1 ----------------------------------------------------------------------------
3 SimGrid (3.28.1) NOT RELEASED YET (v3.29 expected September 22. 2021, 19:21 UTC)
6 - Non-linear resource sharing for decay models:
7 - The total capacity may depend on the number of concurrent usages
8 - For that, resources can take a callback that computes the capacity
9 depending on the idle capacity and the number of concurrent usages
10 - Examples (both cpp and python): io-degradation, network-nonlinear, exec-cpu-nonlinear
12 - Dynamic factors for CPU and disk: similarly to dynamic network factors,
13 allows the user to set a callback which can affect the progress of activities
14 (multiplicative factor applied when updating the amount of work remaining).
15 - Example: examples/cpp/exec-cpu-factors
18 - Added support to programmatic platform creation in Python.
19 Example: examples/python/clusters-multicpu
22 - Disk and Host now have a set_sharing_policy() too, for non-linear sharing.
23 This can only be set through the API, not through XML files.
27 - Multiple fixes to ensure reproducibility of tracing
28 - scan/excan can now be replayed
29 - wait action now uses ranks and not pid, as the other ones.
30 - smpi/init and smpi/finalization-barrier are now valid for replays.
31 - exit() is now intercepted by SMPI to avoid premature shutdown of
32 simulation. First non 0 return codes is returned as simulation return
36 * New section "Release Notes" documenting recent and current developments.
37 * New section "Modeling I/O: the realistic way" presenting how to properly model disks in SimGrid.
38 * Improvements in API Reference for C++ and Python interfaces.
41 - Make wifi creation compatible with ns-3 version 3.34 too.
43 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
44 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
45 - FG#78: Multiple fixes for SMPI replay:
46 - TI tracing of allotallv/w was outputting wrong values
47 - MPI_LOGICAL in fortran is actually 32 bits wide, and not 8.
48 - FG#77: Search feature of doc is broken (update sphinx theme version)
49 ----------------------------------------------------------------------------
51 SimGrid (3.28) July 14. 2021
53 The Victoriadagarna Release.
56 - C++ platform interface: Users can now describe their platform directly in C++.
57 This provides greatly flexibility and performance improvement for complex
58 platforms. Main features:
59 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
60 "zones", instead of single hosts. This feature enables the description
61 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
62 - StarZone: new zone with a Star-like topology. The routes are defined
63 as a set of links used to communicate from node to everybody (node<->ALL).
64 - Split-Duplex links: auxiliary method to create split-duplex links in
65 the platform, easing its utilisation. It automatically creates both UP
66 and DOWN links (similarly as done in XML).
67 - Please refer to the documentation and the examples included:
68 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
69 - New plugin: Producer-Consumer with monitor. Just requires to include the
70 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
71 associated example (examples/cpp/plugin-prodcons).
74 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
75 s4u::Io::wait_any(), s4u::Io::wait_any_for().
76 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
77 parameter by reference, instead of a pointer.
78 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
79 Explicitly cancel the activity to get back to previous behavior.
80 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
83 - The default SMPI compiler flags are no more taken from the environment.
84 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
85 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
87 --cfg=smpi/finalization-barrier: which can be used to add
88 a barrier inside MPI_Finalize. This can help for some codes which cleanup
89 data attached to a process, but still used in other SMPI processes.
90 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
91 is active when set to false, to keep going after a small error
92 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
93 which may or may not be problematic in the end.
95 - fix behaviour, as maximum iteration count could be ignored
96 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
97 use sampling when the same kernel is called with a different set of parameters
98 which have an impact on the timing.
99 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
100 It should now work with smpi/auto-shared-malloc-thresh.
101 - Improve error handling and reporting in multiple places
102 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
103 validity checks, truncated messages are now an error, return errors
104 when explicitely deleted handles are reused, ...)
105 - RMA: multiple fixes and stability improvements.
106 - analysis (-analyze flag in smpirun):
107 - SMPI can now report buffer leaks as well as MPI handles leaks,
108 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
109 - if -trace-call-location is used when compiling, SMPI can report
110 origin of leaked handles/buffers
111 - group leaks by type/origin in output message if possible
112 - New implemented MPI calls: MPI_Comm_test_inter
115 - Changed internal implementation of bandwidth factors in network models.
116 Models affected: CM02, LV08 (default), SMPI, IB.
117 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
118 Bandwidth factors are applied to communications to describe that users
119 cannot use 100% of the available bandwidth. For example, the default network model,
120 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
121 behavior was implemented by limiting the bandwidth available in the LMM
122 system for this flow. This may give the false impression that there is
123 bandwidth available for other flows due to its underutilization, especially
124 for the dynamic bandwidth factors used in SMPI models.
125 To avoid this, we have modified the implementation so that each flow uses the
126 maximum physical bandwidth according to the LMM system.
127 However, the actual throughput of the flow seen by the user is defined by
128 the physical bandwidth multiplied by the bandwidth factor.
129 This change impacts on the simulation results for all network models on
130 which we have bandwidth factors configured.
131 *****************************************
132 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
133 *****************************************
134 This change may impact on the timing of your simulation results.
135 Take care when comparing simulations from different SimGrid's
136 versions. Sorry for the inconvenience.
137 - Dynamic network factors: users can configure a callback to define
138 the network factors dynamically. This API is available at
139 simgrid::kernel::resource::NetworkModelIntf.
140 - Users have access to complete information about the current communication
141 to decide which factor to apply. This includes: message size, source and
142 destination hosts, links and zones traversed.
143 - Dynamic factors for both latency and bandwidth.
144 - For more details, see the example in (examples/cpp/network-factors).
145 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
146 deprecated since version 3.24, are no longer supported. Instead, use
147 "wattage_off" and "wattage_per_state".
150 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
153 - New tutorial: Model-checking and formal assessment
154 - New sections: "Demystifying the routing" and "C++ platforms"
155 - Update and improve the part on visualization in MPI and Algo tutorials.
156 - Improve the section on routing: how to define it, how it's used internally
157 - Fix many issues, broken links and missing references in doxygen and Sphinx
160 - Lua platform files are deprecated. Their support will be dropped after v3.31.
163 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
165 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
166 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
167 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
168 - FG#64: Configuring smpi/IB-penalty-factors
169 - FG#67: Running computation concurrently with MPI_Iallreduce
170 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
171 - FG#71: Segmentation fault on invalid gw_src/gw_dst
172 - GH#322: Issue when an actor kills his host vm
174 ----------------------------------------------------------------------------
176 SimGrid (3.27) March 29. 2021
178 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
180 Main user-visible changes:
181 - All functions marked as deprecated for one year were removed.
182 - Removed all that was related to the "Storage" type in all APIs and bindings.
183 "Disks" have been introduced more than a year ago to replace them for the better.
184 It's time to finish this replacement.
185 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
186 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
187 avoiding "initializer element is not constant" errors with some codes. Please be
188 careful to recompile your MPI codes if switching to this new SimGrid version.
189 - Network model ns-3 now requires version ns-3.28 or higher.
190 - Removed some option aliases that were deprecated since v3.23 (June 2019)
191 - 'verbose-exit' is now 'debug/verbose-exit'
192 - 'simix/breakpoint' is now 'debug/breakpoint'
193 - 'clean-atexit' is now 'debug/clean-atexit'
196 - New functions: Comm::sendto_{init,async} to initiate a communication
197 on between two (possibly remote) hosts.
198 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
199 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
200 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
201 or Mailbox::get_async<void>() if you really want to play with void*.
202 - Unify the interface of Activity::on_{start/activity}
203 - New function: Comm::get_dst_data()
204 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
207 - ptask should now work better with multicore hosts (FG#37).
208 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
209 - Internal cleanups and refactoring around models, to prepare for programmatic
211 - Modernize our tests of the performance models
215 - New section: Configuring the logs from the command line.
216 - New section: Programming API of logs.
217 - New section: Getting realistic results
218 - New section: Examples of SMPI Usage
219 - New part: SimGrid's Internals, with a section on Design goals
220 - Move back to the classical Breathe tool chain, and fix many small glitches.
223 - Handle leak checking: SMPI will now display a message if MPI handles are still
224 present at the end of a simulation. Option smpi/list-leaks:n will display the n
225 first leaked pointers and their types.
226 - smpi/display-allocs option : Display Memory allocations analysis, with hints
227 about the largest allocations. This only works for allocations with malloc/calloc
228 and not compiled with SMPI_NO_OVERRIDE_MALLOC
229 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
230 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
231 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
232 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
233 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
234 - Replay now opens and parses the trace file only once for all ranks when possible.
236 - better handling of MPI_PROC_NULL
237 - MPI_Group_incl was broken in some cases
238 - PAPI tracing now fixed
239 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
240 - SMPI will now fail properly when receive size < sent size, instead of
241 just setting MPI_ERR_TRUNCATE and keep going.
242 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
243 - Remove useless build patches and split build/run patches on the page
244 - Build in parallel when possible
245 - Test on yum-based systems
246 - New nightly tested Proxy-apps:
247 - OpenMPI mpi-test-suite
248 - Coral MCB, LSMS, Qbox, CLOMP
249 - ECP/Mantevo Cloverleaf3D, Ember
250 - Codevault bhtree_mpi
253 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
254 - Start to overhaul the implementation to modernize it (still ongoing).
257 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
258 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
259 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
260 - Deprecate xbt_dynar_to_array().
261 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
263 C binding and interface:
264 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
265 - Constness requirement has been lowered for parameter argv of sg_actor_start()
266 and sg_actor_create() It's now (char* const*). The previous version is still
267 available as sg_actor_start_() and sg_actor_create_().
269 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
270 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
271 - FG#37: Parallel tasks are limited to 1 core per host
272 - FG#62: Running "smpirun -replay" on large networks
273 - FG!46: Fix a few potential memory leaks in SMPI colls
274 - GH#343: Wrong consumed energy when an exec timeouts
276 ----------------------------------------------------------------------------
278 SimGrid (3.26) December 15. 2020
280 The Release release (the French lockdown was eased today).
282 Important user-visible changes:
283 - SimGrid now requires a compiler with C++14 support.
284 Sibling projects should upgrade their FindSimgrid.cmake
285 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
286 several users had difficulties to understand issues when using high bandwidth or
287 small latency events. The new value was already the default for SMPI and
288 should not cause too much performance hit. It can change some simulation
289 timings if simulation was skipping a lot of small events (which was actually
290 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
291 - On some network models (L07, CM02, LV08), an implicit loopback link is created
292 for local communication. Default bandwidth was 500MBps and latency was 15µs,
293 which was rather bad. Default values have been changed to 10GBps and 0 latency.
294 Options network/loopback-lat and network/loopback-bw have been added to adapt
295 these values if needed. Network and routing models already providing separate
296 loopbacks (clusters and such) are not impacted by this change.
297 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
298 metadata that are added to each MPI message by MPI libraries (tag,
299 src/dest, and communicator information). This may change simulation times
300 for very small message sizes, but should be more realistic.
303 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
304 - New plugins: wifi_energy and link_load. See the associated examples.
305 - New examples: wifi networks, and communication suspend.
308 - LTO is now enabled for Intel/clang compilers.
309 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
310 Setting it to "auto" will use all available cores, while setting it to n will
311 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
312 - Remove obsolete runtime option 'contexts/parallel-threshold'.
313 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
314 has been kept for compatibility.
315 - Finally remove obsolete snake_case() aliases for runtime options.
316 - Further improve the documentation.
319 - the on_exit() of each actor is also executed when the simulation deadlocks.
320 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
321 An example is provided for s4u::Comm but it should work with Exec and Io.
324 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
325 https://framagit.org/simgrid/SMPI-proxy-apps
326 - MPI/IO: supports names starting with "./"
327 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
329 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
330 building shared libraries, this may cause issues, so environment variable
331 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
332 - most temporary files should now be created in /tmp dir (or equivalent).
333 If this one does not allow execution of code (noexec flag), this may cause issues.
334 Please use another tmp directory (using TMPDIR or equivalent system variable)
338 - A large refactoring is ongoing, but this should be transparent for users so far.
341 - Drop xbt_str_split().
343 C binding and interface:
344 - The return type of the sg_actor_on_exit() callbacks is now 'void'
345 (int value was previously ignored)
346 - Many C functions were renamed to follow a common naming style
347 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
348 names are kept until version 3.30.
349 - Many MSG tests were converted to the new S4U's interface in C, that
350 was extended for that.
352 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
353 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
354 - FG#41: Add sg_actor_create C interface
355 - FG#43: xbt::random needs some care
356 - FG#48: The Impossible Did Happen (yet again)
357 - FG#50: Suspending an actor executed at the current timestamp fails
358 - FG#52: Zero-seconds timeout : "That's in the past already"
359 - FG#53: Crash while using ns-3 network model
360 - FG#54: How to suspend a comm?
361 - FG!22: Stochastic Profiles
362 - FG!24: Documentation and fix for xbt/random
363 - FG!35: Add a modeling hint for parallel links in doc
364 - FG!36: [xbt/random] Read/Write the state of the RNG
365 - FG!42: Link cumulated load plugin
366 - FG!44: Test condition variable wait_until+notify_all
367 - GF#18137: Allow different stack sizes?
368 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
369 - GH#139: Allow pthread creation in SMPI
370 - GH#336: Packet-level simulation using SMPI?
371 - GH#345: Error in the LMM while migrating a VM
372 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
373 - GH#352: pip install / python setup.py install fail to find pybind11
374 - GH!337: Fix link_energy plugin for wifi platforms
375 - GH!339: Add Mailbox set_receiver method to python binding
376 - GH!344: Cast hugepages macros parameters to int64
378 ----------------------------------------------------------------------------
380 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
382 The Palindrom Day Release.
384 Important user-visible changes:
385 - Improve the Python usability (stability and documentation).
386 - A nasty synchronization bug was ironed out, see also below.
387 - Python's doc was organized as a separate tree, now integrated with C++.
388 - C bindings of S4U were not part of the doc.
389 - The C++ doc was also improved as methods are now split by theme.
390 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
391 - This is mandatory to use the Java bindings.
392 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
393 - SimDAG++: Automatic dependencies on S4U activities (experimental)
394 - Some features are already implemented but not all of them
395 - Cannot block an activity until it's scheduled on a resource
396 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
397 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
400 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
401 - Drop support for Viva/Triva (old visualization tools), hence removed
402 TRACE_get_node_types() and TRACE_get_edge_types() functions.
405 - Require pybind11 version 2.4.
406 - Greatly improve locking during context switches with Python.
407 - Declaring Python callbacks to this_actor::on_exit is now working.
408 - No more tests are failing currently, even on win32.
411 - Actor: Merge signals on_migration_start/end into on_host_change
412 - Actor: Rename migrate() into set_host()
413 - Disk: Allow users to get the read and write nominal bandwidth values
414 - Exec: Implement wait_for(timeout)
415 - Io: Implement wait_for(timeout)
418 - Parse errors now raise a simgrid::ParseError that you may want to catch.
421 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
422 been renamed to fix a spelling error. As usual, the previous definitions are
423 kept with a deprecation warning. However, to avoid runtime errors, it is now
424 forbidden to override these deprecated methods in a derived class. Please use
425 the new names immediately if you need to override them.
428 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
429 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
430 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
431 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
432 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
433 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
434 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
435 - New module: random, providing classical random numbers generators.
438 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
439 allocations' size above which they are considered "shared" by default (as if
440 they were performed through SMPI_SHARED_MALLOC macros).
441 Default = 0 = disabled feature.
442 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
443 This can cause some troubles if codes are already overriding these. If this
444 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
445 help, but will make this feature unusable.
446 - Input check is now performed much more consistently and outputs precise
447 warnings for each error found
448 - New supported calls : MPI_File_set_view, MPI_File_get_view
449 - MPI I/O now only supports disks, as storage is discontinued.
450 - MPI I/O allows opening files without specifying absolute path
451 - Note : SMPI configuration options are now initialized by smpi_init_options()
452 call, which should be called if SMPI is not being used through smpirun.
454 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
455 - FG#9: Python bindings crashing
456 - FG#39: Missing s4u::Comm::wait_any_for example
457 - FG#42: Add support for ThreadSanitizer (TSan)
458 - FG!19: Removing RngStream
459 - FG!20: A module for RNG calls
460 - FG!21: Choice between ad-hoc and standard distributions implementations
462 - FG!25: Fix link in Mutex doc
463 - FG!26: Fix links in SMPI interface doc
464 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
465 - GH#305: Unscheduled tasks are still excuted
466 - GH#313: smpirun: manual outdated w.r.t. --help
467 - GH#321: [S4U] Get task remaining work ratio
468 - GH#323: Crash when an actor turn off his physical host
469 - GH#335: Missing links on dragonfly example svg
470 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
471 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
473 ----------------------------------------------------------------------------
475 SimGrid (3.24) October 9. 2019
477 The Clean Disk Release.
480 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
481 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
482 thus becomes the default resource to execute I/O operations. Examples
483 and the file system plugin have been modified to use disks instead of
484 storage elements. Storage can still be used but is doomed to disappear.
485 Users are thus adviced to update their simulators accordingly.
486 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
487 for consistency with pthread_barrier_wait()
488 - Host::get_englobing_zone() returns the englobing netzone
489 - Actor::on_destruction is now called in the destructor
490 Actor::on_termination new signal called when the actor terminates
492 - Global signals are now part of the Engine:
493 - on_platform_creation: after config settings, before the XML parsing
494 - on_platform_created: right after the XML parsing
495 - on_time_advance: each time the clock advances
496 - on_simulation_end: after simulation, before cleanups
497 - on_deadlock: as the name implies.
499 - sg_{actor,host,link}_{data,data_set}() now all exist.
500 Use them to attach user data to the object and retrieve it.
503 - Introduce an experimental Wifi model. It sounds reasonable
504 according to the state of the art, but it still has to be properly
505 validated, at least against ns-3 if not against reality.
506 - Improved the usability of ns-3. Several bugs were ironed out.
507 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
508 It is now expressed as 'idle:epsilon:allCores' to properly model the
509 consumption of non-whole tasks on mono-core hosts. More info in the doc.
512 - convert a new set of functions to the S4U C interface and move the old MSG
513 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
516 - Fortran bindings for DVFS have been removed.
517 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
518 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
519 - SMPI can now be selected by cmake's find_module(MPI) with
520 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
521 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
522 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
523 - trace-call-location can be used with TI traces, and replayed, and is compatible
524 with smpi/comp-adjustment-file.
525 - sleep events are now correctly traced/replayed.
526 - Default for trace-call-location is now to use file names and not full paths.
527 To revert to previous behavior (in case of collision of filenames), option
528 "smpi/trace-call-use-absolute-path" can be set to yes.
531 - Use the included xxHash as an hash implem when C++14 is usable.
532 - Option model-checker/hash was removed. This is always activated now.
533 - New option smpi/buffering controls the MPI buffering in MC mode.
534 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
535 This is useful to check for MPI compliance.
538 - New section on plugins: how to define a new one, and existing ones.
541 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
542 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
545 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
546 and <mount> tags. The rationale is finish to decouple what should be handle
547 by SimGrid kernel (i.e., resources and activities) from plugins built in
548 user space (e.g., the file system plugin). This new tag currently accepts
549 only three attributes (id, read_bw, and write_bw). All the other information
550 that was declared with the storage related tags now has to be expressed as
551 properties. An example of platform using this new tag is available at
552 examples/platforms/hosts_with_disks.xml
555 - 'expect signal' can now accept more than one potential signal.
557 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
558 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
559 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
560 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
561 - FG#31: per-actor data
562 - FG#34: SG_BARRIER_SERIAL_THREAD?
563 - FG#35: model-checker does not like buster-produced binaries
564 - FG!13: MC: complete workaround in the error msg seen on modern systems
565 - FG!15: execute_flops now logs compute
566 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
567 - FG!17: ns-3: unblock the right number of communications + others issues
568 - FG!18: Improving the performance of the ns-3 bindings
569 - GH#189: Energy consumption of parallel tasks
570 - GH#219: Error in the throughput of TCP transfer
571 - GH!330: Minor polishing of English: Configuring_Simgrid
572 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
573 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
574 - GH!334: Rework energy plugin, again
576 ----------------------------------------------------------------------------
578 SimGrid (3.23.2) July 8. 2019
581 - Nicer introduction page.
582 - Migrate the "Deploy your application" page to the new doc.
583 - Move Java as a subtree of MSG.
586 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
589 - Drop xbt_dynar_sort_strings().
592 - Really fix FG#26: Turning off a link should raise NetworkFailureException
593 - FG#27: Wrong exception thrown to wait_any when link is turned off
594 - GH#328: Java: Canceling multiple tasks in a single vm/host
596 ----------------------------------------------------------------------------
598 SimGrid (3.23) June 25. 2019
600 The Exotic Solstice Release.
603 - SunOS and Haiku OS support. Because exotic platforms are fun.
604 - Stop setting random seed with srand48() at initialization.
605 - Use addr2line as a fallback for stacktraces when backtrace is not available.
606 - Build option -Denable_documentation is now OFF by default.
607 - Network model 'NS3' was renamed into 'ns-3'.
610 - Simgrid can now hopefully be installed with pip.
613 - wait_any can now be used for asynchronous executions too.
616 - New log appenders: stdout and stderr. Use stdout for xbt_help.
617 - Drop xbt_dict_dump.
620 - 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
621 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
622 - MPI interface is now const correct for input parameters
623 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
624 - Fortran bindings for async collectives.
625 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
628 - Remove option 'model-check/record': Paths are recorded in any cases now.
629 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
630 always sparse. That's an immense gain in memory, and can even be
631 faster because of cache effects. So there is no need to clutter the
632 code to allow the user to go for the unefficient mode.
635 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
636 models proved to be more accurate than these old experiments.
638 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
639 - FG#1: Broken link in error messages
640 - FG#2: missing installation documentation
641 - FG#3: missing documentation in smpirun
642 - FG#6: Python bindings not available on PyPI
643 - FG#7: simple cmake call requires doxygen
644 - FG#8: make python bindings an optional dependency
645 - FG#10: Can not use MSG_process_set_data from SMPI any more
646 - FG#11: Auto-restart actors forget their on_exit behavior
647 - FG#12: -Denable_lto=OFF doesn't disable LTO
648 - FG#13: Installs unstripped file 'bin/graphicator'
649 - FG#14: Installs the empty directory 'doc/simgrid/html'
650 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
651 - FG#17: Dead link in doc (pls_ns3)
652 - FG#20: 'tesh --help' should return 0
653 - FG#21: Documentation link on http://simgrid.org/ broken
654 - FG#22: Debian installation instruction are broken
655 - FG#26: Turning off a link should raise NetworkFailureException exceptions
656 - GH#133: Java: a process can run on a VM even if its host is off
657 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
658 - GH#326: Valgrind-detected error for join() when energy plugin is activated
660 ----------------------------------------------------------------------------
662 SimGrid (3.22) April 1. 2019
664 The Easter Christmas Release (this one is somewhat late).
667 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
668 - Not complete yet: asynchronous activities (amongst others) are still missing
669 - Still ongoing: the interface may change in the future. We need more
670 testers! Please report any glitches.
671 - No new project using Java should start now. Please switch to Python.
674 - Some of the internal cleanups may lead to speed improvements:
675 - The hard limitation on the amount of simulated actors with
676 Java+Mac was removed. Now, the available memory is the only limit.
677 - Our refcounting was tidyied, leading to 10% speedups in some cases.
678 - We are still working on making our code robust to the actor kills
679 and hosts' churn. Things are improving, but it's not perfect yet.
680 - Replaced our own code to display a backtrace (that was forking addr2line)
681 with the Boost.Stacktrace library.
682 You won't see your backtraces without this optional dependency.
683 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
684 - Stop setting random seed with srand() at initialization.
687 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
688 XML file version remains 4.2 since old files are still compatible.
691 - Process termination which was broken at version 3.21 has been repaired.
692 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
693 - Hide the examples into examples/deprecated. New users should use Python.
696 - Drop MSG_process_create_from_stdfunc() from the API.
697 This C++ function was a pimple in the C API, made necessary at some
698 point by the Java bindings. This is fixed now.
699 - Hide the examples into examples/deprecated. New users should use S4U.
700 - MSG_process_create and MSG_process_attach now crash if the host used
701 to run this process is off.
702 - Fix the protype of MSG_process_on_exit()
703 Now use (int,void*) callbacks instead of (void*,void*) ones.
704 The implementation was ways too messy to actually work, I guess.
707 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
708 too often and provide more accurate timings.
709 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
710 - MPI_Alltoallw support
711 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
712 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
713 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
714 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
716 - Basic implementation of generalized requests (SMPI doesn't
717 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
720 - Drop sg_cmdline. Please use xbt_cmdline instead.
721 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
722 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
723 OsSemaphore is implemented in a portable way with C++11 threads.
724 This should allow much more threads to be created at the same time,
725 allowing Mac and Java users to simulate many more actors.
726 - Implement the 'thread' factory with std::thread instead of xbt ones.
727 It is not possible to set the stack size with threads anymore, but
728 -fsplit-stack is the way to go nowadays when using threads.
729 - Drop the xbt_os_thread_t module (now unused)
730 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
731 - Drop xbt_str_join_array().
732 - Drop cunit, use Catch2 instead.
735 - Many cleanups in the kernel::activity namespace. This was long
736 overdue, and shall open the path to many future endeavors.
739 - #132: Java: a process can not shut down its own host
740 - #220: S4U: detached send causes exception if sender terminates soon after sending
741 - #261: Document the parameters of parallel execution's constructor
742 - #300: [s4u] BarrierPtr is missing
743 - #314: SMPI args internal cleanup
744 - #316: Fix a bug related to the CPU utilization of multi-core VM
745 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
746 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
747 - #325: Turning off a host has different behavior on sleeping actors and computing actors
749 ----------------------------------------------------------------------------
751 SimGrid (3.21) October 3. 2018
753 The Restarting Documentation (TRD) Release.
756 - Convert the user manual to Sphinx for improved usability.
757 Unstable version now hosted on https://simgrid.frama.io/simgrid
758 Many glitches remain; stay tuned.
759 - Restrict the Doxygen reference API to the installed header files.
760 (documenting all internals this way is overkill + doxygen scales badly)
761 - New tutorial on S4U (highly inspired from the old MSG one, but with
762 a git repository to fork as a starting point, and a docker image)
763 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
764 - The Developper manual is still to be converted (not compiled until then)
765 - Some parts are still missing in this conversion (such as the
766 platform and deployment sections) while others were blindly converted
767 and would need more love (such as the configuration flags).
768 Things will certainly further improve in the future releases.
771 - Fully reimplement auto_restart mechanism that was utterly broken
772 (fix #22, #131 and #281 that were all very related)
773 - Implement semaphores (fix #303 and #291)
774 - When creating an actor from a function and its parameters,
775 move-only parameters are not allowed anymore, as it would prevent
776 the actor to be restartable if its parameters are consumed on run.
777 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
778 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
779 IO operations that can be started, waited for, or canceled as a regular
781 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
785 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
786 - New host variable: 'core_count'
789 - Remove xbt_os_thread_specific features
790 - Remove portability wrapper to condition variables
791 - Remove xbt_os_thread_yield()
794 - MPICH collective selector now mimics MPICH 3.3b
795 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
796 - MPI_Init(NULL,NULL) is now allowed
797 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
798 - Command-line arguments should now be handled properly in Fortran simulations
799 - MPI Attributes and related callbacks should work in Fortran
800 - Apps using C + Fortran should now work
801 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
802 - Support MPI_Error_String
805 - Due to an internal bug, Msg.run() must now be your last line.
806 We hope to fix it in a future release, and we are sorry for the inconvenience.
809 - #22: Process autorestart seem to only work with CAS01 cpus
810 - #56: Feature request: dynamic SMPI replay
811 - #93: simgrid should not eat --help
812 - #111: How to change the loopback link characteristics?
813 - #116: Platform cluster radical
814 - #117: Platform cluster/host order
815 - #129: Replay_multiple on steroid
816 - #131: Java : autostart does not work
817 - #143: Setting a breakpoint at a given time
818 - #235: Network energy models should be integrated
819 - #264: Add ptask L07 resource tracing
820 - #271: Dynamic and manual replay of SMPI traces
821 - #279: Breakpoints option uses time deltas instead of absolute time
822 - #281: Daemonized actor no longer daemonized after an auto-restart
823 - #282: TIT convertor erase traces when using absolute path in the trace list file
824 - #285: segfault when a s4u actor kills itself with msg process tracing activated
825 - #286: Pajé traces are not exposing the number of cores
826 - #287: Command-line parsing should stop when encountering '--'
827 - #288: MPI_Init(NULL, NULL)
828 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
829 - #290: Method to check if a mailbox has a message ready to be consumed
830 - #291: [s4u] semaphores do not seem available yet
831 - #294: Alltoall collective implementation of mpich
832 - #295: Starting a Host as OFF is not supported
833 - #296: DTD too permissive
834 - #299: Add a s4u barrier example
836 ----------------------------------------------------------------------------
838 SimGrid (3.20) Released June 24. 2018
840 The proxy snake_case() release.
843 - The whole API is now using snake_case() instead of mixing it with camelCase.
844 Compatibility wrappers in place for one year.
845 - We took the opportunity to increase the consistency of the API.
846 Please report any remaining glitches.
847 - New function simgrid::s4u::Host::get_actor_count:
848 Returns the number of actors running on a specific host.
851 - SMPI is now tested with ~45 proxy apps from various sources, with none or
852 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
853 - Replay: The replay file has been re-written in C++.
854 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
855 supported. They are stored in the trace and used when replayed.
856 - Basic support of MPI_Cancel. Robustness not guaranteed.
857 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
858 (only for MPI_COMM_TYPE_SHARED).
859 - New option: smpi/privatize-libs, to add external shared libs to be privatized
860 by SMPI. They will be copied locally and loaded separately by each process.
861 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
862 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
863 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
864 with internal index optind with multiple processes. Only works if MPI_Init has
866 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
867 syntax without preprocessing tricks.
870 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
871 behavior tracing is no longer limited to MSG
872 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
873 this means that traces now need to be updated or re-obtained; both Irecv
874 and Isend lines in your traces have now in total 5 mandatory fields:
875 <rankid> <command> <to/from rankid> <tag> <size>
876 To update your traces, it suffices to add a 0 for the tag here.
877 - TIT now also supports waiting for a distinct request via MPI_Wait.
878 Wait/Test now wait for a specific request, not just the last one that was
879 issued. This unfortunately means another update, because we need to
880 identify which request you want to wait for. We do this via the
881 triplet (sender, receiver, tag), which needs to be added:
882 <rankid> <command> <sender> <receiver> <tag>
883 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
884 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
887 - The deprecation of MSG is ongoing (but this should not impact you).
888 Many MSG functions are now simple wrappers on the C API of S4U. If
889 you wish to convert your code to S4U, find the S4U counterparts of
890 your MSG calls in src/msg/msg_legacy.cpp.
891 - MSG can still be used, but won't evolve anymore.
894 - Allow to run the Link energy plugin from the command line with
895 --cfg=plugin:link_energy
896 - Rename Energy plugin into host_energy
897 - Rename Load plugin into host_load
900 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
901 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
902 - Improve performance when handling timeouts of simix synchros.
905 - Config: the C API is now deprecated (will be removed in 3.23), and
906 the C++ API has been slightly improved.
909 - Fix several build issues on OSX.
910 - Move simgrid_config.h to simgrid/config.h (old header still working)
913 - #143: Setting a breakpoint at a given time
914 - #150: Inconsistent event names in SMPI replay
915 - #258: daemonized actors hang after all non-daemonized actors have completed
916 - #267: Linker error on unit_tmgr
917 - #269: SMPI: tracing of MPI_Wait/all/any broken
918 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
920 ----------------------------------------------------------------------------
922 SimGrid (3.19.1) Released March 22 2018
924 The Fixed ABI Release.
926 As you may know, we are currently refactoring SimGrid in deep.
927 Upcoming SimGrid4 will be really different from SimGrid3: modular,
928 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
930 Our goal is to smooth this transition, with backward compatibility
931 and automatic update paths, while still progressing toward SimGrid4.
933 SimGrid remains open during works: The last pure SimGrid3 release was
934 v3.12 while all subsequent versions are usable alpha versions of
935 SimGrid4: Existing interfaces remain unchanged, but the new S4U
936 interface is budding and the internals are deeply reorganized.
938 Since 2015, we work hard to reduce the changes to public APIs. When
939 we need to rename a public library symbol in S4U, we let your
940 compiler issue an explicative warning when you use the deprecated
941 function. These messages remain for four releases, i.e. for one full
942 year, before turning into an error. Starting with v3.15, your can
943 also adapt to API changes with the SIMGRID_VERSION macro, that is
944 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
946 Starting with this v3.19.1, our commitment to reduce the changes to
947 the public interfaces is extended from the API to the ABI: a program
948 using only MSG or SimDag and compiled against a given version of
949 simgrid can probably be used with a later version of SimGrid without
950 recompilation. We will do our best... but don't expect too much of
951 it, that's a really difficult goal during such profund refactoring.
953 The difference between v3.19 and v3.19.1 is that the former was
954 accidentally breaking the ABI of MSG, while the later is restoring
957 S4U and kernel APIs will still evolve until SimGrid4, with one-year
958 deprecation warnings as currently. In fact, cleaning up these
959 interfaces and converting them to snake_case() is one release goal of
960 v3.20. But don't worry, we are working to smooth this upgrade path.
962 In summary, new projects should start with S4U to benefit of the
963 future, but old MSG projects should still be usable with no change.
965 ----------------------------------------------------------------------------
967 SimGrid (3.19) Released March 20 2018
969 The Moscovitly-cold Spring release.
971 SMPI user-visible improvements:
972 - Support of MPI_Type_create_subarray
973 - Unimplemented MPI calls now abort (instead of issuing a warning)
974 Some harmless calls around errhandlers still only issue a warning.
975 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
976 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
977 is now possible (mmap privatization was refactored to this end).
978 Adding new SMPI instances should also be technically doable now.
980 SMPI internal cleanups:
981 - Removed index notion from SMPI -- no more "getPid() - 1"!
982 In tracing files, actors are called rank-<PROCESS-ID> but these are
983 NOT the real ranks (tracing several communicators is not supported yet).
984 - Deprecate smpi_process_index() (will be removed in 3.22)
985 - Better testing on classical proxy apps, through an external project:
986 https://github.com/simgrid/SMPI-proxy-apps/
989 - Use a graphical TOC to make it easier to find the documentation you need
990 - Revamp the MSG tutorial
993 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
994 - Remove parameter of MSG_process_killall().
995 Resetting the PID was bogus anyway (several actors could have the same PID).
998 - Execution->setHost() can be called after start(), to migrate it.
999 - Comm::test_any() is now implemented.
1000 - s4u::Actor now has onCreation() and onDestruction() signals.
1001 - Install some kernel header files for the users' plugins and more.
1002 Warning, their API is really not stable yet.
1003 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
1004 include/simgrid/kernel/routing/*Zone.hpp
1007 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
1009 Build System and other cleanups:
1010 - Remove unused run-time parameter "tracing/onelink-only".
1011 - SimGrid now works with intel compilers (MC still buggy)
1012 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
1013 - tesh kills subprocesses on timeout (but not on windows)
1016 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1019 - #194: Feature request: simgrid::s4u::Comm::test_any()
1020 - #245: migrating an actor does not migrate its execution
1021 - #253: Feature Request: expose clusters as objects
1022 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1023 - #255: Tesh broken on Windows
1024 - #256: Modernize FindSimGrid.cmake
1025 - #257: Fix (ab)use of CMake install
1027 ----------------------------------------------------------------------------
1029 SimGrid (3.18) Released December 24 2017
1031 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1033 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1034 - Support for MSG should not be removed from SimGrid before 2020,
1035 but future evolutions will be limited to the integration of
1036 user-provided patches.
1039 - s4u::onDeadlock() signal, run before stopping the simulation.
1040 - s4u::Exec: asynchronous executions (abstraction of background
1042 - s4u::Host->getLoad() returns the achieved speed in flops/s
1045 - Link::name() is deprecated and will be removed in v3.21, use
1046 Link::getCname() instead.
1047 - Mailbox::getName() changed to return a std::string, use
1048 Mailbox::getCname() instead to get a char*.
1049 - Storage::getName() changed to return a std::string, use
1050 Storage::getCname() instead to get a char*.
1051 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1052 to have both allocation and deallocation of the map in user space.
1055 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1056 The amount of flops remaining to do is a vector, not a scalar.
1057 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1058 name implies on both sequential and parallel tasks.
1059 - Both changes fix GitHub's #223 using PR #237 as a basis.
1060 Thanks Michael Mercier.
1061 - Most examples were converted to S4U and hidden elsewhere at tests.
1062 Rationale: we still want MSG to work; we want newcomers to use S4U.
1065 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1066 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1069 - Switch to the faster dlopen privatization mechanism by default
1070 - Documentation improvments
1073 - Live migration is getting moved to a plugin. Dirty page tracking is
1074 the first part of this plugin. This imply that VM migration is now
1075 only possible if one this function is called:
1076 - C/MSG: MSG_vm_live_migration_plugin_init()
1077 - C/C++: sg_vm_live_migration_plugin_init()
1078 - Java: Msg.liveMigrationInit()
1079 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1083 - Define class simgrid::xbt::Path to manage file names.
1084 - Removed unused functions:
1085 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1086 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1087 xbt_graph_export_graphviz()
1088 - xbt/str.h: xbt_str_join()
1089 - Remove unused datatypes:
1090 - xbt/heap.h: use std::priority_queue or boost::heap instead
1091 - xbt/swag.h: use boost::intrusive::list instead
1094 - New link_energy plugin for the consumption of the links.
1095 - All of the operations on files and storage contents have been
1096 packaged into a plugin (src/plugins/file_system). The current
1097 public interface can be found in
1098 include/simgrid/plugins/file_system.h
1099 To use these functions you now have to initialize the plugin by
1100 calling MSG_storage_file_system_init() just after calling
1101 MSG_init() or sg_storage_file_system_init() just after creating
1105 - Remove the undocumented/untested tag <include>
1108 - Remove viva specific tracing as the tool is no longer maintained
1111 - #248: Kill of finished processes leads to segfault
1112 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1113 - #239: please implement signal s4u::onDeadlock()
1114 - #230: segfaults when exit() without run()
1115 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1116 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1117 - #222: Actor::kill() doesn't really kill and segfaults
1118 - #221: odd LMM warning when killing an actor
1119 - #120: Memory leak when the processes are forcefully killed
1121 ----------------------------------------------------------------------------
1123 SimGrid (3.17) Released October 8 2017
1125 The Drained Leaks release: (almost) no known leaks despite the tests.
1127 Even very long simulations will preserve your memory: our testsuite
1128 stresses SimGrid for over 45mn on fast machines for a coverage of
1129 over 80%, and there is only one single known leak, of about 4kb.
1133 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1134 handler may be given to free resources if the comm cannot be completed.
1135 - this_actor::send(mailbox) is now mailbox->put()
1136 - New: simgrid::s4u::Comm::wait_all()
1137 - New: Host.execute() for remote executions.
1140 - Unused option network/sender-gap is removed.
1143 - Deprecate MSG_task_isend_with_matching(): This unused feature
1144 really complicates our internals. Will be removed in v3.20.
1147 - Improved context termination. It is now possible to free resources, even
1148 when a process is forcibly killed.
1149 - ContextBoost: add support for Boost versions above 1.61.
1152 - Bring back run-time option --cfg=exception/cutpath to remove exception
1154 - Removed unused functions:
1155 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1156 xbt_str_rtrim(), xbt_str_trim().
1157 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1160 - Removed header files obsolete since SimGrid 3.12:
1161 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1163 - #3: SD_exit should be made optional
1164 - #120: Memory leak when the processes are forcefully killed
1165 - #159: Threading test regression in Actor refcounting
1166 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1167 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1168 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1169 - #191: VM migration and pstate
1170 - #192: Updating the energy consumptions of all hosts crashes with VMs
1171 - #195: All actors have PID=0 in the logs
1172 - #204: Sometimes segfault with thread contexts and mmap privatization
1173 - #222: Actor::kill() doesn't really kill and segfaults
1174 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1176 ----------------------------------------------------------------------------
1178 SimGrid (3.16) Released June 22. 2017.
1180 The Blooming Spring Release: developments are budding.
1183 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1184 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1185 - Java JAR file should be finally fully working on Mac OSX too.
1187 XML platforms: Switch to platform v4.1 format.
1188 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1189 - <zone> can be used as a synonym for the now deprecated <as>
1190 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1191 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1192 - <actor> can be used as a synonym for the now deprecated <process>
1193 - state_file and avail_file periodicity is now easier to express
1194 (check the documentation) the old behavior should still work.
1195 * Storage: not backward compatible, but it seems that nobody ever used it
1196 - Remove attribute 'content_type' of <storage_type>: was never used
1197 - Make attribute 'model' of <storage_type> optional: for future usage
1198 - Remove Bconnection model property: was never (in)validated. Replaced by
1199 the maximum of the read and write bandwidth as a resource constraint.
1202 - New and Backwards Compatibility break:
1203 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1204 When one wants to get the list of tasks whose states have changed during a
1205 simulation round, s/he has to allocate and free a dynar and use it as
1206 argument to this function. The former SD_simulate (double how_long)
1210 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1211 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1214 - New function to update the consumption of all hosts at once.
1215 - Fix the model for multi-core hosts, linear in the amount of busy
1216 cores with an abnormality for pIdle (see paper).
1217 BUG: Still not satisfactory for ptask on mono-cores.
1220 - The netzone are now available from the MSG API.
1221 The old names still work, but are now deprecated.
1224 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1225 - New option: smpi/keep-temps to not cleanup temp files
1226 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1227 mallocs mode is used, allows to change the size of the fake file used
1228 (default 1MB), to potentially limit the number of mappings for large runs.
1229 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1230 - Fortran ifort and flang compilers support
1231 - New RMA calls supported (experimental):
1232 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1233 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1234 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1235 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1236 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1237 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1240 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1241 - Behind the scene, ActivityImpl are now automatically refcounted.
1242 Removing this nasty bottleneck will greatly simplify our internals.
1245 - Replay: New function xbt_replay_action_get():
1246 Retrieve the function previously associated to an event type.
1247 - DROPPED FUNCTION: xbt_str_varsubst()
1248 - DROPPED MODULE: strbuff. We don't need it anymore.
1249 - DROPPED MODULE: matrix. We don't need it anymore.
1250 - DROPPED MODULE: lib. We don't need it anymore.
1252 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1254 SimGrid (3.15) Released March 22. 2017
1256 The Spring Release: continuous integration servers become green
1257 We fixed even the transient bugs on all target architectures:
1258 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1259 Windows, FreeBSD, NetBSD.
1261 - Rename NetCards to NetPoints.
1262 This was intended to help NS3 users, but that's not a netcard.
1263 That's a point in the routing algorithm, let's avoid wrong simplifications.
1266 - New: MSG_process_yield(). Stop and yield to other processes.
1267 - New: MSG_process_daemon(). Daemon processes are automatically killed
1268 when the last non-daemon process terminates
1269 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1270 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1271 to make room for the upcoming network energy plugin.
1272 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1275 - Ensure that an actor can kill itself with Process::exit()
1276 - Kill the obscure NativeException. Nobody want to survive the issues
1277 it denotes, so use JniException that is a RuntimeException (not to
1278 be caught explicitly).
1279 - Partial bug fix in initialization. SimGrid flags on command line were
1280 consumed at C level but stayed in the original Java String[] args.
1281 This could mess users' args[i] if SG flags were not put at the end of
1283 The SimGrid flags are now removed from the Java arguments. However,
1284 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1285 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1286 It might be if you have --log or --cfg flags in the command line.
1287 - Fix numerous memleaks all around the place. In particular, around VMs.
1291 - s4u::onPlatformCreated: right before the simulation starts
1292 - s4u::onSimulationEnd: right after the main simulation loop
1293 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1294 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1295 event from the availability_file changes the avail speed.
1296 - Links are now usable from s4u
1297 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1298 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1299 - New examples: The conversion of MSG examples to S4U has begun
1300 - Actors: create, kill, migration, and suspend
1301 - Applications: master-worker and token-ring
1302 - Action replay: communications and storage
1303 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1306 - Backwards Compatibility breaks
1307 - The SD_route_get_list and SD_route_get_size functions have been removed.
1308 They are replaced by the unique
1309 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1310 The route from src to dst is built in the links dynar whose size can
1311 be obtained with xbt_dynar_length.
1312 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1313 sg_host_route_bandwidth and sg_host_route_latency.
1314 Macros ensure the backwards compatibility, but you should fix your code
1317 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1318 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1319 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1320 We don't support 100% of it, but it should be enough. Contact us if not.
1321 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1322 naive and probably inaccurate)
1323 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1324 mapped onto a unique small file using some system magic.
1325 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1326 - Fortran: cleanups, fixes, support of user-added operations
1327 - RMA: MPI_Accumulate are applied in correct order.
1328 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1329 - Better support for MPI_IN_PLACE
1330 - Support for MPI_Win attrs and keyvals.
1331 - Support MPI_Comm_group_create, MPI_Type_size_x
1332 - MPI ops should be properly applied to all allowed types
1333 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1334 algorithm by a non failing one from mpich.
1337 - Kill the fifo data container: we don't use it anymore.
1340 - A <cluster> can now be created with different speed values to represent pstates.
1341 see examples/platforms/energy_cluster.xml for an example.
1343 SimGrid (3.14.159) Released December 28. 2016
1345 The previous release was only a crude approximation of a Pi release;
1346 this one should be a bit better.
1348 - Revert a last minute change that broke on Mac OSX.
1349 - Fix the changelog and NEWS files.
1350 - Improve the documentation, in particular of the routing module.
1351 - Integrate some patches flying around in Debian.
1353 SimGrid (3.14) Released December 24. 2016
1356 * The whole documentation was reworked and reorganized. There is
1357 still much room for improvement here, but we're on it.
1360 * We now need python3 for our testing infrastructure.
1361 * Model checking needs libevent
1364 * Now works on FreeBSD
1367 * Privatization now works on FreeBSD
1369 * Privatization is now activated by default in smpirun
1370 You can turn it off with -no-privatize if something goes wrong.
1372 * Call-location tracing for SMPI.
1373 You can add the exact location (filename / linenumber) of an MPI call to
1374 your trace files and slow down or speed up the simulation between two
1375 consecutive calls by using an adjustment file (see the documentation).
1377 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1378 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1379 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1380 values were ignored in some cases. The timings of these functions can now
1381 be significantly different.
1383 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1384 smpi/running-power is renamed to smpi/host-speed
1386 * smpi/grow-injected-times option to enable or disable multiplication of the
1387 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1388 make simulation less precise (but also much faster).
1390 * smpirun script should be (much) faster for large deployments.
1392 * SMPI tracing: fixed issue with poor matching of send/receives.
1394 * Replay: Fix broken waitall
1396 New functions and features
1397 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1399 Dropped / renamed functions and features
1400 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1401 - MSG_mailbox_is_empty() -> Mailbox::empty()
1402 - MSG_mailbox_front() -> Mailbox::front()
1403 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1404 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1405 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1406 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1407 * Don't pass the free_f parameter to property related functions:
1408 - MSG_host_set_property_value()
1409 - MSG_as_router_set_property_value()
1410 - MSG_storage_set_property_value()
1411 * VM properties. Since msg_vm_t are msg_host_t, just use
1412 MSG_host_get_property_value() and friends
1413 * VM I/O related things:
1414 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1415 - Unimplemented save/restore methods
1416 * MSG_as_router_get_property_value() was redundant with
1417 MSG_environment_as_get_property_value().
1418 - Removed MSG_as_router_*propert*() functions
1419 - Added MSG_environment_as_set_property_value() for consistency
1420 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1421 Well, they are still there for now, but deprecated with a warning.
1422 Please switch to xbt_dict_new_homogeneous() before this is removed
1424 * Task affinity. Its intended behavior (that was very badly tested
1425 and probably not really working) was deceiving what most users
1426 would have hoped here.
1427 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1428 * Option network/coordinates is now useless and should be dropped.
1431 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1434 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1435 * Added Dragonfly topology support
1438 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1439 EiBps, EBps, Eibps, Ebps for bandwidth.
1440 They may become useful to some lucky ones.
1443 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1444 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1445 * ex/async_waitAll: new example, on asynchronous communications
1448 * Memory usage should be decreased for simulations with a large number
1449 of processes. This also helps for SMPI.
1451 SimGrid (3.13) stable; urgency=low
1453 The Half Release, a.k.a. the Zealous Easter Trim.
1455 - We removed half of the lines, that were mostly experimental cruft.
1456 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1457 (not counting blanks and comments -- according to openhub.net)
1458 - The internals are now compiled in C++ (and will soon be clean C++)
1459 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1460 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1462 * Backwards Compatibility breaks
1463 - Removed Lua simulation bindings (switch to C or Java for that).
1464 Lua can still be used to describe platforms
1465 - Removed Java kernel plug-ins.
1466 Will be reintroduced after the ongoing major internals reorg.
1468 - the following functions were removed.
1469 They were too specific and should be reimplemented in a generic
1470 way, with filter function.
1471 - MSG_task_listen_from_host
1472 - MSG_mailbox_get_count_host_waiting_tasks
1473 - MSG_mailbox_put_with_timeout was removed.
1474 Please use MSG_task_send_with_timeout instead.
1476 - the SD_application_reinit function was removed. It has been a noop for a while.
1477 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1478 replaced by a more flexible mechanism.
1479 - The following functions thus do not exist anymore
1480 - SD_workstation_get_access_mode
1481 - SD_workstation_set_access_mode
1482 - SD_workstation_get_current_task
1483 - Basic estimation functions have been removed but can easily be replaced
1484 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1485 - SD_workstation_get_computation_time => amount / sg_host_speed()
1487 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1488 Use VM.getSpeed()*load/100 for the legacy behavior.
1490 - option enable_tracing was removed. It was not doing anything for a while.
1491 - In the ModelChecker:
1492 - the model-checker now ptraces the model-checked process which means
1493 you cannot use a debugger on the latter anymore (we might make this
1494 optional in the feature);
1495 - removed soft-dirty page tracking;
1496 - remove model-checked side snapshot management,
1497 MC_snapshot() and MC_compare_snapshot();
1498 - keep the MC_cut() function as a stub (it was not really working
1499 in the previous release).
1502 * All options are consistently in kebab-case. Old names are kept as alias.
1505 * Switch to platform v4 format.
1506 - Rename from 'power' to 'speed' the attributes describing the amount of
1507 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1508 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1509 - In <host> and <link>, attributes availability and state are gone.
1510 It was redundant with state and availability traces, and with peak values.
1511 - In <cluster>, attributes availability_file and state_file are gone.
1512 It was too complex and unused.
1513 - Kill <gpu>. Was not doing anything.
1514 - The DOCTYPE points to the right URL:
1515 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1516 (the file at this address now documents the changelog since its v1)
1517 - A warning is emitted for unit-less values (they are still accepted).
1518 - speed. Default: 'f' or 'flops'. Also defined:
1519 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1520 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1521 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1522 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1523 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1524 - latency. Default: 's' second. Also defined:
1525 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1527 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1528 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1531 * s4u::Host is now the preferred public interface to the Host features.
1532 sg_host_* functions are C bindings to the exact same behavior
1533 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1536 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1537 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1538 They were deprecated since a few years, and probably did not even compile anymore.
1541 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1542 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1543 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1544 encouraged to update their codes. A list of the modified functions can be found at the end of
1545 include/simgrid/simdag.h
1548 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1549 the lambda or closure passed as a parameter will run in kernel mode.
1550 All the callback functions should be rewritten to that interface at some point.
1553 * Reorganizing and cleaning the internals all around the place.
1556 * Remove old default barrier/bcast buggy algorithms (see #18407)
1557 * Various bug fixes to handle more codes
1558 * Remove the need for the --foreground option of smpirun (it is still
1559 accepted for backward compatibility).
1562 * Kill the setset data container: MC don't use it anymore.
1563 * Kill the queue data container: it made more sense with GRAS.
1564 * Kill the xbt_peer_t data type: it's useless without GRAS.
1565 * Kill rm_cb feature of config sets: it was never useful.
1566 * Kill graphxml parsing feature. It was not used.
1567 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1569 - xbt_dynar_sort_strings(), when the content is char*
1570 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1571 They throw exceptions on invalid input;
1572 * C++ support for declaring CLI flags (simgrid::config::Flag);
1573 * class for abstracting different signal backends (simgrid::xbt::signal).
1574 with no external dependencies (we need very simple signals).
1577 * refactoring and cleanup of the code;
1578 * ongoing process to cleanly separate the model-checking algorithms
1579 from the code model-checking support.
1581 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1583 SimGrid (3.12) stable; urgency=low
1585 The Facelift Release.
1588 * Require g++ v4.7 at least to not speak prehistorical C++.
1589 * Require Boost 1.48 (for signal2 component).
1590 * Java must be version 7 at least when activated.
1591 * Builds on Windows again (including Java bindings).
1592 * Tracing is now always enabled (no way to turn it out)
1593 * Remove GTNetS. It was not working anyway.
1594 * Various cleanups in the cmake scripts.
1595 * Move headers around to sort them out on installed systems:
1596 - instr/instr.h -> simgrid/instr.h
1597 - instr/jedule/* -> simgrid/jedule
1598 - simdag/datatypes.h was removed
1599 - simdag/simdag.h -> simgrid/simdag.h
1600 - msg/datatypes.h was removed
1601 - msg/msg.h -> simgrid/msg.h
1604 * Interface improvement:
1605 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1606 - Sanitize the interface in MSG_task_ module:
1607 - Merge two functions that were close enough but misleading:
1608 set_compute_duration(t) -> set_flops_amount(t)
1609 get_remaining_computation(t) -> get_flops_amount(t)
1610 - set_data_size(t) -> set_bytes_amount(t)
1611 get_data_size(t) -> get_bytes_amount(t)
1612 - Massive cleanups in the functions related to the energy
1613 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1614 - New: MSG_host_get_pstate()
1616 - msg/energy/onoff: switching hosts on and off
1618 * Interface improvement:
1619 - Rename Host.isAvail() to Host.isOn()
1620 - Rename Process.currentProcess() to Process.getCurrentProcess()
1621 - Rename Task.setDataSize() to Task.setBytesAmount()
1622 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1624 - #18874: Actually allows the GC to reclaim tasks
1628 - SIMIX_process_throw: raises an exception in a remote process
1629 * Refactoring: Separate sub-modules
1630 - libsmx: the public interface, as libc in a real system
1631 - popping: the strange dance that converts a user request into a kernel handling
1632 - smx_context_*: the virtualization mechanisms that embed the user code
1633 - smx_*: the handling of each simcalls
1634 * Interface cleanup:
1635 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1636 * Rename smx_action_t into smx_synchro_t, making explicit that these
1637 things are used to synchronize processes with their environment.
1638 For example, a communication is a sort of synchronization involving
1639 the communicating processes (that may block until the exchange) and
1640 the platform. The same can be said from computations, etc.
1642 - #18888: segfault when a process ends before its kill_time
1646 - 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.
1647 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1648 - MPI_Type_set_name, MPI_Type_get_name
1649 - MPI_*_c2f and MPI_*_f2c functions
1650 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1651 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1652 - Activate a lot of new tests from the mpich 3 testsuite
1654 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1655 - InfiniBand network model added: Based on the works of Jerome Vienne
1656 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1657 - When smpi/display_timing is set, also display global simulation time and application times
1658 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1659 * Collective communications
1660 - 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
1661 - 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/).
1662 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1664 - Replay now uses algorithms from wanted collective selector
1665 - Replay can be used with SMP-aware algorithms
1666 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1667 - 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.
1669 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1670 - Correctly use loopback on fat-tree clusters
1671 - Asynchronous small messages shouldn't trigger deadlocks anymore
1672 * Energy/DVFS cleanup and improvement
1673 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1674 - new: smpi_get_host_pstate()
1677 - "Full" network optimization flag was broken since Surf++
1678 - Better handling of precision flags in maxmin
1679 - Fix bug causing sometimes "Impossible" errors
1680 - Properly pass cluster properties to included hosts
1681 * Improvement of the Energy plugin.
1682 - Always update the consumption before returning that value
1683 - New property: watt_off to denote the dissipation when the host is off
1684 - New functions getWattMinAt and getWattMaxAt to retrieve the
1685 dissipation of pstates that we are not currently at.
1686 * Java: class NetworkLink renamed to Link
1687 * New function: simcall_process_get_kill_time()
1688 * Massive rename s/workstation/host/
1689 - That's intrusive, but that's good for the project consistency. Sorry.
1690 - Change config option "workstation/model" into "host/model"
1693 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1694 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1695 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1696 - xbt_dirname and xbt_basename for non-POSIX systems
1698 * The model checker now runs as a separate process.
1699 * The model checker runs must now be launched with the new simgrid-mc program.
1700 * Record/Replay: the MC can display a textual representation of a path in the
1701 execution graph. It can then be replayed outside of the model checker.
1703 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1705 SimGrid (3.11) stable; urgency=low
1710 * Normalizing pointers addresses tool for better diff between logs
1713 * Add cloud examples using new VMs
1714 - examples/msg/cloud/two_tasks_vm.tesh
1715 - examples/msg/cloud/simple_vm.tesh
1716 - examples/java/cloud/cloud.tesh
1717 - examples/java/cloud/migration/migration.tesh
1718 * Add java surf examples:
1719 - examples/java/surfPlugin/surf_plugin.tesh
1720 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1721 - examples/java/surfCpuModel/surf_cpu_model.tesh
1722 * Add SMPI+MSG example:
1723 - examples/smpi/smpi_msg_masterslave/
1728 - msg task destroy cancel test
1729 - msg_host on/off test
1730 * Move all tests in testsuite to teshsuite (adding tesh files)
1731 * Restructure teshsuites
1732 - one folder for each kind of test
1733 * Restructure AddTests.cmake
1735 - structure the order of tests (with sections)
1738 * Add virtual machine
1739 - creation of a VM on a PM
1740 - migration of a VM from a PM to another PM
1742 - MSG_process_join(msg_process_t process, double timeout)
1743 - msg_bar_t MSG_barrier_init(unsigned int count)
1744 - int MSG_barrier_wait(msg_bar_t barrier)
1745 - void MSG_barrier_destroy(msg_bar_t barrier)
1746 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1747 * New option "msg/debug_multiple_use" to help debugging when a task is used
1750 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1752 - Increase code coverage in test suites
1756 * Protect context stacks against stack overflow. The number of protected memory
1757 pages allocated on the top of each stack (1 by default) can be configured
1758 with the new command line option --cfg=contexts/guard_size:<value>.
1759 * Simcalls are now generated by a python script that
1760 - generates files included by SimGrid
1761 - checks that all the functions exist, and proposes prototypes
1763 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1764 host_set_data, host_get_data
1766 - simcall_process_join(smx_process_t process, double timeout)
1767 * Fix bug where sleeping processing could not be suspended.
1770 * Translate surf models from C to C++
1771 - Generic classes for all models: Model, Resource, Action
1772 - A generic interface for each kind of model (CPU, Network, Storage
1773 Workstation, WorkstationVM)
1775 * Translate surf routings from C to C++
1776 * Add callbacks using sigc++ or boost::signals2
1777 - Add callback functions for resource creation/destruction
1778 - Add callback functions for action state change
1779 - Handle Energy as a plugin
1780 * Replace swag by boost::intrusive
1781 * Add new routing models for clusters. For documentation, see
1782 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1783 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1784 parameters for cluster tag
1785 - Fat trees, with topology="FAT_TREE" and
1786 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1787 - see examples/platforms/torus_cluster.xml and
1788 examples/platforms/fat_tree_cluster.xml
1789 * More documentation
1792 * Hostfiles support host:nb_processes construct to deploy several processes on
1794 * Collective communication algorithms should not crash if used with
1795 improper number of nodes and report the error.
1796 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1797 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1798 MPI_Dims_create, MPI_Cart_sub are supported.
1799 * New interface to use SMPI programmatically (still depends on MSG for
1800 some parts, see examples/smpi/smpi_msg_masterslave):
1801 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1805 * Global variables privatization in MPI executables is now performed at runtime
1806 with the option smpi/privatize_global_variables (default:no).
1807 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1808 dynamic libraries loaded by the application are not privatized (static
1809 linking with these libraries is advised in this case)
1812 * Options defined in XML work correctly now.
1815 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1816 copied into simgrid.jar or not (ON by default). Use this option if you want
1817 to reduce the size of the installed simgrid.jar, *and* the native libraries
1818 are kept installed elsewhere.
1819 * Surf binding with SWIG (code generated in maintainer mode only):
1820 - plugin to handle callbacks
1821 - CPU model only for the moment
1824 * Supernovae build mode is definitively removed. It was used to improve
1825 inlining and inter-module optimizations. It is nowadays superseded by
1826 link-time optimizations commonly available in compilers.
1827 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1828 * Add boost dependency for surf++
1829 * Add new macro for tests
1830 - ADD_TESH(name <tesh_args>)
1831 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1835 - xbt_bar_t XBT_barrier_init(unsigned int count)
1836 - int XBT_barrier_wait(xbt_bar_t barrier)
1837 - void XBT_barrier_destroy(xbt_bar_t barrier)
1838 * Make the xbt_os_time module public
1840 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1842 SimGrid (3.10) stable; urgency=low
1844 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1847 * Reintegrate Java to the main archive as desynchronizing these
1848 package is not acceptable anymore (Java is now considered stable)
1849 * Add explicit synchronization facilities through semaphores
1850 * Bug fix: Task.setDataSize() only changed the C world, not the value
1851 cached in the Java world
1854 * Dramatically change the way files are handled. API and internals changed, but
1855 this part of MSG was not considered as production grade either.
1856 * Add explicit synchronization facilities through semaphores
1857 * Add a new function MSG_host_get_process_list()
1858 * Preliminary DVFS support (see examples/msg/energy/ for details)
1861 * SMPI is now included directly in the libsimgrid as the windows
1862 linker doesn't force us on splitting it anymore.
1863 * Improvements of the SMPI replay tool:
1864 - Most of the collective communications are now rooted in the same process as
1865 in the original application.
1866 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1867 used until now). Multiple data types can now be used in a trace.
1868 - The replay tool now supports traces produce either by TAU or a modified
1870 - Bug Fix: the compute part of the reduce action is now taken into account.
1871 - Gatherv collective is now supported
1872 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1873 of smpirun outputs time independent traces for the current run. One file
1874 is created per process. If too many processes are simulated, this behavior
1875 can be changed to one file for all processes by using the
1876 tracing/smpi/format/ti_one_file flag
1877 * smpirun generates the host file if needed (with given host count and platform)
1878 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1879 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1880 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1881 --cfg=smpi/coll_selector:(mpich/ompi)
1882 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1884 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1885 happening inside a collective SMPI call.
1886 * Fix the behavior of complex data types handling.
1887 * Make MPI_Wtime another synchronization point to take computations into
1889 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1890 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1891 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1892 (no privatization of global variables yet)
1893 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1896 * Verification of liveness properties is now available for SMPI applications
1897 (in addition to MSG applications)
1898 * Bugged examples using SMPI in examples/smpi/mc/
1899 * Add --cfg=model-check/visited option. Allows the verification of infinite
1900 programs. Detection of loops in the execution thanks to the system state
1901 comparison and reduction of the state space to explore. Can be combined with
1902 DPOR for safety properties.
1905 * Allow to change SimGrid configuration (see --help) within the code
1906 thanks to SD_config() as it can be done in MSG.
1907 * Add a new function SD_task_set_amount() upon user request.
1910 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1911 * Remove rule based routing (no more PCRE dependency)
1912 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1913 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1914 * Add a loopback_bw and loopback_lat options to cluster tag.
1915 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1916 routing. Now peers are encapsulated in an AS and have their own private
1917 router but this is transparent.
1920 * Our own implementation of getline is renamed xbt_getline, and gets
1921 used even if the OS provide a getline(). This should reduce the
1922 configuration complexity by using the same code on all platforms.
1923 * New type: xbt_cfg_elm_boolean.
1924 * Allow to use yes/no for boolean configuration options in the command line.
1925 * Allow to disable SimGrid cleanups at exit from command line option.
1926 There are situations where one may want a simulation to end with an exit.
1927 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1928 annoying when scripting around the simulator. Adding a
1929 --cfg=clean_atexit:no allows to circumvent this issue.
1932 * Lots of memory leaks were corrected in this release.
1933 There is no known memory leaks anymore, in all of our 600+ tests.
1934 * New command line option --version, to get SimGrid version information.
1935 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1937 * Supernovae builds are deprecated, and expected to be removed in the next
1940 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1942 SimGrid (3.9) stable; urgency=low
1944 The Grasgory release: GRAS is really dead now.
1946 * Complete overhaul of the internal host structures scheme.
1949 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1950 was considered as experimental and badly maintained since then.
1951 * Keeping it was thus a trap to our potential users, that could take
1952 it instead of MSG or SMPI by mistake despite is pity state.
1953 * GRAS seems to have very few users (if any), and no one volunteered
1954 to maintain it further. It also induces a lot of XBT code (for
1955 portability sake), that must be maintained too.
1956 * For all these reasons, we killed GRAS. If someone wants to revive it
1957 in the future, don't cry, our git history still remembers of GRAS.
1960 * Major overhaul. Merge our documentation again as time proved that
1961 splitting it was really not helping our users.
1962 * Further improve the developer documentation to help newcomers
1963 hacking on SimGrid itself. The user documentation (and in
1964 particular, the beginner documentation) is still in a sorry state.
1967 * Now works on Windows too!
1968 * Much more extensive test suite, from MPICH
1971 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1972 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1973 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1974 * Bug fix: let task be scheduled when the last dependency to be solved is
1975 a control dependency.
1976 * Remove SD_load_environment_script function.
1977 Use the C sg_platf function if you want to declare a platform
1981 * New function: MSG_process_get_number()
1982 * Old function documented: MSG_config()
1983 * Remove MSG_load_platform_script function
1984 Use the C sg_platf function if you want to declare a platform
1988 * Change the default value of the TCP_gamma constant (maximal size of TCP
1989 congestion window) to a more realistic 4MiB value. If you notice changes in
1990 your simulation results, you can fall back to the previous 20k tiny window
1991 by adding --cfg=network/TCP_gamma:20000 on command line.
1992 * (Hopefully) fix a bug wrt periodic availability and state traces
1993 * Bug fix: use default values at start when first event in availability/state
1994 trace is not at time 0.
1997 * remove the "new_" part of function name sg_platf_new_trace_connect
1998 (resulting in sg_platf_trace_connect), since it does not create
2002 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
2003 use case where it's really mandatory, and maintaining it was a pain in
2005 * New: xbt_fifo_search(), search an item with a user-provided
2006 comparison function instead of dumb pointer comparison.
2009 * Fix the lua deployment:
2010 Use `simgrid.init_application()` before deployment instead of
2011 `simgrid.msg_register_application()` after.
2014 * Transfer the tracing files into the corresponding modules.
2016 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2018 SimGrid (3.8.1) stable; urgency=low
2020 The "we are told that some people want to also *install* the simgrid
2023 * Add missing file "tesh.1" to the archive.
2025 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2027 SimGrid (3.8) stable; urgency=low
2029 The Psssshiiiit release: SimGrid jumps into the Cloud.
2032 * Add an experimental interface to manipulate VMs. They are mainly
2033 process groups with very few intrinsic semantic, but they should
2034 allow you to build the semantic you want easily.
2035 * New function: MSG_host_set_property_value()
2036 * New function: MSG_process_on_exit(). To clean memory in all cases.
2037 * Bug fixes that made the host (and link) failures unusable.
2038 * Add a way to auto-restart process when the host in which they are
2039 executing comes back (ON_FAILURE="RESTART" on deployment file,
2040 MSG_process_auto_restart_set).
2041 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2042 please stop using the old ones, they are DEPRECATED.
2044 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2045 Please use MSG_init() instead. (reducing the amount of entry
2046 points in the library helps us).
2047 * Make it impossible to link against the wrong version of the lib
2048 * Deprecate MSG_clean(). No need to call it anymore.
2049 * Function MSG_get_host_number() is not deprecated anymore.
2052 * Split the doc into a user guide and a reference guide.
2053 * Start a developper guide to help people hacking on SimGrid.
2056 * Enable tracing by default. This modules rocks you should use it.
2057 * Remove option custom_flags. Now use environment variables CFLAGS
2059 * Use default cmake things to detect lua instead of home grown ones.
2060 * New option "enable_mallocators" to disable mallocators, for debugging
2061 purpose ("on" by default).
2064 * Bug fixes around the resource failures: don't let the processes
2065 survive the host they are running onto.
2066 * Add an interface to auto-restart processes when the host in which they are
2067 executing comes back.
2068 * Ensures that SIMIX_clean is called automatically. It's not part of
2069 the public interface anymore (bindings should be updated).
2072 * Bug fix for when SD_Simulate is called with a positive value: be careful
2073 when comparing doubles. Sometimes they are different for non significant
2075 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2076 parallel task whose initial work is distributed among host according
2077 to the Amdahl's law. Such tasks are created with a parameter alpha
2078 that corresponds to the non-parallelizable part of the computation.
2079 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2080 between two sets of workstations assuming a 1D block distribution (each
2081 workstation owns a similar share of data) on both sides.
2083 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2084 Data redistribution will be automatically scheduled once parent and child
2085 are both scheduled. The filling of computation_amount and
2086 communication_amount structures is now done seamlessly thanks to the chosen
2088 * New function SD_workstation_dump to display various information
2089 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2090 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2091 bandwidth on the route onto which the task is scheduled and the amount of
2093 To divide the nominal bandwidth by 2, the rate then has to be:
2094 rate = bandwidth/(2*amount)
2095 * Compute tasks that have failed can now be rescheduled and executed again
2096 (from their beginning)
2097 * Increasing source code coverage (src/simdag is now covered at 95.8%
2101 * Re-implement time-independent trace replay using SMPI (at the
2102 smpi_smp_* level) instead of MSG. This should replace
2103 examples/msg/actions/actions.c
2104 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2105 hindexed and structs)
2106 * Implement the exchange of non-contiguous data.
2107 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2108 * Correct behavior of smpi/sender_gap and set its default value to 0
2109 * Add option to asynchronously send small messages to allow better
2110 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2111 specifies the size in bytes under which messages will be asynchronously sent.
2112 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2113 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2116 * New C interface to define a platform: XML is now optional.
2117 For more info, please check include/simgrid/platf.h.
2118 * New interface to define random platforms from the C:
2119 For more info, please check include/simgrid/platf_generator.h and
2120 examples/msg/masterslave/masterslave_platfgen.c
2121 * Export a sg_cmdline dynar containing all the arguments we got from
2125 * Two new tracing options for adding comments to trace file so you
2126 can track your experiments (see --help-tracing for details).
2127 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2128 * Adding the SimGrid version that generated the trace file as a comment.
2129 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2130 * Fix to avoid key clashes on Paje links
2131 * Other minor fixes related to the Paje specification
2134 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2135 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2136 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2137 * Greatly improve the robustness of mmalloc to user errors (such as
2138 using an area after freeing it, or freeing it twice)
2140 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2142 SimGrid-java (3.8.1) stable; urgency=low
2144 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2145 Warning: all methods are visible, but only some of them are
2146 implemented so far. Check the source (src/jtrace.c)
2147 for further information.
2148 * New module: org.simgrid.msg.File (SimGrid File management functions)
2149 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2150 * Change the meaning of Process.restart: now restart the process from
2151 the beginning, like MSG_process_restart in C.
2152 * Add Process.setAutoRestart: handling of process restart when failed
2154 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2155 you to retrieve the properties of the processes/hosts
2156 * Deprecate Msg.clean(): you can just forget about it now.
2157 * New function Process.getCount(), that only works when compiling
2158 with the not yet released version 3.9 of the C library.
2160 * New context factory based on Coroutines. It mandates a modified JVM
2161 but then, the simulations run about five times faster, and there is
2162 no limit to the amount of processes (beside of the available memory).
2164 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2166 SimGrid (3.7.1) stable; urgency=low
2169 * Restore the prototype of MSG_process_create_with_environment() to
2170 the pre-3.7 situation by removing the kill_time argument.
2171 * Add a MSG_process_set_kill_time() function instead.
2174 * Fix weird behaviors when dealing with parallel tasks.
2177 * Simgrid is now built as a dll.
2178 * Simgrid-java now works on Windows.
2179 * Simgrid-Java is now included into Windows package.
2182 * First pre-build package for MacOSX.
2185 * Fix compilation when using MSG_USE_DEPRECATED.
2186 * Fix some compilation issues on Macs and Windows.
2187 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2188 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2190 We discovered that the Lua console is broken, but we are missing the
2191 manpower to fix it right now. The problem existed in 3.7 too, so we
2192 are not blocking the release for that. Sorry if you depended on this
2193 feature, any help would be really welcome.
2195 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2198 SimGrid-java (3.7.1) stable; urgency=low
2200 The "Java aint got to be bloated and slow" release
2203 * Various internal cleanups and performance improvement
2204 Simulations are expected to run up to twice faster or so
2205 * Make Process.kill(process) an instance method, not a static one
2206 * User processes are not java.lang.Thread subclasses.
2207 This breaks the compatibility (sorry), but previous API was
2208 brain-dead, making it impossible to have non-trivial
2209 initializations in the process constructor.
2210 * Require a full constructor per Process sub-class.
2211 Kinda breaks the compatibility (sorry), but this allows a much more
2212 efficient way to launch the processes at simulation startup.
2213 * Do not embeed our version of semaphores, java 1.5 can be considered
2214 as sufficiently prevalent for us to not dupplicate its features.
2218 * Add examples for almost every part of the API
2219 We spotted and fixed a lot of bugs in the process
2221 * New module: asynchronous communication API
2222 * New function: Process.sleep()
2223 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2224 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2225 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2227 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2229 SimGrid (3.7) stable; urgency=low
2231 The "spring cleanups (before the next Big Project kicks in)" release.
2234 * We can specify the SMPI latency/bandwidth factor with command line
2235 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2236 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2237 You can also use the "config tag" from platform file by adding this line
2238 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2239 (see "example/platforms/tag_config.xml" to use "config tag").
2240 Note that the command line supersedes the platform file configuration.
2241 * Change the correction factors used in LMM model, according to
2242 the latest experiments described in INRIA RR-7821.
2243 Accuracy should be improved this way.
2244 * Use the partial invalidation optimization by default for the
2245 network too. Should produce the exact same results, only faster.
2246 * Major cleanup in surf to merge models and split some optimization
2247 mechanisms from the core of the models. As a result you can now
2248 specify which model to use (e.g., --cfg=network/model:LV08
2249 --cfg=cpu/model:Cas01) and which optimization mode to use
2250 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2251 Incompatible combinations should err at initialization. See
2252 --help-models for the list of all models and optimization modes.
2253 * The CLM03 workstation model was dropped for simplicity because it
2254 used the deprecated CM02 network model. Use default instead.
2255 * Rename the TCP_gamma configuration option to network/TCP_gamma
2256 * Rename the coordinates configuration option to
2257 network/coordinates, and document it
2258 * Use now crosstraffic keyword instead of the terribly misleading
2259 fullduplex keyword. It is activated by default now in the current
2260 default model, use --cfg=network/crosstraffic:0 to turn it off.
2261 * Ongoing refactoring the model parsing to make XML files optional
2262 See include/simgrid/platf.h for details (still to be completed)
2265 * Major overhaul of the documentation. Almost instructive now :/
2266 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2267 functions and friends. This interface was considered as
2268 deprecated since over 2 years, it's time to inform our users that it is.
2269 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2270 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2271 use these (crappy) functions in your code.
2272 These functions will be removed soon. Stop using them now.
2273 * Deprecate MSG_get_host_{table,number}
2274 Implement MSG_hosts_as_dynar() instead.
2275 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2276 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2279 * Stabilize the parallel execution mode of user contexts
2280 * Introduce configuration variables to control parallel execution:
2281 - contexts/synchro: Synchronization mode to use when running
2282 contexts in parallel (either futex, posix or busy_wait)
2283 - contexts/parallel_threshold: Minimal number of user contexts
2284 that must be part of a scheduling round to switch to parallel
2285 execution mode (raw contexts only)
2286 * Fix bugs that prevented to use suspend/resume along with
2287 synchronization structures.
2288 * Fix bugs in process termination that lead to invalid memory access
2289 in very specific conditions.
2292 * Introduce a parallel mode for the models (controlled by surf/nthreads
2293 configuration item). In our tests, running the models in parallel
2294 never lead to any speedups because they are so fast that the gain
2295 of computing each model in parallel does not amortizes the
2296 synchronization costs, even when ultra fast futexes are used.
2297 This is released anyway because YMMV.
2300 * Performance boost by using a swag internally to compute the set of
2301 tasks that are finished and should constitute the return value of
2305 * Enable it by default now that it is considered rather stable.
2308 * Documentation of the tracing functions.
2309 * Performance gains when tracing categorized/uncategorized resource
2310 utilization by avoiding calls to get route when updating resource
2311 variables. LMM constraints are being used instead.
2312 * API changed to set task categories. Use MSG_task_set_category instead
2313 of TRACE_msg_set_task_category, and SD_task_set_category instead
2314 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2315 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2316 * Examples for link user variables added (see at examples/msg/tracing/)
2317 * Deprecated function TRACE_msg_set_process_category completely removed
2318 * Trace header updated according to the latest Paje file format
2319 * Tracing network lazy updates, no longer obligate users to use full updates
2320 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2321 * Experimental: let user code declare/set/push/pop application states for hosts
2322 * API changed to allow the manual creation of graph configuration files
2323 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2326 * Improve the API of Lua MSG bindings, using the Lua spirit.
2327 * Each simulated process now lives in its own Lua world (globals are
2328 automatically duplicated). It helps writing simulators. Will allow
2329 to run Splay programs within SimGrid in the future.
2330 * Add a Chord example in Lua, equivalent to the MSG one.
2333 * Start the implementation of a solution to express temporal
2334 properties, not only local assertions. This is still an
2335 experimental work in progress, stay clear from it to be safe.
2339 - Add new runtime parameters --help-logs and --help-log-categories
2340 to display information about supported logging parameters and
2342 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2343 don't exists anymore.
2344 * Mallocators: allow value NULL for the reset function.
2346 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2347 create homogeneous dictionaries, where all the elements share the
2348 same free function. Non homogeneous dictionaries will be
2349 deprecated in the next release.
2350 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2351 - Multi-level dictionaries are deprecated.
2353 - new function xbt_dynar_search_or_negative() that is useful when
2354 you have less than 2 million elements in your dynar and don't
2355 want of the extra complexity of catching exceptions when the
2356 element is not found.
2358 - Make xbt_os_thread module (for thread portability) public.
2359 Documentation is still to come, sorry.
2361 - Cleanups and simplifications to make it maintainable again.
2362 - Exotic features (such as memalign and valloc) were removed.
2363 - The metadata were extended and improved so that the
2364 model-checker becomes able to explore and inspect the heaps.
2365 - This may induce a performance drop when enable_model-checking is
2366 ON in cmake (even if it's not used in the simulation), but it is
2367 necessary at this point to get MC working.
2369 Turn model-checking OFF if simulation performance matters to you.
2370 Not enabling it at runtime is not enough, disable it in cmake.
2372 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2375 SimGrid (3.6.2) stable; urgency=low
2377 The "Not coding new stuff allows to polish old things" release.
2380 * New bindings to the NS3 packet level simulator (experimental)
2381 * Use the raw (efficient) execution contexts instead of the sysv
2382 (portable) ones when possible.
2383 * libpcre is now mandatory in any cases since not using it led to
2384 severe performance loss and possibly other issues
2385 * Update the XML platforms:
2386 - G5K: include the latest machine in Nancy
2387 - GridPP and LCG: new platforms
2388 * Documentation was partially updated, at least (more to come)
2390 Bug fixes, cosmetics and small improvements
2391 * Free terminated processes before the end of the simulation to avoid
2392 exhausting the memory of users having very dynamic amount of
2394 * Bug fix and cosmetics about canceling non-running tasks
2395 * Bug fix about the dot loader's issues when using libcgraph
2398 * Create an installer for windows with nsis (amd64 and win32)
2399 - Add an hello world project to illustrate simgrid project creation.
2400 - Embed libpcre into the Simgrid installer to avoid
2401 its compilation burden
2402 * The raw execution contexts should work on Apple now
2403 * Port to Windows 64 bits
2404 - Sysv contexts now have an implementation for this arch
2405 - GRAS communication features now support this arch
2406 * Drop support for borland compiler on windows
2407 - this code was not maintained, and we kinda depend on gcc nowadays
2408 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2409 * Fix portability issue on unstable ubuntu: linker became picky on
2412 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2414 SimGrid (3.6.1) stable; urgency=low
2416 The "Oops, we broke Macs too" release
2419 * Fixed contexts detection so that raw ones are used when possible
2420 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2421 leads to a strange error, with user code segfaulting sometimes when
2422 the generated code is not perfectly aligned (which is not
2423 controllable from the user side, depends on the amount of code)
2426 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2428 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2430 SimGrid-java (3.6) unstable; urgency=low
2433 * Split of every thing from simgrid v3.5 into a separate package.
2435 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2437 SimGrid (3.6) stable; urgency=medium
2439 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2442 * Bindings now constitute their own package, separated from the main one.
2443 Rationale: reduce our maintenance nightmare by reducing the module coupling
2444 They will soon be released on their own on gforge.
2446 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2447 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2449 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2450 * It's not quite deprecated for now because we have no replacement,
2451 but it may soon become the case.
2454 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2455 * Fortran: New user-level cache variable to store the rank of the running
2456 process. This improves performance by an order of magnitude.
2457 * C: New coccinelle script to automatically locate and modify global and
2458 local static variables.
2459 * Improved SMPI network model with a sender-side gap to account for multiple
2463 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2464 only say if a communication is finished, no matter whether it succeeded or
2465 failed. You can call MSG_comm_get_status() to know the status of a finished
2467 * New function MSG_task_dsend() to send a task and detach it. When a
2468 communication is detached, you are never notified of its success or failure
2469 and the memory is released automatically once it is finished. This function
2470 is useful when you don't care about the end nor the success of a
2472 * Change the prototypes of action replay. Sorry for inconvenience,
2473 but this is really more efficient this way (and to adapt your code,
2474 you just have to fix the initialization, that shouldn't be too long)
2475 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2476 ever managed to do anything useful with it.
2477 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2478 takes a dynar of communications. It returns immediately and gives the
2479 index of a finished communication (if any).
2480 * New example: a basic implementation of the Chord P2P algorithm.
2483 * New model for multi-core CPUs. You can now use the core attribute to
2484 precise the number of cores of a host. This is a basic model. Every
2485 process running on the host receives at most the power provided in
2486 the DTD (throughput<=power). Total throughput of process cannot exceed
2488 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2489 router linked by an up-link and a down-link (possibly asymmetrical).
2490 This kind of pattern allows to easily build last-mile model style platforms.
2491 Aggregating such patterns in a rule-based AS is thus the technique of
2492 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2493 * New model for Vivaldi routing. We transformed the Vivaldi network model
2494 into a Vivaldi routing model (based on the rule-based model). This allows to
2495 combine Vivaldi based latencies with last-mile platforms.
2498 * Added a check for NaN of IEEE754 infinite in the double entries of
2500 * Introduce a new context factory "raw", highly inspirated from the
2501 ucontext factory, but using manually crafted functions in assembly to
2502 do the work in an efficient manner.
2503 * Allow to change the used context factory at run time, not only at
2504 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2505 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2506 contexts at runtime (only with raw contexts or ucontexts).
2507 * Completely rewrote this module to allow parallel execution of user
2508 processes. Use --cfg=contexts/nthreads:N to execute user processes
2509 with N parallel threads (the default is 1, meaning no parallelism).
2510 * Allow to decide dynamically between sequential and parallel modes.
2511 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2512 processes in parallel only when their number is greater than or equal to P
2514 * Added a check for NaN of IEEE754 infinite in the double entries of
2518 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2519 won't output the state of processes when interrupted with Ctrl-C
2520 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2521 NULL-terminated array. This may solve backward compatibility issues
2522 due to the change to return type of SD_simulate. See also:
2523 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2524 * Add new macros with variable number of arguments.
2525 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2526 - in xbt/asserts.h: xbt_assert
2527 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2528 - in xbt/ex.h: THROWF and RETHROWF.
2529 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2531 * Change xbt_die() to accept a format string with arguments, just like printf.
2532 * New data structure: xbt_lib_t, like a dict but more general and with better
2536 * New configuration options
2537 Options triva/categorized and triva/uncategorized can be used to generate
2538 graph configuration files for Triva visualization tool.
2539 * Configuration option tracing/platform is renamed to tracing/categorized
2540 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2541 * New TRACE_user_link_variable interface:
2542 User provides the name of the link and the tracing variable to attach to it
2543 * the declaration of tracing categories must be done after the environment
2545 * simpler tracing interface, just one way to declare categories
2546 TRACE_category or TRACE_category_with_color, it is up to you
2547 * links in the trace file are again identified by their names
2548 * trace contains the full platform hierarchy exactly as declared using the ASes
2549 * Options tracing/msg/[task|process]:1 groups the process by hosts
2550 for both cases, tasks and processes must have names that are unique during the simulation
2551 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2552 * The experimental option tracing/msg/volume is deprecated
2553 its functionality may be reincorporated if needed
2555 The tracing generates a trace file with unordered timestamped events,
2556 because of the way the core simulator (surf) works. A script available
2557 at the tools directory (fix-paje-trace.sh) can be used to put the events
2558 in order. We have changed the tracing so it can generate ordered timestamped
2559 events in the final trace, but depending on the simulator (and how much time
2560 is simulated) that can lead to a huge memory utilization. It is deactivated
2561 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2563 Build Infrastructure
2564 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2565 - We are trying hard to keep the API stable, but it may happen that
2566 some things change (we're a research project after all, not a
2567 nuclear plant operating system). If such things should happen, you
2568 could rely on that macro to adapt.
2569 - current value: 30600 for 3.06.00, aka 3.6
2570 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2571 a number that can be compared with SIMGRID_VERSION.
2572 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2573 assertions and verbose/debug logging events are disabled at compile time.
2575 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2577 SimGrid (3.5) stable; urgency=medium
2580 * New feature to any SimGrid-based simulator: Model-Checking
2581 Check SIN#1 for more details.
2584 * New Model SMPI (three-interval linear regression for correction factors)
2585 See RR-7426, available at http://hal.inria.fr/inria-00527150
2586 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2587 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2588 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2589 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2590 * See also src/smpi/README
2595 - Tracing API changes: TRACE_start and TRACE_end should not be called
2596 by user-code. They are automatically called by simulators created
2597 with SimDAG, MSG and SMPI if the toolkit is compiled with
2598 tracing_enabled=ON. Categories declaration and utilization remain the
2599 same for MSG and SimDag.
2600 - A function was added to the tracing API to declare categories with
2602 - TRACE_category_with_color (char *category, char *color)
2603 where color must be in the following format
2604 "%f %f %f", red, green, blue
2605 and red, green, blue are float values in the interval [0, 1]
2606 - User can specify NULL as color parameter, or continue calling
2607 TRACE_category (cat)
2608 On that case, the tracing system will define random colors
2609 - The following command-line options are supported:
2610 --cfg=tracing/filename:msg.trace
2611 --cfg=tracing:1 (activate tracing, needed to use others)
2612 --cfg=tracing/platform:1 (categorized resource use)
2613 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2614 --cfg=tracing/msg/task:1 (task creation)
2615 --cfg=tracing/msg/process:1 (process creation, migration)
2616 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2617 --cfg=tracing/smpi:1 (SMPI interface tracing)
2618 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2619 - examples of examples/msg/tracing updated
2621 - DAXLoader and DOTLoader functions can generate tasks with categories
2622 - A new function to attribute a category to SD tasks:
2623 TRACE_sd_set_task_category (SD_task_t task, char *category)
2624 * Tracing the MPI interface implemented by SMPI
2625 - Collective operations are traced with states
2626 - Point-to-Point operations are traced with states/links
2627 - Tracing activated by a parameter "-trace filename" passed
2628 to smpirun during execution (considering that simgrid
2629 is compiled with tracing enabled)
2630 - To run the simulation with gdb, the simulator
2631 accepts --cfg=tracing/smpi:1 to trace SMPI
2632 - tesh files to check if smpi tracing is ok
2633 - See examples/smpi/NAS/DT-trace
2634 * GTNetS tracing re-worked
2635 - adaptation to the tracing system of GTNets to cope
2636 with modifications regarding the fullduplex mode
2637 - new tesh files to check if gtnets tracing is ok
2640 * Asynchronous communications through the functions:
2641 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2642 * New function: MSG_load_platform_script()
2643 to make possible using a lua script instead of XML files to set up platforms
2644 * New function: MSG_set_function
2645 to associate functions to processes, used when bypassing the parser
2646 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2648 Platforms: Add some more examples in examples/platforms
2649 * Grid'5000: see www.grid5000.fr
2650 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2653 * Change the XML format. This is a very important modification. SimGrid
2654 3.5 introduces a new hierarchical format based on the notion of
2655 Autonomous Systems. Compatibility with old format is ensured
2656 through the perl script provided in the install bin directory
2657 bin/simgrid_update_xml.
2658 It is now possible to build platforms with specific routing
2659 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2660 connect several platforms together. We will try to provide soon
2661 set of realistic platforms exploiting these properties (have a
2662 look at examples/platforms/ for the moment).
2663 * Take the opportunity of the XML format change to be a good XML citizen:
2664 rename link:ctn to link_ctn and similar changes (also dealt with by
2666 * Add a new routing scheme (rule-based) using regular expressions. It
2667 enables to have an extremely low memory footprint when the
2668 underlying routing is simple and can be compactly described. You
2669 need to have libpcre4-dev (perl regular expressions) installed if
2670 you want to use this routing scheme.
2671 * Revive the cluster TAG and allow to easily and efficiently (both in
2672 term of memory and speed) connect clusters together. Have a look
2673 at teshsuite/simdag/platforms/ to see how this can be done. With
2674 this tag, you can create clusters with thousands of tasks at no
2675 cost (have a look at examples/platforms/).
2676 Note: clusters are implemented as ASes, so there is no need for an
2677 enclosing AS tag if you have only one cluster in your platform.
2678 * Add new generic functions in the public interface that allows the user
2679 to call SURF 'create_resource' methods from your code (same
2680 functionality as the XML bypass mechanism but with a much lighter
2682 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2683 that uses a piecewise linear approximation to produce better
2684 results when exchanging small messages.
2685 * Add a new parameter to handle correctly full duplex link and account
2686 for interferences between uplink and downlink communications
2687 (activate with --cfg=fullduplex:1).
2690 * Rename the SD_READY (all dependencies are satisfied and task is
2691 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2692 dependencies are satisfied) state.
2693 This prevents a confusion between the notion of "ready to schedule"
2694 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2695 simulated" (SD_RUNNABLE) used by the simulation kernel.
2696 * Change the way a task is considered as ready. Instead of removing
2697 dependencies when a task is done, a counter is decreased. This way,
2698 it is always possible to reach ancestors thanks to the
2699 SD_taks_get_parents function (even after the end of the simulation.)
2700 * Change the return type of SD_Simulate from (SD_task_t*) into
2701 xbt_dynar_t. This function was in handling a dynar internally and
2702 converted it into a NULL terminated array for historical reasons.
2703 * New function SD_dotload(char*) to load a DAG described in dot
2704 format. This loader and the corresponding examples require the
2705 installation of the graphviz library.
2706 * Fix a bug in the management of tasks of size 0 in the surf network
2707 models. This problem was only visible with SIMDAG and you should
2708 thus disregard results produced with earlier versions if you
2709 relied on this feature (some tasks were blocked because of this).
2710 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2711 even though your applications were doing only point-to-point
2712 communications and sequential computations. Now you can really use any
2713 model you want (of course, if you create real parallel tasks, which are
2714 not implemented in most models beside ptaskL07, this will abort).
2715 * Add an example that schedules a DAX on an heterogeneous platform
2716 using a Min-Min strategy.
2717 * New function SD_workstation_get_current_task() that returns the kind
2718 of task currently running on a workstation in the sequential access
2720 * Raise some warnings when unexecuted tasks remains at the end of the
2721 simulation. This is usually caused by cycles in the DAG.
2724 * New function: SIMIX_process_set_function() called by MSG_set_function
2725 * Change the underlying waiting queue in semaphores so that a process
2726 can wait on several of them simultaneously (as in waitany).
2727 * Fix the way to handle tokens in semaphores so that all access patterns
2728 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2729 * kill the dirty pimple SIMIX_message_sizes_output()
2730 Please use (proper) visualization instead
2733 * New data container: setset (set of sets of elements)
2734 * New module: mmalloc (mapped malloc, allowing to have several
2735 independent segments of malloc)
2736 * New function: xbt_dict_cursor_set_data()
2737 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2738 * New function: xbt_dynar_is_empty()
2739 * New function: xbt_fifo_get_last_item()
2740 * Fix xbt_dynar_shrink(): use the right element size.
2741 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2742 uninitialized areas during expand.
2743 * Fix semaphores: previous implementation was severely broken.
2744 * Use library init/fini functions for our initialization.
2745 - you can use logs and other feature as soon as you want in your
2746 code (even before the xbt_init / MSG_init)
2747 - xbt_exit is now a no-op and produce a warning when used.
2750 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2751 benefit from the latest improvement to the simulation kernel.
2752 * Kill measurement sockets for now. If you rely on them, sorry. This
2753 release is not for you. This feature will be reintroduced in the
2754 future, but we cannot delay the release any further.
2755 * New function: gras_msgtype_get_name().
2756 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2757 * Fix (at last) the pmm example: it should not randomly fail anymore.
2759 Build chain: bug fixes and overall polishing
2760 * Cmake is now stable enough. Hence, we killed the autotools.
2761 * Port to windows ( TM :)
2762 * Fix the 'make install' target.
2763 No need to use 'make install-simgrid' anymore
2764 * Introduce a 'make dist' target compiling a *source* archive
2765 'make package' compiles a binary archive
2766 * Compile java files only on need
2767 * Add --cd and --setenv command line options to tesh
2768 * Out of source builds are not fully supported yet, but we are close
2769 * Enable supernovae and optimization flags by default for our users
2772 * Add layer to set up environment directly from lua, without XML.
2773 * The effect of gras_stub_generator can be achieved through
2774 lua too (check examples/gras/console/ping_generator.lua)
2776 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2778 SimGrid (3.4.1) stable; urgency=low
2780 The "Polishing easter eggs is probably a good idea" release.
2781 This is a bug fixes release only.
2785 * Fix a bug preventing the tasks from begin garbage collected.
2788 * Fix a bug occurring when a host involved in a communication fails.
2789 This was not detected properly by the other peer involved in the
2790 communication. Now, it's reported as a network error.
2793 * Warn the user about loop dependencies in data flow of DAX files
2794 * Obey the control-flow dependencies of DAX files
2797 * Add option "enable_smpi" allowing to not compile SMPI.
2798 Probably useful for the (Mac) users experiencing a build error here
2799 * Improve the detection of lua5.1 and ruby1.8
2801 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2803 SimGrid (3.4) stable; urgency=low
2805 The "Easter in Cargese" release. Also known as (major changes):
2807 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2808 ~> bindings were greatly improved
2809 ~> new tracing infrastructure for better visualization introduced
2811 * the "Welcome to configury modernity" release.
2812 ~> we switched from autotools to cmake, and improved our cdash
2815 A more detailled list of changes follow (full detail in svn log).
2817 Java Bindings: Various Cleanups
2818 * (install java-gcj-compat-dev on debian-like to use them)
2819 * Remove put/get: no need to export deprecated interface in Java
2820 Use send/receive instead.
2821 * Cleanup the examples and add a README per directory
2822 * Remove example autoDestination (that's the only way to go now)
2823 * Remove example explicitDestination (was a plain copy of basic)
2824 * Make JniException a runtime exception, so that there is no need to
2825 declare the fact that you may encounter such a beast. I guess that
2826 nobody will ever want to survive such error.
2827 * Create specific errors for each MSG case of failure:
2828 host failure, transfer failure, timeout, task cancelled
2829 * Cleanup the exceptions that may get thrown by each function
2830 * Other internal cleanups in Java bindings. Performance still bad :/
2831 Ruby and Lua Bindings: create them
2832 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2833 * That's new and great, you should try them out.
2834 Same functionalities than Java bindings, only even less polished
2836 * Kill the useless "rate" argument of SD_task_get_execution_time()
2837 Everyone used to provide -1 as a value, it was not used, and the
2838 semantic of a possible use wasn't even clear.
2839 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2840 either as comm costs or compute costs to mean that there is no
2841 such thing for that specific task.
2842 * Add a SD_task_set_name() function
2843 * Fix SD_task_unschedule() on typed tasks
2844 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2845 * In DAX loader, accept useless 'level' attributes to <job> since
2846 LIGO DAGs have them (seem to be to ease graphical representation).
2848 * Add an example masterslave_mailbox.c using send/receive and not
2849 the deprecated put/get interface.
2850 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2851 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2853 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2854 (declare MSG_USE_DEPRECATED to still have the old name)
2855 * Add a MSG_task_set_data() function
2856 * About trace replay (see examples/msg/actions):
2858 - Allow to work with splitted trace files for each process
2859 Give the specific trace file as argument of each process,
2860 and call MSG_action_trace_run(NULL)
2861 You can still have one merged file for all processes.
2862 - Fix implementation of collective operations
2863 * Allow task_execute() on 0-sized tasks (closes #10063)
2865 * This is the first release of SimGrid where SMPI is not considered
2866 beta anymore (even if some corners should still be improved)
2867 * Port over the new SIMIX_network submodule (internal refactoring)
2868 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2869 * Implement more missing elements of the standard:
2871 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2873 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2874 MPI_Scan() MPI_Get_processor_name()
2875 - Added implementation of missing case for Alltoall (warning: it's
2876 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2877 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2878 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2879 SMPI_MPI_Allgatherv()
2880 * Bug fixes include:
2881 - MPI_Waitsome() was broken
2882 - Allow relative includes in smpicc
2883 - Command line cfg argument 'reference_speed' was ignored...
2884 - Some functions did not properly lead to auto-benching of user code
2885 - smpicc passes -O2 by default (just like openmpi one)
2887 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2888 * Bug fixes about timeouts during communications
2889 * add SIMIX_message_sizes_output() as a pimple to write to file the
2890 amount of messages per size. Use gnuplot to get histogram.
2891 Pimple because that's the only user-visible function of simix,
2892 defined directly in xbt.h (irk, sorry)
2894 - Add a SIMIX_sem_get_capacity() function
2895 - Fix interactions with processe resume/suspende
2896 - release_forever() was stupidly broken
2897 - Fix SIMIX_display_process_status() for processes in a semaphore
2898 - Make SIMIX_sem_block_onto() user-visible
2899 * Refactoring context stuff:
2900 - Use pseudo-OOP for better modularity
2901 - reimplement SIMIX_process_kill() without process_schedule() so
2902 that the latter can take as invariant that it is called from
2904 - Merge context_start into context_new for sake of simplicity
2906 * Add a Vivaldi network model, coded live during SUD'10 ;)
2907 * Rename configuration variables to start a hierarchy:
2908 o cpu_model -> cpu/model
2909 o network_model -> network/model
2910 o workstation_model -> workstation/model
2911 * New configuration variables:
2912 o network/bandwidth_factor: correction to bandwidth
2913 o network/latency_factor: correction to latency
2914 o netwotk/weight_S: correction to the weight of competing streams
2915 * Add a long description to the models, that users can see with such
2916 argument on the command line: --cfg=cpu/model:help
2917 * --help-models display the long description of all known models
2919 * config: add the ability to set a default value after registration
2920 Does not override any previously set value (e.g. from cmd line)
2921 * dict: allow to have integer key and data.
2922 When so, you need to use the following functions
2923 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2924 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2925 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2926 In contrary to regular dicts, the key is not malloced before copy.
2927 Mixing scalar and regular elements in the same dict is not tested
2929 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2930 Tracing for Visualization:
2931 * SimGrid is now instrumented in order to generate a trace file for
2932 visualization analysis: to use it, need to compile SimGrid with the
2933 "tracing" option enabled, and instrument the program using SimGrid with
2934 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2935 (among other functions).
2936 * The instrumentation only traces the platform utilization for now
2937 * Documentation to use the tracing functions and how to analyze the
2938 traces with the Triva tool is written.
2939 * More information about: SimGrid FAQ (in the section Tracing Simulations
2942 * We moved to cmake as default build system. Autotools support will
2943 be dropped soon. Check the FAQ for more info about how to use it.
2944 * Greatly improved our cdash/ctest interactions
2945 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2946 * Added memory checking tests with valgrind; lot of memleak fixing.
2947 This may be the first release of simgrid with so few memory issues
2948 * Added code coverage tests.
2949 Our coverage is still improvable, but at least we see it on cdash.
2951 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2953 SimGrid (3.3.4) stable; urgency=low
2955 The "Desktop Grid needs love too" release (also called Xmas release).
2957 Models improvements:
2958 * Major speedup in the maxmin system solving by using lazy evaluation
2959 Instead of solving completely the maxmin system at each iteration,
2960 only invalidate (and recompute) the modified parts.
2961 This new feature is enabled in default models but you can try to
2962 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2963 * Cas01 IMproved as default CPU model
2964 This CPU model is the same Cas01 model, but it uses the
2965 maxmin-selective-update flag and a heap structure to manage
2966 actions on SURF kernel.
2967 It reduces the complexity to find the next action to finish and,
2968 consequently, it's faster than the old Cas01.
2969 This is the new default CPU model (Cas01).
2970 * Rename the old Cas01 model to Cas01_fullupdate
2971 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2972 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2973 * CpuTI (CPU Trace Integration)
2974 A new CPU model whose objective is simulate faster when using
2975 availability trace files.
2976 Instead of using a full featured, over engineered maxmin system for
2977 CPU modeling, this model does the pre-integration of traces files
2978 to calculate the amount of CPU power available, and so, executes
2979 faster than the old CPU models.
2980 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2981 * Use LV08 as default network model since it gives better accuracy
2982 for small messages and shouldn't change things for big ones.
2983 Use --cfg=network_model:CM02 to get the previous behavior.
2986 ******************************************
2987 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2988 ******************************************
2989 * The new CPU model may changes simulations!
2990 The point is that events occurring at the exact same timestamp
2991 are not scheduled in the same order with the old and new
2992 version. This may be enough to completely change the execution
2993 of simulations in some cases.
2994 * The new network model will change simulations!
2995 This new model is more realistic than the previous one, so you
2996 should consider redoing your old experiments with this model.
2997 Sorry for the inconvenience.
3000 * Introduce the supernovae compilation mode
3001 When compiled that way, the whole SimGrid (or almost) is put in a
3002 single compilation unit and compiled in one shoot.
3003 This is to help gcc which has difficulties to inline stuff from one
3005 The speedup seem to be above 15%, althrough more tests are needed on
3006 amd64 to confirm that gain.
3009 * Port of MSG's mailbox on top of SIMIX network
3010 The put/get mechanism was greatly simplified on the way.
3013 * New SIMIX network module. Provides:
3014 - Mailbox: rendez-vous mechanism to find with who you want to speak
3015 - Synchronous send/recv: easier and hopefully faster since the
3016 logic is handled in the maestro process directly now
3017 - Asynchronous send/recv: you dreamt of it? It's here now
3018 Too bad that nobody cared enough to propagate the change to MSG.
3019 * Add semaphores as SIMIX synchronization mechanism.
3022 * new function SD_daxload(char*) to load a DAX file
3023 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3024 * Introduce typed tasks. Specify its kind and cost at creation.
3025 At scheduling, just give where it should be placed, and the cost
3026 for each involved resource is automatically computed.
3027 Existing constructors so far (more to come of course):
3028 - SD_task_create_comm_e2e() for end-to-end communication
3029 - SD_task_create_comp_seq() for sequential computation
3030 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3031 * new function SD_task_dump() for debugging display
3032 * new function SD_task_dotty(task,FILE*) writing to file the info
3033 about the task in dotty format
3034 * SD_task_dependency_exists() can now cope with having one of its
3035 arguments NULL. If so, it tests whether the other argument has any
3037 * Add getters on list of preceding/following tasks:
3038 SD_task_get_parents(task) and SD_task_get_children(task)
3039 * Add getters on amount of workstations and list:
3040 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3041 * Add getter on task kind: SD_task_get_kind(task)
3042 * Update the start_time and finish_time of tasks on completion/failure
3043 * Bugfix: Remove task from state swags when destroyed
3046 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3049 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3050 dynar with one separate thread per value of the dynar.
3051 * Change the prototype of xbt_thread_create(), sorry.
3052 Added a boolean parameter indicating whether we want to join this
3053 thread (used in SG only for now)
3054 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3057 * GTNetS wrappers should now be usable again (and betterly tested too)
3058 * Fix a major regression from 3.2 where the timeout provided to
3059 MSG_task_put_with_timeout() was used as absolute time before which
3060 the comm should be done.
3061 * Start to fix the <cluster> tag.
3062 - Internal links should be good now (beside of the loopback, which
3063 use the private link instead)
3064 - paths to the external world is still rather broken
3065 - the <route:multi> tag is just broken. Actually that's brain-dead.
3066 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3067 to make it less stupid
3068 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3069 * Fix a source-level compatibility glitch from 3.2: after defining
3070 MSG_USE_DEPRECATED, you can use the old name
3071 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3072 * Allow to compile from the SVN with automake 1.11
3073 * Fix some problems when using the "start_time" tag in deployment XMLs.
3074 * Fix #8569: XBT/synchro.h has redundant declarations
3075 * Fix #8563: MSG return values and exceptions
3076 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3077 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3078 Thanks to Robson Peixoto.
3079 * Fix a vicious bug in dictionaries inducing that some elements were
3080 not freed on xbt_dict_free()
3082 Portability report of this version:
3083 * Main portability targets:
3084 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3085 on (amd64/i386/ia64)
3086 - mac leopard on i386
3087 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3088 but nothing critical.
3089 * Other platforms: windows, AIX and others were not tested for this release
3091 Timing report of this version:
3092 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3093 scenario parameters). From 8h to a few seconds in desktop grid settings.
3094 * Supernovae brings about 25% speedup on i386.
3096 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3098 SimGrid (3.3.3) stable; urgency=low
3100 The "Need for Speed" release.
3102 The timings done to validate the 3.3.2 were faulty.
3103 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3105 The problem was a conversion from a manually handled vector to
3106 xbt_dynar_t on the critical path.
3107 xbt_dynar_foreach calls functions, inducing stack management crap.
3109 We inlined these functions and xbt_dynar_foreach is now breath taking.
3110 We also inlined xbt_swag_belong on the way.
3112 Here are some approximate speedup measurements (on master/slaves
3113 simulations lasting between 10s and 20s each):
3114 3.3.1 -> 3.3.2: about same performance
3115 3.3.2 -> 3.3.3: 40% speedup
3116 3.3.1 -> 3.3.3: 40% speedup
3117 3.3.1 with inline patch -> 3.3.3: 30% speedup
3119 Our reading is that the refactoring which occurred in 3.3.2 made us
3120 suffer much more from the xbt_dynar_foreach low performance, but
3121 once we solved this, this refactoring proved to be very performance
3122 effective. From the 40% speedup, somehow, 10% are due to the
3123 inlining and 30% to the refactoring.
3125 That's a pitty that gcc cannot inline functions placed in other files
3126 alone. We have to choose between:
3127 - break the encapsulation (by putting private data structures and
3128 accessors in headers files to help gcc)
3129 - live with low performance
3130 - switch to a decent compiler such as icc (not quite possible).
3132 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3134 SimGrid (3.3.2) stable; urgency=low
3136 The "Simplicity does not preceed complexity, but follows it" release.
3138 The main contributors of this release were (lexical order):
3139 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3142 * Extract the routing logic into its own object.
3143 (was dupplicated in network.c and workstation_LV07.c;
3144 Allows to implement other ways of storing that info)
3145 => kill now useless network_card concept
3146 - Use dynar to represent routes (instead of void** + int*)
3147 - kill link_set (use surf_network_model->resource_set instead)
3148 - Add a command-line option to choose the routing schema to use
3149 - Add three new models:
3150 * Floyd (shortest path computed at initialization)
3151 * Dijikstra (shortest path recomputed all the time)
3152 * Cached Dijikstra (shortest path computed on need)
3153 All these models where contributed by Silas De Munck, and are
3154 described in his ICCS09 paper.
3156 * Simplify model declaration
3157 (less redirections, less function to write when defining a model)
3158 - Factorize stuff between models:
3161 surf_model_resource_set(model)
3162 surf_model_resource_by_name(model, name)
3163 - Unify the types of models in s_surf_model_t (using an union)
3164 - Embeed fields of common_public directly into s_surf_model_t
3165 - Rename model methods:
3166 action_free ~> action_unref
3167 action_change_state ~> action_state_set
3168 action_get_state ~> action_state_get
3169 - Change model methods into functions:
3170 (model)->common_public->action_use ~> surf_action_ref
3172 * Implement a generic resource; use it as ancestor to specific ones
3173 (allows to kill duplicated code in models)
3174 Drawback: timer command don't need no name nor properties;
3175 workstation_CLM03 don't need no properties
3176 (but I guess we can live with those few bytes wasted)
3178 * Improve the action object model
3179 - implement a constructor avoiding dupplicated code about field
3180 initialization in generic_action part.
3182 * Kill the SDP model: it has an external dependency, is deprecated
3183 in flavor of modern lmm models, and didn't compile since a while
3186 * Relocation of the context module from XBT to SIMIX.
3187 (the context were decoupled from the simix processes, duplicating a lot of code)
3188 => a lot of code was factorized
3189 - less overhead is introduced during scheduling
3190 - simpler API for the context factory
3191 - the logic for process creation,destruction and manipulation was simplified
3192 * Simplification of the s_smx_process_t data structure.
3193 => accesing the simix level data associated to a process is faster now,
3194 and the code is a lot more readable.
3197 * Implement some more MPI primitives:
3198 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3199 -implementation: Bcast: flat or 2-ary tree (default),
3200 Barrier: 4-ary tree,
3202 Allreduce: Reduce then Bcast
3203 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3204 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3205 Alltoallv: flat tree, like ompi
3207 * Add support for optimized collectives (Bcast is now binomial by default)
3208 * Port smpirun and smpicc to OS X
3211 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3212 Shout out if you used it.
3215 * Display the list of still queued messages in SG mode when existing
3219 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3220 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3221 * Add xbt_str_from_file(FILE*)
3222 * Add xbt_dict_get_key achieving a linear reverse search
3223 * Remove the context module
3225 Portability report of this version:
3226 * Main portability targets:
3227 - Linux(debian)/x86/context
3228 - Linux(debian)/x86/pthreads
3229 - Linux(debian)/amd64/context
3230 - Linux(debian)/amd64/pthreads
3231 On these, we still have the eratic breakages of gras/pmm and
3232 amok/saturate_sg reported in previous version. We still think
3233 that the tests are the cause of the fault, not the tested code.
3235 - Mac OSX Leopard/x86/context
3236 Still false negative in tesh autotesting.
3237 Smpi still fails, but this time because readlink does not accept -f
3238 Everything seems to work properly beside of that.
3241 - AIX version 5.3 (only tested contexts this time)
3242 Smpi still fails there because mktemp is not installed.
3243 Everything seems to work properly beside of that.
3245 I managed to compile it for the first time, but several breakages.
3246 Won't delay the release for this exotic platform.
3248 * Windows: it's still lagging behind. If you want to help, please
3251 Timing report of this version:
3252 This version seem to be more than 5% faster than 3.3.1 (on linux
3253 64bits with contextes). The gain is less than expected, we are
3254 investigating this for next release.
3256 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3258 SimGrid (3.3.1) stable; urgency=low
3261 * Implement a --cfg-help to show existing configuration variables
3262 * Build chain do not require doxygen in maintainer mode
3265 * fix a bug on struct sizeof computation, which prevented the
3266 exchange of arrays of structs in some conditions
3267 - added a regression test about this in datadesc_usage
3268 * Allow the exchange of 0-long dynamic vectors.
3269 - for that, use -1 as indicator of dynamic size instead of 0
3270 - This implied to change any size from unsigned long to long,
3271 reducing a bit communication abilities, but I guess that with
3272 64bits being quite common, this is more than enough.
3273 - This also induce a protocol change, thus bumping network protocol
3274 version from 0 to 1 (if we have external users, we have to get
3275 clean on that point too ;)
3276 - added two regression tests about this in datadesc_usage
3277 * Be more verbose when propagating local exceptions
3278 This helps debugging.
3279 * Display the status of simulated processes when receiving SIGINT in
3283 * Allow to control the simulation from a trace file.
3284 New functions MSG_action_register() and MSG_action_trace_run()
3285 The first one allows to associate a function execution to each
3286 kind of action while the second one parses a trace file and
3287 triggers the corresponding actions within the system.
3288 For now, only a toy example is provided in examples/msg/actions
3289 * Add an example of process migration in examples/msg/migration
3290 * Fix a bug in task exchange which broke MSG_task_get_sender()
3291 Add a teshsuite regression test for that.
3292 [Bug: if MSG_task_get_sender() is called after sender exit,
3294 * Fix a bug which prevented suspend/resume to work properly
3295 * Display the status of simulated processes when receiving SIGINT
3296 This fixes a regression of v3.3. due to the introduction of SIMIX
3297 * Bug fixing in failure management:
3298 - trace could not start by a failure at time 0
3299 - failure during communications were not working
3302 * Add SIMIX_process_set_name() to change the name of the current
3303 process in the log messages.
3304 * Store smx_hosts in a dict since we only retrieve them by name
3305 * Move the configuration infrastructure to surf
3308 * Move the configuration infrastructure to surf
3311 * Massive internal cleanups:
3312 - Store internal structures on processes instead of hosts (allows
3313 to have more than one process per host, in addition of being more
3315 - Cleanup the initialization/finalization process
3316 - Kill a whole bunch of unneeded synchronization:
3317 processes run in exclusive manner within the simulator
3318 - Move queues from global tables to process data fields
3320 - now accept -platform and -hostfile arguments
3321 - Pass the right rank value to processes according to the hostfile
3322 * Compile the examples by default, and use them as regression tests
3323 * Implement MPI_Wtime()
3324 * Change the reference speed to a command line option
3327 * TCP_gamma can now be specified as command line option using
3328 --cfg=TCP_gamma:10000000.0
3329 * Change the --surf-path cmd line option into --cfg=path:
3332 * Also include strbuff from xbt.h public header
3333 * xbt_ex_display(): do not free the exception after displaying
3334 This allows to do more with the given exception afterward.
3335 Users should call xbt_ex_free() themselves.
3339 Portability report of this version:
3340 * Main portability targets:
3341 - Linux(debian)/x86/context
3342 - Linux(debian)/x86/pthreads
3343 - Linux(debian)/amd64/context
3344 - Linux(debian)/amd64/pthreads
3345 These targets fail about 1/10 of times on gras/pmm, but we believe
3346 that this is because of the test, not because of simgrid.
3347 amok/saturate_sg fails even more rarely, and the test may not be
3350 - Mac OSX Leopard/x86/context
3351 The test suite still spits tons of errors because some obscure
3352 force prevents us from removing the temporary directories
3353 arguing that they still contain some metadata I've never heard of.
3354 Smpi fails because seq is not installed.
3355 Everything seems to work properly beside of that.
3358 - AIX version 5.3 (both contexts and pthread)
3359 Smpi still fails there because mktemp is not installed.
3360 XML inclusions seems rosty on AIX.
3362 * Windows: it's still lagging behind. If you want to help, please
3365 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3367 SimGrid (3.3) stable; urgency=high
3371 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3372 [Malek Cherier & Mt]
3374 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3377 Until now, GRAS were using MSG as an interface to SURF. It was
3378 quite difficult because both interface have several differences
3379 (MSG channels vs GRAS sockets were the most notable point).
3381 This also opens the gate to SMPI (which should occur soon) and speed
3382 up simulations by to 40% (even if it were not the main goal).
3384 **************************************
3385 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3386 **************************************
3387 The point is that events occurring at the exact same timestamp are
3388 not scheduled in the same order with the old and new version. This
3389 may be enough to completely change the execution of simulations in
3390 some cases. Sorry for the inconvenience.
3392 * Cleanup and upgrade the XML format to push further scalability
3393 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3395 * Improve the testing infrastructure with tesh. Now a very large part of
3396 the code is tested not only by being run but also by checking that the
3397 output match an expected output [Mt].
3399 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3400 is really less memory-demanding, which should allow you to use
3401 larger files in SimGrid [AL].
3403 * Inform valgrind about our contextes, so that it becomes usable
3404 with the default (and more effecient) version of SimGrid
3405 [contributed by Sékou Diakite, many thanks]
3408 * Introduce a listener thread in charge of receiving incoming
3409 messages from the network. It allows to overlap communication and
3410 computation but most notably, it removes some stupid deadlocks due
3411 to the fact that so far, a process could not send and receive at
3412 the same time. This made most non trivial communication schema
3414 * Convert the PIDs from long int to int to match the MSG ones (and
3415 linux ones too) [Mt]
3416 * New function: gras_agent_spawn() to launch a new process on
3417 current host. Only working in simulation for now. [Mt]
3418 * New function: gras_os_hostport() returning a constant form (ie,
3419 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3422 * Make the backtrace of exceptions more human readable [Mt]
3423 * New module: xbt/str [Mt]
3424 a ton of string utility functions (split, join, printf to a newly
3425 allocated buffer, trim, etc)
3426 * New module: xbt/hash [Mt]
3427 SHA1 hashing algorithm (more to come if needed)
3428 * New module: xbt/synchro [Mt]
3429 synchronization tools (mutex and conditions) working the same way
3430 in simulation and in real life (mainly useful for GRAS, but not
3432 * New module: xbt/queue [Mt]
3433 classical producer/consumer synchronization scheme
3434 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3435 (using the classical functions will get serialized) [Mt]
3436 * Make dictionary internal table dynamic. No need to specify its size
3437 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3439 * Make sure the log channels are organized as a tree under windows
3440 (because of ANSI C compatibility issue, any channel were child of
3441 root directly) [Mt].
3444 * Cleaned many thing in surf and fixed a few bugs [AL].
3445 * Add a nice command line configuration mechanism to compose models [AL].
3446 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3447 the previous one (KCCFLN05). It relies on something that looks like
3448 a max-min sharing mechanism but cannot be written as such. A new solver
3449 was thus designed [AL].
3450 * Add a new solver to lmm. Based on Lagrange optimization and
3451 gradient-based descent, it enables to efficiently maximise systems s.a
3453 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3455 This solver enables to propose two new network models for TCP Reno and
3456 TCP Vegas based on Low's work. These models still need to be fully
3457 tested though [Pedro Velho].
3460 * Bug fix in SD_simulate. Now the time bound given as argument is
3462 * Use the new parallel task model (ptask_L07) as default.
3463 * Use the SURF command line configuration mechanism.
3464 * 0-size tasks (for synchronization) should now work.
3466 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3468 SimGrid (3.2) stable; urgency=high
3472 We still experience issues on this platform, but we believe that at
3473 least MSG is usable.
3475 GRAS API BREAKAGE (for simplification purpose, sorry):
3476 * the gras_msgtype_by_name is not used anymore. Instead of
3477 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3478 you can write (and must)
3479 gras_msg_send(toserver, "request", &request);
3480 - If you still want to pass a gras_msgtype_t to the function (to cache
3481 the type and avoid the lookup time), use the gras_msg_send_() variant.
3482 - Impacted functions:
3483 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3484 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3485 * The callbacks are now expected to return 0 when everything went well
3486 (just like the main() function)
3488 GRAS new features and improvements:
3489 * New module mechanism where user code can use per process globals [Mt]
3490 This is similar to gras_userdata_*() functions, but for libraries. It
3491 factorize some code developped over and over in the examples and AMOK.
3492 It has still to be documented and used (only amok/peermanagement is
3494 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3495 loss when they were small enough to fit into the buffer and sent quickly
3496 enough so that they can all get received in one shoot.
3497 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3498 if not found. Use the *_or_null() variant for the old semantic.
3499 * In gras_msg_handle, do not discard messages without callback.
3500 They are probably messages to be explicitly awaited later (ie, proofs of
3501 mis-synchronization in userland since they are sent before being awaited)
3502 No big deal usually.
3503 * gras_socket_meas_send/recv: semantic changed!
3504 The numerical arguments used to be (1) the total amount of data to send
3505 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3506 messages. This was need for the fool willing to send more than MAXINT
3507 bytes on quite fat pipes.
3510 * Do really rename the hostmanagement module to peermanagement. [Mt]
3511 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3512 but this is rather new and this was documented in the module
3513 documentation (poor excuses, I admit)
3514 * Bandwidth measurement semantic changed! This follows the changes to
3515 gras_socket_meas_send/recv explained above.
3518 * A sequential mode has been added to the workstations. When a workstation
3519 is in sequential mode, it can execute only one task, and the other tasks
3520 are waiting in a FIFO. [Christophe Thiery]
3523 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3524 model for SIMDAG. [Christophe Thiery]
3525 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3526 equal to 0, which caused some bad behaviors in
3527 saturated_constraint_set_update. I now use a threshold mechanism like in
3531 * When running manually src/testall, you select specific units [Mt]
3532 testall is the result of our cunit mechanism, and should replace all
3533 the scripty thingy around since bash don't run easily on billware.
3535 * A mallocator system has been added. [Christophe Thiery]
3536 Mallocators allow you to recycle your unused objects instead of freeing them
3537 and allocating new ones.
3539 Documentation update:
3540 * FAQ reworking + New FAQs:
3541 - "Valgrind spits tons of errors!" [Mt]
3542 - "How to repport bugs" [Mt]
3543 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3544 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3545 - Communication time measurement within MSG [AL]
3546 - I experience weird communication times when I change the latency [AL]
3547 * GRAS tutorial [Mt]
3549 - an introduction to the framework and to the used communication model
3550 - an initiatic tour introducing the most proheminent features:
3552 . Lesson 0: Installing GRAS
3553 . Lesson 1: Setting up your own project
3554 o Part 2: Message passing
3555 . Lesson 2: Exchanging simple messages
3556 . Lesson 3: Passing arguments to the processes (in SG)
3557 . Lesson 4: Attaching callbacks to messages
3558 . Lesson 5: Using globals in processes
3559 . Lesson 6: Logging information properly
3560 . Lesson 7: Using internal timers
3561 . Lesson 8: Handling errors through exceptions
3562 . Lesson 9: Exchanging simple data
3563 . Lesson 10: Remote Procedure Calling (RPC)
3564 . Lesson 11: Explicitly waiting for messages
3565 . Recapping of message passing features in GRAS
3566 - A HOWTO section containing:
3567 o HOWTO design a GRAS application
3568 More are due, of course. They will come latter. In the meanwhile, you can
3569 check the examples which are still here.
3571 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3573 SimGrid (3.1) stable; urgency=high
3577 There was a stack corruption somewhere, visible only when optimizing
3578 with these versions. [Vince]
3581 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3582 module that enabled to play with Directed Acyclic Graphs. It is built
3583 directly on top of SURF and provides an API rather close to the old
3584 SG. Some old codes using SG are currently under rewrite to check that
3585 all needful functions are provided. [Christophe Thiery]
3588 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3589 extension of the classical CLM03 model that gracefully handles
3590 failures. This is now the default model for MSG and GRAS. It doesn't
3591 handle parallel tasks yet though. [AL]
3592 * Bug fix: Weights were not correctly set in the network part.
3593 WARNING: This may have resulted in incorrect results with simulations
3594 where there are more than one flow on a given link. [AL]
3597 * After a (long ?) discussion on simgrid-devel, we have decided that the
3598 convention we had on units was stupid. That is why it has been decided
3599 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3600 WARNING: This means that all previous platform files will not work as
3601 such with this version! A warning is issued to ask users to update
3603 A conversion script can be found in the contrib module of the CVS, under
3604 the name contrib/platform_generation/surfxml_update.pl [MQ]
3607 * Bug fix: Processes were started in reverse order, wrt deployment file.
3608 WARNING: if your code relies on this bug, please fix it. [AL]
3609 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3610 being executed on two different locations. [AL]
3611 * Bug fix: Failures are now better supported thanks to Derrick's tests
3612 (there was many failure situations I hadn't thought of and that weren't
3613 correctly handled). [AL]
3614 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3618 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3619 Exception occurring on server-side are propagated back to client (!).
3621 API CHANGE: the callback changed their prototype. Change:
3622 int my_handler(gras_socket_t expeditor, void *payload_data) {
3624 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3625 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3627 * New! function: gras_msg_handleall to deal with all messages arriving
3628 within a given period.
3629 * New! function: gras_socket_server_range to get a server socket in a
3630 range of port numbers (ease to avoid port number conflicts) [MQ]
3631 * New! gras processes display their backtrace when they get a SIGUSR1
3632 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3633 Sweet to debug RL processes [MQ]
3637 - Do not force experiment sizes to be expressed in kb, or it becomes
3638 impossible to measure the latency this way (needs one byte-long tests)
3639 WARNING: this changes the amok_bw_* function semantic. [MQ]
3640 - Implements the link saturation stuff. [MQ]
3641 * Peer management module:
3642 New! module factorizing code that we wrote over and over [MQ].
3645 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3646 - Test units are placed directly into the library code, they get extracted
3647 automatically and placed into the src/testall binary.
3648 - Convert most of the XBT tests to this system.
3649 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3650 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3651 live in the same C code. This required to change a little bit the API
3652 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3653 * New module: sparse graph structure with basic algorithms (this is work
3654 in progress and the API is not considered to be frozen yet). [AL]
3655 * Display more information on backtraces: source line & function names are
3656 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3657 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3658 * Reworked a little bit some #include statements to load only required
3659 headers. Some user code that relied on SimGrid to include stdlib or
3660 stdio may need to include it by themselves. [AL]
3661 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3663 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3664 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3665 but I guess that almost nobody used those parts. [MQ]
3667 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3669 SimGrid (3.0.1) stable; urgency=low
3672 * Unfortunately, I had missed 5 misnamed functions:
3673 xbt_fifo_item_t xbt_fifo_newitem(void);
3674 void xbt_fifo_freeitem(xbt_fifo_item_t);
3675 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3676 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3677 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3678 They're now deprecated. Please use their new versions:
3679 xbt_fifo_item_t xbt_fifo_new_item(void);
3680 void xbt_fifo_free_item(xbt_fifo_item_t);
3681 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3682 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3683 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3685 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3686 * Documentation: xbt_log module unmercifully reworked [MQ]
3687 * Bugfix: there was a problem with the ending of contexts with
3688 the pthread backend. It caused some weird deadlock or behavior
3689 depending on the pthread implementation. [AL]
3690 * Bugfix: get the exceptions raised in the simulator repport where
3691 and why they come from when they are not caught in time [AL, MQ]
3694 * Bugfix: Do repport the error when two non-connected hosts try to
3695 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3698 * Add additionnal checkings on communications. Assert that two
3699 communicating hosts are connected by a set of links... [AL]
3702 * Add additionnal checkings on channel values in communication [AL]
3703 * New: MSG_task_get_source to see on which host a task was generated [HC]
3704 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3705 the number of tasks waiting to be received on channel and sent
3707 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3708 waits for the first task coming from a given host.. [AL]
3710 GRAS new functionnalities: [MQ]
3711 * Enhance the parsing macro to allow the size of multidimentional objects
3712 to be given thru annotations.
3713 * New example (and documentation): Matrix Multiplication a la RPC
3714 (as when I was young!) and fix a bunch of bugs found on the way.
3716 GRAS performance improvements: [MQ]
3718 * Reduce the amount of cbps creation/destruction by making it static to
3719 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3722 * Change libdata to a set so that we can search for stuff by ID (and thus
3723 reduce the insane amount of dict lookups)
3726 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3727 RL communication on top of SG since it's so uneffective.
3728 It may also be used for inter-thread communication in RL, one day.
3729 * Use gras_datadesc_copy() to exchange messages on top of SG
3731 - improve message exchange performance on top of SG
3732 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3733 * Don't exchange on the network the size of the used part of buffer,
3734 instead, specify the possible buffer size to read().
3736 - reduces the amount of read/write calls (one pair per exchange)
3737 - reduces the amount of exchanged data (the size)
3738 - allows to retrieve all arrived data on receiver side, if we don't need
3739 it right now (subsequent read will peek the buffer)
3740 - allows the receiver to proceed with the beginning of the stream before
3741 everything is arrived
3742 - make it possible to build an iov transport (using readv/writev)
3744 - take care of the data with non-stable storage (like stacked data),
3746 * If possible, TCP send uses vector I/O (when writev() is here)
3747 - Don't use it for receive since we send data sizes and data on the
3748 same stream, so we wouldn't be able to chain large amount of chunks
3749 before having to flush the stuff to read the size.
3750 * Rework the transport plugin mechanism to simplify it and reduce the
3751 amount of pointer dereferencement when searching for the right function
3754 * I guess that now, we do almost as few system calls as possible while
3755 doing as few data copy as possible.
3757 To improve it further, we could try to send all the sizes first and then
3758 all the data (to use iov on receiving size), but it's only a partial
3759 solution: when you have 2 dimensional data, the sizes of the second
3760 dimension is data of the first dimension, so you need 3 streams.
3762 I'm not sure the potential performance gains justify the coding burden.
3764 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3766 SimGrid (3.00) stable; urgency=high
3769 * New! Give the possibility to hijack the surf parser and thus bypass
3770 MSG_create_environment and MSG_launch_application. Have a look at
3771 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3773 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3775 SimGrid (2.96) unstable; urgency=low
3780 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3781 This deprecates the xbt_error_t mechanisms.
3782 It modifies (simplifies) all XBT and GRAS API.
3783 MSG API keeps unchanged (exceptions raised by XBT are caught from
3784 within MSG and masked with existing error handling facilities)
3787 * New! Add a FATPIPE model. [AL]
3788 * New! Add a parallel task model. [AL]
3789 * New! Add automatically a loopback interface (in the default
3790 network model) if none was precised.
3793 * Bugfix: MSG_process_resume now works with the current running process.
3795 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3796 * Modification of MSG_task_get_compute_duration. Once a task has been
3797 processed, the value returned by this function is now equal to 0. [AL]
3798 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3799 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3800 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3801 of a m_task. For now, MSG_task_cancel only works with computation
3803 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3804 of the processor (in Mflop/s) regardless of the current load on the
3806 * API Change: use proper naming convention for MSG_getClock and
3807 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3809 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3810 This function changes the priority of a computation task. This priority
3811 doesn't affect the transfer rate. A priority of 2 will make a task
3812 receive two times more cpu power than the other ones. This function
3813 has been added to suit the needs of Nguyen The Loc and hasn't been that
3814 much tested yet. So if it fails, please report it and send me your code.
3816 * API Change: removed all functions and types that were marked "deprecated"
3817 since many months. Renamed MSG_global_init_args to MSG_global_init.
3819 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3821 SimGrid (2.95) unstable; urgency=low
3824 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3825 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3826 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3829 * Bugfix: Allow absolute paths for platform description files [MQ]
3830 * Bugfix: do free the variables after use. Leads to drastic performance
3832 * Implement max_duration (ie, timeouts) on resources [AL]
3835 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3837 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3838 the needed support (timeouts) [AL]
3841 * Implement measurement sockets. You can now get the bandwidth between two
3842 hosts thanks to AMOK (see below). [MQ]
3843 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3844 dynar over the network (yeah) [MQ]
3845 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3846 * simple/static token-ring example. [Alexandre Colucci and MQ]
3847 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3848 in place before [MQ]
3851 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3852 type parsing. "short" and "long" modifiers were also ignored; other
3853 modifier (such as reference level) are still ignored. [MQ]
3854 * Embeed the buffer size within the buffer itself on SG. [MQ]
3855 That way, send() are atomic and cannot get intermixed anymore (at least
3856 the ones which are less than 100k; bigger messages still have the issue)
3857 * Array size pushed by the field, not by the field type (or each
3858 and every long int will push stuff to the cbps) [MQ]
3859 * use select() to sleep since it allows to portably sleep less than one
3862 GRAS (minor cleanups)
3863 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3866 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3867 * type category 'ignored' killed as it was never used and were difficult
3869 * whether a type can cycle or not is now a flag, leaving room for more
3870 flags (as "ignored", if we feel the need one day ;)
3871 * Rename raw sockets to measurement sockets since "raw" has another
3872 meaning in networking community.
3875 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3876 and offers features not belonging to GRAS but that most applications
3877 need. One day, it may be a set of plugins loadable at runtime.
3878 * New module: bandwidth
3879 bandwidth measurement between arbitrary nodes running this module. [MQ]
3881 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3883 SimGrid (2.94) unstable; urgency=low
3885 The first beta release of SimGrid 3 !
3889 * Update the main page and the FAQ. Adding references to gforge.
3892 * Add a gras_os_getpid function.
3895 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3896 * Extend the logs so that they also print PID, hostname, date, ... if
3898 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3899 and kill the old version which were in testsuite/
3900 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3904 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3905 for users. Instead, each programming environment (GRAS, MSG,...) use it
3906 when needed to provide such a feature to users.
3907 Don't shortcut the mechanism or you will also shortcut the virtualization
3908 you need on the simulator.
3912 * Cleanups in configury with regard to compile optimization/warning flags.
3913 Also add -fno-loop-optimize to any powerpc since it's the optimization
3914 killing gcc (< 3.4.0).
3915 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3916 needed by MSG examples complications
3917 * Borrow configury beautifications from PHP
3920 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3921 warning (thanks loris for stumbling into this one).
3922 * Bugfix: stop loading private headers (gras_config.h) from the public
3926 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3927 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3928 RL would imply a DNS resolver, which is *hard* to do in a portable way
3929 (and therefore delayed).
3930 * Implement a real timer mechanism and use it in timing macros. This allows
3931 to avoid rounding errors and get a 0.000005 sec precision in timing
3932 macros. While I was at it, various cleanups:
3933 - allow to declare more than one timed section per file (fix a stupid bug)
3934 - move some private declaration to the right place
3935 - merge conditional execution and timing macros into emulation module
3936 - document the module
3937 - make sure the module cleanups its mess on gras_exit
3938 * Documentation improvements:
3939 - (new) how to compile applications using GRAS
3940 - (new) emulation support (timing macros)
3942 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3944 SimGrid (2.93) unstable; urgency=low
3946 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3949 - Use Paje properly where used. Still to be sanitized properly.
3950 - Portability fix: Add an implementation of the contexts using pthread
3954 - Add xbt_procname(): returns the name of the current process.
3955 Use it to show the current process's name in all logging.
3957 - fix detection of older flex version and the reaction, since we do
3958 depend on modern ones (we use lex_destroy)
3959 - Better separation of SG and RL in the libs: remove all simulation code
3960 from libgras. As a result, this lib is now only 200k when stripped.
3961 Some of the xbt modules may also be duplicated (two sets and such) and
3962 should be cleaned/killed before SG3.
3963 - Insist on using xlC on AIX because of weird problems involving gcc there.
3964 - Cleanup the make remote stuff. This is now done by scripts
3965 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3966 mainly for our private use, but we're working on changing them to user
3969 - Bugfix: flush the socket on close only if there is some *output*.
3970 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3972 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3975 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3977 SimGrid (2.92) unstable; urgency=low
3979 Alpha 3 on the path to SimGrid 3
3983 - New! First try of benchmarking macros.
3984 - New! First try so that gras_stub_generator generate deployment and
3985 remote compilation helpers.
3987 - Bugfix: Initialization fix in msg_test.
3991 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3993 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3995 - API Change: gras_os_sleep to take the amount of seconds as a double.
3996 Accepting an int was error prone since it was the only location where
3997 seconds were coded as such. It leaded to damn rounding errors.
3998 - Bugfix: Hard to belive that timers ever worked before this.
4000 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
4002 SimGrid (2.91) unstable; urgency=low
4004 Alpha 2 on the path to SimGrid 3
4008 - Bug fix in the lmm_solver.
4010 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
4011 through the function MSG_paje_output.
4012 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
4013 - It is possible to bound the rate of a communication in MSG with
4014 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4015 to write it in the changelog).
4016 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4021 - Various cleanups to the autotools stuff
4022 - Begin to move Gras examples to examples/gras/
4023 - Let make distcheck work again (yeah!)
4025 - documentation overhauled using doxygen.
4026 gtk-doc-tools is dead in SimGrid now.
4027 - Automatically extract all existing logging categories, and add the list
4028 to the documentation (long standing one, to say the less)
4030 - Cleanup the known architecture table. Reorder the entries to group what
4031 should be, and use a more consistent naming scheme.
4032 (some of the test dataset are still to be regenerated)
4033 - New! Allow library to register globals on each process just as userdata
4035 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4036 lookup time (for now).
4037 Use it in msg and trp.
4038 This cleans a lot the internals and helps enforcing privacy of the
4039 headers between the gras components.
4040 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4041 - Bugfix: gras_os_time was delirious in RL.
4042 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4044 - Reenable GRAS now that it works.
4046 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4048 SimGrid (2.90) unstable; urgency=low
4050 Alpha 1 on the path to SimGrid 3
4052 * It is a long time since the last release of SimGrid. I'm sorry about
4053 that but as I had told you, I was rewriting a lot of things. I apologize
4054 to those who had been reporting bugs to me and that I had not answered.
4055 If your bug is still in the new version, please tell me. Here is a
4056 summary of the main changes.
4058 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4059 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4060 portability, speed, and a lot more but you'll figure it out later.
4061 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4062 I wanted to release the new MSG as soon as possible and I have
4063 broken GRAS, which is the reason why, for now, only MSG is fully
4064 functional. A laconic description of these projects is available
4065 in the documentation.
4067 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4068 kernel optimized for our needs (called SURF but only the developers
4069 should use it). Hence, MSG is now roughly 30 times faster and I think
4070 that by rewriting a little bit MSG, I could event speed it up a little
4071 bit more. Beside the gain in speed, it is also much easier to encode a
4072 new platform model with SURF than it was with SG. More to come...
4074 * REVOLUTION 3: I have tried to change a little as possible the API of
4075 MSG but a few things really had to disappear. The main differences
4076 with the previous version are:
4077 1) no more m_links_t and the corresponding functions. Platforms are
4078 directly read from an XML description and cannot be hard-coded
4079 anymore. The same format is used for application deployment
4080 description. The new format is described in the documentation.
4081 Have a look in tools/platform_generation. There is a tiny script
4082 that converts from the old platform format to the new one. Concerning
4083 the application deployment format, parsing the old one is tricky.
4084 I think most of you should however be able to convert your files. If
4085 it is really an issue, I can write a C code that does the conversion.
4087 2) the toolbox tbx does not exist anymore. We now have a library
4088 with much more data-structures but without the hash-tables (we have
4089 dictionaries that are much faster).
4091 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4093 *****************************************************************************
4094 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4095 * don't feel like converting the oldies. (Mt) *
4096 *****************************************************************************
4099 Version 2.90: "the long awaited one"
4100 - Finished rewriting and debugging MSG. Rewrote the documentation.
4101 - disable GRAS for now since it needs to be ported to the newest SG
4104 - Finish the port to windows (using mingw32 for cross-compile)
4107 - Main loop and datastructures of SURF. A cpu resource object is
4108 functional. Surf can thus be used to create cpu's with variable
4109 performance on which you can execute some actions.
4111 2004-11-15 Martin Quinson
4112 - Port to ARM. Simply added the alignment and size descriptions. Should
4113 work, but the ARM machines are so slow that I didn't had the opportunity
4114 to 'make check' over there yet.
4116 2004-11-15 Arnaud Legrand
4117 - Trace manager now written. It uses a heap structure and is therefore
4118 expected to be efficient. It may however be speeded up (particularly
4119 when many events occur at the same date) by using red and black
4120 trees. One day maybe...
4121 - Max-min linear system solver written. It uses a sparse matrix
4122 structure taking advantage of its expected use. Most operations are
4123 O(1) and free/calloc are called as few as possible. The computation of
4124 the minimum could however be improved by using a red and black tree
4127 2004-11-03 Arnaud Legrand
4128 - Rename every gras_* function that was in xbt/ to its xbt_
4130 - Add a heap and a doubly-linked list to xbt
4131 - Added a dichotomy to the dictionaries. make check works as well before
4132 so I assume that the patch is correct. I do not know however if things
4133 run effectively faster than before now. :)
4135 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4136 SimGrid, and the version number is bumped to 2.x
4138 2004-10-29 Martin Quinson
4139 - Introduction of the remote errors.
4140 They are the result of a RMI/RPC on the remote machine.
4141 ErrCodes being scalar values, you can't get the host on which those
4142 errors did happen. Extending the error mechanism as in Gnome is possible.
4143 No idea yet whether it is a good idea.
4145 2004-10-28 Martin Quinson
4146 - Interface revolution: the Starred Structure Eradication.
4147 I used to do typedef struct {} toto_t; and then handle *toto_t.
4148 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4149 - ???_t is a valid type (builded with typedef)
4150 - s_toto_t is a structure (access to fields with .)
4151 - s_toto is a structure needing 'struct' keyword to be used
4152 - e_toto_t is an enum
4153 - toto_t is an 'object' (struct*)
4155 typedef struct s_toto {} s_toto_t, *toto_t;
4156 typedef enum {} e_toto_t;
4157 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4158 s_toto_t) is private.
4160 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4161 so that it can set the variable to NULL. It was so for dicts and sets,
4162 it changed for dynars.
4164 - Fix a bunch of memleaks in dict_remove
4165 - Fix a bug in sg/server_socket opening: it failed all the time.
4167 2004-10-07 Martin Quinson
4168 - Speed up dynar lookup operation a bit.
4170 gras_dynar_get is dead.
4172 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4173 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4174 (which gives you the address of the stored data).
4176 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4177 copy of the data using an affectation to do the job and not a memcpy.
4179 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4180 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4182 It does not really speedup the dynar test because they are
4183 setting elements all the time (and look them seldom). But the dict does
4184 far more lookup than setting.
4186 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4188 2004-10-05 Martin Quinson
4189 - Allow to (en/dis)able the cycle detection at run time.
4191 Whether we should check for cycle or not is now a property of each
4192 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4193 datadesc_cycle_unset allow to remove this property when previously set.
4195 Note that the cycle detection is off by default since it impacts the
4196 performance. Watch the data you feed GRAS with ;)
4198 This property is hereditary. Any element embedded in a structure having it
4199 set have it set for the time of this data exchange.
4201 You should set it both on sender and receiver side. If you don't set it on
4202 sender side, it will enter an endless loop. If you forget on receiver
4203 side, the cycles won't be recreated after communication.
4205 - Header reorganization.
4206 Kill gras_private.h, each submodule must load the headers it needs.
4208 2004-10-04 Martin Quinson
4209 - Interface revolution: do not try to survive to malloc failure.
4211 Now, gras_malloc and friends call gras_abort() on failure.
4212 As a conclusion, malloc_error is not a valid error anymore, and all
4213 functions for which it was the only gras_error_t return value are
4214 changed. They now return void, or there result directly.
4215 This simplify the API a lot.
4217 2004-09-29 Martin Quinson
4218 - Re-enable raw sockets.
4219 Created by gras_socket_{client,server}_ext;
4220 Used with gras_raw_{send,recv}
4223 It should allow to kill the last bits of gras first version soon.
4225 This is not completely satisfactory yet (duplicate code with
4226 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4229 - Simplify transport plugin (internal) interface by not passing any
4230 argument to _server and _client, but embedding them in the socket
4233 2004-09-28 Martin Quinson
4234 - Finish the port to AIX.
4235 autoconf was my problem (segfault within the malloc replacement
4236 function. No idea why)
4238 2004-09-16 Martin Quinson
4239 - Fix some size_t madness on 64bit architectures.
4241 2004-09-08 Martin Quinson
4242 - Reduce the number of system headers loaded, overload some more system
4243 calls (such as malloc to cast the result of the system one, and work
4245 - Fix and reintroduce the config support
4247 2004-09-07 Martin Quinson
4248 - Source code reorganization to allow Arnaud to surf all over there.
4249 - Allow to document the logging categories.
4250 - Remove all uppercase from logging categories and useless cleanup in names.
4252 2004-08-18 Martin Quinson
4253 Version 0.6.2 (protocol not changed; API changed)
4254 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4255 gras_error_t), and NULL when not found. Functions expecting a msgtype
4256 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4257 hopefully usefull message.
4258 - Portability to prehistoric sparcs again
4260 2004-08-17 Martin Quinson
4261 Version 0.6.1 (protocol not changed; ABI not changed)
4262 - prealloc some buffers to speed things up
4264 2004-08-11 Martin Quinson
4265 Version 0.6 (protocol not changed; ABI expended)
4266 - The parsing macro can deal with the references, provided that you add
4267 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4269 2004-08-09 Martin Quinson
4270 Version 0.5 (protocol not changed; ABI changed)
4271 - Allow to off turn the cycle detection code in data exchange at
4272 compilation time. It should be at run time, but I'm short of time (and
4273 the config stuff is still broken). That way, we keep dict out of the
4274 critical path, which is good because the performance is poor:
4275 - search not dichotomial yet
4276 - dynar give no way to access their content and memcpy everytime
4277 - In composed data description (struct, ref and so on), stop foolness of
4278 keeping the subtype's ID, but store the type itself. This keeps sets out
4279 of the critical path, which is good since they rely on dynar and
4280 dictionaries. The only loose of that is that we cannot detect the
4281 redeclaration of a structure/union with another content (but I'm not sure
4282 the code detected well this error before anyway). We still can detect
4283 the redefinition discrepancy for the other types.
4284 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4285 it breaks the code because of type-punning used all over the place).
4286 This breaks on all non-gcc architectures (for now).
4288 All those changes (plus the buffer of last time) allow me to gain 2 order
4289 of magnitude on cruel tests consisting of 800000 array of integers on two
4290 level of a hierarchical structure (200 secondes -> 4 secondes)
4293 - the selector of reference must now return the type it points to, not
4294 the ID of this type.
4296 2004-08-06 Martin Quinson
4297 Version 0.4 (protocol changed; ABI not changed)
4298 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4299 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4300 - Implement a buffer transport, and use it by default (it relies on tcp in
4301 real life and on sg in simulation).
4302 That's a bit hackish since I had a new field to the structure to store
4303 its data without interfering with the subtype ones. Inheritance
4304 is tricky in C. And that's a kind of reverse inheritance with one class
4305 derivating two classes. Or maybe a game with java interfaces. Anyway,
4306 that's damn hard in C (at least).
4307 Moreover, I got tired while trying to ensure plugin separation and
4308 genericity in SG mode. MSG wants me to do weird things, so let's go for
4309 cruel hacks (temporarily of course ;).
4310 See comment in transport_private.h:71
4311 - do not include all the _interface headers in private but in the files
4312 which really need them (to cut the compilation time when they are
4315 2004-07-26 Martin Quinson
4316 Version 0.3 (protocol not changed; ABI changed)
4317 - Major overhault of the datadesc interface to simplify it:
4318 - shorted the function names:
4319 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4320 - add a trivial way to push/pop integers into the cbps without malloc.
4321 This allows to make really generic sub_type description, which simply
4322 pop their size of the stack.
4323 - add a function gras_datadesc_ref_pop_arr() which does what users want
4324 most of the time: Declare a dynamic array (which pops its size of the
4325 stack) and declare a reference to it. Poor name, but anyway.
4326 - kill the post-send callback, add a post-receive one
4328 2004-07-23 Martin Quinson
4329 Version 0.2 (protocol changed; ABI changed)
4330 - add some testing for cpbs in the test cases, and fix some more bugs.
4331 This invalidate again the little64 data file, since I cannot regenerate
4333 - remove an awfull optimization in the logging stuff, allowing me to:
4334 - understand it again
4335 - learn gcc how to check that the argument match the provided format
4336 - fix all errors revealed by gcc after that
4337 - internal keys of dict are not \0 terminated. Deal with it properly in
4338 loggings instead of segfaulting when the user want to see the logs :-/
4340 2004-07-22 Martin Quinson
4341 - Fix some stupid bug preventing cbps (callback postit) from working
4343 2004-07-21 Martin Quinson
4344 - Some documentation cleanups
4345 - remove the useless last argument of msgtype_declare
4346 - rename the Virtu functions to fit into the 'os' namespace
4347 - move headers src/include -> src/include/gras/ and stop fooling with
4348 gras -> . symbolic link
4349 - make distcheck is now successful
4351 2004-07-19 Martin Quinson
4353 - Build shared library also
4354 - Install html doc to the right location
4355 - stop removing maintainer files in make clean
4356 - build tests only on make check
4358 2004-07-13 Martin Quinson
4360 - No major issue in previous version => change versioning schema
4361 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4362 corresponding dataset.
4364 2004-07-11 Martin Quinson
4366 - Get it working with any kind of structure (we can compute the padding
4367 bytes remotely for all the architectures I have access to)
4368 - Implement the structure parsing macro (still not quite robust/complete)
4369 - Improvement to the remote testing toysuite
4371 2004-07-10 Martin Quinson
4372 [autoconf mechanism]
4373 - get ride of a bunch of deprecated macros
4374 - actually run the test for two-compliment, not only compile it :-/
4375 - test whether the structures get packed (and bail out if yes. Damn.
4376 Alignment is a serious matter)
4377 - test whether the structures get compacted (but respecting the alignment
4378 constraints of each types)
4379 - test whether the array fields of structures can straddle alignment boundaries
4381 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4382 (took me 2 hours to find that bug, looking at the wrong place)
4383 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4384 before sending/receiving and are used to compute the offsets of fields
4385 - Given that padding size depend even on compiler options, keep track of
4386 alignment and aligned_size only for the current architecture. Not a big
4387 deal since we send structure fields one after the other (seems
4389 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4390 program, let it work on linux/gcc/little32. portability todo.
4392 2004-07-08 Martin Quinson
4393 - import and improve remote compilation support from FAST
4394 - make sure make check works on half a dozen of machines out there
4396 2004-07-07 Martin Quinson
4397 Let's say it's version 0.0.3 ;)
4398 - Implement conversions (yuhu!)
4399 - Let it work on solaris (beside conversion, of course)
4400 - Stupid me, using rand() to generate the conversion datatests in not wise.
4402 2004-07-06 Martin Quinson
4403 - Let make dist work, since I'm gonna need it to compile on remote hosts
4404 - Let Tests/datadesc_usage write the architecture on which the file was
4405 generated as first byte.
4406 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4407 architecture descriptions.
4408 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4409 on those architectures.
4410 - Optimization: send/recv array of scalar in one shoot
4412 2004-07-05 Martin Quinson
4413 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4415 - Plug a whole bunch of memleaks
4416 - each process now have to call gras_{init,exit}. One day, their log
4417 settings will be separated
4418 - Continue the code factorisation between SG, RL and common in Transport.
4420 2004-07-04 Martin Quinson
4422 - Redistribution between SG and RL.
4423 We wanna have to accept in SG, so move accepted related parts of RL in
4424 the common part. (more precisely, the dynar of all known sockets is no
4425 more a static in transport.c, but part of the process_data)
4427 [gras_stub_generator]
4428 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4430 2004-07-03 Martin Quinson
4431 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4432 them all up in one shot)
4434 - Fix the ugly case of reference to dynamic array.
4435 - New (semi-public) function gras_datadesc_size to allow the messaging
4436 layer to malloc the needed space for the buffer.
4438 - gras_socket_close now expect the socket to close (and not its address to
4439 put NULL in it after it). This is because the socket passed to handlers
4440 is one of their argument (=> not writable).
4442 - propagate the interface cleanup from last week in datadesc, ie remove a
4443 superfluous level of indirection. User pass adress of variable
4444 containing data (both when sending and receiving), and not of a variable
4445 being a pointer to the data. Let's say that I like it better ;)
4446 The price for that is constructs like "int msg=*(int*)payload" in
4447 handlers, but it's a fine price, IMHO.
4449 - Let it work in RL (yuhu)
4451 2004-06-21 Martin Quinson
4453 - porting SG plugin and SG select to new standards (works almost).
4454 - plug memleaks and fix bugs around.
4457 - cleanup the prototype of data recv and force users to specify when they
4458 want to handle references to objects. Test case working even for cycles.
4459 - plug memleaks. Valgrind is perfectly ok with this.
4461 2004-06-12 Martin Quinson
4463 - cleanup the separation between plugin and main code in plugin creation
4465 2004-06-11 Martin Quinson
4467 - Reput hook for raw sockets, needed for BW experiments
4468 - kill a few lines of dead code
4469 [Data description] Interface cleanup
4470 - gras_datadesc_by_name returns the searched type or NULL.
4471 That way, no variable is needed to use a type desc once, which makes
4473 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4474 two parameters were strings (field name, type name), leading to
4476 [Dicos] Interface cleanup
4477 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4478 This is consistant with the dynar API.
4480 2004-04-21 Martin Quinson
4482 - Porting to new standards.
4484 - interface cleanup.
4485 There is no bag anymore, no need to take extra provision to mask the
4486 pointers behind "ID".
4487 Better splitup of functions between files create/exchange/convert.
4488 This is still a bit artificial since convert and receive are so
4489 interleaved, but anyway.
4491 - add a queued message list to procdata (the ones not matching criteria
4493 - factorize some more code between SG and RL wrt procdata
4495 - use gras_exit in example to track memleaks
4496 - get rid of gs_example now that GS is properly integrated into gras
4497 - update run_test to integrate the latest tests (datadesc)
4499 - rename WARNINGn macros to WARNn since it prooved error-prone
4501 2004-04-19 Martin Quinson
4503 - register init/exit functions within gras module mechanism
4504 - send/receive function.
4505 Convertion is not implemented, but short-cutted if not needed.
4506 struct/array elements are sent one by one (instead of block-wise), but
4507 nobody really cares (yet). Get a prototype before optimizing.
4508 - tests (using a file socket) for DD send/receive on:
4509 - base types: int, float
4510 - array: fixed size, string (ie ref to dynamic string)
4511 - structure: homogeneous, heterogeneous
4512 - chained list, graph with cycle
4513 Believe it or not, valgrind is not too unhappy with the results. The
4514 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4518 - Bugfix when using a filename explicitly (instead of '-')
4520 2004-04-09 Martin Quinson
4522 - factorize more code between RL and SG in socket creation
4523 - Complete the implementation and tests of:
4525 o file (only in RL, and mainly for debugging)
4527 I lost 3 days to design a portable address resolver, and then decided
4528 that the prototype mainly have to run on my box.
4529 Addressing portability too early may be like optimizing too early :-/
4531 - use gras_init in the Tests instead of the crappy parse_log_opt
4532 (the latter function is removed)
4533 [Conditional execution]
4534 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4535 [Code reorganisation]
4536 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4537 Build examples against the RL library, since there is no way to disable
4538 its creation for now.
4540 For information, the beginning of coding on GRAS was back in june
4541 2003. I guess that every line has been rewritten at least twice since