1 SimGrid (3.32.1) not released yet (target december 22)
4 - Remove the MSG API: its EOL was scheduled for 2020.
5 - Remove the Java bindings: they were limited to the MSG interface.
6 - On Windows, you now need to install WSL2 as the native builds are now disabled.
7 It was not really working anyway.
8 - Support for 32bits architecture is not tested anymore on our CI infrastructure.
9 It may break in the future, but we think that nobody's using SimGrid on 32 bits.
10 - Remove the surf module. It was replaced by the kernel/models module, and that
11 refactoring took almost 10 years to properly complete.
14 - Activity::set_remaining() is not public anymore. Use for example
15 Comm::set_payload_size() to change the size of the simulated data.
16 - New function: Engine::flatify_platform(), to get a fully detailed vision of the
20 - optimize an internal datastructure, leading to a potentially big
21 performance gain (in particular with many detached comms)
24 - New option smpi/barrier-collectives to add a barrier to some collectives
25 to detect dangerous code that /may/ work on some MPI implems.
28 - Write the section of the manual about models, at least.
29 - WiFi: the total capacity of a link depends on the amout of flows on that link.
30 - Use the nonlinear callback feature of LMM to reflect this.
31 - Calibration values can be changed to match different MCS configurations
32 - See the example teshsuite/models/wifi_usage_decay/wifi_usage_decay.cpp
33 - See also "A Flow-Level Wi-Fi Model for Large Scale Network Simulation"
34 https://hal.archives-ouvertes.fr/hal-03777726
35 - Merge parameters network/bandwidth-factor and smpi/bw-factor that serve the same purpose.
36 - Same for the latency
37 - Rewrite the corresponding documentation.
38 - Allow to disable the TCP windowing modeling by setting network/TCP-gamma to 0.
39 - Finally kill the 'compound' host model. You can change the CPU or network model
40 with the default host model, as it should.
41 - Rename option "surf/precision" to "precision/timing" for clarity.
42 - Rename option "maxmin/precision" to "precision/work-amount" for clarity.
45 - Implement pthread_join in MC mode.
48 - New section in the user guide on the provided performance models.
49 - New section presenting some technical good practices for (potential) contributors.
50 - Add a section on errors and exceptions to the API documentation.
51 - Move the s4u examples to a section on their own to ease navigation.
53 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
54 (FG: issues on Framagit; GH: issues on GitHub)
55 - FG#18: Java bindings should be redone or removed
56 - FG!118: Wi-Fi callback mechanism
57 - FG!119: SMPI: add option to inject a barrier before every collective call
58 - GH#383: Segfault when adding a disk after load_platform(xml)
60 ----------------------------------------------------------------------------
62 SimGrid (3.32) October 3. 2022.
64 The Wiedervereinigung release. Germany was reunited 32 years ago.
67 - SimGrid now requires a compiler with C++17 support to compile the lib.
68 Our public headers still allow the user code to be compiled in C++14.
69 - Support graphviz v3 and ns-3 v3.36 (older versions are still supported).
70 - Tested with clang (v11, v13, v14 and v16), gcc (v7 to v13) and IntelCC v2022.2
74 - kill signal Comm::on_completion that was not working anyway.
75 - Expose signals Activity::on_suspend and Activity::on_resume
76 - New macro xbt_enforce(): similar to xbt_assert(), but throws an AssertionError
77 instead of calling abort().
78 - New: s4u::Exec::get_thread_count()
79 - Various cleanups around virtual machines:
80 - host_by_name() and friends now only return hosts. VMs are now excluded.
81 - It is now impossible to search a VM by name globally.
82 You can only search VM by name on a given PM, so either you know
83 the PM on which your VM runs and you can search by name, or you need
84 to manually iterate over all PMs to search this VM.
85 - The s4u::VirtualMachine constructor is now deprecated.
86 Please use s4u::Host::create_vm() instead.
87 - Rename s4u::VirtualMachine::on_creation() to on_vm_creation() to
88 avoid confusion with s4u:Host::on_creation() that is inherited.
89 Also s4u::VirtualMachine::on_destruction -> on_vm_destruction().
91 - One-sided communications (Comm::sendto) can now be detached,
92 and should now be more resilient to network and host faults.
95 - Added the following bindings / examples:
96 - Comm (now 100% covers the C++ interface):
97 - Comm.dst_data_size, Comm.mailbox, Comm.sender, Comm.start_time, Comm.finish_time
98 - Comm.state_str [examples: examples/python/comm-failure/, examples/python/comm-host2host/]
99 - Comm.remaining [examples: examples/python/comm-host2host/, examples/python/comm-suspend/]
100 - Comm.set_payload_size [example: examples/python/comm-host2host/]
101 - Comm.set_rate [example: examples/python/comm-throttling/]
102 - Comm.sendto, Comm.sendto_init, Comm.sendto_async [example: examples/python/comm-host2host/]
103 - Comm.start, Comm.suspend, Comm.resume [example: examples/python/comm-host2host/]
104 - Comm.test_any [example: examples/python/comm-testany/]
105 - Comm.wait_until [example: examples/python/comm-waituntil/]
107 - Engine.host_by_name [example: examples/python/comm-host2host/]
108 - Engine.mailbox_by_name_or_create [example: examples/python/comm-pingpong/]
110 - Mailbox: Mailbox.ready [example: examples/python/comm-ready/]
111 - Ptask [example: examples/python/exec-ptask/]:
112 - this_actor.exec_init
113 - this_actor.parallel_execute
116 - Added an AssertionError exception that may be thrown in case of error.
117 For instance, creating two hosts with the same name will now throw this exception
118 instead of killing the interpreter.
121 - Implement MPI_File_get_type_extent(), MPI_File_s/get_atomicity() and
122 MPI_File_get_byte_offset()
123 - Intercept getpid() calls to return the simulated ones.
124 - Fix various bugs in MPI IO.
126 Platform description & visualization:
127 - More robust sanity checks for platforms, to reject forbidden topologies with
128 a proper error message.
129 - New platform example: supernode.cpp and supernode.py.
130 The Python version generates a nice graphical representation of the platform.
131 - Bug fixes around fat-tree topologies.
132 - Allow to dump the platform topology as a CSV file representing the graph edges
133 with platform_graph_export_csv() (similar to the DOT export).
134 - Fix graphicator for "cluster" topologies (e.g. fat-tree, dragonfly).
137 - Fix a bug when using ptasks with multicores (FG!111).
140 - First bits of sthread, that intercepts pthread operations at runtime.
141 The intend is to use it together with simgrid-mc, but it is TBD.
142 - Sync MBI generators with upstream changes.
143 - Various cosmetics, small bug fixes and inner refactorings
145 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
146 (FG: issues on Framagit; GH: issues on GitHub)
147 - FG#105: "Variable penalty should not be negative!" with in-flight messages and bandwidth profiles
148 - FG#109: Application time reported by --cfg=smpi/display-timing:yes is wrong
149 - FG#110: Wait_any does not trigger new model solve when host events occur
150 - FG#111: Wrong execution time in rare cases when using multicore
151 - FG!98: Re-enable the tests for legacy stochastic profiles
152 - FG!109: Trigger new engine solve upon host events such as host on/off
153 - FG!116: SMPI/replay: Fix issue with recv of size =0
155 ----------------------------------------------------------------------------
157 SimGrid (3.31) March 22. 2022.
159 The ненасильство release. We stand against war.
161 Against the agression by a sick system that forces peoples to take arms against each other.
164 - Rework the internals, for simpler and modern code. This shall unlock many future improvements.
165 - You can now define plugins onto the DFS explorer (previously called SafetyChecker), using the
166 declared signals. See CommunicationDeterminism for an example.
167 - Support mutex, semaphore and barrier in DPOR reduction
168 - Seems to work on Arm64 architectures too.
169 - Display a nice error message when ptrace is not usable.
170 - New test suite, imported from the MPI Bugs Initiative (MBI). Not all MBI generators are integrated yet.
171 - Remove the ISP test suite: it's not free software, and it's superseeded by MBI.
174 - fix for FG#100 by ensuring small asynchronous messages never overtake larger
175 ones, conforming to the standard.
176 - replay: fix waitall behaviour to avoid forgetting requests and leaking
178 - tracing: ensure that we dump the TI traces continuously during execution and
179 not just at the end, reducing memory cost and performance hit.
180 - Update OpenMPI collectives selection logic to match current one (4.1.2)
181 - Add a coherence check for collective operation order and root/MPI_Op
182 coherence. Potentially costly so not activated unless smpi:pendantic is set
183 or -analyze is given.
186 - New signal: Engine::on_simulation_start_cb()
187 - Introduce a new execution mode with this_actor::thread_execute(). This simulate
188 the execution of a certain amount of flops by multiple threads ran by a host. Each
189 thread executes the same number of flops, given as argument. An example of this new
190 function can be found in examples/cpp/exec-threads.
191 - Reimplementation of barriers natively.
192 Previously, they were implemented on top of s4u::Mutex and s4u::ConditionVariable.
193 The new version should be faster (and can be used in the model-checker).
194 - Actor::get_restart_count(): Returns the number of reboots that this actor did.
197 - MSG_barrier_destroy now expects a non-const msg_barrier parameter.
199 New plugin: the Chaos Monkey (killing actors at any time)
200 - Along with the new simgrid-monkey script, it tests whether your simulation
201 resists resource failures at any possible timestamp in your simulation.
202 - It is mostly intended to test the simgrid core in extreme conditions,
203 but some users may find it interesting too.
206 - New solver for parallel task: BMF.
207 - More realistic sharing of heterogeneous resources compared to the fair
208 bottleneck solver used by ptask_L07.
209 - Implement the BMF (Bottleneck max fairness) fairness.
210 - Improved resource sharing for parallel tasks with sub-flows (parallel
211 communications between same source and destination inside the ptask).
213 - "--cfg=host/model:ptask_L07 --cfg=host/solver:bmf": enable the ptask
214 model with BMF solver.
215 - "--cfg=bmf/max-iterations: <N>": maximum number of iterations performed
216 by BMF solver (default: 1000).
217 - "--cfg=bmf/precision: <N>": numerical precision used when computing
218 resource sharing (default: 1e-12).
219 - This model requires Eigen3 library. Make sure Eigen3 is installed to use BMF.
222 - Modifications of the Profile mechanism, with some impact on users
223 - Addition of a new (S4U) method to init profiles from generic functions to improve versatility
224 - Fix initial behaviour of state_profiles
225 - Modify periodicity to behave like a period, and not like a loop delay
228 - Drop xbt_dynar_shrink().
231 - Made the following bindings static (previously member functions):
232 - Actor: Actor.kill_all(), Actor.by_pid()
233 - Host: Host.by_name(), Host.current(), Host.on_creation_cb()
234 - Mailbox: Mailbox.by_name()
235 - Added the following bindings:
236 - this_actor.warning()
237 - Mailbox.put_init() [example: examples/python/comm-waitallfor/]
238 - Comm.detach() [example: examples/python/comm-waitallfor/]
239 - Comm.wait_for() [example: examples/python/comm-waitfor/]
240 - Comm.wait_any_for()
241 - Comm.wait_all_for() [example: examples/python/comm-waitallfor/]
242 - Mutex [example: examples/python/synchro-mutex/]
243 - Barrier [example: examples/python/synchro-barrier/]
244 - Semaphore [example: examples/python/synchro-semaphore/]
247 - Remove target "make uninstall" which was incomplete and no longer maintained.
249 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
250 (FG: issues on Framagit; GH: issues on GitHub)
251 - FG#57: Mc SimGrid should test whether ptrace is usable
252 - FG#87: Smpi scripts fail with spaces in paths
253 - FG#100: [SMPI] Order of the message matching is not guaranteed
254 - FG#101: LGPL 2.1 is deprecated license
255 - FG#104: "make uninstall" not up-to-date
256 - GH#151: Missing mutexes for DPOR.
258 ----------------------------------------------------------------------------
260 SimGrid (3.30) January 30. 2022.
262 The Sunday Bloody Sunday release.
264 Main user-visible changes:
265 - The SimDag API for the simulation of the scheduling of Directed Acyclic
266 Graphs has been dropped. It was marked as deprecated for a couple of years.
267 We finally complete the implementation of what has been called SimDag++
268 internally, i.e., porting the different features of SimDag on top of S4U.
269 The new way to simulate the execution of dependent activities directly by
270 maestro (without any other actor) is details in the examples/cpp/dag-* series
272 - The removal of SimDag led us to also remove the export to Jedule files that
273 was tightly coupled to SimDag. The instrumentation of DAG simulation is still
274 possible through the regular instrumentation API based on the Paje format.
275 - We also dropped the old and clumsy Lua bindings to create platforms in a
276 programmatic way. It can be done in C++ in a much cleaner way now, which
277 motivates this suppression.
280 - Introduce on_X_cb() functions for all signals, to attach a new
281 callback to the signal X. The signal variables are now hidden and
282 only these functions should be used.
283 Rationale: this enables the usual deprecation scheme where functions
284 remain for 4 releases if we need to modify the signals, while the
285 current code with the signal variables directly visible prevents any
287 - New function: Engine::run_until(date), to split the simulation.
288 - New signal: Activity::on_veto, to detect when an activity fails to start.
289 - Signal change: Comm::on_start(Comm&, bool) has been replaced by
290 Comm::on_send and Comm::recv. These two signals respectively correspond to
291 when the sending or receiving side of a Comm is ready. They are raised at
292 the same locations as the former Comm::on_start signal.
293 - New function: Engine::track_vetoed_activities() to interrupt run()
294 when an activity fails to start, and to keep track of such activities.
295 Please see the corresponding example for more info.
296 - New functions: s4u::Comm::{sendto_init, set_source, set_destination} to enable
297 the use of vetoers with direct host-to-host communications. Both source and
298 destination have to set for a comm to start. Each call to these setters check
299 if all vetoes are satisfied. When it is the case, the comm starts. A use case of
300 these functions is given in examples/cpp/dag-scheduling.
301 - New functions: {Exec, Io}::update_priority allow you to modify the priority of
302 these kinds of activities during their execution. Behavior is detailed in
303 examples/cpp/io-priority/.
306 - Dynamic costs for MPI operations: New API to allow users to dynamically
307 change injected costs for MPI_Recv, MPI_Send and MPI_Isend operations.
308 Alternative for smpi/or, smpi/os and smpi/ois configuration options.
309 - Fix some issues with the replay mechanism.
312 - Function xbt::Extendable::get_data() is now templated with the type of the
313 pointee. Untyped function is deprecated. Use get_data<void>() if you still
314 want to retrieve void*.
317 - New section: "SimGrid MPI calibration of a Grid5000 cluster"
318 presenting how to properly calibrate MPI communications in SimGrid.
319 - Complete and reword the platform section, which is now completed.
322 - Thread contexts are used by default with Python bindings. Other kinds of
323 contexts revealed unstable, specially starting with pybind11 v2.8.0.
325 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
326 (FG: issues on Framagit; GH: issues on GitHub)
327 - FG#95: Wrong computation time for multicore execution after pstate change
328 - FG#97: Wrong computation time for ptask+multicore+pstates
329 - FG#98: SMPI offline simulation is inconsistent with the online simulation
330 (deadlocks / message truncation)
331 - FG#99: Weird segfault when not sealing an host
333 ----------------------------------------------------------------------------
335 SimGrid (3.29) October 7. 2021
337 The "Ask a stupid question" release.
339 We wish that every user ask one question about SimGrid to celebrate.
340 On Mattermost, Stack Overflow or using the issues tracker.
342 New modeling features:
343 - Non-linear resource sharing, modeling resources whose performance heavily degrades with contention:
344 - The total capacity may be updated dynamically through a callback
345 and depends mainly on the number of concurrent flows.
346 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
348 - Dynamic factors: model variability in the speed of activities
349 - Each action can now have a factor that affects its progression.
350 This multiplicative factor is applied when updating the amount of work
351 remaining, thereby an activity with factor=0.5 only uses half of the
352 instantaneous power/bandwidth it is allocated and will appear twice
353 slower than what it actually consumes.
354 - This can be used to model a overhead (e.g., there is a 20 bytes
355 header in a 480 bytes TCP packet so the factor 0.9583) but the novelty
356 is this factor can now easily be adjusted depending on activity's and
357 resources characteristics.
358 - This existed for network (e.g., the effective bandwidth depends
359 on the message in SMPI piecewise-linear network model) but it is now
360 more general (the factor may depend on the source and destination and
361 thus account to different behaviors for intra-node communications and
362 extra-node communications) and is available for CPUs (e.g., if you
363 want to model an affinity as in the "Unrelated Machines" problem in
364 scheduling) and disks (e.g., if you want to model a stochastic
366 - For that, resources can be provided with a callback that computes
367 the activity factor when creating the action.
368 - Example: examples/cpp/exec-cpu-factors
369 - The same mechanism is also available for the latency, which
370 allows to easily introduce complex variability patterns.
373 - Added support to programmatic platform creation in Python.
374 Example: examples/python/clusters-multicpu
377 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
378 This can only be set through the API, not through XML files.
382 - Multiple fixes to ensure reproducibility of tracing
383 - scan/excan can now be replayed
384 - wait action now uses ranks and not pid, as the other ones.
385 - smpi/init and smpi/finalization-barrier are now valid for replays.
386 - exit() is now intercepted by SMPI to avoid premature shutdown of
387 simulation. First non 0 return codes is returned as simulation return
391 * New section "Release Notes" documenting recent and current developments.
392 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
393 * Improvements in API Reference for C++ and Python interfaces.
396 - Make wifi creation compatible with ns-3 version 3.34 too.
398 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
399 (FG: issues on Framagit; GH: issues on GitHub)
400 - FG#77: Search feature of doc is broken (update sphinx theme version)
401 - FG#78: Multiple fixes for SMPI replay:
402 - TI tracing of allotallv/w was outputting wrong values
403 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
405 ----------------------------------------------------------------------------
407 SimGrid (3.28) July 14. 2021
409 The Victoriadagarna Release.
412 - C++ platform interface: Users can now describe their platform directly in C++.
413 This provides greatly flexibility and performance improvement for complex
414 platforms. Main features:
415 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
416 "zones", instead of single hosts. This feature enables the description
417 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
418 - StarZone: new zone with a Star-like topology. The routes are defined
419 as a set of links used to communicate from node to everybody (node<->ALL).
420 - Split-Duplex links: auxiliary method to create split-duplex links in
421 the platform, easing its utilisation. It automatically creates both UP
422 and DOWN links (similarly as done in XML).
423 - Please refer to the documentation and the examples included:
424 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
425 - New plugin: Producer-Consumer with monitor. Just requires to include the
426 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
427 associated example (examples/cpp/plugin-prodcons).
430 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
431 s4u::Io::wait_any(), s4u::Io::wait_any_for().
432 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
433 parameter by reference, instead of a pointer.
434 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
435 Explicitly cancel the activity to get back to previous behavior.
436 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
439 - The default SMPI compiler flags are no more taken from the environment.
440 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
441 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
443 --cfg=smpi/finalization-barrier: which can be used to add
444 a barrier inside MPI_Finalize. This can help for some codes which cleanup
445 data attached to a process, but still used in other SMPI processes.
446 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
447 is active when set to false, to keep going after a small error
448 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
449 which may or may not be problematic in the end.
451 - fix behaviour, as maximum iteration count could be ignored
452 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
453 use sampling when the same kernel is called with a different set of parameters
454 which have an impact on the timing.
455 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
456 It should now work with smpi/auto-shared-malloc-thresh.
457 - Improve error handling and reporting in multiple places
458 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
459 validity checks, truncated messages are now an error, return errors
460 when explicitely deleted handles are reused, ...)
461 - RMA: multiple fixes and stability improvements.
462 - analysis (-analyze flag in smpirun):
463 - SMPI can now report buffer leaks as well as MPI handles leaks,
464 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
465 - if -trace-call-location is used when compiling, SMPI can report
466 origin of leaked handles/buffers
467 - group leaks by type/origin in output message if possible
468 - New implemented MPI calls: MPI_Comm_test_inter
471 - Changed internal implementation of bandwidth factors in network models.
472 Models affected: CM02, LV08 (default), SMPI, IB.
473 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
474 Bandwidth factors are applied to communications to describe that users
475 cannot use 100% of the available bandwidth. For example, the default network model,
476 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
477 behavior was implemented by limiting the bandwidth available in the LMM
478 system for this flow. This may give the false impression that there is
479 bandwidth available for other flows due to its underutilization, especially
480 for the dynamic bandwidth factors used in SMPI models.
481 To avoid this, we have modified the implementation so that each flow uses the
482 maximum physical bandwidth according to the LMM system.
483 However, the actual throughput of the flow seen by the user is defined by
484 the physical bandwidth multiplied by the bandwidth factor.
485 This change impacts on the simulation results for all network models on
486 which we have bandwidth factors configured.
487 *****************************************
488 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
489 *****************************************
490 This change may impact on the timing of your simulation results.
491 Take care when comparing simulations from different SimGrid's
492 versions. Sorry for the inconvenience.
493 - Dynamic network factors: users can configure a callback to define
494 the network factors dynamically. This API is available at
495 simgrid::kernel::resource::NetworkModelIntf.
496 - Users have access to complete information about the current communication
497 to decide which factor to apply. This includes: message size, source and
498 destination hosts, links and zones traversed.
499 - Dynamic factors for both latency and bandwidth.
500 - For more details, see the example in (examples/cpp/network-factors).
501 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
502 deprecated since version 3.24, are no longer supported. Instead, use
503 "wattage_off" and "wattage_per_state".
506 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
509 - New tutorial: Model-checking and formal assessment
510 - New sections: "Demystifying the routing" and "C++ platforms"
511 - Update and improve the part on visualization in MPI and Algo tutorials.
512 - Improve the section on routing: how to define it, how it's used internally
513 - Fix many issues, broken links and missing references in doxygen and Sphinx
516 - Lua platform files are deprecated. Their support will be dropped after v3.31.
519 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
521 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
522 (FG: issues on Framagit; GH: issues on GitHub)
523 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
524 - FG#64: Configuring smpi/IB-penalty-factors
525 - FG#67: Running computation concurrently with MPI_Iallreduce
526 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
527 - FG#71: Segmentation fault on invalid gw_src/gw_dst
528 - GH#322: Issue when an actor kills his host vm
530 ----------------------------------------------------------------------------
532 SimGrid (3.27) March 29. 2021
534 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
536 Main user-visible changes:
537 - All functions marked as deprecated for one year were removed.
538 - Removed all that was related to the "Storage" type in all APIs and bindings.
539 "Disks" have been introduced more than a year ago to replace them for the better.
540 It's time to finish this replacement.
541 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
542 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
543 avoiding "initializer element is not constant" errors with some codes. Please be
544 careful to recompile your MPI codes if switching to this new SimGrid version.
545 - Network model ns-3 now requires version ns-3.28 or higher.
546 - Removed some option aliases that were deprecated since v3.23 (June 2019)
547 - 'verbose-exit' is now 'debug/verbose-exit'
548 - 'simix/breakpoint' is now 'debug/breakpoint'
549 - 'clean-atexit' is now 'debug/clean-atexit'
552 - New functions: Comm::sendto_{init,async} to initiate a communication
553 on between two (possibly remote) hosts.
554 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
555 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
556 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
557 or Mailbox::get_async<void>() if you really want to play with void*.
558 - Unify the interface of Activity::on_{start/activity}
559 - New function: Comm::get_dst_data()
560 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
563 - ptask should now work better with multicore hosts (FG#37).
564 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
565 - Internal cleanups and refactoring around models, to prepare for programmatic
567 - Modernize our tests of the performance models
571 - New section: Configuring the logs from the command line.
572 - New section: Programming API of logs.
573 - New section: Getting realistic results
574 - New section: Examples of SMPI Usage
575 - New part: SimGrid's Internals, with a section on Design goals
576 - Move back to the classical Breathe tool chain, and fix many small glitches.
579 - Handle leak checking: SMPI will now display a message if MPI handles are still
580 present at the end of a simulation. Option smpi/list-leaks:n will display the n
581 first leaked pointers and their types.
582 - smpi/display-allocs option : Display Memory allocations analysis, with hints
583 about the largest allocations. This only works for allocations with malloc/calloc
584 and not compiled with SMPI_NO_OVERRIDE_MALLOC
585 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
586 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
587 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
588 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
589 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
590 - Replay now opens and parses the trace file only once for all ranks when possible.
592 - better handling of MPI_PROC_NULL
593 - MPI_Group_incl was broken in some cases
594 - PAPI tracing now fixed
595 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
596 - SMPI will now fail properly when receive size < sent size, instead of
597 just setting MPI_ERR_TRUNCATE and keep going.
598 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
599 - Remove useless build patches and split build/run patches on the page
600 - Build in parallel when possible
601 - Test on yum-based systems
602 - New nightly tested Proxy-apps:
603 - OpenMPI mpi-test-suite
604 - Coral MCB, LSMS, Qbox, CLOMP
605 - ECP/Mantevo Cloverleaf3D, Ember
606 - Codevault bhtree_mpi
609 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
610 - Start to overhaul the implementation to modernize it (still ongoing).
613 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
614 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
615 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
616 - Deprecate xbt_dynar_to_array().
617 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
619 C binding and interface:
620 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
621 - Constness requirement has been lowered for parameter argv of sg_actor_start()
622 and sg_actor_create() It's now (char* const*). The previous version is still
623 available as sg_actor_start_() and sg_actor_create_().
625 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
626 (FG: issues on Framagit; GH: issues on GitHub)
627 - FG#37: Parallel tasks are limited to 1 core per host
628 - FG#62: Running "smpirun -replay" on large networks
629 - FG!46: Fix a few potential memory leaks in SMPI colls
630 - GH#343: Wrong consumed energy when an exec timeouts
632 ----------------------------------------------------------------------------
634 SimGrid (3.26) December 15. 2020
636 The Release release (the French lockdown was eased today).
638 Important user-visible changes:
639 - SimGrid now requires a compiler with C++14 support.
640 Sibling projects should upgrade their FindSimgrid.cmake
641 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
642 several users had difficulties to understand issues when using high bandwidth or
643 small latency events. The new value was already the default for SMPI and
644 should not cause too much performance hit. It can change some simulation
645 timings if simulation was skipping a lot of small events (which was actually
646 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
647 - On some network models (L07, CM02, LV08), an implicit loopback link is created
648 for local communication. Default bandwidth was 500MBps and latency was 15µs,
649 which was rather bad. Default values have been changed to 10GBps and 0 latency.
650 Options network/loopback-lat and network/loopback-bw have been added to adapt
651 these values if needed. Network and routing models already providing separate
652 loopbacks (clusters and such) are not impacted by this change.
653 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
654 metadata that are added to each MPI message by MPI libraries (tag,
655 src/dest, and communicator information). This may change simulation times
656 for very small message sizes, but should be more realistic.
659 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
660 - New plugins: wifi_energy and link_load. See the associated examples.
661 - New examples: wifi networks, and communication suspend.
664 - LTO is now enabled for Intel/clang compilers.
665 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
666 Setting it to "auto" will use all available cores, while setting it to n will
667 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
668 - Remove obsolete runtime option 'contexts/parallel-threshold'.
669 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
670 has been kept for compatibility.
671 - Finally remove obsolete snake_case() aliases for runtime options.
672 - Further improve the documentation.
675 - the on_exit() of each actor is also executed when the simulation deadlocks.
676 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
677 An example is provided for s4u::Comm but it should work with Exec and Io.
680 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
681 https://framagit.org/simgrid/SMPI-proxy-apps
682 - MPI/IO: supports names starting with "./"
683 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
685 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
686 building shared libraries, this may cause issues, so environment variable
687 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
688 - most temporary files should now be created in /tmp dir (or equivalent).
689 If this one does not allow execution of code (noexec flag), this may cause issues.
690 Please use another tmp directory (using TMPDIR or equivalent system variable)
694 - A large refactoring is ongoing, but this should be transparent for users so far.
697 - Drop xbt_str_split().
699 C binding and interface:
700 - The return type of the sg_actor_on_exit() callbacks is now 'void'
701 (int value was previously ignored)
702 - Many C functions were renamed to follow a common naming style
703 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
704 names are kept until version 3.30.
705 - Many MSG tests were converted to the new S4U's interface in C, that
706 was extended for that.
708 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
709 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
710 - FG#41: Add sg_actor_create C interface
711 - FG#43: xbt::random needs some care
712 - FG#48: The Impossible Did Happen (yet again)
713 - FG#50: Suspending an actor executed at the current timestamp fails
714 - FG#52: Zero-seconds timeout : "That's in the past already"
715 - FG#53: Crash while using ns-3 network model
716 - FG#54: How to suspend a comm?
717 - FG!22: Stochastic Profiles
718 - FG!24: Documentation and fix for xbt/random
719 - FG!35: Add a modeling hint for parallel links in doc
720 - FG!36: [xbt/random] Read/Write the state of the RNG
721 - FG!42: Link cumulated load plugin
722 - FG!44: Test condition variable wait_until+notify_all
723 - GF#18137: Allow different stack sizes?
724 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
725 - GH#139: Allow pthread creation in SMPI
726 - GH#336: Packet-level simulation using SMPI?
727 - GH#345: Error in the LMM while migrating a VM
728 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
729 - GH#352: pip install / python setup.py install fail to find pybind11
730 - GH!337: Fix link_energy plugin for wifi platforms
731 - GH!339: Add Mailbox set_receiver method to python binding
732 - GH!344: Cast hugepages macros parameters to int64
734 ----------------------------------------------------------------------------
736 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
738 The Palindrom Day Release.
740 Important user-visible changes:
741 - Improve the Python usability (stability and documentation).
742 - A nasty synchronization bug was ironed out, see also below.
743 - Python's doc was organized as a separate tree, now integrated with C++.
744 - C bindings of S4U were not part of the doc.
745 - The C++ doc was also improved as methods are now split by theme.
746 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
747 - This is mandatory to use the Java bindings.
748 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
749 - SimDAG++: Automatic dependencies on S4U activities (experimental)
750 - Some features are already implemented but not all of them
751 - Cannot block an activity until it's scheduled on a resource
752 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
753 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
756 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
757 - Drop support for Viva/Triva (old visualization tools), hence removed
758 TRACE_get_node_types() and TRACE_get_edge_types() functions.
761 - Require pybind11 version 2.4.
762 - Greatly improve locking during context switches with Python.
763 - Declaring Python callbacks to this_actor::on_exit is now working.
764 - No more tests are failing currently, even on win32.
767 - Actor: Merge signals on_migration_start/end into on_host_change
768 - Actor: Rename migrate() into set_host()
769 - Disk: Allow users to get the read and write nominal bandwidth values
770 - Exec: Implement wait_for(timeout)
771 - Io: Implement wait_for(timeout)
774 - Parse errors now raise a simgrid::ParseError that you may want to catch.
777 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
778 been renamed to fix a spelling error. As usual, the previous definitions are
779 kept with a deprecation warning. However, to avoid runtime errors, it is now
780 forbidden to override these deprecated methods in a derived class. Please use
781 the new names immediately if you need to override them.
784 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
785 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
786 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
787 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
788 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
789 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
790 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
791 - New module: random, providing classical random numbers generators.
794 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
795 allocations' size above which they are considered "shared" by default (as if
796 they were performed through SMPI_SHARED_MALLOC macros).
797 Default = 0 = disabled feature.
798 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
799 This can cause some troubles if codes are already overriding these. If this
800 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
801 help, but will make this feature unusable.
802 - Input check is now performed much more consistently and outputs precise
803 warnings for each error found
804 - New supported calls : MPI_File_set_view, MPI_File_get_view
805 - MPI I/O now only supports disks, as storage is discontinued.
806 - MPI I/O allows opening files without specifying absolute path
807 - Note : SMPI configuration options are now initialized by smpi_init_options()
808 call, which should be called if SMPI is not being used through smpirun.
810 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
811 - FG#9: Python bindings crashing
812 - FG#39: Missing s4u::Comm::wait_any_for example
813 - FG#42: Add support for ThreadSanitizer (TSan)
814 - FG!19: Removing RngStream
815 - FG!20: A module for RNG calls
816 - FG!21: Choice between ad-hoc and standard distributions implementations
818 - FG!25: Fix link in Mutex doc
819 - FG!26: Fix links in SMPI interface doc
820 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
821 - GH#305: Unscheduled tasks are still excuted
822 - GH#313: smpirun: manual outdated w.r.t. --help
823 - GH#321: [S4U] Get task remaining work ratio
824 - GH#323: Crash when an actor turn off his physical host
825 - GH#335: Missing links on dragonfly example svg
826 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
827 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
829 ----------------------------------------------------------------------------
831 SimGrid (3.24) October 9. 2019
833 The Clean Disk Release.
836 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
837 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
838 thus becomes the default resource to execute I/O operations. Examples
839 and the file system plugin have been modified to use disks instead of
840 storage elements. Storage can still be used but is doomed to disappear.
841 Users are thus adviced to update their simulators accordingly.
842 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
843 for consistency with pthread_barrier_wait()
844 - Host::get_englobing_zone() returns the englobing netzone
845 - Actor::on_destruction is now called in the destructor
846 Actor::on_termination new signal called when the actor terminates
848 - Global signals are now part of the Engine:
849 - on_platform_creation: after config settings, before the XML parsing
850 - on_platform_created: right after the XML parsing
851 - on_time_advance: each time the clock advances
852 - on_simulation_end: after simulation, before cleanups
853 - on_deadlock: as the name implies.
855 - sg_{actor,host,link}_{data,data_set}() now all exist.
856 Use them to attach user data to the object and retrieve it.
859 - Introduce an experimental Wifi model. It sounds reasonable
860 according to the state of the art, but it still has to be properly
861 validated, at least against ns-3 if not against reality.
862 - Improved the usability of ns-3. Several bugs were ironed out.
863 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
864 It is now expressed as 'idle:epsilon:allCores' to properly model the
865 consumption of non-whole tasks on mono-core hosts. More info in the doc.
868 - convert a new set of functions to the S4U C interface and move the old MSG
869 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
872 - Fortran bindings for DVFS have been removed.
873 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
874 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
875 - SMPI can now be selected by cmake's find_module(MPI) with
876 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
877 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
878 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
879 - trace-call-location can be used with TI traces, and replayed, and is compatible
880 with smpi/comp-adjustment-file.
881 - sleep events are now correctly traced/replayed.
882 - Default for trace-call-location is now to use file names and not full paths.
883 To revert to previous behavior (in case of collision of filenames), option
884 "smpi/trace-call-use-absolute-path" can be set to yes.
887 - Use the included xxHash as an hash implem when C++14 is usable.
888 - Option model-checker/hash was removed. This is always activated now.
889 - New option smpi/buffering controls the MPI buffering in MC mode.
890 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
891 This is useful to check for MPI compliance.
894 - New section on plugins: how to define a new one, and existing ones.
897 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
898 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
901 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
902 and <mount> tags. The rationale is finish to decouple what should be handle
903 by SimGrid kernel (i.e., resources and activities) from plugins built in
904 user space (e.g., the file system plugin). This new tag currently accepts
905 only three attributes (id, read_bw, and write_bw). All the other information
906 that was declared with the storage related tags now has to be expressed as
907 properties. An example of platform using this new tag is available at
908 examples/platforms/hosts_with_disks.xml
911 - 'expect signal' can now accept more than one potential signal.
913 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
914 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
915 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
916 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
917 - FG#31: per-actor data
918 - FG#34: SG_BARRIER_SERIAL_THREAD?
919 - FG#35: model-checker does not like buster-produced binaries
920 - FG!13: MC: complete workaround in the error msg seen on modern systems
921 - FG!15: execute_flops now logs compute
922 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
923 - FG!17: ns-3: unblock the right number of communications + others issues
924 - FG!18: Improving the performance of the ns-3 bindings
925 - GH#189: Energy consumption of parallel tasks
926 - GH#219: Error in the throughput of TCP transfer
927 - GH!330: Minor polishing of English: Configuring_Simgrid
928 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
929 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
930 - GH!334: Rework energy plugin, again
932 ----------------------------------------------------------------------------
934 SimGrid (3.23.2) July 8. 2019
937 - Nicer introduction page.
938 - Migrate the "Deploy your application" page to the new doc.
939 - Move Java as a subtree of MSG.
942 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
945 - Drop xbt_dynar_sort_strings().
948 - Really fix FG#26: Turning off a link should raise NetworkFailureException
949 - FG#27: Wrong exception thrown to wait_any when link is turned off
950 - GH#328: Java: Canceling multiple tasks in a single vm/host
952 ----------------------------------------------------------------------------
954 SimGrid (3.23) June 25. 2019
956 The Exotic Solstice Release.
959 - SunOS and Haiku OS support. Because exotic platforms are fun.
960 - Stop setting random seed with srand48() at initialization.
961 - Use addr2line as a fallback for stacktraces when backtrace is not available.
962 - Build option -Denable_documentation is now OFF by default.
963 - Network model 'NS3' was renamed into 'ns-3'.
966 - Simgrid can now hopefully be installed with pip.
969 - wait_any can now be used for asynchronous executions too.
972 - New log appenders: stdout and stderr. Use stdout for xbt_help.
973 - Drop xbt_dict_dump.
976 - 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
977 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
978 - MPI interface is now const correct for input parameters
979 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
980 - Fortran bindings for async collectives.
981 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
984 - Remove option 'model-check/record': Paths are recorded in any cases now.
985 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
986 always sparse. That's an immense gain in memory, and can even be
987 faster because of cache effects. So there is no need to clutter the
988 code to allow the user to go for the unefficient mode.
991 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
992 models proved to be more accurate than these old experiments.
994 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
995 - FG#1: Broken link in error messages
996 - FG#2: missing installation documentation
997 - FG#3: missing documentation in smpirun
998 - FG#6: Python bindings not available on PyPI
999 - FG#7: simple cmake call requires doxygen
1000 - FG#8: make python bindings an optional dependency
1001 - FG#10: Can not use MSG_process_set_data from SMPI any more
1002 - FG#11: Auto-restart actors forget their on_exit behavior
1003 - FG#12: -Denable_lto=OFF doesn't disable LTO
1004 - FG#13: Installs unstripped file 'bin/graphicator'
1005 - FG#14: Installs the empty directory 'doc/simgrid/html'
1006 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
1007 - FG#17: Dead link in doc (pls_ns3)
1008 - FG#20: 'tesh --help' should return 0
1009 - FG#21: Documentation link on http://simgrid.org/ broken
1010 - FG#22: Debian installation instruction are broken
1011 - FG#26: Turning off a link should raise NetworkFailureException exceptions
1012 - GH#133: Java: a process can run on a VM even if its host is off
1013 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
1014 - GH#326: Valgrind-detected error for join() when energy plugin is activated
1016 ----------------------------------------------------------------------------
1018 SimGrid (3.22) April 1. 2019
1020 The Easter Christmas Release (this one is somewhat late).
1023 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
1024 - Not complete yet: asynchronous activities (amongst others) are still missing
1025 - Still ongoing: the interface may change in the future. We need more
1026 testers! Please report any glitches.
1027 - No new project using Java should start now. Please switch to Python.
1030 - Some of the internal cleanups may lead to speed improvements:
1031 - The hard limitation on the amount of simulated actors with
1032 Java+Mac was removed. Now, the available memory is the only limit.
1033 - Our refcounting was tidyied, leading to 10% speedups in some cases.
1034 - We are still working on making our code robust to the actor kills
1035 and hosts' churn. Things are improving, but it's not perfect yet.
1036 - Replaced our own code to display a backtrace (that was forking addr2line)
1037 with the Boost.Stacktrace library.
1038 You won't see your backtraces without this optional dependency.
1039 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
1040 - Stop setting random seed with srand() at initialization.
1043 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
1044 XML file version remains 4.2 since old files are still compatible.
1047 - Process termination which was broken at version 3.21 has been repaired.
1048 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
1049 - Hide the examples into examples/deprecated. New users should use Python.
1052 - Drop MSG_process_create_from_stdfunc() from the API.
1053 This C++ function was a pimple in the C API, made necessary at some
1054 point by the Java bindings. This is fixed now.
1055 - Hide the examples into examples/deprecated. New users should use S4U.
1056 - MSG_process_create and MSG_process_attach now crash if the host used
1057 to run this process is off.
1058 - Fix the protype of MSG_process_on_exit()
1059 Now use (int,void*) callbacks instead of (void*,void*) ones.
1060 The implementation was ways too messy to actually work, I guess.
1063 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
1064 too often and provide more accurate timings.
1065 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
1066 - MPI_Alltoallw support
1067 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
1068 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
1069 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
1070 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
1072 - Basic implementation of generalized requests (SMPI doesn't
1073 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
1076 - Drop sg_cmdline. Please use xbt_cmdline instead.
1077 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
1078 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
1079 OsSemaphore is implemented in a portable way with C++11 threads.
1080 This should allow much more threads to be created at the same time,
1081 allowing Mac and Java users to simulate many more actors.
1082 - Implement the 'thread' factory with std::thread instead of xbt ones.
1083 It is not possible to set the stack size with threads anymore, but
1084 -fsplit-stack is the way to go nowadays when using threads.
1085 - Drop the xbt_os_thread_t module (now unused)
1086 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
1087 - Drop xbt_str_join_array().
1088 - Drop cunit, use Catch2 instead.
1091 - Many cleanups in the kernel::activity namespace. This was long
1092 overdue, and shall open the path to many future endeavors.
1095 - #132: Java: a process can not shut down its own host
1096 - #220: S4U: detached send causes exception if sender terminates soon after sending
1097 - #261: Document the parameters of parallel execution's constructor
1098 - #300: [s4u] BarrierPtr is missing
1099 - #314: SMPI args internal cleanup
1100 - #316: Fix a bug related to the CPU utilization of multi-core VM
1101 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
1102 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
1103 - #325: Turning off a host has different behavior on sleeping actors and computing actors
1105 ----------------------------------------------------------------------------
1107 SimGrid (3.21) October 3. 2018
1109 The Restarting Documentation (TRD) Release.
1112 - Convert the user manual to Sphinx for improved usability.
1113 Unstable version now hosted on https://simgrid.frama.io/simgrid
1114 Many glitches remain; stay tuned.
1115 - Restrict the Doxygen reference API to the installed header files.
1116 (documenting all internals this way is overkill + doxygen scales badly)
1117 - New tutorial on S4U (highly inspired from the old MSG one, but with
1118 a git repository to fork as a starting point, and a docker image)
1119 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
1120 - The Developper manual is still to be converted (not compiled until then)
1121 - Some parts are still missing in this conversion (such as the
1122 platform and deployment sections) while others were blindly converted
1123 and would need more love (such as the configuration flags).
1124 Things will certainly further improve in the future releases.
1127 - Fully reimplement auto_restart mechanism that was utterly broken
1128 (fix #22, #131 and #281 that were all very related)
1129 - Implement semaphores (fix #303 and #291)
1130 - When creating an actor from a function and its parameters,
1131 move-only parameters are not allowed anymore, as it would prevent
1132 the actor to be restartable if its parameters are consumed on run.
1133 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
1134 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
1135 IO operations that can be started, waited for, or canceled as a regular
1137 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
1141 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
1142 - New host variable: 'core_count'
1145 - Remove xbt_os_thread_specific features
1146 - Remove portability wrapper to condition variables
1147 - Remove xbt_os_thread_yield()
1150 - MPICH collective selector now mimics MPICH 3.3b
1151 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
1152 - MPI_Init(NULL,NULL) is now allowed
1153 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
1154 - Command-line arguments should now be handled properly in Fortran simulations
1155 - MPI Attributes and related callbacks should work in Fortran
1156 - Apps using C + Fortran should now work
1157 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
1158 - Support MPI_Error_String
1161 - Due to an internal bug, Msg.run() must now be your last line.
1162 We hope to fix it in a future release, and we are sorry for the inconvenience.
1165 - #22: Process autorestart seem to only work with CAS01 cpus
1166 - #56: Feature request: dynamic SMPI replay
1167 - #93: simgrid should not eat --help
1168 - #111: How to change the loopback link characteristics?
1169 - #116: Platform cluster radical
1170 - #117: Platform cluster/host order
1171 - #129: Replay_multiple on steroid
1172 - #131: Java : autostart does not work
1173 - #143: Setting a breakpoint at a given time
1174 - #235: Network energy models should be integrated
1175 - #264: Add ptask L07 resource tracing
1176 - #271: Dynamic and manual replay of SMPI traces
1177 - #279: Breakpoints option uses time deltas instead of absolute time
1178 - #281: Daemonized actor no longer daemonized after an auto-restart
1179 - #282: TIT convertor erase traces when using absolute path in the trace list file
1180 - #285: segfault when a s4u actor kills itself with msg process tracing activated
1181 - #286: Pajé traces are not exposing the number of cores
1182 - #287: Command-line parsing should stop when encountering '--'
1183 - #288: MPI_Init(NULL, NULL)
1184 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
1185 - #290: Method to check if a mailbox has a message ready to be consumed
1186 - #291: [s4u] semaphores do not seem available yet
1187 - #294: Alltoall collective implementation of mpich
1188 - #295: Starting a Host as OFF is not supported
1189 - #296: DTD too permissive
1190 - #299: Add a s4u barrier example
1192 ----------------------------------------------------------------------------
1194 SimGrid (3.20) Released June 24. 2018
1196 The proxy snake_case() release.
1199 - The whole API is now using snake_case() instead of mixing it with camelCase.
1200 Compatibility wrappers in place for one year.
1201 - We took the opportunity to increase the consistency of the API.
1202 Please report any remaining glitches.
1203 - New function simgrid::s4u::Host::get_actor_count:
1204 Returns the number of actors running on a specific host.
1207 - SMPI is now tested with ~45 proxy apps from various sources, with none or
1208 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
1209 - Replay: The replay file has been re-written in C++.
1210 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
1211 supported. They are stored in the trace and used when replayed.
1212 - Basic support of MPI_Cancel. Robustness not guaranteed.
1213 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
1214 (only for MPI_COMM_TYPE_SHARED).
1215 - New option: smpi/privatize-libs, to add external shared libs to be privatized
1216 by SMPI. They will be copied locally and loaded separately by each process.
1217 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
1218 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
1219 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
1220 with internal index optind with multiple processes. Only works if MPI_Init has
1221 already been called.
1222 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
1223 syntax without preprocessing tricks.
1226 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
1227 behavior tracing is no longer limited to MSG
1228 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
1229 this means that traces now need to be updated or re-obtained; both Irecv
1230 and Isend lines in your traces have now in total 5 mandatory fields:
1231 <rankid> <command> <to/from rankid> <tag> <size>
1232 To update your traces, it suffices to add a 0 for the tag here.
1233 - TIT now also supports waiting for a distinct request via MPI_Wait.
1234 Wait/Test now wait for a specific request, not just the last one that was
1235 issued. This unfortunately means another update, because we need to
1236 identify which request you want to wait for. We do this via the
1237 triplet (sender, receiver, tag), which needs to be added:
1238 <rankid> <command> <sender> <receiver> <tag>
1239 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
1240 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
1243 - The deprecation of MSG is ongoing (but this should not impact you).
1244 Many MSG functions are now simple wrappers on the C API of S4U. If
1245 you wish to convert your code to S4U, find the S4U counterparts of
1246 your MSG calls in src/msg/msg_legacy.cpp.
1247 - MSG can still be used, but won't evolve anymore.
1250 - Allow to run the Link energy plugin from the command line with
1251 --cfg=plugin:link_energy
1252 - Rename Energy plugin into host_energy
1253 - Rename Load plugin into host_load
1256 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
1257 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
1258 - Improve performance when handling timeouts of simix synchros.
1261 - Config: the C API is now deprecated (will be removed in 3.23), and
1262 the C++ API has been slightly improved.
1265 - Fix several build issues on OSX.
1266 - Move simgrid_config.h to simgrid/config.h (old header still working)
1269 - #143: Setting a breakpoint at a given time
1270 - #150: Inconsistent event names in SMPI replay
1271 - #258: daemonized actors hang after all non-daemonized actors have completed
1272 - #267: Linker error on unit_tmgr
1273 - #269: SMPI: tracing of MPI_Wait/all/any broken
1274 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
1276 ----------------------------------------------------------------------------
1278 SimGrid (3.19.1) Released March 22 2018
1280 The Fixed ABI Release.
1282 As you may know, we are currently refactoring SimGrid in deep.
1283 Upcoming SimGrid4 will be really different from SimGrid3: modular,
1284 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
1286 Our goal is to smooth this transition, with backward compatibility
1287 and automatic update paths, while still progressing toward SimGrid4.
1289 SimGrid remains open during works: The last pure SimGrid3 release was
1290 v3.12 while all subsequent versions are usable alpha versions of
1291 SimGrid4: Existing interfaces remain unchanged, but the new S4U
1292 interface is budding and the internals are deeply reorganized.
1294 Since 2015, we work hard to reduce the changes to public APIs. When
1295 we need to rename a public library symbol in S4U, we let your
1296 compiler issue an explicative warning when you use the deprecated
1297 function. These messages remain for four releases, i.e. for one full
1298 year, before turning into an error. Starting with v3.15, your can
1299 also adapt to API changes with the SIMGRID_VERSION macro, that is
1300 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
1302 Starting with this v3.19.1, our commitment to reduce the changes to
1303 the public interfaces is extended from the API to the ABI: a program
1304 using only MSG or SimDag and compiled against a given version of
1305 simgrid can probably be used with a later version of SimGrid without
1306 recompilation. We will do our best... but don't expect too much of
1307 it, that's a really difficult goal during such profund refactoring.
1309 The difference between v3.19 and v3.19.1 is that the former was
1310 accidentally breaking the ABI of MSG, while the later is restoring
1313 S4U and kernel APIs will still evolve until SimGrid4, with one-year
1314 deprecation warnings as currently. In fact, cleaning up these
1315 interfaces and converting them to snake_case() is one release goal of
1316 v3.20. But don't worry, we are working to smooth this upgrade path.
1318 In summary, new projects should start with S4U to benefit of the
1319 future, but old MSG projects should still be usable with no change.
1321 ----------------------------------------------------------------------------
1323 SimGrid (3.19) Released March 20 2018
1325 The Moscovitly-cold Spring release.
1327 SMPI user-visible improvements:
1328 - Support of MPI_Type_create_subarray
1329 - Unimplemented MPI calls now abort (instead of issuing a warning)
1330 Some harmless calls around errhandlers still only issue a warning.
1331 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
1332 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
1333 is now possible (mmap privatization was refactored to this end).
1334 Adding new SMPI instances should also be technically doable now.
1336 SMPI internal cleanups:
1337 - Removed index notion from SMPI -- no more "getPid() - 1"!
1338 In tracing files, actors are called rank-<PROCESS-ID> but these are
1339 NOT the real ranks (tracing several communicators is not supported yet).
1340 - Deprecate smpi_process_index() (will be removed in 3.22)
1341 - Better testing on classical proxy apps, through an external project:
1342 https://github.com/simgrid/SMPI-proxy-apps/
1345 - Use a graphical TOC to make it easier to find the documentation you need
1346 - Revamp the MSG tutorial
1349 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
1350 - Remove parameter of MSG_process_killall().
1351 Resetting the PID was bogus anyway (several actors could have the same PID).
1354 - Execution->setHost() can be called after start(), to migrate it.
1355 - Comm::test_any() is now implemented.
1356 - s4u::Actor now has onCreation() and onDestruction() signals.
1357 - Install some kernel header files for the users' plugins and more.
1358 Warning, their API is really not stable yet.
1359 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1360 include/simgrid/kernel/routing/*Zone.hpp
1363 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1365 Build System and other cleanups:
1366 - Remove unused run-time parameter "tracing/onelink-only".
1367 - SimGrid now works with intel compilers (MC still buggy)
1368 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1369 - tesh kills subprocesses on timeout (but not on windows)
1372 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1375 - #194: Feature request: simgrid::s4u::Comm::test_any()
1376 - #245: migrating an actor does not migrate its execution
1377 - #253: Feature Request: expose clusters as objects
1378 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1379 - #255: Tesh broken on Windows
1380 - #256: Modernize FindSimGrid.cmake
1381 - #257: Fix (ab)use of CMake install
1383 ----------------------------------------------------------------------------
1385 SimGrid (3.18) Released December 24 2017
1387 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1389 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1390 - Support for MSG should not be removed from SimGrid before 2020,
1391 but future evolutions will be limited to the integration of
1392 user-provided patches.
1395 - s4u::onDeadlock() signal, run before stopping the simulation.
1396 - s4u::Exec: asynchronous executions (abstraction of background
1398 - s4u::Host->getLoad() returns the achieved speed in flops/s
1401 - Link::name() is deprecated and will be removed in v3.21, use
1402 Link::getCname() instead.
1403 - Mailbox::getName() changed to return a std::string, use
1404 Mailbox::getCname() instead to get a char*.
1405 - Storage::getName() changed to return a std::string, use
1406 Storage::getCname() instead to get a char*.
1407 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1408 to have both allocation and deallocation of the map in user space.
1411 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1412 The amount of flops remaining to do is a vector, not a scalar.
1413 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1414 name implies on both sequential and parallel tasks.
1415 - Both changes fix GitHub's #223 using PR #237 as a basis.
1416 Thanks Michael Mercier.
1417 - Most examples were converted to S4U and hidden elsewhere at tests.
1418 Rationale: we still want MSG to work; we want newcomers to use S4U.
1421 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1422 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1425 - Switch to the faster dlopen privatization mechanism by default
1426 - Documentation improvments
1429 - Live migration is getting moved to a plugin. Dirty page tracking is
1430 the first part of this plugin. This imply that VM migration is now
1431 only possible if one this function is called:
1432 - C/MSG: MSG_vm_live_migration_plugin_init()
1433 - C/C++: sg_vm_live_migration_plugin_init()
1434 - Java: Msg.liveMigrationInit()
1435 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1439 - Define class simgrid::xbt::Path to manage file names.
1440 - Removed unused functions:
1441 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1442 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1443 xbt_graph_export_graphviz()
1444 - xbt/str.h: xbt_str_join()
1445 - Remove unused datatypes:
1446 - xbt/heap.h: use std::priority_queue or boost::heap instead
1447 - xbt/swag.h: use boost::intrusive::list instead
1450 - New link_energy plugin for the consumption of the links.
1451 - All of the operations on files and storage contents have been
1452 packaged into a plugin (src/plugins/file_system). The current
1453 public interface can be found in
1454 include/simgrid/plugins/file_system.h
1455 To use these functions you now have to initialize the plugin by
1456 calling MSG_storage_file_system_init() just after calling
1457 MSG_init() or sg_storage_file_system_init() just after creating
1461 - Remove the undocumented/untested tag <include>
1464 - Remove viva specific tracing as the tool is no longer maintained
1467 - #248: Kill of finished processes leads to segfault
1468 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1469 - #239: please implement signal s4u::onDeadlock()
1470 - #230: segfaults when exit() without run()
1471 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1472 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1473 - #222: Actor::kill() doesn't really kill and segfaults
1474 - #221: odd LMM warning when killing an actor
1475 - #120: Memory leak when the processes are forcefully killed
1477 ----------------------------------------------------------------------------
1479 SimGrid (3.17) Released October 8 2017
1481 The Drained Leaks release: (almost) no known leaks despite the tests.
1483 Even very long simulations will preserve your memory: our testsuite
1484 stresses SimGrid for over 45mn on fast machines for a coverage of
1485 over 80%, and there is only one single known leak, of about 4kb.
1489 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1490 handler may be given to free resources if the comm cannot be completed.
1491 - this_actor::send(mailbox) is now mailbox->put()
1492 - New: simgrid::s4u::Comm::wait_all()
1493 - New: Host.execute() for remote executions.
1496 - Unused option network/sender-gap is removed.
1499 - Deprecate MSG_task_isend_with_matching(): This unused feature
1500 really complicates our internals. Will be removed in v3.20.
1503 - Improved context termination. It is now possible to free resources, even
1504 when a process is forcibly killed.
1505 - ContextBoost: add support for Boost versions above 1.61.
1508 - Bring back run-time option --cfg=exception/cutpath to remove exception
1510 - Removed unused functions:
1511 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1512 xbt_str_rtrim(), xbt_str_trim().
1513 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1516 - Removed header files obsolete since SimGrid 3.12:
1517 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1519 - #3: SD_exit should be made optional
1520 - #120: Memory leak when the processes are forcefully killed
1521 - #159: Threading test regression in Actor refcounting
1522 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1523 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1524 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1525 - #191: VM migration and pstate
1526 - #192: Updating the energy consumptions of all hosts crashes with VMs
1527 - #195: All actors have PID=0 in the logs
1528 - #204: Sometimes segfault with thread contexts and mmap privatization
1529 - #222: Actor::kill() doesn't really kill and segfaults
1530 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1532 ----------------------------------------------------------------------------
1534 SimGrid (3.16) Released June 22. 2017.
1536 The Blooming Spring Release: developments are budding.
1539 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1540 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1541 - Java JAR file should be finally fully working on Mac OSX too.
1543 XML platforms: Switch to platform v4.1 format.
1544 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1545 - <zone> can be used as a synonym for the now deprecated <as>
1546 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1547 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1548 - <actor> can be used as a synonym for the now deprecated <process>
1549 - state_file and avail_file periodicity is now easier to express
1550 (check the documentation) the old behavior should still work.
1551 * Storage: not backward compatible, but it seems that nobody ever used it
1552 - Remove attribute 'content_type' of <storage_type>: was never used
1553 - Make attribute 'model' of <storage_type> optional: for future usage
1554 - Remove Bconnection model property: was never (in)validated. Replaced by
1555 the maximum of the read and write bandwidth as a resource constraint.
1558 - New and Backwards Compatibility break:
1559 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1560 When one wants to get the list of tasks whose states have changed during a
1561 simulation round, s/he has to allocate and free a dynar and use it as
1562 argument to this function. The former SD_simulate (double how_long)
1566 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1567 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1570 - New function to update the consumption of all hosts at once.
1571 - Fix the model for multi-core hosts, linear in the amount of busy
1572 cores with an abnormality for pIdle (see paper).
1573 BUG: Still not satisfactory for ptask on mono-cores.
1576 - The netzone are now available from the MSG API.
1577 The old names still work, but are now deprecated.
1580 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1581 - New option: smpi/keep-temps to not cleanup temp files
1582 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1583 mallocs mode is used, allows to change the size of the fake file used
1584 (default 1MB), to potentially limit the number of mappings for large runs.
1585 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1586 - Fortran ifort and flang compilers support
1587 - New RMA calls supported (experimental):
1588 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1589 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1590 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1591 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1592 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1593 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1596 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1597 - Behind the scene, ActivityImpl are now automatically refcounted.
1598 Removing this nasty bottleneck will greatly simplify our internals.
1601 - Replay: New function xbt_replay_action_get():
1602 Retrieve the function previously associated to an event type.
1603 - DROPPED FUNCTION: xbt_str_varsubst()
1604 - DROPPED MODULE: strbuff. We don't need it anymore.
1605 - DROPPED MODULE: matrix. We don't need it anymore.
1606 - DROPPED MODULE: lib. We don't need it anymore.
1608 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1610 SimGrid (3.15) Released March 22. 2017
1612 The Spring Release: continuous integration servers become green
1613 We fixed even the transient bugs on all target architectures:
1614 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1615 Windows, FreeBSD, NetBSD.
1617 - Rename NetCards to NetPoints.
1618 This was intended to help NS3 users, but that's not a netcard.
1619 That's a point in the routing algorithm, let's avoid wrong simplifications.
1622 - New: MSG_process_yield(). Stop and yield to other processes.
1623 - New: MSG_process_daemon(). Daemon processes are automatically killed
1624 when the last non-daemon process terminates
1625 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1626 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1627 to make room for the upcoming network energy plugin.
1628 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1631 - Ensure that an actor can kill itself with Process::exit()
1632 - Kill the obscure NativeException. Nobody want to survive the issues
1633 it denotes, so use JniException that is a RuntimeException (not to
1634 be caught explicitly).
1635 - Partial bug fix in initialization. SimGrid flags on command line were
1636 consumed at C level but stayed in the original Java String[] args.
1637 This could mess users' args[i] if SG flags were not put at the end of
1639 The SimGrid flags are now removed from the Java arguments. However,
1640 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1641 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1642 It might be if you have --log or --cfg flags in the command line.
1643 - Fix numerous memleaks all around the place. In particular, around VMs.
1647 - s4u::onPlatformCreated: right before the simulation starts
1648 - s4u::onSimulationEnd: right after the main simulation loop
1649 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1650 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1651 event from the availability_file changes the avail speed.
1652 - Links are now usable from s4u
1653 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1654 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1655 - New examples: The conversion of MSG examples to S4U has begun
1656 - Actors: create, kill, migration, and suspend
1657 - Applications: master-worker and token-ring
1658 - Action replay: communications and storage
1659 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1662 - Backwards Compatibility breaks
1663 - The SD_route_get_list and SD_route_get_size functions have been removed.
1664 They are replaced by the unique
1665 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1666 The route from src to dst is built in the links dynar whose size can
1667 be obtained with xbt_dynar_length.
1668 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1669 sg_host_route_bandwidth and sg_host_route_latency.
1670 Macros ensure the backwards compatibility, but you should fix your code
1673 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1674 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1675 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1676 We don't support 100% of it, but it should be enough. Contact us if not.
1677 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1678 naive and probably inaccurate)
1679 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1680 mapped onto a unique small file using some system magic.
1681 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1682 - Fortran: cleanups, fixes, support of user-added operations
1683 - RMA: MPI_Accumulate are applied in correct order.
1684 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1685 - Better support for MPI_IN_PLACE
1686 - Support for MPI_Win attrs and keyvals.
1687 - Support MPI_Comm_group_create, MPI_Type_size_x
1688 - MPI ops should be properly applied to all allowed types
1689 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1690 algorithm by a non failing one from mpich.
1693 - Kill the fifo data container: we don't use it anymore.
1696 - A <cluster> can now be created with different speed values to represent pstates.
1697 see examples/platforms/energy_cluster.xml for an example.
1699 SimGrid (3.14.159) Released December 28. 2016
1701 The previous release was only a crude approximation of a Pi release;
1702 this one should be a bit better.
1704 - Revert a last minute change that broke on Mac OSX.
1705 - Fix the changelog and NEWS files.
1706 - Improve the documentation, in particular of the routing module.
1707 - Integrate some patches flying around in Debian.
1709 SimGrid (3.14) Released December 24. 2016
1712 * The whole documentation was reworked and reorganized. There is
1713 still much room for improvement here, but we're on it.
1716 * We now need python3 for our testing infrastructure.
1717 * Model checking needs libevent
1720 * Now works on FreeBSD
1723 * Privatization now works on FreeBSD
1725 * Privatization is now activated by default in smpirun
1726 You can turn it off with -no-privatize if something goes wrong.
1728 * Call-location tracing for SMPI.
1729 You can add the exact location (filename / linenumber) of an MPI call to
1730 your trace files and slow down or speed up the simulation between two
1731 consecutive calls by using an adjustment file (see the documentation).
1733 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1734 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1735 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1736 values were ignored in some cases. The timings of these functions can now
1737 be significantly different.
1739 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1740 smpi/running-power is renamed to smpi/host-speed
1742 * smpi/grow-injected-times option to enable or disable multiplication of the
1743 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1744 make simulation less precise (but also much faster).
1746 * smpirun script should be (much) faster for large deployments.
1748 * SMPI tracing: fixed issue with poor matching of send/receives.
1750 * Replay: Fix broken waitall
1752 New functions and features
1753 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1755 Dropped / renamed functions and features
1756 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1757 - MSG_mailbox_is_empty() -> Mailbox::empty()
1758 - MSG_mailbox_front() -> Mailbox::front()
1759 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1760 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1761 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1762 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1763 * Don't pass the free_f parameter to property related functions:
1764 - MSG_host_set_property_value()
1765 - MSG_as_router_set_property_value()
1766 - MSG_storage_set_property_value()
1767 * VM properties. Since msg_vm_t are msg_host_t, just use
1768 MSG_host_get_property_value() and friends
1769 * VM I/O related things:
1770 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1771 - Unimplemented save/restore methods
1772 * MSG_as_router_get_property_value() was redundant with
1773 MSG_environment_as_get_property_value().
1774 - Removed MSG_as_router_*propert*() functions
1775 - Added MSG_environment_as_set_property_value() for consistency
1776 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1777 Well, they are still there for now, but deprecated with a warning.
1778 Please switch to xbt_dict_new_homogeneous() before this is removed
1780 * Task affinity. Its intended behavior (that was very badly tested
1781 and probably not really working) was deceiving what most users
1782 would have hoped here.
1783 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1784 * Option network/coordinates is now useless and should be dropped.
1787 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1790 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1791 * Added Dragonfly topology support
1794 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1795 EiBps, EBps, Eibps, Ebps for bandwidth.
1796 They may become useful to some lucky ones.
1799 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1800 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1801 * ex/async_waitAll: new example, on asynchronous communications
1804 * Memory usage should be decreased for simulations with a large number
1805 of processes. This also helps for SMPI.
1807 SimGrid (3.13) stable; urgency=low
1809 The Half Release, a.k.a. the Zealous Easter Trim.
1811 - We removed half of the lines, that were mostly experimental cruft.
1812 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1813 (not counting blanks and comments -- according to openhub.net)
1814 - The internals are now compiled in C++ (and will soon be clean C++)
1815 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1816 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1818 * Backwards Compatibility breaks
1819 - Removed Lua simulation bindings (switch to C or Java for that).
1820 Lua can still be used to describe platforms
1821 - Removed Java kernel plug-ins.
1822 Will be reintroduced after the ongoing major internals reorg.
1824 - the following functions were removed.
1825 They were too specific and should be reimplemented in a generic
1826 way, with filter function.
1827 - MSG_task_listen_from_host
1828 - MSG_mailbox_get_count_host_waiting_tasks
1829 - MSG_mailbox_put_with_timeout was removed.
1830 Please use MSG_task_send_with_timeout instead.
1832 - the SD_application_reinit function was removed. It has been a noop for a while.
1833 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1834 replaced by a more flexible mechanism.
1835 - The following functions thus do not exist anymore
1836 - SD_workstation_get_access_mode
1837 - SD_workstation_set_access_mode
1838 - SD_workstation_get_current_task
1839 - Basic estimation functions have been removed but can easily be replaced
1840 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1841 - SD_workstation_get_computation_time => amount / sg_host_speed()
1843 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1844 Use VM.getSpeed()*load/100 for the legacy behavior.
1846 - option enable_tracing was removed. It was not doing anything for a while.
1847 - In the ModelChecker:
1848 - the model-checker now ptraces the model-checked process which means
1849 you cannot use a debugger on the latter anymore (we might make this
1850 optional in the feature);
1851 - removed soft-dirty page tracking;
1852 - remove model-checked side snapshot management,
1853 MC_snapshot() and MC_compare_snapshot();
1854 - keep the MC_cut() function as a stub (it was not really working
1855 in the previous release).
1858 * All options are consistently in kebab-case. Old names are kept as alias.
1861 * Switch to platform v4 format.
1862 - Rename from 'power' to 'speed' the attributes describing the amount of
1863 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1864 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1865 - In <host> and <link>, attributes availability and state are gone.
1866 It was redundant with state and availability traces, and with peak values.
1867 - In <cluster>, attributes availability_file and state_file are gone.
1868 It was too complex and unused.
1869 - Kill <gpu>. Was not doing anything.
1870 - The DOCTYPE points to the right URL:
1871 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1872 (the file at this address now documents the changelog since its v1)
1873 - A warning is emitted for unit-less values (they are still accepted).
1874 - speed. Default: 'f' or 'flops'. Also defined:
1875 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1876 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1877 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1878 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1879 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1880 - latency. Default: 's' second. Also defined:
1881 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1883 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1884 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1887 * s4u::Host is now the preferred public interface to the Host features.
1888 sg_host_* functions are C bindings to the exact same behavior
1889 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1892 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1893 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1894 They were deprecated since a few years, and probably did not even compile anymore.
1897 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1898 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1899 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1900 encouraged to update their codes. A list of the modified functions can be found at the end of
1901 include/simgrid/simdag.h
1904 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1905 the lambda or closure passed as a parameter will run in kernel mode.
1906 All the callback functions should be rewritten to that interface at some point.
1909 * Reorganizing and cleaning the internals all around the place.
1912 * Remove old default barrier/bcast buggy algorithms (see #18407)
1913 * Various bug fixes to handle more codes
1914 * Remove the need for the --foreground option of smpirun (it is still
1915 accepted for backward compatibility).
1918 * Kill the setset data container: MC don't use it anymore.
1919 * Kill the queue data container: it made more sense with GRAS.
1920 * Kill the xbt_peer_t data type: it's useless without GRAS.
1921 * Kill rm_cb feature of config sets: it was never useful.
1922 * Kill graphxml parsing feature. It was not used.
1923 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1925 - xbt_dynar_sort_strings(), when the content is char*
1926 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1927 They throw exceptions on invalid input;
1928 * C++ support for declaring CLI flags (simgrid::config::Flag);
1929 * class for abstracting different signal backends (simgrid::xbt::signal).
1930 with no external dependencies (we need very simple signals).
1933 * refactoring and cleanup of the code;
1934 * ongoing process to cleanly separate the model-checking algorithms
1935 from the code model-checking support.
1937 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1939 SimGrid (3.12) stable; urgency=low
1941 The Facelift Release.
1944 * Require g++ v4.7 at least to not speak prehistorical C++.
1945 * Require Boost 1.48 (for signal2 component).
1946 * Java must be version 7 at least when activated.
1947 * Builds on Windows again (including Java bindings).
1948 * Tracing is now always enabled (no way to turn it out)
1949 * Remove GTNetS. It was not working anyway.
1950 * Various cleanups in the cmake scripts.
1951 * Move headers around to sort them out on installed systems:
1952 - instr/instr.h -> simgrid/instr.h
1953 - instr/jedule/* -> simgrid/jedule
1954 - simdag/datatypes.h was removed
1955 - simdag/simdag.h -> simgrid/simdag.h
1956 - msg/datatypes.h was removed
1957 - msg/msg.h -> simgrid/msg.h
1960 * Interface improvement:
1961 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1962 - Sanitize the interface in MSG_task_ module:
1963 - Merge two functions that were close enough but misleading:
1964 set_compute_duration(t) -> set_flops_amount(t)
1965 get_remaining_computation(t) -> get_flops_amount(t)
1966 - set_data_size(t) -> set_bytes_amount(t)
1967 get_data_size(t) -> get_bytes_amount(t)
1968 - Massive cleanups in the functions related to the energy
1969 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1970 - New: MSG_host_get_pstate()
1972 - msg/energy/onoff: switching hosts on and off
1974 * Interface improvement:
1975 - Rename Host.isAvail() to Host.isOn()
1976 - Rename Process.currentProcess() to Process.getCurrentProcess()
1977 - Rename Task.setDataSize() to Task.setBytesAmount()
1978 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1980 - #18874: Actually allows the GC to reclaim tasks
1984 - SIMIX_process_throw: raises an exception in a remote process
1985 * Refactoring: Separate sub-modules
1986 - libsmx: the public interface, as libc in a real system
1987 - popping: the strange dance that converts a user request into a kernel handling
1988 - smx_context_*: the virtualization mechanisms that embed the user code
1989 - smx_*: the handling of each simcalls
1990 * Interface cleanup:
1991 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1992 * Rename smx_action_t into smx_synchro_t, making explicit that these
1993 things are used to synchronize processes with their environment.
1994 For example, a communication is a sort of synchronization involving
1995 the communicating processes (that may block until the exchange) and
1996 the platform. The same can be said from computations, etc.
1998 - #18888: segfault when a process ends before its kill_time
2002 - 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.
2003 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
2004 - MPI_Type_set_name, MPI_Type_get_name
2005 - MPI_*_c2f and MPI_*_f2c functions
2006 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
2007 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
2008 - Activate a lot of new tests from the mpich 3 testsuite
2010 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
2011 - InfiniBand network model added: Based on the works of Jerome Vienne
2012 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
2013 - When smpi/display_timing is set, also display global simulation time and application times
2014 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
2015 * Collective communications
2016 - 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
2017 - 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/).
2018 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
2020 - Replay now uses algorithms from wanted collective selector
2021 - Replay can be used with SMP-aware algorithms
2022 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
2023 - 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.
2025 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
2026 - Correctly use loopback on fat-tree clusters
2027 - Asynchronous small messages shouldn't trigger deadlocks anymore
2028 * Energy/DVFS cleanup and improvement
2029 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
2030 - new: smpi_get_host_pstate()
2033 - "Full" network optimization flag was broken since Surf++
2034 - Better handling of precision flags in maxmin
2035 - Fix bug causing sometimes "Impossible" errors
2036 - Properly pass cluster properties to included hosts
2037 * Improvement of the Energy plugin.
2038 - Always update the consumption before returning that value
2039 - New property: watt_off to denote the dissipation when the host is off
2040 - New functions getWattMinAt and getWattMaxAt to retrieve the
2041 dissipation of pstates that we are not currently at.
2042 * Java: class NetworkLink renamed to Link
2043 * New function: simcall_process_get_kill_time()
2044 * Massive rename s/workstation/host/
2045 - That's intrusive, but that's good for the project consistency. Sorry.
2046 - Change config option "workstation/model" into "host/model"
2049 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
2050 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
2051 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
2052 - xbt_dirname and xbt_basename for non-POSIX systems
2054 * The model checker now runs as a separate process.
2055 * The model checker runs must now be launched with the new simgrid-mc program.
2056 * Record/Replay: the MC can display a textual representation of a path in the
2057 execution graph. It can then be replayed outside of the model checker.
2059 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2061 SimGrid (3.11) stable; urgency=low
2066 * Normalizing pointers addresses tool for better diff between logs
2069 * Add cloud examples using new VMs
2070 - examples/msg/cloud/two_tasks_vm.tesh
2071 - examples/msg/cloud/simple_vm.tesh
2072 - examples/java/cloud/cloud.tesh
2073 - examples/java/cloud/migration/migration.tesh
2074 * Add java surf examples:
2075 - examples/java/surfPlugin/surf_plugin.tesh
2076 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
2077 - examples/java/surfCpuModel/surf_cpu_model.tesh
2078 * Add SMPI+MSG example:
2079 - examples/smpi/smpi_msg_masterslave/
2084 - msg task destroy cancel test
2085 - msg_host on/off test
2086 * Move all tests in testsuite to teshsuite (adding tesh files)
2087 * Restructure teshsuites
2088 - one folder for each kind of test
2089 * Restructure AddTests.cmake
2091 - structure the order of tests (with sections)
2094 * Add virtual machine
2095 - creation of a VM on a PM
2096 - migration of a VM from a PM to another PM
2098 - MSG_process_join(msg_process_t process, double timeout)
2099 - msg_bar_t MSG_barrier_init(unsigned int count)
2100 - int MSG_barrier_wait(msg_bar_t barrier)
2101 - void MSG_barrier_destroy(msg_bar_t barrier)
2102 - msg_as_t MSG_environment_as_get_by_name(const char * name)
2103 * New option "msg/debug_multiple_use" to help debugging when a task is used
2106 - Improvements and finalization of MSG_storage, MSG_file APIs and their
2108 - Increase code coverage in test suites
2112 * Protect context stacks against stack overflow. The number of protected memory
2113 pages allocated on the top of each stack (1 by default) can be configured
2114 with the new command line option --cfg=contexts/guard_size:<value>.
2115 * Simcalls are now generated by a python script that
2116 - generates files included by SimGrid
2117 - checks that all the functions exist, and proposes prototypes
2119 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
2120 host_set_data, host_get_data
2122 - simcall_process_join(smx_process_t process, double timeout)
2123 * Fix bug where sleeping processing could not be suspended.
2126 * Translate surf models from C to C++
2127 - Generic classes for all models: Model, Resource, Action
2128 - A generic interface for each kind of model (CPU, Network, Storage
2129 Workstation, WorkstationVM)
2131 * Translate surf routings from C to C++
2132 * Add callbacks using sigc++ or boost::signals2
2133 - Add callback functions for resource creation/destruction
2134 - Add callback functions for action state change
2135 - Handle Energy as a plugin
2136 * Replace swag by boost::intrusive
2137 * Add new routing models for clusters. For documentation, see
2138 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
2139 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
2140 parameters for cluster tag
2141 - Fat trees, with topology="FAT_TREE" and
2142 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
2143 - see examples/platforms/torus_cluster.xml and
2144 examples/platforms/fat_tree_cluster.xml
2145 * More documentation
2148 * Hostfiles support host:nb_processes construct to deploy several processes on
2150 * Collective communication algorithms should not crash if used with
2151 improper number of nodes and report the error.
2152 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
2153 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
2154 MPI_Dims_create, MPI_Cart_sub are supported.
2155 * New interface to use SMPI programmatically (still depends on MSG for
2156 some parts, see examples/smpi/smpi_msg_masterslave):
2157 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
2161 * Global variables privatization in MPI executables is now performed at runtime
2162 with the option smpi/privatize_global_variables (default:no).
2163 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
2164 dynamic libraries loaded by the application are not privatized (static
2165 linking with these libraries is advised in this case)
2168 * Options defined in XML work correctly now.
2171 * New cmake option, enable_lib_in_jar, to control whether native libraries are
2172 copied into simgrid.jar or not (ON by default). Use this option if you want
2173 to reduce the size of the installed simgrid.jar, *and* the native libraries
2174 are kept installed elsewhere.
2175 * Surf binding with SWIG (code generated in maintainer mode only):
2176 - plugin to handle callbacks
2177 - CPU model only for the moment
2180 * Supernovae build mode is definitively removed. It was used to improve
2181 inlining and inter-module optimizations. It is nowadays superseded by
2182 link-time optimizations commonly available in compilers.
2183 * Update ns-3 find lib. Bindings for ns-3 should work again now.
2184 * Add boost dependency for surf++
2185 * Add new macro for tests
2186 - ADD_TESH(name <tesh_args>)
2187 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
2191 - xbt_bar_t XBT_barrier_init(unsigned int count)
2192 - int XBT_barrier_wait(xbt_bar_t barrier)
2193 - void XBT_barrier_destroy(xbt_bar_t barrier)
2194 * Make the xbt_os_time module public
2196 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2198 SimGrid (3.10) stable; urgency=low
2200 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
2203 * Reintegrate Java to the main archive as desynchronizing these
2204 package is not acceptable anymore (Java is now considered stable)
2205 * Add explicit synchronization facilities through semaphores
2206 * Bug fix: Task.setDataSize() only changed the C world, not the value
2207 cached in the Java world
2210 * Dramatically change the way files are handled. API and internals changed, but
2211 this part of MSG was not considered as production grade either.
2212 * Add explicit synchronization facilities through semaphores
2213 * Add a new function MSG_host_get_process_list()
2214 * Preliminary DVFS support (see examples/msg/energy/ for details)
2217 * SMPI is now included directly in the libsimgrid as the windows
2218 linker doesn't force us on splitting it anymore.
2219 * Improvements of the SMPI replay tool:
2220 - Most of the collective communications are now rooted in the same process as
2221 in the original application.
2222 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
2223 used until now). Multiple data types can now be used in a trace.
2224 - The replay tool now supports traces produce either by TAU or a modified
2226 - Bug Fix: the compute part of the reduce action is now taken into account.
2227 - Gatherv collective is now supported
2228 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
2229 of smpirun outputs time independent traces for the current run. One file
2230 is created per process. If too many processes are simulated, this behavior
2231 can be changed to one file for all processes by using the
2232 tracing/smpi/format/ti_one_file flag
2233 * smpirun generates the host file if needed (with given host count and platform)
2234 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
2235 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
2236 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
2237 --cfg=smpi/coll_selector:(mpich/ompi)
2238 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
2240 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
2241 happening inside a collective SMPI call.
2242 * Fix the behavior of complex data types handling.
2243 * Make MPI_Wtime another synchronization point to take computations into
2245 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
2246 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
2247 * Add all missing Fortran bindings, SMPI should work with Fortran 90
2248 (no privatization of global variables yet)
2249 * Preliminary DVFS support (see examples/smpi/energy/ for details)
2252 * Verification of liveness properties is now available for SMPI applications
2253 (in addition to MSG applications)
2254 * Bugged examples using SMPI in examples/smpi/mc/
2255 * Add --cfg=model-check/visited option. Allows the verification of infinite
2256 programs. Detection of loops in the execution thanks to the system state
2257 comparison and reduction of the state space to explore. Can be combined with
2258 DPOR for safety properties.
2261 * Allow to change SimGrid configuration (see --help) within the code
2262 thanks to SD_config() as it can be done in MSG.
2263 * Add a new function SD_task_set_amount() upon user request.
2266 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
2267 * Remove rule based routing (no more PCRE dependency)
2268 * Add a limiter_link option to cluster tag, to specify a maximum reachable
2269 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
2270 * Add a loopback_bw and loopback_lat options to cluster tag.
2271 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
2272 routing. Now peers are encapsulated in an AS and have their own private
2273 router but this is transparent.
2276 * Our own implementation of getline is renamed xbt_getline, and gets
2277 used even if the OS provide a getline(). This should reduce the
2278 configuration complexity by using the same code on all platforms.
2279 * New type: xbt_cfg_elm_boolean.
2280 * Allow to use yes/no for boolean configuration options in the command line.
2281 * Allow to disable SimGrid cleanups at exit from command line option.
2282 There are situations where one may want a simulation to end with an exit.
2283 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
2284 annoying when scripting around the simulator. Adding a
2285 --cfg=clean_atexit:no allows to circumvent this issue.
2288 * Lots of memory leaks were corrected in this release.
2289 There is no known memory leaks anymore, in all of our 600+ tests.
2290 * New command line option --version, to get SimGrid version information.
2291 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
2293 * Supernovae builds are deprecated, and expected to be removed in the next
2296 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2298 SimGrid (3.9) stable; urgency=low
2300 The Grasgory release: GRAS is really dead now.
2302 * Complete overhaul of the internal host structures scheme.
2305 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
2306 was considered as experimental and badly maintained since then.
2307 * Keeping it was thus a trap to our potential users, that could take
2308 it instead of MSG or SMPI by mistake despite is pity state.
2309 * GRAS seems to have very few users (if any), and no one volunteered
2310 to maintain it further. It also induces a lot of XBT code (for
2311 portability sake), that must be maintained too.
2312 * For all these reasons, we killed GRAS. If someone wants to revive it
2313 in the future, don't cry, our git history still remembers of GRAS.
2316 * Major overhaul. Merge our documentation again as time proved that
2317 splitting it was really not helping our users.
2318 * Further improve the developer documentation to help newcomers
2319 hacking on SimGrid itself. The user documentation (and in
2320 particular, the beginner documentation) is still in a sorry state.
2323 * Now works on Windows too!
2324 * Much more extensive test suite, from MPICH
2327 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
2328 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
2329 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
2330 * Bug fix: let task be scheduled when the last dependency to be solved is
2331 a control dependency.
2332 * Remove SD_load_environment_script function.
2333 Use the C sg_platf function if you want to declare a platform
2337 * New function: MSG_process_get_number()
2338 * Old function documented: MSG_config()
2339 * Remove MSG_load_platform_script function
2340 Use the C sg_platf function if you want to declare a platform
2344 * Change the default value of the TCP_gamma constant (maximal size of TCP
2345 congestion window) to a more realistic 4MiB value. If you notice changes in
2346 your simulation results, you can fall back to the previous 20k tiny window
2347 by adding --cfg=network/TCP_gamma:20000 on command line.
2348 * (Hopefully) fix a bug wrt periodic availability and state traces
2349 * Bug fix: use default values at start when first event in availability/state
2350 trace is not at time 0.
2353 * remove the "new_" part of function name sg_platf_new_trace_connect
2354 (resulting in sg_platf_trace_connect), since it does not create
2358 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2359 use case where it's really mandatory, and maintaining it was a pain in
2361 * New: xbt_fifo_search(), search an item with a user-provided
2362 comparison function instead of dumb pointer comparison.
2365 * Fix the lua deployment:
2366 Use `simgrid.init_application()` before deployment instead of
2367 `simgrid.msg_register_application()` after.
2370 * Transfer the tracing files into the corresponding modules.
2372 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2374 SimGrid (3.8.1) stable; urgency=low
2376 The "we are told that some people want to also *install* the simgrid
2379 * Add missing file "tesh.1" to the archive.
2381 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2383 SimGrid (3.8) stable; urgency=low
2385 The Psssshiiiit release: SimGrid jumps into the Cloud.
2388 * Add an experimental interface to manipulate VMs. They are mainly
2389 process groups with very few intrinsic semantic, but they should
2390 allow you to build the semantic you want easily.
2391 * New function: MSG_host_set_property_value()
2392 * New function: MSG_process_on_exit(). To clean memory in all cases.
2393 * Bug fixes that made the host (and link) failures unusable.
2394 * Add a way to auto-restart process when the host in which they are
2395 executing comes back (ON_FAILURE="RESTART" on deployment file,
2396 MSG_process_auto_restart_set).
2397 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2398 please stop using the old ones, they are DEPRECATED.
2400 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2401 Please use MSG_init() instead. (reducing the amount of entry
2402 points in the library helps us).
2403 * Make it impossible to link against the wrong version of the lib
2404 * Deprecate MSG_clean(). No need to call it anymore.
2405 * Function MSG_get_host_number() is not deprecated anymore.
2408 * Split the doc into a user guide and a reference guide.
2409 * Start a developper guide to help people hacking on SimGrid.
2412 * Enable tracing by default. This modules rocks you should use it.
2413 * Remove option custom_flags. Now use environment variables CFLAGS
2415 * Use default cmake things to detect lua instead of home grown ones.
2416 * New option "enable_mallocators" to disable mallocators, for debugging
2417 purpose ("on" by default).
2420 * Bug fixes around the resource failures: don't let the processes
2421 survive the host they are running onto.
2422 * Add an interface to auto-restart processes when the host in which they are
2423 executing comes back.
2424 * Ensures that SIMIX_clean is called automatically. It's not part of
2425 the public interface anymore (bindings should be updated).
2428 * Bug fix for when SD_Simulate is called with a positive value: be careful
2429 when comparing doubles. Sometimes they are different for non significant
2431 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2432 parallel task whose initial work is distributed among host according
2433 to the Amdahl's law. Such tasks are created with a parameter alpha
2434 that corresponds to the non-parallelizable part of the computation.
2435 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2436 between two sets of workstations assuming a 1D block distribution (each
2437 workstation owns a similar share of data) on both sides.
2439 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2440 Data redistribution will be automatically scheduled once parent and child
2441 are both scheduled. The filling of computation_amount and
2442 communication_amount structures is now done seamlessly thanks to the chosen
2444 * New function SD_workstation_dump to display various information
2445 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2446 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2447 bandwidth on the route onto which the task is scheduled and the amount of
2449 To divide the nominal bandwidth by 2, the rate then has to be:
2450 rate = bandwidth/(2*amount)
2451 * Compute tasks that have failed can now be rescheduled and executed again
2452 (from their beginning)
2453 * Increasing source code coverage (src/simdag is now covered at 95.8%
2457 * Re-implement time-independent trace replay using SMPI (at the
2458 smpi_smp_* level) instead of MSG. This should replace
2459 examples/msg/actions/actions.c
2460 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2461 hindexed and structs)
2462 * Implement the exchange of non-contiguous data.
2463 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2464 * Correct behavior of smpi/sender_gap and set its default value to 0
2465 * Add option to asynchronously send small messages to allow better
2466 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2467 specifies the size in bytes under which messages will be asynchronously sent.
2468 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2469 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2472 * New C interface to define a platform: XML is now optional.
2473 For more info, please check include/simgrid/platf.h.
2474 * New interface to define random platforms from the C:
2475 For more info, please check include/simgrid/platf_generator.h and
2476 examples/msg/masterslave/masterslave_platfgen.c
2477 * Export a sg_cmdline dynar containing all the arguments we got from
2481 * Two new tracing options for adding comments to trace file so you
2482 can track your experiments (see --help-tracing for details).
2483 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2484 * Adding the SimGrid version that generated the trace file as a comment.
2485 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2486 * Fix to avoid key clashes on Paje links
2487 * Other minor fixes related to the Paje specification
2490 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2491 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2492 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2493 * Greatly improve the robustness of mmalloc to user errors (such as
2494 using an area after freeing it, or freeing it twice)
2496 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2498 SimGrid-java (3.8.1) stable; urgency=low
2500 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2501 Warning: all methods are visible, but only some of them are
2502 implemented so far. Check the source (src/jtrace.c)
2503 for further information.
2504 * New module: org.simgrid.msg.File (SimGrid File management functions)
2505 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2506 * Change the meaning of Process.restart: now restart the process from
2507 the beginning, like MSG_process_restart in C.
2508 * Add Process.setAutoRestart: handling of process restart when failed
2510 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2511 you to retrieve the properties of the processes/hosts
2512 * Deprecate Msg.clean(): you can just forget about it now.
2513 * New function Process.getCount(), that only works when compiling
2514 with the not yet released version 3.9 of the C library.
2516 * New context factory based on Coroutines. It mandates a modified JVM
2517 but then, the simulations run about five times faster, and there is
2518 no limit to the amount of processes (beside of the available memory).
2520 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2522 SimGrid (3.7.1) stable; urgency=low
2525 * Restore the prototype of MSG_process_create_with_environment() to
2526 the pre-3.7 situation by removing the kill_time argument.
2527 * Add a MSG_process_set_kill_time() function instead.
2530 * Fix weird behaviors when dealing with parallel tasks.
2533 * Simgrid is now built as a dll.
2534 * Simgrid-java now works on Windows.
2535 * Simgrid-Java is now included into Windows package.
2538 * First pre-build package for MacOSX.
2541 * Fix compilation when using MSG_USE_DEPRECATED.
2542 * Fix some compilation issues on Macs and Windows.
2543 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2544 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2546 We discovered that the Lua console is broken, but we are missing the
2547 manpower to fix it right now. The problem existed in 3.7 too, so we
2548 are not blocking the release for that. Sorry if you depended on this
2549 feature, any help would be really welcome.
2551 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2554 SimGrid-java (3.7.1) stable; urgency=low
2556 The "Java aint got to be bloated and slow" release
2559 * Various internal cleanups and performance improvement
2560 Simulations are expected to run up to twice faster or so
2561 * Make Process.kill(process) an instance method, not a static one
2562 * User processes are not java.lang.Thread subclasses.
2563 This breaks the compatibility (sorry), but previous API was
2564 brain-dead, making it impossible to have non-trivial
2565 initializations in the process constructor.
2566 * Require a full constructor per Process sub-class.
2567 Kinda breaks the compatibility (sorry), but this allows a much more
2568 efficient way to launch the processes at simulation startup.
2569 * Do not embeed our version of semaphores, java 1.5 can be considered
2570 as sufficiently prevalent for us to not dupplicate its features.
2574 * Add examples for almost every part of the API
2575 We spotted and fixed a lot of bugs in the process
2577 * New module: asynchronous communication API
2578 * New function: Process.sleep()
2579 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2580 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2581 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2583 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2585 SimGrid (3.7) stable; urgency=low
2587 The "spring cleanups (before the next Big Project kicks in)" release.
2590 * We can specify the SMPI latency/bandwidth factor with command line
2591 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2592 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2593 You can also use the "config tag" from platform file by adding this line
2594 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2595 (see "example/platforms/tag_config.xml" to use "config tag").
2596 Note that the command line supersedes the platform file configuration.
2597 * Change the correction factors used in LMM model, according to
2598 the latest experiments described in INRIA RR-7821.
2599 Accuracy should be improved this way.
2600 * Use the partial invalidation optimization by default for the
2601 network too. Should produce the exact same results, only faster.
2602 * Major cleanup in surf to merge models and split some optimization
2603 mechanisms from the core of the models. As a result you can now
2604 specify which model to use (e.g., --cfg=network/model:LV08
2605 --cfg=cpu/model:Cas01) and which optimization mode to use
2606 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2607 Incompatible combinations should err at initialization. See
2608 --help-models for the list of all models and optimization modes.
2609 * The CLM03 workstation model was dropped for simplicity because it
2610 used the deprecated CM02 network model. Use default instead.
2611 * Rename the TCP_gamma configuration option to network/TCP_gamma
2612 * Rename the coordinates configuration option to
2613 network/coordinates, and document it
2614 * Use now crosstraffic keyword instead of the terribly misleading
2615 fullduplex keyword. It is activated by default now in the current
2616 default model, use --cfg=network/crosstraffic:0 to turn it off.
2617 * Ongoing refactoring the model parsing to make XML files optional
2618 See include/simgrid/platf.h for details (still to be completed)
2621 * Major overhaul of the documentation. Almost instructive now :/
2622 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2623 functions and friends. This interface was considered as
2624 deprecated since over 2 years, it's time to inform our users that it is.
2625 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2626 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2627 use these (crappy) functions in your code.
2628 These functions will be removed soon. Stop using them now.
2629 * Deprecate MSG_get_host_{table,number}
2630 Implement MSG_hosts_as_dynar() instead.
2631 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2632 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2635 * Stabilize the parallel execution mode of user contexts
2636 * Introduce configuration variables to control parallel execution:
2637 - contexts/synchro: Synchronization mode to use when running
2638 contexts in parallel (either futex, posix or busy_wait)
2639 - contexts/parallel_threshold: Minimal number of user contexts
2640 that must be part of a scheduling round to switch to parallel
2641 execution mode (raw contexts only)
2642 * Fix bugs that prevented to use suspend/resume along with
2643 synchronization structures.
2644 * Fix bugs in process termination that lead to invalid memory access
2645 in very specific conditions.
2648 * Introduce a parallel mode for the models (controlled by surf/nthreads
2649 configuration item). In our tests, running the models in parallel
2650 never lead to any speedups because they are so fast that the gain
2651 of computing each model in parallel does not amortizes the
2652 synchronization costs, even when ultra fast futexes are used.
2653 This is released anyway because YMMV.
2656 * Performance boost by using a swag internally to compute the set of
2657 tasks that are finished and should constitute the return value of
2661 * Enable it by default now that it is considered rather stable.
2664 * Documentation of the tracing functions.
2665 * Performance gains when tracing categorized/uncategorized resource
2666 utilization by avoiding calls to get route when updating resource
2667 variables. LMM constraints are being used instead.
2668 * API changed to set task categories. Use MSG_task_set_category instead
2669 of TRACE_msg_set_task_category, and SD_task_set_category instead
2670 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2671 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2672 * Examples for link user variables added (see at examples/msg/tracing/)
2673 * Deprecated function TRACE_msg_set_process_category completely removed
2674 * Trace header updated according to the latest Paje file format
2675 * Tracing network lazy updates, no longer obligate users to use full updates
2676 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2677 * Experimental: let user code declare/set/push/pop application states for hosts
2678 * API changed to allow the manual creation of graph configuration files
2679 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2682 * Improve the API of Lua MSG bindings, using the Lua spirit.
2683 * Each simulated process now lives in its own Lua world (globals are
2684 automatically duplicated). It helps writing simulators. Will allow
2685 to run Splay programs within SimGrid in the future.
2686 * Add a Chord example in Lua, equivalent to the MSG one.
2689 * Start the implementation of a solution to express temporal
2690 properties, not only local assertions. This is still an
2691 experimental work in progress, stay clear from it to be safe.
2695 - Add new runtime parameters --help-logs and --help-log-categories
2696 to display information about supported logging parameters and
2698 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2699 don't exists anymore.
2700 * Mallocators: allow value NULL for the reset function.
2702 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2703 create homogeneous dictionaries, where all the elements share the
2704 same free function. Non homogeneous dictionaries will be
2705 deprecated in the next release.
2706 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2707 - Multi-level dictionaries are deprecated.
2709 - new function xbt_dynar_search_or_negative() that is useful when
2710 you have less than 2 million elements in your dynar and don't
2711 want of the extra complexity of catching exceptions when the
2712 element is not found.
2714 - Make xbt_os_thread module (for thread portability) public.
2715 Documentation is still to come, sorry.
2717 - Cleanups and simplifications to make it maintainable again.
2718 - Exotic features (such as memalign and valloc) were removed.
2719 - The metadata were extended and improved so that the
2720 model-checker becomes able to explore and inspect the heaps.
2721 - This may induce a performance drop when enable_model-checking is
2722 ON in cmake (even if it's not used in the simulation), but it is
2723 necessary at this point to get MC working.
2725 Turn model-checking OFF if simulation performance matters to you.
2726 Not enabling it at runtime is not enough, disable it in cmake.
2728 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2731 SimGrid (3.6.2) stable; urgency=low
2733 The "Not coding new stuff allows to polish old things" release.
2736 * New bindings to the NS3 packet level simulator (experimental)
2737 * Use the raw (efficient) execution contexts instead of the sysv
2738 (portable) ones when possible.
2739 * libpcre is now mandatory in any cases since not using it led to
2740 severe performance loss and possibly other issues
2741 * Update the XML platforms:
2742 - G5K: include the latest machine in Nancy
2743 - GridPP and LCG: new platforms
2744 * Documentation was partially updated, at least (more to come)
2746 Bug fixes, cosmetics and small improvements
2747 * Free terminated processes before the end of the simulation to avoid
2748 exhausting the memory of users having very dynamic amount of
2750 * Bug fix and cosmetics about canceling non-running tasks
2751 * Bug fix about the dot loader's issues when using libcgraph
2754 * Create an installer for windows with nsis (amd64 and win32)
2755 - Add an hello world project to illustrate simgrid project creation.
2756 - Embed libpcre into the Simgrid installer to avoid
2757 its compilation burden
2758 * The raw execution contexts should work on Apple now
2759 * Port to Windows 64 bits
2760 - Sysv contexts now have an implementation for this arch
2761 - GRAS communication features now support this arch
2762 * Drop support for borland compiler on windows
2763 - this code was not maintained, and we kinda depend on gcc nowadays
2764 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2765 * Fix portability issue on unstable ubuntu: linker became picky on
2768 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2770 SimGrid (3.6.1) stable; urgency=low
2772 The "Oops, we broke Macs too" release
2775 * Fixed contexts detection so that raw ones are used when possible
2776 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2777 leads to a strange error, with user code segfaulting sometimes when
2778 the generated code is not perfectly aligned (which is not
2779 controllable from the user side, depends on the amount of code)
2782 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2784 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2786 SimGrid-java (3.6) unstable; urgency=low
2789 * Split of every thing from simgrid v3.5 into a separate package.
2791 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2793 SimGrid (3.6) stable; urgency=medium
2795 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2798 * Bindings now constitute their own package, separated from the main one.
2799 Rationale: reduce our maintenance nightmare by reducing the module coupling
2800 They will soon be released on their own on gforge.
2802 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2803 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2805 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2806 * It's not quite deprecated for now because we have no replacement,
2807 but it may soon become the case.
2810 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2811 * Fortran: New user-level cache variable to store the rank of the running
2812 process. This improves performance by an order of magnitude.
2813 * C: New coccinelle script to automatically locate and modify global and
2814 local static variables.
2815 * Improved SMPI network model with a sender-side gap to account for multiple
2819 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2820 only say if a communication is finished, no matter whether it succeeded or
2821 failed. You can call MSG_comm_get_status() to know the status of a finished
2823 * New function MSG_task_dsend() to send a task and detach it. When a
2824 communication is detached, you are never notified of its success or failure
2825 and the memory is released automatically once it is finished. This function
2826 is useful when you don't care about the end nor the success of a
2828 * Change the prototypes of action replay. Sorry for inconvenience,
2829 but this is really more efficient this way (and to adapt your code,
2830 you just have to fix the initialization, that shouldn't be too long)
2831 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2832 ever managed to do anything useful with it.
2833 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2834 takes a dynar of communications. It returns immediately and gives the
2835 index of a finished communication (if any).
2836 * New example: a basic implementation of the Chord P2P algorithm.
2839 * New model for multi-core CPUs. You can now use the core attribute to
2840 precise the number of cores of a host. This is a basic model. Every
2841 process running on the host receives at most the power provided in
2842 the DTD (throughput<=power). Total throughput of process cannot exceed
2844 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2845 router linked by an up-link and a down-link (possibly asymmetrical).
2846 This kind of pattern allows to easily build last-mile model style platforms.
2847 Aggregating such patterns in a rule-based AS is thus the technique of
2848 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2849 * New model for Vivaldi routing. We transformed the Vivaldi network model
2850 into a Vivaldi routing model (based on the rule-based model). This allows to
2851 combine Vivaldi based latencies with last-mile platforms.
2854 * Added a check for NaN of IEEE754 infinite in the double entries of
2856 * Introduce a new context factory "raw", highly inspirated from the
2857 ucontext factory, but using manually crafted functions in assembly to
2858 do the work in an efficient manner.
2859 * Allow to change the used context factory at run time, not only at
2860 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2861 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2862 contexts at runtime (only with raw contexts or ucontexts).
2863 * Completely rewrote this module to allow parallel execution of user
2864 processes. Use --cfg=contexts/nthreads:N to execute user processes
2865 with N parallel threads (the default is 1, meaning no parallelism).
2866 * Allow to decide dynamically between sequential and parallel modes.
2867 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2868 processes in parallel only when their number is greater than or equal to P
2870 * Added a check for NaN of IEEE754 infinite in the double entries of
2874 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2875 won't output the state of processes when interrupted with Ctrl-C
2876 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2877 NULL-terminated array. This may solve backward compatibility issues
2878 due to the change to return type of SD_simulate. See also:
2879 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2880 * Add new macros with variable number of arguments.
2881 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2882 - in xbt/asserts.h: xbt_assert
2883 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2884 - in xbt/ex.h: THROWF and RETHROWF.
2885 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2887 * Change xbt_die() to accept a format string with arguments, just like printf.
2888 * New data structure: xbt_lib_t, like a dict but more general and with better
2892 * New configuration options
2893 Options triva/categorized and triva/uncategorized can be used to generate
2894 graph configuration files for Triva visualization tool.
2895 * Configuration option tracing/platform is renamed to tracing/categorized
2896 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2897 * New TRACE_user_link_variable interface:
2898 User provides the name of the link and the tracing variable to attach to it
2899 * the declaration of tracing categories must be done after the environment
2901 * simpler tracing interface, just one way to declare categories
2902 TRACE_category or TRACE_category_with_color, it is up to you
2903 * links in the trace file are again identified by their names
2904 * trace contains the full platform hierarchy exactly as declared using the ASes
2905 * Options tracing/msg/[task|process]:1 groups the process by hosts
2906 for both cases, tasks and processes must have names that are unique during the simulation
2907 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2908 * The experimental option tracing/msg/volume is deprecated
2909 its functionality may be reincorporated if needed
2911 The tracing generates a trace file with unordered timestamped events,
2912 because of the way the core simulator (surf) works. A script available
2913 at the tools directory (fix-paje-trace.sh) can be used to put the events
2914 in order. We have changed the tracing so it can generate ordered timestamped
2915 events in the final trace, but depending on the simulator (and how much time
2916 is simulated) that can lead to a huge memory utilization. It is deactivated
2917 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2919 Build Infrastructure
2920 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2921 - We are trying hard to keep the API stable, but it may happen that
2922 some things change (we're a research project after all, not a
2923 nuclear plant operating system). If such things should happen, you
2924 could rely on that macro to adapt.
2925 - current value: 30600 for 3.06.00, aka 3.6
2926 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2927 a number that can be compared with SIMGRID_VERSION.
2928 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2929 assertions and verbose/debug logging events are disabled at compile time.
2931 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2933 SimGrid (3.5) stable; urgency=medium
2936 * New feature to any SimGrid-based simulator: Model-Checking
2937 Check SIN#1 for more details.
2940 * New Model SMPI (three-interval linear regression for correction factors)
2941 See RR-7426, available at http://hal.inria.fr/inria-00527150
2942 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2943 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2944 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2945 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2946 * See also src/smpi/README
2951 - Tracing API changes: TRACE_start and TRACE_end should not be called
2952 by user-code. They are automatically called by simulators created
2953 with SimDAG, MSG and SMPI if the toolkit is compiled with
2954 tracing_enabled=ON. Categories declaration and utilization remain the
2955 same for MSG and SimDag.
2956 - A function was added to the tracing API to declare categories with
2958 - TRACE_category_with_color (char *category, char *color)
2959 where color must be in the following format
2960 "%f %f %f", red, green, blue
2961 and red, green, blue are float values in the interval [0, 1]
2962 - User can specify NULL as color parameter, or continue calling
2963 TRACE_category (cat)
2964 On that case, the tracing system will define random colors
2965 - The following command-line options are supported:
2966 --cfg=tracing/filename:msg.trace
2967 --cfg=tracing:1 (activate tracing, needed to use others)
2968 --cfg=tracing/platform:1 (categorized resource use)
2969 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2970 --cfg=tracing/msg/task:1 (task creation)
2971 --cfg=tracing/msg/process:1 (process creation, migration)
2972 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2973 --cfg=tracing/smpi:1 (SMPI interface tracing)
2974 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2975 - examples of examples/msg/tracing updated
2977 - DAXLoader and DOTLoader functions can generate tasks with categories
2978 - A new function to attribute a category to SD tasks:
2979 TRACE_sd_set_task_category (SD_task_t task, char *category)
2980 * Tracing the MPI interface implemented by SMPI
2981 - Collective operations are traced with states
2982 - Point-to-Point operations are traced with states/links
2983 - Tracing activated by a parameter "-trace filename" passed
2984 to smpirun during execution (considering that simgrid
2985 is compiled with tracing enabled)
2986 - To run the simulation with gdb, the simulator
2987 accepts --cfg=tracing/smpi:1 to trace SMPI
2988 - tesh files to check if smpi tracing is ok
2989 - See examples/smpi/NAS/DT-trace
2990 * GTNetS tracing re-worked
2991 - adaptation to the tracing system of GTNets to cope
2992 with modifications regarding the fullduplex mode
2993 - new tesh files to check if gtnets tracing is ok
2996 * Asynchronous communications through the functions:
2997 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2998 * New function: MSG_load_platform_script()
2999 to make possible using a lua script instead of XML files to set up platforms
3000 * New function: MSG_set_function
3001 to associate functions to processes, used when bypassing the parser
3002 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
3004 Platforms: Add some more examples in examples/platforms
3005 * Grid'5000: see www.grid5000.fr
3006 * *_30000_hosts.xml: various huge files [mainly scalability testing]
3009 * Change the XML format. This is a very important modification. SimGrid
3010 3.5 introduces a new hierarchical format based on the notion of
3011 Autonomous Systems. Compatibility with old format is ensured
3012 through the perl script provided in the install bin directory
3013 bin/simgrid_update_xml.
3014 It is now possible to build platforms with specific routing
3015 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
3016 connect several platforms together. We will try to provide soon
3017 set of realistic platforms exploiting these properties (have a
3018 look at examples/platforms/ for the moment).
3019 * Take the opportunity of the XML format change to be a good XML citizen:
3020 rename link:ctn to link_ctn and similar changes (also dealt with by
3022 * Add a new routing scheme (rule-based) using regular expressions. It
3023 enables to have an extremely low memory footprint when the
3024 underlying routing is simple and can be compactly described. You
3025 need to have libpcre4-dev (perl regular expressions) installed if
3026 you want to use this routing scheme.
3027 * Revive the cluster TAG and allow to easily and efficiently (both in
3028 term of memory and speed) connect clusters together. Have a look
3029 at teshsuite/simdag/platforms/ to see how this can be done. With
3030 this tag, you can create clusters with thousands of tasks at no
3031 cost (have a look at examples/platforms/).
3032 Note: clusters are implemented as ASes, so there is no need for an
3033 enclosing AS tag if you have only one cluster in your platform.
3034 * Add new generic functions in the public interface that allows the user
3035 to call SURF 'create_resource' methods from your code (same
3036 functionality as the XML bypass mechanism but with a much lighter
3038 * Add a new model (enabled through command line --cfg=network/model:SMPI)
3039 that uses a piecewise linear approximation to produce better
3040 results when exchanging small messages.
3041 * Add a new parameter to handle correctly full duplex link and account
3042 for interferences between uplink and downlink communications
3043 (activate with --cfg=fullduplex:1).
3046 * Rename the SD_READY (all dependencies are satisfied and task is
3047 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
3048 dependencies are satisfied) state.
3049 This prevents a confusion between the notion of "ready to schedule"
3050 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
3051 simulated" (SD_RUNNABLE) used by the simulation kernel.
3052 * Change the way a task is considered as ready. Instead of removing
3053 dependencies when a task is done, a counter is decreased. This way,
3054 it is always possible to reach ancestors thanks to the
3055 SD_taks_get_parents function (even after the end of the simulation.)
3056 * Change the return type of SD_Simulate from (SD_task_t*) into
3057 xbt_dynar_t. This function was in handling a dynar internally and
3058 converted it into a NULL terminated array for historical reasons.
3059 * New function SD_dotload(char*) to load a DAG described in dot
3060 format. This loader and the corresponding examples require the
3061 installation of the graphviz library.
3062 * Fix a bug in the management of tasks of size 0 in the surf network
3063 models. This problem was only visible with SIMDAG and you should
3064 thus disregard results produced with earlier versions if you
3065 relied on this feature (some tasks were blocked because of this).
3066 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
3067 even though your applications were doing only point-to-point
3068 communications and sequential computations. Now you can really use any
3069 model you want (of course, if you create real parallel tasks, which are
3070 not implemented in most models beside ptaskL07, this will abort).
3071 * Add an example that schedules a DAX on an heterogeneous platform
3072 using a Min-Min strategy.
3073 * New function SD_workstation_get_current_task() that returns the kind
3074 of task currently running on a workstation in the sequential access
3076 * Raise some warnings when unexecuted tasks remains at the end of the
3077 simulation. This is usually caused by cycles in the DAG.
3080 * New function: SIMIX_process_set_function() called by MSG_set_function
3081 * Change the underlying waiting queue in semaphores so that a process
3082 can wait on several of them simultaneously (as in waitany).
3083 * Fix the way to handle tokens in semaphores so that all access patterns
3084 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
3085 * kill the dirty pimple SIMIX_message_sizes_output()
3086 Please use (proper) visualization instead
3089 * New data container: setset (set of sets of elements)
3090 * New module: mmalloc (mapped malloc, allowing to have several
3091 independent segments of malloc)
3092 * New function: xbt_dict_cursor_set_data()
3093 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
3094 * New function: xbt_dynar_is_empty()
3095 * New function: xbt_fifo_get_last_item()
3096 * Fix xbt_dynar_shrink(): use the right element size.
3097 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
3098 uninitialized areas during expand.
3099 * Fix semaphores: previous implementation was severely broken.
3100 * Use library init/fini functions for our initialization.
3101 - you can use logs and other feature as soon as you want in your
3102 code (even before the xbt_init / MSG_init)
3103 - xbt_exit is now a no-op and produce a warning when used.
3106 * Port GRAS to new SIMIX mechanisms. This allows gras users to
3107 benefit from the latest improvement to the simulation kernel.
3108 * Kill measurement sockets for now. If you rely on them, sorry. This
3109 release is not for you. This feature will be reintroduced in the
3110 future, but we cannot delay the release any further.
3111 * New function: gras_msgtype_get_name().
3112 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
3113 * Fix (at last) the pmm example: it should not randomly fail anymore.
3115 Build chain: bug fixes and overall polishing
3116 * Cmake is now stable enough. Hence, we killed the autotools.
3117 * Port to windows ( TM :)
3118 * Fix the 'make install' target.
3119 No need to use 'make install-simgrid' anymore
3120 * Introduce a 'make dist' target compiling a *source* archive
3121 'make package' compiles a binary archive
3122 * Compile java files only on need
3123 * Add --cd and --setenv command line options to tesh
3124 * Out of source builds are not fully supported yet, but we are close
3125 * Enable supernovae and optimization flags by default for our users
3128 * Add layer to set up environment directly from lua, without XML.
3129 * The effect of gras_stub_generator can be achieved through
3130 lua too (check examples/gras/console/ping_generator.lua)
3132 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
3134 SimGrid (3.4.1) stable; urgency=low
3136 The "Polishing easter eggs is probably a good idea" release.
3137 This is a bug fixes release only.
3141 * Fix a bug preventing the tasks from begin garbage collected.
3144 * Fix a bug occurring when a host involved in a communication fails.
3145 This was not detected properly by the other peer involved in the
3146 communication. Now, it's reported as a network error.
3149 * Warn the user about loop dependencies in data flow of DAX files
3150 * Obey the control-flow dependencies of DAX files
3153 * Add option "enable_smpi" allowing to not compile SMPI.
3154 Probably useful for the (Mac) users experiencing a build error here
3155 * Improve the detection of lua5.1 and ruby1.8
3157 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
3159 SimGrid (3.4) stable; urgency=low
3161 The "Easter in Cargese" release. Also known as (major changes):
3163 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
3164 ~> bindings were greatly improved
3165 ~> new tracing infrastructure for better visualization introduced
3167 * the "Welcome to configury modernity" release.
3168 ~> we switched from autotools to cmake, and improved our cdash
3171 A more detailled list of changes follow (full detail in svn log).
3173 Java Bindings: Various Cleanups
3174 * (install java-gcj-compat-dev on debian-like to use them)
3175 * Remove put/get: no need to export deprecated interface in Java
3176 Use send/receive instead.
3177 * Cleanup the examples and add a README per directory
3178 * Remove example autoDestination (that's the only way to go now)
3179 * Remove example explicitDestination (was a plain copy of basic)
3180 * Make JniException a runtime exception, so that there is no need to
3181 declare the fact that you may encounter such a beast. I guess that
3182 nobody will ever want to survive such error.
3183 * Create specific errors for each MSG case of failure:
3184 host failure, transfer failure, timeout, task cancelled
3185 * Cleanup the exceptions that may get thrown by each function
3186 * Other internal cleanups in Java bindings. Performance still bad :/
3187 Ruby and Lua Bindings: create them
3188 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
3189 * That's new and great, you should try them out.
3190 Same functionalities than Java bindings, only even less polished
3192 * Kill the useless "rate" argument of SD_task_get_execution_time()
3193 Everyone used to provide -1 as a value, it was not used, and the
3194 semantic of a possible use wasn't even clear.
3195 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
3196 either as comm costs or compute costs to mean that there is no
3197 such thing for that specific task.
3198 * Add a SD_task_set_name() function
3199 * Fix SD_task_unschedule() on typed tasks
3200 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
3201 * In DAX loader, accept useless 'level' attributes to <job> since
3202 LIGO DAGs have them (seem to be to ease graphical representation).
3204 * Add an example masterslave_mailbox.c using send/receive and not
3205 the deprecated put/get interface.
3206 * Kill the MSG_paje_output() function. It's a noop since 2 years.
3207 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
3209 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
3210 (declare MSG_USE_DEPRECATED to still have the old name)
3211 * Add a MSG_task_set_data() function
3212 * About trace replay (see examples/msg/actions):
3214 - Allow to work with splitted trace files for each process
3215 Give the specific trace file as argument of each process,
3216 and call MSG_action_trace_run(NULL)
3217 You can still have one merged file for all processes.
3218 - Fix implementation of collective operations
3219 * Allow task_execute() on 0-sized tasks (closes #10063)
3221 * This is the first release of SimGrid where SMPI is not considered
3222 beta anymore (even if some corners should still be improved)
3223 * Port over the new SIMIX_network submodule (internal refactoring)
3224 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
3225 * Implement more missing elements of the standard:
3227 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
3229 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
3230 MPI_Scan() MPI_Get_processor_name()
3231 - Added implementation of missing case for Alltoall (warning: it's
3232 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
3233 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
3234 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
3235 SMPI_MPI_Allgatherv()
3236 * Bug fixes include:
3237 - MPI_Waitsome() was broken
3238 - Allow relative includes in smpicc
3239 - Command line cfg argument 'reference_speed' was ignored...
3240 - Some functions did not properly lead to auto-benching of user code
3241 - smpicc passes -O2 by default (just like openmpi one)
3243 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
3244 * Bug fixes about timeouts during communications
3245 * add SIMIX_message_sizes_output() as a pimple to write to file the
3246 amount of messages per size. Use gnuplot to get histogram.
3247 Pimple because that's the only user-visible function of simix,
3248 defined directly in xbt.h (irk, sorry)
3250 - Add a SIMIX_sem_get_capacity() function
3251 - Fix interactions with processe resume/suspende
3252 - release_forever() was stupidly broken
3253 - Fix SIMIX_display_process_status() for processes in a semaphore
3254 - Make SIMIX_sem_block_onto() user-visible
3255 * Refactoring context stuff:
3256 - Use pseudo-OOP for better modularity
3257 - reimplement SIMIX_process_kill() without process_schedule() so
3258 that the latter can take as invariant that it is called from
3260 - Merge context_start into context_new for sake of simplicity
3262 * Add a Vivaldi network model, coded live during SUD'10 ;)
3263 * Rename configuration variables to start a hierarchy:
3264 o cpu_model -> cpu/model
3265 o network_model -> network/model
3266 o workstation_model -> workstation/model
3267 * New configuration variables:
3268 o network/bandwidth_factor: correction to bandwidth
3269 o network/latency_factor: correction to latency
3270 o netwotk/weight_S: correction to the weight of competing streams
3271 * Add a long description to the models, that users can see with such
3272 argument on the command line: --cfg=cpu/model:help
3273 * --help-models display the long description of all known models
3275 * config: add the ability to set a default value after registration
3276 Does not override any previously set value (e.g. from cmd line)
3277 * dict: allow to have integer key and data.
3278 When so, you need to use the following functions
3279 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
3280 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
3281 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
3282 In contrary to regular dicts, the key is not malloced before copy.
3283 Mixing scalar and regular elements in the same dict is not tested
3285 * Allow to use xbt_dynar_shrink() to expend the dynar instead
3286 Tracing for Visualization:
3287 * SimGrid is now instrumented in order to generate a trace file for
3288 visualization analysis: to use it, need to compile SimGrid with the
3289 "tracing" option enabled, and instrument the program using SimGrid with
3290 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
3291 (among other functions).
3292 * The instrumentation only traces the platform utilization for now
3293 * Documentation to use the tracing functions and how to analyze the
3294 traces with the Triva tool is written.
3295 * More information about: SimGrid FAQ (in the section Tracing Simulations
3298 * We moved to cmake as default build system. Autotools support will
3299 be dropped soon. Check the FAQ for more info about how to use it.
3300 * Greatly improved our cdash/ctest interactions
3301 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
3302 * Added memory checking tests with valgrind; lot of memleak fixing.
3303 This may be the first release of simgrid with so few memory issues
3304 * Added code coverage tests.
3305 Our coverage is still improvable, but at least we see it on cdash.
3307 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
3309 SimGrid (3.3.4) stable; urgency=low
3311 The "Desktop Grid needs love too" release (also called Xmas release).
3313 Models improvements:
3314 * Major speedup in the maxmin system solving by using lazy evaluation
3315 Instead of solving completely the maxmin system at each iteration,
3316 only invalidate (and recompute) the modified parts.
3317 This new feature is enabled in default models but you can try to
3318 turn it on with "--cfg:maxmin-selective-update=1" for other models.
3319 * Cas01 IMproved as default CPU model
3320 This CPU model is the same Cas01 model, but it uses the
3321 maxmin-selective-update flag and a heap structure to manage
3322 actions on SURF kernel.
3323 It reduces the complexity to find the next action to finish and,
3324 consequently, it's faster than the old Cas01.
3325 This is the new default CPU model (Cas01).
3326 * Rename the old Cas01 model to Cas01_fullupdate
3327 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
3328 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
3329 * CpuTI (CPU Trace Integration)
3330 A new CPU model whose objective is simulate faster when using
3331 availability trace files.
3332 Instead of using a full featured, over engineered maxmin system for
3333 CPU modeling, this model does the pre-integration of traces files
3334 to calculate the amount of CPU power available, and so, executes
3335 faster than the old CPU models.
3336 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
3337 * Use LV08 as default network model since it gives better accuracy
3338 for small messages and shouldn't change things for big ones.
3339 Use --cfg=network_model:CM02 to get the previous behavior.
3342 ******************************************
3343 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
3344 ******************************************
3345 * The new CPU model may changes simulations!
3346 The point is that events occurring at the exact same timestamp
3347 are not scheduled in the same order with the old and new
3348 version. This may be enough to completely change the execution
3349 of simulations in some cases.
3350 * The new network model will change simulations!
3351 This new model is more realistic than the previous one, so you
3352 should consider redoing your old experiments with this model.
3353 Sorry for the inconvenience.
3356 * Introduce the supernovae compilation mode
3357 When compiled that way, the whole SimGrid (or almost) is put in a
3358 single compilation unit and compiled in one shoot.
3359 This is to help gcc which has difficulties to inline stuff from one
3361 The speedup seem to be above 15%, althrough more tests are needed on
3362 amd64 to confirm that gain.
3365 * Port of MSG's mailbox on top of SIMIX network
3366 The put/get mechanism was greatly simplified on the way.
3369 * New SIMIX network module. Provides:
3370 - Mailbox: rendez-vous mechanism to find with who you want to speak
3371 - Synchronous send/recv: easier and hopefully faster since the
3372 logic is handled in the maestro process directly now
3373 - Asynchronous send/recv: you dreamt of it? It's here now
3374 Too bad that nobody cared enough to propagate the change to MSG.
3375 * Add semaphores as SIMIX synchronization mechanism.
3378 * new function SD_daxload(char*) to load a DAX file
3379 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3380 * Introduce typed tasks. Specify its kind and cost at creation.
3381 At scheduling, just give where it should be placed, and the cost
3382 for each involved resource is automatically computed.
3383 Existing constructors so far (more to come of course):
3384 - SD_task_create_comm_e2e() for end-to-end communication
3385 - SD_task_create_comp_seq() for sequential computation
3386 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3387 * new function SD_task_dump() for debugging display
3388 * new function SD_task_dotty(task,FILE*) writing to file the info
3389 about the task in dotty format
3390 * SD_task_dependency_exists() can now cope with having one of its
3391 arguments NULL. If so, it tests whether the other argument has any
3393 * Add getters on list of preceding/following tasks:
3394 SD_task_get_parents(task) and SD_task_get_children(task)
3395 * Add getters on amount of workstations and list:
3396 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3397 * Add getter on task kind: SD_task_get_kind(task)
3398 * Update the start_time and finish_time of tasks on completion/failure
3399 * Bugfix: Remove task from state swags when destroyed
3402 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3405 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3406 dynar with one separate thread per value of the dynar.
3407 * Change the prototype of xbt_thread_create(), sorry.
3408 Added a boolean parameter indicating whether we want to join this
3409 thread (used in SG only for now)
3410 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3413 * GTNetS wrappers should now be usable again (and betterly tested too)
3414 * Fix a major regression from 3.2 where the timeout provided to
3415 MSG_task_put_with_timeout() was used as absolute time before which
3416 the comm should be done.
3417 * Start to fix the <cluster> tag.
3418 - Internal links should be good now (beside of the loopback, which
3419 use the private link instead)
3420 - paths to the external world is still rather broken
3421 - the <route:multi> tag is just broken. Actually that's brain-dead.
3422 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3423 to make it less stupid
3424 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3425 * Fix a source-level compatibility glitch from 3.2: after defining
3426 MSG_USE_DEPRECATED, you can use the old name
3427 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3428 * Allow to compile from the SVN with automake 1.11
3429 * Fix some problems when using the "start_time" tag in deployment XMLs.
3430 * Fix #8569: XBT/synchro.h has redundant declarations
3431 * Fix #8563: MSG return values and exceptions
3432 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3433 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3434 Thanks to Robson Peixoto.
3435 * Fix a vicious bug in dictionaries inducing that some elements were
3436 not freed on xbt_dict_free()
3438 Portability report of this version:
3439 * Main portability targets:
3440 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3441 on (amd64/i386/ia64)
3442 - mac leopard on i386
3443 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3444 but nothing critical.
3445 * Other platforms: windows, AIX and others were not tested for this release
3447 Timing report of this version:
3448 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3449 scenario parameters). From 8h to a few seconds in desktop grid settings.
3450 * Supernovae brings about 25% speedup on i386.
3452 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3454 SimGrid (3.3.3) stable; urgency=low
3456 The "Need for Speed" release.
3458 The timings done to validate the 3.3.2 were faulty.
3459 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3461 The problem was a conversion from a manually handled vector to
3462 xbt_dynar_t on the critical path.
3463 xbt_dynar_foreach calls functions, inducing stack management crap.
3465 We inlined these functions and xbt_dynar_foreach is now breath taking.
3466 We also inlined xbt_swag_belong on the way.
3468 Here are some approximate speedup measurements (on master/slaves
3469 simulations lasting between 10s and 20s each):
3470 3.3.1 -> 3.3.2: about same performance
3471 3.3.2 -> 3.3.3: 40% speedup
3472 3.3.1 -> 3.3.3: 40% speedup
3473 3.3.1 with inline patch -> 3.3.3: 30% speedup
3475 Our reading is that the refactoring which occurred in 3.3.2 made us
3476 suffer much more from the xbt_dynar_foreach low performance, but
3477 once we solved this, this refactoring proved to be very performance
3478 effective. From the 40% speedup, somehow, 10% are due to the
3479 inlining and 30% to the refactoring.
3481 That's a pitty that gcc cannot inline functions placed in other files
3482 alone. We have to choose between:
3483 - break the encapsulation (by putting private data structures and
3484 accessors in headers files to help gcc)
3485 - live with low performance
3486 - switch to a decent compiler such as icc (not quite possible).
3488 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3490 SimGrid (3.3.2) stable; urgency=low
3492 The "Simplicity does not preceed complexity, but follows it" release.
3494 The main contributors of this release were (lexical order):
3495 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3498 * Extract the routing logic into its own object.
3499 (was dupplicated in network.c and workstation_LV07.c;
3500 Allows to implement other ways of storing that info)
3501 => kill now useless network_card concept
3502 - Use dynar to represent routes (instead of void** + int*)
3503 - kill link_set (use surf_network_model->resource_set instead)
3504 - Add a command-line option to choose the routing schema to use
3505 - Add three new models:
3506 * Floyd (shortest path computed at initialization)
3507 * Dijikstra (shortest path recomputed all the time)
3508 * Cached Dijikstra (shortest path computed on need)
3509 All these models where contributed by Silas De Munck, and are
3510 described in his ICCS09 paper.
3512 * Simplify model declaration
3513 (less redirections, less function to write when defining a model)
3514 - Factorize stuff between models:
3517 surf_model_resource_set(model)
3518 surf_model_resource_by_name(model, name)
3519 - Unify the types of models in s_surf_model_t (using an union)
3520 - Embeed fields of common_public directly into s_surf_model_t
3521 - Rename model methods:
3522 action_free ~> action_unref
3523 action_change_state ~> action_state_set
3524 action_get_state ~> action_state_get
3525 - Change model methods into functions:
3526 (model)->common_public->action_use ~> surf_action_ref
3528 * Implement a generic resource; use it as ancestor to specific ones
3529 (allows to kill duplicated code in models)
3530 Drawback: timer command don't need no name nor properties;
3531 workstation_CLM03 don't need no properties
3532 (but I guess we can live with those few bytes wasted)
3534 * Improve the action object model
3535 - implement a constructor avoiding dupplicated code about field
3536 initialization in generic_action part.
3538 * Kill the SDP model: it has an external dependency, is deprecated
3539 in flavor of modern lmm models, and didn't compile since a while
3542 * Relocation of the context module from XBT to SIMIX.
3543 (the context were decoupled from the simix processes, duplicating a lot of code)
3544 => a lot of code was factorized
3545 - less overhead is introduced during scheduling
3546 - simpler API for the context factory
3547 - the logic for process creation,destruction and manipulation was simplified
3548 * Simplification of the s_smx_process_t data structure.
3549 => accesing the simix level data associated to a process is faster now,
3550 and the code is a lot more readable.
3553 * Implement some more MPI primitives:
3554 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3555 -implementation: Bcast: flat or 2-ary tree (default),
3556 Barrier: 4-ary tree,
3558 Allreduce: Reduce then Bcast
3559 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3560 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3561 Alltoallv: flat tree, like ompi
3563 * Add support for optimized collectives (Bcast is now binomial by default)
3564 * Port smpirun and smpicc to OS X
3567 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3568 Shout out if you used it.
3571 * Display the list of still queued messages in SG mode when existing
3575 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3576 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3577 * Add xbt_str_from_file(FILE*)
3578 * Add xbt_dict_get_key achieving a linear reverse search
3579 * Remove the context module
3581 Portability report of this version:
3582 * Main portability targets:
3583 - Linux(debian)/x86/context
3584 - Linux(debian)/x86/pthreads
3585 - Linux(debian)/amd64/context
3586 - Linux(debian)/amd64/pthreads
3587 On these, we still have the eratic breakages of gras/pmm and
3588 amok/saturate_sg reported in previous version. We still think
3589 that the tests are the cause of the fault, not the tested code.
3591 - Mac OSX Leopard/x86/context
3592 Still false negative in tesh autotesting.
3593 Smpi still fails, but this time because readlink does not accept -f
3594 Everything seems to work properly beside of that.
3597 - AIX version 5.3 (only tested contexts this time)
3598 Smpi still fails there because mktemp is not installed.
3599 Everything seems to work properly beside of that.
3601 I managed to compile it for the first time, but several breakages.
3602 Won't delay the release for this exotic platform.
3604 * Windows: it's still lagging behind. If you want to help, please
3607 Timing report of this version:
3608 This version seem to be more than 5% faster than 3.3.1 (on linux
3609 64bits with contextes). The gain is less than expected, we are
3610 investigating this for next release.
3612 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3614 SimGrid (3.3.1) stable; urgency=low
3617 * Implement a --cfg-help to show existing configuration variables
3618 * Build chain do not require doxygen in maintainer mode
3621 * fix a bug on struct sizeof computation, which prevented the
3622 exchange of arrays of structs in some conditions
3623 - added a regression test about this in datadesc_usage
3624 * Allow the exchange of 0-long dynamic vectors.
3625 - for that, use -1 as indicator of dynamic size instead of 0
3626 - This implied to change any size from unsigned long to long,
3627 reducing a bit communication abilities, but I guess that with
3628 64bits being quite common, this is more than enough.
3629 - This also induce a protocol change, thus bumping network protocol
3630 version from 0 to 1 (if we have external users, we have to get
3631 clean on that point too ;)
3632 - added two regression tests about this in datadesc_usage
3633 * Be more verbose when propagating local exceptions
3634 This helps debugging.
3635 * Display the status of simulated processes when receiving SIGINT in
3639 * Allow to control the simulation from a trace file.
3640 New functions MSG_action_register() and MSG_action_trace_run()
3641 The first one allows to associate a function execution to each
3642 kind of action while the second one parses a trace file and
3643 triggers the corresponding actions within the system.
3644 For now, only a toy example is provided in examples/msg/actions
3645 * Add an example of process migration in examples/msg/migration
3646 * Fix a bug in task exchange which broke MSG_task_get_sender()
3647 Add a teshsuite regression test for that.
3648 [Bug: if MSG_task_get_sender() is called after sender exit,
3650 * Fix a bug which prevented suspend/resume to work properly
3651 * Display the status of simulated processes when receiving SIGINT
3652 This fixes a regression of v3.3. due to the introduction of SIMIX
3653 * Bug fixing in failure management:
3654 - trace could not start by a failure at time 0
3655 - failure during communications were not working
3658 * Add SIMIX_process_set_name() to change the name of the current
3659 process in the log messages.
3660 * Store smx_hosts in a dict since we only retrieve them by name
3661 * Move the configuration infrastructure to surf
3664 * Move the configuration infrastructure to surf
3667 * Massive internal cleanups:
3668 - Store internal structures on processes instead of hosts (allows
3669 to have more than one process per host, in addition of being more
3671 - Cleanup the initialization/finalization process
3672 - Kill a whole bunch of unneeded synchronization:
3673 processes run in exclusive manner within the simulator
3674 - Move queues from global tables to process data fields
3676 - now accept -platform and -hostfile arguments
3677 - Pass the right rank value to processes according to the hostfile
3678 * Compile the examples by default, and use them as regression tests
3679 * Implement MPI_Wtime()
3680 * Change the reference speed to a command line option
3683 * TCP_gamma can now be specified as command line option using
3684 --cfg=TCP_gamma:10000000.0
3685 * Change the --surf-path cmd line option into --cfg=path:
3688 * Also include strbuff from xbt.h public header
3689 * xbt_ex_display(): do not free the exception after displaying
3690 This allows to do more with the given exception afterward.
3691 Users should call xbt_ex_free() themselves.
3695 Portability report of this version:
3696 * Main portability targets:
3697 - Linux(debian)/x86/context
3698 - Linux(debian)/x86/pthreads
3699 - Linux(debian)/amd64/context
3700 - Linux(debian)/amd64/pthreads
3701 These targets fail about 1/10 of times on gras/pmm, but we believe
3702 that this is because of the test, not because of simgrid.
3703 amok/saturate_sg fails even more rarely, and the test may not be
3706 - Mac OSX Leopard/x86/context
3707 The test suite still spits tons of errors because some obscure
3708 force prevents us from removing the temporary directories
3709 arguing that they still contain some metadata I've never heard of.
3710 Smpi fails because seq is not installed.
3711 Everything seems to work properly beside of that.
3714 - AIX version 5.3 (both contexts and pthread)
3715 Smpi still fails there because mktemp is not installed.
3716 XML inclusions seems rosty on AIX.
3718 * Windows: it's still lagging behind. If you want to help, please
3721 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3723 SimGrid (3.3) stable; urgency=high
3727 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3728 [Malek Cherier & Mt]
3730 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3733 Until now, GRAS were using MSG as an interface to SURF. It was
3734 quite difficult because both interface have several differences
3735 (MSG channels vs GRAS sockets were the most notable point).
3737 This also opens the gate to SMPI (which should occur soon) and speed
3738 up simulations by to 40% (even if it were not the main goal).
3740 **************************************
3741 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3742 **************************************
3743 The point is that events occurring at the exact same timestamp are
3744 not scheduled in the same order with the old and new version. This
3745 may be enough to completely change the execution of simulations in
3746 some cases. Sorry for the inconvenience.
3748 * Cleanup and upgrade the XML format to push further scalability
3749 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3751 * Improve the testing infrastructure with tesh. Now a very large part of
3752 the code is tested not only by being run but also by checking that the
3753 output match an expected output [Mt].
3755 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3756 is really less memory-demanding, which should allow you to use
3757 larger files in SimGrid [AL].
3759 * Inform valgrind about our contextes, so that it becomes usable
3760 with the default (and more effecient) version of SimGrid
3761 [contributed by Sékou Diakite, many thanks]
3764 * Introduce a listener thread in charge of receiving incoming
3765 messages from the network. It allows to overlap communication and
3766 computation but most notably, it removes some stupid deadlocks due
3767 to the fact that so far, a process could not send and receive at
3768 the same time. This made most non trivial communication schema
3770 * Convert the PIDs from long int to int to match the MSG ones (and
3771 linux ones too) [Mt]
3772 * New function: gras_agent_spawn() to launch a new process on
3773 current host. Only working in simulation for now. [Mt]
3774 * New function: gras_os_hostport() returning a constant form (ie,
3775 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3778 * Make the backtrace of exceptions more human readable [Mt]
3779 * New module: xbt/str [Mt]
3780 a ton of string utility functions (split, join, printf to a newly
3781 allocated buffer, trim, etc)
3782 * New module: xbt/hash [Mt]
3783 SHA1 hashing algorithm (more to come if needed)
3784 * New module: xbt/synchro [Mt]
3785 synchronization tools (mutex and conditions) working the same way
3786 in simulation and in real life (mainly useful for GRAS, but not
3788 * New module: xbt/queue [Mt]
3789 classical producer/consumer synchronization scheme
3790 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3791 (using the classical functions will get serialized) [Mt]
3792 * Make dictionary internal table dynamic. No need to specify its size
3793 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3795 * Make sure the log channels are organized as a tree under windows
3796 (because of ANSI C compatibility issue, any channel were child of
3797 root directly) [Mt].
3800 * Cleaned many thing in surf and fixed a few bugs [AL].
3801 * Add a nice command line configuration mechanism to compose models [AL].
3802 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3803 the previous one (KCCFLN05). It relies on something that looks like
3804 a max-min sharing mechanism but cannot be written as such. A new solver
3805 was thus designed [AL].
3806 * Add a new solver to lmm. Based on Lagrange optimization and
3807 gradient-based descent, it enables to efficiently maximise systems s.a
3809 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3811 This solver enables to propose two new network models for TCP Reno and
3812 TCP Vegas based on Low's work. These models still need to be fully
3813 tested though [Pedro Velho].
3816 * Bug fix in SD_simulate. Now the time bound given as argument is
3818 * Use the new parallel task model (ptask_L07) as default.
3819 * Use the SURF command line configuration mechanism.
3820 * 0-size tasks (for synchronization) should now work.
3822 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3824 SimGrid (3.2) stable; urgency=high
3828 We still experience issues on this platform, but we believe that at
3829 least MSG is usable.
3831 GRAS API BREAKAGE (for simplification purpose, sorry):
3832 * the gras_msgtype_by_name is not used anymore. Instead of
3833 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3834 you can write (and must)
3835 gras_msg_send(toserver, "request", &request);
3836 - If you still want to pass a gras_msgtype_t to the function (to cache
3837 the type and avoid the lookup time), use the gras_msg_send_() variant.
3838 - Impacted functions:
3839 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3840 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3841 * The callbacks are now expected to return 0 when everything went well
3842 (just like the main() function)
3844 GRAS new features and improvements:
3845 * New module mechanism where user code can use per process globals [Mt]
3846 This is similar to gras_userdata_*() functions, but for libraries. It
3847 factorize some code developped over and over in the examples and AMOK.
3848 It has still to be documented and used (only amok/peermanagement is
3850 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3851 loss when they were small enough to fit into the buffer and sent quickly
3852 enough so that they can all get received in one shoot.
3853 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3854 if not found. Use the *_or_null() variant for the old semantic.
3855 * In gras_msg_handle, do not discard messages without callback.
3856 They are probably messages to be explicitly awaited later (ie, proofs of
3857 mis-synchronization in userland since they are sent before being awaited)
3858 No big deal usually.
3859 * gras_socket_meas_send/recv: semantic changed!
3860 The numerical arguments used to be (1) the total amount of data to send
3861 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3862 messages. This was need for the fool willing to send more than MAXINT
3863 bytes on quite fat pipes.
3866 * Do really rename the hostmanagement module to peermanagement. [Mt]
3867 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3868 but this is rather new and this was documented in the module
3869 documentation (poor excuses, I admit)
3870 * Bandwidth measurement semantic changed! This follows the changes to
3871 gras_socket_meas_send/recv explained above.
3874 * A sequential mode has been added to the workstations. When a workstation
3875 is in sequential mode, it can execute only one task, and the other tasks
3876 are waiting in a FIFO. [Christophe Thiery]
3879 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3880 model for SIMDAG. [Christophe Thiery]
3881 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3882 equal to 0, which caused some bad behaviors in
3883 saturated_constraint_set_update. I now use a threshold mechanism like in
3887 * When running manually src/testall, you select specific units [Mt]
3888 testall is the result of our cunit mechanism, and should replace all
3889 the scripty thingy around since bash don't run easily on billware.
3891 * A mallocator system has been added. [Christophe Thiery]
3892 Mallocators allow you to recycle your unused objects instead of freeing them
3893 and allocating new ones.
3895 Documentation update:
3896 * FAQ reworking + New FAQs:
3897 - "Valgrind spits tons of errors!" [Mt]
3898 - "How to repport bugs" [Mt]
3899 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3900 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3901 - Communication time measurement within MSG [AL]
3902 - I experience weird communication times when I change the latency [AL]
3903 * GRAS tutorial [Mt]
3905 - an introduction to the framework and to the used communication model
3906 - an initiatic tour introducing the most proheminent features:
3908 . Lesson 0: Installing GRAS
3909 . Lesson 1: Setting up your own project
3910 o Part 2: Message passing
3911 . Lesson 2: Exchanging simple messages
3912 . Lesson 3: Passing arguments to the processes (in SG)
3913 . Lesson 4: Attaching callbacks to messages
3914 . Lesson 5: Using globals in processes
3915 . Lesson 6: Logging information properly
3916 . Lesson 7: Using internal timers
3917 . Lesson 8: Handling errors through exceptions
3918 . Lesson 9: Exchanging simple data
3919 . Lesson 10: Remote Procedure Calling (RPC)
3920 . Lesson 11: Explicitly waiting for messages
3921 . Recapping of message passing features in GRAS
3922 - A HOWTO section containing:
3923 o HOWTO design a GRAS application
3924 More are due, of course. They will come latter. In the meanwhile, you can
3925 check the examples which are still here.
3927 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3929 SimGrid (3.1) stable; urgency=high
3933 There was a stack corruption somewhere, visible only when optimizing
3934 with these versions. [Vince]
3937 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3938 module that enabled to play with Directed Acyclic Graphs. It is built
3939 directly on top of SURF and provides an API rather close to the old
3940 SG. Some old codes using SG are currently under rewrite to check that
3941 all needful functions are provided. [Christophe Thiery]
3944 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3945 extension of the classical CLM03 model that gracefully handles
3946 failures. This is now the default model for MSG and GRAS. It doesn't
3947 handle parallel tasks yet though. [AL]
3948 * Bug fix: Weights were not correctly set in the network part.
3949 WARNING: This may have resulted in incorrect results with simulations
3950 where there are more than one flow on a given link. [AL]
3953 * After a (long ?) discussion on simgrid-devel, we have decided that the
3954 convention we had on units was stupid. That is why it has been decided
3955 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3956 WARNING: This means that all previous platform files will not work as
3957 such with this version! A warning is issued to ask users to update
3959 A conversion script can be found in the contrib module of the CVS, under
3960 the name contrib/platform_generation/surfxml_update.pl [MQ]
3963 * Bug fix: Processes were started in reverse order, wrt deployment file.
3964 WARNING: if your code relies on this bug, please fix it. [AL]
3965 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3966 being executed on two different locations. [AL]
3967 * Bug fix: Failures are now better supported thanks to Derrick's tests
3968 (there was many failure situations I hadn't thought of and that weren't
3969 correctly handled). [AL]
3970 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3974 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3975 Exception occurring on server-side are propagated back to client (!).
3977 API CHANGE: the callback changed their prototype. Change:
3978 int my_handler(gras_socket_t expeditor, void *payload_data) {
3980 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3981 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3983 * New! function: gras_msg_handleall to deal with all messages arriving
3984 within a given period.
3985 * New! function: gras_socket_server_range to get a server socket in a
3986 range of port numbers (ease to avoid port number conflicts) [MQ]
3987 * New! gras processes display their backtrace when they get a SIGUSR1
3988 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3989 Sweet to debug RL processes [MQ]
3993 - Do not force experiment sizes to be expressed in kb, or it becomes
3994 impossible to measure the latency this way (needs one byte-long tests)
3995 WARNING: this changes the amok_bw_* function semantic. [MQ]
3996 - Implements the link saturation stuff. [MQ]
3997 * Peer management module:
3998 New! module factorizing code that we wrote over and over [MQ].
4001 * New module: cunit (my jUnit implementation in ansi C) [MQ]
4002 - Test units are placed directly into the library code, they get extracted
4003 automatically and placed into the src/testall binary.
4004 - Convert most of the XBT tests to this system.
4005 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
4006 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
4007 live in the same C code. This required to change a little bit the API
4008 of surfxml parsing but shouldn't be an issue for end-users. [AL]
4009 * New module: sparse graph structure with basic algorithms (this is work
4010 in progress and the API is not considered to be frozen yet). [AL]
4011 * Display more information on backtraces: source line & function names are
4012 now displayed just like valgrind does (rely on addr2line tool) [MQ]
4013 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
4014 * Reworked a little bit some #include statements to load only required
4015 headers. Some user code that relied on SimGrid to include stdlib or
4016 stdio may need to include it by themselves. [AL]
4017 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
4019 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
4020 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
4021 but I guess that almost nobody used those parts. [MQ]
4023 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
4025 SimGrid (3.0.1) stable; urgency=low
4028 * Unfortunately, I had missed 5 misnamed functions:
4029 xbt_fifo_item_t xbt_fifo_newitem(void);
4030 void xbt_fifo_freeitem(xbt_fifo_item_t);
4031 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
4032 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
4033 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
4034 They're now deprecated. Please use their new versions:
4035 xbt_fifo_item_t xbt_fifo_new_item(void);
4036 void xbt_fifo_free_item(xbt_fifo_item_t);
4037 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
4038 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
4039 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
4041 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
4042 * Documentation: xbt_log module unmercifully reworked [MQ]
4043 * Bugfix: there was a problem with the ending of contexts with
4044 the pthread backend. It caused some weird deadlock or behavior
4045 depending on the pthread implementation. [AL]
4046 * Bugfix: get the exceptions raised in the simulator repport where
4047 and why they come from when they are not caught in time [AL, MQ]
4050 * Bugfix: Do repport the error when two non-connected hosts try to
4051 exchange data (Thanks to Flavien for stumbling into this one) [AL]
4054 * Add additionnal checkings on communications. Assert that two
4055 communicating hosts are connected by a set of links... [AL]
4058 * Add additionnal checkings on channel values in communication [AL]
4059 * New: MSG_task_get_source to see on which host a task was generated [HC]
4060 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
4061 the number of tasks waiting to be received on channel and sent
4063 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
4064 waits for the first task coming from a given host.. [AL]
4066 GRAS new functionnalities: [MQ]
4067 * Enhance the parsing macro to allow the size of multidimentional objects
4068 to be given thru annotations.
4069 * New example (and documentation): Matrix Multiplication a la RPC
4070 (as when I was young!) and fix a bunch of bugs found on the way.
4072 GRAS performance improvements: [MQ]
4074 * Reduce the amount of cbps creation/destruction by making it static to
4075 datadesc_send/recv() and using a (newly created) cbps_reset (based on
4078 * Change libdata to a set so that we can search for stuff by ID (and thus
4079 reduce the insane amount of dict lookups)
4082 * Actually implement gras_datadesc_copy() so that we don't have to mimick
4083 RL communication on top of SG since it's so uneffective.
4084 It may also be used for inter-thread communication in RL, one day.
4085 * Use gras_datadesc_copy() to exchange messages on top of SG
4087 - improve message exchange performance on top of SG
4088 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
4089 * Don't exchange on the network the size of the used part of buffer,
4090 instead, specify the possible buffer size to read().
4092 - reduces the amount of read/write calls (one pair per exchange)
4093 - reduces the amount of exchanged data (the size)
4094 - allows to retrieve all arrived data on receiver side, if we don't need
4095 it right now (subsequent read will peek the buffer)
4096 - allows the receiver to proceed with the beginning of the stream before
4097 everything is arrived
4098 - make it possible to build an iov transport (using readv/writev)
4100 - take care of the data with non-stable storage (like stacked data),
4102 * If possible, TCP send uses vector I/O (when writev() is here)
4103 - Don't use it for receive since we send data sizes and data on the
4104 same stream, so we wouldn't be able to chain large amount of chunks
4105 before having to flush the stuff to read the size.
4106 * Rework the transport plugin mechanism to simplify it and reduce the
4107 amount of pointer dereferencement when searching for the right function
4110 * I guess that now, we do almost as few system calls as possible while
4111 doing as few data copy as possible.
4113 To improve it further, we could try to send all the sizes first and then
4114 all the data (to use iov on receiving size), but it's only a partial
4115 solution: when you have 2 dimensional data, the sizes of the second
4116 dimension is data of the first dimension, so you need 3 streams.
4118 I'm not sure the potential performance gains justify the coding burden.
4120 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
4122 SimGrid (3.00) stable; urgency=high
4125 * New! Give the possibility to hijack the surf parser and thus bypass
4126 MSG_create_environment and MSG_launch_application. Have a look at
4127 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
4129 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
4131 SimGrid (2.96) unstable; urgency=low
4136 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
4137 This deprecates the xbt_error_t mechanisms.
4138 It modifies (simplifies) all XBT and GRAS API.
4139 MSG API keeps unchanged (exceptions raised by XBT are caught from
4140 within MSG and masked with existing error handling facilities)
4143 * New! Add a FATPIPE model. [AL]
4144 * New! Add a parallel task model. [AL]
4145 * New! Add automatically a loopback interface (in the default
4146 network model) if none was precised.
4149 * Bugfix: MSG_process_resume now works with the current running process.
4151 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
4152 * Modification of MSG_task_get_compute_duration. Once a task has been
4153 processed, the value returned by this function is now equal to 0. [AL]
4154 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
4155 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
4156 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
4157 of a m_task. For now, MSG_task_cancel only works with computation
4159 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
4160 of the processor (in Mflop/s) regardless of the current load on the
4162 * API Change: use proper naming convention for MSG_getClock and
4163 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
4165 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
4166 This function changes the priority of a computation task. This priority
4167 doesn't affect the transfer rate. A priority of 2 will make a task
4168 receive two times more cpu power than the other ones. This function
4169 has been added to suit the needs of Nguyen The Loc and hasn't been that
4170 much tested yet. So if it fails, please report it and send me your code.
4172 * API Change: removed all functions and types that were marked "deprecated"
4173 since many months. Renamed MSG_global_init_args to MSG_global_init.
4175 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
4177 SimGrid (2.95) unstable; urgency=low
4180 * Steal some nice code to GNU pth to fix context detection and usage [AL]
4181 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
4182 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
4185 * Bugfix: Allow absolute paths for platform description files [MQ]
4186 * Bugfix: do free the variables after use. Leads to drastic performance
4188 * Implement max_duration (ie, timeouts) on resources [AL]
4191 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
4193 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
4194 the needed support (timeouts) [AL]
4197 * Implement measurement sockets. You can now get the bandwidth between two
4198 hosts thanks to AMOK (see below). [MQ]
4199 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
4200 dynar over the network (yeah) [MQ]
4201 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
4202 * simple/static token-ring example. [Alexandre Colucci and MQ]
4203 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
4204 in place before [MQ]
4207 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
4208 type parsing. "short" and "long" modifiers were also ignored; other
4209 modifier (such as reference level) are still ignored. [MQ]
4210 * Embeed the buffer size within the buffer itself on SG. [MQ]
4211 That way, send() are atomic and cannot get intermixed anymore (at least
4212 the ones which are less than 100k; bigger messages still have the issue)
4213 * Array size pushed by the field, not by the field type (or each
4214 and every long int will push stuff to the cbps) [MQ]
4215 * use select() to sleep since it allows to portably sleep less than one
4218 GRAS (minor cleanups)
4219 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
4222 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
4223 * type category 'ignored' killed as it was never used and were difficult
4225 * whether a type can cycle or not is now a flag, leaving room for more
4226 flags (as "ignored", if we feel the need one day ;)
4227 * Rename raw sockets to measurement sockets since "raw" has another
4228 meaning in networking community.
4231 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
4232 and offers features not belonging to GRAS but that most applications
4233 need. One day, it may be a set of plugins loadable at runtime.
4234 * New module: bandwidth
4235 bandwidth measurement between arbitrary nodes running this module. [MQ]
4237 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
4239 SimGrid (2.94) unstable; urgency=low
4241 The first beta release of SimGrid 3 !
4245 * Update the main page and the FAQ. Adding references to gforge.
4248 * Add a gras_os_getpid function.
4251 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
4252 * Extend the logs so that they also print PID, hostname, date, ... if
4254 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
4255 and kill the old version which were in testsuite/
4256 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
4260 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
4261 for users. Instead, each programming environment (GRAS, MSG,...) use it
4262 when needed to provide such a feature to users.
4263 Don't shortcut the mechanism or you will also shortcut the virtualization
4264 you need on the simulator.
4268 * Cleanups in configury with regard to compile optimization/warning flags.
4269 Also add -fno-loop-optimize to any powerpc since it's the optimization
4270 killing gcc (< 3.4.0).
4271 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
4272 needed by MSG examples complications
4273 * Borrow configury beautifications from PHP
4276 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
4277 warning (thanks loris for stumbling into this one).
4278 * Bugfix: stop loading private headers (gras_config.h) from the public
4282 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
4283 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
4284 RL would imply a DNS resolver, which is *hard* to do in a portable way
4285 (and therefore delayed).
4286 * Implement a real timer mechanism and use it in timing macros. This allows
4287 to avoid rounding errors and get a 0.000005 sec precision in timing
4288 macros. While I was at it, various cleanups:
4289 - allow to declare more than one timed section per file (fix a stupid bug)
4290 - move some private declaration to the right place
4291 - merge conditional execution and timing macros into emulation module
4292 - document the module
4293 - make sure the module cleanups its mess on gras_exit
4294 * Documentation improvements:
4295 - (new) how to compile applications using GRAS
4296 - (new) emulation support (timing macros)
4298 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
4300 SimGrid (2.93) unstable; urgency=low
4302 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
4305 - Use Paje properly where used. Still to be sanitized properly.
4306 - Portability fix: Add an implementation of the contexts using pthread
4310 - Add xbt_procname(): returns the name of the current process.
4311 Use it to show the current process's name in all logging.
4313 - fix detection of older flex version and the reaction, since we do
4314 depend on modern ones (we use lex_destroy)
4315 - Better separation of SG and RL in the libs: remove all simulation code
4316 from libgras. As a result, this lib is now only 200k when stripped.
4317 Some of the xbt modules may also be duplicated (two sets and such) and
4318 should be cleaned/killed before SG3.
4319 - Insist on using xlC on AIX because of weird problems involving gcc there.
4320 - Cleanup the make remote stuff. This is now done by scripts
4321 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
4322 mainly for our private use, but we're working on changing them to user
4325 - Bugfix: flush the socket on close only if there is some *output*.
4326 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
4328 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
4331 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
4333 SimGrid (2.92) unstable; urgency=low
4335 Alpha 3 on the path to SimGrid 3
4339 - New! First try of benchmarking macros.
4340 - New! First try so that gras_stub_generator generate deployment and
4341 remote compilation helpers.
4343 - Bugfix: Initialization fix in msg_test.
4347 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
4349 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
4351 - API Change: gras_os_sleep to take the amount of seconds as a double.
4352 Accepting an int was error prone since it was the only location where
4353 seconds were coded as such. It leaded to damn rounding errors.
4354 - Bugfix: Hard to belive that timers ever worked before this.
4356 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4358 SimGrid (2.91) unstable; urgency=low
4360 Alpha 2 on the path to SimGrid 3
4364 - Bug fix in the lmm_solver.
4366 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4367 through the function MSG_paje_output.
4368 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4369 - It is possible to bound the rate of a communication in MSG with
4370 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4371 to write it in the changelog).
4372 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4377 - Various cleanups to the autotools stuff
4378 - Begin to move Gras examples to examples/gras/
4379 - Let make distcheck work again (yeah!)
4381 - documentation overhauled using doxygen.
4382 gtk-doc-tools is dead in SimGrid now.
4383 - Automatically extract all existing logging categories, and add the list
4384 to the documentation (long standing one, to say the less)
4386 - Cleanup the known architecture table. Reorder the entries to group what
4387 should be, and use a more consistent naming scheme.
4388 (some of the test dataset are still to be regenerated)
4389 - New! Allow library to register globals on each process just as userdata
4391 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4392 lookup time (for now).
4393 Use it in msg and trp.
4394 This cleans a lot the internals and helps enforcing privacy of the
4395 headers between the gras components.
4396 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4397 - Bugfix: gras_os_time was delirious in RL.
4398 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4400 - Reenable GRAS now that it works.
4402 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4404 SimGrid (2.90) unstable; urgency=low
4406 Alpha 1 on the path to SimGrid 3
4408 * It is a long time since the last release of SimGrid. I'm sorry about
4409 that but as I had told you, I was rewriting a lot of things. I apologize
4410 to those who had been reporting bugs to me and that I had not answered.
4411 If your bug is still in the new version, please tell me. Here is a
4412 summary of the main changes.
4414 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4415 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4416 portability, speed, and a lot more but you'll figure it out later.
4417 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4418 I wanted to release the new MSG as soon as possible and I have
4419 broken GRAS, which is the reason why, for now, only MSG is fully
4420 functional. A laconic description of these projects is available
4421 in the documentation.
4423 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4424 kernel optimized for our needs (called SURF but only the developers
4425 should use it). Hence, MSG is now roughly 30 times faster and I think
4426 that by rewriting a little bit MSG, I could event speed it up a little
4427 bit more. Beside the gain in speed, it is also much easier to encode a
4428 new platform model with SURF than it was with SG. More to come...
4430 * REVOLUTION 3: I have tried to change a little as possible the API of
4431 MSG but a few things really had to disappear. The main differences
4432 with the previous version are:
4433 1) no more m_links_t and the corresponding functions. Platforms are
4434 directly read from an XML description and cannot be hard-coded
4435 anymore. The same format is used for application deployment
4436 description. The new format is described in the documentation.
4437 Have a look in tools/platform_generation. There is a tiny script
4438 that converts from the old platform format to the new one. Concerning
4439 the application deployment format, parsing the old one is tricky.
4440 I think most of you should however be able to convert your files. If
4441 it is really an issue, I can write a C code that does the conversion.
4443 2) the toolbox tbx does not exist anymore. We now have a library
4444 with much more data-structures but without the hash-tables (we have
4445 dictionaries that are much faster).
4447 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4449 *****************************************************************************
4450 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4451 * don't feel like converting the oldies. (Mt) *
4452 *****************************************************************************
4455 Version 2.90: "the long awaited one"
4456 - Finished rewriting and debugging MSG. Rewrote the documentation.
4457 - disable GRAS for now since it needs to be ported to the newest SG
4460 - Finish the port to windows (using mingw32 for cross-compile)
4463 - Main loop and datastructures of SURF. A cpu resource object is
4464 functional. Surf can thus be used to create cpu's with variable
4465 performance on which you can execute some actions.
4467 2004-11-15 Martin Quinson
4468 - Port to ARM. Simply added the alignment and size descriptions. Should
4469 work, but the ARM machines are so slow that I didn't had the opportunity
4470 to 'make check' over there yet.
4472 2004-11-15 Arnaud Legrand
4473 - Trace manager now written. It uses a heap structure and is therefore
4474 expected to be efficient. It may however be speeded up (particularly
4475 when many events occur at the same date) by using red and black
4476 trees. One day maybe...
4477 - Max-min linear system solver written. It uses a sparse matrix
4478 structure taking advantage of its expected use. Most operations are
4479 O(1) and free/calloc are called as few as possible. The computation of
4480 the minimum could however be improved by using a red and black tree
4483 2004-11-03 Arnaud Legrand
4484 - Rename every gras_* function that was in xbt/ to its xbt_
4486 - Add a heap and a doubly-linked list to xbt
4487 - Added a dichotomy to the dictionaries. make check works as well before
4488 so I assume that the patch is correct. I do not know however if things
4489 run effectively faster than before now. :)
4491 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4492 SimGrid, and the version number is bumped to 2.x
4494 2004-10-29 Martin Quinson
4495 - Introduction of the remote errors.
4496 They are the result of a RMI/RPC on the remote machine.
4497 ErrCodes being scalar values, you can't get the host on which those
4498 errors did happen. Extending the error mechanism as in Gnome is possible.
4499 No idea yet whether it is a good idea.
4501 2004-10-28 Martin Quinson
4502 - Interface revolution: the Starred Structure Eradication.
4503 I used to do typedef struct {} toto_t; and then handle *toto_t.
4504 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4505 - ???_t is a valid type (builded with typedef)
4506 - s_toto_t is a structure (access to fields with .)
4507 - s_toto is a structure needing 'struct' keyword to be used
4508 - e_toto_t is an enum
4509 - toto_t is an 'object' (struct*)
4511 typedef struct s_toto {} s_toto_t, *toto_t;
4512 typedef enum {} e_toto_t;
4513 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4514 s_toto_t) is private.
4516 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4517 so that it can set the variable to NULL. It was so for dicts and sets,
4518 it changed for dynars.
4520 - Fix a bunch of memleaks in dict_remove
4521 - Fix a bug in sg/server_socket opening: it failed all the time.
4523 2004-10-07 Martin Quinson
4524 - Speed up dynar lookup operation a bit.
4526 gras_dynar_get is dead.
4528 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4529 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4530 (which gives you the address of the stored data).
4532 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4533 copy of the data using an affectation to do the job and not a memcpy.
4535 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4536 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4538 It does not really speedup the dynar test because they are
4539 setting elements all the time (and look them seldom). But the dict does
4540 far more lookup than setting.
4542 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4544 2004-10-05 Martin Quinson
4545 - Allow to (en/dis)able the cycle detection at run time.
4547 Whether we should check for cycle or not is now a property of each
4548 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4549 datadesc_cycle_unset allow to remove this property when previously set.
4551 Note that the cycle detection is off by default since it impacts the
4552 performance. Watch the data you feed GRAS with ;)
4554 This property is hereditary. Any element embedded in a structure having it
4555 set have it set for the time of this data exchange.
4557 You should set it both on sender and receiver side. If you don't set it on
4558 sender side, it will enter an endless loop. If you forget on receiver
4559 side, the cycles won't be recreated after communication.
4561 - Header reorganization.
4562 Kill gras_private.h, each submodule must load the headers it needs.
4564 2004-10-04 Martin Quinson
4565 - Interface revolution: do not try to survive to malloc failure.
4567 Now, gras_malloc and friends call gras_abort() on failure.
4568 As a conclusion, malloc_error is not a valid error anymore, and all
4569 functions for which it was the only gras_error_t return value are
4570 changed. They now return void, or there result directly.
4571 This simplify the API a lot.
4573 2004-09-29 Martin Quinson
4574 - Re-enable raw sockets.
4575 Created by gras_socket_{client,server}_ext;
4576 Used with gras_raw_{send,recv}
4579 It should allow to kill the last bits of gras first version soon.
4581 This is not completely satisfactory yet (duplicate code with
4582 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4585 - Simplify transport plugin (internal) interface by not passing any
4586 argument to _server and _client, but embedding them in the socket
4589 2004-09-28 Martin Quinson
4590 - Finish the port to AIX.
4591 autoconf was my problem (segfault within the malloc replacement
4592 function. No idea why)
4594 2004-09-16 Martin Quinson
4595 - Fix some size_t madness on 64bit architectures.
4597 2004-09-08 Martin Quinson
4598 - Reduce the number of system headers loaded, overload some more system
4599 calls (such as malloc to cast the result of the system one, and work
4601 - Fix and reintroduce the config support
4603 2004-09-07 Martin Quinson
4604 - Source code reorganization to allow Arnaud to surf all over there.
4605 - Allow to document the logging categories.
4606 - Remove all uppercase from logging categories and useless cleanup in names.
4608 2004-08-18 Martin Quinson
4609 Version 0.6.2 (protocol not changed; API changed)
4610 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4611 gras_error_t), and NULL when not found. Functions expecting a msgtype
4612 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4613 hopefully usefull message.
4614 - Portability to prehistoric sparcs again
4616 2004-08-17 Martin Quinson
4617 Version 0.6.1 (protocol not changed; ABI not changed)
4618 - prealloc some buffers to speed things up
4620 2004-08-11 Martin Quinson
4621 Version 0.6 (protocol not changed; ABI expended)
4622 - The parsing macro can deal with the references, provided that you add
4623 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4625 2004-08-09 Martin Quinson
4626 Version 0.5 (protocol not changed; ABI changed)
4627 - Allow to off turn the cycle detection code in data exchange at
4628 compilation time. It should be at run time, but I'm short of time (and
4629 the config stuff is still broken). That way, we keep dict out of the
4630 critical path, which is good because the performance is poor:
4631 - search not dichotomial yet
4632 - dynar give no way to access their content and memcpy everytime
4633 - In composed data description (struct, ref and so on), stop foolness of
4634 keeping the subtype's ID, but store the type itself. This keeps sets out
4635 of the critical path, which is good since they rely on dynar and
4636 dictionaries. The only loose of that is that we cannot detect the
4637 redeclaration of a structure/union with another content (but I'm not sure
4638 the code detected well this error before anyway). We still can detect
4639 the redefinition discrepancy for the other types.
4640 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4641 it breaks the code because of type-punning used all over the place).
4642 This breaks on all non-gcc architectures (for now).
4644 All those changes (plus the buffer of last time) allow me to gain 2 order
4645 of magnitude on cruel tests consisting of 800000 array of integers on two
4646 level of a hierarchical structure (200 secondes -> 4 secondes)
4649 - the selector of reference must now return the type it points to, not
4650 the ID of this type.
4652 2004-08-06 Martin Quinson
4653 Version 0.4 (protocol changed; ABI not changed)
4654 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4655 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4656 - Implement a buffer transport, and use it by default (it relies on tcp in
4657 real life and on sg in simulation).
4658 That's a bit hackish since I had a new field to the structure to store
4659 its data without interfering with the subtype ones. Inheritance
4660 is tricky in C. And that's a kind of reverse inheritance with one class
4661 derivating two classes. Or maybe a game with java interfaces. Anyway,
4662 that's damn hard in C (at least).
4663 Moreover, I got tired while trying to ensure plugin separation and
4664 genericity in SG mode. MSG wants me to do weird things, so let's go for
4665 cruel hacks (temporarily of course ;).
4666 See comment in transport_private.h:71
4667 - do not include all the _interface headers in private but in the files
4668 which really need them (to cut the compilation time when they are
4671 2004-07-26 Martin Quinson
4672 Version 0.3 (protocol not changed; ABI changed)
4673 - Major overhault of the datadesc interface to simplify it:
4674 - shorted the function names:
4675 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4676 - add a trivial way to push/pop integers into the cbps without malloc.
4677 This allows to make really generic sub_type description, which simply
4678 pop their size of the stack.
4679 - add a function gras_datadesc_ref_pop_arr() which does what users want
4680 most of the time: Declare a dynamic array (which pops its size of the
4681 stack) and declare a reference to it. Poor name, but anyway.
4682 - kill the post-send callback, add a post-receive one
4684 2004-07-23 Martin Quinson
4685 Version 0.2 (protocol changed; ABI changed)
4686 - add some testing for cpbs in the test cases, and fix some more bugs.
4687 This invalidate again the little64 data file, since I cannot regenerate
4689 - remove an awfull optimization in the logging stuff, allowing me to:
4690 - understand it again
4691 - learn gcc how to check that the argument match the provided format
4692 - fix all errors revealed by gcc after that
4693 - internal keys of dict are not \0 terminated. Deal with it properly in
4694 loggings instead of segfaulting when the user want to see the logs :-/
4696 2004-07-22 Martin Quinson
4697 - Fix some stupid bug preventing cbps (callback postit) from working
4699 2004-07-21 Martin Quinson
4700 - Some documentation cleanups
4701 - remove the useless last argument of msgtype_declare
4702 - rename the Virtu functions to fit into the 'os' namespace
4703 - move headers src/include -> src/include/gras/ and stop fooling with
4704 gras -> . symbolic link
4705 - make distcheck is now successful
4707 2004-07-19 Martin Quinson
4709 - Build shared library also
4710 - Install html doc to the right location
4711 - stop removing maintainer files in make clean
4712 - build tests only on make check
4714 2004-07-13 Martin Quinson
4716 - No major issue in previous version => change versioning schema
4717 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4718 corresponding dataset.
4720 2004-07-11 Martin Quinson
4722 - Get it working with any kind of structure (we can compute the padding
4723 bytes remotely for all the architectures I have access to)
4724 - Implement the structure parsing macro (still not quite robust/complete)
4725 - Improvement to the remote testing toysuite
4727 2004-07-10 Martin Quinson
4728 [autoconf mechanism]
4729 - get ride of a bunch of deprecated macros
4730 - actually run the test for two-compliment, not only compile it :-/
4731 - test whether the structures get packed (and bail out if yes. Damn.
4732 Alignment is a serious matter)
4733 - test whether the structures get compacted (but respecting the alignment
4734 constraints of each types)
4735 - test whether the array fields of structures can straddle alignment boundaries
4737 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4738 (took me 2 hours to find that bug, looking at the wrong place)
4739 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4740 before sending/receiving and are used to compute the offsets of fields
4741 - Given that padding size depend even on compiler options, keep track of
4742 alignment and aligned_size only for the current architecture. Not a big
4743 deal since we send structure fields one after the other (seems
4745 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4746 program, let it work on linux/gcc/little32. portability todo.
4748 2004-07-08 Martin Quinson
4749 - import and improve remote compilation support from FAST
4750 - make sure make check works on half a dozen of machines out there
4752 2004-07-07 Martin Quinson
4753 Let's say it's version 0.0.3 ;)
4754 - Implement conversions (yuhu!)
4755 - Let it work on solaris (beside conversion, of course)
4756 - Stupid me, using rand() to generate the conversion datatests in not wise.
4758 2004-07-06 Martin Quinson
4759 - Let make dist work, since I'm gonna need it to compile on remote hosts
4760 - Let Tests/datadesc_usage write the architecture on which the file was
4761 generated as first byte.
4762 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4763 architecture descriptions.
4764 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4765 on those architectures.
4766 - Optimization: send/recv array of scalar in one shoot
4768 2004-07-05 Martin Quinson
4769 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4771 - Plug a whole bunch of memleaks
4772 - each process now have to call gras_{init,exit}. One day, their log
4773 settings will be separated
4774 - Continue the code factorisation between SG, RL and common in Transport.
4776 2004-07-04 Martin Quinson
4778 - Redistribution between SG and RL.
4779 We wanna have to accept in SG, so move accepted related parts of RL in
4780 the common part. (more precisely, the dynar of all known sockets is no
4781 more a static in transport.c, but part of the process_data)
4783 [gras_stub_generator]
4784 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4786 2004-07-03 Martin Quinson
4787 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4788 them all up in one shot)
4790 - Fix the ugly case of reference to dynamic array.
4791 - New (semi-public) function gras_datadesc_size to allow the messaging
4792 layer to malloc the needed space for the buffer.
4794 - gras_socket_close now expect the socket to close (and not its address to
4795 put NULL in it after it). This is because the socket passed to handlers
4796 is one of their argument (=> not writable).
4798 - propagate the interface cleanup from last week in datadesc, ie remove a
4799 superfluous level of indirection. User pass adress of variable
4800 containing data (both when sending and receiving), and not of a variable
4801 being a pointer to the data. Let's say that I like it better ;)
4802 The price for that is constructs like "int msg=*(int*)payload" in
4803 handlers, but it's a fine price, IMHO.
4805 - Let it work in RL (yuhu)
4807 2004-06-21 Martin Quinson
4809 - porting SG plugin and SG select to new standards (works almost).
4810 - plug memleaks and fix bugs around.
4813 - cleanup the prototype of data recv and force users to specify when they
4814 want to handle references to objects. Test case working even for cycles.
4815 - plug memleaks. Valgrind is perfectly ok with this.
4817 2004-06-12 Martin Quinson
4819 - cleanup the separation between plugin and main code in plugin creation
4821 2004-06-11 Martin Quinson
4823 - Reput hook for raw sockets, needed for BW experiments
4824 - kill a few lines of dead code
4825 [Data description] Interface cleanup
4826 - gras_datadesc_by_name returns the searched type or NULL.
4827 That way, no variable is needed to use a type desc once, which makes
4829 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4830 two parameters were strings (field name, type name), leading to
4832 [Dicos] Interface cleanup
4833 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4834 This is consistant with the dynar API.
4836 2004-04-21 Martin Quinson
4838 - Porting to new standards.
4840 - interface cleanup.
4841 There is no bag anymore, no need to take extra provision to mask the
4842 pointers behind "ID".
4843 Better splitup of functions between files create/exchange/convert.
4844 This is still a bit artificial since convert and receive are so
4845 interleaved, but anyway.
4847 - add a queued message list to procdata (the ones not matching criteria
4849 - factorize some more code between SG and RL wrt procdata
4851 - use gras_exit in example to track memleaks
4852 - get rid of gs_example now that GS is properly integrated into gras
4853 - update run_test to integrate the latest tests (datadesc)
4855 - rename WARNINGn macros to WARNn since it prooved error-prone
4857 2004-04-19 Martin Quinson
4859 - register init/exit functions within gras module mechanism
4860 - send/receive function.
4861 Convertion is not implemented, but short-cutted if not needed.
4862 struct/array elements are sent one by one (instead of block-wise), but
4863 nobody really cares (yet). Get a prototype before optimizing.
4864 - tests (using a file socket) for DD send/receive on:
4865 - base types: int, float
4866 - array: fixed size, string (ie ref to dynamic string)
4867 - structure: homogeneous, heterogeneous
4868 - chained list, graph with cycle
4869 Believe it or not, valgrind is not too unhappy with the results. The
4870 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4874 - Bugfix when using a filename explicitly (instead of '-')
4876 2004-04-09 Martin Quinson
4878 - factorize more code between RL and SG in socket creation
4879 - Complete the implementation and tests of:
4881 o file (only in RL, and mainly for debugging)
4883 I lost 3 days to design a portable address resolver, and then decided
4884 that the prototype mainly have to run on my box.
4885 Addressing portability too early may be like optimizing too early :-/
4887 - use gras_init in the Tests instead of the crappy parse_log_opt
4888 (the latter function is removed)
4889 [Conditional execution]
4890 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4891 [Code reorganisation]
4892 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4893 Build examples against the RL library, since there is no way to disable
4894 its creation for now.
4896 For information, the beginning of coding on GRAS was back in june
4897 2003. I guess that every line has been rewritten at least twice since