806 variables. LMM constraints are being used instead.
2807 * API changed to set task categories. Use MSG_task_set_category instead
2808 of TRACE_msg_set_task_category, and SD_task_set_category instead
2809 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2810 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2811 * Examples for link user variables added (see at examples/msg/tracing/)
2812 * Deprecated function TRACE_msg_set_process_category completely removed
2813 * Trace header updated according to the latest Paje file format
2814 * Tracing network lazy updates, no longer obligate users to use full updates
2815 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2816 * Experimental: let user code declare/set/push/pop application states for hosts
2817 * API changed to allow the manual creation of graph configuration files
2818 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2821 * Improve the API of Lua MSG bindings, using the Lua spirit.
2822 * Each simulated process now lives in its own Lua world (globals are
2823 automatically duplicated). It helps writing simulators. Will allow
2824 to run Splay programs within SimGrid in the future.
2825 * Add a Chord example in Lua, equivalent to the MSG one.
2828 * Start the implementation of a solution to express temporal
2829 properties, not only local assertions. This is still an
2830 experimental work in progress, stay clear from it to be safe.
2834 - Add new runtime parameters --help-logs and --help-log-categories
2835 to display information about supported logging parameters and
2837 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2838 don't exists anymore.
2839 * Mallocators: allow value NULL for the reset function.
2841 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2842 create homogeneous dictionaries, where all the elements share the
2843 same free function. Non homogeneous dictionaries will be
2844 deprecated in the next release.
2845 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2846 - Multi-level dictionaries are deprecated.
2848 - new function xbt_dynar_search_or_negative() that is useful when
2849 you have less than 2 million elements in your dynar and don't
2850 want of the extra complexity of catching exceptions when the
2851 element is not found.
2853 - Make xbt_os_thread module (for thread portability) public.
2854 Documentation is still to come, sorry.
2856 - Cleanups and simplifications to make it maintainable again.
2857 - Exotic features (such as memalign and valloc) were removed.
2858 - The metadata were extended and improved so that the
2859 model-checker becomes able to explore and inspect the heaps.
2860 - This may induce a performance drop when enable_model-checking is
2861 ON in cmake (even if it's not used in the simulation), but it is
2862 necessary at this point to get MC working.
2864 Turn model-checking OFF if simulation performance matters to you.
2865 Not enabling it at runtime is not enough, disable it in cmake.
2867 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2870 SimGrid (3.6.2) stable; urgency=low
2872 The "Not coding new stuff allows to polish old things" release.
2875 * New bindings to the NS3 packet level simulator (experimental)
2876 * Use the raw (efficient) execution contexts instead of the sysv
2877 (portable) ones when possible.
2878 * libpcre is now mandatory in any cases since not using it led to
2879 severe performance loss and possibly other issues
2880 * Update the XML platforms:
2881 - G5K: include the latest machine in Nancy
2882 - GridPP and LCG: new platforms
2883 * Documentation was partially updated, at least (more to come)
2885 Bug fixes, cosmetics and small improvements
2886 * Free terminated processes before the end of the simulation to avoid
2887 exhausting the memory of users having very dynamic amount of
2889 * Bug fix and cosmetics about canceling non-running tasks
2890 * Bug fix about the dot loader's issues when using libcgraph
2893 * Create an installer for windows with nsis (amd64 and win32)
2894 - Add an hello world project to illustrate SimGrid project creation.
2895 - Embed libpcre into the SimGrid installer to avoid
2896 its compilation burden
2897 * The raw execution contexts should work on Apple now
2898 * Port to Windows 64 bits
2899 - Sysv contexts now have an implementation for this arch
2900 - GRAS communication features now support this arch
2901 * Drop support for borland compiler on windows
2902 - this code was not maintained, and we kinda depend on gcc nowadays
2903 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2904 * Fix portability issue on unstable ubuntu: linker became picky on
2907 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2909 SimGrid (3.6.1) stable; urgency=low
2911 The "Oops, we broke Macs too" release
2914 * Fixed contexts detection so that raw ones are used when possible
2915 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2916 leads to a strange error, with user code segfaulting sometimes when
2917 the generated code is not perfectly aligned (which is not
2918 controllable from the user side, depends on the amount of code)
2921 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2923 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2925 SimGrid-java (3.6) unstable; urgency=low
2928 * Split of every thing from SimGrid v3.5 into a separate package.
2930 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2932 SimGrid (3.6) stable; urgency=medium
2934 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2937 * Bindings now constitute their own package, separated from the main one.
2938 Rationale: reduce our maintenance nightmare by reducing the module coupling
2939 They will soon be released on their own on gforge.
2941 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2942 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2944 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2945 * It's not quite deprecated for now because we have no replacement,
2946 but it may soon become the case.
2949 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2950 * Fortran: New user-level cache variable to store the rank of the running
2951 process. This improves performance by an order of magnitude.
2952 * C: New coccinelle script to automatically locate and modify global and
2953 local static variables.
2954 * Improved SMPI network model with a sender-side gap to account for multiple
2958 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2959 only say if a communication is finished, no matter whether it succeeded or
2960 failed. You can call MSG_comm_get_status() to know the status of a finished
2962 * New function MSG_task_dsend() to send a task and detach it. When a
2963 communication is detached, you are never notified of its success or failure
2964 and the memory is released automatically once it is finished. This function
2965 is useful when you don't care about the end nor the success of a
2967 * Change the prototypes of action replay. Sorry for inconvenience,
2968 but this is really more efficient this way (and to adapt your code,
2969 you just have to fix the initialization, that shouldn't be too long)
2970 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2971 ever managed to do anything useful with it.
2972 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2973 takes a dynar of communications. It returns immediately and gives the
2974 index of a finished communication (if any).
2975 * New example: a basic implementation of the Chord P2P algorithm.
2978 * New model for multi-core CPUs. You can now use the core attribute to
2979 precise the number of cores of a host. This is a basic model. Every
2980 process running on the host receives at most the power provided in
2981 the DTD (throughput<=power). Total throughput of process cannot exceed
2983 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2984 router linked by an up-link and a down-link (possibly asymmetrical).
2985 This kind of pattern allows to easily build last-mile model style platforms.
2986 Aggregating such patterns in a rule-based AS is thus the technique of
2987 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2988 * New model for Vivaldi routing. We transformed the Vivaldi network model
2989 into a Vivaldi routing model (based on the rule-based model). This allows to
2990 combine Vivaldi based latencies with last-mile platforms.
2993 * Added a check for NaN of IEEE754 infinite in the double entries of
2995 * Introduce a new context factory "raw", highly inspirated from the
2996 ucontext factory, but using manually crafted functions in assembly to
2997 do the work in an efficient manner.
2998 * Allow to change the used context factory at run time, not only at
2999 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
3000 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
3001 contexts at runtime (only with raw contexts or ucontexts).
3002 * Completely rewrote this module to allow parallel execution of user
3003 processes. Use --cfg=contexts/nthreads:N to execute user processes
3004 with N parallel threads (the default is 1, meaning no parallelism).
3005 * Allow to decide dynamically between sequential and parallel modes.
3006 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
3007 processes in parallel only when their number is greater than or equal to P
3009 * Added a check for NaN of IEEE754 infinite in the double entries of
3013 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
3014 won't output the state of processes when interrupted with Ctrl-C
3015 * Add a new function xbt_dynar_to_array that transforms a dynar into a
3016 NULL-terminated array. This may solve backward compatibility issues
3017 due to the change to return type of SD_simulate. See also:
3018 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
3019 * Add new macros with variable number of arguments.
3020 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
3021 - in xbt/asserts.h: xbt_assert
3022 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
3023 - in xbt/ex.h: THROWF and RETHROWF.
3024 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
3026 * Change xbt_die() to accept a format string with arguments, just like printf.
3027 * New data structure: xbt_lib_t, like a dict but more general and with better
3031 * New configuration options
3032 Options triva/categorized and triva/uncategorized can be used to generate
3033 graph configuration files for Triva visualization tool.
3034 * Configuration option tracing/platform is renamed to tracing/categorized
3035 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
3036 * New TRACE_user_link_variable interface:
3037 User provides the name of the link and the tracing variable to attach to it
3038 * the declaration of tracing categories must be done after the environment
3040 * simpler tracing interface, just one way to declare categories
3041 TRACE_category or TRACE_category_with_color, it is up to you
3042 * links in the trace file are again identified by their names
3043 * trace contains the full platform hierarchy exactly as declared using the ASes
3044 * Options tracing/msg/[task|process]:1 groups the process by hosts
3045 for both cases, tasks and processes must have names that are unique during the simulation
3046 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
3047 * The experimental option tracing/msg/volume is deprecated
3048 its functionality may be reincorporated if needed
3050 The tracing generates a trace file with unordered timestamped events,
3051 because of the way the core simulator (surf) works. A script available
3052 at the tools directory (fix-paje-trace.sh) can be used to put the events
3053 in order. We have changed the tracing so it can generate ordered timestamped
3054 events in the final trace, but depending on the simulator (and how much time
3055 is simulated) that can lead to a huge memory utilization. It is deactivated
3056 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
3058 Build Infrastructure
3059 * Define a SIMGRID_VERSION macro in simgrid_config.h.
3060 - We are trying hard to keep the API stable, but it may happen that
3061 some things change (we're a research project after all, not a
3062 nuclear plant operating system). If such things should happen, you
3063 could rely on that macro to adapt.
3064 - current value: 30600 for 3.06.00, aka 3.6
3065 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
3066 a number that can be compared with SIMGRID_VERSION.
3067 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
3068 assertions and verbose/debug logging events are disabled at compile time.
3070 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
3072 SimGrid (3.5) stable; urgency=medium
3075 * New feature to any SimGrid-based simulator: Model-Checking
3076 Check SIN#1 for more details.
3079 * New Model SMPI (three-interval linear regression for correction factors)
3080 See RR-7426, available at http://hal.inria.fr/inria-00527150
3081 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
3082 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
3083 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
3084 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
3085 * See also src/smpi/README
3090 - Tracing API changes: TRACE_start and TRACE_end should not be called
3091 by user-code. They are automatically called by simulators created
3092 with SimDAG, MSG and SMPI if the toolkit is compiled with
3093 tracing_enabled=ON. Categories declaration and utilization remain the
3094 same for MSG and SimDag.
3095 - A function was added to the tracing API to declare categories with
3097 - TRACE_category_with_color (char *category, char *color)
3098 where color must be in the following format
3099 "%f %f %f", red, green, blue
3100 and red, green, blue are float values in the interval [0, 1]
3101 - User can specify NULL as color parameter, or continue calling
3102 TRACE_category (cat)
3103 On that case, the tracing system will define random colors
3104 - The following command-line options are supported:
3105 --cfg=tracing/filename:msg.trace
3106 --cfg=tracing:1 (activate tracing, needed to use others)
3107 --cfg=tracing/platform:1 (categorized resource use)
3108 --cfg=tracing/uncategorized:1 (uncategorized resource use)
3109 --cfg=tracing/msg/task:1 (task creation)
3110 --cfg=tracing/msg/process:1 (process creation, migration)
3111 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
3112 --cfg=tracing/smpi:1 (SMPI interface tracing)
3113 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
3114 - examples of examples/msg/tracing updated
3116 - DAXLoader and DOTLoader functions can generate tasks with categories
3117 - A new function to attribute a category to SD tasks:
3118 TRACE_sd_set_task_category (SD_task_t task, char *category)
3119 * Tracing the MPI interface implemented by SMPI
3120 - Collective operations are traced with states
3121 - Point-to-Point operations are traced with states/links
3122 - Tracing activated by a parameter "-trace filename" passed
3123 to smpirun during execution (considering that simgrid
3124 is compiled with tracing enabled)
3125 - To run the simulation with gdb, the simulator
3126 accepts --cfg=tracing/smpi:1 to trace SMPI
3127 - tesh files to check if smpi tracing is ok
3128 - See examples/smpi/NAS/DT-trace
3129 * GTNetS tracing re-worked
3130 - adaptation to the tracing system of GTNets to cope
3131 with modifications regarding the fullduplex mode
3132 - new tesh files to check if gtnets tracing is ok
3135 * Asynchronous communications through the functions:
3136 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
3137 * New function: MSG_load_platform_script()
3138 to make possible using a lua script instead of XML files to set up platforms
3139 * New function: MSG_set_function
3140 to associate functions to processes, used when bypassing the parser
3141 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
3143 Platforms: Add some more examples in examples/platforms
3144 * Grid'5000: see www.grid5000.fr
3145 * *_30000_hosts.xml: various huge files [mainly scalability testing]
3148 * Change the XML format. This is a very important modification. SimGrid
3149 3.5 introduces a new hierarchical format based on the notion of
3150 Autonomous Systems. Compatibility with old format is ensured
3151 through the perl script provided in the install bin directory
3152 bin/simgrid_update_xml.
3153 It is now possible to build platforms with specific routing
3154 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
3155 connect several platforms together. We will try to provide soon
3156 set of realistic platforms exploiting these properties (have a
3157 look at examples/platforms/ for the moment).
3158 * Take the opportunity of the XML format change to be a good XML citizen:
3159 rename link:ctn to link_ctn and similar changes (also dealt with by
3161 * Add a new routing scheme (rule-based) using regular expressions. It
3162 enables to have an extremely low memory footprint when the
3163 underlying routing is simple and can be compactly described. You
3164 need to have libpcre4-dev (perl regular expressions) installed if
3165 you want to use this routing scheme.
3166 * Revive the cluster TAG and allow to easily and efficiently (both in
3167 term of memory and speed) connect clusters together. Have a look
3168 at teshsuite/simdag/platforms/ to see how this can be done. With
3169 this tag, you can create clusters with thousands of tasks at no
3170 cost (have a look at examples/platforms/).
3171 Note: clusters are implemented as ASes, so there is no need for an
3172 enclosing AS tag if you have only one cluster in your platform.
3173 * Add new generic functions in the public interface that allows the user
3174 to call SURF 'create_resource' methods from your code (same
3175 functionality as the XML bypass mechanism but with a much lighter
3177 * Add a new model (enabled through command line --cfg=network/model:SMPI)
3178 that uses a piecewise linear approximation to produce better
3179 results when exchanging small messages.
3180 * Add a new parameter to handle correctly full duplex link and account
3181 for interferences between uplink and downlink communications
3182 (activate with --cfg=fullduplex:1).
3185 * Rename the SD_READY (all dependencies are satisfied and task is
3186 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
3187 dependencies are satisfied) state.
3188 This prevents a confusion between the notion of "ready to schedule"
3189 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
3190 simulated" (SD_RUNNABLE) used by the simulation kernel.
3191 * Change the way a task is considered as ready. Instead of removing
3192 dependencies when a task is done, a counter is decreased. This way,
3193 it is always possible to reach ancestors thanks to the
3194 SD_taks_get_parents function (even after the end of the simulation.)
3195 * Change the return type of SD_Simulate from (SD_task_t*) into
3196 xbt_dynar_t. This function was in handling a dynar internally and
3197 converted it into a NULL terminated array for historical reasons.
3198 * New function SD_dotload(char*) to load a DAG described in dot
3199 format. This loader and the corresponding examples require the
3200 installation of the graphviz library.
3201 * Fix a bug in the management of tasks of size 0 in the surf network
3202 models. This problem was only visible with SIMDAG and you should
3203 thus disregard results produced with earlier versions if you
3204 relied on this feature (some tasks were blocked because of this).
3205 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
3206 even though your applications were doing only point-to-point
3207 communications and sequential computations. Now you can really use any
3208 model you want (of course, if you create real parallel tasks, which are
3209 not implemented in most models beside ptaskL07, this will abort).
3210 * Add an example that schedules a DAX on an heterogeneous platform
3211 using a Min-Min strategy.
3212 * New function SD_workstation_get_current_task() that returns the kind
3213 of task currently running on a workstation in the sequential access
3215 * Raise some warnings when unexecuted tasks remains at the end of the
3216 simulation. This is usually caused by cycles in the DAG.
3219 * New function: SIMIX_process_set_function() called by MSG_set_function
3220 * Change the underlying waiting queue in semaphores so that a process
3221 can wait on several of them simultaneously (as in waitany).
3222 * Fix the way to handle tokens in semaphores so that all access patterns
3223 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
3224 * kill the dirty pimple SIMIX_message_sizes_output()
3225 Please use (proper) visualization instead
3228 * New data container: setset (set of sets of elements)
3229 * New module: mmalloc (mapped malloc, allowing to have several
3230 independent segments of malloc)
3231 * New function: xbt_dict_cursor_set_data()
3232 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
3233 * New function: xbt_dynar_is_empty()
3234 * New function: xbt_fifo_get_last_item()
3235 * Fix xbt_dynar_shrink(): use the right element size.
3236 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
3237 uninitialized areas during expand.
3238 * Fix semaphores: previous implementation was severely broken.
3239 * Use library init/fini functions for our initialization.
3240 - you can use logs and other feature as soon as you want in your
3241 code (even before the xbt_init / MSG_init)
3242 - xbt_exit is now a no-op and produce a warning when used.
3245 * Port GRAS to new SIMIX mechanisms. This allows gras users to
3246 benefit from the latest improvement to the simulation kernel.
3247 * Kill measurement sockets for now. If you rely on them, sorry. This
3248 release is not for you. This feature will be reintroduced in the
3249 future, but we cannot delay the release any further.
3250 * New function: gras_msgtype_get_name().
3251 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
3252 * Fix (at last) the pmm example: it should not randomly fail anymore.
3254 Build chain: bug fixes and overall polishing
3255 * Cmake is now stable enough. Hence, we killed the autotools.
3256 * Port to windows ( TM :)
3257 * Fix the 'make install' target.
3258 No need to use 'make install-simgrid' anymore
3259 * Introduce a 'make dist' target compiling a *source* archive
3260 'make package' compiles a binary archive
3261 * Compile java files only on need
3262 * Add --cd and --setenv command line options to tesh
3263 * Out of source builds are not fully supported yet, but we are close
3264 * Enable supernovae and optimization flags by default for our users
3267 * Add layer to set up environment directly from lua, without XML.
3268 * The effect of gras_stub_generator can be achieved through
3269 lua too (check examples/gras/console/ping_generator.lua)
3271 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
3273 SimGrid (3.4.1) stable; urgency=low
3275 The "Polishing easter eggs is probably a good idea" release.
3276 This is a bug fixes release only.
3280 * Fix a bug preventing the tasks from begin garbage collected.
3283 * Fix a bug occurring when a host involved in a communication fails.
3284 This was not detected properly by the other peer involved in the
3285 communication. Now, it's reported as a network error.
3288 * Warn the user about loop dependencies in data flow of DAX files
3289 * Obey the control-flow dependencies of DAX files
3292 * Add option "enable_smpi" allowing to not compile SMPI.
3293 Probably useful for the (Mac) users experiencing a build error here
3294 * Improve the detection of lua5.1 and ruby1.8
3296 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
3298 SimGrid (3.4) stable; urgency=low
3300 The "Easter in Cargese" release. Also known as (major changes):
3302 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
3303 ~> bindings were greatly improved
3304 ~> new tracing infrastructure for better visualization introduced
3306 * the "Welcome to configury modernity" release.
3307 ~> we switched from autotools to cmake, and improved our cdash
3310 A more detailled list of changes follow (full detail in svn log).
3312 Java Bindings: Various Cleanups
3313 * (install java-gcj-compat-dev on debian-like to use them)
3314 * Remove put/get: no need to export deprecated interface in Java
3315 Use send/receive instead.
3316 * Cleanup the examples and add a README per directory
3317 * Remove example autoDestination (that's the only way to go now)
3318 * Remove example explicitDestination (was a plain copy of basic)
3319 * Make JniException a runtime exception, so that there is no need to
3320 declare the fact that you may encounter such a beast. I guess that
3321 nobody will ever want to survive such error.
3322 * Create specific errors for each MSG case of failure:
3323 host failure, transfer failure, timeout, task cancelled
3324 * Cleanup the exceptions that may get thrown by each function
3325 * Other internal cleanups in Java bindings. Performance still bad :/
3326 Ruby and Lua Bindings: create them
3327 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
3328 * That's new and great, you should try them out.
3329 Same functionalities than Java bindings, only even less polished
3331 * Kill the useless "rate" argument of SD_task_get_execution_time()
3332 Everyone used to provide -1 as a value, it was not used, and the
3333 semantic of a possible use wasn't even clear.
3334 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
3335 either as comm costs or compute costs to mean that there is no
3336 such thing for that specific task.
3337 * Add a SD_task_set_name() function
3338 * Fix SD_task_unschedule() on typed tasks
3339 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
3340 * In DAX loader, accept useless 'level' attributes to <job> since
3341 LIGO DAGs have them (seem to be to ease graphical representation).
3343 * Add an example masterslave_mailbox.c using send/receive and not
3344 the deprecated put/get interface.
3345 * Kill the MSG_paje_output() function. It's a noop since 2 years.
3346 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
3348 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
3349 (declare MSG_USE_DEPRECATED to still have the old name)
3350 * Add a MSG_task_set_data() function
3351 * About trace replay (see examples/msg/actions):
3353 - Allow to work with splitted trace files for each process
3354 Give the specific trace file as argument of each process,
3355 and call MSG_action_trace_run(NULL)
3356 You can still have one merged file for all processes.
3357 - Fix implementation of collective operations
3358 * Allow task_execute() on 0-sized tasks (closes #10063)
3360 * This is the first release of SimGrid where SMPI is not considered
3361 beta anymore (even if some corners should still be improved)
3362 * Port over the new SIMIX_network submodule (internal refactoring)
3363 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
3364 * Implement more missing elements of the standard:
3366 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
3368 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
3369 MPI_Scan() MPI_Get_processor_name()
3370 - Added implementation of missing case for Alltoall (warning: it's
3371 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
3372 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
3373 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
3374 SMPI_MPI_Allgatherv()
3375 * Bug fixes include:
3376 - MPI_Waitsome() was broken
3377 - Allow relative includes in smpicc
3378 - Command line cfg argument 'reference_speed' was ignored...
3379 - Some functions did not properly lead to auto-benching of user code
3380 - smpicc passes -O2 by default (just like openmpi one)
3382 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
3383 * Bug fixes about timeouts during communications
3384 * add SIMIX_message_sizes_output() as a pimple to write to file the
3385 amount of messages per size. Use gnuplot to get histogram.
3386 Pimple because that's the only user-visible function of simix,
3387 defined directly in xbt.h (irk, sorry)
3389 - Add a SIMIX_sem_get_capacity() function
3390 - Fix interactions with processe resume/suspende
3391 - release_forever() was stupidly broken
3392 - Fix SIMIX_display_process_status() for processes in a semaphore
3393 - Make SIMIX_sem_block_onto() user-visible
3394 * Refactoring context stuff:
3395 - Use pseudo-OOP for better modularity
3396 - reimplement SIMIX_process_kill() without process_schedule() so
3397 that the latter can take as invariant that it is called from
3399 - Merge context_start into context_new for sake of simplicity
3401 * Add a Vivaldi network model, coded live during SUD'10 ;)
3402 * Rename configuration variables to start a hierarchy:
3403 o cpu_model -> cpu/model
3404 o network_model -> network/model
3405 o workstation_model -> workstation/model
3406 * New configuration variables:
3407 o network/bandwidth_factor: correction to bandwidth
3408 o network/latency_factor: correction to latency
3409 o netwotk/weight_S: correction to the weight of competing streams
3410 * Add a long description to the models, that users can see with such
3411 argument on the command line: --cfg=cpu/model:help
3412 * --help-models display the long description of all known models
3414 * config: add the ability to set a default value after registration
3415 Does not override any previously set value (e.g. from cmd line)
3416 * dict: allow to have integer key and data.
3417 When so, you need to use the following functions
3418 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
3419 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
3420 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
3421 In contrary to regular dicts, the key is not malloced before copy.
3422 Mixing scalar and regular elements in the same dict is not tested
3424 * Allow to use xbt_dynar_shrink() to expend the dynar instead
3425 Tracing for Visualization:
3426 * SimGrid is now instrumented in order to generate a trace file for
3427 visualization analysis: to use it, need to compile SimGrid with the
3428 "tracing" option enabled, and instrument the program using SimGrid with
3429 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
3430 (among other functions).
3431 * The instrumentation only traces the platform utilization for now
3432 * Documentation to use the tracing functions and how to analyze the
3433 traces with the Triva tool is written.
3434 * More information about: SimGrid FAQ (in the section Tracing Simulations
3437 * We moved to cmake as default build system. Autotools support will
3438 be dropped soon. Check the FAQ for more info about how to use it.
3439 * Greatly improved our cdash/ctest interactions
3440 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
3441 * Added memory checking tests with valgrind; lot of memleak fixing.
3442 This may be the first release of SimGrid with so few memory issues
3443 * Added code coverage tests.
3444 Our coverage is still improvable, but at least we see it on cdash.
3446 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
3448 SimGrid (3.3.4) stable; urgency=low
3450 The "Desktop Grid needs love too" release (also called Xmas release).
3452 Models improvements:
3453 * Major speedup in the maxmin system solving by using lazy evaluation
3454 Instead of solving completely the maxmin system at each iteration,
3455 only invalidate (and recompute) the modified parts.
3456 This new feature is enabled in default models but you can try to
3457 turn it on with "--cfg:maxmin-selective-update=1" for other models.
3458 * Cas01 IMproved as default CPU model
3459 This CPU model is the same Cas01 model, but it uses the
3460 maxmin-selective-update flag and a heap structure to manage
3461 actions on SURF kernel.
3462 It reduces the complexity to find the next action to finish and,
3463 consequently, it's faster than the old Cas01.
3464 This is the new default CPU model (Cas01).
3465 * Rename the old Cas01 model to Cas01_fullupdate
3466 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
3467 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
3468 * CpuTI (CPU Trace Integration)
3469 A new CPU model whose objective is simulate faster when using
3470 availability trace files.
3471 Instead of using a full featured, over engineered maxmin system for
3472 CPU modeling, this model does the pre-integration of traces files
3473 to calculate the amount of CPU power available, and so, executes
3474 faster than the old CPU models.
3475 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
3476 * Use LV08 as default network model since it gives better accuracy
3477 for small messages and shouldn't change things for big ones.
3478 Use --cfg=network_model:CM02 to get the previous behavior.
3481 ******************************************
3482 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
3483 ******************************************
3484 * The new CPU model may changes simulations!
3485 The point is that events occurring at the exact same timestamp
3486 are not scheduled in the same order with the old and new
3487 version. This may be enough to completely change the execution
3488 of simulations in some cases.
3489 * The new network model will change simulations!
3490 This new model is more realistic than the previous one, so you
3491 should consider redoing your old experiments with this model.
3492 Sorry for the inconvenience.
3495 * Introduce the supernovae compilation mode
3496 When compiled that way, the whole SimGrid (or almost) is put in a
3497 single compilation unit and compiled in one shoot.
3498 This is to help gcc which has difficulties to inline stuff from one
3500 The speedup seem to be above 15%, althrough more tests are needed on
3501 amd64 to confirm that gain.
3504 * Port of MSG's mailbox on top of SIMIX network
3505 The put/get mechanism was greatly simplified on the way.
3508 * New SIMIX network module. Provides:
3509 - Mailbox: rendez-vous mechanism to find with who you want to speak
3510 - Synchronous send/recv: easier and hopefully faster since the
3511 logic is handled in the maestro process directly now
3512 - Asynchronous send/recv: you dreamt of it? It's here now
3513 Too bad that nobody cared enough to propagate the change to MSG.
3514 * Add semaphores as SIMIX synchronization mechanism.
3517 * new function SD_daxload(char*) to load a DAX file
3518 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3519 * Introduce typed tasks. Specify its kind and cost at creation.
3520 At scheduling, just give where it should be placed, and the cost
3521 for each involved resource is automatically computed.
3522 Existing constructors so far (more to come of course):
3523 - SD_task_create_comm_e2e() for end-to-end communication
3524 - SD_task_create_comp_seq() for sequential computation
3525 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3526 * new function SD_task_dump() for debugging display
3527 * new function SD_task_dotty(task,FILE*) writing to file the info
3528 about the task in dotty format
3529 * SD_task_dependency_exists() can now cope with having one of its
3530 arguments NULL. If so, it tests whether the other argument has any
3532 * Add getters on list of preceding/following tasks:
3533 SD_task_get_parents(task) and SD_task_get_children(task)
3534 * Add getters on amount of workstations and list:
3535 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3536 * Add getter on task kind: SD_task_get_kind(task)
3537 * Update the start_time and finish_time of tasks on completion/failure
3538 * Bugfix: Remove task from state swags when destroyed
3541 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3544 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3545 dynar with one separate thread per value of the dynar.
3546 * Change the prototype of xbt_thread_create(), sorry.
3547 Added a boolean parameter indicating whether we want to join this
3548 thread (used in SG only for now)
3549 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3552 * GTNetS wrappers should now be usable again (and betterly tested too)
3553 * Fix a major regression from 3.2 where the timeout provided to
3554 MSG_task_put_with_timeout() was used as absolute time before which
3555 the comm should be done.
3556 * Start to fix the <cluster> tag.
3557 - Internal links should be good now (beside of the loopback, which
3558 use the private link instead)
3559 - paths to the external world is still rather broken
3560 - the <route:multi> tag is just broken. Actually that's brain-dead.
3561 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3562 to make it less stupid
3563 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3564 * Fix a source-level compatibility glitch from 3.2: after defining
3565 MSG_USE_DEPRECATED, you can use the old name
3566 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3567 * Allow to compile from the SVN with automake 1.11
3568 * Fix some problems when using the "start_time" tag in deployment XMLs.
3569 * Fix #8569: XBT/synchro.h has redundant declarations
3570 * Fix #8563: MSG return values and exceptions
3571 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3572 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3573 Thanks to Robson Peixoto.
3574 * Fix a vicious bug in dictionaries inducing that some elements were
3575 not freed on xbt_dict_free()
3577 Portability report of this version:
3578 * Main portability targets:
3579 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3580 on (amd64/i386/ia64)
3581 - mac leopard on i386
3582 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3583 but nothing critical.
3584 * Other platforms: windows, AIX and others were not tested for this release
3586 Timing report of this version:
3587 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3588 scenario parameters). From 8h to a few seconds in desktop grid settings.
3589 * Supernovae brings about 25% speedup on i386.
3591 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3593 SimGrid (3.3.3) stable; urgency=low
3595 The "Need for Speed" release.
3597 The timings done to validate the 3.3.2 were faulty.
3598 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3600 The problem was a conversion from a manually handled vector to
3601 xbt_dynar_t on the critical path.
3602 xbt_dynar_foreach calls functions, inducing stack management crap.
3604 We inlined these functions and xbt_dynar_foreach is now breath taking.
3605 We also inlined xbt_swag_belong on the way.
3607 Here are some approximate speedup measurements (on master/slaves
3608 simulations lasting between 10s and 20s each):
3609 3.3.1 -> 3.3.2: about same performance
3610 3.3.2 -> 3.3.3: 40% speedup
3611 3.3.1 -> 3.3.3: 40% speedup
3612 3.3.1 with inline patch -> 3.3.3: 30% speedup
3614 Our reading is that the refactoring which occurred in 3.3.2 made us
3615 suffer much more from the xbt_dynar_foreach low performance, but
3616 once we solved this, this refactoring proved to be very performance
3617 effective. From the 40% speedup, somehow, 10% are due to the
3618 inlining and 30% to the refactoring.
3620 That's a pitty that gcc cannot inline functions placed in other files
3621 alone. We have to choose between:
3622 - break the encapsulation (by putting private data structures and
3623 accessors in headers files to help gcc)
3624 - live with low performance
3625 - switch to a decent compiler such as icc (not quite possible).
3627 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3629 SimGrid (3.3.2) stable; urgency=low
3631 The "Simplicity does not preceed complexity, but follows it" release.
3633 The main contributors of this release were (lexical order):
3634 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3637 * Extract the routing logic into its own object.
3638 (was dupplicated in network.c and workstation_LV07.c;
3639 Allows to implement other ways of storing that info)
3640 => kill now useless network_card concept
3641 - Use dynar to represent routes (instead of void** + int*)
3642 - kill link_set (use surf_network_model->resource_set instead)
3643 - Add a command-line option to choose the routing schema to use
3644 - Add three new models:
3645 * Floyd (shortest path computed at initialization)
3646 * Dijikstra (shortest path recomputed all the time)
3647 * Cached Dijikstra (shortest path computed on need)
3648 All these models where contributed by Silas De Munck, and are
3649 described in his ICCS09 paper.
3651 * Simplify model declaration
3652 (less redirections, less function to write when defining a model)
3653 - Factorize stuff between models:
3656 surf_model_resource_set(model)
3657 surf_model_resource_by_name(model, name)
3658 - Unify the types of models in s_surf_model_t (using an union)
3659 - Embeed fields of common_public directly into s_surf_model_t
3660 - Rename model methods:
3661 action_free ~> action_unref
3662 action_change_state ~> action_state_set
3663 action_get_state ~> action_state_get
3664 - Change model methods into functions:
3665 (model)->common_public->action_use ~> surf_action_ref
3667 * Implement a generic resource; use it as ancestor to specific ones
3668 (allows to kill duplicated code in models)
3669 Drawback: timer command don't need no name nor properties;
3670 workstation_CLM03 don't need no properties
3671 (but I guess we can live with those few bytes wasted)
3673 * Improve the action object model
3674 - implement a constructor avoiding dupplicated code about field
3675 initialization in generic_action part.
3677 * Kill the SDP model: it has an external dependency, is deprecated
3678 in flavor of modern lmm models, and didn't compile since a while
3681 * Relocation of the context module from XBT to SIMIX.
3682 (the context were decoupled from the simix processes, duplicating a lot of code)
3683 => a lot of code was factorized
3684 - less overhead is introduced during scheduling
3685 - simpler API for the context factory
3686 - the logic for process creation,destruction and manipulation was simplified
3687 * Simplification of the s_smx_process_t data structure.
3688 => accesing the simix level data associated to a process is faster now,
3689 and the code is a lot more readable.
3692 * Implement some more MPI primitives:
3693 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3694 -implementation: Bcast: flat or 2-ary tree (default),
3695 Barrier: 4-ary tree,
3697 Allreduce: Reduce then Bcast
3698 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3699 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3700 Alltoallv: flat tree, like ompi
3702 * Add support for optimized collectives (Bcast is now binomial by default)
3703 * Port smpirun and smpicc to OS X
3706 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3707 Shout out if you used it.
3710 * Display the list of still queued messages in SG mode when existing
3714 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3715 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3716 * Add xbt_str_from_file(FILE*)
3717 * Add xbt_dict_get_key achieving a linear reverse search
3718 * Remove the context module
3720 Portability report of this version:
3721 * Main portability targets:
3722 - Linux(debian)/x86/context
3723 - Linux(debian)/x86/pthreads
3724 - Linux(debian)/amd64/context
3725 - Linux(debian)/amd64/pthreads
3726 On these, we still have the eratic breakages of gras/pmm and
3727 amok/saturate_sg reported in previous version. We still think
3728 that the tests are the cause of the fault, not the tested code.
3730 - Mac OSX Leopard/x86/context
3731 Still false negative in tesh autotesting.
3732 Smpi still fails, but this time because readlink does not accept -f
3733 Everything seems to work properly beside of that.
3736 - AIX version 5.3 (only tested contexts this time)
3737 Smpi still fails there because mktemp is not installed.
3738 Everything seems to work properly beside of that.
3740 I managed to compile it for the first time, but several breakages.
3741 Won't delay the release for this exotic platform.
3743 * Windows: it's still lagging behind. If you want to help, please
3746 Timing report of this version:
3747 This version seem to be more than 5% faster than 3.3.1 (on linux
3748 64bits with contexts). The gain is less than expected, we are
3749 investigating this for next release.
3751 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3753 SimGrid (3.3.1) stable; urgency=low
3756 * Implement a --cfg-help to show existing configuration variables
3757 * Build chain do not require doxygen in maintainer mode
3760 * fix a bug on struct sizeof computation, which prevented the
3761 exchange of arrays of structs in some conditions
3762 - added a regression test about this in datadesc_usage
3763 * Allow the exchange of 0-long dynamic vectors.
3764 - for that, use -1 as indicator of dynamic size instead of 0
3765 - This implied to change any size from unsigned long to long,
3766 reducing a bit communication abilities, but I guess that with
3767 64bits being quite common, this is more than enough.
3768 - This also induce a protocol change, thus bumping network protocol
3769 version from 0 to 1 (if we have external users, we have to get
3770 clean on that point too ;)
3771 - added two regression tests about this in datadesc_usage
3772 * Be more verbose when propagating local exceptions
3773 This helps debugging.
3774 * Display the status of simulated processes when receiving SIGINT in
3778 * Allow to control the simulation from a trace file.
3779 New functions MSG_action_register() and MSG_action_trace_run()
3780 The first one allows to associate a function execution to each
3781 kind of action while the second one parses a trace file and
3782 triggers the corresponding actions within the system.
3783 For now, only a toy example is provided in examples/msg/actions
3784 * Add an example of process migration in examples/msg/migration
3785 * Fix a bug in task exchange which broke MSG_task_get_sender()
3786 Add a teshsuite regression test for that.
3787 [Bug: if MSG_task_get_sender() is called after sender exit,
3789 * Fix a bug which prevented suspend/resume to work properly
3790 * Display the status of simulated processes when receiving SIGINT
3791 This fixes a regression of v3.3. due to the introduction of SIMIX
3792 * Bug fixing in failure management:
3793 - trace could not start by a failure at time 0
3794 - failure during communications were not working
3797 * Add SIMIX_process_set_name() to change the name of the current
3798 process in the log messages.
3799 * Store smx_hosts in a dict since we only retrieve them by name
3800 * Move the configuration infrastructure to surf
3803 * Move the configuration infrastructure to surf
3806 * Massive internal cleanups:
3807 - Store internal structures on processes instead of hosts (allows
3808 to have more than one process per host, in addition of being more
3810 - Cleanup the initialization/finalization process
3811 - Kill a whole bunch of unneeded synchronization:
3812 processes run in exclusive manner within the simulator
3813 - Move queues from global tables to process data fields
3815 - now accept -platform and -hostfile arguments
3816 - Pass the right rank value to processes according to the hostfile
3817 * Compile the examples by default, and use them as regression tests
3818 * Implement MPI_Wtime()
3819 * Change the reference speed to a command line option
3822 * TCP_gamma can now be specified as command line option using
3823 --cfg=TCP_gamma:10000000.0
3824 * Change the --surf-path cmd line option into --cfg=path:
3827 * Also include strbuff from xbt.h public header
3828 * xbt_ex_display(): do not free the exception after displaying
3829 This allows to do more with the given exception afterward.
3830 Users should call xbt_ex_free() themselves.
3834 Portability report of this version:
3835 * Main portability targets:
3836 - Linux(debian)/x86/context
3837 - Linux(debian)/x86/pthreads
3838 - Linux(debian)/amd64/context
3839 - Linux(debian)/amd64/pthreads
3840 These targets fail about 1/10 of times on gras/pmm, but we believe
3841 that this is because of the test, not because of SimGrid.
3842 amok/saturate_sg fails even more rarely, and the test may not be
3845 - Mac OSX Leopard/x86/context
3846 The test suite still spits tons of errors because some obscure
3847 force prevents us from removing the temporary directories
3848 arguing that they still contain some metadata I've never heard of.
3849 Smpi fails because seq is not installed.
3850 Everything seems to work properly beside of that.
3853 - AIX version 5.3 (both contexts and pthread)
3854 Smpi still fails there because mktemp is not installed.
3855 XML inclusions seems rosty on AIX.
3857 * Windows: it's still lagging behind. If you want to help, please
3860 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3862 SimGrid (3.3) stable; urgency=high
3866 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3867 [Malek Cherier & Mt]
3869 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3872 Until now, GRAS were using MSG as an interface to SURF. It was
3873 quite difficult because both interface have several differences
3874 (MSG channels vs GRAS sockets were the most notable point).
3876 This also opens the gate to SMPI (which should occur soon) and speed
3877 up simulations by to 40% (even if it were not the main goal).
3879 **************************************
3880 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3881 **************************************
3882 The point is that events occurring at the exact same timestamp are
3883 not scheduled in the same order with the old and new version. This
3884 may be enough to completely change the execution of simulations in
3885 some cases. Sorry for the inconvenience.
3887 * Cleanup and upgrade the XML format to push further scalability
3888 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3890 * Improve the testing infrastructure with tesh. Now a very large part of
3891 the code is tested not only by being run but also by checking that the
3892 output match an expected output [Mt].
3894 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3895 is really less memory-demanding, which should allow you to use
3896 larger files in SimGrid [AL].
3898 * Inform valgrind about our contexts, so that it becomes usable
3899 with the default (and more effecient) version of SimGrid
3900 [contributed by Sékou Diakite, many thanks]
3903 * Introduce a listener thread in charge of receiving incoming
3904 messages from the network. It allows to overlap communication and
3905 computation but most notably, it removes some stupid deadlocks due
3906 to the fact that so far, a process could not send and receive at
3907 the same time. This made most non trivial communication schema
3909 * Convert the PIDs from long int to int to match the MSG ones (and
3910 linux ones too) [Mt]
3911 * New function: gras_agent_spawn() to launch a new process on
3912 current host. Only working in simulation for now. [Mt]
3913 * New function: gras_os_hostport() returning a constant form (ie,
3914 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3917 * Make the backtrace of exceptions more human readable [Mt]
3918 * New module: xbt/str [Mt]
3919 a ton of string utility functions (split, join, printf to a newly
3920 allocated buffer, trim, etc)
3921 * New module: xbt/hash [Mt]
3922 SHA1 hashing algorithm (more to come if needed)
3923 * New module: xbt/synchro [Mt]
3924 synchronization tools (mutex and conditions) working the same way
3925 in simulation and in real life (mainly useful for GRAS, but not
3927 * New module: xbt/queue [Mt]
3928 classical producer/consumer synchronization scheme
3929 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3930 (using the classical functions will get serialized) [Mt]
3931 * Make dictionary internal table dynamic. No need to specify its size
3932 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3934 * Make sure the log channels are organized as a tree under windows
3935 (because of ANSI C compatibility issue, any channel were child of
3936 root directly) [Mt].
3939 * Cleaned many thing in surf and fixed a few bugs [AL].
3940 * Add a nice command line configuration mechanism to compose models [AL].
3941 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3942 the previous one (KCCFLN05). It relies on something that looks like
3943 a max-min sharing mechanism but cannot be written as such. A new solver
3944 was thus designed [AL].
3945 * Add a new solver to lmm. Based on Lagrange optimization and
3946 gradient-based descent, it enables to efficiently maximise systems s.a
3948 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3950 This solver enables to propose two new network models for TCP Reno and
3951 TCP Vegas based on Low's work. These models still need to be fully
3952 tested though [Pedro Velho].
3955 * Bug fix in SD_simulate. Now the time bound given as argument is
3957 * Use the new parallel task model (ptask_L07) as default.
3958 * Use the SURF command line configuration mechanism.
3959 * 0-size tasks (for synchronization) should now work.
3961 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3963 SimGrid (3.2) stable; urgency=high
3967 We still experience issues on this platform, but we believe that at
3968 least MSG is usable.
3970 GRAS API BREAKAGE (for simplification purpose, sorry):
3971 * the gras_msgtype_by_name is not used anymore. Instead of
3972 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3973 you can write (and must)
3974 gras_msg_send(toserver, "request", &request);
3975 - If you still want to pass a gras_msgtype_t to the function (to cache
3976 the type and avoid the lookup time), use the gras_msg_send_() variant.
3977 - Impacted functions:
3978 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3979 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3980 * The callbacks are now expected to return 0 when everything went well
3981 (just like the main() function)
3983 GRAS new features and improvements:
3984 * New module mechanism where user code can use per process globals [Mt]
3985 This is similar to gras_userdata_*() functions, but for libraries. It
3986 factorize some code developped over and over in the examples and AMOK.
3987 It has still to be documented and used (only amok/peermanagement is
3989 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3990 loss when they were small enough to fit into the buffer and sent quickly
3991 enough so that they can all get received in one shoot.
3992 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3993 if not found. Use the *_or_null() variant for the old semantic.
3994 * In gras_msg_handle, do not discard messages without callback.
3995 They are probably messages to be explicitly awaited later (ie, proofs of
3996 mis-synchronization in userland since they are sent before being awaited)
3997 No big deal usually.
3998 * gras_socket_meas_send/recv: semantic changed!
3999 The numerical arguments used to be (1) the total amount of data to send
4000 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
4001 messages. This was need for the fool willing to send more than MAXINT
4002 bytes on quite fat pipes.
4005 * Do really rename the hostmanagement module to peermanagement. [Mt]
4006 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
4007 but this is rather new and this was documented in the module
4008 documentation (poor excuses, I admit)
4009 * Bandwidth measurement semantic changed! This follows the changes to
4010 gras_socket_meas_send/recv explained above.
4013 * A sequential mode has been added to the workstations. When a workstation
4014 is in sequential mode, it can execute only one task, and the other tasks
4015 are waiting in a FIFO. [Christophe Thiery]
4018 * The KCCFLN05 workstation model now handles parallel tasks. It is the
4019 model for SIMDAG. [Christophe Thiery]
4020 * Bug fix in the maxmin solver: Some values were close to 0 instead of
4021 equal to 0, which caused some bad behaviors in
4022 saturated_constraint_set_update. I now use a threshold mechanism like in
4026 * When running manually src/testall, you select specific units [Mt]
4027 testall is the result of our cunit mechanism, and should replace all
4028 the scripty thingy around since bash don't run easily on billware.
4030 * A mallocator system has been added. [Christophe Thiery]
4031 Mallocators allow you to recycle your unused objects instead of freeing them
4032 and allocating new ones.
4034 Documentation update:
4035 * FAQ reworking + New FAQs:
4036 - "Valgrind spits tons of errors!" [Mt]
4037 - "How to repport bugs" [Mt]
4038 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
4039 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
4040 - Communication time measurement within MSG [AL]
4041 - I experience weird communication times when I change the latency [AL]
4042 * GRAS tutorial [Mt]
4044 - an introduction to the framework and to the used communication model
4045 - an initiatic tour introducing the most proheminent features:
4047 . Lesson 0: Installing GRAS
4048 . Lesson 1: Setting up your own project
4049 o Part 2: Message passing
4050 . Lesson 2: Exchanging simple messages
4051 . Lesson 3: Passing arguments to the processes (in SG)
4052 . Lesson 4: Attaching callbacks to messages
4053 . Lesson 5: Using globals in processes
4054 . Lesson 6: Logging information properly
4055 . Lesson 7: Using internal timers
4056 . Lesson 8: Handling errors through exceptions
4057 . Lesson 9: Exchanging simple data
4058 . Lesson 10: Remote Procedure Calling (RPC)
4059 . Lesson 11: Explicitly waiting for messages
4060 . Recapping of message passing features in GRAS
4061 - A HOWTO section containing:
4062 o HOWTO design a GRAS application
4063 More are due, of course. They will come latter. In the meanwhile, you can
4064 check the examples which are still here.
4066 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
4068 SimGrid (3.1) stable; urgency=high
4072 There was a stack corruption somewhere, visible only when optimizing
4073 with these versions. [Vince]
4076 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
4077 module that enabled to play with Directed Acyclic Graphs. It is built
4078 directly on top of SURF and provides an API rather close to the old
4079 SG. Some old codes using SG are currently under rewrite to check that
4080 all needful functions are provided. [Christophe Thiery]
4083 * Complete rewrite of the KCCFLN05 workstation model. It is now an
4084 extension of the classical CLM03 model that gracefully handles
4085 failures. This is now the default model for MSG and GRAS. It doesn't
4086 handle parallel tasks yet though. [AL]
4087 * Bug fix: Weights were not correctly set in the network part.
4088 WARNING: This may have resulted in incorrect results with simulations
4089 where there are more than one flow on a given link. [AL]
4092 * After a (long ?) discussion on simgrid-devel, we have decided that the
4093 convention we had on units was stupid. That is why it has been decided
4094 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
4095 WARNING: This means that all previous platform files will not work as
4096 such with this version! A warning is issued to ask users to update
4098 A conversion script can be found in the contrib module of the CVS, under
4099 the name contrib/platform_generation/surfxml_update.pl [MQ]
4102 * Bug fix: Processes were started in reverse order, wrt deployment file.
4103 WARNING: if your code relies on this bug, please fix it. [AL]
4104 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
4105 being executed on two different locations. [AL]
4106 * Bug fix: Failures are now better supported thanks to Derrick's tests
4107 (there was many failure situations I hadn't thought of and that weren't
4108 correctly handled). [AL]
4109 * New function: MSG_host_is_avail indicates you whether a given m_host_t
4113 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
4114 Exception occurring on server-side are propagated back to client (!).
4116 API CHANGE: the callback changed their prototype. Change:
4117 int my_handler(gras_socket_t expeditor, void *payload_data) {
4119 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
4120 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
4122 * New! function: gras_msg_handleall to deal with all messages arriving
4123 within a given period.
4124 * New! function: gras_socket_server_range to get a server socket in a
4125 range of port numbers (ease to avoid port number conflicts) [MQ]
4126 * New! gras processes display their backtrace when they get a SIGUSR1
4127 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
4128 Sweet to debug RL processes [MQ]
4132 - Do not force experiment sizes to be expressed in kb, or it becomes
4133 impossible to measure the latency this way (needs one byte-long tests)
4134 WARNING: this changes the amok_bw_* function semantic. [MQ]
4135 - Implements the link saturation stuff. [MQ]
4136 * Peer management module:
4137 New! module factorizing code that we wrote over and over [MQ].
4140 * New module: cunit (my jUnit implementation in ansi C) [MQ]
4141 - Test units are placed directly into the library code, they get extracted
4142 automatically and placed into the src/testall binary.
4143 - Convert most of the XBT tests to this system.
4144 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
4145 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
4146 live in the same C code. This required to change a little bit the API
4147 of surfxml parsing but shouldn't be an issue for end-users. [AL]
4148 * New module: sparse graph structure with basic algorithms (this is work
4149 in progress and the API is not considered to be frozen yet). [AL]
4150 * Display more information on backtraces: source line & function names are
4151 now displayed just like valgrind does (rely on addr2line tool) [MQ]
4152 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
4153 * Reworked a little bit some #include statements to load only required
4154 headers. Some user code that relied on SimGrid to include stdlib or
4155 stdio may need to include it by themselves. [AL]
4156 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
4158 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
4159 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
4160 but I guess that almost nobody used those parts. [MQ]
4162 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
4164 SimGrid (3.0.1) stable; urgency=low
4167 * Unfortunately, I had missed 5 misnamed functions:
4168 xbt_fifo_item_t xbt_fifo_newitem(void);
4169 void xbt_fifo_freeitem(xbt_fifo_item_t);
4170 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
4171 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
4172 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
4173 They're now deprecated. Please use their new versions:
4174 xbt_fifo_item_t xbt_fifo_new_item(void);
4175 void xbt_fifo_free_item(xbt_fifo_item_t);
4176 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
4177 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
4178 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
4180 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
4181 * Documentation: xbt_log module unmercifully reworked [MQ]
4182 * Bugfix: there was a problem with the ending of contexts with
4183 the pthread backend. It caused some weird deadlock or behavior
4184 depending on the pthread implementation. [AL]
4185 * Bugfix: get the exceptions raised in the simulator repport where
4186 and why they come from when they are not caught in time [AL, MQ]
4189 * Bugfix: Do repport the error when two non-connected hosts try to
4190 exchange data (Thanks to Flavien for stumbling into this one) [AL]
4193 * Add additionnal checkings on communications. Assert that two
4194 communicating hosts are connected by a set of links... [AL]
4197 * Add additionnal checkings on channel values in communication [AL]
4198 * New: MSG_task_get_source to see on which host a task was generated [HC]
4199 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
4200 the number of tasks waiting to be received on channel and sent
4202 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
4203 waits for the first task coming from a given host.. [AL]
4205 GRAS new functionnalities: [MQ]
4206 * Enhance the parsing macro to allow the size of multidimentional objects
4207 to be given thru annotations.
4208 * New example (and documentation): Matrix Multiplication a la RPC
4209 (as when I was young!) and fix a bunch of bugs found on the way.
4211 GRAS performance improvements: [MQ]
4213 * Reduce the amount of cbps creation/destruction by making it static to
4214 datadesc_send/recv() and using a (newly created) cbps_reset (based on
4217 * Change libdata to a set so that we can search for stuff by ID (and thus
4218 reduce the insane amount of dict lookups)
4221 * Actually implement gras_datadesc_copy() so that we don't have to mimick
4222 RL communication on top of SG since it's so uneffective.
4223 It may also be used for inter-thread communication in RL, one day.
4224 * Use gras_datadesc_copy() to exchange messages on top of SG
4226 - improve message exchange performance on top of SG
4227 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
4228 * Don't exchange on the network the size of the used part of buffer,
4229 instead, specify the possible buffer size to read().
4231 - reduces the amount of read/write calls (one pair per exchange)
4232 - reduces the amount of exchanged data (the size)
4233 - allows to retrieve all arrived data on receiver side, if we don't need
4234 it right now (subsequent read will peek the buffer)
4235 - allows the receiver to proceed with the beginning of the stream before
4236 everything is arrived
4237 - make it possible to build an iov transport (using readv/writev)
4239 - take care of the data with non-stable storage (like stacked data),
4241 * If possible, TCP send uses vector I/O (when writev() is here)
4242 - Don't use it for receive since we send data sizes and data on the
4243 same stream, so we wouldn't be able to chain large amount of chunks
4244 before having to flush the stuff to read the size.
4245 * Rework the transport plugin mechanism to simplify it and reduce the
4246 amount of pointer dereferencement when searching for the right function
4249 * I guess that now, we do almost as few system calls as possible while
4250 doing as few data copy as possible.
4252 To improve it further, we could try to send all the sizes first and then
4253 all the data (to use iov on receiving size), but it's only a partial
4254 solution: when you have 2 dimensional data, the sizes of the second
4255 dimension is data of the first dimension, so you need 3 streams.
4257 I'm not sure the potential performance gains justify the coding burden.
4259 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
4261 SimGrid (3.00) stable; urgency=high
4264 * New! Give the possibility to hijack the surf parser and thus bypass
4265 MSG_create_environment and MSG_launch_application. Have a look at
4266 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
4268 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
4270 SimGrid (2.96) unstable; urgency=low
4275 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
4276 This deprecates the xbt_error_t mechanisms.
4277 It modifies (simplifies) all XBT and GRAS API.
4278 MSG API keeps unchanged (exceptions raised by XBT are caught from
4279 within MSG and masked with existing error handling facilities)
4282 * New! Add a FATPIPE model. [AL]
4283 * New! Add a parallel task model. [AL]
4284 * New! Add automatically a loopback interface (in the default
4285 network model) if none was precised.
4288 * Bugfix: MSG_process_resume now works with the current running process.
4290 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
4291 * Modification of MSG_task_get_compute_duration. Once a task has been
4292 processed, the value returned by this function is now equal to 0. [AL]
4293 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
4294 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
4295 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
4296 of a m_task. For now, MSG_task_cancel only works with computation
4298 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
4299 of the processor (in Mflop/s) regardless of the current load on the
4301 * API Change: use proper naming convention for MSG_getClock and
4302 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
4304 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
4305 This function changes the priority of a computation task. This priority
4306 doesn't affect the transfer rate. A priority of 2 will make a task
4307 receive two times more cpu power than the other ones. This function
4308 has been added to suit the needs of Nguyen The Loc and hasn't been that
4309 much tested yet. So if it fails, please report it and send me your code.
4311 * API Change: removed all functions and types that were marked "deprecated"
4312 since many months. Renamed MSG_global_init_args to MSG_global_init.
4314 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
4316 SimGrid (2.95) unstable; urgency=low
4319 * Steal some nice code to GNU pth to fix context detection and usage [AL]
4320 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
4321 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
4324 * Bugfix: Allow absolute paths for platform description files [MQ]
4325 * Bugfix: do free the variables after use. Leads to drastic performance
4327 * Implement max_duration (ie, timeouts) on resources [AL]
4330 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
4332 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
4333 the needed support (timeouts) [AL]
4336 * Implement measurement sockets. You can now get the bandwidth between two
4337 hosts thanks to AMOK (see below). [MQ]
4338 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
4339 dynar over the network (yeah) [MQ]
4340 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
4341 * simple/static token-ring example. [Alexandre Colucci and MQ]
4342 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
4343 in place before [MQ]
4346 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
4347 type parsing. "short" and "long" modifiers were also ignored; other
4348 modifier (such as reference level) are still ignored. [MQ]
4349 * Embeed the buffer size within the buffer itself on SG. [MQ]
4350 That way, send() are atomic and cannot get intermixed anymore (at least
4351 the ones which are less than 100k; bigger messages still have the issue)
4352 * Array size pushed by the field, not by the field type (or each
4353 and every long int will push stuff to the cbps) [MQ]
4354 * use select() to sleep since it allows to portably sleep less than one
4357 GRAS (minor cleanups)
4358 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
4361 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
4362 * type category 'ignored' killed as it was never used and were difficult
4364 * whether a type can cycle or not is now a flag, leaving room for more
4365 flags (as "ignored", if we feel the need one day ;)
4366 * Rename raw sockets to measurement sockets since "raw" has another
4367 meaning in networking community.
4370 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
4371 and offers features not belonging to GRAS but that most applications
4372 need. One day, it may be a set of plugins loadable at runtime.
4373 * New module: bandwidth
4374 bandwidth measurement between arbitrary nodes running this module. [MQ]
4376 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
4378 SimGrid (2.94) unstable; urgency=low
4380 The first beta release of SimGrid 3 !
4384 * Update the main page and the FAQ. Adding references to gforge.
4387 * Add a gras_os_getpid function.
4390 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
4391 * Extend the logs so that they also print PID, hostname, date, ... if
4393 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
4394 and kill the old version which were in testsuite/
4395 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
4399 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
4400 for users. Instead, each programming environment (GRAS, MSG,...) use it
4401 when needed to provide such a feature to users.
4402 Don't shortcut the mechanism or you will also shortcut the virtualization
4403 you need on the simulator.
4407 * Cleanups in configury with regard to compile optimization/warning flags.
4408 Also add -fno-loop-optimize to any powerpc since it's the optimization
4409 killing gcc (< 3.4.0).
4410 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
4411 needed by MSG examples complications
4412 * Borrow configury beautifications from PHP
4415 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
4416 warning (thanks loris for stumbling into this one).
4417 * Bugfix: stop loading private headers (gras_config.h) from the public
4421 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
4422 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
4423 RL would imply a DNS resolver, which is *hard* to do in a portable way
4424 (and therefore delayed).
4425 * Implement a real timer mechanism and use it in timing macros. This allows
4426 to avoid rounding errors and get a 0.000005 sec precision in timing
4427 macros. While I was at it, various cleanups:
4428 - allow to declare more than one timed section per file (fix a stupid bug)
4429 - move some private declaration to the right place
4430 - merge conditional execution and timing macros into emulation module
4431 - document the module
4432 - make sure the module cleanups its mess on gras_exit
4433 * Documentation improvements:
4434 - (new) how to compile applications using GRAS
4435 - (new) emulation support (timing macros)
4437 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
4439 SimGrid (2.93) unstable; urgency=low
4441 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
4444 - Use Paje properly where used. Still to be sanitized properly.
4445 - Portability fix: Add an implementation of the contexts using pthread
4449 - Add xbt_procname(): returns the name of the current process.
4450 Use it to show the current process's name in all logging.
4452 - fix detection of older flex version and the reaction, since we do
4453 depend on modern ones (we use lex_destroy)
4454 - Better separation of SG and RL in the libs: remove all simulation code
4455 from libgras. As a result, this lib is now only 200k when stripped.
4456 Some of the xbt modules may also be duplicated (two sets and such) and
4457 should be cleaned/killed before SG3.
4458 - Insist on using xlC on AIX because of weird problems involving gcc there.
4459 - Cleanup the make remote stuff. This is now done by scripts
4460 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
4461 mainly for our private use, but we're working on changing them to user
4464 - Bugfix: flush the socket on close only if there is some *output*.
4465 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
4467 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
4470 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
4472 SimGrid (2.92) unstable; urgency=low
4474 Alpha 3 on the path to SimGrid 3
4478 - New! First try of benchmarking macros.
4479 - New! First try so that gras_stub_generator generate deployment and
4480 remote compilation helpers.
4482 - Bugfix: Initialization fix in msg_test.
4486 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
4488 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
4490 - API Change: gras_os_sleep to take the amount of seconds as a double.
4491 Accepting an int was error prone since it was the only location where
4492 seconds were coded as such. It leaded to damn rounding errors.
4493 - Bugfix: Hard to belive that timers ever worked before this.
4495 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4497 SimGrid (2.91) unstable; urgency=low
4499 Alpha 2 on the path to SimGrid 3
4503 - Bug fix in the lmm_solver.
4505 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4506 through the function MSG_paje_output.
4507 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4508 - It is possible to bound the rate of a communication in MSG with
4509 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4510 to write it in the changelog).
4511 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4516 - Various cleanups to the autotools stuff
4517 - Begin to move Gras examples to examples/gras/
4518 - Let make distcheck work again (yeah!)
4520 - documentation overhauled using doxygen.
4521 gtk-doc-tools is dead in SimGrid now.
4522 - Automatically extract all existing logging categories, and add the list
4523 to the documentation (long standing one, to say the less)
4525 - Cleanup the known architecture table. Reorder the entries to group what
4526 should be, and use a more consistent naming scheme.
4527 (some of the test dataset are still to be regenerated)
4528 - New! Allow library to register globals on each process just as userdata
4530 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4531 lookup time (for now).
4532 Use it in msg and trp.
4533 This cleans a lot the internals and helps enforcing privacy of the
4534 headers between the gras components.
4535 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4536 - Bugfix: gras_os_time was delirious in RL.
4537 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4539 - Reenable GRAS now that it works.
4541 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4543 SimGrid (2.90) unstable; urgency=low
4545 Alpha 1 on the path to SimGrid 3
4547 * It is a long time since the last release of SimGrid. I'm sorry about
4548 that but as I had told you, I was rewriting a lot of things. I apologize
4549 to those who had been reporting bugs to me and that I had not answered.
4550 If your bug is still in the new version, please tell me. Here is a
4551 summary of the main changes.
4553 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4554 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4555 portability, speed, and a lot more but you'll figure it out later.
4556 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4557 I wanted to release the new MSG as soon as possible and I have
4558 broken GRAS, which is the reason why, for now, only MSG is fully
4559 functional. A laconic description of these projects is available
4560 in the documentation.
4562 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4563 kernel optimized for our needs (called SURF but only the developers
4564 should use it). Hence, MSG is now roughly 30 times faster and I think
4565 that by rewriting a little bit MSG, I could event speed it up a little
4566 bit more. Beside the gain in speed, it is also much easier to encode a
4567 new platform model with SURF than it was with SG. More to come...
4569 * REVOLUTION 3: I have tried to change a little as possible the API of
4570 MSG but a few things really had to disappear. The main differences
4571 with the previous version are:
4572 1) no more m_links_t and the corresponding functions. Platforms are
4573 directly read from an XML description and cannot be hard-coded
4574 anymore. The same format is used for application deployment
4575 description. The new format is described in the documentation.
4576 Have a look in tools/platform_generation. There is a tiny script
4577 that converts from the old platform format to the new one. Concerning
4578 the application deployment format, parsing the old one is tricky.
4579 I think most of you should however be able to convert your files. If
4580 it is really an issue, I can write a C code that does the conversion.
4582 2) the toolbox tbx does not exist anymore. We now have a library
4583 with much more data-structures but without the hash-tables (we have
4584 dictionaries that are much faster).
4586 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4588 *****************************************************************************
4589 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4590 * don't feel like converting the oldies. (Mt) *
4591 *****************************************************************************
4594 Version 2.90: "the long awaited one"
4595 - Finished rewriting and debugging MSG. Rewrote the documentation.
4596 - disable GRAS for now since it needs to be ported to the newest SG
4599 - Finish the port to windows (using mingw32 for cross-compile)
4602 - Main loop and datastructures of SURF. A cpu resource object is
4603 functional. Surf can thus be used to create cpu's with variable
4604 performance on which you can execute some actions.
4606 2004-11-15 Martin Quinson
4607 - Port to ARM. Simply added the alignment and size descriptions. Should
4608 work, but the ARM machines are so slow that I didn't had the opportunity
4609 to 'make check' over there yet.
4611 2004-11-15 Arnaud Legrand
4612 - Trace manager now written. It uses a heap structure and is therefore
4613 expected to be efficient. It may however be speeded up (particularly
4614 when many events occur at the same date) by using red and black
4615 trees. One day maybe...
4616 - Max-min linear system solver written. It uses a sparse matrix
4617 structure taking advantage of its expected use. Most operations are
4618 O(1) and free/calloc are called as few as possible. The computation of
4619 the minimum could however be improved by using a red and black tree
4622 2004-11-03 Arnaud Legrand
4623 - Rename every gras_* function that was in xbt/ to its xbt_
4625 - Add a heap and a doubly-linked list to xbt
4626 - Added a dichotomy to the dictionaries. make check works as well before
4627 so I assume that the patch is correct. I do not know however if things
4628 run effectively faster than before now. :)
4630 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4631 SimGrid, and the version number is bumped to 2.x
4633 2004-10-29 Martin Quinson
4634 - Introduction of the remote errors.
4635 They are the result of a RMI/RPC on the remote machine.
4636 ErrCodes being scalar values, you can't get the host on which those
4637 errors did happen. Extending the error mechanism as in Gnome is possible.
4638 No idea yet whether it is a good idea.
4640 2004-10-28 Martin Quinson
4641 - Interface revolution: the Starred Structure Eradication.
4642 I used to do typedef struct {} toto_t; and then handle *toto_t.
4643 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4644 - ???_t is a valid type (builded with typedef)
4645 - s_toto_t is a structure (access to fields with .)
4646 - s_toto is a structure needing 'struct' keyword to be used
4647 - e_toto_t is an enum
4648 - toto_t is an 'object' (struct*)
4650 typedef struct s_toto {} s_toto_t, *toto_t;
4651 typedef enum {} e_toto_t;
4652 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4653 s_toto_t) is private.
4655 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4656 so that it can set the variable to NULL. It was so for dicts and sets,
4657 it changed for dynars.
4659 - Fix a bunch of memleaks in dict_remove
4660 - Fix a bug in sg/server_socket opening: it failed all the time.
4662 2004-10-07 Martin Quinson
4663 - Speed up dynar lookup operation a bit.
4665 gras_dynar_get is dead.
4667 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4668 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4669 (which gives you the address of the stored data).
4671 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4672 copy of the data using an affectation to do the job and not a memcpy.
4674 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4675 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4677 It does not really speedup the dynar test because they are
4678 setting elements all the time (and look them seldom). But the dict does
4679 far more lookup than setting.
4681 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4683 2004-10-05 Martin Quinson
4684 - Allow to (en/dis)able the cycle detection at run time.
4686 Whether we should check for cycle or not is now a property of each
4687 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4688 datadesc_cycle_unset allow to remove this property when previously set.
4690 Note that the cycle detection is off by default since it impacts the
4691 performance. Watch the data you feed GRAS with ;)
4693 This property is hereditary. Any element embedded in a structure having it
4694 set have it set for the time of this data exchange.
4696 You should set it both on sender and receiver side. If you don't set it on
4697 sender side, it will enter an endless loop. If you forget on receiver
4698 side, the cycles won't be recreated after communication.
4700 - Header reorganization.
4701 Kill gras_private.h, each submodule must load the headers it needs.
4703 2004-10-04 Martin Quinson
4704 - Interface revolution: do not try to survive to malloc failure.
4706 Now, gras_malloc and friends call gras_abort() on failure.
4707 As a conclusion, malloc_error is not a valid error anymore, and all
4708 functions for which it was the only gras_error_t return value are
4709 changed. They now return void, or there result directly.
4710 This simplify the API a lot.
4712 2004-09-29 Martin Quinson
4713 - Re-enable raw sockets.
4714 Created by gras_socket_{client,server}_ext;
4715 Used with gras_raw_{send,recv}
4718 It should allow to kill the last bits of gras first version soon.
4720 This is not completely satisfactory yet (duplicate code with
4721 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4724 - Simplify transport plugin (internal) interface by not passing any
4725 argument to _server and _client, but embedding them in the socket
4728 2004-09-28 Martin Quinson
4729 - Finish the port to AIX.
4730 autoconf was my problem (segfault within the malloc replacement
4731 function. No idea why)
4733 2004-09-16 Martin Quinson
4734 - Fix some size_t madness on 64bit architectures.
4736 2004-09-08 Martin Quinson
4737 - Reduce the number of system headers loaded, overload some more system
4738 calls (such as malloc to cast the result of the system one, and work
4740 - Fix and reintroduce the config support
4742 2004-09-07 Martin Quinson
4743 - Source code reorganization to allow Arnaud to surf all over there.
4744 - Allow to document the logging categories.
4745 - Remove all uppercase from logging categories and useless cleanup in names.
4747 2004-08-18 Martin Quinson
4748 Version 0.6.2 (protocol not changed; API changed)
4749 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4750 gras_error_t), and NULL when not found. Functions expecting a msgtype
4751 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4752 hopefully usefull message.
4753 - Portability to prehistoric sparcs again
4755 2004-08-17 Martin Quinson
4756 Version 0.6.1 (protocol not changed; ABI not changed)
4757 - prealloc some buffers to speed things up
4759 2004-08-11 Martin Quinson
4760 Version 0.6 (protocol not changed; ABI expended)
4761 - The parsing macro can deal with the references, provided that you add
4762 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4764 2004-08-09 Martin Quinson
4765 Version 0.5 (protocol not changed; ABI changed)
4766 - Allow to off turn the cycle detection code in data exchange at
4767 compilation time. It should be at run time, but I'm short of time (and
4768 the config stuff is still broken). That way, we keep dict out of the
4769 critical path, which is good because the performance is poor:
4770 - search not dichotomial yet
4771 - dynar give no way to access their content and memcpy everytime
4772 - In composed data description (struct, ref and so on), stop foolness of
4773 keeping the subtype's ID, but store the type itself. This keeps sets out
4774 of the critical path, which is good since they rely on dynar and
4775 dictionaries. The only loose of that is that we cannot detect the
4776 redeclaration of a structure/union with another content (but I'm not sure
4777 the code detected well this error before anyway). We still can detect
4778 the redefinition discrepancy for the other types.
4779 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4780 it breaks the code because of type-punning used all over the place).
4781 This breaks on all non-gcc architectures (for now).
4783 All those changes (plus the buffer of last time) allow me to gain 2 order
4784 of magnitude on cruel tests consisting of 800000 array of integers on two
4785 level of a hierarchical structure (200 secondes -> 4 secondes)
4788 - the selector of reference must now return the type it points to, not
4789 the ID of this type.
4791 2004-08-06 Martin Quinson
4792 Version 0.4 (protocol changed; ABI not changed)
4793 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4794 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4795 - Implement a buffer transport, and use it by default (it relies on tcp in
4796 real life and on sg in simulation).
4797 That's a bit hackish since I had a new field to the structure to store
4798 its data without interfering with the subtype ones. Inheritance
4799 is tricky in C. And that's a kind of reverse inheritance with one class
4800 derivating two classes. Or maybe a game with java interfaces. Anyway,
4801 that's damn hard in C (at least).
4802 Moreover, I got tired while trying to ensure plugin separation and
4803 genericity in SG mode. MSG wants me to do weird things, so let's go for
4804 cruel hacks (temporarily of course ;).
4805 See comment in transport_private.h:71
4806 - do not include all the _interface headers in private but in the files
4807 which really need them (to cut the compilation time when they are
4810 2004-07-26 Martin Quinson
4811 Version 0.3 (protocol not changed; ABI changed)
4812 - Major overhault of the datadesc interface to simplify it:
4813 - shorted the function names:
4814 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4815 - add a trivial way to push/pop integers into the cbps without malloc.
4816 This allows to make really generic sub_type description, which simply
4817 pop their size of the stack.
4818 - add a function gras_datadesc_ref_pop_arr() which does what users want
4819 most of the time: Declare a dynamic array (which pops its size of the
4820 stack) and declare a reference to it. Poor name, but anyway.
4821 - kill the post-send callback, add a post-receive one
4823 2004-07-23 Martin Quinson
4824 Version 0.2 (protocol changed; ABI changed)
4825 - add some testing for cpbs in the test cases, and fix some more bugs.
4826 This invalidate again the little64 data file, since I cannot regenerate
4828 - remove an awfull optimization in the logging stuff, allowing me to:
4829 - understand it again
4830 - learn gcc how to check that the argument match the provided format
4831 - fix all errors revealed by gcc after that
4832 - internal keys of dict are not \0 terminated. Deal with it properly in
4833 loggings instead of segfaulting when the user want to see the logs :-/
4835 2004-07-22 Martin Quinson
4836 - Fix some stupid bug preventing cbps (callback postit) from working
4838 2004-07-21 Martin Quinson
4839 - Some documentation cleanups
4840 - remove the useless last argument of msgtype_declare
4841 - rename the Virtu functions to fit into the 'os' namespace
4842 - move headers src/include -> src/include/gras/ and stop fooling with
4843 gras -> . symbolic link
4844 - make distcheck is now successful
4846 2004-07-19 Martin Quinson
4848 - Build shared library also
4849 - Install html doc to the right location
4850 - stop removing maintainer files in make clean
4851 - build tests only on make check
4853 2004-07-13 Martin Quinson
4855 - No major issue in previous version => change versioning schema
4856 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4857 corresponding dataset.
4859 2004-07-11 Martin Quinson
4861 - Get it working with any kind of structure (we can compute the padding
4862 bytes remotely for all the architectures I have access to)
4863 - Implement the structure parsing macro (still not quite robust/complete)
4864 - Improvement to the remote testing toysuite
4866 2004-07-10 Martin Quinson
4867 [autoconf mechanism]
4868 - get ride of a bunch of deprecated macros
4869 - actually run the test for two-compliment, not only compile it :-/
4870 - test whether the structures get packed (and bail out if yes. Damn.
4871 Alignment is a serious matter)
4872 - test whether the structures get compacted (but respecting the alignment
4873 constraints of each types)
4874 - test whether the array fields of structures can straddle alignment boundaries
4876 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4877 (took me 2 hours to find that bug, looking at the wrong place)
4878 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4879 before sending/receiving and are used to compute the offsets of fields
4880 - Given that padding size depend even on compiler options, keep track of
4881 alignment and aligned_size only for the current architecture. Not a big
4882 deal since we send structure fields one after the other (seems
4884 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4885 program, let it work on linux/gcc/little32. portability todo.
4887 2004-07-08 Martin Quinson
4888 - import and improve remote compilation support from FAST
4889 - make sure make check works on half a dozen of machines out there
4891 2004-07-07 Martin Quinson
4892 Let's say it's version 0.0.3 ;)
4893 - Implement conversions (yuhu!)
4894 - Let it work on solaris (beside conversion, of course)
4895 - Stupid me, using rand() to generate the conversion datatests in not wise.
4897 2004-07-06 Martin Quinson
4898 - Let make dist work, since I'm gonna need it to compile on remote hosts
4899 - Let Tests/datadesc_usage write the architecture on which the file was
4900 generated as first byte.
4901 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4902 architecture descriptions.
4903 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4904 on those architectures.
4905 - Optimization: send/recv array of scalar in one shoot
4907 2004-07-05 Martin Quinson
4908 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4910 - Plug a whole bunch of memleaks
4911 - each process now have to call gras_{init,exit}. One day, their log
4912 settings will be separated
4913 - Continue the code factorisation between SG, RL and common in Transport.
4915 2004-07-04 Martin Quinson
4917 - Redistribution between SG and RL.
4918 We wanna have to accept in SG, so move accepted related parts of RL in
4919 the common part. (more precisely, the dynar of all known sockets is no
4920 more a static in transport.c, but part of the process_data)
4922 [gras_stub_generator]
4923 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4925 2004-07-03 Martin Quinson
4926 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4927 them all up in one shot)
4929 - Fix the ugly case of reference to dynamic array.
4930 - New (semi-public) function gras_datadesc_size to allow the messaging
4931 layer to malloc the needed space for the buffer.
4933 - gras_socket_close now expect the socket to close (and not its address to
4934 put NULL in it after it). This is because the socket passed to handlers
4935 is one of their argument (=> not writable).
4937 - propagate the interface cleanup from last week in datadesc, ie remove a
4938 superfluous level of indirection. User pass adress of variable
4939 containing data (both when sending and receiving), and not of a variable
4940 being a pointer to the data. Let's say that I like it better ;)
4941 The price for that is constructs like "int msg=*(int*)payload" in
4942 handlers, but it's a fine price, IMHO.
4944 - Let it work in RL (yuhu)
4946 2004-06-21 Martin Quinson
4948 - porting SG plugin and SG select to new standards (works almost).
4949 - plug memleaks and fix bugs around.
4952 - cleanup the prototype of data recv and force users to specify when they
4953 want to handle references to objects. Test case working even for cycles.
4954 - plug memleaks. Valgrind is perfectly ok with this.
4956 2004-06-12 Martin Quinson
4958 - cleanup the separation between plugin and main code in plugin creation
4960 2004-06-11 Martin Quinson
4962 - Reput hook for raw sockets, needed for BW experiments
4963 - kill a few lines of dead code
4964 [Data description] Interface cleanup
4965 - gras_datadesc_by_name returns the searched type or NULL.
4966 That way, no variable is needed to use a type desc once, which makes
4968 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4969 two parameters were strings (field name, type name), leading to
4971 [Dicos] Interface cleanup
4972 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4973 This is consistant with the dynar API.
4975 2004-04-21 Martin Quinson
4977 - Porting to new standards.
4979 - interface cleanup.
4980 There is no bag anymore, no need to take extra provision to mask the
4981 pointers behind "ID".
4982 Better splitup of functions between files create/exchange/convert.
4983 This is still a bit artificial since convert and receive are so
4984 interleaved, but anyway.
4986 - add a queued message list to procdata (the ones not matching criteria
4988 - factorize some more code between SG and RL wrt procdata
4990 - use gras_exit in example to track memleaks
4991 - get rid of gs_example now that GS is properly integrated into gras
4992 - update run_test to integrate the latest tests (datadesc)
4994 - rename WARNINGn macros to WARNn since it prooved error-prone
4996 2004-04-19 Martin Quinson
4998 - register init/exit functions within gras module mechanism
4999 - send/receive function.
5000 Convertion is not implemented, but short-cutted if not needed.
5001 struct/array elements are sent one by one (instead of block-wise), but
5002 nobody really cares (yet). Get a prototype before optimizing.
5003 - tests (using a file socket) for DD send/receive on:
5004 - base types: int, float
5005 - array: fixed size, string (ie ref to dynamic string)
5006 - structure: homogeneous, heterogeneous
5007 - chained list, graph with cycle
5008 Believe it or not, valgrind is not too unhappy with the results. The
5009 cycle happily segfaults, but the others are ok. And I'm sick of pointers
5013 - Bugfix when using a filename explicitly (instead of '-')
5015 2004-04-09 Martin Quinson
5017 - factorize more code between RL and SG in socket creation
5018 - Complete the implementation and tests of:
5020 o file (only in RL, and mainly for debugging)
5022 I lost 3 days to design a portable address resolver, and then decided
5023 that the prototype mainly have to run on my box.
5024 Addressing portability too early may be like optimizing too early :-/
5026 - use gras_init in the Tests instead of the crappy parse_log_opt
5027 (the latter function is removed)
5028 [Conditional execution]
5029 - New functions: gras_if_RL/gras_if_SG (basic support for this)
5030 [Code reorganisation]
5031 - Get rid of libgrasutils.a since it makes more trouble than it solves.
5032 Build examples against the RL library, since there is no way to disable
5033 its creation for now.
5035 For information, the beginning of coding on GRAS was back in june
5036 2003. I guess that every line has been rewritten at least twice since