1 SimGrid (3.30.1) NOT RELEASED YET (v3.31 expected March 20. 2022, 15:33 UTC)
4 - Rework the internals, for simpler and modern code. This shall unlock many future improvements.
5 - You can now define plugins onto the DFS explorer (previously called SafetyChecker), using the
6 declared signals. See CommunicationDeterminism for an example.
7 - Support mutex, semaphore and barrier in DPOR reduction
8 - Seems to work on Arm64 architectures too.
9 - Display a nice error message when ptrace is not usable.
10 - New test suite, imported from the MPI Bugs Initiative (MBI). Not all MBI generators are integrated yet.
11 - Remove the ISP test suite: it's not free software, and it's superseeded by MBI.
14 - fix for FG#100 by ensuring small asynchronous messages never overtake larger
15 ones, conforming to the standard.
16 - replay: fix waitall behaviour to avoid forgetting requests and leaking
18 - tracing: ensure that we dump the TI traces continuously during execution and
19 not just at the end, reducing memory cost and performance hit.
20 - Update OpenMPI collectives selection logic to match current one (4.1.2)
23 - New signal: Engine::on_simulation_start_cb()
24 - Introduce a new execution mode with this_actor::thread_execute(). This simulate
25 the execution of a certain amount of flops by multiple threads ran by a host. Each
26 thread executes the same number of flops, given as argument. An example of this new
27 function can be found in examples/cpp/exec-threads.
28 - Reimplementation of barriers natively.
29 Previously, they were implemented on top of s4u::Mutex and s4u::ConditionVariable.
30 The new version should be faster (and can be used in the model-checker).
31 - Actor::get_restart_count(): Returns the number of reboots that this actor did.
34 - MSG_barrier_destroy now expects a non-const msg_barrier parameter.
36 New plugin: the Chaos Monkey (killing actors at any time)
37 - Along with the new simgrid-monkey script, it tests whether your simulation
38 resists resource failures at any possible timestamp in your simulation.
39 - It is mostly intended to test the simgrid core in extreme conditions,
40 but some users may find it interesting too.
43 - New model for parallel task: ptask_BMF.
44 - More realistic sharing of heterogeneous resources compared to ptask_L07.
45 - Implement the BMF (Bottleneck max fairness) fairness.
46 - Improved resource sharing for parallel tasks with sub-flows (parallel
47 communications between same source and destination inside the ptask).
49 - "--cfg=host/model:ptask_BMF": enable the model.
50 - "--cfg=bmf/max-iterations: <N>" - maximum number of iterations performed
51 by BMF solver (default: 1000).
52 - "--cfg=bmf/selective-update:<true/false>" - enable/disable the
53 selective-update optimization. Only invalidates and recomputes modified
54 parts of inequations system. May speed up simulation if sparse resource
55 utilization (default: false).
56 - ATTENTION: this model requires Eigen3 library. If you install SimGrid
57 from source, please see the "Installing from source" section:
58 https://simgrid.org/doc/latest/Installing_SimGrid.html#installing-from-the-source.
59 No action is required if you use pre-compiled packages.
62 - Drop xbt_dynar_shrink().
65 - Added the following bindings: Comm.wait_for() and Comm.wait_any_for()
66 Example: examples/python/comm-waitfor/
68 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
69 (FG: issues on Framagit; GH: issues on GitHub)
70 - FG#57: Mc SimGrid should test whether ptrace is usable
71 - FG#87: Smpi scripts fail with spaces in paths
72 - FG#100: [SMPI] Order of the message matching is not guaranteed
73 - FG#101: LGPL 2.1 is deprecated license
74 - GH#151: Missing mutexes for DPOR.
76 ----------------------------------------------------------------------------
78 SimGrid (3.30) January 30. 2022.
80 The Sunday Bloody Sunday release.
82 Main user-visible changes:
83 - The SimDag API for the simulation of the scheduling of Directed Acyclic
84 Graphs has been dropped. It was marked as deprecated for a couple of years.
85 We finally complete the implementation of what has been called SimDag++
86 internally, i.e., porting the different features of SimDag on top of S4U.
87 The new way to simulate the execution of dependent activities directly by
88 maestro (without any other actor) is details in the examples/cpp/dag-* series
90 - The removal of SimDag led us to also remove the export to Jedule files that
91 was tightly coupled to SimDag. The instrumentation of DAG simulation is still
92 possible through the regular instrumentation API based on the Paje format.
93 - We also dropped the old and clumsy Lua bindings to create platforms in a
94 programmatic way. It can be done in C++ in a much cleaner way now, which
95 motivates this suppression.
98 - Introduce on_X_cb() functions for all signals, to attach a new
99 callback to the signal X. The signal variables are now hidden and
100 only these functions should be used.
101 Rationale: this enables the usual deprecation scheme where functions
102 remain for 4 releases if we need to modify the signals, while the
103 current code with the signal variables directly visible prevents any
105 - New function: Engine::run_until(date), to split the simulation.
106 - New signal: Activity::on_veto, to detect when an activity fails to start.
107 - Signal change: Comm::on_start(Comm&, bool) has been replaced by
108 Comm::on_send and Comm::recv. These two signals respectively correspond to
109 when the sending or receiving side of a Comm is ready. They are raised at
110 the same locations as the former Comm::on_start signal.
111 - New function: Engine::track_vetoed_activities() to interrupt run()
112 when an activity fails to start, and to keep track of such activities.
113 Please see the corresponding example for more info.
114 - New functions: s4u::Comm::{sendto_init, set_source, set_destination} to enable
115 the use of vetoers with direct host-to-host communications. Both source and
116 destination have to set for a comm to start. Each call to these setters check
117 if all vetoes are satisfied. When it is the case, the comm starts. A use case of
118 these functions is given in examples/cpp/dag-scheduling.
119 - New functions: {Exec, Io}::update_priority allow you to modify the priority of
120 these kinds of activities during their execution. Behavior is detailed in
121 examples/cpp/io-priority/.
124 - Dynamic costs for MPI operations: New API to allow users to dynamically
125 change injected costs for MPI_Recv, MPI_Send and MPI_Isend operations.
126 Alternative for smpi/or, smpi/os and smpi/ois configuration options.
127 - Fix some issues with the replay mechanism.
130 - Function xbt::Extendable::get_data() is now templated with the type of the
131 pointee. Untyped function is deprecated. Use get_data<void>() if you still
132 want to retrieve void*.
135 - New section: "SimGrid MPI calibration of a Grid5000 cluster"
136 presenting how to properly calibrate MPI communications in SimGrid.
137 - Complete and reword the platform section, which is now completed.
140 - Thread contexts are used by default with Python bindings. Other kinds of
141 contexts revealed unstable, specially starting with pybind11 v2.8.0.
143 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
144 (FG: issues on Framagit; GH: issues on GitHub)
145 - FG#95: Wrong computation time for multicore execution after pstate change
146 - FG#97: Wrong computation time for ptask+multicore+pstates
147 - FG#98: SMPI offline simulation is inconsistent with the online simulation
148 (deadlocks / message truncation)
149 - FG#99: Weird segfault when not sealing an host
151 ----------------------------------------------------------------------------
153 SimGrid (3.29) October 7. 2021
155 The "Ask a stupid question" release.
157 We wish that every user ask one question about SimGrid to celebrate.
158 On Mattermost, Stack Overflow or using the issues tracker.
161 New modeling features:
162 - Non-linear resource sharing, modeling resources whose performance heavily degrades with contention:
163 - The total capacity may be updated dynamically through a callback
164 and depends mainly on the number of concurrent flows.
165 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
167 - Dynamic factors: model variability in the speed of activities
168 - Each action can now have a factor that affects its progression.
169 This multiplicative factor is applied when updating the amount of work
170 remaining, thereby an activity with factor=0.5 only uses half of the
171 instantaneous power/bandwidth it is allocated and will appear twice
172 slower than what it actually consumes.
173 - This can be used to model a overhead (e.g., there is a 20 bytes
174 header in a 480 bytes TCP packet so the factor 0.9583) but the novelty
175 is this factor can now easily be adjusted depending on activity's and
176 resources characteristics.
177 - This existed for network (e.g., the effective bandwidth depends
178 on the message in SMPI piecewise-linear network model) but it is now
179 more general (the factor may depend on the source and destination and
180 thus account to different behaviors for intra-node communications and
181 extra-node communications) and is available for CPUs (e.g., if you
182 want to model an affinity as in the "Unrelated Machines" problem in
183 scheduling) and disks (e.g., if you want to model a stochastic
185 - For that, resources can be provided with a callback that computes
186 the activity factor when creating the action.
187 - Example: examples/cpp/exec-cpu-factors
188 - The same mechanism is also available for the latency, which
189 allows to easily introduce complex variability patterns.
192 - Added support to programmatic platform creation in Python.
193 Example: examples/python/clusters-multicpu
196 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
197 This can only be set through the API, not through XML files.
201 - Multiple fixes to ensure reproducibility of tracing
202 - scan/excan can now be replayed
203 - wait action now uses ranks and not pid, as the other ones.
204 - smpi/init and smpi/finalization-barrier are now valid for replays.
205 - exit() is now intercepted by SMPI to avoid premature shutdown of
206 simulation. First non 0 return codes is returned as simulation return
210 * New section "Release Notes" documenting recent and current developments.
211 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
212 * Improvements in API Reference for C++ and Python interfaces.
215 - Make wifi creation compatible with ns-3 version 3.34 too.
217 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
218 (FG: issues on Framagit; GH: issues on GitHub)
219 - FG#77: Search feature of doc is broken (update sphinx theme version)
220 - FG#78: Multiple fixes for SMPI replay:
221 - TI tracing of allotallv/w was outputting wrong values
222 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
224 ----------------------------------------------------------------------------
226 SimGrid (3.28) July 14. 2021
228 The Victoriadagarna Release.
231 - C++ platform interface: Users can now describe their platform directly in C++.
232 This provides greatly flexibility and performance improvement for complex
233 platforms. Main features:
234 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
235 "zones", instead of single hosts. This feature enables the description
236 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
237 - StarZone: new zone with a Star-like topology. The routes are defined
238 as a set of links used to communicate from node to everybody (node<->ALL).
239 - Split-Duplex links: auxiliary method to create split-duplex links in
240 the platform, easing its utilisation. It automatically creates both UP
241 and DOWN links (similarly as done in XML).
242 - Please refer to the documentation and the examples included:
243 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
244 - New plugin: Producer-Consumer with monitor. Just requires to include the
245 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
246 associated example (examples/cpp/plugin-prodcons).
249 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
250 s4u::Io::wait_any(), s4u::Io::wait_any_for().
251 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
252 parameter by reference, instead of a pointer.
253 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
254 Explicitly cancel the activity to get back to previous behavior.
255 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
258 - The default SMPI compiler flags are no more taken from the environment.
259 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
260 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
262 --cfg=smpi/finalization-barrier: which can be used to add
263 a barrier inside MPI_Finalize. This can help for some codes which cleanup
264 data attached to a process, but still used in other SMPI processes.
265 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
266 is active when set to false, to keep going after a small error
267 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
268 which may or may not be problematic in the end.
270 - fix behaviour, as maximum iteration count could be ignored
271 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
272 use sampling when the same kernel is called with a different set of parameters
273 which have an impact on the timing.
274 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
275 It should now work with smpi/auto-shared-malloc-thresh.
276 - Improve error handling and reporting in multiple places
277 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
278 validity checks, truncated messages are now an error, return errors
279 when explicitely deleted handles are reused, ...)
280 - RMA: multiple fixes and stability improvements.
281 - analysis (-analyze flag in smpirun):
282 - SMPI can now report buffer leaks as well as MPI handles leaks,
283 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
284 - if -trace-call-location is used when compiling, SMPI can report
285 origin of leaked handles/buffers
286 - group leaks by type/origin in output message if possible
287 - New implemented MPI calls: MPI_Comm_test_inter
290 - Changed internal implementation of bandwidth factors in network models.
291 Models affected: CM02, LV08 (default), SMPI, IB.
292 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
293 Bandwidth factors are applied to communications to describe that users
294 cannot use 100% of the available bandwidth. For example, the default network model,
295 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
296 behavior was implemented by limiting the bandwidth available in the LMM
297 system for this flow. This may give the false impression that there is
298 bandwidth available for other flows due to its underutilization, especially
299 for the dynamic bandwidth factors used in SMPI models.
300 To avoid this, we have modified the implementation so that each flow uses the
301 maximum physical bandwidth according to the LMM system.
302 However, the actual throughput of the flow seen by the user is defined by
303 the physical bandwidth multiplied by the bandwidth factor.
304 This change impacts on the simulation results for all network models on
305 which we have bandwidth factors configured.
306 *****************************************
307 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
308 *****************************************
309 This change may impact on the timing of your simulation results.
310 Take care when comparing simulations from different SimGrid's
311 versions. Sorry for the inconvenience.
312 - Dynamic network factors: users can configure a callback to define
313 the network factors dynamically. This API is available at
314 simgrid::kernel::resource::NetworkModelIntf.
315 - Users have access to complete information about the current communication
316 to decide which factor to apply. This includes: message size, source and
317 destination hosts, links and zones traversed.
318 - Dynamic factors for both latency and bandwidth.
319 - For more details, see the example in (examples/cpp/network-factors).
320 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
321 deprecated since version 3.24, are no longer supported. Instead, use
322 "wattage_off" and "wattage_per_state".
325 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
328 - New tutorial: Model-checking and formal assessment
329 - New sections: "Demystifying the routing" and "C++ platforms"
330 - Update and improve the part on visualization in MPI and Algo tutorials.
331 - Improve the section on routing: how to define it, how it's used internally
332 - Fix many issues, broken links and missing references in doxygen and Sphinx
335 - Lua platform files are deprecated. Their support will be dropped after v3.31.
338 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
340 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
341 (FG: issues on Framagit; GH: issues on GitHub)
342 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
343 - FG#64: Configuring smpi/IB-penalty-factors
344 - FG#67: Running computation concurrently with MPI_Iallreduce
345 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
346 - FG#71: Segmentation fault on invalid gw_src/gw_dst
347 - GH#322: Issue when an actor kills his host vm
349 ----------------------------------------------------------------------------
351 SimGrid (3.27) March 29. 2021
353 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
355 Main user-visible changes:
356 - All functions marked as deprecated for one year were removed.
357 - Removed all that was related to the "Storage" type in all APIs and bindings.
358 "Disks" have been introduced more than a year ago to replace them for the better.
359 It's time to finish this replacement.
360 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
361 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
362 avoiding "initializer element is not constant" errors with some codes. Please be
363 careful to recompile your MPI codes if switching to this new SimGrid version.
364 - Network model ns-3 now requires version ns-3.28 or higher.
365 - Removed some option aliases that were deprecated since v3.23 (June 2019)
366 - 'verbose-exit' is now 'debug/verbose-exit'
367 - 'simix/breakpoint' is now 'debug/breakpoint'
368 - 'clean-atexit' is now 'debug/clean-atexit'
371 - New functions: Comm::sendto_{init,async} to initiate a communication
372 on between two (possibly remote) hosts.
373 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
374 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
375 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
376 or Mailbox::get_async<void>() if you really want to play with void*.
377 - Unify the interface of Activity::on_{start/activity}
378 - New function: Comm::get_dst_data()
379 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
382 - ptask should now work better with multicore hosts (FG#37).
383 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
384 - Internal cleanups and refactoring around models, to prepare for programmatic
386 - Modernize our tests of the performance models
390 - New section: Configuring the logs from the command line.
391 - New section: Programming API of logs.
392 - New section: Getting realistic results
393 - New section: Examples of SMPI Usage
394 - New part: SimGrid's Internals, with a section on Design goals
395 - Move back to the classical Breathe tool chain, and fix many small glitches.
398 - Handle leak checking: SMPI will now display a message if MPI handles are still
399 present at the end of a simulation. Option smpi/list-leaks:n will display the n
400 first leaked pointers and their types.
401 - smpi/display-allocs option : Display Memory allocations analysis, with hints
402 about the largest allocations. This only works for allocations with malloc/calloc
403 and not compiled with SMPI_NO_OVERRIDE_MALLOC
404 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
405 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
406 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
407 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
408 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
409 - Replay now opens and parses the trace file only once for all ranks when possible.
411 - better handling of MPI_PROC_NULL
412 - MPI_Group_incl was broken in some cases
413 - PAPI tracing now fixed
414 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
415 - SMPI will now fail properly when receive size < sent size, instead of
416 just setting MPI_ERR_TRUNCATE and keep going.
417 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
418 - Remove useless build patches and split build/run patches on the page
419 - Build in parallel when possible
420 - Test on yum-based systems
421 - New nightly tested Proxy-apps:
422 - OpenMPI mpi-test-suite
423 - Coral MCB, LSMS, Qbox, CLOMP
424 - ECP/Mantevo Cloverleaf3D, Ember
425 - Codevault bhtree_mpi
428 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
429 - Start to overhaul the implementation to modernize it (still ongoing).
432 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
433 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
434 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
435 - Deprecate xbt_dynar_to_array().
436 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
438 C binding and interface:
439 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
440 - Constness requirement has been lowered for parameter argv of sg_actor_start()
441 and sg_actor_create() It's now (char* const*). The previous version is still
442 available as sg_actor_start_() and sg_actor_create_().
444 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
445 (FG: issues on Framagit; GH: issues on GitHub)
446 - FG#37: Parallel tasks are limited to 1 core per host
447 - FG#62: Running "smpirun -replay" on large networks
448 - FG!46: Fix a few potential memory leaks in SMPI colls
449 - GH#343: Wrong consumed energy when an exec timeouts
451 ----------------------------------------------------------------------------
453 SimGrid (3.26) December 15. 2020
455 The Release release (the French lockdown was eased today).
457 Important user-visible changes:
458 - SimGrid now requires a compiler with C++14 support.
459 Sibling projects should upgrade their FindSimgrid.cmake
460 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
461 several users had difficulties to understand issues when using high bandwidth or
462 small latency events. The new value was already the default for SMPI and
463 should not cause too much performance hit. It can change some simulation
464 timings if simulation was skipping a lot of small events (which was actually
465 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
466 - On some network models (L07, CM02, LV08), an implicit loopback link is created
467 for local communication. Default bandwidth was 500MBps and latency was 15µs,
468 which was rather bad. Default values have been changed to 10GBps and 0 latency.
469 Options network/loopback-lat and network/loopback-bw have been added to adapt
470 these values if needed. Network and routing models already providing separate
471 loopbacks (clusters and such) are not impacted by this change.
472 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
473 metadata that are added to each MPI message by MPI libraries (tag,
474 src/dest, and communicator information). This may change simulation times
475 for very small message sizes, but should be more realistic.
478 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
479 - New plugins: wifi_energy and link_load. See the associated examples.
480 - New examples: wifi networks, and communication suspend.
483 - LTO is now enabled for Intel/clang compilers.
484 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
485 Setting it to "auto" will use all available cores, while setting it to n will
486 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
487 - Remove obsolete runtime option 'contexts/parallel-threshold'.
488 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
489 has been kept for compatibility.
490 - Finally remove obsolete snake_case() aliases for runtime options.
491 - Further improve the documentation.
494 - the on_exit() of each actor is also executed when the simulation deadlocks.
495 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
496 An example is provided for s4u::Comm but it should work with Exec and Io.
499 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
500 https://framagit.org/simgrid/SMPI-proxy-apps
501 - MPI/IO: supports names starting with "./"
502 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
504 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
505 building shared libraries, this may cause issues, so environment variable
506 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
507 - most temporary files should now be created in /tmp dir (or equivalent).
508 If this one does not allow execution of code (noexec flag), this may cause issues.
509 Please use another tmp directory (using TMPDIR or equivalent system variable)
513 - A large refactoring is ongoing, but this should be transparent for users so far.
516 - Drop xbt_str_split().
518 C binding and interface:
519 - The return type of the sg_actor_on_exit() callbacks is now 'void'
520 (int value was previously ignored)
521 - Many C functions were renamed to follow a common naming style
522 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
523 names are kept until version 3.30.
524 - Many MSG tests were converted to the new S4U's interface in C, that
525 was extended for that.
527 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
528 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
529 - FG#41: Add sg_actor_create C interface
530 - FG#43: xbt::random needs some care
531 - FG#48: The Impossible Did Happen (yet again)
532 - FG#50: Suspending an actor executed at the current timestamp fails
533 - FG#52: Zero-seconds timeout : "That's in the past already"
534 - FG#53: Crash while using ns-3 network model
535 - FG#54: How to suspend a comm?
536 - FG!22: Stochastic Profiles
537 - FG!24: Documentation and fix for xbt/random
538 - FG!35: Add a modeling hint for parallel links in doc
539 - FG!36: [xbt/random] Read/Write the state of the RNG
540 - FG!42: Link cumulated load plugin
541 - FG!44: Test condition variable wait_until+notify_all
542 - GF#18137: Allow different stack sizes?
543 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
544 - GH#139: Allow pthread creation in SMPI
545 - GH#336: Packet-level simulation using SMPI?
546 - GH#345: Error in the LMM while migrating a VM
547 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
548 - GH#352: pip install / python setup.py install fail to find pybind11
549 - GH!337: Fix link_energy plugin for wifi platforms
550 - GH!339: Add Mailbox set_receiver method to python binding
551 - GH!344: Cast hugepages macros parameters to int64
553 ----------------------------------------------------------------------------
555 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
557 The Palindrom Day Release.
559 Important user-visible changes:
560 - Improve the Python usability (stability and documentation).
561 - A nasty synchronization bug was ironed out, see also below.
562 - Python's doc was organized as a separate tree, now integrated with C++.
563 - C bindings of S4U were not part of the doc.
564 - The C++ doc was also improved as methods are now split by theme.
565 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
566 - This is mandatory to use the Java bindings.
567 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
568 - SimDAG++: Automatic dependencies on S4U activities (experimental)
569 - Some features are already implemented but not all of them
570 - Cannot block an activity until it's scheduled on a resource
571 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
572 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
575 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
576 - Drop support for Viva/Triva (old visualization tools), hence removed
577 TRACE_get_node_types() and TRACE_get_edge_types() functions.
580 - Require pybind11 version 2.4.
581 - Greatly improve locking during context switches with Python.
582 - Declaring Python callbacks to this_actor::on_exit is now working.
583 - No more tests are failing currently, even on win32.
586 - Actor: Merge signals on_migration_start/end into on_host_change
587 - Actor: Rename migrate() into set_host()
588 - Disk: Allow users to get the read and write nominal bandwidth values
589 - Exec: Implement wait_for(timeout)
590 - Io: Implement wait_for(timeout)
593 - Parse errors now raise a simgrid::ParseError that you may want to catch.
596 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
597 been renamed to fix a spelling error. As usual, the previous definitions are
598 kept with a deprecation warning. However, to avoid runtime errors, it is now
599 forbidden to override these deprecated methods in a derived class. Please use
600 the new names immediately if you need to override them.
603 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
604 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
605 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
606 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
607 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
608 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
609 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
610 - New module: random, providing classical random numbers generators.
613 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
614 allocations' size above which they are considered "shared" by default (as if
615 they were performed through SMPI_SHARED_MALLOC macros).
616 Default = 0 = disabled feature.
617 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
618 This can cause some troubles if codes are already overriding these. If this
619 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
620 help, but will make this feature unusable.
621 - Input check is now performed much more consistently and outputs precise
622 warnings for each error found
623 - New supported calls : MPI_File_set_view, MPI_File_get_view
624 - MPI I/O now only supports disks, as storage is discontinued.
625 - MPI I/O allows opening files without specifying absolute path
626 - Note : SMPI configuration options are now initialized by smpi_init_options()
627 call, which should be called if SMPI is not being used through smpirun.
629 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
630 - FG#9: Python bindings crashing
631 - FG#39: Missing s4u::Comm::wait_any_for example
632 - FG#42: Add support for ThreadSanitizer (TSan)
633 - FG!19: Removing RngStream
634 - FG!20: A module for RNG calls
635 - FG!21: Choice between ad-hoc and standard distributions implementations
637 - FG!25: Fix link in Mutex doc
638 - FG!26: Fix links in SMPI interface doc
639 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
640 - GH#305: Unscheduled tasks are still excuted
641 - GH#313: smpirun: manual outdated w.r.t. --help
642 - GH#321: [S4U] Get task remaining work ratio
643 - GH#323: Crash when an actor turn off his physical host
644 - GH#335: Missing links on dragonfly example svg
645 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
646 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
648 ----------------------------------------------------------------------------
650 SimGrid (3.24) October 9. 2019
652 The Clean Disk Release.
655 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
656 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
657 thus becomes the default resource to execute I/O operations. Examples
658 and the file system plugin have been modified to use disks instead of
659 storage elements. Storage can still be used but is doomed to disappear.
660 Users are thus adviced to update their simulators accordingly.
661 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
662 for consistency with pthread_barrier_wait()
663 - Host::get_englobing_zone() returns the englobing netzone
664 - Actor::on_destruction is now called in the destructor
665 Actor::on_termination new signal called when the actor terminates
667 - Global signals are now part of the Engine:
668 - on_platform_creation: after config settings, before the XML parsing
669 - on_platform_created: right after the XML parsing
670 - on_time_advance: each time the clock advances
671 - on_simulation_end: after simulation, before cleanups
672 - on_deadlock: as the name implies.
674 - sg_{actor,host,link}_{data,data_set}() now all exist.
675 Use them to attach user data to the object and retrieve it.
678 - Introduce an experimental Wifi model. It sounds reasonable
679 according to the state of the art, but it still has to be properly
680 validated, at least against ns-3 if not against reality.
681 - Improved the usability of ns-3. Several bugs were ironed out.
682 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
683 It is now expressed as 'idle:epsilon:allCores' to properly model the
684 consumption of non-whole tasks on mono-core hosts. More info in the doc.
687 - convert a new set of functions to the S4U C interface and move the old MSG
688 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
691 - Fortran bindings for DVFS have been removed.
692 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
693 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
694 - SMPI can now be selected by cmake's find_module(MPI) with
695 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
696 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
697 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
698 - trace-call-location can be used with TI traces, and replayed, and is compatible
699 with smpi/comp-adjustment-file.
700 - sleep events are now correctly traced/replayed.
701 - Default for trace-call-location is now to use file names and not full paths.
702 To revert to previous behavior (in case of collision of filenames), option
703 "smpi/trace-call-use-absolute-path" can be set to yes.
706 - Use the included xxHash as an hash implem when C++14 is usable.
707 - Option model-checker/hash was removed. This is always activated now.
708 - New option smpi/buffering controls the MPI buffering in MC mode.
709 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
710 This is useful to check for MPI compliance.
713 - New section on plugins: how to define a new one, and existing ones.
716 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
717 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
720 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
721 and <mount> tags. The rationale is finish to decouple what should be handle
722 by SimGrid kernel (i.e., resources and activities) from plugins built in
723 user space (e.g., the file system plugin). This new tag currently accepts
724 only three attributes (id, read_bw, and write_bw). All the other information
725 that was declared with the storage related tags now has to be expressed as
726 properties. An example of platform using this new tag is available at
727 examples/platforms/hosts_with_disks.xml
730 - 'expect signal' can now accept more than one potential signal.
732 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
733 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
734 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
735 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
736 - FG#31: per-actor data
737 - FG#34: SG_BARRIER_SERIAL_THREAD?
738 - FG#35: model-checker does not like buster-produced binaries
739 - FG!13: MC: complete workaround in the error msg seen on modern systems
740 - FG!15: execute_flops now logs compute
741 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
742 - FG!17: ns-3: unblock the right number of communications + others issues
743 - FG!18: Improving the performance of the ns-3 bindings
744 - GH#189: Energy consumption of parallel tasks
745 - GH#219: Error in the throughput of TCP transfer
746 - GH!330: Minor polishing of English: Configuring_Simgrid
747 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
748 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
749 - GH!334: Rework energy plugin, again
751 ----------------------------------------------------------------------------
753 SimGrid (3.23.2) July 8. 2019
756 - Nicer introduction page.
757 - Migrate the "Deploy your application" page to the new doc.
758 - Move Java as a subtree of MSG.
761 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
764 - Drop xbt_dynar_sort_strings().
767 - Really fix FG#26: Turning off a link should raise NetworkFailureException
768 - FG#27: Wrong exception thrown to wait_any when link is turned off
769 - GH#328: Java: Canceling multiple tasks in a single vm/host
771 ----------------------------------------------------------------------------
773 SimGrid (3.23) June 25. 2019
775 The Exotic Solstice Release.
778 - SunOS and Haiku OS support. Because exotic platforms are fun.
779 - Stop setting random seed with srand48() at initialization.
780 - Use addr2line as a fallback for stacktraces when backtrace is not available.
781 - Build option -Denable_documentation is now OFF by default.
782 - Network model 'NS3' was renamed into 'ns-3'.
785 - Simgrid can now hopefully be installed with pip.
788 - wait_any can now be used for asynchronous executions too.
791 - New log appenders: stdout and stderr. Use stdout for xbt_help.
792 - Drop xbt_dict_dump.
795 - 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
796 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
797 - MPI interface is now const correct for input parameters
798 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
799 - Fortran bindings for async collectives.
800 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
803 - Remove option 'model-check/record': Paths are recorded in any cases now.
804 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
805 always sparse. That's an immense gain in memory, and can even be
806 faster because of cache effects. So there is no need to clutter the
807 code to allow the user to go for the unefficient mode.
810 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
811 models proved to be more accurate than these old experiments.
813 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
814 - FG#1: Broken link in error messages
815 - FG#2: missing installation documentation
816 - FG#3: missing documentation in smpirun
817 - FG#6: Python bindings not available on PyPI
818 - FG#7: simple cmake call requires doxygen
819 - FG#8: make python bindings an optional dependency
820 - FG#10: Can not use MSG_process_set_data from SMPI any more
821 - FG#11: Auto-restart actors forget their on_exit behavior
822 - FG#12: -Denable_lto=OFF doesn't disable LTO
823 - FG#13: Installs unstripped file 'bin/graphicator'
824 - FG#14: Installs the empty directory 'doc/simgrid/html'
825 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
826 - FG#17: Dead link in doc (pls_ns3)
827 - FG#20: 'tesh --help' should return 0
828 - FG#21: Documentation link on http://simgrid.org/ broken
829 - FG#22: Debian installation instruction are broken
830 - FG#26: Turning off a link should raise NetworkFailureException exceptions
831 - GH#133: Java: a process can run on a VM even if its host is off
832 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
833 - GH#326: Valgrind-detected error for join() when energy plugin is activated
835 ----------------------------------------------------------------------------
837 SimGrid (3.22) April 1. 2019
839 The Easter Christmas Release (this one is somewhat late).
842 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
843 - Not complete yet: asynchronous activities (amongst others) are still missing
844 - Still ongoing: the interface may change in the future. We need more
845 testers! Please report any glitches.
846 - No new project using Java should start now. Please switch to Python.
849 - Some of the internal cleanups may lead to speed improvements:
850 - The hard limitation on the amount of simulated actors with
851 Java+Mac was removed. Now, the available memory is the only limit.
852 - Our refcounting was tidyied, leading to 10% speedups in some cases.
853 - We are still working on making our code robust to the actor kills
854 and hosts' churn. Things are improving, but it's not perfect yet.
855 - Replaced our own code to display a backtrace (that was forking addr2line)
856 with the Boost.Stacktrace library.
857 You won't see your backtraces without this optional dependency.
858 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
859 - Stop setting random seed with srand() at initialization.
862 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
863 XML file version remains 4.2 since old files are still compatible.
866 - Process termination which was broken at version 3.21 has been repaired.
867 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
868 - Hide the examples into examples/deprecated. New users should use Python.
871 - Drop MSG_process_create_from_stdfunc() from the API.
872 This C++ function was a pimple in the C API, made necessary at some
873 point by the Java bindings. This is fixed now.
874 - Hide the examples into examples/deprecated. New users should use S4U.
875 - MSG_process_create and MSG_process_attach now crash if the host used
876 to run this process is off.
877 - Fix the protype of MSG_process_on_exit()
878 Now use (int,void*) callbacks instead of (void*,void*) ones.
879 The implementation was ways too messy to actually work, I guess.
882 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
883 too often and provide more accurate timings.
884 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
885 - MPI_Alltoallw support
886 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
887 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
888 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
889 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
891 - Basic implementation of generalized requests (SMPI doesn't
892 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
895 - Drop sg_cmdline. Please use xbt_cmdline instead.
896 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
897 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
898 OsSemaphore is implemented in a portable way with C++11 threads.
899 This should allow much more threads to be created at the same time,
900 allowing Mac and Java users to simulate many more actors.
901 - Implement the 'thread' factory with std::thread instead of xbt ones.
902 It is not possible to set the stack size with threads anymore, but
903 -fsplit-stack is the way to go nowadays when using threads.
904 - Drop the xbt_os_thread_t module (now unused)
905 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
906 - Drop xbt_str_join_array().
907 - Drop cunit, use Catch2 instead.
910 - Many cleanups in the kernel::activity namespace. This was long
911 overdue, and shall open the path to many future endeavors.
914 - #132: Java: a process can not shut down its own host
915 - #220: S4U: detached send causes exception if sender terminates soon after sending
916 - #261: Document the parameters of parallel execution's constructor
917 - #300: [s4u] BarrierPtr is missing
918 - #314: SMPI args internal cleanup
919 - #316: Fix a bug related to the CPU utilization of multi-core VM
920 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
921 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
922 - #325: Turning off a host has different behavior on sleeping actors and computing actors
924 ----------------------------------------------------------------------------
926 SimGrid (3.21) October 3. 2018
928 The Restarting Documentation (TRD) Release.
931 - Convert the user manual to Sphinx for improved usability.
932 Unstable version now hosted on https://simgrid.frama.io/simgrid
933 Many glitches remain; stay tuned.
934 - Restrict the Doxygen reference API to the installed header files.
935 (documenting all internals this way is overkill + doxygen scales badly)
936 - New tutorial on S4U (highly inspired from the old MSG one, but with
937 a git repository to fork as a starting point, and a docker image)
938 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
939 - The Developper manual is still to be converted (not compiled until then)
940 - Some parts are still missing in this conversion (such as the
941 platform and deployment sections) while others were blindly converted
942 and would need more love (such as the configuration flags).
943 Things will certainly further improve in the future releases.
946 - Fully reimplement auto_restart mechanism that was utterly broken
947 (fix #22, #131 and #281 that were all very related)
948 - Implement semaphores (fix #303 and #291)
949 - When creating an actor from a function and its parameters,
950 move-only parameters are not allowed anymore, as it would prevent
951 the actor to be restartable if its parameters are consumed on run.
952 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
953 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
954 IO operations that can be started, waited for, or canceled as a regular
956 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
960 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
961 - New host variable: 'core_count'
964 - Remove xbt_os_thread_specific features
965 - Remove portability wrapper to condition variables
966 - Remove xbt_os_thread_yield()
969 - MPICH collective selector now mimics MPICH 3.3b
970 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
971 - MPI_Init(NULL,NULL) is now allowed
972 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
973 - Command-line arguments should now be handled properly in Fortran simulations
974 - MPI Attributes and related callbacks should work in Fortran
975 - Apps using C + Fortran should now work
976 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
977 - Support MPI_Error_String
980 - Due to an internal bug, Msg.run() must now be your last line.
981 We hope to fix it in a future release, and we are sorry for the inconvenience.
984 - #22: Process autorestart seem to only work with CAS01 cpus
985 - #56: Feature request: dynamic SMPI replay
986 - #93: simgrid should not eat --help
987 - #111: How to change the loopback link characteristics?
988 - #116: Platform cluster radical
989 - #117: Platform cluster/host order
990 - #129: Replay_multiple on steroid
991 - #131: Java : autostart does not work
992 - #143: Setting a breakpoint at a given time
993 - #235: Network energy models should be integrated
994 - #264: Add ptask L07 resource tracing
995 - #271: Dynamic and manual replay of SMPI traces
996 - #279: Breakpoints option uses time deltas instead of absolute time
997 - #281: Daemonized actor no longer daemonized after an auto-restart
998 - #282: TIT convertor erase traces when using absolute path in the trace list file
999 - #285: segfault when a s4u actor kills itself with msg process tracing activated
1000 - #286: Pajé traces are not exposing the number of cores
1001 - #287: Command-line parsing should stop when encountering '--'
1002 - #288: MPI_Init(NULL, NULL)
1003 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
1004 - #290: Method to check if a mailbox has a message ready to be consumed
1005 - #291: [s4u] semaphores do not seem available yet
1006 - #294: Alltoall collective implementation of mpich
1007 - #295: Starting a Host as OFF is not supported
1008 - #296: DTD too permissive
1009 - #299: Add a s4u barrier example
1011 ----------------------------------------------------------------------------
1013 SimGrid (3.20) Released June 24. 2018
1015 The proxy snake_case() release.
1018 - The whole API is now using snake_case() instead of mixing it with camelCase.
1019 Compatibility wrappers in place for one year.
1020 - We took the opportunity to increase the consistency of the API.
1021 Please report any remaining glitches.
1022 - New function simgrid::s4u::Host::get_actor_count:
1023 Returns the number of actors running on a specific host.
1026 - SMPI is now tested with ~45 proxy apps from various sources, with none or
1027 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
1028 - Replay: The replay file has been re-written in C++.
1029 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
1030 supported. They are stored in the trace and used when replayed.
1031 - Basic support of MPI_Cancel. Robustness not guaranteed.
1032 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
1033 (only for MPI_COMM_TYPE_SHARED).
1034 - New option: smpi/privatize-libs, to add external shared libs to be privatized
1035 by SMPI. They will be copied locally and loaded separately by each process.
1036 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
1037 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
1038 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
1039 with internal index optind with multiple processes. Only works if MPI_Init has
1040 already been called.
1041 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
1042 syntax without preprocessing tricks.
1045 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
1046 behavior tracing is no longer limited to MSG
1047 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
1048 this means that traces now need to be updated or re-obtained; both Irecv
1049 and Isend lines in your traces have now in total 5 mandatory fields:
1050 <rankid> <command> <to/from rankid> <tag> <size>
1051 To update your traces, it suffices to add a 0 for the tag here.
1052 - TIT now also supports waiting for a distinct request via MPI_Wait.
1053 Wait/Test now wait for a specific request, not just the last one that was
1054 issued. This unfortunately means another update, because we need to
1055 identify which request you want to wait for. We do this via the
1056 triplet (sender, receiver, tag), which needs to be added:
1057 <rankid> <command> <sender> <receiver> <tag>
1058 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
1059 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
1062 - The deprecation of MSG is ongoing (but this should not impact you).
1063 Many MSG functions are now simple wrappers on the C API of S4U. If
1064 you wish to convert your code to S4U, find the S4U counterparts of
1065 your MSG calls in src/msg/msg_legacy.cpp.
1066 - MSG can still be used, but won't evolve anymore.
1069 - Allow to run the Link energy plugin from the command line with
1070 --cfg=plugin:link_energy
1071 - Rename Energy plugin into host_energy
1072 - Rename Load plugin into host_load
1075 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
1076 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
1077 - Improve performance when handling timeouts of simix synchros.
1080 - Config: the C API is now deprecated (will be removed in 3.23), and
1081 the C++ API has been slightly improved.
1084 - Fix several build issues on OSX.
1085 - Move simgrid_config.h to simgrid/config.h (old header still working)
1088 - #143: Setting a breakpoint at a given time
1089 - #150: Inconsistent event names in SMPI replay
1090 - #258: daemonized actors hang after all non-daemonized actors have completed
1091 - #267: Linker error on unit_tmgr
1092 - #269: SMPI: tracing of MPI_Wait/all/any broken
1093 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
1095 ----------------------------------------------------------------------------
1097 SimGrid (3.19.1) Released March 22 2018
1099 The Fixed ABI Release.
1101 As you may know, we are currently refactoring SimGrid in deep.
1102 Upcoming SimGrid4 will be really different from SimGrid3: modular,
1103 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
1105 Our goal is to smooth this transition, with backward compatibility
1106 and automatic update paths, while still progressing toward SimGrid4.
1108 SimGrid remains open during works: The last pure SimGrid3 release was
1109 v3.12 while all subsequent versions are usable alpha versions of
1110 SimGrid4: Existing interfaces remain unchanged, but the new S4U
1111 interface is budding and the internals are deeply reorganized.
1113 Since 2015, we work hard to reduce the changes to public APIs. When
1114 we need to rename a public library symbol in S4U, we let your
1115 compiler issue an explicative warning when you use the deprecated
1116 function. These messages remain for four releases, i.e. for one full
1117 year, before turning into an error. Starting with v3.15, your can
1118 also adapt to API changes with the SIMGRID_VERSION macro, that is
1119 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
1121 Starting with this v3.19.1, our commitment to reduce the changes to
1122 the public interfaces is extended from the API to the ABI: a program
1123 using only MSG or SimDag and compiled against a given version of
1124 simgrid can probably be used with a later version of SimGrid without
1125 recompilation. We will do our best... but don't expect too much of
1126 it, that's a really difficult goal during such profund refactoring.
1128 The difference between v3.19 and v3.19.1 is that the former was
1129 accidentally breaking the ABI of MSG, while the later is restoring
1132 S4U and kernel APIs will still evolve until SimGrid4, with one-year
1133 deprecation warnings as currently. In fact, cleaning up these
1134 interfaces and converting them to snake_case() is one release goal of
1135 v3.20. But don't worry, we are working to smooth this upgrade path.
1137 In summary, new projects should start with S4U to benefit of the
1138 future, but old MSG projects should still be usable with no change.
1140 ----------------------------------------------------------------------------
1142 SimGrid (3.19) Released March 20 2018
1144 The Moscovitly-cold Spring release.
1146 SMPI user-visible improvements:
1147 - Support of MPI_Type_create_subarray
1148 - Unimplemented MPI calls now abort (instead of issuing a warning)
1149 Some harmless calls around errhandlers still only issue a warning.
1150 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
1151 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
1152 is now possible (mmap privatization was refactored to this end).
1153 Adding new SMPI instances should also be technically doable now.
1155 SMPI internal cleanups:
1156 - Removed index notion from SMPI -- no more "getPid() - 1"!
1157 In tracing files, actors are called rank-<PROCESS-ID> but these are
1158 NOT the real ranks (tracing several communicators is not supported yet).
1159 - Deprecate smpi_process_index() (will be removed in 3.22)
1160 - Better testing on classical proxy apps, through an external project:
1161 https://github.com/simgrid/SMPI-proxy-apps/
1164 - Use a graphical TOC to make it easier to find the documentation you need
1165 - Revamp the MSG tutorial
1168 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
1169 - Remove parameter of MSG_process_killall().
1170 Resetting the PID was bogus anyway (several actors could have the same PID).
1173 - Execution->setHost() can be called after start(), to migrate it.
1174 - Comm::test_any() is now implemented.
1175 - s4u::Actor now has onCreation() and onDestruction() signals.
1176 - Install some kernel header files for the users' plugins and more.
1177 Warning, their API is really not stable yet.
1178 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1179 include/simgrid/kernel/routing/*Zone.hpp
1182 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1184 Build System and other cleanups:
1185 - Remove unused run-time parameter "tracing/onelink-only".
1186 - SimGrid now works with intel compilers (MC still buggy)
1187 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1188 - tesh kills subprocesses on timeout (but not on windows)
1191 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1194 - #194: Feature request: simgrid::s4u::Comm::test_any()
1195 - #245: migrating an actor does not migrate its execution
1196 - #253: Feature Request: expose clusters as objects
1197 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1198 - #255: Tesh broken on Windows
1199 - #256: Modernize FindSimGrid.cmake
1200 - #257: Fix (ab)use of CMake install
1202 ----------------------------------------------------------------------------
1204 SimGrid (3.18) Released December 24 2017
1206 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1208 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1209 - Support for MSG should not be removed from SimGrid before 2020,
1210 but future evolutions will be limited to the integration of
1211 user-provided patches.
1214 - s4u::onDeadlock() signal, run before stopping the simulation.
1215 - s4u::Exec: asynchronous executions (abstraction of background
1217 - s4u::Host->getLoad() returns the achieved speed in flops/s
1220 - Link::name() is deprecated and will be removed in v3.21, use
1221 Link::getCname() instead.
1222 - Mailbox::getName() changed to return a std::string, use
1223 Mailbox::getCname() instead to get a char*.
1224 - Storage::getName() changed to return a std::string, use
1225 Storage::getCname() instead to get a char*.
1226 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1227 to have both allocation and deallocation of the map in user space.
1230 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1231 The amount of flops remaining to do is a vector, not a scalar.
1232 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1233 name implies on both sequential and parallel tasks.
1234 - Both changes fix GitHub's #223 using PR #237 as a basis.
1235 Thanks Michael Mercier.
1236 - Most examples were converted to S4U and hidden elsewhere at tests.
1237 Rationale: we still want MSG to work; we want newcomers to use S4U.
1240 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1241 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1244 - Switch to the faster dlopen privatization mechanism by default
1245 - Documentation improvments
1248 - Live migration is getting moved to a plugin. Dirty page tracking is
1249 the first part of this plugin. This imply that VM migration is now
1250 only possible if one this function is called:
1251 - C/MSG: MSG_vm_live_migration_plugin_init()
1252 - C/C++: sg_vm_live_migration_plugin_init()
1253 - Java: Msg.liveMigrationInit()
1254 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1258 - Define class simgrid::xbt::Path to manage file names.
1259 - Removed unused functions:
1260 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1261 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1262 xbt_graph_export_graphviz()
1263 - xbt/str.h: xbt_str_join()
1264 - Remove unused datatypes:
1265 - xbt/heap.h: use std::priority_queue or boost::heap instead
1266 - xbt/swag.h: use boost::intrusive::list instead
1269 - New link_energy plugin for the consumption of the links.
1270 - All of the operations on files and storage contents have been
1271 packaged into a plugin (src/plugins/file_system). The current
1272 public interface can be found in
1273 include/simgrid/plugins/file_system.h
1274 To use these functions you now have to initialize the plugin by
1275 calling MSG_storage_file_system_init() just after calling
1276 MSG_init() or sg_storage_file_system_init() just after creating
1280 - Remove the undocumented/untested tag <include>
1283 - Remove viva specific tracing as the tool is no longer maintained
1286 - #248: Kill of finished processes leads to segfault
1287 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1288 - #239: please implement signal s4u::onDeadlock()
1289 - #230: segfaults when exit() without run()
1290 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1291 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1292 - #222: Actor::kill() doesn't really kill and segfaults
1293 - #221: odd LMM warning when killing an actor
1294 - #120: Memory leak when the processes are forcefully killed
1296 ----------------------------------------------------------------------------
1298 SimGrid (3.17) Released October 8 2017
1300 The Drained Leaks release: (almost) no known leaks despite the tests.
1302 Even very long simulations will preserve your memory: our testsuite
1303 stresses SimGrid for over 45mn on fast machines for a coverage of
1304 over 80%, and there is only one single known leak, of about 4kb.
1308 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1309 handler may be given to free resources if the comm cannot be completed.
1310 - this_actor::send(mailbox) is now mailbox->put()
1311 - New: simgrid::s4u::Comm::wait_all()
1312 - New: Host.execute() for remote executions.
1315 - Unused option network/sender-gap is removed.
1318 - Deprecate MSG_task_isend_with_matching(): This unused feature
1319 really complicates our internals. Will be removed in v3.20.
1322 - Improved context termination. It is now possible to free resources, even
1323 when a process is forcibly killed.
1324 - ContextBoost: add support for Boost versions above 1.61.
1327 - Bring back run-time option --cfg=exception/cutpath to remove exception
1329 - Removed unused functions:
1330 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1331 xbt_str_rtrim(), xbt_str_trim().
1332 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1335 - Removed header files obsolete since SimGrid 3.12:
1336 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1338 - #3: SD_exit should be made optional
1339 - #120: Memory leak when the processes are forcefully killed
1340 - #159: Threading test regression in Actor refcounting
1341 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1342 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1343 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1344 - #191: VM migration and pstate
1345 - #192: Updating the energy consumptions of all hosts crashes with VMs
1346 - #195: All actors have PID=0 in the logs
1347 - #204: Sometimes segfault with thread contexts and mmap privatization
1348 - #222: Actor::kill() doesn't really kill and segfaults
1349 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1351 ----------------------------------------------------------------------------
1353 SimGrid (3.16) Released June 22. 2017.
1355 The Blooming Spring Release: developments are budding.
1358 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1359 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1360 - Java JAR file should be finally fully working on Mac OSX too.
1362 XML platforms: Switch to platform v4.1 format.
1363 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1364 - <zone> can be used as a synonym for the now deprecated <as>
1365 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1366 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1367 - <actor> can be used as a synonym for the now deprecated <process>
1368 - state_file and avail_file periodicity is now easier to express
1369 (check the documentation) the old behavior should still work.
1370 * Storage: not backward compatible, but it seems that nobody ever used it
1371 - Remove attribute 'content_type' of <storage_type>: was never used
1372 - Make attribute 'model' of <storage_type> optional: for future usage
1373 - Remove Bconnection model property: was never (in)validated. Replaced by
1374 the maximum of the read and write bandwidth as a resource constraint.
1377 - New and Backwards Compatibility break:
1378 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1379 When one wants to get the list of tasks whose states have changed during a
1380 simulation round, s/he has to allocate and free a dynar and use it as
1381 argument to this function. The former SD_simulate (double how_long)
1385 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1386 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1389 - New function to update the consumption of all hosts at once.
1390 - Fix the model for multi-core hosts, linear in the amount of busy
1391 cores with an abnormality for pIdle (see paper).
1392 BUG: Still not satisfactory for ptask on mono-cores.
1395 - The netzone are now available from the MSG API.
1396 The old names still work, but are now deprecated.
1399 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1400 - New option: smpi/keep-temps to not cleanup temp files
1401 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1402 mallocs mode is used, allows to change the size of the fake file used
1403 (default 1MB), to potentially limit the number of mappings for large runs.
1404 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1405 - Fortran ifort and flang compilers support
1406 - New RMA calls supported (experimental):
1407 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1408 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1409 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1410 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1411 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1412 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1415 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1416 - Behind the scene, ActivityImpl are now automatically refcounted.
1417 Removing this nasty bottleneck will greatly simplify our internals.
1420 - Replay: New function xbt_replay_action_get():
1421 Retrieve the function previously associated to an event type.
1422 - DROPPED FUNCTION: xbt_str_varsubst()
1423 - DROPPED MODULE: strbuff. We don't need it anymore.
1424 - DROPPED MODULE: matrix. We don't need it anymore.
1425 - DROPPED MODULE: lib. We don't need it anymore.
1427 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1429 SimGrid (3.15) Released March 22. 2017
1431 The Spring Release: continuous integration servers become green
1432 We fixed even the transient bugs on all target architectures:
1433 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1434 Windows, FreeBSD, NetBSD.
1436 - Rename NetCards to NetPoints.
1437 This was intended to help NS3 users, but that's not a netcard.
1438 That's a point in the routing algorithm, let's avoid wrong simplifications.
1441 - New: MSG_process_yield(). Stop and yield to other processes.
1442 - New: MSG_process_daemon(). Daemon processes are automatically killed
1443 when the last non-daemon process terminates
1444 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1445 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1446 to make room for the upcoming network energy plugin.
1447 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1450 - Ensure that an actor can kill itself with Process::exit()
1451 - Kill the obscure NativeException. Nobody want to survive the issues
1452 it denotes, so use JniException that is a RuntimeException (not to
1453 be caught explicitly).
1454 - Partial bug fix in initialization. SimGrid flags on command line were
1455 consumed at C level but stayed in the original Java String[] args.
1456 This could mess users' args[i] if SG flags were not put at the end of
1458 The SimGrid flags are now removed from the Java arguments. However,
1459 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1460 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1461 It might be if you have --log or --cfg flags in the command line.
1462 - Fix numerous memleaks all around the place. In particular, around VMs.
1466 - s4u::onPlatformCreated: right before the simulation starts
1467 - s4u::onSimulationEnd: right after the main simulation loop
1468 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1469 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1470 event from the availability_file changes the avail speed.
1471 - Links are now usable from s4u
1472 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1473 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1474 - New examples: The conversion of MSG examples to S4U has begun
1475 - Actors: create, kill, migration, and suspend
1476 - Applications: master-worker and token-ring
1477 - Action replay: communications and storage
1478 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1481 - Backwards Compatibility breaks
1482 - The SD_route_get_list and SD_route_get_size functions have been removed.
1483 They are replaced by the unique
1484 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1485 The route from src to dst is built in the links dynar whose size can
1486 be obtained with xbt_dynar_length.
1487 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1488 sg_host_route_bandwidth and sg_host_route_latency.
1489 Macros ensure the backwards compatibility, but you should fix your code
1492 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1493 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1494 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1495 We don't support 100% of it, but it should be enough. Contact us if not.
1496 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1497 naive and probably inaccurate)
1498 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1499 mapped onto a unique small file using some system magic.
1500 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1501 - Fortran: cleanups, fixes, support of user-added operations
1502 - RMA: MPI_Accumulate are applied in correct order.
1503 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1504 - Better support for MPI_IN_PLACE
1505 - Support for MPI_Win attrs and keyvals.
1506 - Support MPI_Comm_group_create, MPI_Type_size_x
1507 - MPI ops should be properly applied to all allowed types
1508 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1509 algorithm by a non failing one from mpich.
1512 - Kill the fifo data container: we don't use it anymore.
1515 - A <cluster> can now be created with different speed values to represent pstates.
1516 see examples/platforms/energy_cluster.xml for an example.
1518 SimGrid (3.14.159) Released December 28. 2016
1520 The previous release was only a crude approximation of a Pi release;
1521 this one should be a bit better.
1523 - Revert a last minute change that broke on Mac OSX.
1524 - Fix the changelog and NEWS files.
1525 - Improve the documentation, in particular of the routing module.
1526 - Integrate some patches flying around in Debian.
1528 SimGrid (3.14) Released December 24. 2016
1531 * The whole documentation was reworked and reorganized. There is
1532 still much room for improvement here, but we're on it.
1535 * We now need python3 for our testing infrastructure.
1536 * Model checking needs libevent
1539 * Now works on FreeBSD
1542 * Privatization now works on FreeBSD
1544 * Privatization is now activated by default in smpirun
1545 You can turn it off with -no-privatize if something goes wrong.
1547 * Call-location tracing for SMPI.
1548 You can add the exact location (filename / linenumber) of an MPI call to
1549 your trace files and slow down or speed up the simulation between two
1550 consecutive calls by using an adjustment file (see the documentation).
1552 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1553 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1554 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1555 values were ignored in some cases. The timings of these functions can now
1556 be significantly different.
1558 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1559 smpi/running-power is renamed to smpi/host-speed
1561 * smpi/grow-injected-times option to enable or disable multiplication of the
1562 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1563 make simulation less precise (but also much faster).
1565 * smpirun script should be (much) faster for large deployments.
1567 * SMPI tracing: fixed issue with poor matching of send/receives.
1569 * Replay: Fix broken waitall
1571 New functions and features
1572 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1574 Dropped / renamed functions and features
1575 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1576 - MSG_mailbox_is_empty() -> Mailbox::empty()
1577 - MSG_mailbox_front() -> Mailbox::front()
1578 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1579 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1580 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1581 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1582 * Don't pass the free_f parameter to property related functions:
1583 - MSG_host_set_property_value()
1584 - MSG_as_router_set_property_value()
1585 - MSG_storage_set_property_value()
1586 * VM properties. Since msg_vm_t are msg_host_t, just use
1587 MSG_host_get_property_value() and friends
1588 * VM I/O related things:
1589 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1590 - Unimplemented save/restore methods
1591 * MSG_as_router_get_property_value() was redundant with
1592 MSG_environment_as_get_property_value().
1593 - Removed MSG_as_router_*propert*() functions
1594 - Added MSG_environment_as_set_property_value() for consistency
1595 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1596 Well, they are still there for now, but deprecated with a warning.
1597 Please switch to xbt_dict_new_homogeneous() before this is removed
1599 * Task affinity. Its intended behavior (that was very badly tested
1600 and probably not really working) was deceiving what most users
1601 would have hoped here.
1602 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1603 * Option network/coordinates is now useless and should be dropped.
1606 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1609 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1610 * Added Dragonfly topology support
1613 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1614 EiBps, EBps, Eibps, Ebps for bandwidth.
1615 They may become useful to some lucky ones.
1618 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1619 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1620 * ex/async_waitAll: new example, on asynchronous communications
1623 * Memory usage should be decreased for simulations with a large number
1624 of processes. This also helps for SMPI.
1626 SimGrid (3.13) stable; urgency=low
1628 The Half Release, a.k.a. the Zealous Easter Trim.
1630 - We removed half of the lines, that were mostly experimental cruft.
1631 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1632 (not counting blanks and comments -- according to openhub.net)
1633 - The internals are now compiled in C++ (and will soon be clean C++)
1634 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1635 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1637 * Backwards Compatibility breaks
1638 - Removed Lua simulation bindings (switch to C or Java for that).
1639 Lua can still be used to describe platforms
1640 - Removed Java kernel plug-ins.
1641 Will be reintroduced after the ongoing major internals reorg.
1643 - the following functions were removed.
1644 They were too specific and should be reimplemented in a generic
1645 way, with filter function.
1646 - MSG_task_listen_from_host
1647 - MSG_mailbox_get_count_host_waiting_tasks
1648 - MSG_mailbox_put_with_timeout was removed.
1649 Please use MSG_task_send_with_timeout instead.
1651 - the SD_application_reinit function was removed. It has been a noop for a while.
1652 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1653 replaced by a more flexible mechanism.
1654 - The following functions thus do not exist anymore
1655 - SD_workstation_get_access_mode
1656 - SD_workstation_set_access_mode
1657 - SD_workstation_get_current_task
1658 - Basic estimation functions have been removed but can easily be replaced
1659 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1660 - SD_workstation_get_computation_time => amount / sg_host_speed()
1662 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1663 Use VM.getSpeed()*load/100 for the legacy behavior.
1665 - option enable_tracing was removed. It was not doing anything for a while.
1666 - In the ModelChecker:
1667 - the model-checker now ptraces the model-checked process which means
1668 you cannot use a debugger on the latter anymore (we might make this
1669 optional in the feature);
1670 - removed soft-dirty page tracking;
1671 - remove model-checked side snapshot management,
1672 MC_snapshot() and MC_compare_snapshot();
1673 - keep the MC_cut() function as a stub (it was not really working
1674 in the previous release).
1677 * All options are consistently in kebab-case. Old names are kept as alias.
1680 * Switch to platform v4 format.
1681 - Rename from 'power' to 'speed' the attributes describing the amount of
1682 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1683 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1684 - In <host> and <link>, attributes availability and state are gone.
1685 It was redundant with state and availability traces, and with peak values.
1686 - In <cluster>, attributes availability_file and state_file are gone.
1687 It was too complex and unused.
1688 - Kill <gpu>. Was not doing anything.
1689 - The DOCTYPE points to the right URL:
1690 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1691 (the file at this address now documents the changelog since its v1)
1692 - A warning is emitted for unit-less values (they are still accepted).
1693 - speed. Default: 'f' or 'flops'. Also defined:
1694 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1695 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1696 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1697 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1698 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1699 - latency. Default: 's' second. Also defined:
1700 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1702 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1703 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1706 * s4u::Host is now the preferred public interface to the Host features.
1707 sg_host_* functions are C bindings to the exact same behavior
1708 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1711 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1712 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1713 They were deprecated since a few years, and probably did not even compile anymore.
1716 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1717 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1718 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1719 encouraged to update their codes. A list of the modified functions can be found at the end of
1720 include/simgrid/simdag.h
1723 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1724 the lambda or closure passed as a parameter will run in kernel mode.
1725 All the callback functions should be rewritten to that interface at some point.
1728 * Reorganizing and cleaning the internals all around the place.
1731 * Remove old default barrier/bcast buggy algorithms (see #18407)
1732 * Various bug fixes to handle more codes
1733 * Remove the need for the --foreground option of smpirun (it is still
1734 accepted for backward compatibility).
1737 * Kill the setset data container: MC don't use it anymore.
1738 * Kill the queue data container: it made more sense with GRAS.
1739 * Kill the xbt_peer_t data type: it's useless without GRAS.
1740 * Kill rm_cb feature of config sets: it was never useful.
1741 * Kill graphxml parsing feature. It was not used.
1742 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1744 - xbt_dynar_sort_strings(), when the content is char*
1745 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1746 They throw exceptions on invalid input;
1747 * C++ support for declaring CLI flags (simgrid::config::Flag);
1748 * class for abstracting different signal backends (simgrid::xbt::signal).
1749 with no external dependencies (we need very simple signals).
1752 * refactoring and cleanup of the code;
1753 * ongoing process to cleanly separate the model-checking algorithms
1754 from the code model-checking support.
1756 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1758 SimGrid (3.12) stable; urgency=low
1760 The Facelift Release.
1763 * Require g++ v4.7 at least to not speak prehistorical C++.
1764 * Require Boost 1.48 (for signal2 component).
1765 * Java must be version 7 at least when activated.
1766 * Builds on Windows again (including Java bindings).
1767 * Tracing is now always enabled (no way to turn it out)
1768 * Remove GTNetS. It was not working anyway.
1769 * Various cleanups in the cmake scripts.
1770 * Move headers around to sort them out on installed systems:
1771 - instr/instr.h -> simgrid/instr.h
1772 - instr/jedule/* -> simgrid/jedule
1773 - simdag/datatypes.h was removed
1774 - simdag/simdag.h -> simgrid/simdag.h
1775 - msg/datatypes.h was removed
1776 - msg/msg.h -> simgrid/msg.h
1779 * Interface improvement:
1780 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1781 - Sanitize the interface in MSG_task_ module:
1782 - Merge two functions that were close enough but misleading:
1783 set_compute_duration(t) -> set_flops_amount(t)
1784 get_remaining_computation(t) -> get_flops_amount(t)
1785 - set_data_size(t) -> set_bytes_amount(t)
1786 get_data_size(t) -> get_bytes_amount(t)
1787 - Massive cleanups in the functions related to the energy
1788 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1789 - New: MSG_host_get_pstate()
1791 - msg/energy/onoff: switching hosts on and off
1793 * Interface improvement:
1794 - Rename Host.isAvail() to Host.isOn()
1795 - Rename Process.currentProcess() to Process.getCurrentProcess()
1796 - Rename Task.setDataSize() to Task.setBytesAmount()
1797 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1799 - #18874: Actually allows the GC to reclaim tasks
1803 - SIMIX_process_throw: raises an exception in a remote process
1804 * Refactoring: Separate sub-modules
1805 - libsmx: the public interface, as libc in a real system
1806 - popping: the strange dance that converts a user request into a kernel handling
1807 - smx_context_*: the virtualization mechanisms that embed the user code
1808 - smx_*: the handling of each simcalls
1809 * Interface cleanup:
1810 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1811 * Rename smx_action_t into smx_synchro_t, making explicit that these
1812 things are used to synchronize processes with their environment.
1813 For example, a communication is a sort of synchronization involving
1814 the communicating processes (that may block until the exchange) and
1815 the platform. The same can be said from computations, etc.
1817 - #18888: segfault when a process ends before its kill_time
1821 - 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.
1822 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1823 - MPI_Type_set_name, MPI_Type_get_name
1824 - MPI_*_c2f and MPI_*_f2c functions
1825 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1826 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1827 - Activate a lot of new tests from the mpich 3 testsuite
1829 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1830 - InfiniBand network model added: Based on the works of Jerome Vienne
1831 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1832 - When smpi/display_timing is set, also display global simulation time and application times
1833 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1834 * Collective communications
1835 - 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
1836 - 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/).
1837 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1839 - Replay now uses algorithms from wanted collective selector
1840 - Replay can be used with SMP-aware algorithms
1841 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1842 - 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.
1844 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1845 - Correctly use loopback on fat-tree clusters
1846 - Asynchronous small messages shouldn't trigger deadlocks anymore
1847 * Energy/DVFS cleanup and improvement
1848 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1849 - new: smpi_get_host_pstate()
1852 - "Full" network optimization flag was broken since Surf++
1853 - Better handling of precision flags in maxmin
1854 - Fix bug causing sometimes "Impossible" errors
1855 - Properly pass cluster properties to included hosts
1856 * Improvement of the Energy plugin.
1857 - Always update the consumption before returning that value
1858 - New property: watt_off to denote the dissipation when the host is off
1859 - New functions getWattMinAt and getWattMaxAt to retrieve the
1860 dissipation of pstates that we are not currently at.
1861 * Java: class NetworkLink renamed to Link
1862 * New function: simcall_process_get_kill_time()
1863 * Massive rename s/workstation/host/
1864 - That's intrusive, but that's good for the project consistency. Sorry.
1865 - Change config option "workstation/model" into "host/model"
1868 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1869 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1870 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1871 - xbt_dirname and xbt_basename for non-POSIX systems
1873 * The model checker now runs as a separate process.
1874 * The model checker runs must now be launched with the new simgrid-mc program.
1875 * Record/Replay: the MC can display a textual representation of a path in the
1876 execution graph. It can then be replayed outside of the model checker.
1878 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1880 SimGrid (3.11) stable; urgency=low
1885 * Normalizing pointers addresses tool for better diff between logs
1888 * Add cloud examples using new VMs
1889 - examples/msg/cloud/two_tasks_vm.tesh
1890 - examples/msg/cloud/simple_vm.tesh
1891 - examples/java/cloud/cloud.tesh
1892 - examples/java/cloud/migration/migration.tesh
1893 * Add java surf examples:
1894 - examples/java/surfPlugin/surf_plugin.tesh
1895 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1896 - examples/java/surfCpuModel/surf_cpu_model.tesh
1897 * Add SMPI+MSG example:
1898 - examples/smpi/smpi_msg_masterslave/
1903 - msg task destroy cancel test
1904 - msg_host on/off test
1905 * Move all tests in testsuite to teshsuite (adding tesh files)
1906 * Restructure teshsuites
1907 - one folder for each kind of test
1908 * Restructure AddTests.cmake
1910 - structure the order of tests (with sections)
1913 * Add virtual machine
1914 - creation of a VM on a PM
1915 - migration of a VM from a PM to another PM
1917 - MSG_process_join(msg_process_t process, double timeout)
1918 - msg_bar_t MSG_barrier_init(unsigned int count)
1919 - int MSG_barrier_wait(msg_bar_t barrier)
1920 - void MSG_barrier_destroy(msg_bar_t barrier)
1921 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1922 * New option "msg/debug_multiple_use" to help debugging when a task is used
1925 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1927 - Increase code coverage in test suites
1931 * Protect context stacks against stack overflow. The number of protected memory
1932 pages allocated on the top of each stack (1 by default) can be configured
1933 with the new command line option --cfg=contexts/guard_size:<value>.
1934 * Simcalls are now generated by a python script that
1935 - generates files included by SimGrid
1936 - checks that all the functions exist, and proposes prototypes
1938 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1939 host_set_data, host_get_data
1941 - simcall_process_join(smx_process_t process, double timeout)
1942 * Fix bug where sleeping processing could not be suspended.
1945 * Translate surf models from C to C++
1946 - Generic classes for all models: Model, Resource, Action
1947 - A generic interface for each kind of model (CPU, Network, Storage
1948 Workstation, WorkstationVM)
1950 * Translate surf routings from C to C++
1951 * Add callbacks using sigc++ or boost::signals2
1952 - Add callback functions for resource creation/destruction
1953 - Add callback functions for action state change
1954 - Handle Energy as a plugin
1955 * Replace swag by boost::intrusive
1956 * Add new routing models for clusters. For documentation, see
1957 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1958 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1959 parameters for cluster tag
1960 - Fat trees, with topology="FAT_TREE" and
1961 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1962 - see examples/platforms/torus_cluster.xml and
1963 examples/platforms/fat_tree_cluster.xml
1964 * More documentation
1967 * Hostfiles support host:nb_processes construct to deploy several processes on
1969 * Collective communication algorithms should not crash if used with
1970 improper number of nodes and report the error.
1971 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1972 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1973 MPI_Dims_create, MPI_Cart_sub are supported.
1974 * New interface to use SMPI programmatically (still depends on MSG for
1975 some parts, see examples/smpi/smpi_msg_masterslave):
1976 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1980 * Global variables privatization in MPI executables is now performed at runtime
1981 with the option smpi/privatize_global_variables (default:no).
1982 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1983 dynamic libraries loaded by the application are not privatized (static
1984 linking with these libraries is advised in this case)
1987 * Options defined in XML work correctly now.
1990 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1991 copied into simgrid.jar or not (ON by default). Use this option if you want
1992 to reduce the size of the installed simgrid.jar, *and* the native libraries
1993 are kept installed elsewhere.
1994 * Surf binding with SWIG (code generated in maintainer mode only):
1995 - plugin to handle callbacks
1996 - CPU model only for the moment
1999 * Supernovae build mode is definitively removed. It was used to improve
2000 inlining and inter-module optimizations. It is nowadays superseded by
2001 link-time optimizations commonly available in compilers.
2002 * Update ns-3 find lib. Bindings for ns-3 should work again now.
2003 * Add boost dependency for surf++
2004 * Add new macro for tests
2005 - ADD_TESH(name <tesh_args>)
2006 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
2010 - xbt_bar_t XBT_barrier_init(unsigned int count)
2011 - int XBT_barrier_wait(xbt_bar_t barrier)
2012 - void XBT_barrier_destroy(xbt_bar_t barrier)
2013 * Make the xbt_os_time module public
2015 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2017 SimGrid (3.10) stable; urgency=low
2019 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
2022 * Reintegrate Java to the main archive as desynchronizing these
2023 package is not acceptable anymore (Java is now considered stable)
2024 * Add explicit synchronization facilities through semaphores
2025 * Bug fix: Task.setDataSize() only changed the C world, not the value
2026 cached in the Java world
2029 * Dramatically change the way files are handled. API and internals changed, but
2030 this part of MSG was not considered as production grade either.
2031 * Add explicit synchronization facilities through semaphores
2032 * Add a new function MSG_host_get_process_list()
2033 * Preliminary DVFS support (see examples/msg/energy/ for details)
2036 * SMPI is now included directly in the libsimgrid as the windows
2037 linker doesn't force us on splitting it anymore.
2038 * Improvements of the SMPI replay tool:
2039 - Most of the collective communications are now rooted in the same process as
2040 in the original application.
2041 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
2042 used until now). Multiple data types can now be used in a trace.
2043 - The replay tool now supports traces produce either by TAU or a modified
2045 - Bug Fix: the compute part of the reduce action is now taken into account.
2046 - Gatherv collective is now supported
2047 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
2048 of smpirun outputs time independent traces for the current run. One file
2049 is created per process. If too many processes are simulated, this behavior
2050 can be changed to one file for all processes by using the
2051 tracing/smpi/format/ti_one_file flag
2052 * smpirun generates the host file if needed (with given host count and platform)
2053 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
2054 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
2055 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
2056 --cfg=smpi/coll_selector:(mpich/ompi)
2057 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
2059 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
2060 happening inside a collective SMPI call.
2061 * Fix the behavior of complex data types handling.
2062 * Make MPI_Wtime another synchronization point to take computations into
2064 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
2065 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
2066 * Add all missing Fortran bindings, SMPI should work with Fortran 90
2067 (no privatization of global variables yet)
2068 * Preliminary DVFS support (see examples/smpi/energy/ for details)
2071 * Verification of liveness properties is now available for SMPI applications
2072 (in addition to MSG applications)
2073 * Bugged examples using SMPI in examples/smpi/mc/
2074 * Add --cfg=model-check/visited option. Allows the verification of infinite
2075 programs. Detection of loops in the execution thanks to the system state
2076 comparison and reduction of the state space to explore. Can be combined with
2077 DPOR for safety properties.
2080 * Allow to change SimGrid configuration (see --help) within the code
2081 thanks to SD_config() as it can be done in MSG.
2082 * Add a new function SD_task_set_amount() upon user request.
2085 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
2086 * Remove rule based routing (no more PCRE dependency)
2087 * Add a limiter_link option to cluster tag, to specify a maximum reachable
2088 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
2089 * Add a loopback_bw and loopback_lat options to cluster tag.
2090 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
2091 routing. Now peers are encapsulated in an AS and have their own private
2092 router but this is transparent.
2095 * Our own implementation of getline is renamed xbt_getline, and gets
2096 used even if the OS provide a getline(). This should reduce the
2097 configuration complexity by using the same code on all platforms.
2098 * New type: xbt_cfg_elm_boolean.
2099 * Allow to use yes/no for boolean configuration options in the command line.
2100 * Allow to disable SimGrid cleanups at exit from command line option.
2101 There are situations where one may want a simulation to end with an exit.
2102 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
2103 annoying when scripting around the simulator. Adding a
2104 --cfg=clean_atexit:no allows to circumvent this issue.
2107 * Lots of memory leaks were corrected in this release.
2108 There is no known memory leaks anymore, in all of our 600+ tests.
2109 * New command line option --version, to get SimGrid version information.
2110 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
2112 * Supernovae builds are deprecated, and expected to be removed in the next
2115 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2117 SimGrid (3.9) stable; urgency=low
2119 The Grasgory release: GRAS is really dead now.
2121 * Complete overhaul of the internal host structures scheme.
2124 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
2125 was considered as experimental and badly maintained since then.
2126 * Keeping it was thus a trap to our potential users, that could take
2127 it instead of MSG or SMPI by mistake despite is pity state.
2128 * GRAS seems to have very few users (if any), and no one volunteered
2129 to maintain it further. It also induces a lot of XBT code (for
2130 portability sake), that must be maintained too.
2131 * For all these reasons, we killed GRAS. If someone wants to revive it
2132 in the future, don't cry, our git history still remembers of GRAS.
2135 * Major overhaul. Merge our documentation again as time proved that
2136 splitting it was really not helping our users.
2137 * Further improve the developer documentation to help newcomers
2138 hacking on SimGrid itself. The user documentation (and in
2139 particular, the beginner documentation) is still in a sorry state.
2142 * Now works on Windows too!
2143 * Much more extensive test suite, from MPICH
2146 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
2147 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
2148 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
2149 * Bug fix: let task be scheduled when the last dependency to be solved is
2150 a control dependency.
2151 * Remove SD_load_environment_script function.
2152 Use the C sg_platf function if you want to declare a platform
2156 * New function: MSG_process_get_number()
2157 * Old function documented: MSG_config()
2158 * Remove MSG_load_platform_script function
2159 Use the C sg_platf function if you want to declare a platform
2163 * Change the default value of the TCP_gamma constant (maximal size of TCP
2164 congestion window) to a more realistic 4MiB value. If you notice changes in
2165 your simulation results, you can fall back to the previous 20k tiny window
2166 by adding --cfg=network/TCP_gamma:20000 on command line.
2167 * (Hopefully) fix a bug wrt periodic availability and state traces
2168 * Bug fix: use default values at start when first event in availability/state
2169 trace is not at time 0.
2172 * remove the "new_" part of function name sg_platf_new_trace_connect
2173 (resulting in sg_platf_trace_connect), since it does not create
2177 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2178 use case where it's really mandatory, and maintaining it was a pain in
2180 * New: xbt_fifo_search(), search an item with a user-provided
2181 comparison function instead of dumb pointer comparison.
2184 * Fix the lua deployment:
2185 Use `simgrid.init_application()` before deployment instead of
2186 `simgrid.msg_register_application()` after.
2189 * Transfer the tracing files into the corresponding modules.
2191 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2193 SimGrid (3.8.1) stable; urgency=low
2195 The "we are told that some people want to also *install* the simgrid
2198 * Add missing file "tesh.1" to the archive.
2200 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2202 SimGrid (3.8) stable; urgency=low
2204 The Psssshiiiit release: SimGrid jumps into the Cloud.
2207 * Add an experimental interface to manipulate VMs. They are mainly
2208 process groups with very few intrinsic semantic, but they should
2209 allow you to build the semantic you want easily.
2210 * New function: MSG_host_set_property_value()
2211 * New function: MSG_process_on_exit(). To clean memory in all cases.
2212 * Bug fixes that made the host (and link) failures unusable.
2213 * Add a way to auto-restart process when the host in which they are
2214 executing comes back (ON_FAILURE="RESTART" on deployment file,
2215 MSG_process_auto_restart_set).
2216 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2217 please stop using the old ones, they are DEPRECATED.
2219 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2220 Please use MSG_init() instead. (reducing the amount of entry
2221 points in the library helps us).
2222 * Make it impossible to link against the wrong version of the lib
2223 * Deprecate MSG_clean(). No need to call it anymore.
2224 * Function MSG_get_host_number() is not deprecated anymore.
2227 * Split the doc into a user guide and a reference guide.
2228 * Start a developper guide to help people hacking on SimGrid.
2231 * Enable tracing by default. This modules rocks you should use it.
2232 * Remove option custom_flags. Now use environment variables CFLAGS
2234 * Use default cmake things to detect lua instead of home grown ones.
2235 * New option "enable_mallocators" to disable mallocators, for debugging
2236 purpose ("on" by default).
2239 * Bug fixes around the resource failures: don't let the processes
2240 survive the host they are running onto.
2241 * Add an interface to auto-restart processes when the host in which they are
2242 executing comes back.
2243 * Ensures that SIMIX_clean is called automatically. It's not part of
2244 the public interface anymore (bindings should be updated).
2247 * Bug fix for when SD_Simulate is called with a positive value: be careful
2248 when comparing doubles. Sometimes they are different for non significant
2250 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2251 parallel task whose initial work is distributed among host according
2252 to the Amdahl's law. Such tasks are created with a parameter alpha
2253 that corresponds to the non-parallelizable part of the computation.
2254 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2255 between two sets of workstations assuming a 1D block distribution (each
2256 workstation owns a similar share of data) on both sides.
2258 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2259 Data redistribution will be automatically scheduled once parent and child
2260 are both scheduled. The filling of computation_amount and
2261 communication_amount structures is now done seamlessly thanks to the chosen
2263 * New function SD_workstation_dump to display various information
2264 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2265 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2266 bandwidth on the route onto which the task is scheduled and the amount of
2268 To divide the nominal bandwidth by 2, the rate then has to be:
2269 rate = bandwidth/(2*amount)
2270 * Compute tasks that have failed can now be rescheduled and executed again
2271 (from their beginning)
2272 * Increasing source code coverage (src/simdag is now covered at 95.8%
2276 * Re-implement time-independent trace replay using SMPI (at the
2277 smpi_smp_* level) instead of MSG. This should replace
2278 examples/msg/actions/actions.c
2279 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2280 hindexed and structs)
2281 * Implement the exchange of non-contiguous data.
2282 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2283 * Correct behavior of smpi/sender_gap and set its default value to 0
2284 * Add option to asynchronously send small messages to allow better
2285 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2286 specifies the size in bytes under which messages will be asynchronously sent.
2287 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2288 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2291 * New C interface to define a platform: XML is now optional.
2292 For more info, please check include/simgrid/platf.h.
2293 * New interface to define random platforms from the C:
2294 For more info, please check include/simgrid/platf_generator.h and
2295 examples/msg/masterslave/masterslave_platfgen.c
2296 * Export a sg_cmdline dynar containing all the arguments we got from
2300 * Two new tracing options for adding comments to trace file so you
2301 can track your experiments (see --help-tracing for details).
2302 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2303 * Adding the SimGrid version that generated the trace file as a comment.
2304 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2305 * Fix to avoid key clashes on Paje links
2306 * Other minor fixes related to the Paje specification
2309 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2310 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2311 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2312 * Greatly improve the robustness of mmalloc to user errors (such as
2313 using an area after freeing it, or freeing it twice)
2315 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2317 SimGrid-java (3.8.1) stable; urgency=low
2319 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2320 Warning: all methods are visible, but only some of them are
2321 implemented so far. Check the source (src/jtrace.c)
2322 for further information.
2323 * New module: org.simgrid.msg.File (SimGrid File management functions)
2324 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2325 * Change the meaning of Process.restart: now restart the process from
2326 the beginning, like MSG_process_restart in C.
2327 * Add Process.setAutoRestart: handling of process restart when failed
2329 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2330 you to retrieve the properties of the processes/hosts
2331 * Deprecate Msg.clean(): you can just forget about it now.
2332 * New function Process.getCount(), that only works when compiling
2333 with the not yet released version 3.9 of the C library.
2335 * New context factory based on Coroutines. It mandates a modified JVM
2336 but then, the simulations run about five times faster, and there is
2337 no limit to the amount of processes (beside of the available memory).
2339 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2341 SimGrid (3.7.1) stable; urgency=low
2344 * Restore the prototype of MSG_process_create_with_environment() to
2345 the pre-3.7 situation by removing the kill_time argument.
2346 * Add a MSG_process_set_kill_time() function instead.
2349 * Fix weird behaviors when dealing with parallel tasks.
2352 * Simgrid is now built as a dll.
2353 * Simgrid-java now works on Windows.
2354 * Simgrid-Java is now included into Windows package.
2357 * First pre-build package for MacOSX.
2360 * Fix compilation when using MSG_USE_DEPRECATED.
2361 * Fix some compilation issues on Macs and Windows.
2362 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2363 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2365 We discovered that the Lua console is broken, but we are missing the
2366 manpower to fix it right now. The problem existed in 3.7 too, so we
2367 are not blocking the release for that. Sorry if you depended on this
2368 feature, any help would be really welcome.
2370 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2373 SimGrid-java (3.7.1) stable; urgency=low
2375 The "Java aint got to be bloated and slow" release
2378 * Various internal cleanups and performance improvement
2379 Simulations are expected to run up to twice faster or so
2380 * Make Process.kill(process) an instance method, not a static one
2381 * User processes are not java.lang.Thread subclasses.
2382 This breaks the compatibility (sorry), but previous API was
2383 brain-dead, making it impossible to have non-trivial
2384 initializations in the process constructor.
2385 * Require a full constructor per Process sub-class.
2386 Kinda breaks the compatibility (sorry), but this allows a much more
2387 efficient way to launch the processes at simulation startup.
2388 * Do not embeed our version of semaphores, java 1.5 can be considered
2389 as sufficiently prevalent for us to not dupplicate its features.
2393 * Add examples for almost every part of the API
2394 We spotted and fixed a lot of bugs in the process
2396 * New module: asynchronous communication API
2397 * New function: Process.sleep()
2398 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2399 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2400 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2402 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2404 SimGrid (3.7) stable; urgency=low
2406 The "spring cleanups (before the next Big Project kicks in)" release.
2409 * We can specify the SMPI latency/bandwidth factor with command line
2410 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2411 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2412 You can also use the "config tag" from platform file by adding this line
2413 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2414 (see "example/platforms/tag_config.xml" to use "config tag").
2415 Note that the command line supersedes the platform file configuration.
2416 * Change the correction factors used in LMM model, according to
2417 the latest experiments described in INRIA RR-7821.
2418 Accuracy should be improved this way.
2419 * Use the partial invalidation optimization by default for the
2420 network too. Should produce the exact same results, only faster.
2421 * Major cleanup in surf to merge models and split some optimization
2422 mechanisms from the core of the models. As a result you can now
2423 specify which model to use (e.g., --cfg=network/model:LV08
2424 --cfg=cpu/model:Cas01) and which optimization mode to use
2425 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2426 Incompatible combinations should err at initialization. See
2427 --help-models for the list of all models and optimization modes.
2428 * The CLM03 workstation model was dropped for simplicity because it
2429 used the deprecated CM02 network model. Use default instead.
2430 * Rename the TCP_gamma configuration option to network/TCP_gamma
2431 * Rename the coordinates configuration option to
2432 network/coordinates, and document it
2433 * Use now crosstraffic keyword instead of the terribly misleading
2434 fullduplex keyword. It is activated by default now in the current
2435 default model, use --cfg=network/crosstraffic:0 to turn it off.
2436 * Ongoing refactoring the model parsing to make XML files optional
2437 See include/simgrid/platf.h for details (still to be completed)
2440 * Major overhaul of the documentation. Almost instructive now :/
2441 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2442 functions and friends. This interface was considered as
2443 deprecated since over 2 years, it's time to inform our users that it is.
2444 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2445 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2446 use these (crappy) functions in your code.
2447 These functions will be removed soon. Stop using them now.
2448 * Deprecate MSG_get_host_{table,number}
2449 Implement MSG_hosts_as_dynar() instead.
2450 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2451 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2454 * Stabilize the parallel execution mode of user contexts
2455 * Introduce configuration variables to control parallel execution:
2456 - contexts/synchro: Synchronization mode to use when running
2457 contexts in parallel (either futex, posix or busy_wait)
2458 - contexts/parallel_threshold: Minimal number of user contexts
2459 that must be part of a scheduling round to switch to parallel
2460 execution mode (raw contexts only)
2461 * Fix bugs that prevented to use suspend/resume along with
2462 synchronization structures.
2463 * Fix bugs in process termination that lead to invalid memory access
2464 in very specific conditions.
2467 * Introduce a parallel mode for the models (controlled by surf/nthreads
2468 configuration item). In our tests, running the models in parallel
2469 never lead to any speedups because they are so fast that the gain
2470 of computing each model in parallel does not amortizes the
2471 synchronization costs, even when ultra fast futexes are used.
2472 This is released anyway because YMMV.
2475 * Performance boost by using a swag internally to compute the set of
2476 tasks that are finished and should constitute the return value of
2480 * Enable it by default now that it is considered rather stable.
2483 * Documentation of the tracing functions.
2484 * Performance gains when tracing categorized/uncategorized resource
2485 utilization by avoiding calls to get route when updating resource
2486 variables. LMM constraints are being used instead.
2487 * API changed to set task categories. Use MSG_task_set_category instead
2488 of TRACE_msg_set_task_category, and SD_task_set_category instead
2489 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2490 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2491 * Examples for link user variables added (see at examples/msg/tracing/)
2492 * Deprecated function TRACE_msg_set_process_category completely removed
2493 * Trace header updated according to the latest Paje file format
2494 * Tracing network lazy updates, no longer obligate users to use full updates
2495 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2496 * Experimental: let user code declare/set/push/pop application states for hosts
2497 * API changed to allow the manual creation of graph configuration files
2498 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2501 * Improve the API of Lua MSG bindings, using the Lua spirit.
2502 * Each simulated process now lives in its own Lua world (globals are
2503 automatically duplicated). It helps writing simulators. Will allow
2504 to run Splay programs within SimGrid in the future.
2505 * Add a Chord example in Lua, equivalent to the MSG one.
2508 * Start the implementation of a solution to express temporal
2509 properties, not only local assertions. This is still an
2510 experimental work in progress, stay clear from it to be safe.
2514 - Add new runtime parameters --help-logs and --help-log-categories
2515 to display information about supported logging parameters and
2517 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2518 don't exists anymore.
2519 * Mallocators: allow value NULL for the reset function.
2521 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2522 create homogeneous dictionaries, where all the elements share the
2523 same free function. Non homogeneous dictionaries will be
2524 deprecated in the next release.
2525 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2526 - Multi-level dictionaries are deprecated.
2528 - new function xbt_dynar_search_or_negative() that is useful when
2529 you have less than 2 million elements in your dynar and don't
2530 want of the extra complexity of catching exceptions when the
2531 element is not found.
2533 - Make xbt_os_thread module (for thread portability) public.
2534 Documentation is still to come, sorry.
2536 - Cleanups and simplifications to make it maintainable again.
2537 - Exotic features (such as memalign and valloc) were removed.
2538 - The metadata were extended and improved so that the
2539 model-checker becomes able to explore and inspect the heaps.
2540 - This may induce a performance drop when enable_model-checking is
2541 ON in cmake (even if it's not used in the simulation), but it is
2542 necessary at this point to get MC working.
2544 Turn model-checking OFF if simulation performance matters to you.
2545 Not enabling it at runtime is not enough, disable it in cmake.
2547 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2550 SimGrid (3.6.2) stable; urgency=low
2552 The "Not coding new stuff allows to polish old things" release.
2555 * New bindings to the NS3 packet level simulator (experimental)
2556 * Use the raw (efficient) execution contexts instead of the sysv
2557 (portable) ones when possible.
2558 * libpcre is now mandatory in any cases since not using it led to
2559 severe performance loss and possibly other issues
2560 * Update the XML platforms:
2561 - G5K: include the latest machine in Nancy
2562 - GridPP and LCG: new platforms
2563 * Documentation was partially updated, at least (more to come)
2565 Bug fixes, cosmetics and small improvements
2566 * Free terminated processes before the end of the simulation to avoid
2567 exhausting the memory of users having very dynamic amount of
2569 * Bug fix and cosmetics about canceling non-running tasks
2570 * Bug fix about the dot loader's issues when using libcgraph
2573 * Create an installer for windows with nsis (amd64 and win32)
2574 - Add an hello world project to illustrate simgrid project creation.
2575 - Embed libpcre into the Simgrid installer to avoid
2576 its compilation burden
2577 * The raw execution contexts should work on Apple now
2578 * Port to Windows 64 bits
2579 - Sysv contexts now have an implementation for this arch
2580 - GRAS communication features now support this arch
2581 * Drop support for borland compiler on windows
2582 - this code was not maintained, and we kinda depend on gcc nowadays
2583 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2584 * Fix portability issue on unstable ubuntu: linker became picky on
2587 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2589 SimGrid (3.6.1) stable; urgency=low
2591 The "Oops, we broke Macs too" release
2594 * Fixed contexts detection so that raw ones are used when possible
2595 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2596 leads to a strange error, with user code segfaulting sometimes when
2597 the generated code is not perfectly aligned (which is not
2598 controllable from the user side, depends on the amount of code)
2601 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2603 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2605 SimGrid-java (3.6) unstable; urgency=low
2608 * Split of every thing from simgrid v3.5 into a separate package.
2610 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2612 SimGrid (3.6) stable; urgency=medium
2614 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2617 * Bindings now constitute their own package, separated from the main one.
2618 Rationale: reduce our maintenance nightmare by reducing the module coupling
2619 They will soon be released on their own on gforge.
2621 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2622 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2624 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2625 * It's not quite deprecated for now because we have no replacement,
2626 but it may soon become the case.
2629 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2630 * Fortran: New user-level cache variable to store the rank of the running
2631 process. This improves performance by an order of magnitude.
2632 * C: New coccinelle script to automatically locate and modify global and
2633 local static variables.
2634 * Improved SMPI network model with a sender-side gap to account for multiple
2638 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2639 only say if a communication is finished, no matter whether it succeeded or
2640 failed. You can call MSG_comm_get_status() to know the status of a finished
2642 * New function MSG_task_dsend() to send a task and detach it. When a
2643 communication is detached, you are never notified of its success or failure
2644 and the memory is released automatically once it is finished. This function
2645 is useful when you don't care about the end nor the success of a
2647 * Change the prototypes of action replay. Sorry for inconvenience,
2648 but this is really more efficient this way (and to adapt your code,
2649 you just have to fix the initialization, that shouldn't be too long)
2650 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2651 ever managed to do anything useful with it.
2652 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2653 takes a dynar of communications. It returns immediately and gives the
2654 index of a finished communication (if any).
2655 * New example: a basic implementation of the Chord P2P algorithm.
2658 * New model for multi-core CPUs. You can now use the core attribute to
2659 precise the number of cores of a host. This is a basic model. Every
2660 process running on the host receives at most the power provided in
2661 the DTD (throughput<=power). Total throughput of process cannot exceed
2663 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2664 router linked by an up-link and a down-link (possibly asymmetrical).
2665 This kind of pattern allows to easily build last-mile model style platforms.
2666 Aggregating such patterns in a rule-based AS is thus the technique of
2667 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2668 * New model for Vivaldi routing. We transformed the Vivaldi network model
2669 into a Vivaldi routing model (based on the rule-based model). This allows to
2670 combine Vivaldi based latencies with last-mile platforms.
2673 * Added a check for NaN of IEEE754 infinite in the double entries of
2675 * Introduce a new context factory "raw", highly inspirated from the
2676 ucontext factory, but using manually crafted functions in assembly to
2677 do the work in an efficient manner.
2678 * Allow to change the used context factory at run time, not only at
2679 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2680 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2681 contexts at runtime (only with raw contexts or ucontexts).
2682 * Completely rewrote this module to allow parallel execution of user
2683 processes. Use --cfg=contexts/nthreads:N to execute user processes
2684 with N parallel threads (the default is 1, meaning no parallelism).
2685 * Allow to decide dynamically between sequential and parallel modes.
2686 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2687 processes in parallel only when their number is greater than or equal to P
2689 * Added a check for NaN of IEEE754 infinite in the double entries of
2693 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2694 won't output the state of processes when interrupted with Ctrl-C
2695 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2696 NULL-terminated array. This may solve backward compatibility issues
2697 due to the change to return type of SD_simulate. See also:
2698 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2699 * Add new macros with variable number of arguments.
2700 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2701 - in xbt/asserts.h: xbt_assert
2702 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2703 - in xbt/ex.h: THROWF and RETHROWF.
2704 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2706 * Change xbt_die() to accept a format string with arguments, just like printf.
2707 * New data structure: xbt_lib_t, like a dict but more general and with better
2711 * New configuration options
2712 Options triva/categorized and triva/uncategorized can be used to generate
2713 graph configuration files for Triva visualization tool.
2714 * Configuration option tracing/platform is renamed to tracing/categorized
2715 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2716 * New TRACE_user_link_variable interface:
2717 User provides the name of the link and the tracing variable to attach to it
2718 * the declaration of tracing categories must be done after the environment
2720 * simpler tracing interface, just one way to declare categories
2721 TRACE_category or TRACE_category_with_color, it is up to you
2722 * links in the trace file are again identified by their names
2723 * trace contains the full platform hierarchy exactly as declared using the ASes
2724 * Options tracing/msg/[task|process]:1 groups the process by hosts
2725 for both cases, tasks and processes must have names that are unique during the simulation
2726 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2727 * The experimental option tracing/msg/volume is deprecated
2728 its functionality may be reincorporated if needed
2730 The tracing generates a trace file with unordered timestamped events,
2731 because of the way the core simulator (surf) works. A script available
2732 at the tools directory (fix-paje-trace.sh) can be used to put the events
2733 in order. We have changed the tracing so it can generate ordered timestamped
2734 events in the final trace, but depending on the simulator (and how much time
2735 is simulated) that can lead to a huge memory utilization. It is deactivated
2736 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2738 Build Infrastructure
2739 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2740 - We are trying hard to keep the API stable, but it may happen that
2741 some things change (we're a research project after all, not a
2742 nuclear plant operating system). If such things should happen, you
2743 could rely on that macro to adapt.
2744 - current value: 30600 for 3.06.00, aka 3.6
2745 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2746 a number that can be compared with SIMGRID_VERSION.
2747 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2748 assertions and verbose/debug logging events are disabled at compile time.
2750 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2752 SimGrid (3.5) stable; urgency=medium
2755 * New feature to any SimGrid-based simulator: Model-Checking
2756 Check SIN#1 for more details.
2759 * New Model SMPI (three-interval linear regression for correction factors)
2760 See RR-7426, available at http://hal.inria.fr/inria-00527150
2761 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2762 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2763 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2764 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2765 * See also src/smpi/README
2770 - Tracing API changes: TRACE_start and TRACE_end should not be called
2771 by user-code. They are automatically called by simulators created
2772 with SimDAG, MSG and SMPI if the toolkit is compiled with
2773 tracing_enabled=ON. Categories declaration and utilization remain the
2774 same for MSG and SimDag.
2775 - A function was added to the tracing API to declare categories with
2777 - TRACE_category_with_color (char *category, char *color)
2778 where color must be in the following format
2779 "%f %f %f", red, green, blue
2780 and red, green, blue are float values in the interval [0, 1]
2781 - User can specify NULL as color parameter, or continue calling
2782 TRACE_category (cat)
2783 On that case, the tracing system will define random colors
2784 - The following command-line options are supported:
2785 --cfg=tracing/filename:msg.trace
2786 --cfg=tracing:1 (activate tracing, needed to use others)
2787 --cfg=tracing/platform:1 (categorized resource use)
2788 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2789 --cfg=tracing/msg/task:1 (task creation)
2790 --cfg=tracing/msg/process:1 (process creation, migration)
2791 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2792 --cfg=tracing/smpi:1 (SMPI interface tracing)
2793 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2794 - examples of examples/msg/tracing updated
2796 - DAXLoader and DOTLoader functions can generate tasks with categories
2797 - A new function to attribute a category to SD tasks:
2798 TRACE_sd_set_task_category (SD_task_t task, char *category)
2799 * Tracing the MPI interface implemented by SMPI
2800 - Collective operations are traced with states
2801 - Point-to-Point operations are traced with states/links
2802 - Tracing activated by a parameter "-trace filename" passed
2803 to smpirun during execution (considering that simgrid
2804 is compiled with tracing enabled)
2805 - To run the simulation with gdb, the simulator
2806 accepts --cfg=tracing/smpi:1 to trace SMPI
2807 - tesh files to check if smpi tracing is ok
2808 - See examples/smpi/NAS/DT-trace
2809 * GTNetS tracing re-worked
2810 - adaptation to the tracing system of GTNets to cope
2811 with modifications regarding the fullduplex mode
2812 - new tesh files to check if gtnets tracing is ok
2815 * Asynchronous communications through the functions:
2816 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2817 * New function: MSG_load_platform_script()
2818 to make possible using a lua script instead of XML files to set up platforms
2819 * New function: MSG_set_function
2820 to associate functions to processes, used when bypassing the parser
2821 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2823 Platforms: Add some more examples in examples/platforms
2824 * Grid'5000: see www.grid5000.fr
2825 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2828 * Change the XML format. This is a very important modification. SimGrid
2829 3.5 introduces a new hierarchical format based on the notion of
2830 Autonomous Systems. Compatibility with old format is ensured
2831 through the perl script provided in the install bin directory
2832 bin/simgrid_update_xml.
2833 It is now possible to build platforms with specific routing
2834 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2835 connect several platforms together. We will try to provide soon
2836 set of realistic platforms exploiting these properties (have a
2837 look at examples/platforms/ for the moment).
2838 * Take the opportunity of the XML format change to be a good XML citizen:
2839 rename link:ctn to link_ctn and similar changes (also dealt with by
2841 * Add a new routing scheme (rule-based) using regular expressions. It
2842 enables to have an extremely low memory footprint when the
2843 underlying routing is simple and can be compactly described. You
2844 need to have libpcre4-dev (perl regular expressions) installed if
2845 you want to use this routing scheme.
2846 * Revive the cluster TAG and allow to easily and efficiently (both in
2847 term of memory and speed) connect clusters together. Have a look
2848 at teshsuite/simdag/platforms/ to see how this can be done. With
2849 this tag, you can create clusters with thousands of tasks at no
2850 cost (have a look at examples/platforms/).
2851 Note: clusters are implemented as ASes, so there is no need for an
2852 enclosing AS tag if you have only one cluster in your platform.
2853 * Add new generic functions in the public interface that allows the user
2854 to call SURF 'create_resource' methods from your code (same
2855 functionality as the XML bypass mechanism but with a much lighter
2857 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2858 that uses a piecewise linear approximation to produce better
2859 results when exchanging small messages.
2860 * Add a new parameter to handle correctly full duplex link and account
2861 for interferences between uplink and downlink communications
2862 (activate with --cfg=fullduplex:1).
2865 * Rename the SD_READY (all dependencies are satisfied and task is
2866 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2867 dependencies are satisfied) state.
2868 This prevents a confusion between the notion of "ready to schedule"
2869 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2870 simulated" (SD_RUNNABLE) used by the simulation kernel.
2871 * Change the way a task is considered as ready. Instead of removing
2872 dependencies when a task is done, a counter is decreased. This way,
2873 it is always possible to reach ancestors thanks to the
2874 SD_taks_get_parents function (even after the end of the simulation.)
2875 * Change the return type of SD_Simulate from (SD_task_t*) into
2876 xbt_dynar_t. This function was in handling a dynar internally and
2877 converted it into a NULL terminated array for historical reasons.
2878 * New function SD_dotload(char*) to load a DAG described in dot
2879 format. This loader and the corresponding examples require the
2880 installation of the graphviz library.
2881 * Fix a bug in the management of tasks of size 0 in the surf network
2882 models. This problem was only visible with SIMDAG and you should
2883 thus disregard results produced with earlier versions if you
2884 relied on this feature (some tasks were blocked because of this).
2885 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2886 even though your applications were doing only point-to-point
2887 communications and sequential computations. Now you can really use any
2888 model you want (of course, if you create real parallel tasks, which are
2889 not implemented in most models beside ptaskL07, this will abort).
2890 * Add an example that schedules a DAX on an heterogeneous platform
2891 using a Min-Min strategy.
2892 * New function SD_workstation_get_current_task() that returns the kind
2893 of task currently running on a workstation in the sequential access
2895 * Raise some warnings when unexecuted tasks remains at the end of the
2896 simulation. This is usually caused by cycles in the DAG.
2899 * New function: SIMIX_process_set_function() called by MSG_set_function
2900 * Change the underlying waiting queue in semaphores so that a process
2901 can wait on several of them simultaneously (as in waitany).
2902 * Fix the way to handle tokens in semaphores so that all access patterns
2903 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2904 * kill the dirty pimple SIMIX_message_sizes_output()
2905 Please use (proper) visualization instead
2908 * New data container: setset (set of sets of elements)
2909 * New module: mmalloc (mapped malloc, allowing to have several
2910 independent segments of malloc)
2911 * New function: xbt_dict_cursor_set_data()
2912 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2913 * New function: xbt_dynar_is_empty()
2914 * New function: xbt_fifo_get_last_item()
2915 * Fix xbt_dynar_shrink(): use the right element size.
2916 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2917 uninitialized areas during expand.
2918 * Fix semaphores: previous implementation was severely broken.
2919 * Use library init/fini functions for our initialization.
2920 - you can use logs and other feature as soon as you want in your
2921 code (even before the xbt_init / MSG_init)
2922 - xbt_exit is now a no-op and produce a warning when used.
2925 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2926 benefit from the latest improvement to the simulation kernel.
2927 * Kill measurement sockets for now. If you rely on them, sorry. This
2928 release is not for you. This feature will be reintroduced in the
2929 future, but we cannot delay the release any further.
2930 * New function: gras_msgtype_get_name().
2931 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2932 * Fix (at last) the pmm example: it should not randomly fail anymore.
2934 Build chain: bug fixes and overall polishing
2935 * Cmake is now stable enough. Hence, we killed the autotools.
2936 * Port to windows ( TM :)
2937 * Fix the 'make install' target.
2938 No need to use 'make install-simgrid' anymore
2939 * Introduce a 'make dist' target compiling a *source* archive
2940 'make package' compiles a binary archive
2941 * Compile java files only on need
2942 * Add --cd and --setenv command line options to tesh
2943 * Out of source builds are not fully supported yet, but we are close
2944 * Enable supernovae and optimization flags by default for our users
2947 * Add layer to set up environment directly from lua, without XML.
2948 * The effect of gras_stub_generator can be achieved through
2949 lua too (check examples/gras/console/ping_generator.lua)
2951 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2953 SimGrid (3.4.1) stable; urgency=low
2955 The "Polishing easter eggs is probably a good idea" release.
2956 This is a bug fixes release only.
2960 * Fix a bug preventing the tasks from begin garbage collected.
2963 * Fix a bug occurring when a host involved in a communication fails.
2964 This was not detected properly by the other peer involved in the
2965 communication. Now, it's reported as a network error.
2968 * Warn the user about loop dependencies in data flow of DAX files
2969 * Obey the control-flow dependencies of DAX files
2972 * Add option "enable_smpi" allowing to not compile SMPI.
2973 Probably useful for the (Mac) users experiencing a build error here
2974 * Improve the detection of lua5.1 and ruby1.8
2976 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2978 SimGrid (3.4) stable; urgency=low
2980 The "Easter in Cargese" release. Also known as (major changes):
2982 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2983 ~> bindings were greatly improved
2984 ~> new tracing infrastructure for better visualization introduced
2986 * the "Welcome to configury modernity" release.
2987 ~> we switched from autotools to cmake, and improved our cdash
2990 A more detailled list of changes follow (full detail in svn log).
2992 Java Bindings: Various Cleanups
2993 * (install java-gcj-compat-dev on debian-like to use them)
2994 * Remove put/get: no need to export deprecated interface in Java
2995 Use send/receive instead.
2996 * Cleanup the examples and add a README per directory
2997 * Remove example autoDestination (that's the only way to go now)
2998 * Remove example explicitDestination (was a plain copy of basic)
2999 * Make JniException a runtime exception, so that there is no need to
3000 declare the fact that you may encounter such a beast. I guess that
3001 nobody will ever want to survive such error.
3002 * Create specific errors for each MSG case of failure:
3003 host failure, transfer failure, timeout, task cancelled
3004 * Cleanup the exceptions that may get thrown by each function
3005 * Other internal cleanups in Java bindings. Performance still bad :/
3006 Ruby and Lua Bindings: create them
3007 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
3008 * That's new and great, you should try them out.
3009 Same functionalities than Java bindings, only even less polished
3011 * Kill the useless "rate" argument of SD_task_get_execution_time()
3012 Everyone used to provide -1 as a value, it was not used, and the
3013 semantic of a possible use wasn't even clear.
3014 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
3015 either as comm costs or compute costs to mean that there is no
3016 such thing for that specific task.
3017 * Add a SD_task_set_name() function
3018 * Fix SD_task_unschedule() on typed tasks
3019 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
3020 * In DAX loader, accept useless 'level' attributes to <job> since
3021 LIGO DAGs have them (seem to be to ease graphical representation).
3023 * Add an example masterslave_mailbox.c using send/receive and not
3024 the deprecated put/get interface.
3025 * Kill the MSG_paje_output() function. It's a noop since 2 years.
3026 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
3028 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
3029 (declare MSG_USE_DEPRECATED to still have the old name)
3030 * Add a MSG_task_set_data() function
3031 * About trace replay (see examples/msg/actions):
3033 - Allow to work with splitted trace files for each process
3034 Give the specific trace file as argument of each process,
3035 and call MSG_action_trace_run(NULL)
3036 You can still have one merged file for all processes.
3037 - Fix implementation of collective operations
3038 * Allow task_execute() on 0-sized tasks (closes #10063)
3040 * This is the first release of SimGrid where SMPI is not considered
3041 beta anymore (even if some corners should still be improved)
3042 * Port over the new SIMIX_network submodule (internal refactoring)
3043 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
3044 * Implement more missing elements of the standard:
3046 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
3048 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
3049 MPI_Scan() MPI_Get_processor_name()
3050 - Added implementation of missing case for Alltoall (warning: it's
3051 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
3052 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
3053 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
3054 SMPI_MPI_Allgatherv()
3055 * Bug fixes include:
3056 - MPI_Waitsome() was broken
3057 - Allow relative includes in smpicc
3058 - Command line cfg argument 'reference_speed' was ignored...
3059 - Some functions did not properly lead to auto-benching of user code
3060 - smpicc passes -O2 by default (just like openmpi one)
3062 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
3063 * Bug fixes about timeouts during communications
3064 * add SIMIX_message_sizes_output() as a pimple to write to file the
3065 amount of messages per size. Use gnuplot to get histogram.
3066 Pimple because that's the only user-visible function of simix,
3067 defined directly in xbt.h (irk, sorry)
3069 - Add a SIMIX_sem_get_capacity() function
3070 - Fix interactions with processe resume/suspende
3071 - release_forever() was stupidly broken
3072 - Fix SIMIX_display_process_status() for processes in a semaphore
3073 - Make SIMIX_sem_block_onto() user-visible
3074 * Refactoring context stuff:
3075 - Use pseudo-OOP for better modularity
3076 - reimplement SIMIX_process_kill() without process_schedule() so
3077 that the latter can take as invariant that it is called from
3079 - Merge context_start into context_new for sake of simplicity
3081 * Add a Vivaldi network model, coded live during SUD'10 ;)
3082 * Rename configuration variables to start a hierarchy:
3083 o cpu_model -> cpu/model
3084 o network_model -> network/model
3085 o workstation_model -> workstation/model
3086 * New configuration variables:
3087 o network/bandwidth_factor: correction to bandwidth
3088 o network/latency_factor: correction to latency
3089 o netwotk/weight_S: correction to the weight of competing streams
3090 * Add a long description to the models, that users can see with such
3091 argument on the command line: --cfg=cpu/model:help
3092 * --help-models display the long description of all known models
3094 * config: add the ability to set a default value after registration
3095 Does not override any previously set value (e.g. from cmd line)
3096 * dict: allow to have integer key and data.
3097 When so, you need to use the following functions
3098 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
3099 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
3100 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
3101 In contrary to regular dicts, the key is not malloced before copy.
3102 Mixing scalar and regular elements in the same dict is not tested
3104 * Allow to use xbt_dynar_shrink() to expend the dynar instead
3105 Tracing for Visualization:
3106 * SimGrid is now instrumented in order to generate a trace file for
3107 visualization analysis: to use it, need to compile SimGrid with the
3108 "tracing" option enabled, and instrument the program using SimGrid with
3109 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
3110 (among other functions).
3111 * The instrumentation only traces the platform utilization for now
3112 * Documentation to use the tracing functions and how to analyze the
3113 traces with the Triva tool is written.
3114 * More information about: SimGrid FAQ (in the section Tracing Simulations
3117 * We moved to cmake as default build system. Autotools support will
3118 be dropped soon. Check the FAQ for more info about how to use it.
3119 * Greatly improved our cdash/ctest interactions
3120 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
3121 * Added memory checking tests with valgrind; lot of memleak fixing.
3122 This may be the first release of simgrid with so few memory issues
3123 * Added code coverage tests.
3124 Our coverage is still improvable, but at least we see it on cdash.
3126 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
3128 SimGrid (3.3.4) stable; urgency=low
3130 The "Desktop Grid needs love too" release (also called Xmas release).
3132 Models improvements:
3133 * Major speedup in the maxmin system solving by using lazy evaluation
3134 Instead of solving completely the maxmin system at each iteration,
3135 only invalidate (and recompute) the modified parts.
3136 This new feature is enabled in default models but you can try to
3137 turn it on with "--cfg:maxmin-selective-update=1" for other models.
3138 * Cas01 IMproved as default CPU model
3139 This CPU model is the same Cas01 model, but it uses the
3140 maxmin-selective-update flag and a heap structure to manage
3141 actions on SURF kernel.
3142 It reduces the complexity to find the next action to finish and,
3143 consequently, it's faster than the old Cas01.
3144 This is the new default CPU model (Cas01).
3145 * Rename the old Cas01 model to Cas01_fullupdate
3146 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
3147 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
3148 * CpuTI (CPU Trace Integration)
3149 A new CPU model whose objective is simulate faster when using
3150 availability trace files.
3151 Instead of using a full featured, over engineered maxmin system for
3152 CPU modeling, this model does the pre-integration of traces files
3153 to calculate the amount of CPU power available, and so, executes
3154 faster than the old CPU models.
3155 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
3156 * Use LV08 as default network model since it gives better accuracy
3157 for small messages and shouldn't change things for big ones.
3158 Use --cfg=network_model:CM02 to get the previous behavior.
3161 ******************************************
3162 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
3163 ******************************************
3164 * The new CPU model may changes simulations!
3165 The point is that events occurring at the exact same timestamp
3166 are not scheduled in the same order with the old and new
3167 version. This may be enough to completely change the execution
3168 of simulations in some cases.
3169 * The new network model will change simulations!
3170 This new model is more realistic than the previous one, so you
3171 should consider redoing your old experiments with this model.
3172 Sorry for the inconvenience.
3175 * Introduce the supernovae compilation mode
3176 When compiled that way, the whole SimGrid (or almost) is put in a
3177 single compilation unit and compiled in one shoot.
3178 This is to help gcc which has difficulties to inline stuff from one
3180 The speedup seem to be above 15%, althrough more tests are needed on
3181 amd64 to confirm that gain.
3184 * Port of MSG's mailbox on top of SIMIX network
3185 The put/get mechanism was greatly simplified on the way.
3188 * New SIMIX network module. Provides:
3189 - Mailbox: rendez-vous mechanism to find with who you want to speak
3190 - Synchronous send/recv: easier and hopefully faster since the
3191 logic is handled in the maestro process directly now
3192 - Asynchronous send/recv: you dreamt of it? It's here now
3193 Too bad that nobody cared enough to propagate the change to MSG.
3194 * Add semaphores as SIMIX synchronization mechanism.
3197 * new function SD_daxload(char*) to load a DAX file
3198 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3199 * Introduce typed tasks. Specify its kind and cost at creation.
3200 At scheduling, just give where it should be placed, and the cost
3201 for each involved resource is automatically computed.
3202 Existing constructors so far (more to come of course):
3203 - SD_task_create_comm_e2e() for end-to-end communication
3204 - SD_task_create_comp_seq() for sequential computation
3205 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3206 * new function SD_task_dump() for debugging display
3207 * new function SD_task_dotty(task,FILE*) writing to file the info
3208 about the task in dotty format
3209 * SD_task_dependency_exists() can now cope with having one of its
3210 arguments NULL. If so, it tests whether the other argument has any
3212 * Add getters on list of preceding/following tasks:
3213 SD_task_get_parents(task) and SD_task_get_children(task)
3214 * Add getters on amount of workstations and list:
3215 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3216 * Add getter on task kind: SD_task_get_kind(task)
3217 * Update the start_time and finish_time of tasks on completion/failure
3218 * Bugfix: Remove task from state swags when destroyed
3221 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3224 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3225 dynar with one separate thread per value of the dynar.
3226 * Change the prototype of xbt_thread_create(), sorry.
3227 Added a boolean parameter indicating whether we want to join this
3228 thread (used in SG only for now)
3229 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3232 * GTNetS wrappers should now be usable again (and betterly tested too)
3233 * Fix a major regression from 3.2 where the timeout provided to
3234 MSG_task_put_with_timeout() was used as absolute time before which
3235 the comm should be done.
3236 * Start to fix the <cluster> tag.
3237 - Internal links should be good now (beside of the loopback, which
3238 use the private link instead)
3239 - paths to the external world is still rather broken
3240 - the <route:multi> tag is just broken. Actually that's brain-dead.
3241 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3242 to make it less stupid
3243 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3244 * Fix a source-level compatibility glitch from 3.2: after defining
3245 MSG_USE_DEPRECATED, you can use the old name
3246 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3247 * Allow to compile from the SVN with automake 1.11
3248 * Fix some problems when using the "start_time" tag in deployment XMLs.
3249 * Fix #8569: XBT/synchro.h has redundant declarations
3250 * Fix #8563: MSG return values and exceptions
3251 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3252 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3253 Thanks to Robson Peixoto.
3254 * Fix a vicious bug in dictionaries inducing that some elements were
3255 not freed on xbt_dict_free()
3257 Portability report of this version:
3258 * Main portability targets:
3259 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3260 on (amd64/i386/ia64)
3261 - mac leopard on i386
3262 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3263 but nothing critical.
3264 * Other platforms: windows, AIX and others were not tested for this release
3266 Timing report of this version:
3267 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3268 scenario parameters). From 8h to a few seconds in desktop grid settings.
3269 * Supernovae brings about 25% speedup on i386.
3271 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3273 SimGrid (3.3.3) stable; urgency=low
3275 The "Need for Speed" release.
3277 The timings done to validate the 3.3.2 were faulty.
3278 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3280 The problem was a conversion from a manually handled vector to
3281 xbt_dynar_t on the critical path.
3282 xbt_dynar_foreach calls functions, inducing stack management crap.
3284 We inlined these functions and xbt_dynar_foreach is now breath taking.
3285 We also inlined xbt_swag_belong on the way.
3287 Here are some approximate speedup measurements (on master/slaves
3288 simulations lasting between 10s and 20s each):
3289 3.3.1 -> 3.3.2: about same performance
3290 3.3.2 -> 3.3.3: 40% speedup
3291 3.3.1 -> 3.3.3: 40% speedup
3292 3.3.1 with inline patch -> 3.3.3: 30% speedup
3294 Our reading is that the refactoring which occurred in 3.3.2 made us
3295 suffer much more from the xbt_dynar_foreach low performance, but
3296 once we solved this, this refactoring proved to be very performance
3297 effective. From the 40% speedup, somehow, 10% are due to the
3298 inlining and 30% to the refactoring.
3300 That's a pitty that gcc cannot inline functions placed in other files
3301 alone. We have to choose between:
3302 - break the encapsulation (by putting private data structures and
3303 accessors in headers files to help gcc)
3304 - live with low performance
3305 - switch to a decent compiler such as icc (not quite possible).
3307 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3309 SimGrid (3.3.2) stable; urgency=low
3311 The "Simplicity does not preceed complexity, but follows it" release.
3313 The main contributors of this release were (lexical order):
3314 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3317 * Extract the routing logic into its own object.
3318 (was dupplicated in network.c and workstation_LV07.c;
3319 Allows to implement other ways of storing that info)
3320 => kill now useless network_card concept
3321 - Use dynar to represent routes (instead of void** + int*)
3322 - kill link_set (use surf_network_model->resource_set instead)
3323 - Add a command-line option to choose the routing schema to use
3324 - Add three new models:
3325 * Floyd (shortest path computed at initialization)
3326 * Dijikstra (shortest path recomputed all the time)
3327 * Cached Dijikstra (shortest path computed on need)
3328 All these models where contributed by Silas De Munck, and are
3329 described in his ICCS09 paper.
3331 * Simplify model declaration
3332 (less redirections, less function to write when defining a model)
3333 - Factorize stuff between models:
3336 surf_model_resource_set(model)
3337 surf_model_resource_by_name(model, name)
3338 - Unify the types of models in s_surf_model_t (using an union)
3339 - Embeed fields of common_public directly into s_surf_model_t
3340 - Rename model methods:
3341 action_free ~> action_unref
3342 action_change_state ~> action_state_set
3343 action_get_state ~> action_state_get
3344 - Change model methods into functions:
3345 (model)->common_public->action_use ~> surf_action_ref
3347 * Implement a generic resource; use it as ancestor to specific ones
3348 (allows to kill duplicated code in models)
3349 Drawback: timer command don't need no name nor properties;
3350 workstation_CLM03 don't need no properties
3351 (but I guess we can live with those few bytes wasted)
3353 * Improve the action object model
3354 - implement a constructor avoiding dupplicated code about field
3355 initialization in generic_action part.
3357 * Kill the SDP model: it has an external dependency, is deprecated
3358 in flavor of modern lmm models, and didn't compile since a while
3361 * Relocation of the context module from XBT to SIMIX.
3362 (the context were decoupled from the simix processes, duplicating a lot of code)
3363 => a lot of code was factorized
3364 - less overhead is introduced during scheduling
3365 - simpler API for the context factory
3366 - the logic for process creation,destruction and manipulation was simplified
3367 * Simplification of the s_smx_process_t data structure.
3368 => accesing the simix level data associated to a process is faster now,
3369 and the code is a lot more readable.
3372 * Implement some more MPI primitives:
3373 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3374 -implementation: Bcast: flat or 2-ary tree (default),
3375 Barrier: 4-ary tree,
3377 Allreduce: Reduce then Bcast
3378 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3379 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3380 Alltoallv: flat tree, like ompi
3382 * Add support for optimized collectives (Bcast is now binomial by default)
3383 * Port smpirun and smpicc to OS X
3386 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3387 Shout out if you used it.
3390 * Display the list of still queued messages in SG mode when existing
3394 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3395 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3396 * Add xbt_str_from_file(FILE*)
3397 * Add xbt_dict_get_key achieving a linear reverse search
3398 * Remove the context module
3400 Portability report of this version:
3401 * Main portability targets:
3402 - Linux(debian)/x86/context
3403 - Linux(debian)/x86/pthreads
3404 - Linux(debian)/amd64/context
3405 - Linux(debian)/amd64/pthreads
3406 On these, we still have the eratic breakages of gras/pmm and
3407 amok/saturate_sg reported in previous version. We still think
3408 that the tests are the cause of the fault, not the tested code.
3410 - Mac OSX Leopard/x86/context
3411 Still false negative in tesh autotesting.
3412 Smpi still fails, but this time because readlink does not accept -f
3413 Everything seems to work properly beside of that.
3416 - AIX version 5.3 (only tested contexts this time)
3417 Smpi still fails there because mktemp is not installed.
3418 Everything seems to work properly beside of that.
3420 I managed to compile it for the first time, but several breakages.
3421 Won't delay the release for this exotic platform.
3423 * Windows: it's still lagging behind. If you want to help, please
3426 Timing report of this version:
3427 This version seem to be more than 5% faster than 3.3.1 (on linux
3428 64bits with contextes). The gain is less than expected, we are
3429 investigating this for next release.
3431 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3433 SimGrid (3.3.1) stable; urgency=low
3436 * Implement a --cfg-help to show existing configuration variables
3437 * Build chain do not require doxygen in maintainer mode
3440 * fix a bug on struct sizeof computation, which prevented the
3441 exchange of arrays of structs in some conditions
3442 - added a regression test about this in datadesc_usage
3443 * Allow the exchange of 0-long dynamic vectors.
3444 - for that, use -1 as indicator of dynamic size instead of 0
3445 - This implied to change any size from unsigned long to long,
3446 reducing a bit communication abilities, but I guess that with
3447 64bits being quite common, this is more than enough.
3448 - This also induce a protocol change, thus bumping network protocol
3449 version from 0 to 1 (if we have external users, we have to get
3450 clean on that point too ;)
3451 - added two regression tests about this in datadesc_usage
3452 * Be more verbose when propagating local exceptions
3453 This helps debugging.
3454 * Display the status of simulated processes when receiving SIGINT in
3458 * Allow to control the simulation from a trace file.
3459 New functions MSG_action_register() and MSG_action_trace_run()
3460 The first one allows to associate a function execution to each
3461 kind of action while the second one parses a trace file and
3462 triggers the corresponding actions within the system.
3463 For now, only a toy example is provided in examples/msg/actions
3464 * Add an example of process migration in examples/msg/migration
3465 * Fix a bug in task exchange which broke MSG_task_get_sender()
3466 Add a teshsuite regression test for that.
3467 [Bug: if MSG_task_get_sender() is called after sender exit,
3469 * Fix a bug which prevented suspend/resume to work properly
3470 * Display the status of simulated processes when receiving SIGINT
3471 This fixes a regression of v3.3. due to the introduction of SIMIX
3472 * Bug fixing in failure management:
3473 - trace could not start by a failure at time 0
3474 - failure during communications were not working
3477 * Add SIMIX_process_set_name() to change the name of the current
3478 process in the log messages.
3479 * Store smx_hosts in a dict since we only retrieve them by name
3480 * Move the configuration infrastructure to surf
3483 * Move the configuration infrastructure to surf
3486 * Massive internal cleanups:
3487 - Store internal structures on processes instead of hosts (allows
3488 to have more than one process per host, in addition of being more
3490 - Cleanup the initialization/finalization process
3491 - Kill a whole bunch of unneeded synchronization:
3492 processes run in exclusive manner within the simulator
3493 - Move queues from global tables to process data fields
3495 - now accept -platform and -hostfile arguments
3496 - Pass the right rank value to processes according to the hostfile
3497 * Compile the examples by default, and use them as regression tests
3498 * Implement MPI_Wtime()
3499 * Change the reference speed to a command line option
3502 * TCP_gamma can now be specified as command line option using
3503 --cfg=TCP_gamma:10000000.0
3504 * Change the --surf-path cmd line option into --cfg=path:
3507 * Also include strbuff from xbt.h public header
3508 * xbt_ex_display(): do not free the exception after displaying
3509 This allows to do more with the given exception afterward.
3510 Users should call xbt_ex_free() themselves.
3514 Portability report of this version:
3515 * Main portability targets:
3516 - Linux(debian)/x86/context
3517 - Linux(debian)/x86/pthreads
3518 - Linux(debian)/amd64/context
3519 - Linux(debian)/amd64/pthreads
3520 These targets fail about 1/10 of times on gras/pmm, but we believe
3521 that this is because of the test, not because of simgrid.
3522 amok/saturate_sg fails even more rarely, and the test may not be
3525 - Mac OSX Leopard/x86/context
3526 The test suite still spits tons of errors because some obscure
3527 force prevents us from removing the temporary directories
3528 arguing that they still contain some metadata I've never heard of.
3529 Smpi fails because seq is not installed.
3530 Everything seems to work properly beside of that.
3533 - AIX version 5.3 (both contexts and pthread)
3534 Smpi still fails there because mktemp is not installed.
3535 XML inclusions seems rosty on AIX.
3537 * Windows: it's still lagging behind. If you want to help, please
3540 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3542 SimGrid (3.3) stable; urgency=high
3546 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3547 [Malek Cherier & Mt]
3549 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3552 Until now, GRAS were using MSG as an interface to SURF. It was
3553 quite difficult because both interface have several differences
3554 (MSG channels vs GRAS sockets were the most notable point).
3556 This also opens the gate to SMPI (which should occur soon) and speed
3557 up simulations by to 40% (even if it were not the main goal).
3559 **************************************
3560 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3561 **************************************
3562 The point is that events occurring at the exact same timestamp are
3563 not scheduled in the same order with the old and new version. This
3564 may be enough to completely change the execution of simulations in
3565 some cases. Sorry for the inconvenience.
3567 * Cleanup and upgrade the XML format to push further scalability
3568 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3570 * Improve the testing infrastructure with tesh. Now a very large part of
3571 the code is tested not only by being run but also by checking that the
3572 output match an expected output [Mt].
3574 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3575 is really less memory-demanding, which should allow you to use
3576 larger files in SimGrid [AL].
3578 * Inform valgrind about our contextes, so that it becomes usable
3579 with the default (and more effecient) version of SimGrid
3580 [contributed by Sékou Diakite, many thanks]
3583 * Introduce a listener thread in charge of receiving incoming
3584 messages from the network. It allows to overlap communication and
3585 computation but most notably, it removes some stupid deadlocks due
3586 to the fact that so far, a process could not send and receive at
3587 the same time. This made most non trivial communication schema
3589 * Convert the PIDs from long int to int to match the MSG ones (and
3590 linux ones too) [Mt]
3591 * New function: gras_agent_spawn() to launch a new process on
3592 current host. Only working in simulation for now. [Mt]
3593 * New function: gras_os_hostport() returning a constant form (ie,
3594 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3597 * Make the backtrace of exceptions more human readable [Mt]
3598 * New module: xbt/str [Mt]
3599 a ton of string utility functions (split, join, printf to a newly
3600 allocated buffer, trim, etc)
3601 * New module: xbt/hash [Mt]
3602 SHA1 hashing algorithm (more to come if needed)
3603 * New module: xbt/synchro [Mt]
3604 synchronization tools (mutex and conditions) working the same way
3605 in simulation and in real life (mainly useful for GRAS, but not
3607 * New module: xbt/queue [Mt]
3608 classical producer/consumer synchronization scheme
3609 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3610 (using the classical functions will get serialized) [Mt]
3611 * Make dictionary internal table dynamic. No need to specify its size
3612 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3614 * Make sure the log channels are organized as a tree under windows
3615 (because of ANSI C compatibility issue, any channel were child of
3616 root directly) [Mt].
3619 * Cleaned many thing in surf and fixed a few bugs [AL].
3620 * Add a nice command line configuration mechanism to compose models [AL].
3621 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3622 the previous one (KCCFLN05). It relies on something that looks like
3623 a max-min sharing mechanism but cannot be written as such. A new solver
3624 was thus designed [AL].
3625 * Add a new solver to lmm. Based on Lagrange optimization and
3626 gradient-based descent, it enables to efficiently maximise systems s.a
3628 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3630 This solver enables to propose two new network models for TCP Reno and
3631 TCP Vegas based on Low's work. These models still need to be fully
3632 tested though [Pedro Velho].
3635 * Bug fix in SD_simulate. Now the time bound given as argument is
3637 * Use the new parallel task model (ptask_L07) as default.
3638 * Use the SURF command line configuration mechanism.
3639 * 0-size tasks (for synchronization) should now work.
3641 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3643 SimGrid (3.2) stable; urgency=high
3647 We still experience issues on this platform, but we believe that at
3648 least MSG is usable.
3650 GRAS API BREAKAGE (for simplification purpose, sorry):
3651 * the gras_msgtype_by_name is not used anymore. Instead of
3652 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3653 you can write (and must)
3654 gras_msg_send(toserver, "request", &request);
3655 - If you still want to pass a gras_msgtype_t to the function (to cache
3656 the type and avoid the lookup time), use the gras_msg_send_() variant.
3657 - Impacted functions:
3658 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3659 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3660 * The callbacks are now expected to return 0 when everything went well
3661 (just like the main() function)
3663 GRAS new features and improvements:
3664 * New module mechanism where user code can use per process globals [Mt]
3665 This is similar to gras_userdata_*() functions, but for libraries. It
3666 factorize some code developped over and over in the examples and AMOK.
3667 It has still to be documented and used (only amok/peermanagement is
3669 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3670 loss when they were small enough to fit into the buffer and sent quickly
3671 enough so that they can all get received in one shoot.
3672 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3673 if not found. Use the *_or_null() variant for the old semantic.
3674 * In gras_msg_handle, do not discard messages without callback.
3675 They are probably messages to be explicitly awaited later (ie, proofs of
3676 mis-synchronization in userland since they are sent before being awaited)
3677 No big deal usually.
3678 * gras_socket_meas_send/recv: semantic changed!
3679 The numerical arguments used to be (1) the total amount of data to send
3680 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3681 messages. This was need for the fool willing to send more than MAXINT
3682 bytes on quite fat pipes.
3685 * Do really rename the hostmanagement module to peermanagement. [Mt]
3686 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3687 but this is rather new and this was documented in the module
3688 documentation (poor excuses, I admit)
3689 * Bandwidth measurement semantic changed! This follows the changes to
3690 gras_socket_meas_send/recv explained above.
3693 * A sequential mode has been added to the workstations. When a workstation
3694 is in sequential mode, it can execute only one task, and the other tasks
3695 are waiting in a FIFO. [Christophe Thiery]
3698 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3699 model for SIMDAG. [Christophe Thiery]
3700 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3701 equal to 0, which caused some bad behaviors in
3702 saturated_constraint_set_update. I now use a threshold mechanism like in
3706 * When running manually src/testall, you select specific units [Mt]
3707 testall is the result of our cunit mechanism, and should replace all
3708 the scripty thingy around since bash don't run easily on billware.
3710 * A mallocator system has been added. [Christophe Thiery]
3711 Mallocators allow you to recycle your unused objects instead of freeing them
3712 and allocating new ones.
3714 Documentation update:
3715 * FAQ reworking + New FAQs:
3716 - "Valgrind spits tons of errors!" [Mt]
3717 - "How to repport bugs" [Mt]
3718 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3719 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3720 - Communication time measurement within MSG [AL]
3721 - I experience weird communication times when I change the latency [AL]
3722 * GRAS tutorial [Mt]
3724 - an introduction to the framework and to the used communication model
3725 - an initiatic tour introducing the most proheminent features:
3727 . Lesson 0: Installing GRAS
3728 . Lesson 1: Setting up your own project
3729 o Part 2: Message passing
3730 . Lesson 2: Exchanging simple messages
3731 . Lesson 3: Passing arguments to the processes (in SG)
3732 . Lesson 4: Attaching callbacks to messages
3733 . Lesson 5: Using globals in processes
3734 . Lesson 6: Logging information properly
3735 . Lesson 7: Using internal timers
3736 . Lesson 8: Handling errors through exceptions
3737 . Lesson 9: Exchanging simple data
3738 . Lesson 10: Remote Procedure Calling (RPC)
3739 . Lesson 11: Explicitly waiting for messages
3740 . Recapping of message passing features in GRAS
3741 - A HOWTO section containing:
3742 o HOWTO design a GRAS application
3743 More are due, of course. They will come latter. In the meanwhile, you can
3744 check the examples which are still here.
3746 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3748 SimGrid (3.1) stable; urgency=high
3752 There was a stack corruption somewhere, visible only when optimizing
3753 with these versions. [Vince]
3756 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3757 module that enabled to play with Directed Acyclic Graphs. It is built
3758 directly on top of SURF and provides an API rather close to the old
3759 SG. Some old codes using SG are currently under rewrite to check that
3760 all needful functions are provided. [Christophe Thiery]
3763 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3764 extension of the classical CLM03 model that gracefully handles
3765 failures. This is now the default model for MSG and GRAS. It doesn't
3766 handle parallel tasks yet though. [AL]
3767 * Bug fix: Weights were not correctly set in the network part.
3768 WARNING: This may have resulted in incorrect results with simulations
3769 where there are more than one flow on a given link. [AL]
3772 * After a (long ?) discussion on simgrid-devel, we have decided that the
3773 convention we had on units was stupid. That is why it has been decided
3774 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3775 WARNING: This means that all previous platform files will not work as
3776 such with this version! A warning is issued to ask users to update
3778 A conversion script can be found in the contrib module of the CVS, under
3779 the name contrib/platform_generation/surfxml_update.pl [MQ]
3782 * Bug fix: Processes were started in reverse order, wrt deployment file.
3783 WARNING: if your code relies on this bug, please fix it. [AL]
3784 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3785 being executed on two different locations. [AL]
3786 * Bug fix: Failures are now better supported thanks to Derrick's tests
3787 (there was many failure situations I hadn't thought of and that weren't
3788 correctly handled). [AL]
3789 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3793 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3794 Exception occurring on server-side are propagated back to client (!).
3796 API CHANGE: the callback changed their prototype. Change:
3797 int my_handler(gras_socket_t expeditor, void *payload_data) {
3799 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3800 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3802 * New! function: gras_msg_handleall to deal with all messages arriving
3803 within a given period.
3804 * New! function: gras_socket_server_range to get a server socket in a
3805 range of port numbers (ease to avoid port number conflicts) [MQ]
3806 * New! gras processes display their backtrace when they get a SIGUSR1
3807 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3808 Sweet to debug RL processes [MQ]
3812 - Do not force experiment sizes to be expressed in kb, or it becomes
3813 impossible to measure the latency this way (needs one byte-long tests)
3814 WARNING: this changes the amok_bw_* function semantic. [MQ]
3815 - Implements the link saturation stuff. [MQ]
3816 * Peer management module:
3817 New! module factorizing code that we wrote over and over [MQ].
3820 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3821 - Test units are placed directly into the library code, they get extracted
3822 automatically and placed into the src/testall binary.
3823 - Convert most of the XBT tests to this system.
3824 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3825 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3826 live in the same C code. This required to change a little bit the API
3827 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3828 * New module: sparse graph structure with basic algorithms (this is work
3829 in progress and the API is not considered to be frozen yet). [AL]
3830 * Display more information on backtraces: source line & function names are
3831 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3832 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3833 * Reworked a little bit some #include statements to load only required
3834 headers. Some user code that relied on SimGrid to include stdlib or
3835 stdio may need to include it by themselves. [AL]
3836 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3838 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3839 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3840 but I guess that almost nobody used those parts. [MQ]
3842 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3844 SimGrid (3.0.1) stable; urgency=low
3847 * Unfortunately, I had missed 5 misnamed functions:
3848 xbt_fifo_item_t xbt_fifo_newitem(void);
3849 void xbt_fifo_freeitem(xbt_fifo_item_t);
3850 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3851 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3852 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3853 They're now deprecated. Please use their new versions:
3854 xbt_fifo_item_t xbt_fifo_new_item(void);
3855 void xbt_fifo_free_item(xbt_fifo_item_t);
3856 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3857 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3858 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3860 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3861 * Documentation: xbt_log module unmercifully reworked [MQ]
3862 * Bugfix: there was a problem with the ending of contexts with
3863 the pthread backend. It caused some weird deadlock or behavior
3864 depending on the pthread implementation. [AL]
3865 * Bugfix: get the exceptions raised in the simulator repport where
3866 and why they come from when they are not caught in time [AL, MQ]
3869 * Bugfix: Do repport the error when two non-connected hosts try to
3870 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3873 * Add additionnal checkings on communications. Assert that two
3874 communicating hosts are connected by a set of links... [AL]
3877 * Add additionnal checkings on channel values in communication [AL]
3878 * New: MSG_task_get_source to see on which host a task was generated [HC]
3879 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3880 the number of tasks waiting to be received on channel and sent
3882 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3883 waits for the first task coming from a given host.. [AL]
3885 GRAS new functionnalities: [MQ]
3886 * Enhance the parsing macro to allow the size of multidimentional objects
3887 to be given thru annotations.
3888 * New example (and documentation): Matrix Multiplication a la RPC
3889 (as when I was young!) and fix a bunch of bugs found on the way.
3891 GRAS performance improvements: [MQ]
3893 * Reduce the amount of cbps creation/destruction by making it static to
3894 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3897 * Change libdata to a set so that we can search for stuff by ID (and thus
3898 reduce the insane amount of dict lookups)
3901 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3902 RL communication on top of SG since it's so uneffective.
3903 It may also be used for inter-thread communication in RL, one day.
3904 * Use gras_datadesc_copy() to exchange messages on top of SG
3906 - improve message exchange performance on top of SG
3907 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3908 * Don't exchange on the network the size of the used part of buffer,
3909 instead, specify the possible buffer size to read().
3911 - reduces the amount of read/write calls (one pair per exchange)
3912 - reduces the amount of exchanged data (the size)
3913 - allows to retrieve all arrived data on receiver side, if we don't need
3914 it right now (subsequent read will peek the buffer)
3915 - allows the receiver to proceed with the beginning of the stream before
3916 everything is arrived
3917 - make it possible to build an iov transport (using readv/writev)
3919 - take care of the data with non-stable storage (like stacked data),
3921 * If possible, TCP send uses vector I/O (when writev() is here)
3922 - Don't use it for receive since we send data sizes and data on the
3923 same stream, so we wouldn't be able to chain large amount of chunks
3924 before having to flush the stuff to read the size.
3925 * Rework the transport plugin mechanism to simplify it and reduce the
3926 amount of pointer dereferencement when searching for the right function
3929 * I guess that now, we do almost as few system calls as possible while
3930 doing as few data copy as possible.
3932 To improve it further, we could try to send all the sizes first and then
3933 all the data (to use iov on receiving size), but it's only a partial
3934 solution: when you have 2 dimensional data, the sizes of the second
3935 dimension is data of the first dimension, so you need 3 streams.
3937 I'm not sure the potential performance gains justify the coding burden.
3939 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3941 SimGrid (3.00) stable; urgency=high
3944 * New! Give the possibility to hijack the surf parser and thus bypass
3945 MSG_create_environment and MSG_launch_application. Have a look at
3946 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3948 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3950 SimGrid (2.96) unstable; urgency=low
3955 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3956 This deprecates the xbt_error_t mechanisms.
3957 It modifies (simplifies) all XBT and GRAS API.
3958 MSG API keeps unchanged (exceptions raised by XBT are caught from
3959 within MSG and masked with existing error handling facilities)
3962 * New! Add a FATPIPE model. [AL]
3963 * New! Add a parallel task model. [AL]
3964 * New! Add automatically a loopback interface (in the default
3965 network model) if none was precised.
3968 * Bugfix: MSG_process_resume now works with the current running process.
3970 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3971 * Modification of MSG_task_get_compute_duration. Once a task has been
3972 processed, the value returned by this function is now equal to 0. [AL]
3973 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3974 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3975 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3976 of a m_task. For now, MSG_task_cancel only works with computation
3978 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3979 of the processor (in Mflop/s) regardless of the current load on the
3981 * API Change: use proper naming convention for MSG_getClock and
3982 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3984 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3985 This function changes the priority of a computation task. This priority
3986 doesn't affect the transfer rate. A priority of 2 will make a task
3987 receive two times more cpu power than the other ones. This function
3988 has been added to suit the needs of Nguyen The Loc and hasn't been that
3989 much tested yet. So if it fails, please report it and send me your code.
3991 * API Change: removed all functions and types that were marked "deprecated"
3992 since many months. Renamed MSG_global_init_args to MSG_global_init.
3994 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3996 SimGrid (2.95) unstable; urgency=low
3999 * Steal some nice code to GNU pth to fix context detection and usage [AL]
4000 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
4001 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
4004 * Bugfix: Allow absolute paths for platform description files [MQ]
4005 * Bugfix: do free the variables after use. Leads to drastic performance
4007 * Implement max_duration (ie, timeouts) on resources [AL]
4010 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
4012 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
4013 the needed support (timeouts) [AL]
4016 * Implement measurement sockets. You can now get the bandwidth between two
4017 hosts thanks to AMOK (see below). [MQ]
4018 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
4019 dynar over the network (yeah) [MQ]
4020 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
4021 * simple/static token-ring example. [Alexandre Colucci and MQ]
4022 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
4023 in place before [MQ]
4026 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
4027 type parsing. "short" and "long" modifiers were also ignored; other
4028 modifier (such as reference level) are still ignored. [MQ]
4029 * Embeed the buffer size within the buffer itself on SG. [MQ]
4030 That way, send() are atomic and cannot get intermixed anymore (at least
4031 the ones which are less than 100k; bigger messages still have the issue)
4032 * Array size pushed by the field, not by the field type (or each
4033 and every long int will push stuff to the cbps) [MQ]
4034 * use select() to sleep since it allows to portably sleep less than one
4037 GRAS (minor cleanups)
4038 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
4041 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
4042 * type category 'ignored' killed as it was never used and were difficult
4044 * whether a type can cycle or not is now a flag, leaving room for more
4045 flags (as "ignored", if we feel the need one day ;)
4046 * Rename raw sockets to measurement sockets since "raw" has another
4047 meaning in networking community.
4050 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
4051 and offers features not belonging to GRAS but that most applications
4052 need. One day, it may be a set of plugins loadable at runtime.
4053 * New module: bandwidth
4054 bandwidth measurement between arbitrary nodes running this module. [MQ]
4056 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
4058 SimGrid (2.94) unstable; urgency=low
4060 The first beta release of SimGrid 3 !
4064 * Update the main page and the FAQ. Adding references to gforge.
4067 * Add a gras_os_getpid function.
4070 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
4071 * Extend the logs so that they also print PID, hostname, date, ... if
4073 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
4074 and kill the old version which were in testsuite/
4075 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
4079 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
4080 for users. Instead, each programming environment (GRAS, MSG,...) use it
4081 when needed to provide such a feature to users.
4082 Don't shortcut the mechanism or you will also shortcut the virtualization
4083 you need on the simulator.
4087 * Cleanups in configury with regard to compile optimization/warning flags.
4088 Also add -fno-loop-optimize to any powerpc since it's the optimization
4089 killing gcc (< 3.4.0).
4090 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
4091 needed by MSG examples complications
4092 * Borrow configury beautifications from PHP
4095 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
4096 warning (thanks loris for stumbling into this one).
4097 * Bugfix: stop loading private headers (gras_config.h) from the public
4101 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
4102 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
4103 RL would imply a DNS resolver, which is *hard* to do in a portable way
4104 (and therefore delayed).
4105 * Implement a real timer mechanism and use it in timing macros. This allows
4106 to avoid rounding errors and get a 0.000005 sec precision in timing
4107 macros. While I was at it, various cleanups:
4108 - allow to declare more than one timed section per file (fix a stupid bug)
4109 - move some private declaration to the right place
4110 - merge conditional execution and timing macros into emulation module
4111 - document the module
4112 - make sure the module cleanups its mess on gras_exit
4113 * Documentation improvements:
4114 - (new) how to compile applications using GRAS
4115 - (new) emulation support (timing macros)
4117 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
4119 SimGrid (2.93) unstable; urgency=low
4121 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
4124 - Use Paje properly where used. Still to be sanitized properly.
4125 - Portability fix: Add an implementation of the contexts using pthread
4129 - Add xbt_procname(): returns the name of the current process.
4130 Use it to show the current process's name in all logging.
4132 - fix detection of older flex version and the reaction, since we do
4133 depend on modern ones (we use lex_destroy)
4134 - Better separation of SG and RL in the libs: remove all simulation code
4135 from libgras. As a result, this lib is now only 200k when stripped.
4136 Some of the xbt modules may also be duplicated (two sets and such) and
4137 should be cleaned/killed before SG3.
4138 - Insist on using xlC on AIX because of weird problems involving gcc there.
4139 - Cleanup the make remote stuff. This is now done by scripts
4140 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
4141 mainly for our private use, but we're working on changing them to user
4144 - Bugfix: flush the socket on close only if there is some *output*.
4145 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
4147 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
4150 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
4152 SimGrid (2.92) unstable; urgency=low
4154 Alpha 3 on the path to SimGrid 3
4158 - New! First try of benchmarking macros.
4159 - New! First try so that gras_stub_generator generate deployment and
4160 remote compilation helpers.
4162 - Bugfix: Initialization fix in msg_test.
4166 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
4168 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
4170 - API Change: gras_os_sleep to take the amount of seconds as a double.
4171 Accepting an int was error prone since it was the only location where
4172 seconds were coded as such. It leaded to damn rounding errors.
4173 - Bugfix: Hard to belive that timers ever worked before this.
4175 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4177 SimGrid (2.91) unstable; urgency=low
4179 Alpha 2 on the path to SimGrid 3
4183 - Bug fix in the lmm_solver.
4185 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4186 through the function MSG_paje_output.
4187 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4188 - It is possible to bound the rate of a communication in MSG with
4189 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4190 to write it in the changelog).
4191 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4196 - Various cleanups to the autotools stuff
4197 - Begin to move Gras examples to examples/gras/
4198 - Let make distcheck work again (yeah!)
4200 - documentation overhauled using doxygen.
4201 gtk-doc-tools is dead in SimGrid now.
4202 - Automatically extract all existing logging categories, and add the list
4203 to the documentation (long standing one, to say the less)
4205 - Cleanup the known architecture table. Reorder the entries to group what
4206 should be, and use a more consistent naming scheme.
4207 (some of the test dataset are still to be regenerated)
4208 - New! Allow library to register globals on each process just as userdata
4210 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4211 lookup time (for now).
4212 Use it in msg and trp.
4213 This cleans a lot the internals and helps enforcing privacy of the
4214 headers between the gras components.
4215 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4216 - Bugfix: gras_os_time was delirious in RL.
4217 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4219 - Reenable GRAS now that it works.
4221 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4223 SimGrid (2.90) unstable; urgency=low
4225 Alpha 1 on the path to SimGrid 3
4227 * It is a long time since the last release of SimGrid. I'm sorry about
4228 that but as I had told you, I was rewriting a lot of things. I apologize
4229 to those who had been reporting bugs to me and that I had not answered.
4230 If your bug is still in the new version, please tell me. Here is a
4231 summary of the main changes.
4233 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4234 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4235 portability, speed, and a lot more but you'll figure it out later.
4236 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4237 I wanted to release the new MSG as soon as possible and I have
4238 broken GRAS, which is the reason why, for now, only MSG is fully
4239 functional. A laconic description of these projects is available
4240 in the documentation.
4242 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4243 kernel optimized for our needs (called SURF but only the developers
4244 should use it). Hence, MSG is now roughly 30 times faster and I think
4245 that by rewriting a little bit MSG, I could event speed it up a little
4246 bit more. Beside the gain in speed, it is also much easier to encode a
4247 new platform model with SURF than it was with SG. More to come...
4249 * REVOLUTION 3: I have tried to change a little as possible the API of
4250 MSG but a few things really had to disappear. The main differences
4251 with the previous version are:
4252 1) no more m_links_t and the corresponding functions. Platforms are
4253 directly read from an XML description and cannot be hard-coded
4254 anymore. The same format is used for application deployment
4255 description. The new format is described in the documentation.
4256 Have a look in tools/platform_generation. There is a tiny script
4257 that converts from the old platform format to the new one. Concerning
4258 the application deployment format, parsing the old one is tricky.
4259 I think most of you should however be able to convert your files. If
4260 it is really an issue, I can write a C code that does the conversion.
4262 2) the toolbox tbx does not exist anymore. We now have a library
4263 with much more data-structures but without the hash-tables (we have
4264 dictionaries that are much faster).
4266 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4268 *****************************************************************************
4269 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4270 * don't feel like converting the oldies. (Mt) *
4271 *****************************************************************************
4274 Version 2.90: "the long awaited one"
4275 - Finished rewriting and debugging MSG. Rewrote the documentation.
4276 - disable GRAS for now since it needs to be ported to the newest SG
4279 - Finish the port to windows (using mingw32 for cross-compile)
4282 - Main loop and datastructures of SURF. A cpu resource object is
4283 functional. Surf can thus be used to create cpu's with variable
4284 performance on which you can execute some actions.
4286 2004-11-15 Martin Quinson
4287 - Port to ARM. Simply added the alignment and size descriptions. Should
4288 work, but the ARM machines are so slow that I didn't had the opportunity
4289 to 'make check' over there yet.
4291 2004-11-15 Arnaud Legrand
4292 - Trace manager now written. It uses a heap structure and is therefore
4293 expected to be efficient. It may however be speeded up (particularly
4294 when many events occur at the same date) by using red and black
4295 trees. One day maybe...
4296 - Max-min linear system solver written. It uses a sparse matrix
4297 structure taking advantage of its expected use. Most operations are
4298 O(1) and free/calloc are called as few as possible. The computation of
4299 the minimum could however be improved by using a red and black tree
4302 2004-11-03 Arnaud Legrand
4303 - Rename every gras_* function that was in xbt/ to its xbt_
4305 - Add a heap and a doubly-linked list to xbt
4306 - Added a dichotomy to the dictionaries. make check works as well before
4307 so I assume that the patch is correct. I do not know however if things
4308 run effectively faster than before now. :)
4310 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4311 SimGrid, and the version number is bumped to 2.x
4313 2004-10-29 Martin Quinson
4314 - Introduction of the remote errors.
4315 They are the result of a RMI/RPC on the remote machine.
4316 ErrCodes being scalar values, you can't get the host on which those
4317 errors did happen. Extending the error mechanism as in Gnome is possible.
4318 No idea yet whether it is a good idea.
4320 2004-10-28 Martin Quinson
4321 - Interface revolution: the Starred Structure Eradication.
4322 I used to do typedef struct {} toto_t; and then handle *toto_t.
4323 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4324 - ???_t is a valid type (builded with typedef)
4325 - s_toto_t is a structure (access to fields with .)
4326 - s_toto is a structure needing 'struct' keyword to be used
4327 - e_toto_t is an enum
4328 - toto_t is an 'object' (struct*)
4330 typedef struct s_toto {} s_toto_t, *toto_t;
4331 typedef enum {} e_toto_t;
4332 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4333 s_toto_t) is private.
4335 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4336 so that it can set the variable to NULL. It was so for dicts and sets,
4337 it changed for dynars.
4339 - Fix a bunch of memleaks in dict_remove
4340 - Fix a bug in sg/server_socket opening: it failed all the time.
4342 2004-10-07 Martin Quinson
4343 - Speed up dynar lookup operation a bit.
4345 gras_dynar_get is dead.
4347 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4348 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4349 (which gives you the address of the stored data).
4351 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4352 copy of the data using an affectation to do the job and not a memcpy.
4354 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4355 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4357 It does not really speedup the dynar test because they are
4358 setting elements all the time (and look them seldom). But the dict does
4359 far more lookup than setting.
4361 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4363 2004-10-05 Martin Quinson
4364 - Allow to (en/dis)able the cycle detection at run time.
4366 Whether we should check for cycle or not is now a property of each
4367 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4368 datadesc_cycle_unset allow to remove this property when previously set.
4370 Note that the cycle detection is off by default since it impacts the
4371 performance. Watch the data you feed GRAS with ;)
4373 This property is hereditary. Any element embedded in a structure having it
4374 set have it set for the time of this data exchange.
4376 You should set it both on sender and receiver side. If you don't set it on
4377 sender side, it will enter an endless loop. If you forget on receiver
4378 side, the cycles won't be recreated after communication.
4380 - Header reorganization.
4381 Kill gras_private.h, each submodule must load the headers it needs.
4383 2004-10-04 Martin Quinson
4384 - Interface revolution: do not try to survive to malloc failure.
4386 Now, gras_malloc and friends call gras_abort() on failure.
4387 As a conclusion, malloc_error is not a valid error anymore, and all
4388 functions for which it was the only gras_error_t return value are
4389 changed. They now return void, or there result directly.
4390 This simplify the API a lot.
4392 2004-09-29 Martin Quinson
4393 - Re-enable raw sockets.
4394 Created by gras_socket_{client,server}_ext;
4395 Used with gras_raw_{send,recv}
4398 It should allow to kill the last bits of gras first version soon.
4400 This is not completely satisfactory yet (duplicate code with
4401 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4404 - Simplify transport plugin (internal) interface by not passing any
4405 argument to _server and _client, but embedding them in the socket
4408 2004-09-28 Martin Quinson
4409 - Finish the port to AIX.
4410 autoconf was my problem (segfault within the malloc replacement
4411 function. No idea why)
4413 2004-09-16 Martin Quinson
4414 - Fix some size_t madness on 64bit architectures.
4416 2004-09-08 Martin Quinson
4417 - Reduce the number of system headers loaded, overload some more system
4418 calls (such as malloc to cast the result of the system one, and work
4420 - Fix and reintroduce the config support
4422 2004-09-07 Martin Quinson
4423 - Source code reorganization to allow Arnaud to surf all over there.
4424 - Allow to document the logging categories.
4425 - Remove all uppercase from logging categories and useless cleanup in names.
4427 2004-08-18 Martin Quinson
4428 Version 0.6.2 (protocol not changed; API changed)
4429 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4430 gras_error_t), and NULL when not found. Functions expecting a msgtype
4431 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4432 hopefully usefull message.
4433 - Portability to prehistoric sparcs again
4435 2004-08-17 Martin Quinson
4436 Version 0.6.1 (protocol not changed; ABI not changed)
4437 - prealloc some buffers to speed things up
4439 2004-08-11 Martin Quinson
4440 Version 0.6 (protocol not changed; ABI expended)
4441 - The parsing macro can deal with the references, provided that you add
4442 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4444 2004-08-09 Martin Quinson
4445 Version 0.5 (protocol not changed; ABI changed)
4446 - Allow to off turn the cycle detection code in data exchange at
4447 compilation time. It should be at run time, but I'm short of time (and
4448 the config stuff is still broken). That way, we keep dict out of the
4449 critical path, which is good because the performance is poor:
4450 - search not dichotomial yet
4451 - dynar give no way to access their content and memcpy everytime
4452 - In composed data description (struct, ref and so on), stop foolness of
4453 keeping the subtype's ID, but store the type itself. This keeps sets out
4454 of the critical path, which is good since they rely on dynar and
4455 dictionaries. The only loose of that is that we cannot detect the
4456 redeclaration of a structure/union with another content (but I'm not sure
4457 the code detected well this error before anyway). We still can detect
4458 the redefinition discrepancy for the other types.
4459 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4460 it breaks the code because of type-punning used all over the place).
4461 This breaks on all non-gcc architectures (for now).
4463 All those changes (plus the buffer of last time) allow me to gain 2 order
4464 of magnitude on cruel tests consisting of 800000 array of integers on two
4465 level of a hierarchical structure (200 secondes -> 4 secondes)
4468 - the selector of reference must now return the type it points to, not
4469 the ID of this type.
4471 2004-08-06 Martin Quinson
4472 Version 0.4 (protocol changed; ABI not changed)
4473 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4474 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4475 - Implement a buffer transport, and use it by default (it relies on tcp in
4476 real life and on sg in simulation).
4477 That's a bit hackish since I had a new field to the structure to store
4478 its data without interfering with the subtype ones. Inheritance
4479 is tricky in C. And that's a kind of reverse inheritance with one class
4480 derivating two classes. Or maybe a game with java interfaces. Anyway,
4481 that's damn hard in C (at least).
4482 Moreover, I got tired while trying to ensure plugin separation and
4483 genericity in SG mode. MSG wants me to do weird things, so let's go for
4484 cruel hacks (temporarily of course ;).
4485 See comment in transport_private.h:71
4486 - do not include all the _interface headers in private but in the files
4487 which really need them (to cut the compilation time when they are
4490 2004-07-26 Martin Quinson
4491 Version 0.3 (protocol not changed; ABI changed)
4492 - Major overhault of the datadesc interface to simplify it:
4493 - shorted the function names:
4494 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4495 - add a trivial way to push/pop integers into the cbps without malloc.
4496 This allows to make really generic sub_type description, which simply
4497 pop their size of the stack.
4498 - add a function gras_datadesc_ref_pop_arr() which does what users want
4499 most of the time: Declare a dynamic array (which pops its size of the
4500 stack) and declare a reference to it. Poor name, but anyway.
4501 - kill the post-send callback, add a post-receive one
4503 2004-07-23 Martin Quinson
4504 Version 0.2 (protocol changed; ABI changed)
4505 - add some testing for cpbs in the test cases, and fix some more bugs.
4506 This invalidate again the little64 data file, since I cannot regenerate
4508 - remove an awfull optimization in the logging stuff, allowing me to:
4509 - understand it again
4510 - learn gcc how to check that the argument match the provided format
4511 - fix all errors revealed by gcc after that
4512 - internal keys of dict are not \0 terminated. Deal with it properly in
4513 loggings instead of segfaulting when the user want to see the logs :-/
4515 2004-07-22 Martin Quinson
4516 - Fix some stupid bug preventing cbps (callback postit) from working
4518 2004-07-21 Martin Quinson
4519 - Some documentation cleanups
4520 - remove the useless last argument of msgtype_declare
4521 - rename the Virtu functions to fit into the 'os' namespace
4522 - move headers src/include -> src/include/gras/ and stop fooling with
4523 gras -> . symbolic link
4524 - make distcheck is now successful
4526 2004-07-19 Martin Quinson
4528 - Build shared library also
4529 - Install html doc to the right location
4530 - stop removing maintainer files in make clean
4531 - build tests only on make check
4533 2004-07-13 Martin Quinson
4535 - No major issue in previous version => change versioning schema
4536 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4537 corresponding dataset.
4539 2004-07-11 Martin Quinson
4541 - Get it working with any kind of structure (we can compute the padding
4542 bytes remotely for all the architectures I have access to)
4543 - Implement the structure parsing macro (still not quite robust/complete)
4544 - Improvement to the remote testing toysuite
4546 2004-07-10 Martin Quinson
4547 [autoconf mechanism]
4548 - get ride of a bunch of deprecated macros
4549 - actually run the test for two-compliment, not only compile it :-/
4550 - test whether the structures get packed (and bail out if yes. Damn.
4551 Alignment is a serious matter)
4552 - test whether the structures get compacted (but respecting the alignment
4553 constraints of each types)
4554 - test whether the array fields of structures can straddle alignment boundaries
4556 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4557 (took me 2 hours to find that bug, looking at the wrong place)
4558 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4559 before sending/receiving and are used to compute the offsets of fields
4560 - Given that padding size depend even on compiler options, keep track of
4561 alignment and aligned_size only for the current architecture. Not a big
4562 deal since we send structure fields one after the other (seems
4564 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4565 program, let it work on linux/gcc/little32. portability todo.
4567 2004-07-08 Martin Quinson
4568 - import and improve remote compilation support from FAST
4569 - make sure make check works on half a dozen of machines out there
4571 2004-07-07 Martin Quinson
4572 Let's say it's version 0.0.3 ;)
4573 - Implement conversions (yuhu!)
4574 - Let it work on solaris (beside conversion, of course)
4575 - Stupid me, using rand() to generate the conversion datatests in not wise.
4577 2004-07-06 Martin Quinson
4578 - Let make dist work, since I'm gonna need it to compile on remote hosts
4579 - Let Tests/datadesc_usage write the architecture on which the file was
4580 generated as first byte.
4581 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4582 architecture descriptions.
4583 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4584 on those architectures.
4585 - Optimization: send/recv array of scalar in one shoot
4587 2004-07-05 Martin Quinson
4588 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4590 - Plug a whole bunch of memleaks
4591 - each process now have to call gras_{init,exit}. One day, their log
4592 settings will be separated
4593 - Continue the code factorisation between SG, RL and common in Transport.
4595 2004-07-04 Martin Quinson
4597 - Redistribution between SG and RL.
4598 We wanna have to accept in SG, so move accepted related parts of RL in
4599 the common part. (more precisely, the dynar of all known sockets is no
4600 more a static in transport.c, but part of the process_data)
4602 [gras_stub_generator]
4603 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4605 2004-07-03 Martin Quinson
4606 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4607 them all up in one shot)
4609 - Fix the ugly case of reference to dynamic array.
4610 - New (semi-public) function gras_datadesc_size to allow the messaging
4611 layer to malloc the needed space for the buffer.
4613 - gras_socket_close now expect the socket to close (and not its address to
4614 put NULL in it after it). This is because the socket passed to handlers
4615 is one of their argument (=> not writable).
4617 - propagate the interface cleanup from last week in datadesc, ie remove a
4618 superfluous level of indirection. User pass adress of variable
4619 containing data (both when sending and receiving), and not of a variable
4620 being a pointer to the data. Let's say that I like it better ;)
4621 The price for that is constructs like "int msg=*(int*)payload" in
4622 handlers, but it's a fine price, IMHO.
4624 - Let it work in RL (yuhu)
4626 2004-06-21 Martin Quinson
4628 - porting SG plugin and SG select to new standards (works almost).
4629 - plug memleaks and fix bugs around.
4632 - cleanup the prototype of data recv and force users to specify when they
4633 want to handle references to objects. Test case working even for cycles.
4634 - plug memleaks. Valgrind is perfectly ok with this.
4636 2004-06-12 Martin Quinson
4638 - cleanup the separation between plugin and main code in plugin creation
4640 2004-06-11 Martin Quinson
4642 - Reput hook for raw sockets, needed for BW experiments
4643 - kill a few lines of dead code
4644 [Data description] Interface cleanup
4645 - gras_datadesc_by_name returns the searched type or NULL.
4646 That way, no variable is needed to use a type desc once, which makes
4648 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4649 two parameters were strings (field name, type name), leading to
4651 [Dicos] Interface cleanup
4652 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4653 This is consistant with the dynar API.
4655 2004-04-21 Martin Quinson
4657 - Porting to new standards.
4659 - interface cleanup.
4660 There is no bag anymore, no need to take extra provision to mask the
4661 pointers behind "ID".
4662 Better splitup of functions between files create/exchange/convert.
4663 This is still a bit artificial since convert and receive are so
4664 interleaved, but anyway.
4666 - add a queued message list to procdata (the ones not matching criteria
4668 - factorize some more code between SG and RL wrt procdata
4670 - use gras_exit in example to track memleaks
4671 - get rid of gs_example now that GS is properly integrated into gras
4672 - update run_test to integrate the latest tests (datadesc)
4674 - rename WARNINGn macros to WARNn since it prooved error-prone
4676 2004-04-19 Martin Quinson
4678 - register init/exit functions within gras module mechanism
4679 - send/receive function.
4680 Convertion is not implemented, but short-cutted if not needed.
4681 struct/array elements are sent one by one (instead of block-wise), but
4682 nobody really cares (yet). Get a prototype before optimizing.
4683 - tests (using a file socket) for DD send/receive on:
4684 - base types: int, float
4685 - array: fixed size, string (ie ref to dynamic string)
4686 - structure: homogeneous, heterogeneous
4687 - chained list, graph with cycle
4688 Believe it or not, valgrind is not too unhappy with the results. The
4689 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4693 - Bugfix when using a filename explicitly (instead of '-')
4695 2004-04-09 Martin Quinson
4697 - factorize more code between RL and SG in socket creation
4698 - Complete the implementation and tests of:
4700 o file (only in RL, and mainly for debugging)
4702 I lost 3 days to design a portable address resolver, and then decided
4703 that the prototype mainly have to run on my box.
4704 Addressing portability too early may be like optimizing too early :-/
4706 - use gras_init in the Tests instead of the crappy parse_log_opt
4707 (the latter function is removed)
4708 [Conditional execution]
4709 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4710 [Code reorganisation]
4711 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4712 Build examples against the RL library, since there is no way to disable
4713 its creation for now.
4715 For information, the beginning of coding on GRAS was back in june
4716 2003. I guess that every line has been rewritten at least twice since