1 SimGrid (3.34.1) not released (Target: fall 2023)
4 - New class ActivitySet to ease wait_any()/test_any()/wait_all()
5 - Deprecate {Comm,Io,Exec}::{wait_any,wait_all,test_any} and friends
6 - Simplify a bit the declaration of multi-zoned platforms from C++
7 - New function NetZone::add_route(host1, host2, links) when you don't need gateways
8 - Also add a variant with s4u::Link, when you don't want to specify the directions
10 - Zone's gateways can now be controlled directly.
11 - Add NetZone::add_route(zone1, zone2, links) specifying the route between zones
12 - Introduce a Mailbox::get_async() with no payload parameter. You can use the new
13 Comm::get_payload() once the communication is over to retrieve the payload.
14 - Implement recursive mutexes. Simply pass true to the constructor to get one.
15 - Simplify the expression of horizontal scaling of Tasks.
16 - Each Task now consists of a dispatcher, a collector and one or more instances.
17 - The parallelism degree of each of these can be set.
18 - Several examples have been added or modified accordingly.
19 - Update s4u::create_DAG_from_json() to support wfformat 1.4.
20 - Introduce a new MessageQueue abstraction and associated Mess simulated object.
21 The behavior of a MessageQueue is similar to that of a Mailbox, but intended for
22 control messages that do not incur any simulated cost. Information is automagically
23 transported over thin air between producer and consumer. See examples/cpp/mess-wait
24 - New function: Mutex::get_owner()
27 - New: Add a JBOD (just a bunch of disks) concept. It's a sort of host with many disks.
28 - Revamp the battery plugin: rewrite completely the API, for a better usability.
29 The examples were updated accordingly.
30 The battery can now act as a simple connector (see battery-connector example).
31 - Revamp of the photovoltaic plugin: now called SolarPanel and complete rewrite of the API
32 - Add chiller plugin: enable the management of chillers consuming electrical energy
33 to compensate heat generated by hosts.
34 - Add a battery-chiller-solar example combining several plugins to evaluate the amount
35 of brown energy (from the electrical grid) and green energy (from the solar panel)
36 during a given computation.
39 - New SMPI_app_instance_join(): wait for the completion of a started MPI instance
40 - MPI_UNIVERSE_SIZE now initialized to the total amount of hosts in the platform
41 - Memory usage due to SMPI for non-MPI actors greatly reduced.
42 - New implemented calls: MPI_Isendrecv, MPI_Isendrecv_replace
45 - Allow to use on valgrind-observed or gdb-observed processes.
46 - Install sthread on user's disk.
47 - Implement recursive pthreads.
48 - Implement pthread_barrier and pthread_cond (but conditional are not supported by the MC yet).
49 - Add some McMini codes to test sthread further (controlled with enable_testsuite_McMini).
52 - Remove stateful model-checking. This was not used, not really working, and very hard to fix.
53 Liveness properties cannot be verified anymore.
54 - More informative backtraces on assertion failure.
55 - Fix dependency bugs for mutex and other transitions
56 - Fix some reversible_race definitions, and also the rest of ODPOR.
59 - Make the host_load plugin available from Python. See examples/python/plugin-host-load
60 - Mailbox::get_async() does not return a pair anymore. Use comm.get_payload() instead.
61 - Comm::waitall/waitany/testany() are gone. Please use ActivitySet() instead.
62 - Comm::waitallfor() is gone too. Its semantic was unclear on timeout anyway.
63 - Io::waitany() and waitanyfor() are gone. Please use ActivitySet() instead.
64 - Do not export the values of enums. So you need to write e.g. SharingPolicy.LINEAR
65 while it should have been possible to write LINEAR alone before. This is the advised
66 behavior for modern C++ code.
69 - Introduce sg_activity_set_t and deprecate wait_all/wait_any/test_any for
73 - optimize an internal data structure (replace boost::circular_buffer_space_optimized by
74 std::deque to store pending and unmatched Comms in Mailboxes). It is actually a revert
75 to what was used a few years back. The boost structure had a lower memory footprint than
76 deques, but it appeared that their "space_optimized" character was generating a huge lot
77 of refcount changes on the stored Comms.
80 - Fix errors with ns-3 v3.36+
81 - Many other small bug fixes, in particular in MC and sthread.
83 ----------------------------------------------------------------------------
85 SimGrid (3.34) June 26. 2023
87 Save the planet, skip a release: 3.33 was due 6 months ago, so skip directly to 3.34.
90 - SimGrid now requires a compiler with C++17 support for public headers too.
91 Sibling projects should upgrade their FindSimGrid.cmake
92 - Remove the MSG API: its EOL was scheduled for 2020.
93 - Remove the Java bindings: they were limited to the MSG interface.
94 - On Windows, you now need to install WSL2 as the native builds are now disabled.
95 It was not really working anyway.
96 - Support for 32bits architecture is not tested anymore on our CI infrastructure.
97 It may break in the future, but we think that nobody's using SimGrid on 32 bits.
98 - Remove the surf module. It was replaced by the kernel/models module, and that
99 refactoring took almost 10 years to properly complete.
102 - Activity::set_remaining() is not public anymore. Use for example
103 Comm::set_payload_size() to change the size of the simulated data.
104 - New function: Engine::flatify_platform(), to get a fully detailed vision of the
106 - New Task abstraction: They are designed to represent dataflows, i.e, graphs of repeatable Activities.
107 See the examples under examples/cpp/task-* and the associated documentation.
108 - Full simDAG integration: Activity::start() actually starts only when all dependencies
109 are fulfilled. If it cannot be started right away, it will start as soon as it becomes
111 - Allow to set a concurrency limit on disks and hosts, as it was already the case for links.
112 - Rename Link::get_usage() to Link::get_load() for consistency with Host::
113 - Every signal now come with a static version that is invoked for every object of that class,
114 and an instance version that is invoked for this specific object only. For example,
115 s4u::Actor::on_suspend_cb() adds a callback that is invoked for the suspend of any actor while
116 s4u::Actor::on_this_suspend_cb() adds a callback for this specific actor only.
117 - Activity::on_suspended_cb() is renamed to Activity::on_suspend_cb(), and fired right before the suspend.
118 - Activity::on_resumed_cb() is renamed to Activity::on_resume_cb(), and fired right before the resume.
119 - Resource::on_state_change_cb() is renamed to Resource::on_onoff_cb() to distinguish from the
120 Activity::on_state_change_cb() that is related to the activity state machine, not on/off.
121 - Activity signals (veto, suspend, resume, completion) are now specialized by activity class.
122 That is, callbacks registered in Exec::on_suspend_cb will not be fired for Comms nor Ios.
125 - Battery: Enable the management of batteries on hosts.
126 See the examples under examples/cpp/battery-* and the documentation in the Plugins page.
127 - Photovoltaic: Enable the management of photovoltaic panels on hosts.
128 See the examples under examples/cpp/photovoltaic-* and the documentation in the Plugins page.
131 - optimize an internal data structure (use a set instead of a list for ongoing activities),
132 leading to a potentially big performance gain, in particular with many detached comms.
135 - New option smpi/barrier-collectives to add a barrier to some collectives
136 to detect dangerous code that /may/ work on some MPI implems.
137 - New function SMPI_app_instance_start() to easily start a MPI instance in your S4U simulation.
140 - Write the section of the manual about models, at least.
141 - WiFi: the total capacity of a link depends on the amount of flows on that link.
142 - Use the nonlinear callback feature of LMM to reflect this.
143 - Calibration values can be changed to match different MCS configurations
144 - See the example teshsuite/models/wifi_usage_decay/wifi_usage_decay.cpp
145 - See also "A Flow-Level Wi-Fi Model for Large Scale Network Simulation"
146 https://hal.archives-ouvertes.fr/hal-03777726
147 - Merge parameters network/bandwidth-factor and smpi/bw-factor that serve the same purpose.
148 - Same for the latency
149 - Rewrite the corresponding documentation.
150 - Allow to disable the TCP windowing modeling by setting network/TCP-gamma to 0.
151 - Finally kill the 'compound' host model. You can change the CPU or network model
152 with the default host model, as it should.
153 - Rename option "surf/precision" to "precision/timing" for clarity.
154 - Rename option "maxmin/precision" to "precision/work-amount" for clarity.
155 - New function: Engine::flatify_platform() to debug your platform.
158 - Implement pthread_join in MC mode.
159 - Implement semaphore functions in sthread.
160 - Add an intricated way to verify the access to non-reentrant data structures
161 It requires code annotation, as shown in examples/sthread/stdobject/stdobject.cpp
164 - Stateless model-checking is now usable on any system, including Mac OSX and ARM processors.
165 - The stateless aspects of the MC are now enabled by default in all SimGrid builds.
166 Liveness and stateful aspects are still controlled by the enabling_model-checking
167 configuration option.
168 - Introducing ODPOR and SDPOR reduction strategies
169 - Introducing guiding heuristics, trying to find bugs faster than DFS in reduced state space.
170 - Synchronize the MBI tests with upstream.
171 - Show the full actor backtraces when replaying a MC trace (with model-check/replay)
172 and the status of all actors on deadlocks in MC mode.
175 - simgrid::xbt::cmdline and simgrid::xbt::binary_name are gone.
176 Please use simgrid::s4u::Engine::get_cmdline() instead.
179 - New tutorial on simulating DAGs.
180 - New section in the user guide on the provided performance models.
181 - New section presenting some technical good practices for (potential) contributors.
182 - Add a section on errors and exceptions to the API documentation.
183 - Move the s4u examples to a section on their own to ease navigation.
185 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
186 (FG: issues on Framagit; GH: issues on GitHub)
187 - FG#18: Java bindings should be redone or removed
188 - FG!118: Wi-Fi callback mechanism
189 - FG!119: SMPI: add option to inject a barrier before every collective call
190 - GH#383: Segfault when adding a disk after load_platform(xml)
192 ----------------------------------------------------------------------------
194 SimGrid (3.32) October 3. 2022.
196 The Wiedervereinigung release. Germany was reunited 32 years ago.
199 - SimGrid now requires a compiler with C++17 support to compile the lib.
200 Our public headers still allow the user code to be compiled in C++14.
201 - Support graphviz v3 and ns-3 v3.36 (older versions are still supported).
202 - Tested with clang (v11, v13, v14 and v16), gcc (v7 to v13) and IntelCC v2022.2
206 - kill signal Comm::on_completion that was not working anyway.
207 - Expose signals Activity::on_suspend and Activity::on_resume
208 - New macro xbt_enforce(): similar to xbt_assert(), but throws an AssertionError
209 instead of calling abort().
210 - New: s4u::Exec::get_thread_count()
211 - Various cleanups around virtual machines:
212 - host_by_name() and friends now only return hosts. VMs are now excluded.
213 - It is now impossible to search a VM by name globally.
214 You can only search VM by name on a given PM, so either you know
215 the PM on which your VM runs and you can search by name, or you need
216 to manually iterate over all PMs to search this VM.
217 - The s4u::VirtualMachine constructor is now deprecated.
218 Please use s4u::Host::create_vm() instead.
219 - Rename s4u::VirtualMachine::on_creation() to on_vm_creation() to
220 avoid confusion with s4u:Host::on_creation() that is inherited.
221 Also s4u::VirtualMachine::on_destruction -> on_vm_destruction().
223 - One-sided communications (Comm::sendto) can now be detached,
224 and should now be more resilient to network and host faults.
227 - Added the following bindings / examples:
228 - Comm (now 100% covers the C++ interface):
229 - Comm.dst_data_size, Comm.mailbox, Comm.sender, Comm.start_time, Comm.finish_time
230 - Comm.state_str [examples: examples/python/comm-failure/, examples/python/comm-host2host/]
231 - Comm.remaining [examples: examples/python/comm-host2host/, examples/python/comm-suspend/]
232 - Comm.set_payload_size [example: examples/python/comm-host2host/]
233 - Comm.set_rate [example: examples/python/comm-throttling/]
234 - Comm.sendto, Comm.sendto_init, Comm.sendto_async [example: examples/python/comm-host2host/]
235 - Comm.start, Comm.suspend, Comm.resume [example: examples/python/comm-host2host/]
236 - Comm.test_any [example: examples/python/comm-testany/]
237 - Comm.wait_until [example: examples/python/comm-waituntil/]
239 - Engine.host_by_name [example: examples/python/comm-host2host/]
240 - Engine.mailbox_by_name_or_create [example: examples/python/comm-pingpong/]
242 - Mailbox: Mailbox.ready [example: examples/python/comm-ready/]
243 - Ptask [example: examples/python/exec-ptask/]:
244 - this_actor.exec_init
245 - this_actor.parallel_execute
248 - Added an AssertionError exception that may be thrown in case of error.
249 For instance, creating two hosts with the same name will now throw this exception
250 instead of killing the interpreter.
253 - Implement MPI_File_get_type_extent(), MPI_File_s/get_atomicity() and
254 MPI_File_get_byte_offset()
255 - Intercept getpid() calls to return the simulated ones.
256 - Fix various bugs in MPI IO.
258 Platform description & visualization:
259 - More robust sanity checks for platforms, to reject forbidden topologies with
260 a proper error message.
261 - New platform example: supernode.cpp and supernode.py.
262 The Python version generates a nice graphical representation of the platform.
263 - Bug fixes around fat-tree topologies.
264 - Allow to dump the platform topology as a CSV file representing the graph edges
265 with platform_graph_export_csv() (similar to the DOT export).
266 - Fix graphicator for "cluster" topologies (e.g. fat-tree, dragonfly).
269 - Fix a bug when using ptasks with multicores (FG!111).
272 - First bits of sthread, that intercepts pthread operations at runtime.
273 The intend is to use it together with simgrid-mc, but it is TBD.
274 - Sync MBI generators with upstream changes.
275 - Various cosmetics, small bug fixes and inner refactorings
277 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
278 (FG: issues on Framagit; GH: issues on GitHub)
279 - FG#105: "Variable penalty should not be negative!" with in-flight messages and bandwidth profiles
280 - FG#109: Application time reported by --cfg=smpi/display-timing:yes is wrong
281 - FG#110: Wait_any does not trigger new model solve when host events occur
282 - FG#111: Wrong execution time in rare cases when using multicore
283 - FG!98: Re-enable the tests for legacy stochastic profiles
284 - FG!109: Trigger new engine solve upon host events such as host on/off
285 - FG!116: SMPI/replay: Fix issue with recv of size =0
287 ----------------------------------------------------------------------------
289 SimGrid (3.31) March 22. 2022.
291 The ненасильство release. We stand against war.
293 Against the agression by a sick system that forces peoples to take arms against each other.
296 - Rework the internals, for simpler and modern code. This shall unlock many future improvements.
297 - You can now define plugins onto the DFS explorer (previously called SafetyChecker), using the
298 declared signals. See CommunicationDeterminism for an example.
299 - Support mutex, semaphore and barrier in DPOR reduction
300 - Seems to work on Arm64 architectures too.
301 - Display a nice error message when ptrace is not usable.
302 - New test suite, imported from the MPI Bugs Initiative (MBI). Not all MBI generators are integrated yet.
303 - Remove the ISP test suite: it's not free software, and it's superseeded by MBI.
306 - fix for FG#100 by ensuring small asynchronous messages never overtake larger
307 ones, conforming to the standard.
308 - replay: fix waitall behaviour to avoid forgetting requests and leaking
310 - tracing: ensure that we dump the TI traces continuously during execution and
311 not just at the end, reducing memory cost and performance hit.
312 - Update OpenMPI collectives selection logic to match current one (4.1.2)
313 - Add a coherence check for collective operation order and root/MPI_Op
314 coherence. Potentially costly so not activated unless smpi:pendantic is set
315 or -analyze is given.
318 - New signal: Engine::on_simulation_start_cb()
319 - Introduce a new execution mode with this_actor::thread_execute(). This simulate
320 the execution of a certain amount of flops by multiple threads ran by a host. Each
321 thread executes the same number of flops, given as argument. An example of this new
322 function can be found in examples/cpp/exec-threads.
323 - Reimplementation of barriers natively.
324 Previously, they were implemented on top of s4u::Mutex and s4u::ConditionVariable.
325 The new version should be faster (and can be used in the model-checker).
326 - Actor::get_restart_count(): Returns the number of reboots that this actor did.
329 - MSG_barrier_destroy now expects a non-const msg_barrier parameter.
331 New plugin: the Chaos Monkey (killing actors at any time)
332 - Along with the new simgrid-monkey script, it tests whether your simulation
333 resists resource failures at any possible timestamp in your simulation.
334 - It is mostly intended to test the SimGrid core in extreme conditions,
335 but some users may find it interesting too.
338 - New solver for parallel task: BMF.
339 - More realistic sharing of heterogeneous resources compared to the fair
340 bottleneck solver used by ptask_L07.
341 - Implement the BMF (Bottleneck max fairness) fairness.
342 - Improved resource sharing for parallel tasks with sub-flows (parallel
343 communications between same source and destination inside the ptask).
345 - "--cfg=host/model:ptask_L07 --cfg=host/solver:bmf": enable the ptask
346 model with BMF solver.
347 - "--cfg=bmf/max-iterations: <N>": maximum number of iterations performed
348 by BMF solver (default: 1000).
349 - "--cfg=bmf/precision: <N>": numerical precision used when computing
350 resource sharing (default: 1e-12).
351 - This model requires Eigen3 library. Make sure Eigen3 is installed to use BMF.
354 - Modifications of the Profile mechanism, with some impact on users
355 - Addition of a new (S4U) method to init profiles from generic functions to improve versatility
356 - Fix initial behaviour of state_profiles
357 - Modify periodicity to behave like a period, and not like a loop delay
360 - Drop xbt_dynar_shrink().
363 - Made the following bindings static (previously member functions):
364 - Actor: Actor.kill_all(), Actor.by_pid()
365 - Host: Host.by_name(), Host.current(), Host.on_creation_cb()
366 - Mailbox: Mailbox.by_name()
367 - Added the following bindings:
368 - this_actor.warning()
369 - Mailbox.put_init() [example: examples/python/comm-waitallfor/]
370 - Comm.detach() [example: examples/python/comm-waitallfor/]
371 - Comm.wait_for() [example: examples/python/comm-waitfor/]
372 - Comm.wait_any_for()
373 - Comm.wait_all_for() [example: examples/python/comm-waitallfor/]
374 - Mutex [example: examples/python/synchro-mutex/]
375 - Barrier [example: examples/python/synchro-barrier/]
376 - Semaphore [example: examples/python/synchro-semaphore/]
379 - Remove target "make uninstall" which was incomplete and no longer maintained.
381 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
382 (FG: issues on Framagit; GH: issues on GitHub)
383 - FG#57: Mc SimGrid should test whether ptrace is usable
384 - FG#87: Smpi scripts fail with spaces in paths
385 - FG#100: [SMPI] Order of the message matching is not guaranteed
386 - FG#101: LGPL 2.1 is deprecated license
387 - FG#104: "make uninstall" not up-to-date
388 - GH#151: Missing mutexes for DPOR.
390 ----------------------------------------------------------------------------
392 SimGrid (3.30) January 30. 2022.
394 The Sunday Bloody Sunday release.
396 Main user-visible changes:
397 - The SimDag API for the simulation of the scheduling of Directed Acyclic
398 Graphs has been dropped. It was marked as deprecated for a couple of years.
399 We finally complete the implementation of what has been called SimDag++
400 internally, i.e., porting the different features of SimDag on top of S4U.
401 The new way to simulate the execution of dependent activities directly by
402 maestro (without any other actor) is details in the examples/cpp/dag-* series
404 - The removal of SimDag led us to also remove the export to Jedule files that
405 was tightly coupled to SimDag. The instrumentation of DAG simulation is still
406 possible through the regular instrumentation API based on the Paje format.
407 - We also dropped the old and clumsy Lua bindings to create platforms in a
408 programmatic way. It can be done in C++ in a much cleaner way now, which
409 motivates this suppression.
412 - Introduce on_X_cb() functions for all signals, to attach a new
413 callback to the signal X. The signal variables are now hidden and
414 only these functions should be used.
415 Rationale: this enables the usual deprecation scheme where functions
416 remain for 4 releases if we need to modify the signals, while the
417 current code with the signal variables directly visible prevents any
419 - New function: Engine::run_until(date), to split the simulation.
420 - New signal: Activity::on_veto, to detect when an activity fails to start.
421 - Signal change: Comm::on_start(Comm&, bool) has been replaced by
422 Comm::on_send and Comm::recv. These two signals respectively correspond to
423 when the sending or receiving side of a Comm is ready. They are raised at
424 the same locations as the former Comm::on_start signal.
425 - New function: Engine::track_vetoed_activities() to interrupt run()
426 when an activity fails to start, and to keep track of such activities.
427 Please see the corresponding example for more info.
428 - New functions: s4u::Comm::{sendto_init, set_source, set_destination} to enable
429 the use of vetoers with direct host-to-host communications. Both source and
430 destination have to set for a comm to start. Each call to these setters check
431 if all vetoes are satisfied. When it is the case, the comm starts. A use case of
432 these functions is given in examples/cpp/dag-scheduling.
433 - New functions: {Exec, Io}::update_priority allow you to modify the priority of
434 these kinds of activities during their execution. Behavior is detailed in
435 examples/cpp/io-priority/.
438 - Dynamic costs for MPI operations: New API to allow users to dynamically
439 change injected costs for MPI_Recv, MPI_Send and MPI_Isend operations.
440 Alternative for smpi/or, smpi/os and smpi/ois configuration options.
441 - Fix some issues with the replay mechanism.
444 - Function xbt::Extendable::get_data() is now templated with the type of the
445 pointee. Untyped function is deprecated. Use get_data<void>() if you still
446 want to retrieve void*.
449 - New section: "SimGrid MPI calibration of a Grid5000 cluster"
450 presenting how to properly calibrate MPI communications in SimGrid.
451 - Complete and reword the platform section, which is now completed.
454 - Thread contexts are used by default with Python bindings. Other kinds of
455 contexts revealed unstable, specially starting with pybind11 v2.8.0.
457 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
458 (FG: issues on Framagit; GH: issues on GitHub)
459 - FG#95: Wrong computation time for multicore execution after pstate change
460 - FG#97: Wrong computation time for ptask+multicore+pstates
461 - FG#98: SMPI offline simulation is inconsistent with the online simulation
462 (deadlocks / message truncation)
463 - FG#99: Weird segfault when not sealing an host
465 ----------------------------------------------------------------------------
467 SimGrid (3.29) October 7. 2021
469 The "Ask a stupid question" release.
471 We wish that every user ask one question about SimGrid to celebrate.
472 On Mattermost, Stack Overflow or using the issues tracker.
474 New modeling features:
475 - Non-linear resource sharing, modeling resources whose performance heavily degrades with contention:
476 - The total capacity may be updated dynamically through a callback
477 and depends mainly on the number of concurrent flows.
478 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
480 - Dynamic factors: model variability in the speed of activities
481 - Each action can now have a factor that affects its progression.
482 This multiplicative factor is applied when updating the amount of work
483 remaining, thereby an activity with factor=0.5 only uses half of the
484 instantaneous power/bandwidth it is allocated and will appear twice
485 slower than what it actually consumes.
486 - This can be used to model a overhead (e.g., there is a 20 bytes
487 header in a 480 bytes TCP packet so the factor 0.9583) but the novelty
488 is this factor can now easily be adjusted depending on activity's and
489 resources characteristics.
490 - This existed for network (e.g., the effective bandwidth depends
491 on the message in SMPI piecewise-linear network model) but it is now
492 more general (the factor may depend on the source and destination and
493 thus account to different behaviors for intra-node communications and
494 extra-node communications) and is available for CPUs (e.g., if you
495 want to model an affinity as in the "Unrelated Machines" problem in
496 scheduling) and disks (e.g., if you want to model a stochastic
498 - For that, resources can be provided with a callback that computes
499 the activity factor when creating the action.
500 - Example: examples/cpp/exec-cpu-factors
501 - The same mechanism is also available for the latency, which
502 allows to easily introduce complex variability patterns.
505 - Added support to programmatic platform creation in Python.
506 Example: examples/python/clusters-multicpu
509 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
510 This can only be set through the API, not through XML files.
514 - Multiple fixes to ensure reproducibility of tracing
515 - scan/excan can now be replayed
516 - wait action now uses ranks and not pid, as the other ones.
517 - smpi/init and smpi/finalization-barrier are now valid for replays.
518 - exit() is now intercepted by SMPI to avoid premature shutdown of
519 simulation. First non 0 return codes is returned as simulation return
523 * New section "Release Notes" documenting recent and current developments.
524 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
525 * Improvements in API Reference for C++ and Python interfaces.
528 - Make wifi creation compatible with ns-3 version 3.34 too.
530 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
531 (FG: issues on Framagit; GH: issues on GitHub)
532 - FG#77: Search feature of doc is broken (update sphinx theme version)
533 - FG#78: Multiple fixes for SMPI replay:
534 - TI tracing of allotallv/w was outputting wrong values
535 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
537 ----------------------------------------------------------------------------
539 SimGrid (3.28) July 14. 2021
541 The Victoriadagarna Release.
544 - C++ platform interface: Users can now describe their platform directly in C++.
545 This provides greatly flexibility and performance improvement for complex
546 platforms. Main features:
547 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
548 "zones", instead of single hosts. This feature enables the description
549 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
550 - StarZone: new zone with a Star-like topology. The routes are defined
551 as a set of links used to communicate from node to everybody (node<->ALL).
552 - Split-Duplex links: auxiliary method to create split-duplex links in
553 the platform, easing its utilisation. It automatically creates both UP
554 and DOWN links (similarly as done in XML).
555 - Please refer to the documentation and the examples included:
556 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
557 - New plugin: Producer-Consumer with monitor. Just requires to include the
558 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
559 associated example (examples/cpp/plugin-prodcons).
562 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
563 s4u::Io::wait_any(), s4u::Io::wait_any_for().
564 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
565 parameter by reference, instead of a pointer.
566 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
567 Explicitly cancel the activity to get back to previous behavior.
568 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
571 - The default SMPI compiler flags are no more taken from the environment.
572 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
573 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
575 --cfg=smpi/finalization-barrier: which can be used to add
576 a barrier inside MPI_Finalize. This can help for some codes which cleanup
577 data attached to a process, but still used in other SMPI processes.
578 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
579 is active when set to false, to keep going after a small error
580 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
581 which may or may not be problematic in the end.
583 - fix behaviour, as maximum iteration count could be ignored
584 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
585 use sampling when the same kernel is called with a different set of parameters
586 which have an impact on the timing.
587 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
588 It should now work with smpi/auto-shared-malloc-thresh.
589 - Improve error handling and reporting in multiple places
590 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
591 validity checks, truncated messages are now an error, return errors
592 when explicitely deleted handles are reused, ...)
593 - RMA: multiple fixes and stability improvements.
594 - analysis (-analyze flag in smpirun):
595 - SMPI can now report buffer leaks as well as MPI handles leaks,
596 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
597 - if -trace-call-location is used when compiling, SMPI can report
598 origin of leaked handles/buffers
599 - group leaks by type/origin in output message if possible
600 - New implemented MPI calls: MPI_Comm_test_inter
603 - Changed internal implementation of bandwidth factors in network models.
604 Models affected: CM02, LV08 (default), SMPI, IB.
605 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
606 Bandwidth factors are applied to communications to describe that users
607 cannot use 100% of the available bandwidth. For example, the default network model,
608 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
609 behavior was implemented by limiting the bandwidth available in the LMM
610 system for this flow. This may give the false impression that there is
611 bandwidth available for other flows due to its underutilization, especially
612 for the dynamic bandwidth factors used in SMPI models.
613 To avoid this, we have modified the implementation so that each flow uses the
614 maximum physical bandwidth according to the LMM system.
615 However, the actual throughput of the flow seen by the user is defined by
616 the physical bandwidth multiplied by the bandwidth factor.
617 This change impacts on the simulation results for all network models on
618 which we have bandwidth factors configured.
619 *****************************************
620 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
621 *****************************************
622 This change may impact on the timing of your simulation results.
623 Take care when comparing simulations from different SimGrid's
624 versions. Sorry for the inconvenience.
625 - Dynamic network factors: users can configure a callback to define
626 the network factors dynamically. This API is available at
627 simgrid::kernel::resource::NetworkModelIntf.
628 - Users have access to complete information about the current communication
629 to decide which factor to apply. This includes: message size, source and
630 destination hosts, links and zones traversed.
631 - Dynamic factors for both latency and bandwidth.
632 - For more details, see the example in (examples/cpp/network-factors).
633 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
634 deprecated since version 3.24, are no longer supported. Instead, use
635 "wattage_off" and "wattage_per_state".
638 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
641 - New tutorial: Model-checking and formal assessment
642 - New sections: "Demystifying the routing" and "C++ platforms"
643 - Update and improve the part on visualization in MPI and Algo tutorials.
644 - Improve the section on routing: how to define it, how it's used internally
645 - Fix many issues, broken links and missing references in doxygen and Sphinx
648 - Lua platform files are deprecated. Their support will be dropped after v3.31.
651 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
653 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
654 (FG: issues on Framagit; GH: issues on GitHub)
655 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
656 - FG#64: Configuring smpi/IB-penalty-factors
657 - FG#67: Running computation concurrently with MPI_Iallreduce
658 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
659 - FG#71: Segmentation fault on invalid gw_src/gw_dst
660 - GH#322: Issue when an actor kills his host vm
662 ----------------------------------------------------------------------------
664 SimGrid (3.27) March 29. 2021
666 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
668 Main user-visible changes:
669 - All functions marked as deprecated for one year were removed.
670 - Removed all that was related to the "Storage" type in all APIs and bindings.
671 "Disks" have been introduced more than a year ago to replace them for the better.
672 It's time to finish this replacement.
673 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
674 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
675 avoiding "initializer element is not constant" errors with some codes. Please be
676 careful to recompile your MPI codes if switching to this new SimGrid version.
677 - Network model ns-3 now requires version ns-3.28 or higher.
678 - Removed some option aliases that were deprecated since v3.23 (June 2019)
679 - 'verbose-exit' is now 'debug/verbose-exit'
680 - 'simix/breakpoint' is now 'debug/breakpoint'
681 - 'clean-atexit' is now 'debug/clean-atexit'
684 - New functions: Comm::sendto_{init,async} to initiate a communication
685 on between two (possibly remote) hosts.
686 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
687 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
688 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
689 or Mailbox::get_async<void>() if you really want to play with void*.
690 - Unify the interface of Activity::on_{start/activity}
691 - New function: Comm::get_dst_data()
692 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
695 - ptask should now work better with multicore hosts (FG#37).
696 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
697 - Internal cleanups and refactoring around models, to prepare for programmatic
699 - Modernize our tests of the performance models
703 - New section: Configuring the logs from the command line.
704 - New section: Programming API of logs.
705 - New section: Getting realistic results
706 - New section: Examples of SMPI Usage
707 - New part: SimGrid's Internals, with a section on Design goals
708 - Move back to the classical Breathe tool chain, and fix many small glitches.
711 - Handle leak checking: SMPI will now display a message if MPI handles are still
712 present at the end of a simulation. Option smpi/list-leaks:n will display the n
713 first leaked pointers and their types.
714 - smpi/display-allocs option : Display Memory allocations analysis, with hints
715 about the largest allocations. This only works for allocations with malloc/calloc
716 and not compiled with SMPI_NO_OVERRIDE_MALLOC
717 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
718 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
719 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
720 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
721 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
722 - Replay now opens and parses the trace file only once for all ranks when possible.
724 - better handling of MPI_PROC_NULL
725 - MPI_Group_incl was broken in some cases
726 - PAPI tracing now fixed
727 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
728 - SMPI will now fail properly when receive size < sent size, instead of
729 just setting MPI_ERR_TRUNCATE and keep going.
730 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
731 - Remove useless build patches and split build/run patches on the page
732 - Build in parallel when possible
733 - Test on yum-based systems
734 - New nightly tested Proxy-apps:
735 - OpenMPI mpi-test-suite
736 - Coral MCB, LSMS, Qbox, CLOMP
737 - ECP/Mantevo Cloverleaf3D, Ember
738 - Codevault bhtree_mpi
741 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
742 - Start to overhaul the implementation to modernize it (still ongoing).
745 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
746 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
747 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
748 - Deprecate xbt_dynar_to_array().
749 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
751 C binding and interface:
752 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
753 - Constness requirement has been lowered for parameter argv of sg_actor_start()
754 and sg_actor_create() It's now (char* const*). The previous version is still
755 available as sg_actor_start_() and sg_actor_create_().
757 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
758 (FG: issues on Framagit; GH: issues on GitHub)
759 - FG#37: Parallel tasks are limited to 1 core per host
760 - FG#62: Running "smpirun -replay" on large networks
761 - FG!46: Fix a few potential memory leaks in SMPI colls
762 - GH#343: Wrong consumed energy when an exec timeouts
764 ----------------------------------------------------------------------------
766 SimGrid (3.26) December 15. 2020
768 The Release release (the French lockdown was eased today).
770 Important user-visible changes:
771 - SimGrid now requires a compiler with C++14 support.
772 Sibling projects should upgrade their FindSimGrid.cmake
773 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
774 several users had difficulties to understand issues when using high bandwidth or
775 small latency events. The new value was already the default for SMPI and
776 should not cause too much performance hit. It can change some simulation
777 timings if simulation was skipping a lot of small events (which was actually
778 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
779 - On some network models (L07, CM02, LV08), an implicit loopback link is created
780 for local communication. Default bandwidth was 500MBps and latency was 15µs,
781 which was rather bad. Default values have been changed to 10GBps and 0 latency.
782 Options network/loopback-lat and network/loopback-bw have been added to adapt
783 these values if needed. Network and routing models already providing separate
784 loopbacks (clusters and such) are not impacted by this change.
785 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
786 metadata that are added to each MPI message by MPI libraries (tag,
787 src/dest, and communicator information). This may change simulation times
788 for very small message sizes, but should be more realistic.
791 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
792 - New plugins: wifi_energy and link_load. See the associated examples.
793 - New examples: wifi networks, and communication suspend.
796 - LTO is now enabled for Intel/clang compilers.
797 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
798 Setting it to "auto" will use all available cores, while setting it to n will
799 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
800 - Remove obsolete runtime option 'contexts/parallel-threshold'.
801 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
802 has been kept for compatibility.
803 - Finally remove obsolete snake_case() aliases for runtime options.
804 - Further improve the documentation.
807 - the on_exit() of each actor is also executed when the simulation deadlocks.
808 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
809 An example is provided for s4u::Comm but it should work with Exec and Io.
812 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
813 https://framagit.org/simgrid/SMPI-proxy-apps
814 - MPI/IO: supports names starting with "./"
815 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
817 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
818 building shared libraries, this may cause issues, so environment variable
819 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
820 - most temporary files should now be created in /tmp dir (or equivalent).
821 If this one does not allow execution of code (noexec flag), this may cause issues.
822 Please use another tmp directory (using TMPDIR or equivalent system variable)
826 - A large refactoring is ongoing, but this should be transparent for users so far.
829 - Drop xbt_str_split().
831 C binding and interface:
832 - The return type of the sg_actor_on_exit() callbacks is now 'void'
833 (int value was previously ignored)
834 - Many C functions were renamed to follow a common naming style
835 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
836 names are kept until version 3.30.
837 - Many MSG tests were converted to the new S4U's interface in C, that
838 was extended for that.
840 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
841 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
842 - FG#41: Add sg_actor_create C interface
843 - FG#43: xbt::random needs some care
844 - FG#48: The Impossible Did Happen (yet again)
845 - FG#50: Suspending an actor executed at the current timestamp fails
846 - FG#52: Zero-seconds timeout : "That's in the past already"
847 - FG#53: Crash while using ns-3 network model
848 - FG#54: How to suspend a comm?
849 - FG!22: Stochastic Profiles
850 - FG!24: Documentation and fix for xbt/random
851 - FG!35: Add a modeling hint for parallel links in doc
852 - FG!36: [xbt/random] Read/Write the state of the RNG
853 - FG!42: Link cumulated load plugin
854 - FG!44: Test condition variable wait_until+notify_all
855 - GF#18137: Allow different stack sizes?
856 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
857 - GH#139: Allow pthread creation in SMPI
858 - GH#336: Packet-level simulation using SMPI?
859 - GH#345: Error in the LMM while migrating a VM
860 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
861 - GH#352: pip install / python setup.py install fail to find pybind11
862 - GH!337: Fix link_energy plugin for wifi platforms
863 - GH!339: Add Mailbox set_receiver method to python binding
864 - GH!344: Cast hugepages macros parameters to int64
866 ----------------------------------------------------------------------------
868 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
870 The Palindrom Day Release.
872 Important user-visible changes:
873 - Improve the Python usability (stability and documentation).
874 - A nasty synchronization bug was ironed out, see also below.
875 - Python's doc was organized as a separate tree, now integrated with C++.
876 - C bindings of S4U were not part of the doc.
877 - The C++ doc was also improved as methods are now split by theme.
878 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
879 - This is mandatory to use the Java bindings.
880 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
881 - SimDAG++: Automatic dependencies on S4U activities (experimental)
882 - Some features are already implemented but not all of them
883 - Cannot block an activity until it's scheduled on a resource
884 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
885 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
888 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
889 - Drop support for Viva/Triva (old visualization tools), hence removed
890 TRACE_get_node_types() and TRACE_get_edge_types() functions.
893 - Require pybind11 version 2.4.
894 - Greatly improve locking during context switches with Python.
895 - Declaring Python callbacks to this_actor::on_exit is now working.
896 - No more tests are failing currently, even on win32.
899 - Actor: Merge signals on_migration_start/end into on_host_change
900 - Actor: Rename migrate() into set_host()
901 - Disk: Allow users to get the read and write nominal bandwidth values
902 - Exec: Implement wait_for(timeout)
903 - Io: Implement wait_for(timeout)
906 - Parse errors now raise a simgrid::ParseError that you may want to catch.
909 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
910 been renamed to fix a spelling error. As usual, the previous definitions are
911 kept with a deprecation warning. However, to avoid runtime errors, it is now
912 forbidden to override these deprecated methods in a derived class. Please use
913 the new names immediately if you need to override them.
916 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
917 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
918 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
919 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
920 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
921 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
922 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
923 - New module: random, providing classical random numbers generators.
926 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
927 allocations' size above which they are considered "shared" by default (as if
928 they were performed through SMPI_SHARED_MALLOC macros).
929 Default = 0 = disabled feature.
930 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
931 This can cause some troubles if codes are already overriding these. If this
932 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
933 help, but will make this feature unusable.
934 - Input check is now performed much more consistently and outputs precise
935 warnings for each error found
936 - New supported calls : MPI_File_set_view, MPI_File_get_view
937 - MPI I/O now only supports disks, as storage is discontinued.
938 - MPI I/O allows opening files without specifying absolute path
939 - Note : SMPI configuration options are now initialized by smpi_init_options()
940 call, which should be called if SMPI is not being used through smpirun.
942 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
943 - FG#9: Python bindings crashing
944 - FG#39: Missing s4u::Comm::wait_any_for example
945 - FG#42: Add support for ThreadSanitizer (TSan)
946 - FG!19: Removing RngStream
947 - FG!20: A module for RNG calls
948 - FG!21: Choice between ad-hoc and standard distributions implementations
950 - FG!25: Fix link in Mutex doc
951 - FG!26: Fix links in SMPI interface doc
952 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
953 - GH#305: Unscheduled tasks are still excuted
954 - GH#313: smpirun: manual outdated w.r.t. --help
955 - GH#321: [S4U] Get task remaining work ratio
956 - GH#323: Crash when an actor turn off his physical host
957 - GH#335: Missing links on dragonfly example svg
958 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
959 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
961 ----------------------------------------------------------------------------
963 SimGrid (3.24) October 9. 2019
965 The Clean Disk Release.
968 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
969 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
970 thus becomes the default resource to execute I/O operations. Examples
971 and the file system plugin have been modified to use disks instead of
972 storage elements. Storage can still be used but is doomed to disappear.
973 Users are thus adviced to update their simulators accordingly.
974 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
975 for consistency with pthread_barrier_wait()
976 - Host::get_englobing_zone() returns the englobing netzone
977 - Actor::on_destruction is now called in the destructor
978 Actor::on_termination new signal called when the actor terminates
980 - Global signals are now part of the Engine:
981 - on_platform_creation: after config settings, before the XML parsing
982 - on_platform_created: right after the XML parsing
983 - on_time_advance: each time the clock advances
984 - on_simulation_end: after simulation, before cleanups
985 - on_deadlock: as the name implies.
987 - sg_{actor,host,link}_{data,data_set}() now all exist.
988 Use them to attach user data to the object and retrieve it.
991 - Introduce an experimental Wifi model. It sounds reasonable
992 according to the state of the art, but it still has to be properly
993 validated, at least against ns-3 if not against reality.
994 - Improved the usability of ns-3. Several bugs were ironed out.
995 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
996 It is now expressed as 'idle:epsilon:allCores' to properly model the
997 consumption of non-whole tasks on mono-core hosts. More info in the doc.
1000 - convert a new set of functions to the S4U C interface and move the old MSG
1001 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
1004 - Fortran bindings for DVFS have been removed.
1005 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
1006 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
1007 - SMPI can now be selected by cmake's find_module(MPI) with
1008 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
1009 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
1010 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
1011 - trace-call-location can be used with TI traces, and replayed, and is compatible
1012 with smpi/comp-adjustment-file.
1013 - sleep events are now correctly traced/replayed.
1014 - Default for trace-call-location is now to use file names and not full paths.
1015 To revert to previous behavior (in case of collision of filenames), option
1016 "smpi/trace-call-use-absolute-path" can be set to yes.
1019 - Use the included xxHash as an hash implem when C++14 is usable.
1020 - Option model-checker/hash was removed. This is always activated now.
1021 - New option smpi/buffering controls the MPI buffering in MC mode.
1022 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
1023 This is useful to check for MPI compliance.
1026 - New section on plugins: how to define a new one, and existing ones.
1029 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
1030 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
1033 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
1034 and <mount> tags. The rationale is finish to decouple what should be handle
1035 by SimGrid kernel (i.e., resources and activities) from plugins built in
1036 user space (e.g., the file system plugin). This new tag currently accepts
1037 only three attributes (id, read_bw, and write_bw). All the other information
1038 that was declared with the storage related tags now has to be expressed as
1039 properties. An example of platform using this new tag is available at
1040 examples/platforms/hosts_with_disks.xml
1043 - 'expect signal' can now accept more than one potential signal.
1045 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
1046 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
1047 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
1048 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
1049 - FG#31: per-actor data
1050 - FG#34: SG_BARRIER_SERIAL_THREAD?
1051 - FG#35: model-checker does not like buster-produced binaries
1052 - FG!13: MC: complete workaround in the error msg seen on modern systems
1053 - FG!15: execute_flops now logs compute
1054 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
1055 - FG!17: ns-3: unblock the right number of communications + others issues
1056 - FG!18: Improving the performance of the ns-3 bindings
1057 - GH#189: Energy consumption of parallel tasks
1058 - GH#219: Error in the throughput of TCP transfer
1059 - GH!330: Minor polishing of English: Configuring_Simgrid
1060 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
1061 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
1062 - GH!334: Rework energy plugin, again
1064 ----------------------------------------------------------------------------
1066 SimGrid (3.23.2) July 8. 2019
1069 - Nicer introduction page.
1070 - Migrate the "Deploy your application" page to the new doc.
1071 - Move Java as a subtree of MSG.
1074 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
1077 - Drop xbt_dynar_sort_strings().
1080 - Really fix FG#26: Turning off a link should raise NetworkFailureException
1081 - FG#27: Wrong exception thrown to wait_any when link is turned off
1082 - GH#328: Java: Canceling multiple tasks in a single vm/host
1084 ----------------------------------------------------------------------------
1086 SimGrid (3.23) June 25. 2019
1088 The Exotic Solstice Release.
1091 - SunOS and Haiku OS support. Because exotic platforms are fun.
1092 - Stop setting random seed with srand48() at initialization.
1093 - Use addr2line as a fallback for stacktraces when backtrace is not available.
1094 - Build option -Denable_documentation is now OFF by default.
1095 - Network model 'NS3' was renamed into 'ns-3'.
1098 - SimGrid can now hopefully be installed with pip.
1101 - wait_any can now be used for asynchronous executions too.
1104 - New log appenders: stdout and stderr. Use stdout for xbt_help.
1105 - Drop xbt_dict_dump.
1108 - 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
1109 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
1110 - MPI interface is now const correct for input parameters
1111 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
1112 - Fortran bindings for async collectives.
1113 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
1116 - Remove option 'model-check/record': Paths are recorded in any cases now.
1117 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
1118 always sparse. That's an immense gain in memory, and can even be
1119 faster because of cache effects. So there is no need to clutter the
1120 code to allow the user to go for the unefficient mode.
1123 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
1124 models proved to be more accurate than these old experiments.
1126 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
1127 - FG#1: Broken link in error messages
1128 - FG#2: missing installation documentation
1129 - FG#3: missing documentation in smpirun
1130 - FG#6: Python bindings not available on PyPI
1131 - FG#7: simple cmake call requires doxygen
1132 - FG#8: make python bindings an optional dependency
1133 - FG#10: Can not use MSG_process_set_data from SMPI any more
1134 - FG#11: Auto-restart actors forget their on_exit behavior
1135 - FG#12: -Denable_lto=OFF doesn't disable LTO
1136 - FG#13: Installs unstripped file 'bin/graphicator'
1137 - FG#14: Installs the empty directory 'doc/simgrid/html'
1138 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
1139 - FG#17: Dead link in doc (pls_ns3)
1140 - FG#20: 'tesh --help' should return 0
1141 - FG#21: Documentation link on http://simgrid.org/ broken
1142 - FG#22: Debian installation instruction are broken
1143 - FG#26: Turning off a link should raise NetworkFailureException exceptions
1144 - GH#133: Java: a process can run on a VM even if its host is off
1145 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
1146 - GH#326: Valgrind-detected error for join() when energy plugin is activated
1148 ----------------------------------------------------------------------------
1150 SimGrid (3.22) April 1. 2019
1152 The Easter Christmas Release (this one is somewhat late).
1155 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
1156 - Not complete yet: asynchronous activities (amongst others) are still missing
1157 - Still ongoing: the interface may change in the future. We need more
1158 testers! Please report any glitches.
1159 - No new project using Java should start now. Please switch to Python.
1162 - Some of the internal cleanups may lead to speed improvements:
1163 - The hard limitation on the amount of simulated actors with
1164 Java+Mac was removed. Now, the available memory is the only limit.
1165 - Our refcounting was tidyied, leading to 10% speedups in some cases.
1166 - We are still working on making our code robust to the actor kills
1167 and hosts' churn. Things are improving, but it's not perfect yet.
1168 - Replaced our own code to display a backtrace (that was forking addr2line)
1169 with the Boost.Stacktrace library.
1170 You won't see your backtraces without this optional dependency.
1171 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
1172 - Stop setting random seed with srand() at initialization.
1175 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
1176 XML file version remains 4.2 since old files are still compatible.
1179 - Process termination which was broken at version 3.21 has been repaired.
1180 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
1181 - Hide the examples into examples/deprecated. New users should use Python.
1184 - Drop MSG_process_create_from_stdfunc() from the API.
1185 This C++ function was a pimple in the C API, made necessary at some
1186 point by the Java bindings. This is fixed now.
1187 - Hide the examples into examples/deprecated. New users should use S4U.
1188 - MSG_process_create and MSG_process_attach now crash if the host used
1189 to run this process is off.
1190 - Fix the protype of MSG_process_on_exit()
1191 Now use (int,void*) callbacks instead of (void*,void*) ones.
1192 The implementation was ways too messy to actually work, I guess.
1195 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
1196 too often and provide more accurate timings.
1197 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
1198 - MPI_Alltoallw support
1199 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
1200 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
1201 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
1202 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
1204 - Basic implementation of generalized requests (SMPI doesn't
1205 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
1208 - Drop sg_cmdline. Please use xbt_cmdline instead.
1209 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
1210 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
1211 OsSemaphore is implemented in a portable way with C++11 threads.
1212 This should allow much more threads to be created at the same time,
1213 allowing Mac and Java users to simulate many more actors.
1214 - Implement the 'thread' factory with std::thread instead of xbt ones.
1215 It is not possible to set the stack size with threads anymore, but
1216 -fsplit-stack is the way to go nowadays when using threads.
1217 - Drop the xbt_os_thread_t module (now unused)
1218 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
1219 - Drop xbt_str_join_array().
1220 - Drop cunit, use Catch2 instead.
1223 - Many cleanups in the kernel::activity namespace. This was long
1224 overdue, and shall open the path to many future endeavors.
1227 - #132: Java: a process can not shut down its own host
1228 - #220: S4U: detached send causes exception if sender terminates soon after sending
1229 - #261: Document the parameters of parallel execution's constructor
1230 - #300: [s4u] BarrierPtr is missing
1231 - #314: SMPI args internal cleanup
1232 - #316: Fix a bug related to the CPU utilization of multi-core VM
1233 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
1234 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
1235 - #325: Turning off a host has different behavior on sleeping actors and computing actors
1237 ----------------------------------------------------------------------------
1239 SimGrid (3.21) October 3. 2018
1241 The Restarting Documentation (TRD) Release.
1244 - Convert the user manual to Sphinx for improved usability.
1245 Unstable version now hosted on https://simgrid.frama.io/simgrid
1246 Many glitches remain; stay tuned.
1247 - Restrict the Doxygen reference API to the installed header files.
1248 (documenting all internals this way is overkill + doxygen scales badly)
1249 - New tutorial on S4U (highly inspired from the old MSG one, but with
1250 a git repository to fork as a starting point, and a docker image)
1251 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
1252 - The Developper manual is still to be converted (not compiled until then)
1253 - Some parts are still missing in this conversion (such as the
1254 platform and deployment sections) while others were blindly converted
1255 and would need more love (such as the configuration flags).
1256 Things will certainly further improve in the future releases.
1259 - Fully reimplement auto_restart mechanism that was utterly broken
1260 (fix #22, #131 and #281 that were all very related)
1261 - Implement semaphores (fix #303 and #291)
1262 - When creating an actor from a function and its parameters,
1263 move-only parameters are not allowed anymore, as it would prevent
1264 the actor to be restartable if its parameters are consumed on run.
1265 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
1266 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
1267 IO operations that can be started, waited for, or canceled as a regular
1269 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
1273 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
1274 - New host variable: 'core_count'
1277 - Remove xbt_os_thread_specific features
1278 - Remove portability wrapper to condition variables
1279 - Remove xbt_os_thread_yield()
1282 - MPICH collective selector now mimics MPICH 3.3b
1283 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
1284 - MPI_Init(NULL,NULL) is now allowed
1285 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
1286 - Command-line arguments should now be handled properly in Fortran simulations
1287 - MPI Attributes and related callbacks should work in Fortran
1288 - Apps using C + Fortran should now work
1289 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
1290 - Support MPI_Error_String
1293 - Due to an internal bug, Msg.run() must now be your last line.
1294 We hope to fix it in a future release, and we are sorry for the inconvenience.
1297 - #22: Process autorestart seem to only work with CAS01 cpus
1298 - #56: Feature request: dynamic SMPI replay
1299 - #93: simgrid should not eat --help
1300 - #111: How to change the loopback link characteristics?
1301 - #116: Platform cluster radical
1302 - #117: Platform cluster/host order
1303 - #129: Replay_multiple on steroid
1304 - #131: Java : autostart does not work
1305 - #143: Setting a breakpoint at a given time
1306 - #235: Network energy models should be integrated
1307 - #264: Add ptask L07 resource tracing
1308 - #271: Dynamic and manual replay of SMPI traces
1309 - #279: Breakpoints option uses time deltas instead of absolute time
1310 - #281: Daemonized actor no longer daemonized after an auto-restart
1311 - #282: TIT convertor erase traces when using absolute path in the trace list file
1312 - #285: segfault when a s4u actor kills itself with msg process tracing activated
1313 - #286: Pajé traces are not exposing the number of cores
1314 - #287: Command-line parsing should stop when encountering '--'
1315 - #288: MPI_Init(NULL, NULL)
1316 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
1317 - #290: Method to check if a mailbox has a message ready to be consumed
1318 - #291: [s4u] semaphores do not seem available yet
1319 - #294: Alltoall collective implementation of mpich
1320 - #295: Starting a Host as OFF is not supported
1321 - #296: DTD too permissive
1322 - #299: Add a s4u barrier example
1324 ----------------------------------------------------------------------------
1326 SimGrid (3.20) Released June 24. 2018
1328 The proxy snake_case() release.
1331 - The whole API is now using snake_case() instead of mixing it with camelCase.
1332 Compatibility wrappers in place for one year.
1333 - We took the opportunity to increase the consistency of the API.
1334 Please report any remaining glitches.
1335 - New function simgrid::s4u::Host::get_actor_count:
1336 Returns the number of actors running on a specific host.
1339 - SMPI is now tested with ~45 proxy apps from various sources, with none or
1340 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
1341 - Replay: The replay file has been re-written in C++.
1342 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
1343 supported. They are stored in the trace and used when replayed.
1344 - Basic support of MPI_Cancel. Robustness not guaranteed.
1345 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
1346 (only for MPI_COMM_TYPE_SHARED).
1347 - New option: smpi/privatize-libs, to add external shared libs to be privatized
1348 by SMPI. They will be copied locally and loaded separately by each process.
1349 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
1350 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
1351 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
1352 with internal index optind with multiple processes. Only works if MPI_Init has
1353 already been called.
1354 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
1355 syntax without preprocessing tricks.
1358 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
1359 behavior tracing is no longer limited to MSG
1360 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
1361 this means that traces now need to be updated or re-obtained; both Irecv
1362 and Isend lines in your traces have now in total 5 mandatory fields:
1363 <rankid> <command> <to/from rankid> <tag> <size>
1364 To update your traces, it suffices to add a 0 for the tag here.
1365 - TIT now also supports waiting for a distinct request via MPI_Wait.
1366 Wait/Test now wait for a specific request, not just the last one that was
1367 issued. This unfortunately means another update, because we need to
1368 identify which request you want to wait for. We do this via the
1369 triplet (sender, receiver, tag), which needs to be added:
1370 <rankid> <command> <sender> <receiver> <tag>
1371 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
1372 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
1375 - The deprecation of MSG is ongoing (but this should not impact you).
1376 Many MSG functions are now simple wrappers on the C API of S4U. If
1377 you wish to convert your code to S4U, find the S4U counterparts of
1378 your MSG calls in src/msg/msg_legacy.cpp.
1379 - MSG can still be used, but won't evolve anymore.
1382 - Allow to run the Link energy plugin from the command line with
1383 --cfg=plugin:link_energy
1384 - Rename Energy plugin into host_energy
1385 - Rename Load plugin into host_load
1388 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
1389 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
1390 - Improve performance when handling timeouts of simix synchros.
1393 - Config: the C API is now deprecated (will be removed in 3.23), and
1394 the C++ API has been slightly improved.
1397 - Fix several build issues on OSX.
1398 - Move simgrid_config.h to simgrid/config.h (old header still working)
1401 - #143: Setting a breakpoint at a given time
1402 - #150: Inconsistent event names in SMPI replay
1403 - #258: daemonized actors hang after all non-daemonized actors have completed
1404 - #267: Linker error on unit_tmgr
1405 - #269: SMPI: tracing of MPI_Wait/all/any broken
1406 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
1408 ----------------------------------------------------------------------------
1410 SimGrid (3.19.1) Released March 22 2018
1412 The Fixed ABI Release.
1414 As you may know, we are currently refactoring SimGrid in deep.
1415 Upcoming SimGrid4 will be really different from SimGrid3: modular,
1416 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
1418 Our goal is to smooth this transition, with backward compatibility
1419 and automatic update paths, while still progressing toward SimGrid4.
1421 SimGrid remains open during works: The last pure SimGrid3 release was
1422 v3.12 while all subsequent versions are usable alpha versions of
1423 SimGrid4: Existing interfaces remain unchanged, but the new S4U
1424 interface is budding and the internals are deeply reorganized.
1426 Since 2015, we work hard to reduce the changes to public APIs. When
1427 we need to rename a public library symbol in S4U, we let your
1428 compiler issue an explicative warning when you use the deprecated
1429 function. These messages remain for four releases, i.e. for one full
1430 year, before turning into an error. Starting with v3.15, your can
1431 also adapt to API changes with the SIMGRID_VERSION macro, that is
1432 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
1434 Starting with this v3.19.1, our commitment to reduce the changes to
1435 the public interfaces is extended from the API to the ABI: a program
1436 using only MSG or SimDag and compiled against a given version of
1437 simgrid can probably be used with a later version of SimGrid without
1438 recompilation. We will do our best... but don't expect too much of
1439 it, that's a really difficult goal during such profund refactoring.
1441 The difference between v3.19 and v3.19.1 is that the former was
1442 accidentally breaking the ABI of MSG, while the later is restoring
1445 S4U and kernel APIs will still evolve until SimGrid4, with one-year
1446 deprecation warnings as currently. In fact, cleaning up these
1447 interfaces and converting them to snake_case() is one release goal of
1448 v3.20. But don't worry, we are working to smooth this upgrade path.
1450 In summary, new projects should start with S4U to benefit of the
1451 future, but old MSG projects should still be usable with no change.
1453 ----------------------------------------------------------------------------
1455 SimGrid (3.19) Released March 20 2018
1457 The Moscovitly-cold Spring release.
1459 SMPI user-visible improvements:
1460 - Support of MPI_Type_create_subarray
1461 - Unimplemented MPI calls now abort (instead of issuing a warning)
1462 Some harmless calls around errhandlers still only issue a warning.
1463 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
1464 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
1465 is now possible (mmap privatization was refactored to this end).
1466 Adding new SMPI instances should also be technically doable now.
1468 SMPI internal cleanups:
1469 - Removed index notion from SMPI -- no more "getPid() - 1"!
1470 In tracing files, actors are called rank-<PROCESS-ID> but these are
1471 NOT the real ranks (tracing several communicators is not supported yet).
1472 - Deprecate smpi_process_index() (will be removed in 3.22)
1473 - Better testing on classical proxy apps, through an external project:
1474 https://github.com/simgrid/SMPI-proxy-apps/
1477 - Use a graphical TOC to make it easier to find the documentation you need
1478 - Revamp the MSG tutorial
1481 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
1482 - Remove parameter of MSG_process_killall().
1483 Resetting the PID was bogus anyway (several actors could have the same PID).
1486 - Execution->setHost() can be called after start(), to migrate it.
1487 - Comm::test_any() is now implemented.
1488 - s4u::Actor now has onCreation() and onDestruction() signals.
1489 - Install some kernel header files for the users' plugins and more.
1490 Warning, their API is really not stable yet.
1491 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1492 include/simgrid/kernel/routing/*Zone.hpp
1495 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1497 Build System and other cleanups:
1498 - Remove unused run-time parameter "tracing/onelink-only".
1499 - SimGrid now works with intel compilers (MC still buggy)
1500 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1501 - tesh kills subprocesses on timeout (but not on windows)
1504 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1507 - #194: Feature request: simgrid::s4u::Comm::test_any()
1508 - #245: migrating an actor does not migrate its execution
1509 - #253: Feature Request: expose clusters as objects
1510 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1511 - #255: Tesh broken on Windows
1512 - #256: Modernize FindSimGrid.cmake
1513 - #257: Fix (ab)use of CMake install
1515 ----------------------------------------------------------------------------
1517 SimGrid (3.18) Released December 24 2017
1519 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1521 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1522 - Support for MSG should not be removed from SimGrid before 2020,
1523 but future evolutions will be limited to the integration of
1524 user-provided patches.
1527 - s4u::onDeadlock() signal, run before stopping the simulation.
1528 - s4u::Exec: asynchronous executions (abstraction of background
1530 - s4u::Host->getLoad() returns the achieved speed in flops/s
1533 - Link::name() is deprecated and will be removed in v3.21, use
1534 Link::getCname() instead.
1535 - Mailbox::getName() changed to return a std::string, use
1536 Mailbox::getCname() instead to get a char*.
1537 - Storage::getName() changed to return a std::string, use
1538 Storage::getCname() instead to get a char*.
1539 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1540 to have both allocation and deallocation of the map in user space.
1543 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1544 The amount of flops remaining to do is a vector, not a scalar.
1545 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1546 name implies on both sequential and parallel tasks.
1547 - Both changes fix GitHub's #223 using PR #237 as a basis.
1548 Thanks Michael Mercier.
1549 - Most examples were converted to S4U and hidden elsewhere at tests.
1550 Rationale: we still want MSG to work; we want newcomers to use S4U.
1553 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1554 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1557 - Switch to the faster dlopen privatization mechanism by default
1558 - Documentation improvments
1561 - Live migration is getting moved to a plugin. Dirty page tracking is
1562 the first part of this plugin. This imply that VM migration is now
1563 only possible if one this function is called:
1564 - C/MSG: MSG_vm_live_migration_plugin_init()
1565 - C/C++: sg_vm_live_migration_plugin_init()
1566 - Java: Msg.liveMigrationInit()
1567 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1571 - Define class simgrid::xbt::Path to manage file names.
1572 - Removed unused functions:
1573 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1574 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1575 xbt_graph_export_graphviz()
1576 - xbt/str.h: xbt_str_join()
1577 - Remove unused datatypes:
1578 - xbt/heap.h: use std::priority_queue or boost::heap instead
1579 - xbt/swag.h: use boost::intrusive::list instead
1582 - New link_energy plugin for the consumption of the links.
1583 - All of the operations on files and storage contents have been
1584 packaged into a plugin (src/plugins/file_system). The current
1585 public interface can be found in
1586 include/simgrid/plugins/file_system.h
1587 To use these functions you now have to initialize the plugin by
1588 calling MSG_storage_file_system_init() just after calling
1589 MSG_init() or sg_storage_file_system_init() just after creating
1593 - Remove the undocumented/untested tag <include>
1596 - Remove viva specific tracing as the tool is no longer maintained
1599 - #248: Kill of finished processes leads to segfault
1600 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1601 - #239: please implement signal s4u::onDeadlock()
1602 - #230: segfaults when exit() without run()
1603 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1604 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1605 - #222: Actor::kill() doesn't really kill and segfaults
1606 - #221: odd LMM warning when killing an actor
1607 - #120: Memory leak when the processes are forcefully killed
1609 ----------------------------------------------------------------------------
1611 SimGrid (3.17) Released October 8 2017
1613 The Drained Leaks release: (almost) no known leaks despite the tests.
1615 Even very long simulations will preserve your memory: our testsuite
1616 stresses SimGrid for over 45mn on fast machines for a coverage of
1617 over 80%, and there is only one single known leak, of about 4kb.
1621 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1622 handler may be given to free resources if the comm cannot be completed.
1623 - this_actor::send(mailbox) is now mailbox->put()
1624 - New: simgrid::s4u::Comm::wait_all()
1625 - New: Host.execute() for remote executions.
1628 - Unused option network/sender-gap is removed.
1631 - Deprecate MSG_task_isend_with_matching(): This unused feature
1632 really complicates our internals. Will be removed in v3.20.
1635 - Improved context termination. It is now possible to free resources, even
1636 when a process is forcibly killed.
1637 - ContextBoost: add support for Boost versions above 1.61.
1640 - Bring back run-time option --cfg=exception/cutpath to remove exception
1642 - Removed unused functions:
1643 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1644 xbt_str_rtrim(), xbt_str_trim().
1645 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1648 - Removed header files obsolete since SimGrid 3.12:
1649 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1651 - #3: SD_exit should be made optional
1652 - #120: Memory leak when the processes are forcefully killed
1653 - #159: Threading test regression in Actor refcounting
1654 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1655 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1656 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1657 - #191: VM migration and pstate
1658 - #192: Updating the energy consumptions of all hosts crashes with VMs
1659 - #195: All actors have PID=0 in the logs
1660 - #204: Sometimes segfault with thread contexts and mmap privatization
1661 - #222: Actor::kill() doesn't really kill and segfaults
1662 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1664 ----------------------------------------------------------------------------
1666 SimGrid (3.16) Released June 22. 2017.
1668 The Blooming Spring Release: developments are budding.
1671 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1672 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1673 - Java JAR file should be finally fully working on Mac OSX too.
1675 XML platforms: Switch to platform v4.1 format.
1676 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1677 - <zone> can be used as a synonym for the now deprecated <as>
1678 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1679 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1680 - <actor> can be used as a synonym for the now deprecated <process>
1681 - state_file and avail_file periodicity is now easier to express
1682 (check the documentation) the old behavior should still work.
1683 * Storage: not backward compatible, but it seems that nobody ever used it
1684 - Remove attribute 'content_type' of <storage_type>: was never used
1685 - Make attribute 'model' of <storage_type> optional: for future usage
1686 - Remove Bconnection model property: was never (in)validated. Replaced by
1687 the maximum of the read and write bandwidth as a resource constraint.
1690 - New and Backwards Compatibility break:
1691 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1692 When one wants to get the list of tasks whose states have changed during a
1693 simulation round, s/he has to allocate and free a dynar and use it as
1694 argument to this function. The former SD_simulate (double how_long)
1698 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1699 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1702 - New function to update the consumption of all hosts at once.
1703 - Fix the model for multi-core hosts, linear in the amount of busy
1704 cores with an abnormality for pIdle (see paper).
1705 BUG: Still not satisfactory for ptask on mono-cores.
1708 - The netzone are now available from the MSG API.
1709 The old names still work, but are now deprecated.
1712 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1713 - New option: smpi/keep-temps to not cleanup temp files
1714 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1715 mallocs mode is used, allows to change the size of the fake file used
1716 (default 1MB), to potentially limit the number of mappings for large runs.
1717 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1718 - Fortran ifort and flang compilers support
1719 - New RMA calls supported (experimental):
1720 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1721 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1722 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1723 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1724 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1725 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1728 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1729 - Behind the scene, ActivityImpl are now automatically refcounted.
1730 Removing this nasty bottleneck will greatly simplify our internals.
1733 - Replay: New function xbt_replay_action_get():
1734 Retrieve the function previously associated to an event type.
1735 - DROPPED FUNCTION: xbt_str_varsubst()
1736 - DROPPED MODULE: strbuff. We don't need it anymore.
1737 - DROPPED MODULE: matrix. We don't need it anymore.
1738 - DROPPED MODULE: lib. We don't need it anymore.
1740 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1742 SimGrid (3.15) Released March 22. 2017
1744 The Spring Release: continuous integration servers become green
1745 We fixed even the transient bugs on all target architectures:
1746 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1747 Windows, FreeBSD, NetBSD.
1749 - Rename NetCards to NetPoints.
1750 This was intended to help NS3 users, but that's not a netcard.
1751 That's a point in the routing algorithm, let's avoid wrong simplifications.
1754 - New: MSG_process_yield(). Stop and yield to other processes.
1755 - New: MSG_process_daemon(). Daemon processes are automatically killed
1756 when the last non-daemon process terminates
1757 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1758 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1759 to make room for the upcoming network energy plugin.
1760 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1763 - Ensure that an actor can kill itself with Process::exit()
1764 - Kill the obscure NativeException. Nobody want to survive the issues
1765 it denotes, so use JniException that is a RuntimeException (not to
1766 be caught explicitly).
1767 - Partial bug fix in initialization. SimGrid flags on command line were
1768 consumed at C level but stayed in the original Java String[] args.
1769 This could mess users' args[i] if SG flags were not put at the end of
1771 The SimGrid flags are now removed from the Java arguments. However,
1772 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1773 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1774 It might be if you have --log or --cfg flags in the command line.
1775 - Fix numerous memleaks all around the place. In particular, around VMs.
1779 - s4u::onPlatformCreated: right before the simulation starts
1780 - s4u::onSimulationEnd: right after the main simulation loop
1781 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1782 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1783 event from the availability_file changes the avail speed.
1784 - Links are now usable from s4u
1785 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1786 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1787 - New examples: The conversion of MSG examples to S4U has begun
1788 - Actors: create, kill, migration, and suspend
1789 - Applications: master-worker and token-ring
1790 - Action replay: communications and storage
1791 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1794 - Backwards Compatibility breaks
1795 - The SD_route_get_list and SD_route_get_size functions have been removed.
1796 They are replaced by the unique
1797 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1798 The route from src to dst is built in the links dynar whose size can
1799 be obtained with xbt_dynar_length.
1800 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1801 sg_host_route_bandwidth and sg_host_route_latency.
1802 Macros ensure the backwards compatibility, but you should fix your code
1805 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1806 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1807 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1808 We don't support 100% of it, but it should be enough. Contact us if not.
1809 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1810 naive and probably inaccurate)
1811 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1812 mapped onto a unique small file using some system magic.
1813 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1814 - Fortran: cleanups, fixes, support of user-added operations
1815 - RMA: MPI_Accumulate are applied in correct order.
1816 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1817 - Better support for MPI_IN_PLACE
1818 - Support for MPI_Win attrs and keyvals.
1819 - Support MPI_Comm_group_create, MPI_Type_size_x
1820 - MPI ops should be properly applied to all allowed types
1821 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1822 algorithm by a non failing one from mpich.
1825 - Kill the fifo data container: we don't use it anymore.
1828 - A <cluster> can now be created with different speed values to represent pstates.
1829 see examples/platforms/energy_cluster.xml for an example.
1831 SimGrid (3.14.159) Released December 28. 2016
1833 The previous release was only a crude approximation of a Pi release;
1834 this one should be a bit better.
1836 - Revert a last minute change that broke on Mac OSX.
1837 - Fix the changelog and NEWS files.
1838 - Improve the documentation, in particular of the routing module.
1839 - Integrate some patches flying around in Debian.
1841 SimGrid (3.14) Released December 24. 2016
1844 * The whole documentation was reworked and reorganized. There is
1845 still much room for improvement here, but we're on it.
1848 * We now need python3 for our testing infrastructure.
1849 * Model checking needs libevent
1852 * Now works on FreeBSD
1855 * Privatization now works on FreeBSD
1857 * Privatization is now activated by default in smpirun
1858 You can turn it off with -no-privatize if something goes wrong.
1860 * Call-location tracing for SMPI.
1861 You can add the exact location (filename / linenumber) of an MPI call to
1862 your trace files and slow down or speed up the simulation between two
1863 consecutive calls by using an adjustment file (see the documentation).
1865 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1866 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1867 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1868 values were ignored in some cases. The timings of these functions can now
1869 be significantly different.
1871 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1872 smpi/running-power is renamed to smpi/host-speed
1874 * smpi/grow-injected-times option to enable or disable multiplication of the
1875 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1876 make simulation less precise (but also much faster).
1878 * smpirun script should be (much) faster for large deployments.
1880 * SMPI tracing: fixed issue with poor matching of send/receives.
1882 * Replay: Fix broken waitall
1884 New functions and features
1885 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1887 Dropped / renamed functions and features
1888 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1889 - MSG_mailbox_is_empty() -> Mailbox::empty()
1890 - MSG_mailbox_front() -> Mailbox::front()
1891 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1892 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1893 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1894 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1895 * Don't pass the free_f parameter to property related functions:
1896 - MSG_host_set_property_value()
1897 - MSG_as_router_set_property_value()
1898 - MSG_storage_set_property_value()
1899 * VM properties. Since msg_vm_t are msg_host_t, just use
1900 MSG_host_get_property_value() and friends
1901 * VM I/O related things:
1902 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1903 - Unimplemented save/restore methods
1904 * MSG_as_router_get_property_value() was redundant with
1905 MSG_environment_as_get_property_value().
1906 - Removed MSG_as_router_*propert*() functions
1907 - Added MSG_environment_as_set_property_value() for consistency
1908 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1909 Well, they are still there for now, but deprecated with a warning.
1910 Please switch to xbt_dict_new_homogeneous() before this is removed
1912 * Task affinity. Its intended behavior (that was very badly tested
1913 and probably not really working) was deceiving what most users
1914 would have hoped here.
1915 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1916 * Option network/coordinates is now useless and should be dropped.
1919 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1922 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1923 * Added Dragonfly topology support
1926 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1927 EiBps, EBps, Eibps, Ebps for bandwidth.
1928 They may become useful to some lucky ones.
1931 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1932 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1933 * ex/async_waitAll: new example, on asynchronous communications
1936 * Memory usage should be decreased for simulations with a large number
1937 of processes. This also helps for SMPI.
1939 SimGrid (3.13) stable; urgency=low
1941 The Half Release, a.k.a. the Zealous Easter Trim.
1943 - We removed half of the lines, that were mostly experimental cruft.
1944 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1945 (not counting blanks and comments -- according to openhub.net)
1946 - The internals are now compiled in C++ (and will soon be clean C++)
1947 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1948 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1950 * Backwards Compatibility breaks
1951 - Removed Lua simulation bindings (switch to C or Java for that).
1952 Lua can still be used to describe platforms
1953 - Removed Java kernel plug-ins.
1954 Will be reintroduced after the ongoing major internals reorg.
1956 - the following functions were removed.
1957 They were too specific and should be reimplemented in a generic
1958 way, with filter function.
1959 - MSG_task_listen_from_host
1960 - MSG_mailbox_get_count_host_waiting_tasks
1961 - MSG_mailbox_put_with_timeout was removed.
1962 Please use MSG_task_send_with_timeout instead.
1964 - the SD_application_reinit function was removed. It has been a noop for a while.
1965 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1966 replaced by a more flexible mechanism.
1967 - The following functions thus do not exist anymore
1968 - SD_workstation_get_access_mode
1969 - SD_workstation_set_access_mode
1970 - SD_workstation_get_current_task
1971 - Basic estimation functions have been removed but can easily be replaced
1972 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1973 - SD_workstation_get_computation_time => amount / sg_host_speed()
1975 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1976 Use VM.getSpeed()*load/100 for the legacy behavior.
1978 - option enable_tracing was removed. It was not doing anything for a while.
1979 - In the ModelChecker:
1980 - the model-checker now ptraces the model-checked process which means
1981 you cannot use a debugger on the latter anymore (we might make this
1982 optional in the feature);
1983 - removed soft-dirty page tracking;
1984 - remove model-checked side snapshot management,
1985 MC_snapshot() and MC_compare_snapshot();
1986 - keep the MC_cut() function as a stub (it was not really working
1987 in the previous release).
1990 * All options are consistently in kebab-case. Old names are kept as alias.
1993 * Switch to platform v4 format.
1994 - Rename from 'power' to 'speed' the attributes describing the amount of
1995 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1996 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1997 - In <host> and <link>, attributes availability and state are gone.
1998 It was redundant with state and availability traces, and with peak values.
1999 - In <cluster>, attributes availability_file and state_file are gone.
2000 It was too complex and unused.
2001 - Kill <gpu>. Was not doing anything.
2002 - The DOCTYPE points to the right URL:
2003 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
2004 (the file at this address now documents the changelog since its v1)
2005 - A warning is emitted for unit-less values (they are still accepted).
2006 - speed. Default: 'f' or 'flops'. Also defined:
2007 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
2008 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
2009 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
2010 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
2011 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
2012 - latency. Default: 's' second. Also defined:
2013 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
2015 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
2016 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
2019 * s4u::Host is now the preferred public interface to the Host features.
2020 sg_host_* functions are C bindings to the exact same behavior
2021 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
2024 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
2025 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
2026 They were deprecated since a few years, and probably did not even compile anymore.
2029 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
2030 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
2031 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
2032 encouraged to update their codes. A list of the modified functions can be found at the end of
2033 include/simgrid/simdag.h
2036 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
2037 the lambda or closure passed as a parameter will run in kernel mode.
2038 All the callback functions should be rewritten to that interface at some point.
2041 * Reorganizing and cleaning the internals all around the place.
2044 * Remove old default barrier/bcast buggy algorithms (see #18407)
2045 * Various bug fixes to handle more codes
2046 * Remove the need for the --foreground option of smpirun (it is still
2047 accepted for backward compatibility).
2050 * Kill the setset data container: MC don't use it anymore.
2051 * Kill the queue data container: it made more sense with GRAS.
2052 * Kill the xbt_peer_t data type: it's useless without GRAS.
2053 * Kill rm_cb feature of config sets: it was never useful.
2054 * Kill graphxml parsing feature. It was not used.
2055 * Kill the deprecated code protected by XBT_USE_DEPRECATED
2057 - xbt_dynar_sort_strings(), when the content is char*
2058 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
2059 They throw exceptions on invalid input;
2060 * C++ support for declaring CLI flags (simgrid::config::Flag);
2061 * class for abstracting different signal backends (simgrid::xbt::signal).
2062 with no external dependencies (we need very simple signals).
2065 * refactoring and cleanup of the code;
2066 * ongoing process to cleanly separate the model-checking algorithms
2067 from the code model-checking support.
2069 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2071 SimGrid (3.12) stable; urgency=low
2073 The Facelift Release.
2076 * Require g++ v4.7 at least to not speak prehistorical C++.
2077 * Require Boost 1.48 (for signal2 component).
2078 * Java must be version 7 at least when activated.
2079 * Builds on Windows again (including Java bindings).
2080 * Tracing is now always enabled (no way to turn it out)
2081 * Remove GTNetS. It was not working anyway.
2082 * Various cleanups in the cmake scripts.
2083 * Move headers around to sort them out on installed systems:
2084 - instr/instr.h -> simgrid/instr.h
2085 - instr/jedule/* -> simgrid/jedule
2086 - simdag/datatypes.h was removed
2087 - simdag/simdag.h -> simgrid/simdag.h
2088 - msg/datatypes.h was removed
2089 - msg/msg.h -> simgrid/msg.h
2092 * Interface improvement:
2093 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
2094 - Sanitize the interface in MSG_task_ module:
2095 - Merge two functions that were close enough but misleading:
2096 set_compute_duration(t) -> set_flops_amount(t)
2097 get_remaining_computation(t) -> get_flops_amount(t)
2098 - set_data_size(t) -> set_bytes_amount(t)
2099 get_data_size(t) -> get_bytes_amount(t)
2100 - Massive cleanups in the functions related to the energy
2101 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
2102 - New: MSG_host_get_pstate()
2104 - msg/energy/onoff: switching hosts on and off
2106 * Interface improvement:
2107 - Rename Host.isAvail() to Host.isOn()
2108 - Rename Process.currentProcess() to Process.getCurrentProcess()
2109 - Rename Task.setDataSize() to Task.setBytesAmount()
2110 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
2112 - #18874: Actually allows the GC to reclaim tasks
2116 - SIMIX_process_throw: raises an exception in a remote process
2117 * Refactoring: Separate sub-modules
2118 - libsmx: the public interface, as libc in a real system
2119 - popping: the strange dance that converts a user request into a kernel handling
2120 - smx_context_*: the virtualization mechanisms that embed the user code
2121 - smx_*: the handling of each simcalls
2122 * Interface cleanup:
2123 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
2124 * Rename smx_action_t into smx_synchro_t, making explicit that these
2125 things are used to synchronize processes with their environment.
2126 For example, a communication is a sort of synchronization involving
2127 the communicating processes (that may block until the exchange) and
2128 the platform. The same can be said from computations, etc.
2130 - #18888: segfault when a process ends before its kill_time
2134 - 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.
2135 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
2136 - MPI_Type_set_name, MPI_Type_get_name
2137 - MPI_*_c2f and MPI_*_f2c functions
2138 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
2139 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
2140 - Activate a lot of new tests from the mpich 3 testsuite
2142 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
2143 - InfiniBand network model added: Based on the works of Jerome Vienne
2144 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
2145 - When smpi/display_timing is set, also display global simulation time and application times
2146 - Have smpirun, smpicc and friends display the SimGrid git hash version on --git-version
2147 * Collective communications
2148 - 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
2149 - 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/).
2150 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
2152 - Replay now uses algorithms from wanted collective selector
2153 - Replay can be used with SMP-aware algorithms
2154 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
2155 - 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.
2157 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
2158 - Correctly use loopback on fat-tree clusters
2159 - Asynchronous small messages shouldn't trigger deadlocks anymore
2160 * Energy/DVFS cleanup and improvement
2161 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
2162 - new: smpi_get_host_pstate()
2165 - "Full" network optimization flag was broken since Surf++
2166 - Better handling of precision flags in maxmin
2167 - Fix bug causing sometimes "Impossible" errors
2168 - Properly pass cluster properties to included hosts
2169 * Improvement of the Energy plugin.
2170 - Always update the consumption before returning that value
2171 - New property: watt_off to denote the dissipation when the host is off
2172 - New functions getWattMinAt and getWattMaxAt to retrieve the
2173 dissipation of pstates that we are not currently at.
2174 * Java: class NetworkLink renamed to Link
2175 * New function: simcall_process_get_kill_time()
2176 * Massive rename s/workstation/host/
2177 - That's intrusive, but that's good for the project consistency. Sorry.
2178 - Change config option "workstation/model" into "host/model"
2181 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
2182 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
2183 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
2184 - xbt_dirname and xbt_basename for non-POSIX systems
2186 * The model checker now runs as a separate process.
2187 * The model checker runs must now be launched with the new simgrid-mc program.
2188 * Record/Replay: the MC can display a textual representation of a path in the
2189 execution graph. It can then be replayed outside of the model checker.
2191 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2193 SimGrid (3.11) stable; urgency=low
2198 * Normalizing pointers addresses tool for better diff between logs
2201 * Add cloud examples using new VMs
2202 - examples/msg/cloud/two_tasks_vm.tesh
2203 - examples/msg/cloud/simple_vm.tesh
2204 - examples/java/cloud/cloud.tesh
2205 - examples/java/cloud/migration/migration.tesh
2206 * Add java surf examples:
2207 - examples/java/surfPlugin/surf_plugin.tesh
2208 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
2209 - examples/java/surfCpuModel/surf_cpu_model.tesh
2210 * Add SMPI+MSG example:
2211 - examples/smpi/smpi_msg_masterslave/
2216 - msg task destroy cancel test
2217 - msg_host on/off test
2218 * Move all tests in testsuite to teshsuite (adding tesh files)
2219 * Restructure teshsuites
2220 - one folder for each kind of test
2221 * Restructure AddTests.cmake
2223 - structure the order of tests (with sections)
2226 * Add virtual machine
2227 - creation of a VM on a PM
2228 - migration of a VM from a PM to another PM
2230 - MSG_process_join(msg_process_t process, double timeout)
2231 - msg_bar_t MSG_barrier_init(unsigned int count)
2232 - int MSG_barrier_wait(msg_bar_t barrier)
2233 - void MSG_barrier_destroy(msg_bar_t barrier)
2234 - msg_as_t MSG_environment_as_get_by_name(const char * name)
2235 * New option "msg/debug_multiple_use" to help debugging when a task is used
2238 - Improvements and finalization of MSG_storage, MSG_file APIs and their
2240 - Increase code coverage in test suites
2244 * Protect context stacks against stack overflow. The number of protected memory
2245 pages allocated on the top of each stack (1 by default) can be configured
2246 with the new command line option --cfg=contexts/guard_size:<value>.
2247 * Simcalls are now generated by a python script that
2248 - generates files included by SimGrid
2249 - checks that all the functions exist, and proposes prototypes
2251 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
2252 host_set_data, host_get_data
2254 - simcall_process_join(smx_process_t process, double timeout)
2255 * Fix bug where sleeping processing could not be suspended.
2258 * Translate surf models from C to C++
2259 - Generic classes for all models: Model, Resource, Action
2260 - A generic interface for each kind of model (CPU, Network, Storage
2261 Workstation, WorkstationVM)
2263 * Translate surf routings from C to C++
2264 * Add callbacks using sigc++ or boost::signals2
2265 - Add callback functions for resource creation/destruction
2266 - Add callback functions for action state change
2267 - Handle Energy as a plugin
2268 * Replace swag by boost::intrusive
2269 * Add new routing models for clusters. For documentation, see
2270 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
2271 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
2272 parameters for cluster tag
2273 - Fat trees, with topology="FAT_TREE" and
2274 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
2275 - see examples/platforms/torus_cluster.xml and
2276 examples/platforms/fat_tree_cluster.xml
2277 * More documentation
2280 * Hostfiles support host:nb_processes construct to deploy several processes on
2282 * Collective communication algorithms should not crash if used with
2283 improper number of nodes and report the error.
2284 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
2285 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
2286 MPI_Dims_create, MPI_Cart_sub are supported.
2287 * New interface to use SMPI programmatically (still depends on MSG for
2288 some parts, see examples/smpi/smpi_msg_masterslave):
2289 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
2293 * Global variables privatization in MPI executables is now performed at runtime
2294 with the option smpi/privatize_global_variables (default:no).
2295 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
2296 dynamic libraries loaded by the application are not privatized (static
2297 linking with these libraries is advised in this case)
2300 * Options defined in XML work correctly now.
2303 * New cmake option, enable_lib_in_jar, to control whether native libraries are
2304 copied into simgrid.jar or not (ON by default). Use this option if you want
2305 to reduce the size of the installed simgrid.jar, *and* the native libraries
2306 are kept installed elsewhere.
2307 * Surf binding with SWIG (code generated in maintainer mode only):
2308 - plugin to handle callbacks
2309 - CPU model only for the moment
2312 * Supernovae build mode is definitively removed. It was used to improve
2313 inlining and inter-module optimizations. It is nowadays superseded by
2314 link-time optimizations commonly available in compilers.
2315 * Update ns-3 find lib. Bindings for ns-3 should work again now.
2316 * Add boost dependency for surf++
2317 * Add new macro for tests
2318 - ADD_TESH(name <tesh_args>)
2319 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
2323 - xbt_bar_t XBT_barrier_init(unsigned int count)
2324 - int XBT_barrier_wait(xbt_bar_t barrier)
2325 - void XBT_barrier_destroy(xbt_bar_t barrier)
2326 * Make the xbt_os_time module public
2328 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2330 SimGrid (3.10) stable; urgency=low
2332 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
2335 * Reintegrate Java to the main archive as desynchronizing these
2336 package is not acceptable anymore (Java is now considered stable)
2337 * Add explicit synchronization facilities through semaphores
2338 * Bug fix: Task.setDataSize() only changed the C world, not the value
2339 cached in the Java world
2342 * Dramatically change the way files are handled. API and internals changed, but
2343 this part of MSG was not considered as production grade either.
2344 * Add explicit synchronization facilities through semaphores
2345 * Add a new function MSG_host_get_process_list()
2346 * Preliminary DVFS support (see examples/msg/energy/ for details)
2349 * SMPI is now included directly in the libsimgrid as the windows
2350 linker doesn't force us on splitting it anymore.
2351 * Improvements of the SMPI replay tool:
2352 - Most of the collective communications are now rooted in the same process as
2353 in the original application.
2354 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
2355 used until now). Multiple data types can now be used in a trace.
2356 - The replay tool now supports traces produce either by TAU or a modified
2358 - Bug Fix: the compute part of the reduce action is now taken into account.
2359 - Gatherv collective is now supported
2360 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
2361 of smpirun outputs time independent traces for the current run. One file
2362 is created per process. If too many processes are simulated, this behavior
2363 can be changed to one file for all processes by using the
2364 tracing/smpi/format/ti_one_file flag
2365 * smpirun generates the host file if needed (with given host count and platform)
2366 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
2367 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
2368 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
2369 --cfg=smpi/coll_selector:(mpich/ompi)
2370 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
2372 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
2373 happening inside a collective SMPI call.
2374 * Fix the behavior of complex data types handling.
2375 * Make MPI_Wtime another synchronization point to take computations into
2377 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
2378 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
2379 * Add all missing Fortran bindings, SMPI should work with Fortran 90
2380 (no privatization of global variables yet)
2381 * Preliminary DVFS support (see examples/smpi/energy/ for details)
2384 * Verification of liveness properties is now available for SMPI applications
2385 (in addition to MSG applications)
2386 * Bugged examples using SMPI in examples/smpi/mc/
2387 * Add --cfg=model-check/visited option. Allows the verification of infinite
2388 programs. Detection of loops in the execution thanks to the system state
2389 comparison and reduction of the state space to explore. Can be combined with
2390 DPOR for safety properties.
2393 * Allow to change SimGrid configuration (see --help) within the code
2394 thanks to SD_config() as it can be done in MSG.
2395 * Add a new function SD_task_set_amount() upon user request.
2398 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
2399 * Remove rule based routing (no more PCRE dependency)
2400 * Add a limiter_link option to cluster tag, to specify a maximum reachable
2401 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
2402 * Add a loopback_bw and loopback_lat options to cluster tag.
2403 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
2404 routing. Now peers are encapsulated in an AS and have their own private
2405 router but this is transparent.
2408 * Our own implementation of getline is renamed xbt_getline, and gets
2409 used even if the OS provide a getline(). This should reduce the
2410 configuration complexity by using the same code on all platforms.
2411 * New type: xbt_cfg_elm_boolean.
2412 * Allow to use yes/no for boolean configuration options in the command line.
2413 * Allow to disable SimGrid cleanups at exit from command line option.
2414 There are situations where one may want a simulation to end with an exit.
2415 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
2416 annoying when scripting around the simulator. Adding a
2417 --cfg=clean_atexit:no allows to circumvent this issue.
2420 * Lots of memory leaks were corrected in this release.
2421 There is no known memory leaks anymore, in all of our 600+ tests.
2422 * New command line option --version, to get SimGrid version information.
2423 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
2425 * Supernovae builds are deprecated, and expected to be removed in the next
2428 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2430 SimGrid (3.9) stable; urgency=low
2432 The Grasgory release: GRAS is really dead now.
2434 * Complete overhaul of the internal host structures scheme.
2437 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
2438 was considered as experimental and badly maintained since then.
2439 * Keeping it was thus a trap to our potential users, that could take
2440 it instead of MSG or SMPI by mistake despite is pity state.
2441 * GRAS seems to have very few users (if any), and no one volunteered
2442 to maintain it further. It also induces a lot of XBT code (for
2443 portability sake), that must be maintained too.
2444 * For all these reasons, we killed GRAS. If someone wants to revive it
2445 in the future, don't cry, our git history still remembers of GRAS.
2448 * Major overhaul. Merge our documentation again as time proved that
2449 splitting it was really not helping our users.
2450 * Further improve the developer documentation to help newcomers
2451 hacking on SimGrid itself. The user documentation (and in
2452 particular, the beginner documentation) is still in a sorry state.
2455 * Now works on Windows too!
2456 * Much more extensive test suite, from MPICH
2459 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
2460 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
2461 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
2462 * Bug fix: let task be scheduled when the last dependency to be solved is
2463 a control dependency.
2464 * Remove SD_load_environment_script function.
2465 Use the C sg_platf function if you want to declare a platform
2469 * New function: MSG_process_get_number()
2470 * Old function documented: MSG_config()
2471 * Remove MSG_load_platform_script function
2472 Use the C sg_platf function if you want to declare a platform
2476 * Change the default value of the TCP_gamma constant (maximal size of TCP
2477 congestion window) to a more realistic 4MiB value. If you notice changes in
2478 your simulation results, you can fall back to the previous 20k tiny window
2479 by adding --cfg=network/TCP_gamma:20000 on command line.
2480 * (Hopefully) fix a bug wrt periodic availability and state traces
2481 * Bug fix: use default values at start when first event in availability/state
2482 trace is not at time 0.
2485 * remove the "new_" part of function name sg_platf_new_trace_connect
2486 (resulting in sg_platf_trace_connect), since it does not create
2490 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2491 use case where it's really mandatory, and maintaining it was a pain in
2493 * New: xbt_fifo_search(), search an item with a user-provided
2494 comparison function instead of dumb pointer comparison.
2497 * Fix the lua deployment:
2498 Use `simgrid.init_application()` before deployment instead of
2499 `simgrid.msg_register_application()` after.
2502 * Transfer the tracing files into the corresponding modules.
2504 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2506 SimGrid (3.8.1) stable; urgency=low
2508 The "we are told that some people want to also *install* the simgrid
2511 * Add missing file "tesh.1" to the archive.
2513 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2515 SimGrid (3.8) stable; urgency=low
2517 The Psssshiiiit release: SimGrid jumps into the Cloud.
2520 * Add an experimental interface to manipulate VMs. They are mainly
2521 process groups with very few intrinsic semantic, but they should
2522 allow you to build the semantic you want easily.
2523 * New function: MSG_host_set_property_value()
2524 * New function: MSG_process_on_exit(). To clean memory in all cases.
2525 * Bug fixes that made the host (and link) failures unusable.
2526 * Add a way to auto-restart process when the host in which they are
2527 executing comes back (ON_FAILURE="RESTART" on deployment file,
2528 MSG_process_auto_restart_set).
2529 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2530 please stop using the old ones, they are DEPRECATED.
2532 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2533 Please use MSG_init() instead. (reducing the amount of entry
2534 points in the library helps us).
2535 * Make it impossible to link against the wrong version of the lib
2536 * Deprecate MSG_clean(). No need to call it anymore.
2537 * Function MSG_get_host_number() is not deprecated anymore.
2540 * Split the doc into a user guide and a reference guide.
2541 * Start a developper guide to help people hacking on SimGrid.
2544 * Enable tracing by default. This modules rocks you should use it.
2545 * Remove option custom_flags. Now use environment variables CFLAGS
2547 * Use default cmake things to detect lua instead of home grown ones.
2548 * New option "enable_mallocators" to disable mallocators, for debugging
2549 purpose ("on" by default).
2552 * Bug fixes around the resource failures: don't let the processes
2553 survive the host they are running onto.
2554 * Add an interface to auto-restart processes when the host in which they are
2555 executing comes back.
2556 * Ensures that SIMIX_clean is called automatically. It's not part of
2557 the public interface anymore (bindings should be updated).
2560 * Bug fix for when SD_Simulate is called with a positive value: be careful
2561 when comparing doubles. Sometimes they are different for non significant
2563 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2564 parallel task whose initial work is distributed among host according
2565 to the Amdahl's law. Such tasks are created with a parameter alpha
2566 that corresponds to the non-parallelizable part of the computation.
2567 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2568 between two sets of workstations assuming a 1D block distribution (each
2569 workstation owns a similar share of data) on both sides.
2571 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2572 Data redistribution will be automatically scheduled once parent and child
2573 are both scheduled. The filling of computation_amount and
2574 communication_amount structures is now done seamlessly thanks to the chosen
2576 * New function SD_workstation_dump to display various information
2577 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2578 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2579 bandwidth on the route onto which the task is scheduled and the amount of
2581 To divide the nominal bandwidth by 2, the rate then has to be:
2582 rate = bandwidth/(2*amount)
2583 * Compute tasks that have failed can now be rescheduled and executed again
2584 (from their beginning)
2585 * Increasing source code coverage (src/simdag is now covered at 95.8%
2589 * Re-implement time-independent trace replay using SMPI (at the
2590 smpi_smp_* level) instead of MSG. This should replace
2591 examples/msg/actions/actions.c
2592 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2593 hindexed and structs)
2594 * Implement the exchange of non-contiguous data.
2595 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2596 * Correct behavior of smpi/sender_gap and set its default value to 0
2597 * Add option to asynchronously send small messages to allow better
2598 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2599 specifies the size in bytes under which messages will be asynchronously sent.
2600 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2601 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2604 * New C interface to define a platform: XML is now optional.
2605 For more info, please check include/simgrid/platf.h.
2606 * New interface to define random platforms from the C:
2607 For more info, please check include/simgrid/platf_generator.h and
2608 examples/msg/masterslave/masterslave_platfgen.c
2609 * Export a sg_cmdline dynar containing all the arguments we got from
2613 * Two new tracing options for adding comments to trace file so you
2614 can track your experiments (see --help-tracing for details).
2615 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2616 * Adding the SimGrid version that generated the trace file as a comment.
2617 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2618 * Fix to avoid key clashes on Paje links
2619 * Other minor fixes related to the Paje specification
2622 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2623 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2624 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2625 * Greatly improve the robustness of mmalloc to user errors (such as
2626 using an area after freeing it, or freeing it twice)
2628 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2630 SimGrid-java (3.8.1) stable; urgency=low
2632 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2633 Warning: all methods are visible, but only some of them are
2634 implemented so far. Check the source (src/jtrace.c)
2635 for further information.
2636 * New module: org.simgrid.msg.File (SimGrid File management functions)
2637 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2638 * Change the meaning of Process.restart: now restart the process from
2639 the beginning, like MSG_process_restart in C.
2640 * Add Process.setAutoRestart: handling of process restart when failed
2642 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2643 you to retrieve the properties of the processes/hosts
2644 * Deprecate Msg.clean(): you can just forget about it now.
2645 * New function Process.getCount(), that only works when compiling
2646 with the not yet released version 3.9 of the C library.
2648 * New context factory based on Coroutines. It mandates a modified JVM
2649 but then, the simulations run about five times faster, and there is
2650 no limit to the amount of processes (beside of the available memory).
2652 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2654 SimGrid (3.7.1) stable; urgency=low
2657 * Restore the prototype of MSG_process_create_with_environment() to
2658 the pre-3.7 situation by removing the kill_time argument.
2659 * Add a MSG_process_set_kill_time() function instead.
2662 * Fix weird behaviors when dealing with parallel tasks.
2665 * Simgrid is now built as a dll.
2666 * Simgrid-java now works on Windows.
2667 * Simgrid-Java is now included into Windows package.
2670 * First pre-build package for MacOSX.
2673 * Fix compilation when using MSG_USE_DEPRECATED.
2674 * Fix some compilation issues on Macs and Windows.
2675 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2676 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2678 We discovered that the Lua console is broken, but we are missing the
2679 manpower to fix it right now. The problem existed in 3.7 too, so we
2680 are not blocking the release for that. Sorry if you depended on this
2681 feature, any help would be really welcome.
2683 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2686 SimGrid-java (3.7.1) stable; urgency=low
2688 The "Java aint got to be bloated and slow" release
2691 * Various internal cleanups and performance improvement
2692 Simulations are expected to run up to twice faster or so
2693 * Make Process.kill(process) an instance method, not a static one
2694 * User processes are not java.lang.Thread subclasses.
2695 This breaks the compatibility (sorry), but previous API was
2696 brain-dead, making it impossible to have non-trivial
2697 initializations in the process constructor.
2698 * Require a full constructor per Process sub-class.
2699 Kinda breaks the compatibility (sorry), but this allows a much more
2700 efficient way to launch the processes at simulation startup.
2701 * Do not embeed our version of semaphores, java 1.5 can be considered
2702 as sufficiently prevalent for us to not dupplicate its features.
2706 * Add examples for almost every part of the API
2707 We spotted and fixed a lot of bugs in the process
2709 * New module: asynchronous communication API
2710 * New function: Process.sleep()
2711 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2712 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2713 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2715 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2717 SimGrid (3.7) stable; urgency=low
2719 The "spring cleanups (before the next Big Project kicks in)" release.
2722 * We can specify the SMPI latency/bandwidth factor with command line
2723 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2724 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2725 You can also use the "config tag" from platform file by adding this line
2726 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2727 (see "example/platforms/tag_config.xml" to use "config tag").
2728 Note that the command line supersedes the platform file configuration.
2729 * Change the correction factors used in LMM model, according to
2730 the latest experiments described in INRIA RR-7821.
2731 Accuracy should be improved this way.
2732 * Use the partial invalidation optimization by default for the
2733 network too. Should produce the exact same results, only faster.
2734 * Major cleanup in surf to merge models and split some optimization
2735 mechanisms from the core of the models. As a result you can now
2736 specify which model to use (e.g., --cfg=network/model:LV08
2737 --cfg=cpu/model:Cas01) and which optimization mode to use
2738 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2739 Incompatible combinations should err at initialization. See
2740 --help-models for the list of all models and optimization modes.
2741 * The CLM03 workstation model was dropped for simplicity because it
2742 used the deprecated CM02 network model. Use default instead.
2743 * Rename the TCP_gamma configuration option to network/TCP_gamma
2744 * Rename the coordinates configuration option to
2745 network/coordinates, and document it
2746 * Use now crosstraffic keyword instead of the terribly misleading
2747 fullduplex keyword. It is activated by default now in the current
2748 default model, use --cfg=network/crosstraffic:0 to turn it off.
2749 * Ongoing refactoring the model parsing to make XML files optional
2750 See include/simgrid/platf.h for details (still to be completed)
2753 * Major overhaul of the documentation. Almost instructive now :/
2754 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2755 functions and friends. This interface was considered as
2756 deprecated since over 2 years, it's time to inform our users that it is.
2757 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2758 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2759 use these (crappy) functions in your code.
2760 These functions will be removed soon. Stop using them now.
2761 * Deprecate MSG_get_host_{table,number}
2762 Implement MSG_hosts_as_dynar() instead.
2763 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2764 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2767 * Stabilize the parallel execution mode of user contexts
2768 * Introduce configuration variables to control parallel execution:
2769 - contexts/synchro: Synchronization mode to use when running
2770 contexts in parallel (either futex, posix or busy_wait)
2771 - contexts/parallel_threshold: Minimal number of user contexts
2772 that must be part of a scheduling round to switch to parallel
2773 execution mode (raw contexts only)
2774 * Fix bugs that prevented to use suspend/resume along with
2775 synchronization structures.
2776 * Fix bugs in process termination that lead to invalid memory access
2777 in very specific conditions.
2780 * Introduce a parallel mode for the models (controlled by surf/nthreads
2781 configuration item). In our tests, running the models in parallel
2782 never lead to any speedups because they are so fast that the gain
2783 of computing each model in parallel does not amortizes the
2784 synchronization costs, even when ultra fast futexes are used.
2785 This is released anyway because YMMV.
2788 * Performance boost by using a swag internally to compute the set of
2789 tasks that are finished and should constitute the return value of
2793 * Enable it by default now that it is considered rather stable.
2796 * Documentation of the tracing functions.
2797 * Performance gains when tracing categorized/uncategorized resource
2798 utilization by avoiding calls to get route when updating resource
2799 variables. LMM constraints are being used instead.
2800 * API changed to set task categories. Use MSG_task_set_category instead
2801 of TRACE_msg_set_task_category, and SD_task_set_category instead
2802 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2803 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2804 * Examples for link user variables added (see at examples/msg/tracing/)
2805 * Deprecated function TRACE_msg_set_process_category completely removed
2806 * Trace header updated according to the latest Paje file format
2807 * Tracing network lazy updates, no longer obligate users to use full updates
2808 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2809 * Experimental: let user code declare/set/push/pop application states for hosts
2810 * API changed to allow the manual creation of graph configuration files
2811 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2814 * Improve the API of Lua MSG bindings, using the Lua spirit.
2815 * Each simulated process now lives in its own Lua world (globals are
2816 automatically duplicated). It helps writing simulators. Will allow
2817 to run Splay programs within SimGrid in the future.
2818 * Add a Chord example in Lua, equivalent to the MSG one.
2821 * Start the implementation of a solution to express temporal
2822 properties, not only local assertions. This is still an
2823 experimental work in progress, stay clear from it to be safe.
2827 - Add new runtime parameters --help-logs and --help-log-categories
2828 to display information about supported logging parameters and
2830 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2831 don't exists anymore.
2832 * Mallocators: allow value NULL for the reset function.
2834 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2835 create homogeneous dictionaries, where all the elements share the
2836 same free function. Non homogeneous dictionaries will be
2837 deprecated in the next release.
2838 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2839 - Multi-level dictionaries are deprecated.
2841 - new function xbt_dynar_search_or_negative() that is useful when
2842 you have less than 2 million elements in your dynar and don't
2843 want of the extra complexity of catching exceptions when the
2844 element is not found.
2846 - Make xbt_os_thread module (for thread portability) public.
2847 Documentation is still to come, sorry.
2849 - Cleanups and simplifications to make it maintainable again.
2850 - Exotic features (such as memalign and valloc) were removed.
2851 - The metadata were extended and improved so that the
2852 model-checker becomes able to explore and inspect the heaps.
2853 - This may induce a performance drop when enable_model-checking is
2854 ON in cmake (even if it's not used in the simulation), but it is
2855 necessary at this point to get MC working.
2857 Turn model-checking OFF if simulation performance matters to you.
2858 Not enabling it at runtime is not enough, disable it in cmake.
2860 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2863 SimGrid (3.6.2) stable; urgency=low
2865 The "Not coding new stuff allows to polish old things" release.
2868 * New bindings to the NS3 packet level simulator (experimental)
2869 * Use the raw (efficient) execution contexts instead of the sysv
2870 (portable) ones when possible.
2871 * libpcre is now mandatory in any cases since not using it led to
2872 severe performance loss and possibly other issues
2873 * Update the XML platforms:
2874 - G5K: include the latest machine in Nancy
2875 - GridPP and LCG: new platforms
2876 * Documentation was partially updated, at least (more to come)
2878 Bug fixes, cosmetics and small improvements
2879 * Free terminated processes before the end of the simulation to avoid
2880 exhausting the memory of users having very dynamic amount of
2882 * Bug fix and cosmetics about canceling non-running tasks
2883 * Bug fix about the dot loader's issues when using libcgraph
2886 * Create an installer for windows with nsis (amd64 and win32)
2887 - Add an hello world project to illustrate SimGrid project creation.
2888 - Embed libpcre into the SimGrid installer to avoid
2889 its compilation burden
2890 * The raw execution contexts should work on Apple now
2891 * Port to Windows 64 bits
2892 - Sysv contexts now have an implementation for this arch
2893 - GRAS communication features now support this arch
2894 * Drop support for borland compiler on windows
2895 - this code was not maintained, and we kinda depend on gcc nowadays
2896 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2897 * Fix portability issue on unstable ubuntu: linker became picky on
2900 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2902 SimGrid (3.6.1) stable; urgency=low
2904 The "Oops, we broke Macs too" release
2907 * Fixed contexts detection so that raw ones are used when possible
2908 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2909 leads to a strange error, with user code segfaulting sometimes when
2910 the generated code is not perfectly aligned (which is not
2911 controllable from the user side, depends on the amount of code)
2914 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2916 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2918 SimGrid-java (3.6) unstable; urgency=low
2921 * Split of every thing from SimGrid v3.5 into a separate package.
2923 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2925 SimGrid (3.6) stable; urgency=medium
2927 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2930 * Bindings now constitute their own package, separated from the main one.
2931 Rationale: reduce our maintenance nightmare by reducing the module coupling
2932 They will soon be released on their own on gforge.
2934 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2935 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2937 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2938 * It's not quite deprecated for now because we have no replacement,
2939 but it may soon become the case.
2942 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2943 * Fortran: New user-level cache variable to store the rank of the running
2944 process. This improves performance by an order of magnitude.
2945 * C: New coccinelle script to automatically locate and modify global and
2946 local static variables.
2947 * Improved SMPI network model with a sender-side gap to account for multiple
2951 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2952 only say if a communication is finished, no matter whether it succeeded or
2953 failed. You can call MSG_comm_get_status() to know the status of a finished
2955 * New function MSG_task_dsend() to send a task and detach it. When a
2956 communication is detached, you are never notified of its success or failure
2957 and the memory is released automatically once it is finished. This function
2958 is useful when you don't care about the end nor the success of a
2960 * Change the prototypes of action replay. Sorry for inconvenience,
2961 but this is really more efficient this way (and to adapt your code,
2962 you just have to fix the initialization, that shouldn't be too long)
2963 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2964 ever managed to do anything useful with it.
2965 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2966 takes a dynar of communications. It returns immediately and gives the
2967 index of a finished communication (if any).
2968 * New example: a basic implementation of the Chord P2P algorithm.
2971 * New model for multi-core CPUs. You can now use the core attribute to
2972 precise the number of cores of a host. This is a basic model. Every
2973 process running on the host receives at most the power provided in
2974 the DTD (throughput<=power). Total throughput of process cannot exceed
2976 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2977 router linked by an up-link and a down-link (possibly asymmetrical).
2978 This kind of pattern allows to easily build last-mile model style platforms.
2979 Aggregating such patterns in a rule-based AS is thus the technique of
2980 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2981 * New model for Vivaldi routing. We transformed the Vivaldi network model
2982 into a Vivaldi routing model (based on the rule-based model). This allows to
2983 combine Vivaldi based latencies with last-mile platforms.
2986 * Added a check for NaN of IEEE754 infinite in the double entries of
2988 * Introduce a new context factory "raw", highly inspirated from the
2989 ucontext factory, but using manually crafted functions in assembly to
2990 do the work in an efficient manner.
2991 * Allow to change the used context factory at run time, not only at
2992 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2993 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2994 contexts at runtime (only with raw contexts or ucontexts).
2995 * Completely rewrote this module to allow parallel execution of user
2996 processes. Use --cfg=contexts/nthreads:N to execute user processes
2997 with N parallel threads (the default is 1, meaning no parallelism).
2998 * Allow to decide dynamically between sequential and parallel modes.
2999 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
3000 processes in parallel only when their number is greater than or equal to P
3002 * Added a check for NaN of IEEE754 infinite in the double entries of
3006 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
3007 won't output the state of processes when interrupted with Ctrl-C
3008 * Add a new function xbt_dynar_to_array that transforms a dynar into a
3009 NULL-terminated array. This may solve backward compatibility issues
3010 due to the change to return type of SD_simulate. See also:
3011 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
3012 * Add new macros with variable number of arguments.
3013 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
3014 - in xbt/asserts.h: xbt_assert
3015 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
3016 - in xbt/ex.h: THROWF and RETHROWF.
3017 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
3019 * Change xbt_die() to accept a format string with arguments, just like printf.
3020 * New data structure: xbt_lib_t, like a dict but more general and with better
3024 * New configuration options
3025 Options triva/categorized and triva/uncategorized can be used to generate
3026 graph configuration files for Triva visualization tool.
3027 * Configuration option tracing/platform is renamed to tracing/categorized
3028 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
3029 * New TRACE_user_link_variable interface:
3030 User provides the name of the link and the tracing variable to attach to it
3031 * the declaration of tracing categories must be done after the environment
3033 * simpler tracing interface, just one way to declare categories
3034 TRACE_category or TRACE_category_with_color, it is up to you
3035 * links in the trace file are again identified by their names
3036 * trace contains the full platform hierarchy exactly as declared using the ASes
3037 * Options tracing/msg/[task|process]:1 groups the process by hosts
3038 for both cases, tasks and processes must have names that are unique during the simulation
3039 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
3040 * The experimental option tracing/msg/volume is deprecated
3041 its functionality may be reincorporated if needed
3043 The tracing generates a trace file with unordered timestamped events,
3044 because of the way the core simulator (surf) works. A script available
3045 at the tools directory (fix-paje-trace.sh) can be used to put the events
3046 in order. We have changed the tracing so it can generate ordered timestamped
3047 events in the final trace, but depending on the simulator (and how much time
3048 is simulated) that can lead to a huge memory utilization. It is deactivated
3049 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
3051 Build Infrastructure
3052 * Define a SIMGRID_VERSION macro in simgrid_config.h.
3053 - We are trying hard to keep the API stable, but it may happen that
3054 some things change (we're a research project after all, not a
3055 nuclear plant operating system). If such things should happen, you
3056 could rely on that macro to adapt.
3057 - current value: 30600 for 3.06.00, aka 3.6
3058 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
3059 a number that can be compared with SIMGRID_VERSION.
3060 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
3061 assertions and verbose/debug logging events are disabled at compile time.
3063 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
3065 SimGrid (3.5) stable; urgency=medium
3068 * New feature to any SimGrid-based simulator: Model-Checking
3069 Check SIN#1 for more details.
3072 * New Model SMPI (three-interval linear regression for correction factors)
3073 See RR-7426, available at http://hal.inria.fr/inria-00527150
3074 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
3075 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
3076 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
3077 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
3078 * See also src/smpi/README
3083 - Tracing API changes: TRACE_start and TRACE_end should not be called
3084 by user-code. They are automatically called by simulators created
3085 with SimDAG, MSG and SMPI if the toolkit is compiled with
3086 tracing_enabled=ON. Categories declaration and utilization remain the
3087 same for MSG and SimDag.
3088 - A function was added to the tracing API to declare categories with
3090 - TRACE_category_with_color (char *category, char *color)
3091 where color must be in the following format
3092 "%f %f %f", red, green, blue
3093 and red, green, blue are float values in the interval [0, 1]
3094 - User can specify NULL as color parameter, or continue calling
3095 TRACE_category (cat)
3096 On that case, the tracing system will define random colors
3097 - The following command-line options are supported:
3098 --cfg=tracing/filename:msg.trace
3099 --cfg=tracing:1 (activate tracing, needed to use others)
3100 --cfg=tracing/platform:1 (categorized resource use)
3101 --cfg=tracing/uncategorized:1 (uncategorized resource use)
3102 --cfg=tracing/msg/task:1 (task creation)
3103 --cfg=tracing/msg/process:1 (process creation, migration)
3104 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
3105 --cfg=tracing/smpi:1 (SMPI interface tracing)
3106 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
3107 - examples of examples/msg/tracing updated
3109 - DAXLoader and DOTLoader functions can generate tasks with categories
3110 - A new function to attribute a category to SD tasks:
3111 TRACE_sd_set_task_category (SD_task_t task, char *category)
3112 * Tracing the MPI interface implemented by SMPI
3113 - Collective operations are traced with states
3114 - Point-to-Point operations are traced with states/links
3115 - Tracing activated by a parameter "-trace filename" passed
3116 to smpirun during execution (considering that simgrid
3117 is compiled with tracing enabled)
3118 - To run the simulation with gdb, the simulator
3119 accepts --cfg=tracing/smpi:1 to trace SMPI
3120 - tesh files to check if smpi tracing is ok
3121 - See examples/smpi/NAS/DT-trace
3122 * GTNetS tracing re-worked
3123 - adaptation to the tracing system of GTNets to cope
3124 with modifications regarding the fullduplex mode
3125 - new tesh files to check if gtnets tracing is ok
3128 * Asynchronous communications through the functions:
3129 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
3130 * New function: MSG_load_platform_script()
3131 to make possible using a lua script instead of XML files to set up platforms
3132 * New function: MSG_set_function
3133 to associate functions to processes, used when bypassing the parser
3134 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
3136 Platforms: Add some more examples in examples/platforms
3137 * Grid'5000: see www.grid5000.fr
3138 * *_30000_hosts.xml: various huge files [mainly scalability testing]
3141 * Change the XML format. This is a very important modification. SimGrid
3142 3.5 introduces a new hierarchical format based on the notion of
3143 Autonomous Systems. Compatibility with old format is ensured
3144 through the perl script provided in the install bin directory
3145 bin/simgrid_update_xml.
3146 It is now possible to build platforms with specific routing
3147 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
3148 connect several platforms together. We will try to provide soon
3149 set of realistic platforms exploiting these properties (have a
3150 look at examples/platforms/ for the moment).
3151 * Take the opportunity of the XML format change to be a good XML citizen:
3152 rename link:ctn to link_ctn and similar changes (also dealt with by
3154 * Add a new routing scheme (rule-based) using regular expressions. It
3155 enables to have an extremely low memory footprint when the
3156 underlying routing is simple and can be compactly described. You
3157 need to have libpcre4-dev (perl regular expressions) installed if
3158 you want to use this routing scheme.
3159 * Revive the cluster TAG and allow to easily and efficiently (both in
3160 term of memory and speed) connect clusters together. Have a look
3161 at teshsuite/simdag/platforms/ to see how this can be done. With
3162 this tag, you can create clusters with thousands of tasks at no
3163 cost (have a look at examples/platforms/).
3164 Note: clusters are implemented as ASes, so there is no need for an
3165 enclosing AS tag if you have only one cluster in your platform.
3166 * Add new generic functions in the public interface that allows the user
3167 to call SURF 'create_resource' methods from your code (same
3168 functionality as the XML bypass mechanism but with a much lighter
3170 * Add a new model (enabled through command line --cfg=network/model:SMPI)
3171 that uses a piecewise linear approximation to produce better
3172 results when exchanging small messages.
3173 * Add a new parameter to handle correctly full duplex link and account
3174 for interferences between uplink and downlink communications
3175 (activate with --cfg=fullduplex:1).
3178 * Rename the SD_READY (all dependencies are satisfied and task is
3179 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
3180 dependencies are satisfied) state.
3181 This prevents a confusion between the notion of "ready to schedule"
3182 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
3183 simulated" (SD_RUNNABLE) used by the simulation kernel.
3184 * Change the way a task is considered as ready. Instead of removing
3185 dependencies when a task is done, a counter is decreased. This way,
3186 it is always possible to reach ancestors thanks to the
3187 SD_taks_get_parents function (even after the end of the simulation.)
3188 * Change the return type of SD_Simulate from (SD_task_t*) into
3189 xbt_dynar_t. This function was in handling a dynar internally and
3190 converted it into a NULL terminated array for historical reasons.
3191 * New function SD_dotload(char*) to load a DAG described in dot
3192 format. This loader and the corresponding examples require the
3193 installation of the graphviz library.
3194 * Fix a bug in the management of tasks of size 0 in the surf network
3195 models. This problem was only visible with SIMDAG and you should
3196 thus disregard results produced with earlier versions if you
3197 relied on this feature (some tasks were blocked because of this).
3198 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
3199 even though your applications were doing only point-to-point
3200 communications and sequential computations. Now you can really use any
3201 model you want (of course, if you create real parallel tasks, which are
3202 not implemented in most models beside ptaskL07, this will abort).
3203 * Add an example that schedules a DAX on an heterogeneous platform
3204 using a Min-Min strategy.
3205 * New function SD_workstation_get_current_task() that returns the kind
3206 of task currently running on a workstation in the sequential access
3208 * Raise some warnings when unexecuted tasks remains at the end of the
3209 simulation. This is usually caused by cycles in the DAG.
3212 * New function: SIMIX_process_set_function() called by MSG_set_function
3213 * Change the underlying waiting queue in semaphores so that a process
3214 can wait on several of them simultaneously (as in waitany).
3215 * Fix the way to handle tokens in semaphores so that all access patterns
3216 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
3217 * kill the dirty pimple SIMIX_message_sizes_output()
3218 Please use (proper) visualization instead
3221 * New data container: setset (set of sets of elements)
3222 * New module: mmalloc (mapped malloc, allowing to have several
3223 independent segments of malloc)
3224 * New function: xbt_dict_cursor_set_data()
3225 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
3226 * New function: xbt_dynar_is_empty()
3227 * New function: xbt_fifo_get_last_item()
3228 * Fix xbt_dynar_shrink(): use the right element size.
3229 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
3230 uninitialized areas during expand.
3231 * Fix semaphores: previous implementation was severely broken.
3232 * Use library init/fini functions for our initialization.
3233 - you can use logs and other feature as soon as you want in your
3234 code (even before the xbt_init / MSG_init)
3235 - xbt_exit is now a no-op and produce a warning when used.
3238 * Port GRAS to new SIMIX mechanisms. This allows gras users to
3239 benefit from the latest improvement to the simulation kernel.
3240 * Kill measurement sockets for now. If you rely on them, sorry. This
3241 release is not for you. This feature will be reintroduced in the
3242 future, but we cannot delay the release any further.
3243 * New function: gras_msgtype_get_name().
3244 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
3245 * Fix (at last) the pmm example: it should not randomly fail anymore.
3247 Build chain: bug fixes and overall polishing
3248 * Cmake is now stable enough. Hence, we killed the autotools.
3249 * Port to windows ( TM :)
3250 * Fix the 'make install' target.
3251 No need to use 'make install-simgrid' anymore
3252 * Introduce a 'make dist' target compiling a *source* archive
3253 'make package' compiles a binary archive
3254 * Compile java files only on need
3255 * Add --cd and --setenv command line options to tesh
3256 * Out of source builds are not fully supported yet, but we are close
3257 * Enable supernovae and optimization flags by default for our users
3260 * Add layer to set up environment directly from lua, without XML.
3261 * The effect of gras_stub_generator can be achieved through
3262 lua too (check examples/gras/console/ping_generator.lua)
3264 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
3266 SimGrid (3.4.1) stable; urgency=low
3268 The "Polishing easter eggs is probably a good idea" release.
3269 This is a bug fixes release only.
3273 * Fix a bug preventing the tasks from begin garbage collected.
3276 * Fix a bug occurring when a host involved in a communication fails.
3277 This was not detected properly by the other peer involved in the
3278 communication. Now, it's reported as a network error.
3281 * Warn the user about loop dependencies in data flow of DAX files
3282 * Obey the control-flow dependencies of DAX files
3285 * Add option "enable_smpi" allowing to not compile SMPI.
3286 Probably useful for the (Mac) users experiencing a build error here
3287 * Improve the detection of lua5.1 and ruby1.8
3289 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
3291 SimGrid (3.4) stable; urgency=low
3293 The "Easter in Cargese" release. Also known as (major changes):
3295 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
3296 ~> bindings were greatly improved
3297 ~> new tracing infrastructure for better visualization introduced
3299 * the "Welcome to configury modernity" release.
3300 ~> we switched from autotools to cmake, and improved our cdash
3303 A more detailled list of changes follow (full detail in svn log).
3305 Java Bindings: Various Cleanups
3306 * (install java-gcj-compat-dev on debian-like to use them)
3307 * Remove put/get: no need to export deprecated interface in Java
3308 Use send/receive instead.
3309 * Cleanup the examples and add a README per directory
3310 * Remove example autoDestination (that's the only way to go now)
3311 * Remove example explicitDestination (was a plain copy of basic)
3312 * Make JniException a runtime exception, so that there is no need to
3313 declare the fact that you may encounter such a beast. I guess that
3314 nobody will ever want to survive such error.
3315 * Create specific errors for each MSG case of failure:
3316 host failure, transfer failure, timeout, task cancelled
3317 * Cleanup the exceptions that may get thrown by each function
3318 * Other internal cleanups in Java bindings. Performance still bad :/
3319 Ruby and Lua Bindings: create them
3320 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
3321 * That's new and great, you should try them out.
3322 Same functionalities than Java bindings, only even less polished
3324 * Kill the useless "rate" argument of SD_task_get_execution_time()
3325 Everyone used to provide -1 as a value, it was not used, and the
3326 semantic of a possible use wasn't even clear.
3327 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
3328 either as comm costs or compute costs to mean that there is no
3329 such thing for that specific task.
3330 * Add a SD_task_set_name() function
3331 * Fix SD_task_unschedule() on typed tasks
3332 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
3333 * In DAX loader, accept useless 'level' attributes to <job> since
3334 LIGO DAGs have them (seem to be to ease graphical representation).
3336 * Add an example masterslave_mailbox.c using send/receive and not
3337 the deprecated put/get interface.
3338 * Kill the MSG_paje_output() function. It's a noop since 2 years.
3339 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
3341 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
3342 (declare MSG_USE_DEPRECATED to still have the old name)
3343 * Add a MSG_task_set_data() function
3344 * About trace replay (see examples/msg/actions):
3346 - Allow to work with splitted trace files for each process
3347 Give the specific trace file as argument of each process,
3348 and call MSG_action_trace_run(NULL)
3349 You can still have one merged file for all processes.
3350 - Fix implementation of collective operations
3351 * Allow task_execute() on 0-sized tasks (closes #10063)
3353 * This is the first release of SimGrid where SMPI is not considered
3354 beta anymore (even if some corners should still be improved)
3355 * Port over the new SIMIX_network submodule (internal refactoring)
3356 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
3357 * Implement more missing elements of the standard:
3359 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
3361 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
3362 MPI_Scan() MPI_Get_processor_name()
3363 - Added implementation of missing case for Alltoall (warning: it's
3364 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
3365 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
3366 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
3367 SMPI_MPI_Allgatherv()
3368 * Bug fixes include:
3369 - MPI_Waitsome() was broken
3370 - Allow relative includes in smpicc
3371 - Command line cfg argument 'reference_speed' was ignored...
3372 - Some functions did not properly lead to auto-benching of user code
3373 - smpicc passes -O2 by default (just like openmpi one)
3375 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
3376 * Bug fixes about timeouts during communications
3377 * add SIMIX_message_sizes_output() as a pimple to write to file the
3378 amount of messages per size. Use gnuplot to get histogram.
3379 Pimple because that's the only user-visible function of simix,
3380 defined directly in xbt.h (irk, sorry)
3382 - Add a SIMIX_sem_get_capacity() function
3383 - Fix interactions with processe resume/suspende
3384 - release_forever() was stupidly broken
3385 - Fix SIMIX_display_process_status() for processes in a semaphore
3386 - Make SIMIX_sem_block_onto() user-visible
3387 * Refactoring context stuff:
3388 - Use pseudo-OOP for better modularity
3389 - reimplement SIMIX_process_kill() without process_schedule() so
3390 that the latter can take as invariant that it is called from
3392 - Merge context_start into context_new for sake of simplicity
3394 * Add a Vivaldi network model, coded live during SUD'10 ;)
3395 * Rename configuration variables to start a hierarchy:
3396 o cpu_model -> cpu/model
3397 o network_model -> network/model
3398 o workstation_model -> workstation/model
3399 * New configuration variables:
3400 o network/bandwidth_factor: correction to bandwidth
3401 o network/latency_factor: correction to latency
3402 o netwotk/weight_S: correction to the weight of competing streams
3403 * Add a long description to the models, that users can see with such
3404 argument on the command line: --cfg=cpu/model:help
3405 * --help-models display the long description of all known models
3407 * config: add the ability to set a default value after registration
3408 Does not override any previously set value (e.g. from cmd line)
3409 * dict: allow to have integer key and data.
3410 When so, you need to use the following functions
3411 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
3412 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
3413 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
3414 In contrary to regular dicts, the key is not malloced before copy.
3415 Mixing scalar and regular elements in the same dict is not tested
3417 * Allow to use xbt_dynar_shrink() to expend the dynar instead
3418 Tracing for Visualization:
3419 * SimGrid is now instrumented in order to generate a trace file for
3420 visualization analysis: to use it, need to compile SimGrid with the
3421 "tracing" option enabled, and instrument the program using SimGrid with
3422 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
3423 (among other functions).
3424 * The instrumentation only traces the platform utilization for now
3425 * Documentation to use the tracing functions and how to analyze the
3426 traces with the Triva tool is written.
3427 * More information about: SimGrid FAQ (in the section Tracing Simulations
3430 * We moved to cmake as default build system. Autotools support will
3431 be dropped soon. Check the FAQ for more info about how to use it.
3432 * Greatly improved our cdash/ctest interactions
3433 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
3434 * Added memory checking tests with valgrind; lot of memleak fixing.
3435 This may be the first release of SimGrid with so few memory issues
3436 * Added code coverage tests.
3437 Our coverage is still improvable, but at least we see it on cdash.
3439 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
3441 SimGrid (3.3.4) stable; urgency=low
3443 The "Desktop Grid needs love too" release (also called Xmas release).
3445 Models improvements:
3446 * Major speedup in the maxmin system solving by using lazy evaluation
3447 Instead of solving completely the maxmin system at each iteration,
3448 only invalidate (and recompute) the modified parts.
3449 This new feature is enabled in default models but you can try to
3450 turn it on with "--cfg:maxmin-selective-update=1" for other models.
3451 * Cas01 IMproved as default CPU model
3452 This CPU model is the same Cas01 model, but it uses the
3453 maxmin-selective-update flag and a heap structure to manage
3454 actions on SURF kernel.
3455 It reduces the complexity to find the next action to finish and,
3456 consequently, it's faster than the old Cas01.
3457 This is the new default CPU model (Cas01).
3458 * Rename the old Cas01 model to Cas01_fullupdate
3459 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
3460 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
3461 * CpuTI (CPU Trace Integration)
3462 A new CPU model whose objective is simulate faster when using
3463 availability trace files.
3464 Instead of using a full featured, over engineered maxmin system for
3465 CPU modeling, this model does the pre-integration of traces files
3466 to calculate the amount of CPU power available, and so, executes
3467 faster than the old CPU models.
3468 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
3469 * Use LV08 as default network model since it gives better accuracy
3470 for small messages and shouldn't change things for big ones.
3471 Use --cfg=network_model:CM02 to get the previous behavior.
3474 ******************************************
3475 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
3476 ******************************************
3477 * The new CPU model may changes simulations!
3478 The point is that events occurring at the exact same timestamp
3479 are not scheduled in the same order with the old and new
3480 version. This may be enough to completely change the execution
3481 of simulations in some cases.
3482 * The new network model will change simulations!
3483 This new model is more realistic than the previous one, so you
3484 should consider redoing your old experiments with this model.
3485 Sorry for the inconvenience.
3488 * Introduce the supernovae compilation mode
3489 When compiled that way, the whole SimGrid (or almost) is put in a
3490 single compilation unit and compiled in one shoot.
3491 This is to help gcc which has difficulties to inline stuff from one
3493 The speedup seem to be above 15%, althrough more tests are needed on
3494 amd64 to confirm that gain.
3497 * Port of MSG's mailbox on top of SIMIX network
3498 The put/get mechanism was greatly simplified on the way.
3501 * New SIMIX network module. Provides:
3502 - Mailbox: rendez-vous mechanism to find with who you want to speak
3503 - Synchronous send/recv: easier and hopefully faster since the
3504 logic is handled in the maestro process directly now
3505 - Asynchronous send/recv: you dreamt of it? It's here now
3506 Too bad that nobody cared enough to propagate the change to MSG.
3507 * Add semaphores as SIMIX synchronization mechanism.
3510 * new function SD_daxload(char*) to load a DAX file
3511 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3512 * Introduce typed tasks. Specify its kind and cost at creation.
3513 At scheduling, just give where it should be placed, and the cost
3514 for each involved resource is automatically computed.
3515 Existing constructors so far (more to come of course):
3516 - SD_task_create_comm_e2e() for end-to-end communication
3517 - SD_task_create_comp_seq() for sequential computation
3518 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3519 * new function SD_task_dump() for debugging display
3520 * new function SD_task_dotty(task,FILE*) writing to file the info
3521 about the task in dotty format
3522 * SD_task_dependency_exists() can now cope with having one of its
3523 arguments NULL. If so, it tests whether the other argument has any
3525 * Add getters on list of preceding/following tasks:
3526 SD_task_get_parents(task) and SD_task_get_children(task)
3527 * Add getters on amount of workstations and list:
3528 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3529 * Add getter on task kind: SD_task_get_kind(task)
3530 * Update the start_time and finish_time of tasks on completion/failure
3531 * Bugfix: Remove task from state swags when destroyed
3534 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3537 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3538 dynar with one separate thread per value of the dynar.
3539 * Change the prototype of xbt_thread_create(), sorry.
3540 Added a boolean parameter indicating whether we want to join this
3541 thread (used in SG only for now)
3542 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3545 * GTNetS wrappers should now be usable again (and betterly tested too)
3546 * Fix a major regression from 3.2 where the timeout provided to
3547 MSG_task_put_with_timeout() was used as absolute time before which
3548 the comm should be done.
3549 * Start to fix the <cluster> tag.
3550 - Internal links should be good now (beside of the loopback, which
3551 use the private link instead)
3552 - paths to the external world is still rather broken
3553 - the <route:multi> tag is just broken. Actually that's brain-dead.
3554 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3555 to make it less stupid
3556 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3557 * Fix a source-level compatibility glitch from 3.2: after defining
3558 MSG_USE_DEPRECATED, you can use the old name
3559 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3560 * Allow to compile from the SVN with automake 1.11
3561 * Fix some problems when using the "start_time" tag in deployment XMLs.
3562 * Fix #8569: XBT/synchro.h has redundant declarations
3563 * Fix #8563: MSG return values and exceptions
3564 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3565 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3566 Thanks to Robson Peixoto.
3567 * Fix a vicious bug in dictionaries inducing that some elements were
3568 not freed on xbt_dict_free()
3570 Portability report of this version:
3571 * Main portability targets:
3572 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3573 on (amd64/i386/ia64)
3574 - mac leopard on i386
3575 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3576 but nothing critical.
3577 * Other platforms: windows, AIX and others were not tested for this release
3579 Timing report of this version:
3580 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3581 scenario parameters). From 8h to a few seconds in desktop grid settings.
3582 * Supernovae brings about 25% speedup on i386.
3584 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3586 SimGrid (3.3.3) stable; urgency=low
3588 The "Need for Speed" release.
3590 The timings done to validate the 3.3.2 were faulty.
3591 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3593 The problem was a conversion from a manually handled vector to
3594 xbt_dynar_t on the critical path.
3595 xbt_dynar_foreach calls functions, inducing stack management crap.
3597 We inlined these functions and xbt_dynar_foreach is now breath taking.
3598 We also inlined xbt_swag_belong on the way.
3600 Here are some approximate speedup measurements (on master/slaves
3601 simulations lasting between 10s and 20s each):
3602 3.3.1 -> 3.3.2: about same performance
3603 3.3.2 -> 3.3.3: 40% speedup
3604 3.3.1 -> 3.3.3: 40% speedup
3605 3.3.1 with inline patch -> 3.3.3: 30% speedup
3607 Our reading is that the refactoring which occurred in 3.3.2 made us
3608 suffer much more from the xbt_dynar_foreach low performance, but
3609 once we solved this, this refactoring proved to be very performance
3610 effective. From the 40% speedup, somehow, 10% are due to the
3611 inlining and 30% to the refactoring.
3613 That's a pitty that gcc cannot inline functions placed in other files
3614 alone. We have to choose between:
3615 - break the encapsulation (by putting private data structures and
3616 accessors in headers files to help gcc)
3617 - live with low performance
3618 - switch to a decent compiler such as icc (not quite possible).
3620 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3622 SimGrid (3.3.2) stable; urgency=low
3624 The "Simplicity does not preceed complexity, but follows it" release.
3626 The main contributors of this release were (lexical order):
3627 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3630 * Extract the routing logic into its own object.
3631 (was dupplicated in network.c and workstation_LV07.c;
3632 Allows to implement other ways of storing that info)
3633 => kill now useless network_card concept
3634 - Use dynar to represent routes (instead of void** + int*)
3635 - kill link_set (use surf_network_model->resource_set instead)
3636 - Add a command-line option to choose the routing schema to use
3637 - Add three new models:
3638 * Floyd (shortest path computed at initialization)
3639 * Dijikstra (shortest path recomputed all the time)
3640 * Cached Dijikstra (shortest path computed on need)
3641 All these models where contributed by Silas De Munck, and are
3642 described in his ICCS09 paper.
3644 * Simplify model declaration
3645 (less redirections, less function to write when defining a model)
3646 - Factorize stuff between models:
3649 surf_model_resource_set(model)
3650 surf_model_resource_by_name(model, name)
3651 - Unify the types of models in s_surf_model_t (using an union)
3652 - Embeed fields of common_public directly into s_surf_model_t
3653 - Rename model methods:
3654 action_free ~> action_unref
3655 action_change_state ~> action_state_set
3656 action_get_state ~> action_state_get
3657 - Change model methods into functions:
3658 (model)->common_public->action_use ~> surf_action_ref
3660 * Implement a generic resource; use it as ancestor to specific ones
3661 (allows to kill duplicated code in models)
3662 Drawback: timer command don't need no name nor properties;
3663 workstation_CLM03 don't need no properties
3664 (but I guess we can live with those few bytes wasted)
3666 * Improve the action object model
3667 - implement a constructor avoiding dupplicated code about field
3668 initialization in generic_action part.
3670 * Kill the SDP model: it has an external dependency, is deprecated
3671 in flavor of modern lmm models, and didn't compile since a while
3674 * Relocation of the context module from XBT to SIMIX.
3675 (the context were decoupled from the simix processes, duplicating a lot of code)
3676 => a lot of code was factorized
3677 - less overhead is introduced during scheduling
3678 - simpler API for the context factory
3679 - the logic for process creation,destruction and manipulation was simplified
3680 * Simplification of the s_smx_process_t data structure.
3681 => accesing the simix level data associated to a process is faster now,
3682 and the code is a lot more readable.
3685 * Implement some more MPI primitives:
3686 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3687 -implementation: Bcast: flat or 2-ary tree (default),
3688 Barrier: 4-ary tree,
3690 Allreduce: Reduce then Bcast
3691 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3692 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3693 Alltoallv: flat tree, like ompi
3695 * Add support for optimized collectives (Bcast is now binomial by default)
3696 * Port smpirun and smpicc to OS X
3699 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3700 Shout out if you used it.
3703 * Display the list of still queued messages in SG mode when existing
3707 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3708 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3709 * Add xbt_str_from_file(FILE*)
3710 * Add xbt_dict_get_key achieving a linear reverse search
3711 * Remove the context module
3713 Portability report of this version:
3714 * Main portability targets:
3715 - Linux(debian)/x86/context
3716 - Linux(debian)/x86/pthreads
3717 - Linux(debian)/amd64/context
3718 - Linux(debian)/amd64/pthreads
3719 On these, we still have the eratic breakages of gras/pmm and
3720 amok/saturate_sg reported in previous version. We still think
3721 that the tests are the cause of the fault, not the tested code.
3723 - Mac OSX Leopard/x86/context
3724 Still false negative in tesh autotesting.
3725 Smpi still fails, but this time because readlink does not accept -f
3726 Everything seems to work properly beside of that.
3729 - AIX version 5.3 (only tested contexts this time)
3730 Smpi still fails there because mktemp is not installed.
3731 Everything seems to work properly beside of that.
3733 I managed to compile it for the first time, but several breakages.
3734 Won't delay the release for this exotic platform.
3736 * Windows: it's still lagging behind. If you want to help, please
3739 Timing report of this version:
3740 This version seem to be more than 5% faster than 3.3.1 (on linux
3741 64bits with contexts). The gain is less than expected, we are
3742 investigating this for next release.
3744 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3746 SimGrid (3.3.1) stable; urgency=low
3749 * Implement a --cfg-help to show existing configuration variables
3750 * Build chain do not require doxygen in maintainer mode
3753 * fix a bug on struct sizeof computation, which prevented the
3754 exchange of arrays of structs in some conditions
3755 - added a regression test about this in datadesc_usage
3756 * Allow the exchange of 0-long dynamic vectors.
3757 - for that, use -1 as indicator of dynamic size instead of 0
3758 - This implied to change any size from unsigned long to long,
3759 reducing a bit communication abilities, but I guess that with
3760 64bits being quite common, this is more than enough.
3761 - This also induce a protocol change, thus bumping network protocol
3762 version from 0 to 1 (if we have external users, we have to get
3763 clean on that point too ;)
3764 - added two regression tests about this in datadesc_usage
3765 * Be more verbose when propagating local exceptions
3766 This helps debugging.
3767 * Display the status of simulated processes when receiving SIGINT in
3771 * Allow to control the simulation from a trace file.
3772 New functions MSG_action_register() and MSG_action_trace_run()
3773 The first one allows to associate a function execution to each
3774 kind of action while the second one parses a trace file and
3775 triggers the corresponding actions within the system.
3776 For now, only a toy example is provided in examples/msg/actions
3777 * Add an example of process migration in examples/msg/migration
3778 * Fix a bug in task exchange which broke MSG_task_get_sender()
3779 Add a teshsuite regression test for that.
3780 [Bug: if MSG_task_get_sender() is called after sender exit,
3782 * Fix a bug which prevented suspend/resume to work properly
3783 * Display the status of simulated processes when receiving SIGINT
3784 This fixes a regression of v3.3. due to the introduction of SIMIX
3785 * Bug fixing in failure management:
3786 - trace could not start by a failure at time 0
3787 - failure during communications were not working
3790 * Add SIMIX_process_set_name() to change the name of the current
3791 process in the log messages.
3792 * Store smx_hosts in a dict since we only retrieve them by name
3793 * Move the configuration infrastructure to surf
3796 * Move the configuration infrastructure to surf
3799 * Massive internal cleanups:
3800 - Store internal structures on processes instead of hosts (allows
3801 to have more than one process per host, in addition of being more
3803 - Cleanup the initialization/finalization process
3804 - Kill a whole bunch of unneeded synchronization:
3805 processes run in exclusive manner within the simulator
3806 - Move queues from global tables to process data fields
3808 - now accept -platform and -hostfile arguments
3809 - Pass the right rank value to processes according to the hostfile
3810 * Compile the examples by default, and use them as regression tests
3811 * Implement MPI_Wtime()
3812 * Change the reference speed to a command line option
3815 * TCP_gamma can now be specified as command line option using
3816 --cfg=TCP_gamma:10000000.0
3817 * Change the --surf-path cmd line option into --cfg=path:
3820 * Also include strbuff from xbt.h public header
3821 * xbt_ex_display(): do not free the exception after displaying
3822 This allows to do more with the given exception afterward.
3823 Users should call xbt_ex_free() themselves.
3827 Portability report of this version:
3828 * Main portability targets:
3829 - Linux(debian)/x86/context
3830 - Linux(debian)/x86/pthreads
3831 - Linux(debian)/amd64/context
3832 - Linux(debian)/amd64/pthreads
3833 These targets fail about 1/10 of times on gras/pmm, but we believe
3834 that this is because of the test, not because of SimGrid.
3835 amok/saturate_sg fails even more rarely, and the test may not be
3838 - Mac OSX Leopard/x86/context
3839 The test suite still spits tons of errors because some obscure
3840 force prevents us from removing the temporary directories
3841 arguing that they still contain some metadata I've never heard of.
3842 Smpi fails because seq is not installed.
3843 Everything seems to work properly beside of that.
3846 - AIX version 5.3 (both contexts and pthread)
3847 Smpi still fails there because mktemp is not installed.
3848 XML inclusions seems rosty on AIX.
3850 * Windows: it's still lagging behind. If you want to help, please
3853 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3855 SimGrid (3.3) stable; urgency=high
3859 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3860 [Malek Cherier & Mt]
3862 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3865 Until now, GRAS were using MSG as an interface to SURF. It was
3866 quite difficult because both interface have several differences
3867 (MSG channels vs GRAS sockets were the most notable point).
3869 This also opens the gate to SMPI (which should occur soon) and speed
3870 up simulations by to 40% (even if it were not the main goal).
3872 **************************************
3873 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3874 **************************************
3875 The point is that events occurring at the exact same timestamp are
3876 not scheduled in the same order with the old and new version. This
3877 may be enough to completely change the execution of simulations in
3878 some cases. Sorry for the inconvenience.
3880 * Cleanup and upgrade the XML format to push further scalability
3881 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3883 * Improve the testing infrastructure with tesh. Now a very large part of
3884 the code is tested not only by being run but also by checking that the
3885 output match an expected output [Mt].
3887 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3888 is really less memory-demanding, which should allow you to use
3889 larger files in SimGrid [AL].
3891 * Inform valgrind about our contexts, so that it becomes usable
3892 with the default (and more effecient) version of SimGrid
3893 [contributed by Sékou Diakite, many thanks]
3896 * Introduce a listener thread in charge of receiving incoming
3897 messages from the network. It allows to overlap communication and
3898 computation but most notably, it removes some stupid deadlocks due
3899 to the fact that so far, a process could not send and receive at
3900 the same time. This made most non trivial communication schema
3902 * Convert the PIDs from long int to int to match the MSG ones (and
3903 linux ones too) [Mt]
3904 * New function: gras_agent_spawn() to launch a new process on
3905 current host. Only working in simulation for now. [Mt]
3906 * New function: gras_os_hostport() returning a constant form (ie,
3907 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3910 * Make the backtrace of exceptions more human readable [Mt]
3911 * New module: xbt/str [Mt]
3912 a ton of string utility functions (split, join, printf to a newly
3913 allocated buffer, trim, etc)
3914 * New module: xbt/hash [Mt]
3915 SHA1 hashing algorithm (more to come if needed)
3916 * New module: xbt/synchro [Mt]
3917 synchronization tools (mutex and conditions) working the same way
3918 in simulation and in real life (mainly useful for GRAS, but not
3920 * New module: xbt/queue [Mt]
3921 classical producer/consumer synchronization scheme
3922 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3923 (using the classical functions will get serialized) [Mt]
3924 * Make dictionary internal table dynamic. No need to specify its size
3925 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3927 * Make sure the log channels are organized as a tree under windows
3928 (because of ANSI C compatibility issue, any channel were child of
3929 root directly) [Mt].
3932 * Cleaned many thing in surf and fixed a few bugs [AL].
3933 * Add a nice command line configuration mechanism to compose models [AL].
3934 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3935 the previous one (KCCFLN05). It relies on something that looks like
3936 a max-min sharing mechanism but cannot be written as such. A new solver
3937 was thus designed [AL].
3938 * Add a new solver to lmm. Based on Lagrange optimization and
3939 gradient-based descent, it enables to efficiently maximise systems s.a
3941 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3943 This solver enables to propose two new network models for TCP Reno and
3944 TCP Vegas based on Low's work. These models still need to be fully
3945 tested though [Pedro Velho].
3948 * Bug fix in SD_simulate. Now the time bound given as argument is
3950 * Use the new parallel task model (ptask_L07) as default.
3951 * Use the SURF command line configuration mechanism.
3952 * 0-size tasks (for synchronization) should now work.
3954 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3956 SimGrid (3.2) stable; urgency=high
3960 We still experience issues on this platform, but we believe that at
3961 least MSG is usable.
3963 GRAS API BREAKAGE (for simplification purpose, sorry):
3964 * the gras_msgtype_by_name is not used anymore. Instead of
3965 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3966 you can write (and must)
3967 gras_msg_send(toserver, "request", &request);
3968 - If you still want to pass a gras_msgtype_t to the function (to cache
3969 the type and avoid the lookup time), use the gras_msg_send_() variant.
3970 - Impacted functions:
3971 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3972 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3973 * The callbacks are now expected to return 0 when everything went well
3974 (just like the main() function)
3976 GRAS new features and improvements:
3977 * New module mechanism where user code can use per process globals [Mt]
3978 This is similar to gras_userdata_*() functions, but for libraries. It
3979 factorize some code developped over and over in the examples and AMOK.
3980 It has still to be documented and used (only amok/peermanagement is
3982 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3983 loss when they were small enough to fit into the buffer and sent quickly
3984 enough so that they can all get received in one shoot.
3985 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3986 if not found. Use the *_or_null() variant for the old semantic.
3987 * In gras_msg_handle, do not discard messages without callback.
3988 They are probably messages to be explicitly awaited later (ie, proofs of
3989 mis-synchronization in userland since they are sent before being awaited)
3990 No big deal usually.
3991 * gras_socket_meas_send/recv: semantic changed!
3992 The numerical arguments used to be (1) the total amount of data to send
3993 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3994 messages. This was need for the fool willing to send more than MAXINT
3995 bytes on quite fat pipes.
3998 * Do really rename the hostmanagement module to peermanagement. [Mt]
3999 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
4000 but this is rather new and this was documented in the module
4001 documentation (poor excuses, I admit)
4002 * Bandwidth measurement semantic changed! This follows the changes to
4003 gras_socket_meas_send/recv explained above.
4006 * A sequential mode has been added to the workstations. When a workstation
4007 is in sequential mode, it can execute only one task, and the other tasks
4008 are waiting in a FIFO. [Christophe Thiery]
4011 * The KCCFLN05 workstation model now handles parallel tasks. It is the
4012 model for SIMDAG. [Christophe Thiery]
4013 * Bug fix in the maxmin solver: Some values were close to 0 instead of
4014 equal to 0, which caused some bad behaviors in
4015 saturated_constraint_set_update. I now use a threshold mechanism like in
4019 * When running manually src/testall, you select specific units [Mt]
4020 testall is the result of our cunit mechanism, and should replace all
4021 the scripty thingy around since bash don't run easily on billware.
4023 * A mallocator system has been added. [Christophe Thiery]
4024 Mallocators allow you to recycle your unused objects instead of freeing them
4025 and allocating new ones.
4027 Documentation update:
4028 * FAQ reworking + New FAQs:
4029 - "Valgrind spits tons of errors!" [Mt]
4030 - "How to repport bugs" [Mt]
4031 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
4032 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
4033 - Communication time measurement within MSG [AL]
4034 - I experience weird communication times when I change the latency [AL]
4035 * GRAS tutorial [Mt]
4037 - an introduction to the framework and to the used communication model
4038 - an initiatic tour introducing the most proheminent features:
4040 . Lesson 0: Installing GRAS
4041 . Lesson 1: Setting up your own project
4042 o Part 2: Message passing
4043 . Lesson 2: Exchanging simple messages
4044 . Lesson 3: Passing arguments to the processes (in SG)
4045 . Lesson 4: Attaching callbacks to messages
4046 . Lesson 5: Using globals in processes
4047 . Lesson 6: Logging information properly
4048 . Lesson 7: Using internal timers
4049 . Lesson 8: Handling errors through exceptions
4050 . Lesson 9: Exchanging simple data
4051 . Lesson 10: Remote Procedure Calling (RPC)
4052 . Lesson 11: Explicitly waiting for messages
4053 . Recapping of message passing features in GRAS
4054 - A HOWTO section containing:
4055 o HOWTO design a GRAS application
4056 More are due, of course. They will come latter. In the meanwhile, you can
4057 check the examples which are still here.
4059 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
4061 SimGrid (3.1) stable; urgency=high
4065 There was a stack corruption somewhere, visible only when optimizing
4066 with these versions. [Vince]
4069 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
4070 module that enabled to play with Directed Acyclic Graphs. It is built
4071 directly on top of SURF and provides an API rather close to the old
4072 SG. Some old codes using SG are currently under rewrite to check that
4073 all needful functions are provided. [Christophe Thiery]
4076 * Complete rewrite of the KCCFLN05 workstation model. It is now an
4077 extension of the classical CLM03 model that gracefully handles
4078 failures. This is now the default model for MSG and GRAS. It doesn't
4079 handle parallel tasks yet though. [AL]
4080 * Bug fix: Weights were not correctly set in the network part.
4081 WARNING: This may have resulted in incorrect results with simulations
4082 where there are more than one flow on a given link. [AL]
4085 * After a (long ?) discussion on simgrid-devel, we have decided that the
4086 convention we had on units was stupid. That is why it has been decided
4087 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
4088 WARNING: This means that all previous platform files will not work as
4089 such with this version! A warning is issued to ask users to update
4091 A conversion script can be found in the contrib module of the CVS, under
4092 the name contrib/platform_generation/surfxml_update.pl [MQ]
4095 * Bug fix: Processes were started in reverse order, wrt deployment file.
4096 WARNING: if your code relies on this bug, please fix it. [AL]
4097 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
4098 being executed on two different locations. [AL]
4099 * Bug fix: Failures are now better supported thanks to Derrick's tests
4100 (there was many failure situations I hadn't thought of and that weren't
4101 correctly handled). [AL]
4102 * New function: MSG_host_is_avail indicates you whether a given m_host_t
4106 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
4107 Exception occurring on server-side are propagated back to client (!).
4109 API CHANGE: the callback changed their prototype. Change:
4110 int my_handler(gras_socket_t expeditor, void *payload_data) {
4112 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
4113 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
4115 * New! function: gras_msg_handleall to deal with all messages arriving
4116 within a given period.
4117 * New! function: gras_socket_server_range to get a server socket in a
4118 range of port numbers (ease to avoid port number conflicts) [MQ]
4119 * New! gras processes display their backtrace when they get a SIGUSR1
4120 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
4121 Sweet to debug RL processes [MQ]
4125 - Do not force experiment sizes to be expressed in kb, or it becomes
4126 impossible to measure the latency this way (needs one byte-long tests)
4127 WARNING: this changes the amok_bw_* function semantic. [MQ]
4128 - Implements the link saturation stuff. [MQ]
4129 * Peer management module:
4130 New! module factorizing code that we wrote over and over [MQ].
4133 * New module: cunit (my jUnit implementation in ansi C) [MQ]
4134 - Test units are placed directly into the library code, they get extracted
4135 automatically and placed into the src/testall binary.
4136 - Convert most of the XBT tests to this system.
4137 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
4138 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
4139 live in the same C code. This required to change a little bit the API
4140 of surfxml parsing but shouldn't be an issue for end-users. [AL]
4141 * New module: sparse graph structure with basic algorithms (this is work
4142 in progress and the API is not considered to be frozen yet). [AL]
4143 * Display more information on backtraces: source line & function names are
4144 now displayed just like valgrind does (rely on addr2line tool) [MQ]
4145 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
4146 * Reworked a little bit some #include statements to load only required
4147 headers. Some user code that relied on SimGrid to include stdlib or
4148 stdio may need to include it by themselves. [AL]
4149 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
4151 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
4152 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
4153 but I guess that almost nobody used those parts. [MQ]
4155 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
4157 SimGrid (3.0.1) stable; urgency=low
4160 * Unfortunately, I had missed 5 misnamed functions:
4161 xbt_fifo_item_t xbt_fifo_newitem(void);
4162 void xbt_fifo_freeitem(xbt_fifo_item_t);
4163 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
4164 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
4165 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
4166 They're now deprecated. Please use their new versions:
4167 xbt_fifo_item_t xbt_fifo_new_item(void);
4168 void xbt_fifo_free_item(xbt_fifo_item_t);
4169 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
4170 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
4171 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
4173 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
4174 * Documentation: xbt_log module unmercifully reworked [MQ]
4175 * Bugfix: there was a problem with the ending of contexts with
4176 the pthread backend. It caused some weird deadlock or behavior
4177 depending on the pthread implementation. [AL]
4178 * Bugfix: get the exceptions raised in the simulator repport where
4179 and why they come from when they are not caught in time [AL, MQ]
4182 * Bugfix: Do repport the error when two non-connected hosts try to
4183 exchange data (Thanks to Flavien for stumbling into this one) [AL]
4186 * Add additionnal checkings on communications. Assert that two
4187 communicating hosts are connected by a set of links... [AL]
4190 * Add additionnal checkings on channel values in communication [AL]
4191 * New: MSG_task_get_source to see on which host a task was generated [HC]
4192 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
4193 the number of tasks waiting to be received on channel and sent
4195 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
4196 waits for the first task coming from a given host.. [AL]
4198 GRAS new functionnalities: [MQ]
4199 * Enhance the parsing macro to allow the size of multidimentional objects
4200 to be given thru annotations.
4201 * New example (and documentation): Matrix Multiplication a la RPC
4202 (as when I was young!) and fix a bunch of bugs found on the way.
4204 GRAS performance improvements: [MQ]
4206 * Reduce the amount of cbps creation/destruction by making it static to
4207 datadesc_send/recv() and using a (newly created) cbps_reset (based on
4210 * Change libdata to a set so that we can search for stuff by ID (and thus
4211 reduce the insane amount of dict lookups)
4214 * Actually implement gras_datadesc_copy() so that we don't have to mimick
4215 RL communication on top of SG since it's so uneffective.
4216 It may also be used for inter-thread communication in RL, one day.
4217 * Use gras_datadesc_copy() to exchange messages on top of SG
4219 - improve message exchange performance on top of SG
4220 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
4221 * Don't exchange on the network the size of the used part of buffer,
4222 instead, specify the possible buffer size to read().
4224 - reduces the amount of read/write calls (one pair per exchange)
4225 - reduces the amount of exchanged data (the size)
4226 - allows to retrieve all arrived data on receiver side, if we don't need
4227 it right now (subsequent read will peek the buffer)
4228 - allows the receiver to proceed with the beginning of the stream before
4229 everything is arrived
4230 - make it possible to build an iov transport (using readv/writev)
4232 - take care of the data with non-stable storage (like stacked data),
4234 * If possible, TCP send uses vector I/O (when writev() is here)
4235 - Don't use it for receive since we send data sizes and data on the
4236 same stream, so we wouldn't be able to chain large amount of chunks
4237 before having to flush the stuff to read the size.
4238 * Rework the transport plugin mechanism to simplify it and reduce the
4239 amount of pointer dereferencement when searching for the right function
4242 * I guess that now, we do almost as few system calls as possible while
4243 doing as few data copy as possible.
4245 To improve it further, we could try to send all the sizes first and then
4246 all the data (to use iov on receiving size), but it's only a partial
4247 solution: when you have 2 dimensional data, the sizes of the second
4248 dimension is data of the first dimension, so you need 3 streams.
4250 I'm not sure the potential performance gains justify the coding burden.
4252 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
4254 SimGrid (3.00) stable; urgency=high
4257 * New! Give the possibility to hijack the surf parser and thus bypass
4258 MSG_create_environment and MSG_launch_application. Have a look at
4259 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
4261 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
4263 SimGrid (2.96) unstable; urgency=low
4268 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
4269 This deprecates the xbt_error_t mechanisms.
4270 It modifies (simplifies) all XBT and GRAS API.
4271 MSG API keeps unchanged (exceptions raised by XBT are caught from
4272 within MSG and masked with existing error handling facilities)
4275 * New! Add a FATPIPE model. [AL]
4276 * New! Add a parallel task model. [AL]
4277 * New! Add automatically a loopback interface (in the default
4278 network model) if none was precised.
4281 * Bugfix: MSG_process_resume now works with the current running process.
4283 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
4284 * Modification of MSG_task_get_compute_duration. Once a task has been
4285 processed, the value returned by this function is now equal to 0. [AL]
4286 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
4287 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
4288 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
4289 of a m_task. For now, MSG_task_cancel only works with computation
4291 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
4292 of the processor (in Mflop/s) regardless of the current load on the
4294 * API Change: use proper naming convention for MSG_getClock and
4295 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
4297 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
4298 This function changes the priority of a computation task. This priority
4299 doesn't affect the transfer rate. A priority of 2 will make a task
4300 receive two times more cpu power than the other ones. This function
4301 has been added to suit the needs of Nguyen The Loc and hasn't been that
4302 much tested yet. So if it fails, please report it and send me your code.
4304 * API Change: removed all functions and types that were marked "deprecated"
4305 since many months. Renamed MSG_global_init_args to MSG_global_init.
4307 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
4309 SimGrid (2.95) unstable; urgency=low
4312 * Steal some nice code to GNU pth to fix context detection and usage [AL]
4313 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
4314 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
4317 * Bugfix: Allow absolute paths for platform description files [MQ]
4318 * Bugfix: do free the variables after use. Leads to drastic performance
4320 * Implement max_duration (ie, timeouts) on resources [AL]
4323 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
4325 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
4326 the needed support (timeouts) [AL]
4329 * Implement measurement sockets. You can now get the bandwidth between two
4330 hosts thanks to AMOK (see below). [MQ]
4331 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
4332 dynar over the network (yeah) [MQ]
4333 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
4334 * simple/static token-ring example. [Alexandre Colucci and MQ]
4335 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
4336 in place before [MQ]
4339 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
4340 type parsing. "short" and "long" modifiers were also ignored; other
4341 modifier (such as reference level) are still ignored. [MQ]
4342 * Embeed the buffer size within the buffer itself on SG. [MQ]
4343 That way, send() are atomic and cannot get intermixed anymore (at least
4344 the ones which are less than 100k; bigger messages still have the issue)
4345 * Array size pushed by the field, not by the field type (or each
4346 and every long int will push stuff to the cbps) [MQ]
4347 * use select() to sleep since it allows to portably sleep less than one
4350 GRAS (minor cleanups)
4351 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
4354 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
4355 * type category 'ignored' killed as it was never used and were difficult
4357 * whether a type can cycle or not is now a flag, leaving room for more
4358 flags (as "ignored", if we feel the need one day ;)
4359 * Rename raw sockets to measurement sockets since "raw" has another
4360 meaning in networking community.
4363 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
4364 and offers features not belonging to GRAS but that most applications
4365 need. One day, it may be a set of plugins loadable at runtime.
4366 * New module: bandwidth
4367 bandwidth measurement between arbitrary nodes running this module. [MQ]
4369 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
4371 SimGrid (2.94) unstable; urgency=low
4373 The first beta release of SimGrid 3 !
4377 * Update the main page and the FAQ. Adding references to gforge.
4380 * Add a gras_os_getpid function.
4383 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
4384 * Extend the logs so that they also print PID, hostname, date, ... if
4386 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
4387 and kill the old version which were in testsuite/
4388 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
4392 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
4393 for users. Instead, each programming environment (GRAS, MSG,...) use it
4394 when needed to provide such a feature to users.
4395 Don't shortcut the mechanism or you will also shortcut the virtualization
4396 you need on the simulator.
4400 * Cleanups in configury with regard to compile optimization/warning flags.
4401 Also add -fno-loop-optimize to any powerpc since it's the optimization
4402 killing gcc (< 3.4.0).
4403 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
4404 needed by MSG examples complications
4405 * Borrow configury beautifications from PHP
4408 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
4409 warning (thanks loris for stumbling into this one).
4410 * Bugfix: stop loading private headers (gras_config.h) from the public
4414 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
4415 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
4416 RL would imply a DNS resolver, which is *hard* to do in a portable way
4417 (and therefore delayed).
4418 * Implement a real timer mechanism and use it in timing macros. This allows
4419 to avoid rounding errors and get a 0.000005 sec precision in timing
4420 macros. While I was at it, various cleanups:
4421 - allow to declare more than one timed section per file (fix a stupid bug)
4422 - move some private declaration to the right place
4423 - merge conditional execution and timing macros into emulation module
4424 - document the module
4425 - make sure the module cleanups its mess on gras_exit
4426 * Documentation improvements:
4427 - (new) how to compile applications using GRAS
4428 - (new) emulation support (timing macros)
4430 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
4432 SimGrid (2.93) unstable; urgency=low
4434 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
4437 - Use Paje properly where used. Still to be sanitized properly.
4438 - Portability fix: Add an implementation of the contexts using pthread
4442 - Add xbt_procname(): returns the name of the current process.
4443 Use it to show the current process's name in all logging.
4445 - fix detection of older flex version and the reaction, since we do
4446 depend on modern ones (we use lex_destroy)
4447 - Better separation of SG and RL in the libs: remove all simulation code
4448 from libgras. As a result, this lib is now only 200k when stripped.
4449 Some of the xbt modules may also be duplicated (two sets and such) and
4450 should be cleaned/killed before SG3.
4451 - Insist on using xlC on AIX because of weird problems involving gcc there.
4452 - Cleanup the make remote stuff. This is now done by scripts
4453 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
4454 mainly for our private use, but we're working on changing them to user
4457 - Bugfix: flush the socket on close only if there is some *output*.
4458 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
4460 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
4463 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
4465 SimGrid (2.92) unstable; urgency=low
4467 Alpha 3 on the path to SimGrid 3
4471 - New! First try of benchmarking macros.
4472 - New! First try so that gras_stub_generator generate deployment and
4473 remote compilation helpers.
4475 - Bugfix: Initialization fix in msg_test.
4479 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
4481 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
4483 - API Change: gras_os_sleep to take the amount of seconds as a double.
4484 Accepting an int was error prone since it was the only location where
4485 seconds were coded as such. It leaded to damn rounding errors.
4486 - Bugfix: Hard to belive that timers ever worked before this.
4488 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4490 SimGrid (2.91) unstable; urgency=low
4492 Alpha 2 on the path to SimGrid 3
4496 - Bug fix in the lmm_solver.
4498 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4499 through the function MSG_paje_output.
4500 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4501 - It is possible to bound the rate of a communication in MSG with
4502 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4503 to write it in the changelog).
4504 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4509 - Various cleanups to the autotools stuff
4510 - Begin to move Gras examples to examples/gras/
4511 - Let make distcheck work again (yeah!)
4513 - documentation overhauled using doxygen.
4514 gtk-doc-tools is dead in SimGrid now.
4515 - Automatically extract all existing logging categories, and add the list
4516 to the documentation (long standing one, to say the less)
4518 - Cleanup the known architecture table. Reorder the entries to group what
4519 should be, and use a more consistent naming scheme.
4520 (some of the test dataset are still to be regenerated)
4521 - New! Allow library to register globals on each process just as userdata
4523 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4524 lookup time (for now).
4525 Use it in msg and trp.
4526 This cleans a lot the internals and helps enforcing privacy of the
4527 headers between the gras components.
4528 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4529 - Bugfix: gras_os_time was delirious in RL.
4530 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4532 - Reenable GRAS now that it works.
4534 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4536 SimGrid (2.90) unstable; urgency=low
4538 Alpha 1 on the path to SimGrid 3
4540 * It is a long time since the last release of SimGrid. I'm sorry about
4541 that but as I had told you, I was rewriting a lot of things. I apologize
4542 to those who had been reporting bugs to me and that I had not answered.
4543 If your bug is still in the new version, please tell me. Here is a
4544 summary of the main changes.
4546 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4547 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4548 portability, speed, and a lot more but you'll figure it out later.
4549 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4550 I wanted to release the new MSG as soon as possible and I have
4551 broken GRAS, which is the reason why, for now, only MSG is fully
4552 functional. A laconic description of these projects is available
4553 in the documentation.
4555 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4556 kernel optimized for our needs (called SURF but only the developers
4557 should use it). Hence, MSG is now roughly 30 times faster and I think
4558 that by rewriting a little bit MSG, I could event speed it up a little
4559 bit more. Beside the gain in speed, it is also much easier to encode a
4560 new platform model with SURF than it was with SG. More to come...
4562 * REVOLUTION 3: I have tried to change a little as possible the API of
4563 MSG but a few things really had to disappear. The main differences
4564 with the previous version are:
4565 1) no more m_links_t and the corresponding functions. Platforms are
4566 directly read from an XML description and cannot be hard-coded
4567 anymore. The same format is used for application deployment
4568 description. The new format is described in the documentation.
4569 Have a look in tools/platform_generation. There is a tiny script
4570 that converts from the old platform format to the new one. Concerning
4571 the application deployment format, parsing the old one is tricky.
4572 I think most of you should however be able to convert your files. If
4573 it is really an issue, I can write a C code that does the conversion.
4575 2) the toolbox tbx does not exist anymore. We now have a library
4576 with much more data-structures but without the hash-tables (we have
4577 dictionaries that are much faster).
4579 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4581 *****************************************************************************
4582 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4583 * don't feel like converting the oldies. (Mt) *
4584 *****************************************************************************
4587 Version 2.90: "the long awaited one"
4588 - Finished rewriting and debugging MSG. Rewrote the documentation.
4589 - disable GRAS for now since it needs to be ported to the newest SG
4592 - Finish the port to windows (using mingw32 for cross-compile)
4595 - Main loop and datastructures of SURF. A cpu resource object is
4596 functional. Surf can thus be used to create cpu's with variable
4597 performance on which you can execute some actions.
4599 2004-11-15 Martin Quinson
4600 - Port to ARM. Simply added the alignment and size descriptions. Should
4601 work, but the ARM machines are so slow that I didn't had the opportunity
4602 to 'make check' over there yet.
4604 2004-11-15 Arnaud Legrand
4605 - Trace manager now written. It uses a heap structure and is therefore
4606 expected to be efficient. It may however be speeded up (particularly
4607 when many events occur at the same date) by using red and black
4608 trees. One day maybe...
4609 - Max-min linear system solver written. It uses a sparse matrix
4610 structure taking advantage of its expected use. Most operations are
4611 O(1) and free/calloc are called as few as possible. The computation of
4612 the minimum could however be improved by using a red and black tree
4615 2004-11-03 Arnaud Legrand
4616 - Rename every gras_* function that was in xbt/ to its xbt_
4618 - Add a heap and a doubly-linked list to xbt
4619 - Added a dichotomy to the dictionaries. make check works as well before
4620 so I assume that the patch is correct. I do not know however if things
4621 run effectively faster than before now. :)
4623 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4624 SimGrid, and the version number is bumped to 2.x
4626 2004-10-29 Martin Quinson
4627 - Introduction of the remote errors.
4628 They are the result of a RMI/RPC on the remote machine.
4629 ErrCodes being scalar values, you can't get the host on which those
4630 errors did happen. Extending the error mechanism as in Gnome is possible.
4631 No idea yet whether it is a good idea.
4633 2004-10-28 Martin Quinson
4634 - Interface revolution: the Starred Structure Eradication.
4635 I used to do typedef struct {} toto_t; and then handle *toto_t.
4636 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4637 - ???_t is a valid type (builded with typedef)
4638 - s_toto_t is a structure (access to fields with .)
4639 - s_toto is a structure needing 'struct' keyword to be used
4640 - e_toto_t is an enum
4641 - toto_t is an 'object' (struct*)
4643 typedef struct s_toto {} s_toto_t, *toto_t;
4644 typedef enum {} e_toto_t;
4645 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4646 s_toto_t) is private.
4648 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4649 so that it can set the variable to NULL. It was so for dicts and sets,
4650 it changed for dynars.
4652 - Fix a bunch of memleaks in dict_remove
4653 - Fix a bug in sg/server_socket opening: it failed all the time.
4655 2004-10-07 Martin Quinson
4656 - Speed up dynar lookup operation a bit.
4658 gras_dynar_get is dead.
4660 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4661 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4662 (which gives you the address of the stored data).
4664 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4665 copy of the data using an affectation to do the job and not a memcpy.
4667 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4668 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4670 It does not really speedup the dynar test because they are
4671 setting elements all the time (and look them seldom). But the dict does
4672 far more lookup than setting.
4674 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4676 2004-10-05 Martin Quinson
4677 - Allow to (en/dis)able the cycle detection at run time.
4679 Whether we should check for cycle or not is now a property of each
4680 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4681 datadesc_cycle_unset allow to remove this property when previously set.
4683 Note that the cycle detection is off by default since it impacts the
4684 performance. Watch the data you feed GRAS with ;)
4686 This property is hereditary. Any element embedded in a structure having it
4687 set have it set for the time of this data exchange.
4689 You should set it both on sender and receiver side. If you don't set it on
4690 sender side, it will enter an endless loop. If you forget on receiver
4691 side, the cycles won't be recreated after communication.
4693 - Header reorganization.
4694 Kill gras_private.h, each submodule must load the headers it needs.
4696 2004-10-04 Martin Quinson
4697 - Interface revolution: do not try to survive to malloc failure.
4699 Now, gras_malloc and friends call gras_abort() on failure.
4700 As a conclusion, malloc_error is not a valid error anymore, and all
4701 functions for which it was the only gras_error_t return value are
4702 changed. They now return void, or there result directly.
4703 This simplify the API a lot.
4705 2004-09-29 Martin Quinson
4706 - Re-enable raw sockets.
4707 Created by gras_socket_{client,server}_ext;
4708 Used with gras_raw_{send,recv}
4711 It should allow to kill the last bits of gras first version soon.
4713 This is not completely satisfactory yet (duplicate code with
4714 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4717 - Simplify transport plugin (internal) interface by not passing any
4718 argument to _server and _client, but embedding them in the socket
4721 2004-09-28 Martin Quinson
4722 - Finish the port to AIX.
4723 autoconf was my problem (segfault within the malloc replacement
4724 function. No idea why)
4726 2004-09-16 Martin Quinson
4727 - Fix some size_t madness on 64bit architectures.
4729 2004-09-08 Martin Quinson
4730 - Reduce the number of system headers loaded, overload some more system
4731 calls (such as malloc to cast the result of the system one, and work
4733 - Fix and reintroduce the config support
4735 2004-09-07 Martin Quinson
4736 - Source code reorganization to allow Arnaud to surf all over there.
4737 - Allow to document the logging categories.
4738 - Remove all uppercase from logging categories and useless cleanup in names.
4740 2004-08-18 Martin Quinson
4741 Version 0.6.2 (protocol not changed; API changed)
4742 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4743 gras_error_t), and NULL when not found. Functions expecting a msgtype
4744 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4745 hopefully usefull message.
4746 - Portability to prehistoric sparcs again
4748 2004-08-17 Martin Quinson
4749 Version 0.6.1 (protocol not changed; ABI not changed)
4750 - prealloc some buffers to speed things up
4752 2004-08-11 Martin Quinson
4753 Version 0.6 (protocol not changed; ABI expended)
4754 - The parsing macro can deal with the references, provided that you add
4755 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4757 2004-08-09 Martin Quinson
4758 Version 0.5 (protocol not changed; ABI changed)
4759 - Allow to off turn the cycle detection code in data exchange at
4760 compilation time. It should be at run time, but I'm short of time (and
4761 the config stuff is still broken). That way, we keep dict out of the
4762 critical path, which is good because the performance is poor:
4763 - search not dichotomial yet
4764 - dynar give no way to access their content and memcpy everytime
4765 - In composed data description (struct, ref and so on), stop foolness of
4766 keeping the subtype's ID, but store the type itself. This keeps sets out
4767 of the critical path, which is good since they rely on dynar and
4768 dictionaries. The only loose of that is that we cannot detect the
4769 redeclaration of a structure/union with another content (but I'm not sure
4770 the code detected well this error before anyway). We still can detect
4771 the redefinition discrepancy for the other types.
4772 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4773 it breaks the code because of type-punning used all over the place).
4774 This breaks on all non-gcc architectures (for now).
4776 All those changes (plus the buffer of last time) allow me to gain 2 order
4777 of magnitude on cruel tests consisting of 800000 array of integers on two
4778 level of a hierarchical structure (200 secondes -> 4 secondes)
4781 - the selector of reference must now return the type it points to, not
4782 the ID of this type.
4784 2004-08-06 Martin Quinson
4785 Version 0.4 (protocol changed; ABI not changed)
4786 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4787 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4788 - Implement a buffer transport, and use it by default (it relies on tcp in
4789 real life and on sg in simulation).
4790 That's a bit hackish since I had a new field to the structure to store
4791 its data without interfering with the subtype ones. Inheritance
4792 is tricky in C. And that's a kind of reverse inheritance with one class
4793 derivating two classes. Or maybe a game with java interfaces. Anyway,
4794 that's damn hard in C (at least).
4795 Moreover, I got tired while trying to ensure plugin separation and
4796 genericity in SG mode. MSG wants me to do weird things, so let's go for
4797 cruel hacks (temporarily of course ;).
4798 See comment in transport_private.h:71
4799 - do not include all the _interface headers in private but in the files
4800 which really need them (to cut the compilation time when they are
4803 2004-07-26 Martin Quinson
4804 Version 0.3 (protocol not changed; ABI changed)
4805 - Major overhault of the datadesc interface to simplify it:
4806 - shorted the function names:
4807 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4808 - add a trivial way to push/pop integers into the cbps without malloc.
4809 This allows to make really generic sub_type description, which simply
4810 pop their size of the stack.
4811 - add a function gras_datadesc_ref_pop_arr() which does what users want
4812 most of the time: Declare a dynamic array (which pops its size of the
4813 stack) and declare a reference to it. Poor name, but anyway.
4814 - kill the post-send callback, add a post-receive one
4816 2004-07-23 Martin Quinson
4817 Version 0.2 (protocol changed; ABI changed)
4818 - add some testing for cpbs in the test cases, and fix some more bugs.
4819 This invalidate again the little64 data file, since I cannot regenerate
4821 - remove an awfull optimization in the logging stuff, allowing me to:
4822 - understand it again
4823 - learn gcc how to check that the argument match the provided format
4824 - fix all errors revealed by gcc after that
4825 - internal keys of dict are not \0 terminated. Deal with it properly in
4826 loggings instead of segfaulting when the user want to see the logs :-/
4828 2004-07-22 Martin Quinson
4829 - Fix some stupid bug preventing cbps (callback postit) from working
4831 2004-07-21 Martin Quinson
4832 - Some documentation cleanups
4833 - remove the useless last argument of msgtype_declare
4834 - rename the Virtu functions to fit into the 'os' namespace
4835 - move headers src/include -> src/include/gras/ and stop fooling with
4836 gras -> . symbolic link
4837 - make distcheck is now successful
4839 2004-07-19 Martin Quinson
4841 - Build shared library also
4842 - Install html doc to the right location
4843 - stop removing maintainer files in make clean
4844 - build tests only on make check
4846 2004-07-13 Martin Quinson
4848 - No major issue in previous version => change versioning schema
4849 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4850 corresponding dataset.
4852 2004-07-11 Martin Quinson
4854 - Get it working with any kind of structure (we can compute the padding
4855 bytes remotely for all the architectures I have access to)
4856 - Implement the structure parsing macro (still not quite robust/complete)
4857 - Improvement to the remote testing toysuite
4859 2004-07-10 Martin Quinson
4860 [autoconf mechanism]
4861 - get ride of a bunch of deprecated macros
4862 - actually run the test for two-compliment, not only compile it :-/
4863 - test whether the structures get packed (and bail out if yes. Damn.
4864 Alignment is a serious matter)
4865 - test whether the structures get compacted (but respecting the alignment
4866 constraints of each types)
4867 - test whether the array fields of structures can straddle alignment boundaries
4869 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4870 (took me 2 hours to find that bug, looking at the wrong place)
4871 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4872 before sending/receiving and are used to compute the offsets of fields
4873 - Given that padding size depend even on compiler options, keep track of
4874 alignment and aligned_size only for the current architecture. Not a big
4875 deal since we send structure fields one after the other (seems
4877 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4878 program, let it work on linux/gcc/little32. portability todo.
4880 2004-07-08 Martin Quinson
4881 - import and improve remote compilation support from FAST
4882 - make sure make check works on half a dozen of machines out there
4884 2004-07-07 Martin Quinson
4885 Let's say it's version 0.0.3 ;)
4886 - Implement conversions (yuhu!)
4887 - Let it work on solaris (beside conversion, of course)
4888 - Stupid me, using rand() to generate the conversion datatests in not wise.
4890 2004-07-06 Martin Quinson
4891 - Let make dist work, since I'm gonna need it to compile on remote hosts
4892 - Let Tests/datadesc_usage write the architecture on which the file was
4893 generated as first byte.
4894 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4895 architecture descriptions.
4896 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4897 on those architectures.
4898 - Optimization: send/recv array of scalar in one shoot
4900 2004-07-05 Martin Quinson
4901 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4903 - Plug a whole bunch of memleaks
4904 - each process now have to call gras_{init,exit}. One day, their log
4905 settings will be separated
4906 - Continue the code factorisation between SG, RL and common in Transport.
4908 2004-07-04 Martin Quinson
4910 - Redistribution between SG and RL.
4911 We wanna have to accept in SG, so move accepted related parts of RL in
4912 the common part. (more precisely, the dynar of all known sockets is no
4913 more a static in transport.c, but part of the process_data)
4915 [gras_stub_generator]
4916 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4918 2004-07-03 Martin Quinson
4919 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4920 them all up in one shot)
4922 - Fix the ugly case of reference to dynamic array.
4923 - New (semi-public) function gras_datadesc_size to allow the messaging
4924 layer to malloc the needed space for the buffer.
4926 - gras_socket_close now expect the socket to close (and not its address to
4927 put NULL in it after it). This is because the socket passed to handlers
4928 is one of their argument (=> not writable).
4930 - propagate the interface cleanup from last week in datadesc, ie remove a
4931 superfluous level of indirection. User pass adress of variable
4932 containing data (both when sending and receiving), and not of a variable
4933 being a pointer to the data. Let's say that I like it better ;)
4934 The price for that is constructs like "int msg=*(int*)payload" in
4935 handlers, but it's a fine price, IMHO.
4937 - Let it work in RL (yuhu)
4939 2004-06-21 Martin Quinson
4941 - porting SG plugin and SG select to new standards (works almost).
4942 - plug memleaks and fix bugs around.
4945 - cleanup the prototype of data recv and force users to specify when they
4946 want to handle references to objects. Test case working even for cycles.
4947 - plug memleaks. Valgrind is perfectly ok with this.
4949 2004-06-12 Martin Quinson
4951 - cleanup the separation between plugin and main code in plugin creation
4953 2004-06-11 Martin Quinson
4955 - Reput hook for raw sockets, needed for BW experiments
4956 - kill a few lines of dead code
4957 [Data description] Interface cleanup
4958 - gras_datadesc_by_name returns the searched type or NULL.
4959 That way, no variable is needed to use a type desc once, which makes
4961 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4962 two parameters were strings (field name, type name), leading to
4964 [Dicos] Interface cleanup
4965 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4966 This is consistant with the dynar API.
4968 2004-04-21 Martin Quinson
4970 - Porting to new standards.
4972 - interface cleanup.
4973 There is no bag anymore, no need to take extra provision to mask the
4974 pointers behind "ID".
4975 Better splitup of functions between files create/exchange/convert.
4976 This is still a bit artificial since convert and receive are so
4977 interleaved, but anyway.
4979 - add a queued message list to procdata (the ones not matching criteria
4981 - factorize some more code between SG and RL wrt procdata
4983 - use gras_exit in example to track memleaks
4984 - get rid of gs_example now that GS is properly integrated into gras
4985 - update run_test to integrate the latest tests (datadesc)
4987 - rename WARNINGn macros to WARNn since it prooved error-prone
4989 2004-04-19 Martin Quinson
4991 - register init/exit functions within gras module mechanism
4992 - send/receive function.
4993 Convertion is not implemented, but short-cutted if not needed.
4994 struct/array elements are sent one by one (instead of block-wise), but
4995 nobody really cares (yet). Get a prototype before optimizing.
4996 - tests (using a file socket) for DD send/receive on:
4997 - base types: int, float
4998 - array: fixed size, string (ie ref to dynamic string)
4999 - structure: homogeneous, heterogeneous
5000 - chained list, graph with cycle
5001 Believe it or not, valgrind is not too unhappy with the results. The
5002 cycle happily segfaults, but the others are ok. And I'm sick of pointers
5006 - Bugfix when using a filename explicitly (instead of '-')
5008 2004-04-09 Martin Quinson
5010 - factorize more code between RL and SG in socket creation
5011 - Complete the implementation and tests of:
5013 o file (only in RL, and mainly for debugging)
5015 I lost 3 days to design a portable address resolver, and then decided
5016 that the prototype mainly have to run on my box.
5017 Addressing portability too early may be like optimizing too early :-/
5019 - use gras_init in the Tests instead of the crappy parse_log_opt
5020 (the latter function is removed)
5021 [Conditional execution]
5022 - New functions: gras_if_RL/gras_if_SG (basic support for this)
5023 [Code reorganisation]
5024 - Get rid of libgrasutils.a since it makes more trouble than it solves.
5025 Build examples against the RL library, since there is no way to disable
5026 its creation for now.
5028 For information, the beginning of coding on GRAS was back in june
5029 2003. I guess that every line has been rewritten at least twice since