1 ----------------------------------------------------------------------------
3 SimGrid (3.28.1) NOT RELEASED YET (v3.29 expected September 22. 2021, 19:21 UTC)
6 - Non-linear resource sharing: allows to dynamically change the resource
7 capacity. Users can change this capacity by setting a callback which will
8 be called when SimGrid is sharing the resource between the active
9 activities. An activity can be a task running, a read/write IO operation or a
11 - Callback signature: double (double capacity, int n_activities): given the
12 current resource's capacity and number of activities running on it,
13 returns the new capacity.
14 - Note that this callback is in the critical path of the system solve and
15 should not take long to execute.
17 - Disk: examples/cpp/io-degradation
18 - Link: examples/cpp/network-nonlinear
19 - CPU: examples/cpp/exec-cpu-nonlinear
20 - Dynamic factors for CPU and disk: similarly to dynamic network factors,
21 allows the user to set a callback which can affect the progress of activities
22 (multiplicative factor applied when updating the amount of work remaining).
23 - Example: examples/cpp/exec-cpu-factors
26 - New: s4u::Disk::set_sharing_policy() and s4u::Host::set_sharing_policy().
27 Allows the configuration of non-linear resource sharing for hosts and
31 * New section "Release Notes" documenting recent and current developments.
32 * New section "Modeling I/O: the realistic way" presenting how to properly
33 model disks in SimGrid.
35 ----------------------------------------------------------------------------
37 SimGrid (3.28) July 14. 2021
39 The Victoriadagarna Release.
42 - C++ platform interface: Users can now describe their platform directly in C++.
43 This provides greatly flexibility and performance improvement for complex
44 platforms. Main features:
45 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
46 "zones", instead of single hosts. This feature enables the description
47 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
48 - StarZone: new zone with a Star-like topology. The routes are defined
49 as a set of links used to communicate from node to everybody (node<->ALL).
50 - Split-Duplex links: auxiliary method to create split-duplex links in
51 the platform, easing its utilisation. It automatically creates both UP
52 and DOWN links (similarly as done in XML).
53 - Please refer to the documentation and the examples included:
54 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
55 - New plugin: Producer-Consumer with monitor. Just requires to include the
56 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
57 associated example (examples/cpp/plugin-prodcons).
60 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
61 s4u::Io::wait_any(), s4u::Io::wait_any_for().
62 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
63 parameter by reference, instead of a pointer.
64 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
65 Explicitly cancel the activity to get back to previous behavior.
66 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
69 - The default SMPI compiler flags are no more taken from the environment.
70 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
71 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
73 --cfg=smpi/finalization-barrier: which can be used to add
74 a barrier inside MPI_Finalize. This can help for some codes which cleanup
75 data attached to a process, but still used in other SMPI processes.
76 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
77 is active when set to false, to keep going after a small error
78 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
79 which may or may not be problematic in the end.
81 - fix behaviour, as maximum iteration count could be ignored
82 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
83 use sampling when the same kernel is called with a different set of parameters
84 which have an impact on the timing.
85 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
86 It should now work with smpi/auto-shared-malloc-thresh.
87 - Improve error handling and reporting in multiple places
88 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
89 validity checks, truncated messages are now an error, return errors
90 when explicitely deleted handles are reused, ...)
91 - RMA: multiple fixes and stability improvements.
92 - analysis (-analyze flag in smpirun):
93 - SMPI can now report buffer leaks as well as MPI handles leaks,
94 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
95 - if -trace-call-location is used when compiling, SMPI can report
96 origin of leaked handles/buffers
97 - group leaks by type/origin in output message if possible
98 - New implemented MPI calls: MPI_Comm_test_inter
101 - Changed internal implementation of bandwidth factors in network models.
102 Models affected: CM02, LV08 (default), SMPI, IB.
103 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
104 Bandwidth factors are applied to communications to describe that users
105 cannot use 100% of the available bandwidth. For example, the default network model,
106 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
107 behavior was implemented by limiting the bandwidth available in the LMM
108 system for this flow. This may give the false impression that there is
109 bandwidth available for other flows due to its underutilization, especially
110 for the dynamic bandwidth factors used in SMPI models.
111 To avoid this, we have modified the implementation so that each flow uses the
112 maximum physical bandwidth according to the LMM system.
113 However, the actual throughput of the flow seen by the user is defined by
114 the physical bandwidth multiplied by the bandwidth factor.
115 This change impacts on the simulation results for all network models on
116 which we have bandwidth factors configured.
117 *****************************************
118 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
119 *****************************************
120 This change may impact on the timing of your simulation results.
121 Take care when comparing simulations from different SimGrid's
122 versions. Sorry for the inconvenience.
123 - Dynamic network factors: users can configure a callback to define
124 the network factors dynamically. This API is available at
125 simgrid::kernel::resource::NetworkModelIntf.
126 - Users have access to complete information about the current communication
127 to decide which factor to apply. This includes: message size, source and
128 destination hosts, links and zones traversed.
129 - Dynamic factors for both latency and bandwidth.
130 - For more details, see the example in (examples/cpp/network-factors).
131 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
132 deprecated since version 3.24, are no longer supported. Instead, use
133 "wattage_off" and "wattage_per_state".
136 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
139 - New tutorial: Model-checking and formal assessment
140 - New sections: "Demystifying the routing" and "C++ platforms"
141 - Update and improve the part on visualization in MPI and Algo tutorials.
142 - Improve the section on routing: how to define it, how it's used internally
143 - Fix many issues, broken links and missing references in doxygen and Sphinx
146 - Lua platform files are deprecated. Their support will be dropped after v3.31.
149 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
151 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
152 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
153 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
154 - FG#64: Configuring smpi/IB-penalty-factors
155 - FG#67: Running computation concurrently with MPI_Iallreduce
156 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
157 - FG#71: Segmentation fault on invalid gw_src/gw_dst
158 - GH#322: Issue when an actor kills his host vm
160 ----------------------------------------------------------------------------
162 SimGrid (3.27) March 29. 2021
164 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
166 Main user-visible changes:
167 - All functions marked as deprecated for one year were removed.
168 - Removed all that was related to the "Storage" type in all APIs and bindings.
169 "Disks" have been introduced more than a year ago to replace them for the better.
170 It's time to finish this replacement.
171 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
172 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
173 avoiding "initializer element is not constant" errors with some codes. Please be
174 careful to recompile your MPI codes if switching to this new SimGrid version.
175 - Network model ns-3 now requires version ns-3.28 or higher.
176 - Removed some option aliases that were deprecated since v3.23 (June 2019)
177 - 'verbose-exit' is now 'debug/verbose-exit'
178 - 'simix/breakpoint' is now 'debug/breakpoint'
179 - 'clean-atexit' is now 'debug/clean-atexit'
182 - New functions: Comm::sendto_{init,async} to initiate a communication
183 on between two (possibly remote) hosts.
184 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
185 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
186 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
187 or Mailbox::get_async<void>() if you really want to play with void*.
188 - Unify the interface of Activity::on_{start/activity}
189 - New function: Comm::get_dst_data()
190 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
193 - ptask should now work better with multicore hosts (FG#37).
194 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
195 - Internal cleanups and refactoring around models, to prepare for programmatic
197 - Modernize our tests of the performance models
201 - New section: Configuring the logs from the command line.
202 - New section: Programming API of logs.
203 - New section: Getting realistic results
204 - New section: Examples of SMPI Usage
205 - New part: SimGrid's Internals, with a section on Design goals
206 - Move back to the classical Breathe tool chain, and fix many small glitches.
209 - Handle leak checking: SMPI will now display a message if MPI handles are still
210 present at the end of a simulation. Option smpi/list-leaks:n will display the n
211 first leaked pointers and their types.
212 - smpi/display-allocs option : Display Memory allocations analysis, with hints
213 about the largest allocations. This only works for allocations with malloc/calloc
214 and not compiled with SMPI_NO_OVERRIDE_MALLOC
215 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
216 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
217 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
218 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
219 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
220 - Replay now opens and parses the trace file only once for all ranks when possible.
222 - better handling of MPI_PROC_NULL
223 - MPI_Group_incl was broken in some cases
224 - PAPI tracing now fixed
225 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
226 - SMPI will now fail properly when receive size < sent size, instead of
227 just setting MPI_ERR_TRUNCATE and keep going.
228 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
229 - Remove useless build patches and split build/run patches on the page
230 - Build in parallel when possible
231 - Test on yum-based systems
232 - New nightly tested Proxy-apps:
233 - OpenMPI mpi-test-suite
234 - Coral MCB, LSMS, Qbox, CLOMP
235 - ECP/Mantevo Cloverleaf3D, Ember
236 - Codevault bhtree_mpi
239 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
240 - Start to overhaul the implementation to modernize it (still ongoing).
243 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
244 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
245 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
246 - Deprecate xbt_dynar_to_array().
247 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
249 C binding and interface:
250 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
251 - Constness requirement has been lowered for parameter argv of sg_actor_start()
252 and sg_actor_create() It's now (char* const*). The previous version is still
253 available as sg_actor_start_() and sg_actor_create_().
255 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
256 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
257 - FG#37: Parallel tasks are limited to 1 core per host
258 - FG#62: Running "smpirun -replay" on large networks
259 - FG!46: Fix a few potential memory leaks in SMPI colls
260 - GH#343: Wrong consumed energy when an exec timeouts
262 ----------------------------------------------------------------------------
264 SimGrid (3.26) December 15. 2020
266 The Release release (the French lockdown was eased today).
268 Important user-visible changes:
269 - SimGrid now requires a compiler with C++14 support.
270 Sibling projects should upgrade their FindSimgrid.cmake
271 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
272 several users had difficulties to understand issues when using high bandwidth or
273 small latency events. The new value was already the default for SMPI and
274 should not cause too much performance hit. It can change some simulation
275 timings if simulation was skipping a lot of small events (which was actually
276 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
277 - On some network models (L07, CM02, LV08), an implicit loopback link is created
278 for local communication. Default bandwidth was 500MBps and latency was 15µs,
279 which was rather bad. Default values have been changed to 10GBps and 0 latency.
280 Options network/loopback-lat and network/loopback-bw have been added to adapt
281 these values if needed. Network and routing models already providing separate
282 loopbacks (clusters and such) are not impacted by this change.
283 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
284 metadata that are added to each MPI message by MPI libraries (tag,
285 src/dest, and communicator information). This may change simulation times
286 for very small message sizes, but should be more realistic.
289 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
290 - New plugins: wifi_energy and link_load. See the associated examples.
291 - New examples: wifi networks, and communication suspend.
294 - LTO is now enabled for Intel/clang compilers.
295 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
296 Setting it to "auto" will use all available cores, while setting it to n will
297 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
298 - Remove obsolete runtime option 'contexts/parallel-threshold'.
299 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
300 has been kept for compatibility.
301 - Finally remove obsolete snake_case() aliases for runtime options.
302 - Further improve the documentation.
305 - the on_exit() of each actor is also executed when the simulation deadlocks.
306 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
307 An example is provided for s4u::Comm but it should work with Exec and Io.
310 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
311 https://framagit.org/simgrid/SMPI-proxy-apps
312 - MPI/IO: supports names starting with "./"
313 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
315 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
316 building shared libraries, this may cause issues, so environment variable
317 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
318 - most temporary files should now be created in /tmp dir (or equivalent).
319 If this one does not allow execution of code (noexec flag), this may cause issues.
320 Please use another tmp directory (using TMPDIR or equivalent system variable)
324 - A large refactoring is ongoing, but this should be transparent for users so far.
327 - Drop xbt_str_split().
329 C binding and interface:
330 - The return type of the sg_actor_on_exit() callbacks is now 'void'
331 (int value was previously ignored)
332 - Many C functions were renamed to follow a common naming style
333 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
334 names are kept until version 3.30.
335 - Many MSG tests were converted to the new S4U's interface in C, that
336 was extended for that.
338 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
339 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
340 - FG#41: Add sg_actor_create C interface
341 - FG#43: xbt::random needs some care
342 - FG#48: The Impossible Did Happen (yet again)
343 - FG#50: Suspending an actor executed at the current timestamp fails
344 - FG#52: Zero-seconds timeout : "That's in the past already"
345 - FG#53: Crash while using ns-3 network model
346 - FG#54: How to suspend a comm?
347 - FG!22: Stochastic Profiles
348 - FG!24: Documentation and fix for xbt/random
349 - FG!35: Add a modeling hint for parallel links in doc
350 - FG!36: [xbt/random] Read/Write the state of the RNG
351 - FG!42: Link cumulated load plugin
352 - FG!44: Test condition variable wait_until+notify_all
353 - GF#18137: Allow different stack sizes?
354 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
355 - GH#139: Allow pthread creation in SMPI
356 - GH#336: Packet-level simulation using SMPI?
357 - GH#345: Error in the LMM while migrating a VM
358 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
359 - GH#352: pip install / python setup.py install fail to find pybind11
360 - GH!337: Fix link_energy plugin for wifi platforms
361 - GH!339: Add Mailbox set_receiver method to python binding
362 - GH!344: Cast hugepages macros parameters to int64
364 ----------------------------------------------------------------------------
366 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
368 The Palindrom Day Release.
370 Important user-visible changes:
371 - Improve the Python usability (stability and documentation).
372 - A nasty synchronization bug was ironed out, see also below.
373 - Python's doc was organized as a separate tree, now integrated with C++.
374 - C bindings of S4U were not part of the doc.
375 - The C++ doc was also improved as methods are now split by theme.
376 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
377 - This is mandatory to use the Java bindings.
378 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
379 - SimDAG++: Automatic dependencies on S4U activities (experimental)
380 - Some features are already implemented but not all of them
381 - Cannot block an activity until it's scheduled on a resource
382 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
383 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
386 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
387 - Drop support for Viva/Triva (old visualization tools), hence removed
388 TRACE_get_node_types() and TRACE_get_edge_types() functions.
391 - Require pybind11 version 2.4.
392 - Greatly improve locking during context switches with Python.
393 - Declaring Python callbacks to this_actor::on_exit is now working.
394 - No more tests are failing currently, even on win32.
397 - Actor: Merge signals on_migration_start/end into on_host_change
398 - Actor: Rename migrate() into set_host()
399 - Disk: Allow users to get the read and write nominal bandwidth values
400 - Exec: Implement wait_for(timeout)
401 - Io: Implement wait_for(timeout)
404 - Parse errors now raise a simgrid::ParseError that you may want to catch.
407 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
408 been renamed to fix a spelling error. As usual, the previous definitions are
409 kept with a deprecation warning. However, to avoid runtime errors, it is now
410 forbidden to override these deprecated methods in a derived class. Please use
411 the new names immediately if you need to override them.
414 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
415 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
416 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
417 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
418 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
419 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
420 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
421 - New module: random, providing classical random numbers generators.
424 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
425 allocations' size above which they are considered "shared" by default (as if
426 they were performed through SMPI_SHARED_MALLOC macros).
427 Default = 0 = disabled feature.
428 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
429 This can cause some troubles if codes are already overriding these. If this
430 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
431 help, but will make this feature unusable.
432 - Input check is now performed much more consistently and outputs precise
433 warnings for each error found
434 - New supported calls : MPI_File_set_view, MPI_File_get_view
435 - MPI I/O now only supports disks, as storage is discontinued.
436 - MPI I/O allows opening files without specifying absolute path
437 - Note : SMPI configuration options are now initialized by smpi_init_options()
438 call, which should be called if SMPI is not being used through smpirun.
440 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
441 - FG#9: Python bindings crashing
442 - FG#39: Missing s4u::Comm::wait_any_for example
443 - FG#42: Add support for ThreadSanitizer (TSan)
444 - FG!19: Removing RngStream
445 - FG!20: A module for RNG calls
446 - FG!21: Choice between ad-hoc and standard distributions implementations
448 - FG!25: Fix link in Mutex doc
449 - FG!26: Fix links in SMPI interface doc
450 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
451 - GH#305: Unscheduled tasks are still excuted
452 - GH#313: smpirun: manual outdated w.r.t. --help
453 - GH#321: [S4U] Get task remaining work ratio
454 - GH#323: Crash when an actor turn off his physical host
455 - GH#335: Missing links on dragonfly example svg
456 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
457 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
459 ----------------------------------------------------------------------------
461 SimGrid (3.24) October 9. 2019
463 The Clean Disk Release.
466 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
467 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
468 thus becomes the default resource to execute I/O operations. Examples
469 and the file system plugin have been modified to use disks instead of
470 storage elements. Storage can still be used but is doomed to disappear.
471 Users are thus adviced to update their simulators accordingly.
472 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
473 for consistency with pthread_barrier_wait()
474 - Host::get_englobing_zone() returns the englobing netzone
475 - Actor::on_destruction is now called in the destructor
476 Actor::on_termination new signal called when the actor terminates
478 - Global signals are now part of the Engine:
479 - on_platform_creation: after config settings, before the XML parsing
480 - on_platform_created: right after the XML parsing
481 - on_time_advance: each time the clock advances
482 - on_simulation_end: after simulation, before cleanups
483 - on_deadlock: as the name implies.
485 - sg_{actor,host,link}_{data,data_set}() now all exist.
486 Use them to attach user data to the object and retrieve it.
489 - Introduce an experimental Wifi model. It sounds reasonable
490 according to the state of the art, but it still has to be properly
491 validated, at least against ns-3 if not against reality.
492 - Improved the usability of ns-3. Several bugs were ironed out.
493 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
494 It is now expressed as 'idle:epsilon:allCores' to properly model the
495 consumption of non-whole tasks on mono-core hosts. More info in the doc.
498 - convert a new set of functions to the S4U C interface and move the old MSG
499 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
502 - Fortran bindings for DVFS have been removed.
503 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
504 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
505 - SMPI can now be selected by cmake's find_module(MPI) with
506 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
507 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
508 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
509 - trace-call-location can be used with TI traces, and replayed, and is compatible
510 with smpi/comp-adjustment-file.
511 - sleep events are now correctly traced/replayed.
512 - Default for trace-call-location is now to use file names and not full paths.
513 To revert to previous behavior (in case of collision of filenames), option
514 "smpi/trace-call-use-absolute-path" can be set to yes.
517 - Use the included xxHash as an hash implem when C++14 is usable.
518 - Option model-checker/hash was removed. This is always activated now.
519 - New option smpi/buffering controls the MPI buffering in MC mode.
520 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
521 This is useful to check for MPI compliance.
524 - New section on plugins: how to define a new one, and existing ones.
527 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
528 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
531 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
532 and <mount> tags. The rationale is finish to decouple what should be handle
533 by SimGrid kernel (i.e., resources and activities) from plugins built in
534 user space (e.g., the file system plugin). This new tag currently accepts
535 only three attributes (id, read_bw, and write_bw). All the other information
536 that was declared with the storage related tags now has to be expressed as
537 properties. An example of platform using this new tag is available at
538 examples/platforms/hosts_with_disks.xml
541 - 'expect signal' can now accept more than one potential signal.
543 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
544 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
545 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
546 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
547 - FG#31: per-actor data
548 - FG#34: SG_BARRIER_SERIAL_THREAD?
549 - FG#35: model-checker does not like buster-produced binaries
550 - FG!13: MC: complete workaround in the error msg seen on modern systems
551 - FG!15: execute_flops now logs compute
552 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
553 - FG!17: ns-3: unblock the right number of communications + others issues
554 - FG!18: Improving the performance of the ns-3 bindings
555 - GH#189: Energy consumption of parallel tasks
556 - GH#219: Error in the throughput of TCP transfer
557 - GH!330: Minor polishing of English: Configuring_Simgrid
558 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
559 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
560 - GH!334: Rework energy plugin, again
562 ----------------------------------------------------------------------------
564 SimGrid (3.23.2) July 8. 2019
567 - Nicer introduction page.
568 - Migrate the "Deploy your application" page to the new doc.
569 - Move Java as a subtree of MSG.
572 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
575 - Drop xbt_dynar_sort_strings().
578 - Really fix FG#26: Turning off a link should raise NetworkFailureException
579 - FG#27: Wrong exception thrown to wait_any when link is turned off
580 - GH#328: Java: Canceling multiple tasks in a single vm/host
582 ----------------------------------------------------------------------------
584 SimGrid (3.23) June 25. 2019
586 The Exotic Solstice Release.
589 - SunOS and Haiku OS support. Because exotic platforms are fun.
590 - Stop setting random seed with srand48() at initialization.
591 - Use addr2line as a fallback for stacktraces when backtrace is not available.
592 - Build option -Denable_documentation is now OFF by default.
593 - Network model 'NS3' was renamed into 'ns-3'.
596 - Simgrid can now hopefully be installed with pip.
599 - wait_any can now be used for asynchronous executions too.
602 - New log appenders: stdout and stderr. Use stdout for xbt_help.
603 - Drop xbt_dict_dump.
606 - 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
607 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
608 - MPI interface is now const correct for input parameters
609 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
610 - Fortran bindings for async collectives.
611 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
614 - Remove option 'model-check/record': Paths are recorded in any cases now.
615 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
616 always sparse. That's an immense gain in memory, and can even be
617 faster because of cache effects. So there is no need to clutter the
618 code to allow the user to go for the unefficient mode.
621 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
622 models proved to be more accurate than these old experiments.
624 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
625 - FG#1: Broken link in error messages
626 - FG#2: missing installation documentation
627 - FG#3: missing documentation in smpirun
628 - FG#6: Python bindings not available on PyPI
629 - FG#7: simple cmake call requires doxygen
630 - FG#8: make python bindings an optional dependency
631 - FG#10: Can not use MSG_process_set_data from SMPI any more
632 - FG#11: Auto-restart actors forget their on_exit behavior
633 - FG#12: -Denable_lto=OFF doesn't disable LTO
634 - FG#13: Installs unstripped file 'bin/graphicator'
635 - FG#14: Installs the empty directory 'doc/simgrid/html'
636 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
637 - FG#17: Dead link in doc (pls_ns3)
638 - FG#20: 'tesh --help' should return 0
639 - FG#21: Documentation link on http://simgrid.org/ broken
640 - FG#22: Debian installation instruction are broken
641 - FG#26: Turning off a link should raise NetworkFailureException exceptions
642 - GH#133: Java: a process can run on a VM even if its host is off
643 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
644 - GH#326: Valgrind-detected error for join() when energy plugin is activated
646 ----------------------------------------------------------------------------
648 SimGrid (3.22) April 1. 2019
650 The Easter Christmas Release (this one is somewhat late).
653 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
654 - Not complete yet: asynchronous activities (amongst others) are still missing
655 - Still ongoing: the interface may change in the future. We need more
656 testers! Please report any glitches.
657 - No new project using Java should start now. Please switch to Python.
660 - Some of the internal cleanups may lead to speed improvements:
661 - The hard limitation on the amount of simulated actors with
662 Java+Mac was removed. Now, the available memory is the only limit.
663 - Our refcounting was tidyied, leading to 10% speedups in some cases.
664 - We are still working on making our code robust to the actor kills
665 and hosts' churn. Things are improving, but it's not perfect yet.
666 - Replaced our own code to display a backtrace (that was forking addr2line)
667 with the Boost.Stacktrace library.
668 You won't see your backtraces without this optional dependency.
669 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
670 - Stop setting random seed with srand() at initialization.
673 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
674 XML file version remains 4.2 since old files are still compatible.
677 - Process termination which was broken at version 3.21 has been repaired.
678 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
679 - Hide the examples into examples/deprecated. New users should use Python.
682 - Drop MSG_process_create_from_stdfunc() from the API.
683 This C++ function was a pimple in the C API, made necessary at some
684 point by the Java bindings. This is fixed now.
685 - Hide the examples into examples/deprecated. New users should use S4U.
686 - MSG_process_create and MSG_process_attach now crash if the host used
687 to run this process is off.
688 - Fix the protype of MSG_process_on_exit()
689 Now use (int,void*) callbacks instead of (void*,void*) ones.
690 The implementation was ways too messy to actually work, I guess.
693 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
694 too often and provide more accurate timings.
695 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
696 - MPI_Alltoallw support
697 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
698 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
699 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
700 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
702 - Basic implementation of generalized requests (SMPI doesn't
703 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
706 - Drop sg_cmdline. Please use xbt_cmdline instead.
707 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
708 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
709 OsSemaphore is implemented in a portable way with C++11 threads.
710 This should allow much more threads to be created at the same time,
711 allowing Mac and Java users to simulate many more actors.
712 - Implement the 'thread' factory with std::thread instead of xbt ones.
713 It is not possible to set the stack size with threads anymore, but
714 -fsplit-stack is the way to go nowadays when using threads.
715 - Drop the xbt_os_thread_t module (now unused)
716 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
717 - Drop xbt_str_join_array().
718 - Drop cunit, use Catch2 instead.
721 - Many cleanups in the kernel::activity namespace. This was long
722 overdue, and shall open the path to many future endeavors.
725 - #132: Java: a process can not shut down its own host
726 - #220: S4U: detached send causes exception if sender terminates soon after sending
727 - #261: Document the parameters of parallel execution's constructor
728 - #300: [s4u] BarrierPtr is missing
729 - #314: SMPI args internal cleanup
730 - #316: Fix a bug related to the CPU utilization of multi-core VM
731 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
732 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
733 - #325: Turning off a host has different behavior on sleeping actors and computing actors
735 ----------------------------------------------------------------------------
737 SimGrid (3.21) October 3. 2018
739 The Restarting Documentation (TRD) Release.
742 - Convert the user manual to Sphinx for improved usability.
743 Unstable version now hosted on https://simgrid.frama.io/simgrid
744 Many glitches remain; stay tuned.
745 - Restrict the Doxygen reference API to the installed header files.
746 (documenting all internals this way is overkill + doxygen scales badly)
747 - New tutorial on S4U (highly inspired from the old MSG one, but with
748 a git repository to fork as a starting point, and a docker image)
749 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
750 - The Developper manual is still to be converted (not compiled until then)
751 - Some parts are still missing in this conversion (such as the
752 platform and deployment sections) while others were blindly converted
753 and would need more love (such as the configuration flags).
754 Things will certainly further improve in the future releases.
757 - Fully reimplement auto_restart mechanism that was utterly broken
758 (fix #22, #131 and #281 that were all very related)
759 - Implement semaphores (fix #303 and #291)
760 - When creating an actor from a function and its parameters,
761 move-only parameters are not allowed anymore, as it would prevent
762 the actor to be restartable if its parameters are consumed on run.
763 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
764 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
765 IO operations that can be started, waited for, or canceled as a regular
767 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
771 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
772 - New host variable: 'core_count'
775 - Remove xbt_os_thread_specific features
776 - Remove portability wrapper to condition variables
777 - Remove xbt_os_thread_yield()
780 - MPICH collective selector now mimics MPICH 3.3b
781 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
782 - MPI_Init(NULL,NULL) is now allowed
783 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
784 - Command-line arguments should now be handled properly in Fortran simulations
785 - MPI Attributes and related callbacks should work in Fortran
786 - Apps using C + Fortran should now work
787 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
788 - Support MPI_Error_String
791 - Due to an internal bug, Msg.run() must now be your last line.
792 We hope to fix it in a future release, and we are sorry for the inconvenience.
795 - #22: Process autorestart seem to only work with CAS01 cpus
796 - #56: Feature request: dynamic SMPI replay
797 - #93: simgrid should not eat --help
798 - #111: How to change the loopback link characteristics?
799 - #116: Platform cluster radical
800 - #117: Platform cluster/host order
801 - #129: Replay_multiple on steroid
802 - #131: Java : autostart does not work
803 - #143: Setting a breakpoint at a given time
804 - #235: Network energy models should be integrated
805 - #264: Add ptask L07 resource tracing
806 - #271: Dynamic and manual replay of SMPI traces
807 - #279: Breakpoints option uses time deltas instead of absolute time
808 - #281: Daemonized actor no longer daemonized after an auto-restart
809 - #282: TIT convertor erase traces when using absolute path in the trace list file
810 - #285: segfault when a s4u actor kills itself with msg process tracing activated
811 - #286: Pajé traces are not exposing the number of cores
812 - #287: Command-line parsing should stop when encountering '--'
813 - #288: MPI_Init(NULL, NULL)
814 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
815 - #290: Method to check if a mailbox has a message ready to be consumed
816 - #291: [s4u] semaphores do not seem available yet
817 - #294: Alltoall collective implementation of mpich
818 - #295: Starting a Host as OFF is not supported
819 - #296: DTD too permissive
820 - #299: Add a s4u barrier example
822 ----------------------------------------------------------------------------
824 SimGrid (3.20) Released June 24. 2018
826 The proxy snake_case() release.
829 - The whole API is now using snake_case() instead of mixing it with camelCase.
830 Compatibility wrappers in place for one year.
831 - We took the opportunity to increase the consistency of the API.
832 Please report any remaining glitches.
833 - New function simgrid::s4u::Host::get_actor_count:
834 Returns the number of actors running on a specific host.
837 - SMPI is now tested with ~45 proxy apps from various sources, with none or
838 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
839 - Replay: The replay file has been re-written in C++.
840 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
841 supported. They are stored in the trace and used when replayed.
842 - Basic support of MPI_Cancel. Robustness not guaranteed.
843 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
844 (only for MPI_COMM_TYPE_SHARED).
845 - New option: smpi/privatize-libs, to add external shared libs to be privatized
846 by SMPI. They will be copied locally and loaded separately by each process.
847 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
848 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
849 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
850 with internal index optind with multiple processes. Only works if MPI_Init has
852 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
853 syntax without preprocessing tricks.
856 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
857 behavior tracing is no longer limited to MSG
858 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
859 this means that traces now need to be updated or re-obtained; both Irecv
860 and Isend lines in your traces have now in total 5 mandatory fields:
861 <rankid> <command> <to/from rankid> <tag> <size>
862 To update your traces, it suffices to add a 0 for the tag here.
863 - TIT now also supports waiting for a distinct request via MPI_Wait.
864 Wait/Test now wait for a specific request, not just the last one that was
865 issued. This unfortunately means another update, because we need to
866 identify which request you want to wait for. We do this via the
867 triplet (sender, receiver, tag), which needs to be added:
868 <rankid> <command> <sender> <receiver> <tag>
869 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
870 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
873 - The deprecation of MSG is ongoing (but this should not impact you).
874 Many MSG functions are now simple wrappers on the C API of S4U. If
875 you wish to convert your code to S4U, find the S4U counterparts of
876 your MSG calls in src/msg/msg_legacy.cpp.
877 - MSG can still be used, but won't evolve anymore.
880 - Allow to run the Link energy plugin from the command line with
881 --cfg=plugin:link_energy
882 - Rename Energy plugin into host_energy
883 - Rename Load plugin into host_load
886 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
887 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
888 - Improve performance when handling timeouts of simix synchros.
891 - Config: the C API is now deprecated (will be removed in 3.23), and
892 the C++ API has been slightly improved.
895 - Fix several build issues on OSX.
896 - Move simgrid_config.h to simgrid/config.h (old header still working)
899 - #143: Setting a breakpoint at a given time
900 - #150: Inconsistent event names in SMPI replay
901 - #258: daemonized actors hang after all non-daemonized actors have completed
902 - #267: Linker error on unit_tmgr
903 - #269: SMPI: tracing of MPI_Wait/all/any broken
904 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
906 ----------------------------------------------------------------------------
908 SimGrid (3.19.1) Released March 22 2018
910 The Fixed ABI Release.
912 As you may know, we are currently refactoring SimGrid in deep.
913 Upcoming SimGrid4 will be really different from SimGrid3: modular,
914 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
916 Our goal is to smooth this transition, with backward compatibility
917 and automatic update paths, while still progressing toward SimGrid4.
919 SimGrid remains open during works: The last pure SimGrid3 release was
920 v3.12 while all subsequent versions are usable alpha versions of
921 SimGrid4: Existing interfaces remain unchanged, but the new S4U
922 interface is budding and the internals are deeply reorganized.
924 Since 2015, we work hard to reduce the changes to public APIs. When
925 we need to rename a public library symbol in S4U, we let your
926 compiler issue an explicative warning when you use the deprecated
927 function. These messages remain for four releases, i.e. for one full
928 year, before turning into an error. Starting with v3.15, your can
929 also adapt to API changes with the SIMGRID_VERSION macro, that is
930 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
932 Starting with this v3.19.1, our commitment to reduce the changes to
933 the public interfaces is extended from the API to the ABI: a program
934 using only MSG or SimDag and compiled against a given version of
935 simgrid can probably be used with a later version of SimGrid without
936 recompilation. We will do our best... but don't expect too much of
937 it, that's a really difficult goal during such profund refactoring.
939 The difference between v3.19 and v3.19.1 is that the former was
940 accidentally breaking the ABI of MSG, while the later is restoring
943 S4U and kernel APIs will still evolve until SimGrid4, with one-year
944 deprecation warnings as currently. In fact, cleaning up these
945 interfaces and converting them to snake_case() is one release goal of
946 v3.20. But don't worry, we are working to smooth this upgrade path.
948 In summary, new projects should start with S4U to benefit of the
949 future, but old MSG projects should still be usable with no change.
951 ----------------------------------------------------------------------------
953 SimGrid (3.19) Released March 20 2018
955 The Moscovitly-cold Spring release.
957 SMPI user-visible improvements:
958 - Support of MPI_Type_create_subarray
959 - Unimplemented MPI calls now abort (instead of issuing a warning)
960 Some harmless calls around errhandlers still only issue a warning.
961 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
962 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
963 is now possible (mmap privatization was refactored to this end).
964 Adding new SMPI instances should also be technically doable now.
966 SMPI internal cleanups:
967 - Removed index notion from SMPI -- no more "getPid() - 1"!
968 In tracing files, actors are called rank-<PROCESS-ID> but these are
969 NOT the real ranks (tracing several communicators is not supported yet).
970 - Deprecate smpi_process_index() (will be removed in 3.22)
971 - Better testing on classical proxy apps, through an external project:
972 https://github.com/simgrid/SMPI-proxy-apps/
975 - Use a graphical TOC to make it easier to find the documentation you need
976 - Revamp the MSG tutorial
979 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
980 - Remove parameter of MSG_process_killall().
981 Resetting the PID was bogus anyway (several actors could have the same PID).
984 - Execution->setHost() can be called after start(), to migrate it.
985 - Comm::test_any() is now implemented.
986 - s4u::Actor now has onCreation() and onDestruction() signals.
987 - Install some kernel header files for the users' plugins and more.
988 Warning, their API is really not stable yet.
989 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
990 include/simgrid/kernel/routing/*Zone.hpp
993 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
995 Build System and other cleanups:
996 - Remove unused run-time parameter "tracing/onelink-only".
997 - SimGrid now works with intel compilers (MC still buggy)
998 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
999 - tesh kills subprocesses on timeout (but not on windows)
1002 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
1005 - #194: Feature request: simgrid::s4u::Comm::test_any()
1006 - #245: migrating an actor does not migrate its execution
1007 - #253: Feature Request: expose clusters as objects
1008 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1009 - #255: Tesh broken on Windows
1010 - #256: Modernize FindSimGrid.cmake
1011 - #257: Fix (ab)use of CMake install
1013 ----------------------------------------------------------------------------
1015 SimGrid (3.18) Released December 24 2017
1017 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1019 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1020 - Support for MSG should not be removed from SimGrid before 2020,
1021 but future evolutions will be limited to the integration of
1022 user-provided patches.
1025 - s4u::onDeadlock() signal, run before stopping the simulation.
1026 - s4u::Exec: asynchronous executions (abstraction of background
1028 - s4u::Host->getLoad() returns the achieved speed in flops/s
1031 - Link::name() is deprecated and will be removed in v3.21, use
1032 Link::getCname() instead.
1033 - Mailbox::getName() changed to return a std::string, use
1034 Mailbox::getCname() instead to get a char*.
1035 - Storage::getName() changed to return a std::string, use
1036 Storage::getCname() instead to get a char*.
1037 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1038 to have both allocation and deallocation of the map in user space.
1041 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1042 The amount of flops remaining to do is a vector, not a scalar.
1043 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1044 name implies on both sequential and parallel tasks.
1045 - Both changes fix GitHub's #223 using PR #237 as a basis.
1046 Thanks Michael Mercier.
1047 - Most examples were converted to S4U and hidden elsewhere at tests.
1048 Rationale: we still want MSG to work; we want newcomers to use S4U.
1051 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1052 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1055 - Switch to the faster dlopen privatization mechanism by default
1056 - Documentation improvments
1059 - Live migration is getting moved to a plugin. Dirty page tracking is
1060 the first part of this plugin. This imply that VM migration is now
1061 only possible if one this function is called:
1062 - C/MSG: MSG_vm_live_migration_plugin_init()
1063 - C/C++: sg_vm_live_migration_plugin_init()
1064 - Java: Msg.liveMigrationInit()
1065 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1069 - Define class simgrid::xbt::Path to manage file names.
1070 - Removed unused functions:
1071 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1072 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1073 xbt_graph_export_graphviz()
1074 - xbt/str.h: xbt_str_join()
1075 - Remove unused datatypes:
1076 - xbt/heap.h: use std::priority_queue or boost::heap instead
1077 - xbt/swag.h: use boost::intrusive::list instead
1080 - New link_energy plugin for the consumption of the links.
1081 - All of the operations on files and storage contents have been
1082 packaged into a plugin (src/plugins/file_system). The current
1083 public interface can be found in
1084 include/simgrid/plugins/file_system.h
1085 To use these functions you now have to initialize the plugin by
1086 calling MSG_storage_file_system_init() just after calling
1087 MSG_init() or sg_storage_file_system_init() just after creating
1091 - Remove the undocumented/untested tag <include>
1094 - Remove viva specific tracing as the tool is no longer maintained
1097 - #248: Kill of finished processes leads to segfault
1098 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1099 - #239: please implement signal s4u::onDeadlock()
1100 - #230: segfaults when exit() without run()
1101 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1102 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1103 - #222: Actor::kill() doesn't really kill and segfaults
1104 - #221: odd LMM warning when killing an actor
1105 - #120: Memory leak when the processes are forcefully killed
1107 ----------------------------------------------------------------------------
1109 SimGrid (3.17) Released October 8 2017
1111 The Drained Leaks release: (almost) no known leaks despite the tests.
1113 Even very long simulations will preserve your memory: our testsuite
1114 stresses SimGrid for over 45mn on fast machines for a coverage of
1115 over 80%, and there is only one single known leak, of about 4kb.
1119 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1120 handler may be given to free resources if the comm cannot be completed.
1121 - this_actor::send(mailbox) is now mailbox->put()
1122 - New: simgrid::s4u::Comm::wait_all()
1123 - New: Host.execute() for remote executions.
1126 - Unused option network/sender-gap is removed.
1129 - Deprecate MSG_task_isend_with_matching(): This unused feature
1130 really complicates our internals. Will be removed in v3.20.
1133 - Improved context termination. It is now possible to free resources, even
1134 when a process is forcibly killed.
1135 - ContextBoost: add support for Boost versions above 1.61.
1138 - Bring back run-time option --cfg=exception/cutpath to remove exception
1140 - Removed unused functions:
1141 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1142 xbt_str_rtrim(), xbt_str_trim().
1143 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1146 - Removed header files obsolete since SimGrid 3.12:
1147 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1149 - #3: SD_exit should be made optional
1150 - #120: Memory leak when the processes are forcefully killed
1151 - #159: Threading test regression in Actor refcounting
1152 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1153 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1154 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1155 - #191: VM migration and pstate
1156 - #192: Updating the energy consumptions of all hosts crashes with VMs
1157 - #195: All actors have PID=0 in the logs
1158 - #204: Sometimes segfault with thread contexts and mmap privatization
1159 - #222: Actor::kill() doesn't really kill and segfaults
1160 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1162 ----------------------------------------------------------------------------
1164 SimGrid (3.16) Released June 22. 2017.
1166 The Blooming Spring Release: developments are budding.
1169 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1170 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1171 - Java JAR file should be finally fully working on Mac OSX too.
1173 XML platforms: Switch to platform v4.1 format.
1174 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1175 - <zone> can be used as a synonym for the now deprecated <as>
1176 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1177 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1178 - <actor> can be used as a synonym for the now deprecated <process>
1179 - state_file and avail_file periodicity is now easier to express
1180 (check the documentation) the old behavior should still work.
1181 * Storage: not backward compatible, but it seems that nobody ever used it
1182 - Remove attribute 'content_type' of <storage_type>: was never used
1183 - Make attribute 'model' of <storage_type> optional: for future usage
1184 - Remove Bconnection model property: was never (in)validated. Replaced by
1185 the maximum of the read and write bandwidth as a resource constraint.
1188 - New and Backwards Compatibility break:
1189 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1190 When one wants to get the list of tasks whose states have changed during a
1191 simulation round, s/he has to allocate and free a dynar and use it as
1192 argument to this function. The former SD_simulate (double how_long)
1196 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1197 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1200 - New function to update the consumption of all hosts at once.
1201 - Fix the model for multi-core hosts, linear in the amount of busy
1202 cores with an abnormality for pIdle (see paper).
1203 BUG: Still not satisfactory for ptask on mono-cores.
1206 - The netzone are now available from the MSG API.
1207 The old names still work, but are now deprecated.
1210 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1211 - New option: smpi/keep-temps to not cleanup temp files
1212 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1213 mallocs mode is used, allows to change the size of the fake file used
1214 (default 1MB), to potentially limit the number of mappings for large runs.
1215 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1216 - Fortran ifort and flang compilers support
1217 - New RMA calls supported (experimental):
1218 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1219 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1220 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1221 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1222 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1223 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1226 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1227 - Behind the scene, ActivityImpl are now automatically refcounted.
1228 Removing this nasty bottleneck will greatly simplify our internals.
1231 - Replay: New function xbt_replay_action_get():
1232 Retrieve the function previously associated to an event type.
1233 - DROPPED FUNCTION: xbt_str_varsubst()
1234 - DROPPED MODULE: strbuff. We don't need it anymore.
1235 - DROPPED MODULE: matrix. We don't need it anymore.
1236 - DROPPED MODULE: lib. We don't need it anymore.
1238 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1240 SimGrid (3.15) Released March 22. 2017
1242 The Spring Release: continuous integration servers become green
1243 We fixed even the transient bugs on all target architectures:
1244 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1245 Windows, FreeBSD, NetBSD.
1247 - Rename NetCards to NetPoints.
1248 This was intended to help NS3 users, but that's not a netcard.
1249 That's a point in the routing algorithm, let's avoid wrong simplifications.
1252 - New: MSG_process_yield(). Stop and yield to other processes.
1253 - New: MSG_process_daemon(). Daemon processes are automatically killed
1254 when the last non-daemon process terminates
1255 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1256 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1257 to make room for the upcoming network energy plugin.
1258 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1261 - Ensure that an actor can kill itself with Process::exit()
1262 - Kill the obscure NativeException. Nobody want to survive the issues
1263 it denotes, so use JniException that is a RuntimeException (not to
1264 be caught explicitly).
1265 - Partial bug fix in initialization. SimGrid flags on command line were
1266 consumed at C level but stayed in the original Java String[] args.
1267 This could mess users' args[i] if SG flags were not put at the end of
1269 The SimGrid flags are now removed from the Java arguments. However,
1270 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1271 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1272 It might be if you have --log or --cfg flags in the command line.
1273 - Fix numerous memleaks all around the place. In particular, around VMs.
1277 - s4u::onPlatformCreated: right before the simulation starts
1278 - s4u::onSimulationEnd: right after the main simulation loop
1279 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1280 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1281 event from the availability_file changes the avail speed.
1282 - Links are now usable from s4u
1283 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1284 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1285 - New examples: The conversion of MSG examples to S4U has begun
1286 - Actors: create, kill, migration, and suspend
1287 - Applications: master-worker and token-ring
1288 - Action replay: communications and storage
1289 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1292 - Backwards Compatibility breaks
1293 - The SD_route_get_list and SD_route_get_size functions have been removed.
1294 They are replaced by the unique
1295 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1296 The route from src to dst is built in the links dynar whose size can
1297 be obtained with xbt_dynar_length.
1298 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1299 sg_host_route_bandwidth and sg_host_route_latency.
1300 Macros ensure the backwards compatibility, but you should fix your code
1303 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1304 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1305 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1306 We don't support 100% of it, but it should be enough. Contact us if not.
1307 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1308 naive and probably inaccurate)
1309 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1310 mapped onto a unique small file using some system magic.
1311 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1312 - Fortran: cleanups, fixes, support of user-added operations
1313 - RMA: MPI_Accumulate are applied in correct order.
1314 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1315 - Better support for MPI_IN_PLACE
1316 - Support for MPI_Win attrs and keyvals.
1317 - Support MPI_Comm_group_create, MPI_Type_size_x
1318 - MPI ops should be properly applied to all allowed types
1319 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1320 algorithm by a non failing one from mpich.
1323 - Kill the fifo data container: we don't use it anymore.
1326 - A <cluster> can now be created with different speed values to represent pstates.
1327 see examples/platforms/energy_cluster.xml for an example.
1329 SimGrid (3.14.159) Released December 28. 2016
1331 The previous release was only a crude approximation of a Pi release;
1332 this one should be a bit better.
1334 - Revert a last minute change that broke on Mac OSX.
1335 - Fix the changelog and NEWS files.
1336 - Improve the documentation, in particular of the routing module.
1337 - Integrate some patches flying around in Debian.
1339 SimGrid (3.14) Released December 24. 2016
1342 * The whole documentation was reworked and reorganized. There is
1343 still much room for improvement here, but we're on it.
1346 * We now need python3 for our testing infrastructure.
1347 * Model checking needs libevent
1350 * Now works on FreeBSD
1353 * Privatization now works on FreeBSD
1355 * Privatization is now activated by default in smpirun
1356 You can turn it off with -no-privatize if something goes wrong.
1358 * Call-location tracing for SMPI.
1359 You can add the exact location (filename / linenumber) of an MPI call to
1360 your trace files and slow down or speed up the simulation between two
1361 consecutive calls by using an adjustment file (see the documentation).
1363 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1364 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1365 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1366 values were ignored in some cases. The timings of these functions can now
1367 be significantly different.
1369 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1370 smpi/running-power is renamed to smpi/host-speed
1372 * smpi/grow-injected-times option to enable or disable multiplication of the
1373 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1374 make simulation less precise (but also much faster).
1376 * smpirun script should be (much) faster for large deployments.
1378 * SMPI tracing: fixed issue with poor matching of send/receives.
1380 * Replay: Fix broken waitall
1382 New functions and features
1383 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1385 Dropped / renamed functions and features
1386 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1387 - MSG_mailbox_is_empty() -> Mailbox::empty()
1388 - MSG_mailbox_front() -> Mailbox::front()
1389 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1390 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1391 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1392 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1393 * Don't pass the free_f parameter to property related functions:
1394 - MSG_host_set_property_value()
1395 - MSG_as_router_set_property_value()
1396 - MSG_storage_set_property_value()
1397 * VM properties. Since msg_vm_t are msg_host_t, just use
1398 MSG_host_get_property_value() and friends
1399 * VM I/O related things:
1400 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1401 - Unimplemented save/restore methods
1402 * MSG_as_router_get_property_value() was redundant with
1403 MSG_environment_as_get_property_value().
1404 - Removed MSG_as_router_*propert*() functions
1405 - Added MSG_environment_as_set_property_value() for consistency
1406 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1407 Well, they are still there for now, but deprecated with a warning.
1408 Please switch to xbt_dict_new_homogeneous() before this is removed
1410 * Task affinity. Its intended behavior (that was very badly tested
1411 and probably not really working) was deceiving what most users
1412 would have hoped here.
1413 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1414 * Option network/coordinates is now useless and should be dropped.
1417 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1420 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1421 * Added Dragonfly topology support
1424 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1425 EiBps, EBps, Eibps, Ebps for bandwidth.
1426 They may become useful to some lucky ones.
1429 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1430 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1431 * ex/async_waitAll: new example, on asynchronous communications
1434 * Memory usage should be decreased for simulations with a large number
1435 of processes. This also helps for SMPI.
1437 SimGrid (3.13) stable; urgency=low
1439 The Half Release, a.k.a. the Zealous Easter Trim.
1441 - We removed half of the lines, that were mostly experimental cruft.
1442 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1443 (not counting blanks and comments -- according to openhub.net)
1444 - The internals are now compiled in C++ (and will soon be clean C++)
1445 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1446 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1448 * Backwards Compatibility breaks
1449 - Removed Lua simulation bindings (switch to C or Java for that).
1450 Lua can still be used to describe platforms
1451 - Removed Java kernel plug-ins.
1452 Will be reintroduced after the ongoing major internals reorg.
1454 - the following functions were removed.
1455 They were too specific and should be reimplemented in a generic
1456 way, with filter function.
1457 - MSG_task_listen_from_host
1458 - MSG_mailbox_get_count_host_waiting_tasks
1459 - MSG_mailbox_put_with_timeout was removed.
1460 Please use MSG_task_send_with_timeout instead.
1462 - the SD_application_reinit function was removed. It has been a noop for a while.
1463 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1464 replaced by a more flexible mechanism.
1465 - The following functions thus do not exist anymore
1466 - SD_workstation_get_access_mode
1467 - SD_workstation_set_access_mode
1468 - SD_workstation_get_current_task
1469 - Basic estimation functions have been removed but can easily be replaced
1470 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1471 - SD_workstation_get_computation_time => amount / sg_host_speed()
1473 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1474 Use VM.getSpeed()*load/100 for the legacy behavior.
1476 - option enable_tracing was removed. It was not doing anything for a while.
1477 - In the ModelChecker:
1478 - the model-checker now ptraces the model-checked process which means
1479 you cannot use a debugger on the latter anymore (we might make this
1480 optional in the feature);
1481 - removed soft-dirty page tracking;
1482 - remove model-checked side snapshot management,
1483 MC_snapshot() and MC_compare_snapshot();
1484 - keep the MC_cut() function as a stub (it was not really working
1485 in the previous release).
1488 * All options are consistently in kebab-case. Old names are kept as alias.
1491 * Switch to platform v4 format.
1492 - Rename from 'power' to 'speed' the attributes describing the amount of
1493 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1494 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1495 - In <host> and <link>, attributes availability and state are gone.
1496 It was redundant with state and availability traces, and with peak values.
1497 - In <cluster>, attributes availability_file and state_file are gone.
1498 It was too complex and unused.
1499 - Kill <gpu>. Was not doing anything.
1500 - The DOCTYPE points to the right URL:
1501 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1502 (the file at this address now documents the changelog since its v1)
1503 - A warning is emitted for unit-less values (they are still accepted).
1504 - speed. Default: 'f' or 'flops'. Also defined:
1505 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1506 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1507 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1508 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1509 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1510 - latency. Default: 's' second. Also defined:
1511 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1513 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1514 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1517 * s4u::Host is now the preferred public interface to the Host features.
1518 sg_host_* functions are C bindings to the exact same behavior
1519 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1522 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1523 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1524 They were deprecated since a few years, and probably did not even compile anymore.
1527 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1528 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1529 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1530 encouraged to update their codes. A list of the modified functions can be found at the end of
1531 include/simgrid/simdag.h
1534 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1535 the lambda or closure passed as a parameter will run in kernel mode.
1536 All the callback functions should be rewritten to that interface at some point.
1539 * Reorganizing and cleaning the internals all around the place.
1542 * Remove old default barrier/bcast buggy algorithms (see #18407)
1543 * Various bug fixes to handle more codes
1544 * Remove the need for the --foreground option of smpirun (it is still
1545 accepted for backward compatibility).
1548 * Kill the setset data container: MC don't use it anymore.
1549 * Kill the queue data container: it made more sense with GRAS.
1550 * Kill the xbt_peer_t data type: it's useless without GRAS.
1551 * Kill rm_cb feature of config sets: it was never useful.
1552 * Kill graphxml parsing feature. It was not used.
1553 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1555 - xbt_dynar_sort_strings(), when the content is char*
1556 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1557 They throw exceptions on invalid input;
1558 * C++ support for declaring CLI flags (simgrid::config::Flag);
1559 * class for abstracting different signal backends (simgrid::xbt::signal).
1560 with no external dependencies (we need very simple signals).
1563 * refactoring and cleanup of the code;
1564 * ongoing process to cleanly separate the model-checking algorithms
1565 from the code model-checking support.
1567 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1569 SimGrid (3.12) stable; urgency=low
1571 The Facelift Release.
1574 * Require g++ v4.7 at least to not speak prehistorical C++.
1575 * Require Boost 1.48 (for signal2 component).
1576 * Java must be version 7 at least when activated.
1577 * Builds on Windows again (including Java bindings).
1578 * Tracing is now always enabled (no way to turn it out)
1579 * Remove GTNetS. It was not working anyway.
1580 * Various cleanups in the cmake scripts.
1581 * Move headers around to sort them out on installed systems:
1582 - instr/instr.h -> simgrid/instr.h
1583 - instr/jedule/* -> simgrid/jedule
1584 - simdag/datatypes.h was removed
1585 - simdag/simdag.h -> simgrid/simdag.h
1586 - msg/datatypes.h was removed
1587 - msg/msg.h -> simgrid/msg.h
1590 * Interface improvement:
1591 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1592 - Sanitize the interface in MSG_task_ module:
1593 - Merge two functions that were close enough but misleading:
1594 set_compute_duration(t) -> set_flops_amount(t)
1595 get_remaining_computation(t) -> get_flops_amount(t)
1596 - set_data_size(t) -> set_bytes_amount(t)
1597 get_data_size(t) -> get_bytes_amount(t)
1598 - Massive cleanups in the functions related to the energy
1599 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1600 - New: MSG_host_get_pstate()
1602 - msg/energy/onoff: switching hosts on and off
1604 * Interface improvement:
1605 - Rename Host.isAvail() to Host.isOn()
1606 - Rename Process.currentProcess() to Process.getCurrentProcess()
1607 - Rename Task.setDataSize() to Task.setBytesAmount()
1608 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1610 - #18874: Actually allows the GC to reclaim tasks
1614 - SIMIX_process_throw: raises an exception in a remote process
1615 * Refactoring: Separate sub-modules
1616 - libsmx: the public interface, as libc in a real system
1617 - popping: the strange dance that converts a user request into a kernel handling
1618 - smx_context_*: the virtualization mechanisms that embed the user code
1619 - smx_*: the handling of each simcalls
1620 * Interface cleanup:
1621 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1622 * Rename smx_action_t into smx_synchro_t, making explicit that these
1623 things are used to synchronize processes with their environment.
1624 For example, a communication is a sort of synchronization involving
1625 the communicating processes (that may block until the exchange) and
1626 the platform. The same can be said from computations, etc.
1628 - #18888: segfault when a process ends before its kill_time
1632 - 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.
1633 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1634 - MPI_Type_set_name, MPI_Type_get_name
1635 - MPI_*_c2f and MPI_*_f2c functions
1636 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1637 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1638 - Activate a lot of new tests from the mpich 3 testsuite
1640 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1641 - InfiniBand network model added: Based on the works of Jerome Vienne
1642 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1643 - When smpi/display_timing is set, also display global simulation time and application times
1644 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1645 * Collective communications
1646 - 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
1647 - 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/).
1648 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1650 - Replay now uses algorithms from wanted collective selector
1651 - Replay can be used with SMP-aware algorithms
1652 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1653 - 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.
1655 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1656 - Correctly use loopback on fat-tree clusters
1657 - Asynchronous small messages shouldn't trigger deadlocks anymore
1658 * Energy/DVFS cleanup and improvement
1659 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1660 - new: smpi_get_host_pstate()
1663 - "Full" network optimization flag was broken since Surf++
1664 - Better handling of precision flags in maxmin
1665 - Fix bug causing sometimes "Impossible" errors
1666 - Properly pass cluster properties to included hosts
1667 * Improvement of the Energy plugin.
1668 - Always update the consumption before returning that value
1669 - New property: watt_off to denote the dissipation when the host is off
1670 - New functions getWattMinAt and getWattMaxAt to retrieve the
1671 dissipation of pstates that we are not currently at.
1672 * Java: class NetworkLink renamed to Link
1673 * New function: simcall_process_get_kill_time()
1674 * Massive rename s/workstation/host/
1675 - That's intrusive, but that's good for the project consistency. Sorry.
1676 - Change config option "workstation/model" into "host/model"
1679 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1680 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1681 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1682 - xbt_dirname and xbt_basename for non-POSIX systems
1684 * The model checker now runs as a separate process.
1685 * The model checker runs must now be launched with the new simgrid-mc program.
1686 * Record/Replay: the MC can display a textual representation of a path in the
1687 execution graph. It can then be replayed outside of the model checker.
1689 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1691 SimGrid (3.11) stable; urgency=low
1696 * Normalizing pointers addresses tool for better diff between logs
1699 * Add cloud examples using new VMs
1700 - examples/msg/cloud/two_tasks_vm.tesh
1701 - examples/msg/cloud/simple_vm.tesh
1702 - examples/java/cloud/cloud.tesh
1703 - examples/java/cloud/migration/migration.tesh
1704 * Add java surf examples:
1705 - examples/java/surfPlugin/surf_plugin.tesh
1706 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1707 - examples/java/surfCpuModel/surf_cpu_model.tesh
1708 * Add SMPI+MSG example:
1709 - examples/smpi/smpi_msg_masterslave/
1714 - msg task destroy cancel test
1715 - msg_host on/off test
1716 * Move all tests in testsuite to teshsuite (adding tesh files)
1717 * Restructure teshsuites
1718 - one folder for each kind of test
1719 * Restructure AddTests.cmake
1721 - structure the order of tests (with sections)
1724 * Add virtual machine
1725 - creation of a VM on a PM
1726 - migration of a VM from a PM to another PM
1728 - MSG_process_join(msg_process_t process, double timeout)
1729 - msg_bar_t MSG_barrier_init(unsigned int count)
1730 - int MSG_barrier_wait(msg_bar_t barrier)
1731 - void MSG_barrier_destroy(msg_bar_t barrier)
1732 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1733 * New option "msg/debug_multiple_use" to help debugging when a task is used
1736 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1738 - Increase code coverage in test suites
1742 * Protect context stacks against stack overflow. The number of protected memory
1743 pages allocated on the top of each stack (1 by default) can be configured
1744 with the new command line option --cfg=contexts/guard_size:<value>.
1745 * Simcalls are now generated by a python script that
1746 - generates files included by SimGrid
1747 - checks that all the functions exist, and proposes prototypes
1749 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1750 host_set_data, host_get_data
1752 - simcall_process_join(smx_process_t process, double timeout)
1753 * Fix bug where sleeping processing could not be suspended.
1756 * Translate surf models from C to C++
1757 - Generic classes for all models: Model, Resource, Action
1758 - A generic interface for each kind of model (CPU, Network, Storage
1759 Workstation, WorkstationVM)
1761 * Translate surf routings from C to C++
1762 * Add callbacks using sigc++ or boost::signals2
1763 - Add callback functions for resource creation/destruction
1764 - Add callback functions for action state change
1765 - Handle Energy as a plugin
1766 * Replace swag by boost::intrusive
1767 * Add new routing models for clusters. For documentation, see
1768 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1769 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1770 parameters for cluster tag
1771 - Fat trees, with topology="FAT_TREE" and
1772 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1773 - see examples/platforms/torus_cluster.xml and
1774 examples/platforms/fat_tree_cluster.xml
1775 * More documentation
1778 * Hostfiles support host:nb_processes construct to deploy several processes on
1780 * Collective communication algorithms should not crash if used with
1781 improper number of nodes and report the error.
1782 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1783 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1784 MPI_Dims_create, MPI_Cart_sub are supported.
1785 * New interface to use SMPI programmatically (still depends on MSG for
1786 some parts, see examples/smpi/smpi_msg_masterslave):
1787 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1791 * Global variables privatization in MPI executables is now performed at runtime
1792 with the option smpi/privatize_global_variables (default:no).
1793 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1794 dynamic libraries loaded by the application are not privatized (static
1795 linking with these libraries is advised in this case)
1798 * Options defined in XML work correctly now.
1801 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1802 copied into simgrid.jar or not (ON by default). Use this option if you want
1803 to reduce the size of the installed simgrid.jar, *and* the native libraries
1804 are kept installed elsewhere.
1805 * Surf binding with SWIG (code generated in maintainer mode only):
1806 - plugin to handle callbacks
1807 - CPU model only for the moment
1810 * Supernovae build mode is definitively removed. It was used to improve
1811 inlining and inter-module optimizations. It is nowadays superseded by
1812 link-time optimizations commonly available in compilers.
1813 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1814 * Add boost dependency for surf++
1815 * Add new macro for tests
1816 - ADD_TESH(name <tesh_args>)
1817 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1821 - xbt_bar_t XBT_barrier_init(unsigned int count)
1822 - int XBT_barrier_wait(xbt_bar_t barrier)
1823 - void XBT_barrier_destroy(xbt_bar_t barrier)
1824 * Make the xbt_os_time module public
1826 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1828 SimGrid (3.10) stable; urgency=low
1830 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1833 * Reintegrate Java to the main archive as desynchronizing these
1834 package is not acceptable anymore (Java is now considered stable)
1835 * Add explicit synchronization facilities through semaphores
1836 * Bug fix: Task.setDataSize() only changed the C world, not the value
1837 cached in the Java world
1840 * Dramatically change the way files are handled. API and internals changed, but
1841 this part of MSG was not considered as production grade either.
1842 * Add explicit synchronization facilities through semaphores
1843 * Add a new function MSG_host_get_process_list()
1844 * Preliminary DVFS support (see examples/msg/energy/ for details)
1847 * SMPI is now included directly in the libsimgrid as the windows
1848 linker doesn't force us on splitting it anymore.
1849 * Improvements of the SMPI replay tool:
1850 - Most of the collective communications are now rooted in the same process as
1851 in the original application.
1852 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1853 used until now). Multiple data types can now be used in a trace.
1854 - The replay tool now supports traces produce either by TAU or a modified
1856 - Bug Fix: the compute part of the reduce action is now taken into account.
1857 - Gatherv collective is now supported
1858 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1859 of smpirun outputs time independent traces for the current run. One file
1860 is created per process. If too many processes are simulated, this behavior
1861 can be changed to one file for all processes by using the
1862 tracing/smpi/format/ti_one_file flag
1863 * smpirun generates the host file if needed (with given host count and platform)
1864 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1865 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1866 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1867 --cfg=smpi/coll_selector:(mpich/ompi)
1868 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1870 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1871 happening inside a collective SMPI call.
1872 * Fix the behavior of complex data types handling.
1873 * Make MPI_Wtime another synchronization point to take computations into
1875 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1876 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1877 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1878 (no privatization of global variables yet)
1879 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1882 * Verification of liveness properties is now available for SMPI applications
1883 (in addition to MSG applications)
1884 * Bugged examples using SMPI in examples/smpi/mc/
1885 * Add --cfg=model-check/visited option. Allows the verification of infinite
1886 programs. Detection of loops in the execution thanks to the system state
1887 comparison and reduction of the state space to explore. Can be combined with
1888 DPOR for safety properties.
1891 * Allow to change SimGrid configuration (see --help) within the code
1892 thanks to SD_config() as it can be done in MSG.
1893 * Add a new function SD_task_set_amount() upon user request.
1896 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1897 * Remove rule based routing (no more PCRE dependency)
1898 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1899 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1900 * Add a loopback_bw and loopback_lat options to cluster tag.
1901 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1902 routing. Now peers are encapsulated in an AS and have their own private
1903 router but this is transparent.
1906 * Our own implementation of getline is renamed xbt_getline, and gets
1907 used even if the OS provide a getline(). This should reduce the
1908 configuration complexity by using the same code on all platforms.
1909 * New type: xbt_cfg_elm_boolean.
1910 * Allow to use yes/no for boolean configuration options in the command line.
1911 * Allow to disable SimGrid cleanups at exit from command line option.
1912 There are situations where one may want a simulation to end with an exit.
1913 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1914 annoying when scripting around the simulator. Adding a
1915 --cfg=clean_atexit:no allows to circumvent this issue.
1918 * Lots of memory leaks were corrected in this release.
1919 There is no known memory leaks anymore, in all of our 600+ tests.
1920 * New command line option --version, to get SimGrid version information.
1921 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1923 * Supernovae builds are deprecated, and expected to be removed in the next
1926 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1928 SimGrid (3.9) stable; urgency=low
1930 The Grasgory release: GRAS is really dead now.
1932 * Complete overhaul of the internal host structures scheme.
1935 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1936 was considered as experimental and badly maintained since then.
1937 * Keeping it was thus a trap to our potential users, that could take
1938 it instead of MSG or SMPI by mistake despite is pity state.
1939 * GRAS seems to have very few users (if any), and no one volunteered
1940 to maintain it further. It also induces a lot of XBT code (for
1941 portability sake), that must be maintained too.
1942 * For all these reasons, we killed GRAS. If someone wants to revive it
1943 in the future, don't cry, our git history still remembers of GRAS.
1946 * Major overhaul. Merge our documentation again as time proved that
1947 splitting it was really not helping our users.
1948 * Further improve the developer documentation to help newcomers
1949 hacking on SimGrid itself. The user documentation (and in
1950 particular, the beginner documentation) is still in a sorry state.
1953 * Now works on Windows too!
1954 * Much more extensive test suite, from MPICH
1957 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1958 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1959 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1960 * Bug fix: let task be scheduled when the last dependency to be solved is
1961 a control dependency.
1962 * Remove SD_load_environment_script function.
1963 Use the C sg_platf function if you want to declare a platform
1967 * New function: MSG_process_get_number()
1968 * Old function documented: MSG_config()
1969 * Remove MSG_load_platform_script function
1970 Use the C sg_platf function if you want to declare a platform
1974 * Change the default value of the TCP_gamma constant (maximal size of TCP
1975 congestion window) to a more realistic 4MiB value. If you notice changes in
1976 your simulation results, you can fall back to the previous 20k tiny window
1977 by adding --cfg=network/TCP_gamma:20000 on command line.
1978 * (Hopefully) fix a bug wrt periodic availability and state traces
1979 * Bug fix: use default values at start when first event in availability/state
1980 trace is not at time 0.
1983 * remove the "new_" part of function name sg_platf_new_trace_connect
1984 (resulting in sg_platf_trace_connect), since it does not create
1988 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1989 use case where it's really mandatory, and maintaining it was a pain in
1991 * New: xbt_fifo_search(), search an item with a user-provided
1992 comparison function instead of dumb pointer comparison.
1995 * Fix the lua deployment:
1996 Use `simgrid.init_application()` before deployment instead of
1997 `simgrid.msg_register_application()` after.
2000 * Transfer the tracing files into the corresponding modules.
2002 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2004 SimGrid (3.8.1) stable; urgency=low
2006 The "we are told that some people want to also *install* the simgrid
2009 * Add missing file "tesh.1" to the archive.
2011 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2013 SimGrid (3.8) stable; urgency=low
2015 The Psssshiiiit release: SimGrid jumps into the Cloud.
2018 * Add an experimental interface to manipulate VMs. They are mainly
2019 process groups with very few intrinsic semantic, but they should
2020 allow you to build the semantic you want easily.
2021 * New function: MSG_host_set_property_value()
2022 * New function: MSG_process_on_exit(). To clean memory in all cases.
2023 * Bug fixes that made the host (and link) failures unusable.
2024 * Add a way to auto-restart process when the host in which they are
2025 executing comes back (ON_FAILURE="RESTART" on deployment file,
2026 MSG_process_auto_restart_set).
2027 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2028 please stop using the old ones, they are DEPRECATED.
2030 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2031 Please use MSG_init() instead. (reducing the amount of entry
2032 points in the library helps us).
2033 * Make it impossible to link against the wrong version of the lib
2034 * Deprecate MSG_clean(). No need to call it anymore.
2035 * Function MSG_get_host_number() is not deprecated anymore.
2038 * Split the doc into a user guide and a reference guide.
2039 * Start a developper guide to help people hacking on SimGrid.
2042 * Enable tracing by default. This modules rocks you should use it.
2043 * Remove option custom_flags. Now use environment variables CFLAGS
2045 * Use default cmake things to detect lua instead of home grown ones.
2046 * New option "enable_mallocators" to disable mallocators, for debugging
2047 purpose ("on" by default).
2050 * Bug fixes around the resource failures: don't let the processes
2051 survive the host they are running onto.
2052 * Add an interface to auto-restart processes when the host in which they are
2053 executing comes back.
2054 * Ensures that SIMIX_clean is called automatically. It's not part of
2055 the public interface anymore (bindings should be updated).
2058 * Bug fix for when SD_Simulate is called with a positive value: be careful
2059 when comparing doubles. Sometimes they are different for non significant
2061 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2062 parallel task whose initial work is distributed among host according
2063 to the Amdahl's law. Such tasks are created with a parameter alpha
2064 that corresponds to the non-parallelizable part of the computation.
2065 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2066 between two sets of workstations assuming a 1D block distribution (each
2067 workstation owns a similar share of data) on both sides.
2069 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2070 Data redistribution will be automatically scheduled once parent and child
2071 are both scheduled. The filling of computation_amount and
2072 communication_amount structures is now done seamlessly thanks to the chosen
2074 * New function SD_workstation_dump to display various information
2075 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2076 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2077 bandwidth on the route onto which the task is scheduled and the amount of
2079 To divide the nominal bandwidth by 2, the rate then has to be:
2080 rate = bandwidth/(2*amount)
2081 * Compute tasks that have failed can now be rescheduled and executed again
2082 (from their beginning)
2083 * Increasing source code coverage (src/simdag is now covered at 95.8%
2087 * Re-implement time-independent trace replay using SMPI (at the
2088 smpi_smp_* level) instead of MSG. This should replace
2089 examples/msg/actions/actions.c
2090 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2091 hindexed and structs)
2092 * Implement the exchange of non-contiguous data.
2093 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2094 * Correct behavior of smpi/sender_gap and set its default value to 0
2095 * Add option to asynchronously send small messages to allow better
2096 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2097 specifies the size in bytes under which messages will be asynchronously sent.
2098 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2099 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2102 * New C interface to define a platform: XML is now optional.
2103 For more info, please check include/simgrid/platf.h.
2104 * New interface to define random platforms from the C:
2105 For more info, please check include/simgrid/platf_generator.h and
2106 examples/msg/masterslave/masterslave_platfgen.c
2107 * Export a sg_cmdline dynar containing all the arguments we got from
2111 * Two new tracing options for adding comments to trace file so you
2112 can track your experiments (see --help-tracing for details).
2113 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2114 * Adding the SimGrid version that generated the trace file as a comment.
2115 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2116 * Fix to avoid key clashes on Paje links
2117 * Other minor fixes related to the Paje specification
2120 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2121 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2122 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2123 * Greatly improve the robustness of mmalloc to user errors (such as
2124 using an area after freeing it, or freeing it twice)
2126 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2128 SimGrid-java (3.8.1) stable; urgency=low
2130 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2131 Warning: all methods are visible, but only some of them are
2132 implemented so far. Check the source (src/jtrace.c)
2133 for further information.
2134 * New module: org.simgrid.msg.File (SimGrid File management functions)
2135 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2136 * Change the meaning of Process.restart: now restart the process from
2137 the beginning, like MSG_process_restart in C.
2138 * Add Process.setAutoRestart: handling of process restart when failed
2140 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2141 you to retrieve the properties of the processes/hosts
2142 * Deprecate Msg.clean(): you can just forget about it now.
2143 * New function Process.getCount(), that only works when compiling
2144 with the not yet released version 3.9 of the C library.
2146 * New context factory based on Coroutines. It mandates a modified JVM
2147 but then, the simulations run about five times faster, and there is
2148 no limit to the amount of processes (beside of the available memory).
2150 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2152 SimGrid (3.7.1) stable; urgency=low
2155 * Restore the prototype of MSG_process_create_with_environment() to
2156 the pre-3.7 situation by removing the kill_time argument.
2157 * Add a MSG_process_set_kill_time() function instead.
2160 * Fix weird behaviors when dealing with parallel tasks.
2163 * Simgrid is now built as a dll.
2164 * Simgrid-java now works on Windows.
2165 * Simgrid-Java is now included into Windows package.
2168 * First pre-build package for MacOSX.
2171 * Fix compilation when using MSG_USE_DEPRECATED.
2172 * Fix some compilation issues on Macs and Windows.
2173 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2174 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2176 We discovered that the Lua console is broken, but we are missing the
2177 manpower to fix it right now. The problem existed in 3.7 too, so we
2178 are not blocking the release for that. Sorry if you depended on this
2179 feature, any help would be really welcome.
2181 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2184 SimGrid-java (3.7.1) stable; urgency=low
2186 The "Java aint got to be bloated and slow" release
2189 * Various internal cleanups and performance improvement
2190 Simulations are expected to run up to twice faster or so
2191 * Make Process.kill(process) an instance method, not a static one
2192 * User processes are not java.lang.Thread subclasses.
2193 This breaks the compatibility (sorry), but previous API was
2194 brain-dead, making it impossible to have non-trivial
2195 initializations in the process constructor.
2196 * Require a full constructor per Process sub-class.
2197 Kinda breaks the compatibility (sorry), but this allows a much more
2198 efficient way to launch the processes at simulation startup.
2199 * Do not embeed our version of semaphores, java 1.5 can be considered
2200 as sufficiently prevalent for us to not dupplicate its features.
2204 * Add examples for almost every part of the API
2205 We spotted and fixed a lot of bugs in the process
2207 * New module: asynchronous communication API
2208 * New function: Process.sleep()
2209 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2210 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2211 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2213 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2215 SimGrid (3.7) stable; urgency=low
2217 The "spring cleanups (before the next Big Project kicks in)" release.
2220 * We can specify the SMPI latency/bandwidth factor with command line
2221 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2222 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2223 You can also use the "config tag" from platform file by adding this line
2224 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2225 (see "example/platforms/tag_config.xml" to use "config tag").
2226 Note that the command line supersedes the platform file configuration.
2227 * Change the correction factors used in LMM model, according to
2228 the latest experiments described in INRIA RR-7821.
2229 Accuracy should be improved this way.
2230 * Use the partial invalidation optimization by default for the
2231 network too. Should produce the exact same results, only faster.
2232 * Major cleanup in surf to merge models and split some optimization
2233 mechanisms from the core of the models. As a result you can now
2234 specify which model to use (e.g., --cfg=network/model:LV08
2235 --cfg=cpu/model:Cas01) and which optimization mode to use
2236 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2237 Incompatible combinations should err at initialization. See
2238 --help-models for the list of all models and optimization modes.
2239 * The CLM03 workstation model was dropped for simplicity because it
2240 used the deprecated CM02 network model. Use default instead.
2241 * Rename the TCP_gamma configuration option to network/TCP_gamma
2242 * Rename the coordinates configuration option to
2243 network/coordinates, and document it
2244 * Use now crosstraffic keyword instead of the terribly misleading
2245 fullduplex keyword. It is activated by default now in the current
2246 default model, use --cfg=network/crosstraffic:0 to turn it off.
2247 * Ongoing refactoring the model parsing to make XML files optional
2248 See include/simgrid/platf.h for details (still to be completed)
2251 * Major overhaul of the documentation. Almost instructive now :/
2252 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2253 functions and friends. This interface was considered as
2254 deprecated since over 2 years, it's time to inform our users that it is.
2255 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2256 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2257 use these (crappy) functions in your code.
2258 These functions will be removed soon. Stop using them now.
2259 * Deprecate MSG_get_host_{table,number}
2260 Implement MSG_hosts_as_dynar() instead.
2261 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2262 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2265 * Stabilize the parallel execution mode of user contexts
2266 * Introduce configuration variables to control parallel execution:
2267 - contexts/synchro: Synchronization mode to use when running
2268 contexts in parallel (either futex, posix or busy_wait)
2269 - contexts/parallel_threshold: Minimal number of user contexts
2270 that must be part of a scheduling round to switch to parallel
2271 execution mode (raw contexts only)
2272 * Fix bugs that prevented to use suspend/resume along with
2273 synchronization structures.
2274 * Fix bugs in process termination that lead to invalid memory access
2275 in very specific conditions.
2278 * Introduce a parallel mode for the models (controlled by surf/nthreads
2279 configuration item). In our tests, running the models in parallel
2280 never lead to any speedups because they are so fast that the gain
2281 of computing each model in parallel does not amortizes the
2282 synchronization costs, even when ultra fast futexes are used.
2283 This is released anyway because YMMV.
2286 * Performance boost by using a swag internally to compute the set of
2287 tasks that are finished and should constitute the return value of
2291 * Enable it by default now that it is considered rather stable.
2294 * Documentation of the tracing functions.
2295 * Performance gains when tracing categorized/uncategorized resource
2296 utilization by avoiding calls to get route when updating resource
2297 variables. LMM constraints are being used instead.
2298 * API changed to set task categories. Use MSG_task_set_category instead
2299 of TRACE_msg_set_task_category, and SD_task_set_category instead
2300 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2301 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2302 * Examples for link user variables added (see at examples/msg/tracing/)
2303 * Deprecated function TRACE_msg_set_process_category completely removed
2304 * Trace header updated according to the latest Paje file format
2305 * Tracing network lazy updates, no longer obligate users to use full updates
2306 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2307 * Experimental: let user code declare/set/push/pop application states for hosts
2308 * API changed to allow the manual creation of graph configuration files
2309 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2312 * Improve the API of Lua MSG bindings, using the Lua spirit.
2313 * Each simulated process now lives in its own Lua world (globals are
2314 automatically duplicated). It helps writing simulators. Will allow
2315 to run Splay programs within SimGrid in the future.
2316 * Add a Chord example in Lua, equivalent to the MSG one.
2319 * Start the implementation of a solution to express temporal
2320 properties, not only local assertions. This is still an
2321 experimental work in progress, stay clear from it to be safe.
2325 - Add new runtime parameters --help-logs and --help-log-categories
2326 to display information about supported logging parameters and
2328 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2329 don't exists anymore.
2330 * Mallocators: allow value NULL for the reset function.
2332 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2333 create homogeneous dictionaries, where all the elements share the
2334 same free function. Non homogeneous dictionaries will be
2335 deprecated in the next release.
2336 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2337 - Multi-level dictionaries are deprecated.
2339 - new function xbt_dynar_search_or_negative() that is useful when
2340 you have less than 2 million elements in your dynar and don't
2341 want of the extra complexity of catching exceptions when the
2342 element is not found.
2344 - Make xbt_os_thread module (for thread portability) public.
2345 Documentation is still to come, sorry.
2347 - Cleanups and simplifications to make it maintainable again.
2348 - Exotic features (such as memalign and valloc) were removed.
2349 - The metadata were extended and improved so that the
2350 model-checker becomes able to explore and inspect the heaps.
2351 - This may induce a performance drop when enable_model-checking is
2352 ON in cmake (even if it's not used in the simulation), but it is
2353 necessary at this point to get MC working.
2355 Turn model-checking OFF if simulation performance matters to you.
2356 Not enabling it at runtime is not enough, disable it in cmake.
2358 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2361 SimGrid (3.6.2) stable; urgency=low
2363 The "Not coding new stuff allows to polish old things" release.
2366 * New bindings to the NS3 packet level simulator (experimental)
2367 * Use the raw (efficient) execution contexts instead of the sysv
2368 (portable) ones when possible.
2369 * libpcre is now mandatory in any cases since not using it led to
2370 severe performance loss and possibly other issues
2371 * Update the XML platforms:
2372 - G5K: include the latest machine in Nancy
2373 - GridPP and LCG: new platforms
2374 * Documentation was partially updated, at least (more to come)
2376 Bug fixes, cosmetics and small improvements
2377 * Free terminated processes before the end of the simulation to avoid
2378 exhausting the memory of users having very dynamic amount of
2380 * Bug fix and cosmetics about canceling non-running tasks
2381 * Bug fix about the dot loader's issues when using libcgraph
2384 * Create an installer for windows with nsis (amd64 and win32)
2385 - Add an hello world project to illustrate simgrid project creation.
2386 - Embed libpcre into the Simgrid installer to avoid
2387 its compilation burden
2388 * The raw execution contexts should work on Apple now
2389 * Port to Windows 64 bits
2390 - Sysv contexts now have an implementation for this arch
2391 - GRAS communication features now support this arch
2392 * Drop support for borland compiler on windows
2393 - this code was not maintained, and we kinda depend on gcc nowadays
2394 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2395 * Fix portability issue on unstable ubuntu: linker became picky on
2398 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2400 SimGrid (3.6.1) stable; urgency=low
2402 The "Oops, we broke Macs too" release
2405 * Fixed contexts detection so that raw ones are used when possible
2406 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2407 leads to a strange error, with user code segfaulting sometimes when
2408 the generated code is not perfectly aligned (which is not
2409 controllable from the user side, depends on the amount of code)
2412 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2414 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2416 SimGrid-java (3.6) unstable; urgency=low
2419 * Split of every thing from simgrid v3.5 into a separate package.
2421 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2423 SimGrid (3.6) stable; urgency=medium
2425 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2428 * Bindings now constitute their own package, separated from the main one.
2429 Rationale: reduce our maintenance nightmare by reducing the module coupling
2430 They will soon be released on their own on gforge.
2432 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2433 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2435 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2436 * It's not quite deprecated for now because we have no replacement,
2437 but it may soon become the case.
2440 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2441 * Fortran: New user-level cache variable to store the rank of the running
2442 process. This improves performance by an order of magnitude.
2443 * C: New coccinelle script to automatically locate and modify global and
2444 local static variables.
2445 * Improved SMPI network model with a sender-side gap to account for multiple
2449 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2450 only say if a communication is finished, no matter whether it succeeded or
2451 failed. You can call MSG_comm_get_status() to know the status of a finished
2453 * New function MSG_task_dsend() to send a task and detach it. When a
2454 communication is detached, you are never notified of its success or failure
2455 and the memory is released automatically once it is finished. This function
2456 is useful when you don't care about the end nor the success of a
2458 * Change the prototypes of action replay. Sorry for inconvenience,
2459 but this is really more efficient this way (and to adapt your code,
2460 you just have to fix the initialization, that shouldn't be too long)
2461 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2462 ever managed to do anything useful with it.
2463 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2464 takes a dynar of communications. It returns immediately and gives the
2465 index of a finished communication (if any).
2466 * New example: a basic implementation of the Chord P2P algorithm.
2469 * New model for multi-core CPUs. You can now use the core attribute to
2470 precise the number of cores of a host. This is a basic model. Every
2471 process running on the host receives at most the power provided in
2472 the DTD (throughput<=power). Total throughput of process cannot exceed
2474 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2475 router linked by an up-link and a down-link (possibly asymmetrical).
2476 This kind of pattern allows to easily build last-mile model style platforms.
2477 Aggregating such patterns in a rule-based AS is thus the technique of
2478 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2479 * New model for Vivaldi routing. We transformed the Vivaldi network model
2480 into a Vivaldi routing model (based on the rule-based model). This allows to
2481 combine Vivaldi based latencies with last-mile platforms.
2484 * Added a check for NaN of IEEE754 infinite in the double entries of
2486 * Introduce a new context factory "raw", highly inspirated from the
2487 ucontext factory, but using manually crafted functions in assembly to
2488 do the work in an efficient manner.
2489 * Allow to change the used context factory at run time, not only at
2490 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2491 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2492 contexts at runtime (only with raw contexts or ucontexts).
2493 * Completely rewrote this module to allow parallel execution of user
2494 processes. Use --cfg=contexts/nthreads:N to execute user processes
2495 with N parallel threads (the default is 1, meaning no parallelism).
2496 * Allow to decide dynamically between sequential and parallel modes.
2497 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2498 processes in parallel only when their number is greater than or equal to P
2500 * Added a check for NaN of IEEE754 infinite in the double entries of
2504 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2505 won't output the state of processes when interrupted with Ctrl-C
2506 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2507 NULL-terminated array. This may solve backward compatibility issues
2508 due to the change to return type of SD_simulate. See also:
2509 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2510 * Add new macros with variable number of arguments.
2511 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2512 - in xbt/asserts.h: xbt_assert
2513 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2514 - in xbt/ex.h: THROWF and RETHROWF.
2515 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2517 * Change xbt_die() to accept a format string with arguments, just like printf.
2518 * New data structure: xbt_lib_t, like a dict but more general and with better
2522 * New configuration options
2523 Options triva/categorized and triva/uncategorized can be used to generate
2524 graph configuration files for Triva visualization tool.
2525 * Configuration option tracing/platform is renamed to tracing/categorized
2526 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2527 * New TRACE_user_link_variable interface:
2528 User provides the name of the link and the tracing variable to attach to it
2529 * the declaration of tracing categories must be done after the environment
2531 * simpler tracing interface, just one way to declare categories
2532 TRACE_category or TRACE_category_with_color, it is up to you
2533 * links in the trace file are again identified by their names
2534 * trace contains the full platform hierarchy exactly as declared using the ASes
2535 * Options tracing/msg/[task|process]:1 groups the process by hosts
2536 for both cases, tasks and processes must have names that are unique during the simulation
2537 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2538 * The experimental option tracing/msg/volume is deprecated
2539 its functionality may be reincorporated if needed
2541 The tracing generates a trace file with unordered timestamped events,
2542 because of the way the core simulator (surf) works. A script available
2543 at the tools directory (fix-paje-trace.sh) can be used to put the events
2544 in order. We have changed the tracing so it can generate ordered timestamped
2545 events in the final trace, but depending on the simulator (and how much time
2546 is simulated) that can lead to a huge memory utilization. It is deactivated
2547 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2549 Build Infrastructure
2550 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2551 - We are trying hard to keep the API stable, but it may happen that
2552 some things change (we're a research project after all, not a
2553 nuclear plant operating system). If such things should happen, you
2554 could rely on that macro to adapt.
2555 - current value: 30600 for 3.06.00, aka 3.6
2556 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2557 a number that can be compared with SIMGRID_VERSION.
2558 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2559 assertions and verbose/debug logging events are disabled at compile time.
2561 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2563 SimGrid (3.5) stable; urgency=medium
2566 * New feature to any SimGrid-based simulator: Model-Checking
2567 Check SIN#1 for more details.
2570 * New Model SMPI (three-interval linear regression for correction factors)
2571 See RR-7426, available at http://hal.inria.fr/inria-00527150
2572 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2573 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2574 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2575 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2576 * See also src/smpi/README
2581 - Tracing API changes: TRACE_start and TRACE_end should not be called
2582 by user-code. They are automatically called by simulators created
2583 with SimDAG, MSG and SMPI if the toolkit is compiled with
2584 tracing_enabled=ON. Categories declaration and utilization remain the
2585 same for MSG and SimDag.
2586 - A function was added to the tracing API to declare categories with
2588 - TRACE_category_with_color (char *category, char *color)
2589 where color must be in the following format
2590 "%f %f %f", red, green, blue
2591 and red, green, blue are float values in the interval [0, 1]
2592 - User can specify NULL as color parameter, or continue calling
2593 TRACE_category (cat)
2594 On that case, the tracing system will define random colors
2595 - The following command-line options are supported:
2596 --cfg=tracing/filename:msg.trace
2597 --cfg=tracing:1 (activate tracing, needed to use others)
2598 --cfg=tracing/platform:1 (categorized resource use)
2599 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2600 --cfg=tracing/msg/task:1 (task creation)
2601 --cfg=tracing/msg/process:1 (process creation, migration)
2602 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2603 --cfg=tracing/smpi:1 (SMPI interface tracing)
2604 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2605 - examples of examples/msg/tracing updated
2607 - DAXLoader and DOTLoader functions can generate tasks with categories
2608 - A new function to attribute a category to SD tasks:
2609 TRACE_sd_set_task_category (SD_task_t task, char *category)
2610 * Tracing the MPI interface implemented by SMPI
2611 - Collective operations are traced with states
2612 - Point-to-Point operations are traced with states/links
2613 - Tracing activated by a parameter "-trace filename" passed
2614 to smpirun during execution (considering that simgrid
2615 is compiled with tracing enabled)
2616 - To run the simulation with gdb, the simulator
2617 accepts --cfg=tracing/smpi:1 to trace SMPI
2618 - tesh files to check if smpi tracing is ok
2619 - See examples/smpi/NAS/DT-trace
2620 * GTNetS tracing re-worked
2621 - adaptation to the tracing system of GTNets to cope
2622 with modifications regarding the fullduplex mode
2623 - new tesh files to check if gtnets tracing is ok
2626 * Asynchronous communications through the functions:
2627 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2628 * New function: MSG_load_platform_script()
2629 to make possible using a lua script instead of XML files to set up platforms
2630 * New function: MSG_set_function
2631 to associate functions to processes, used when bypassing the parser
2632 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2634 Platforms: Add some more examples in examples/platforms
2635 * Grid'5000: see www.grid5000.fr
2636 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2639 * Change the XML format. This is a very important modification. SimGrid
2640 3.5 introduces a new hierarchical format based on the notion of
2641 Autonomous Systems. Compatibility with old format is ensured
2642 through the perl script provided in the install bin directory
2643 bin/simgrid_update_xml.
2644 It is now possible to build platforms with specific routing
2645 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2646 connect several platforms together. We will try to provide soon
2647 set of realistic platforms exploiting these properties (have a
2648 look at examples/platforms/ for the moment).
2649 * Take the opportunity of the XML format change to be a good XML citizen:
2650 rename link:ctn to link_ctn and similar changes (also dealt with by
2652 * Add a new routing scheme (rule-based) using regular expressions. It
2653 enables to have an extremely low memory footprint when the
2654 underlying routing is simple and can be compactly described. You
2655 need to have libpcre4-dev (perl regular expressions) installed if
2656 you want to use this routing scheme.
2657 * Revive the cluster TAG and allow to easily and efficiently (both in
2658 term of memory and speed) connect clusters together. Have a look
2659 at teshsuite/simdag/platforms/ to see how this can be done. With
2660 this tag, you can create clusters with thousands of tasks at no
2661 cost (have a look at examples/platforms/).
2662 Note: clusters are implemented as ASes, so there is no need for an
2663 enclosing AS tag if you have only one cluster in your platform.
2664 * Add new generic functions in the public interface that allows the user
2665 to call SURF 'create_resource' methods from your code (same
2666 functionality as the XML bypass mechanism but with a much lighter
2668 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2669 that uses a piecewise linear approximation to produce better
2670 results when exchanging small messages.
2671 * Add a new parameter to handle correctly full duplex link and account
2672 for interferences between uplink and downlink communications
2673 (activate with --cfg=fullduplex:1).
2676 * Rename the SD_READY (all dependencies are satisfied and task is
2677 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2678 dependencies are satisfied) state.
2679 This prevents a confusion between the notion of "ready to schedule"
2680 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2681 simulated" (SD_RUNNABLE) used by the simulation kernel.
2682 * Change the way a task is considered as ready. Instead of removing
2683 dependencies when a task is done, a counter is decreased. This way,
2684 it is always possible to reach ancestors thanks to the
2685 SD_taks_get_parents function (even after the end of the simulation.)
2686 * Change the return type of SD_Simulate from (SD_task_t*) into
2687 xbt_dynar_t. This function was in handling a dynar internally and
2688 converted it into a NULL terminated array for historical reasons.
2689 * New function SD_dotload(char*) to load a DAG described in dot
2690 format. This loader and the corresponding examples require the
2691 installation of the graphviz library.
2692 * Fix a bug in the management of tasks of size 0 in the surf network
2693 models. This problem was only visible with SIMDAG and you should
2694 thus disregard results produced with earlier versions if you
2695 relied on this feature (some tasks were blocked because of this).
2696 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2697 even though your applications were doing only point-to-point
2698 communications and sequential computations. Now you can really use any
2699 model you want (of course, if you create real parallel tasks, which are
2700 not implemented in most models beside ptaskL07, this will abort).
2701 * Add an example that schedules a DAX on an heterogeneous platform
2702 using a Min-Min strategy.
2703 * New function SD_workstation_get_current_task() that returns the kind
2704 of task currently running on a workstation in the sequential access
2706 * Raise some warnings when unexecuted tasks remains at the end of the
2707 simulation. This is usually caused by cycles in the DAG.
2710 * New function: SIMIX_process_set_function() called by MSG_set_function
2711 * Change the underlying waiting queue in semaphores so that a process
2712 can wait on several of them simultaneously (as in waitany).
2713 * Fix the way to handle tokens in semaphores so that all access patterns
2714 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2715 * kill the dirty pimple SIMIX_message_sizes_output()
2716 Please use (proper) visualization instead
2719 * New data container: setset (set of sets of elements)
2720 * New module: mmalloc (mapped malloc, allowing to have several
2721 independent segments of malloc)
2722 * New function: xbt_dict_cursor_set_data()
2723 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2724 * New function: xbt_dynar_is_empty()
2725 * New function: xbt_fifo_get_last_item()
2726 * Fix xbt_dynar_shrink(): use the right element size.
2727 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2728 uninitialized areas during expand.
2729 * Fix semaphores: previous implementation was severely broken.
2730 * Use library init/fini functions for our initialization.
2731 - you can use logs and other feature as soon as you want in your
2732 code (even before the xbt_init / MSG_init)
2733 - xbt_exit is now a no-op and produce a warning when used.
2736 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2737 benefit from the latest improvement to the simulation kernel.
2738 * Kill measurement sockets for now. If you rely on them, sorry. This
2739 release is not for you. This feature will be reintroduced in the
2740 future, but we cannot delay the release any further.
2741 * New function: gras_msgtype_get_name().
2742 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2743 * Fix (at last) the pmm example: it should not randomly fail anymore.
2745 Build chain: bug fixes and overall polishing
2746 * Cmake is now stable enough. Hence, we killed the autotools.
2747 * Port to windows ( TM :)
2748 * Fix the 'make install' target.
2749 No need to use 'make install-simgrid' anymore
2750 * Introduce a 'make dist' target compiling a *source* archive
2751 'make package' compiles a binary archive
2752 * Compile java files only on need
2753 * Add --cd and --setenv command line options to tesh
2754 * Out of source builds are not fully supported yet, but we are close
2755 * Enable supernovae and optimization flags by default for our users
2758 * Add layer to set up environment directly from lua, without XML.
2759 * The effect of gras_stub_generator can be achieved through
2760 lua too (check examples/gras/console/ping_generator.lua)
2762 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2764 SimGrid (3.4.1) stable; urgency=low
2766 The "Polishing easter eggs is probably a good idea" release.
2767 This is a bug fixes release only.
2771 * Fix a bug preventing the tasks from begin garbage collected.
2774 * Fix a bug occurring when a host involved in a communication fails.
2775 This was not detected properly by the other peer involved in the
2776 communication. Now, it's reported as a network error.
2779 * Warn the user about loop dependencies in data flow of DAX files
2780 * Obey the control-flow dependencies of DAX files
2783 * Add option "enable_smpi" allowing to not compile SMPI.
2784 Probably useful for the (Mac) users experiencing a build error here
2785 * Improve the detection of lua5.1 and ruby1.8
2787 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2789 SimGrid (3.4) stable; urgency=low
2791 The "Easter in Cargese" release. Also known as (major changes):
2793 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2794 ~> bindings were greatly improved
2795 ~> new tracing infrastructure for better visualization introduced
2797 * the "Welcome to configury modernity" release.
2798 ~> we switched from autotools to cmake, and improved our cdash
2801 A more detailled list of changes follow (full detail in svn log).
2803 Java Bindings: Various Cleanups
2804 * (install java-gcj-compat-dev on debian-like to use them)
2805 * Remove put/get: no need to export deprecated interface in Java
2806 Use send/receive instead.
2807 * Cleanup the examples and add a README per directory
2808 * Remove example autoDestination (that's the only way to go now)
2809 * Remove example explicitDestination (was a plain copy of basic)
2810 * Make JniException a runtime exception, so that there is no need to
2811 declare the fact that you may encounter such a beast. I guess that
2812 nobody will ever want to survive such error.
2813 * Create specific errors for each MSG case of failure:
2814 host failure, transfer failure, timeout, task cancelled
2815 * Cleanup the exceptions that may get thrown by each function
2816 * Other internal cleanups in Java bindings. Performance still bad :/
2817 Ruby and Lua Bindings: create them
2818 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2819 * That's new and great, you should try them out.
2820 Same functionalities than Java bindings, only even less polished
2822 * Kill the useless "rate" argument of SD_task_get_execution_time()
2823 Everyone used to provide -1 as a value, it was not used, and the
2824 semantic of a possible use wasn't even clear.
2825 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2826 either as comm costs or compute costs to mean that there is no
2827 such thing for that specific task.
2828 * Add a SD_task_set_name() function
2829 * Fix SD_task_unschedule() on typed tasks
2830 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2831 * In DAX loader, accept useless 'level' attributes to <job> since
2832 LIGO DAGs have them (seem to be to ease graphical representation).
2834 * Add an example masterslave_mailbox.c using send/receive and not
2835 the deprecated put/get interface.
2836 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2837 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2839 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2840 (declare MSG_USE_DEPRECATED to still have the old name)
2841 * Add a MSG_task_set_data() function
2842 * About trace replay (see examples/msg/actions):
2844 - Allow to work with splitted trace files for each process
2845 Give the specific trace file as argument of each process,
2846 and call MSG_action_trace_run(NULL)
2847 You can still have one merged file for all processes.
2848 - Fix implementation of collective operations
2849 * Allow task_execute() on 0-sized tasks (closes #10063)
2851 * This is the first release of SimGrid where SMPI is not considered
2852 beta anymore (even if some corners should still be improved)
2853 * Port over the new SIMIX_network submodule (internal refactoring)
2854 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2855 * Implement more missing elements of the standard:
2857 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2859 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2860 MPI_Scan() MPI_Get_processor_name()
2861 - Added implementation of missing case for Alltoall (warning: it's
2862 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2863 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2864 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2865 SMPI_MPI_Allgatherv()
2866 * Bug fixes include:
2867 - MPI_Waitsome() was broken
2868 - Allow relative includes in smpicc
2869 - Command line cfg argument 'reference_speed' was ignored...
2870 - Some functions did not properly lead to auto-benching of user code
2871 - smpicc passes -O2 by default (just like openmpi one)
2873 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2874 * Bug fixes about timeouts during communications
2875 * add SIMIX_message_sizes_output() as a pimple to write to file the
2876 amount of messages per size. Use gnuplot to get histogram.
2877 Pimple because that's the only user-visible function of simix,
2878 defined directly in xbt.h (irk, sorry)
2880 - Add a SIMIX_sem_get_capacity() function
2881 - Fix interactions with processe resume/suspende
2882 - release_forever() was stupidly broken
2883 - Fix SIMIX_display_process_status() for processes in a semaphore
2884 - Make SIMIX_sem_block_onto() user-visible
2885 * Refactoring context stuff:
2886 - Use pseudo-OOP for better modularity
2887 - reimplement SIMIX_process_kill() without process_schedule() so
2888 that the latter can take as invariant that it is called from
2890 - Merge context_start into context_new for sake of simplicity
2892 * Add a Vivaldi network model, coded live during SUD'10 ;)
2893 * Rename configuration variables to start a hierarchy:
2894 o cpu_model -> cpu/model
2895 o network_model -> network/model
2896 o workstation_model -> workstation/model
2897 * New configuration variables:
2898 o network/bandwidth_factor: correction to bandwidth
2899 o network/latency_factor: correction to latency
2900 o netwotk/weight_S: correction to the weight of competing streams
2901 * Add a long description to the models, that users can see with such
2902 argument on the command line: --cfg=cpu/model:help
2903 * --help-models display the long description of all known models
2905 * config: add the ability to set a default value after registration
2906 Does not override any previously set value (e.g. from cmd line)
2907 * dict: allow to have integer key and data.
2908 When so, you need to use the following functions
2909 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2910 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2911 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2912 In contrary to regular dicts, the key is not malloced before copy.
2913 Mixing scalar and regular elements in the same dict is not tested
2915 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2916 Tracing for Visualization:
2917 * SimGrid is now instrumented in order to generate a trace file for
2918 visualization analysis: to use it, need to compile SimGrid with the
2919 "tracing" option enabled, and instrument the program using SimGrid with
2920 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2921 (among other functions).
2922 * The instrumentation only traces the platform utilization for now
2923 * Documentation to use the tracing functions and how to analyze the
2924 traces with the Triva tool is written.
2925 * More information about: SimGrid FAQ (in the section Tracing Simulations
2928 * We moved to cmake as default build system. Autotools support will
2929 be dropped soon. Check the FAQ for more info about how to use it.
2930 * Greatly improved our cdash/ctest interactions
2931 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2932 * Added memory checking tests with valgrind; lot of memleak fixing.
2933 This may be the first release of simgrid with so few memory issues
2934 * Added code coverage tests.
2935 Our coverage is still improvable, but at least we see it on cdash.
2937 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2939 SimGrid (3.3.4) stable; urgency=low
2941 The "Desktop Grid needs love too" release (also called Xmas release).
2943 Models improvements:
2944 * Major speedup in the maxmin system solving by using lazy evaluation
2945 Instead of solving completely the maxmin system at each iteration,
2946 only invalidate (and recompute) the modified parts.
2947 This new feature is enabled in default models but you can try to
2948 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2949 * Cas01 IMproved as default CPU model
2950 This CPU model is the same Cas01 model, but it uses the
2951 maxmin-selective-update flag and a heap structure to manage
2952 actions on SURF kernel.
2953 It reduces the complexity to find the next action to finish and,
2954 consequently, it's faster than the old Cas01.
2955 This is the new default CPU model (Cas01).
2956 * Rename the old Cas01 model to Cas01_fullupdate
2957 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2958 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2959 * CpuTI (CPU Trace Integration)
2960 A new CPU model whose objective is simulate faster when using
2961 availability trace files.
2962 Instead of using a full featured, over engineered maxmin system for
2963 CPU modeling, this model does the pre-integration of traces files
2964 to calculate the amount of CPU power available, and so, executes
2965 faster than the old CPU models.
2966 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2967 * Use LV08 as default network model since it gives better accuracy
2968 for small messages and shouldn't change things for big ones.
2969 Use --cfg=network_model:CM02 to get the previous behavior.
2972 ******************************************
2973 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2974 ******************************************
2975 * The new CPU model may changes simulations!
2976 The point is that events occurring at the exact same timestamp
2977 are not scheduled in the same order with the old and new
2978 version. This may be enough to completely change the execution
2979 of simulations in some cases.
2980 * The new network model will change simulations!
2981 This new model is more realistic than the previous one, so you
2982 should consider redoing your old experiments with this model.
2983 Sorry for the inconvenience.
2986 * Introduce the supernovae compilation mode
2987 When compiled that way, the whole SimGrid (or almost) is put in a
2988 single compilation unit and compiled in one shoot.
2989 This is to help gcc which has difficulties to inline stuff from one
2991 The speedup seem to be above 15%, althrough more tests are needed on
2992 amd64 to confirm that gain.
2995 * Port of MSG's mailbox on top of SIMIX network
2996 The put/get mechanism was greatly simplified on the way.
2999 * New SIMIX network module. Provides:
3000 - Mailbox: rendez-vous mechanism to find with who you want to speak
3001 - Synchronous send/recv: easier and hopefully faster since the
3002 logic is handled in the maestro process directly now
3003 - Asynchronous send/recv: you dreamt of it? It's here now
3004 Too bad that nobody cared enough to propagate the change to MSG.
3005 * Add semaphores as SIMIX synchronization mechanism.
3008 * new function SD_daxload(char*) to load a DAX file
3009 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3010 * Introduce typed tasks. Specify its kind and cost at creation.
3011 At scheduling, just give where it should be placed, and the cost
3012 for each involved resource is automatically computed.
3013 Existing constructors so far (more to come of course):
3014 - SD_task_create_comm_e2e() for end-to-end communication
3015 - SD_task_create_comp_seq() for sequential computation
3016 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3017 * new function SD_task_dump() for debugging display
3018 * new function SD_task_dotty(task,FILE*) writing to file the info
3019 about the task in dotty format
3020 * SD_task_dependency_exists() can now cope with having one of its
3021 arguments NULL. If so, it tests whether the other argument has any
3023 * Add getters on list of preceding/following tasks:
3024 SD_task_get_parents(task) and SD_task_get_children(task)
3025 * Add getters on amount of workstations and list:
3026 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3027 * Add getter on task kind: SD_task_get_kind(task)
3028 * Update the start_time and finish_time of tasks on completion/failure
3029 * Bugfix: Remove task from state swags when destroyed
3032 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3035 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3036 dynar with one separate thread per value of the dynar.
3037 * Change the prototype of xbt_thread_create(), sorry.
3038 Added a boolean parameter indicating whether we want to join this
3039 thread (used in SG only for now)
3040 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3043 * GTNetS wrappers should now be usable again (and betterly tested too)
3044 * Fix a major regression from 3.2 where the timeout provided to
3045 MSG_task_put_with_timeout() was used as absolute time before which
3046 the comm should be done.
3047 * Start to fix the <cluster> tag.
3048 - Internal links should be good now (beside of the loopback, which
3049 use the private link instead)
3050 - paths to the external world is still rather broken
3051 - the <route:multi> tag is just broken. Actually that's brain-dead.
3052 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3053 to make it less stupid
3054 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3055 * Fix a source-level compatibility glitch from 3.2: after defining
3056 MSG_USE_DEPRECATED, you can use the old name
3057 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3058 * Allow to compile from the SVN with automake 1.11
3059 * Fix some problems when using the "start_time" tag in deployment XMLs.
3060 * Fix #8569: XBT/synchro.h has redundant declarations
3061 * Fix #8563: MSG return values and exceptions
3062 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3063 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3064 Thanks to Robson Peixoto.
3065 * Fix a vicious bug in dictionaries inducing that some elements were
3066 not freed on xbt_dict_free()
3068 Portability report of this version:
3069 * Main portability targets:
3070 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3071 on (amd64/i386/ia64)
3072 - mac leopard on i386
3073 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3074 but nothing critical.
3075 * Other platforms: windows, AIX and others were not tested for this release
3077 Timing report of this version:
3078 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3079 scenario parameters). From 8h to a few seconds in desktop grid settings.
3080 * Supernovae brings about 25% speedup on i386.
3082 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3084 SimGrid (3.3.3) stable; urgency=low
3086 The "Need for Speed" release.
3088 The timings done to validate the 3.3.2 were faulty.
3089 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3091 The problem was a conversion from a manually handled vector to
3092 xbt_dynar_t on the critical path.
3093 xbt_dynar_foreach calls functions, inducing stack management crap.
3095 We inlined these functions and xbt_dynar_foreach is now breath taking.
3096 We also inlined xbt_swag_belong on the way.
3098 Here are some approximate speedup measurements (on master/slaves
3099 simulations lasting between 10s and 20s each):
3100 3.3.1 -> 3.3.2: about same performance
3101 3.3.2 -> 3.3.3: 40% speedup
3102 3.3.1 -> 3.3.3: 40% speedup
3103 3.3.1 with inline patch -> 3.3.3: 30% speedup
3105 Our reading is that the refactoring which occurred in 3.3.2 made us
3106 suffer much more from the xbt_dynar_foreach low performance, but
3107 once we solved this, this refactoring proved to be very performance
3108 effective. From the 40% speedup, somehow, 10% are due to the
3109 inlining and 30% to the refactoring.
3111 That's a pitty that gcc cannot inline functions placed in other files
3112 alone. We have to choose between:
3113 - break the encapsulation (by putting private data structures and
3114 accessors in headers files to help gcc)
3115 - live with low performance
3116 - switch to a decent compiler such as icc (not quite possible).
3118 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3120 SimGrid (3.3.2) stable; urgency=low
3122 The "Simplicity does not preceed complexity, but follows it" release.
3124 The main contributors of this release were (lexical order):
3125 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3128 * Extract the routing logic into its own object.
3129 (was dupplicated in network.c and workstation_LV07.c;
3130 Allows to implement other ways of storing that info)
3131 => kill now useless network_card concept
3132 - Use dynar to represent routes (instead of void** + int*)
3133 - kill link_set (use surf_network_model->resource_set instead)
3134 - Add a command-line option to choose the routing schema to use
3135 - Add three new models:
3136 * Floyd (shortest path computed at initialization)
3137 * Dijikstra (shortest path recomputed all the time)
3138 * Cached Dijikstra (shortest path computed on need)
3139 All these models where contributed by Silas De Munck, and are
3140 described in his ICCS09 paper.
3142 * Simplify model declaration
3143 (less redirections, less function to write when defining a model)
3144 - Factorize stuff between models:
3147 surf_model_resource_set(model)
3148 surf_model_resource_by_name(model, name)
3149 - Unify the types of models in s_surf_model_t (using an union)
3150 - Embeed fields of common_public directly into s_surf_model_t
3151 - Rename model methods:
3152 action_free ~> action_unref
3153 action_change_state ~> action_state_set
3154 action_get_state ~> action_state_get
3155 - Change model methods into functions:
3156 (model)->common_public->action_use ~> surf_action_ref
3158 * Implement a generic resource; use it as ancestor to specific ones
3159 (allows to kill duplicated code in models)
3160 Drawback: timer command don't need no name nor properties;
3161 workstation_CLM03 don't need no properties
3162 (but I guess we can live with those few bytes wasted)
3164 * Improve the action object model
3165 - implement a constructor avoiding dupplicated code about field
3166 initialization in generic_action part.
3168 * Kill the SDP model: it has an external dependency, is deprecated
3169 in flavor of modern lmm models, and didn't compile since a while
3172 * Relocation of the context module from XBT to SIMIX.
3173 (the context were decoupled from the simix processes, duplicating a lot of code)
3174 => a lot of code was factorized
3175 - less overhead is introduced during scheduling
3176 - simpler API for the context factory
3177 - the logic for process creation,destruction and manipulation was simplified
3178 * Simplification of the s_smx_process_t data structure.
3179 => accesing the simix level data associated to a process is faster now,
3180 and the code is a lot more readable.
3183 * Implement some more MPI primitives:
3184 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3185 -implementation: Bcast: flat or 2-ary tree (default),
3186 Barrier: 4-ary tree,
3188 Allreduce: Reduce then Bcast
3189 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3190 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3191 Alltoallv: flat tree, like ompi
3193 * Add support for optimized collectives (Bcast is now binomial by default)
3194 * Port smpirun and smpicc to OS X
3197 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3198 Shout out if you used it.
3201 * Display the list of still queued messages in SG mode when existing
3205 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3206 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3207 * Add xbt_str_from_file(FILE*)
3208 * Add xbt_dict_get_key achieving a linear reverse search
3209 * Remove the context module
3211 Portability report of this version:
3212 * Main portability targets:
3213 - Linux(debian)/x86/context
3214 - Linux(debian)/x86/pthreads
3215 - Linux(debian)/amd64/context
3216 - Linux(debian)/amd64/pthreads
3217 On these, we still have the eratic breakages of gras/pmm and
3218 amok/saturate_sg reported in previous version. We still think
3219 that the tests are the cause of the fault, not the tested code.
3221 - Mac OSX Leopard/x86/context
3222 Still false negative in tesh autotesting.
3223 Smpi still fails, but this time because readlink does not accept -f
3224 Everything seems to work properly beside of that.
3227 - AIX version 5.3 (only tested contexts this time)
3228 Smpi still fails there because mktemp is not installed.
3229 Everything seems to work properly beside of that.
3231 I managed to compile it for the first time, but several breakages.
3232 Won't delay the release for this exotic platform.
3234 * Windows: it's still lagging behind. If you want to help, please
3237 Timing report of this version:
3238 This version seem to be more than 5% faster than 3.3.1 (on linux
3239 64bits with contextes). The gain is less than expected, we are
3240 investigating this for next release.
3242 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3244 SimGrid (3.3.1) stable; urgency=low
3247 * Implement a --cfg-help to show existing configuration variables
3248 * Build chain do not require doxygen in maintainer mode
3251 * fix a bug on struct sizeof computation, which prevented the
3252 exchange of arrays of structs in some conditions
3253 - added a regression test about this in datadesc_usage
3254 * Allow the exchange of 0-long dynamic vectors.
3255 - for that, use -1 as indicator of dynamic size instead of 0
3256 - This implied to change any size from unsigned long to long,
3257 reducing a bit communication abilities, but I guess that with
3258 64bits being quite common, this is more than enough.
3259 - This also induce a protocol change, thus bumping network protocol
3260 version from 0 to 1 (if we have external users, we have to get
3261 clean on that point too ;)
3262 - added two regression tests about this in datadesc_usage
3263 * Be more verbose when propagating local exceptions
3264 This helps debugging.
3265 * Display the status of simulated processes when receiving SIGINT in
3269 * Allow to control the simulation from a trace file.
3270 New functions MSG_action_register() and MSG_action_trace_run()
3271 The first one allows to associate a function execution to each
3272 kind of action while the second one parses a trace file and
3273 triggers the corresponding actions within the system.
3274 For now, only a toy example is provided in examples/msg/actions
3275 * Add an example of process migration in examples/msg/migration
3276 * Fix a bug in task exchange which broke MSG_task_get_sender()
3277 Add a teshsuite regression test for that.
3278 [Bug: if MSG_task_get_sender() is called after sender exit,
3280 * Fix a bug which prevented suspend/resume to work properly
3281 * Display the status of simulated processes when receiving SIGINT
3282 This fixes a regression of v3.3. due to the introduction of SIMIX
3283 * Bug fixing in failure management:
3284 - trace could not start by a failure at time 0
3285 - failure during communications were not working
3288 * Add SIMIX_process_set_name() to change the name of the current
3289 process in the log messages.
3290 * Store smx_hosts in a dict since we only retrieve them by name
3291 * Move the configuration infrastructure to surf
3294 * Move the configuration infrastructure to surf
3297 * Massive internal cleanups:
3298 - Store internal structures on processes instead of hosts (allows
3299 to have more than one process per host, in addition of being more
3301 - Cleanup the initialization/finalization process
3302 - Kill a whole bunch of unneeded synchronization:
3303 processes run in exclusive manner within the simulator
3304 - Move queues from global tables to process data fields
3306 - now accept -platform and -hostfile arguments
3307 - Pass the right rank value to processes according to the hostfile
3308 * Compile the examples by default, and use them as regression tests
3309 * Implement MPI_Wtime()
3310 * Change the reference speed to a command line option
3313 * TCP_gamma can now be specified as command line option using
3314 --cfg=TCP_gamma:10000000.0
3315 * Change the --surf-path cmd line option into --cfg=path:
3318 * Also include strbuff from xbt.h public header
3319 * xbt_ex_display(): do not free the exception after displaying
3320 This allows to do more with the given exception afterward.
3321 Users should call xbt_ex_free() themselves.
3325 Portability report of this version:
3326 * Main portability targets:
3327 - Linux(debian)/x86/context
3328 - Linux(debian)/x86/pthreads
3329 - Linux(debian)/amd64/context
3330 - Linux(debian)/amd64/pthreads
3331 These targets fail about 1/10 of times on gras/pmm, but we believe
3332 that this is because of the test, not because of simgrid.
3333 amok/saturate_sg fails even more rarely, and the test may not be
3336 - Mac OSX Leopard/x86/context
3337 The test suite still spits tons of errors because some obscure
3338 force prevents us from removing the temporary directories
3339 arguing that they still contain some metadata I've never heard of.
3340 Smpi fails because seq is not installed.
3341 Everything seems to work properly beside of that.
3344 - AIX version 5.3 (both contexts and pthread)
3345 Smpi still fails there because mktemp is not installed.
3346 XML inclusions seems rosty on AIX.
3348 * Windows: it's still lagging behind. If you want to help, please
3351 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3353 SimGrid (3.3) stable; urgency=high
3357 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3358 [Malek Cherier & Mt]
3360 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3363 Until now, GRAS were using MSG as an interface to SURF. It was
3364 quite difficult because both interface have several differences
3365 (MSG channels vs GRAS sockets were the most notable point).
3367 This also opens the gate to SMPI (which should occur soon) and speed
3368 up simulations by to 40% (even if it were not the main goal).
3370 **************************************
3371 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3372 **************************************
3373 The point is that events occurring at the exact same timestamp are
3374 not scheduled in the same order with the old and new version. This
3375 may be enough to completely change the execution of simulations in
3376 some cases. Sorry for the inconvenience.
3378 * Cleanup and upgrade the XML format to push further scalability
3379 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3381 * Improve the testing infrastructure with tesh. Now a very large part of
3382 the code is tested not only by being run but also by checking that the
3383 output match an expected output [Mt].
3385 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3386 is really less memory-demanding, which should allow you to use
3387 larger files in SimGrid [AL].
3389 * Inform valgrind about our contextes, so that it becomes usable
3390 with the default (and more effecient) version of SimGrid
3391 [contributed by Sékou Diakite, many thanks]
3394 * Introduce a listener thread in charge of receiving incoming
3395 messages from the network. It allows to overlap communication and
3396 computation but most notably, it removes some stupid deadlocks due
3397 to the fact that so far, a process could not send and receive at
3398 the same time. This made most non trivial communication schema
3400 * Convert the PIDs from long int to int to match the MSG ones (and
3401 linux ones too) [Mt]
3402 * New function: gras_agent_spawn() to launch a new process on
3403 current host. Only working in simulation for now. [Mt]
3404 * New function: gras_os_hostport() returning a constant form (ie,
3405 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3408 * Make the backtrace of exceptions more human readable [Mt]
3409 * New module: xbt/str [Mt]
3410 a ton of string utility functions (split, join, printf to a newly
3411 allocated buffer, trim, etc)
3412 * New module: xbt/hash [Mt]
3413 SHA1 hashing algorithm (more to come if needed)
3414 * New module: xbt/synchro [Mt]
3415 synchronization tools (mutex and conditions) working the same way
3416 in simulation and in real life (mainly useful for GRAS, but not
3418 * New module: xbt/queue [Mt]
3419 classical producer/consumer synchronization scheme
3420 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3421 (using the classical functions will get serialized) [Mt]
3422 * Make dictionary internal table dynamic. No need to specify its size
3423 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3425 * Make sure the log channels are organized as a tree under windows
3426 (because of ANSI C compatibility issue, any channel were child of
3427 root directly) [Mt].
3430 * Cleaned many thing in surf and fixed a few bugs [AL].
3431 * Add a nice command line configuration mechanism to compose models [AL].
3432 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3433 the previous one (KCCFLN05). It relies on something that looks like
3434 a max-min sharing mechanism but cannot be written as such. A new solver
3435 was thus designed [AL].
3436 * Add a new solver to lmm. Based on Lagrange optimization and
3437 gradient-based descent, it enables to efficiently maximise systems s.a
3439 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3441 This solver enables to propose two new network models for TCP Reno and
3442 TCP Vegas based on Low's work. These models still need to be fully
3443 tested though [Pedro Velho].
3446 * Bug fix in SD_simulate. Now the time bound given as argument is
3448 * Use the new parallel task model (ptask_L07) as default.
3449 * Use the SURF command line configuration mechanism.
3450 * 0-size tasks (for synchronization) should now work.
3452 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3454 SimGrid (3.2) stable; urgency=high
3458 We still experience issues on this platform, but we believe that at
3459 least MSG is usable.
3461 GRAS API BREAKAGE (for simplification purpose, sorry):
3462 * the gras_msgtype_by_name is not used anymore. Instead of
3463 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3464 you can write (and must)
3465 gras_msg_send(toserver, "request", &request);
3466 - If you still want to pass a gras_msgtype_t to the function (to cache
3467 the type and avoid the lookup time), use the gras_msg_send_() variant.
3468 - Impacted functions:
3469 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3470 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3471 * The callbacks are now expected to return 0 when everything went well
3472 (just like the main() function)
3474 GRAS new features and improvements:
3475 * New module mechanism where user code can use per process globals [Mt]
3476 This is similar to gras_userdata_*() functions, but for libraries. It
3477 factorize some code developped over and over in the examples and AMOK.
3478 It has still to be documented and used (only amok/peermanagement is
3480 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3481 loss when they were small enough to fit into the buffer and sent quickly
3482 enough so that they can all get received in one shoot.
3483 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3484 if not found. Use the *_or_null() variant for the old semantic.
3485 * In gras_msg_handle, do not discard messages without callback.
3486 They are probably messages to be explicitly awaited later (ie, proofs of
3487 mis-synchronization in userland since they are sent before being awaited)
3488 No big deal usually.
3489 * gras_socket_meas_send/recv: semantic changed!
3490 The numerical arguments used to be (1) the total amount of data to send
3491 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3492 messages. This was need for the fool willing to send more than MAXINT
3493 bytes on quite fat pipes.
3496 * Do really rename the hostmanagement module to peermanagement. [Mt]
3497 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3498 but this is rather new and this was documented in the module
3499 documentation (poor excuses, I admit)
3500 * Bandwidth measurement semantic changed! This follows the changes to
3501 gras_socket_meas_send/recv explained above.
3504 * A sequential mode has been added to the workstations. When a workstation
3505 is in sequential mode, it can execute only one task, and the other tasks
3506 are waiting in a FIFO. [Christophe Thiery]
3509 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3510 model for SIMDAG. [Christophe Thiery]
3511 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3512 equal to 0, which caused some bad behaviors in
3513 saturated_constraint_set_update. I now use a threshold mechanism like in
3517 * When running manually src/testall, you select specific units [Mt]
3518 testall is the result of our cunit mechanism, and should replace all
3519 the scripty thingy around since bash don't run easily on billware.
3521 * A mallocator system has been added. [Christophe Thiery]
3522 Mallocators allow you to recycle your unused objects instead of freeing them
3523 and allocating new ones.
3525 Documentation update:
3526 * FAQ reworking + New FAQs:
3527 - "Valgrind spits tons of errors!" [Mt]
3528 - "How to repport bugs" [Mt]
3529 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3530 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3531 - Communication time measurement within MSG [AL]
3532 - I experience weird communication times when I change the latency [AL]
3533 * GRAS tutorial [Mt]
3535 - an introduction to the framework and to the used communication model
3536 - an initiatic tour introducing the most proheminent features:
3538 . Lesson 0: Installing GRAS
3539 . Lesson 1: Setting up your own project
3540 o Part 2: Message passing
3541 . Lesson 2: Exchanging simple messages
3542 . Lesson 3: Passing arguments to the processes (in SG)
3543 . Lesson 4: Attaching callbacks to messages
3544 . Lesson 5: Using globals in processes
3545 . Lesson 6: Logging information properly
3546 . Lesson 7: Using internal timers
3547 . Lesson 8: Handling errors through exceptions
3548 . Lesson 9: Exchanging simple data
3549 . Lesson 10: Remote Procedure Calling (RPC)
3550 . Lesson 11: Explicitly waiting for messages
3551 . Recapping of message passing features in GRAS
3552 - A HOWTO section containing:
3553 o HOWTO design a GRAS application
3554 More are due, of course. They will come latter. In the meanwhile, you can
3555 check the examples which are still here.
3557 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3559 SimGrid (3.1) stable; urgency=high
3563 There was a stack corruption somewhere, visible only when optimizing
3564 with these versions. [Vince]
3567 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3568 module that enabled to play with Directed Acyclic Graphs. It is built
3569 directly on top of SURF and provides an API rather close to the old
3570 SG. Some old codes using SG are currently under rewrite to check that
3571 all needful functions are provided. [Christophe Thiery]
3574 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3575 extension of the classical CLM03 model that gracefully handles
3576 failures. This is now the default model for MSG and GRAS. It doesn't
3577 handle parallel tasks yet though. [AL]
3578 * Bug fix: Weights were not correctly set in the network part.
3579 WARNING: This may have resulted in incorrect results with simulations
3580 where there are more than one flow on a given link. [AL]
3583 * After a (long ?) discussion on simgrid-devel, we have decided that the
3584 convention we had on units was stupid. That is why it has been decided
3585 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3586 WARNING: This means that all previous platform files will not work as
3587 such with this version! A warning is issued to ask users to update
3589 A conversion script can be found in the contrib module of the CVS, under
3590 the name contrib/platform_generation/surfxml_update.pl [MQ]
3593 * Bug fix: Processes were started in reverse order, wrt deployment file.
3594 WARNING: if your code relies on this bug, please fix it. [AL]
3595 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3596 being executed on two different locations. [AL]
3597 * Bug fix: Failures are now better supported thanks to Derrick's tests
3598 (there was many failure situations I hadn't thought of and that weren't
3599 correctly handled). [AL]
3600 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3604 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3605 Exception occurring on server-side are propagated back to client (!).
3607 API CHANGE: the callback changed their prototype. Change:
3608 int my_handler(gras_socket_t expeditor, void *payload_data) {
3610 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3611 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3613 * New! function: gras_msg_handleall to deal with all messages arriving
3614 within a given period.
3615 * New! function: gras_socket_server_range to get a server socket in a
3616 range of port numbers (ease to avoid port number conflicts) [MQ]
3617 * New! gras processes display their backtrace when they get a SIGUSR1
3618 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3619 Sweet to debug RL processes [MQ]
3623 - Do not force experiment sizes to be expressed in kb, or it becomes
3624 impossible to measure the latency this way (needs one byte-long tests)
3625 WARNING: this changes the amok_bw_* function semantic. [MQ]
3626 - Implements the link saturation stuff. [MQ]
3627 * Peer management module:
3628 New! module factorizing code that we wrote over and over [MQ].
3631 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3632 - Test units are placed directly into the library code, they get extracted
3633 automatically and placed into the src/testall binary.
3634 - Convert most of the XBT tests to this system.
3635 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3636 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3637 live in the same C code. This required to change a little bit the API
3638 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3639 * New module: sparse graph structure with basic algorithms (this is work
3640 in progress and the API is not considered to be frozen yet). [AL]
3641 * Display more information on backtraces: source line & function names are
3642 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3643 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3644 * Reworked a little bit some #include statements to load only required
3645 headers. Some user code that relied on SimGrid to include stdlib or
3646 stdio may need to include it by themselves. [AL]
3647 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3649 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3650 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3651 but I guess that almost nobody used those parts. [MQ]
3653 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3655 SimGrid (3.0.1) stable; urgency=low
3658 * Unfortunately, I had missed 5 misnamed functions:
3659 xbt_fifo_item_t xbt_fifo_newitem(void);
3660 void xbt_fifo_freeitem(xbt_fifo_item_t);
3661 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3662 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3663 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3664 They're now deprecated. Please use their new versions:
3665 xbt_fifo_item_t xbt_fifo_new_item(void);
3666 void xbt_fifo_free_item(xbt_fifo_item_t);
3667 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3668 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3669 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3671 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3672 * Documentation: xbt_log module unmercifully reworked [MQ]
3673 * Bugfix: there was a problem with the ending of contexts with
3674 the pthread backend. It caused some weird deadlock or behavior
3675 depending on the pthread implementation. [AL]
3676 * Bugfix: get the exceptions raised in the simulator repport where
3677 and why they come from when they are not caught in time [AL, MQ]
3680 * Bugfix: Do repport the error when two non-connected hosts try to
3681 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3684 * Add additionnal checkings on communications. Assert that two
3685 communicating hosts are connected by a set of links... [AL]
3688 * Add additionnal checkings on channel values in communication [AL]
3689 * New: MSG_task_get_source to see on which host a task was generated [HC]
3690 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3691 the number of tasks waiting to be received on channel and sent
3693 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3694 waits for the first task coming from a given host.. [AL]
3696 GRAS new functionnalities: [MQ]
3697 * Enhance the parsing macro to allow the size of multidimentional objects
3698 to be given thru annotations.
3699 * New example (and documentation): Matrix Multiplication a la RPC
3700 (as when I was young!) and fix a bunch of bugs found on the way.
3702 GRAS performance improvements: [MQ]
3704 * Reduce the amount of cbps creation/destruction by making it static to
3705 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3708 * Change libdata to a set so that we can search for stuff by ID (and thus
3709 reduce the insane amount of dict lookups)
3712 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3713 RL communication on top of SG since it's so uneffective.
3714 It may also be used for inter-thread communication in RL, one day.
3715 * Use gras_datadesc_copy() to exchange messages on top of SG
3717 - improve message exchange performance on top of SG
3718 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3719 * Don't exchange on the network the size of the used part of buffer,
3720 instead, specify the possible buffer size to read().
3722 - reduces the amount of read/write calls (one pair per exchange)
3723 - reduces the amount of exchanged data (the size)
3724 - allows to retrieve all arrived data on receiver side, if we don't need
3725 it right now (subsequent read will peek the buffer)
3726 - allows the receiver to proceed with the beginning of the stream before
3727 everything is arrived
3728 - make it possible to build an iov transport (using readv/writev)
3730 - take care of the data with non-stable storage (like stacked data),
3732 * If possible, TCP send uses vector I/O (when writev() is here)
3733 - Don't use it for receive since we send data sizes and data on the
3734 same stream, so we wouldn't be able to chain large amount of chunks
3735 before having to flush the stuff to read the size.
3736 * Rework the transport plugin mechanism to simplify it and reduce the
3737 amount of pointer dereferencement when searching for the right function
3740 * I guess that now, we do almost as few system calls as possible while
3741 doing as few data copy as possible.
3743 To improve it further, we could try to send all the sizes first and then
3744 all the data (to use iov on receiving size), but it's only a partial
3745 solution: when you have 2 dimensional data, the sizes of the second
3746 dimension is data of the first dimension, so you need 3 streams.
3748 I'm not sure the potential performance gains justify the coding burden.
3750 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3752 SimGrid (3.00) stable; urgency=high
3755 * New! Give the possibility to hijack the surf parser and thus bypass
3756 MSG_create_environment and MSG_launch_application. Have a look at
3757 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3759 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3761 SimGrid (2.96) unstable; urgency=low
3766 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3767 This deprecates the xbt_error_t mechanisms.
3768 It modifies (simplifies) all XBT and GRAS API.
3769 MSG API keeps unchanged (exceptions raised by XBT are caught from
3770 within MSG and masked with existing error handling facilities)
3773 * New! Add a FATPIPE model. [AL]
3774 * New! Add a parallel task model. [AL]
3775 * New! Add automatically a loopback interface (in the default
3776 network model) if none was precised.
3779 * Bugfix: MSG_process_resume now works with the current running process.
3781 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3782 * Modification of MSG_task_get_compute_duration. Once a task has been
3783 processed, the value returned by this function is now equal to 0. [AL]
3784 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3785 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3786 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3787 of a m_task. For now, MSG_task_cancel only works with computation
3789 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3790 of the processor (in Mflop/s) regardless of the current load on the
3792 * API Change: use proper naming convention for MSG_getClock and
3793 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3795 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3796 This function changes the priority of a computation task. This priority
3797 doesn't affect the transfer rate. A priority of 2 will make a task
3798 receive two times more cpu power than the other ones. This function
3799 has been added to suit the needs of Nguyen The Loc and hasn't been that
3800 much tested yet. So if it fails, please report it and send me your code.
3802 * API Change: removed all functions and types that were marked "deprecated"
3803 since many months. Renamed MSG_global_init_args to MSG_global_init.
3805 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3807 SimGrid (2.95) unstable; urgency=low
3810 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3811 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3812 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3815 * Bugfix: Allow absolute paths for platform description files [MQ]
3816 * Bugfix: do free the variables after use. Leads to drastic performance
3818 * Implement max_duration (ie, timeouts) on resources [AL]
3821 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3823 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3824 the needed support (timeouts) [AL]
3827 * Implement measurement sockets. You can now get the bandwidth between two
3828 hosts thanks to AMOK (see below). [MQ]
3829 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3830 dynar over the network (yeah) [MQ]
3831 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3832 * simple/static token-ring example. [Alexandre Colucci and MQ]
3833 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3834 in place before [MQ]
3837 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3838 type parsing. "short" and "long" modifiers were also ignored; other
3839 modifier (such as reference level) are still ignored. [MQ]
3840 * Embeed the buffer size within the buffer itself on SG. [MQ]
3841 That way, send() are atomic and cannot get intermixed anymore (at least
3842 the ones which are less than 100k; bigger messages still have the issue)
3843 * Array size pushed by the field, not by the field type (or each
3844 and every long int will push stuff to the cbps) [MQ]
3845 * use select() to sleep since it allows to portably sleep less than one
3848 GRAS (minor cleanups)
3849 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3852 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3853 * type category 'ignored' killed as it was never used and were difficult
3855 * whether a type can cycle or not is now a flag, leaving room for more
3856 flags (as "ignored", if we feel the need one day ;)
3857 * Rename raw sockets to measurement sockets since "raw" has another
3858 meaning in networking community.
3861 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3862 and offers features not belonging to GRAS but that most applications
3863 need. One day, it may be a set of plugins loadable at runtime.
3864 * New module: bandwidth
3865 bandwidth measurement between arbitrary nodes running this module. [MQ]
3867 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3869 SimGrid (2.94) unstable; urgency=low
3871 The first beta release of SimGrid 3 !
3875 * Update the main page and the FAQ. Adding references to gforge.
3878 * Add a gras_os_getpid function.
3881 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3882 * Extend the logs so that they also print PID, hostname, date, ... if
3884 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3885 and kill the old version which were in testsuite/
3886 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3890 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3891 for users. Instead, each programming environment (GRAS, MSG,...) use it
3892 when needed to provide such a feature to users.
3893 Don't shortcut the mechanism or you will also shortcut the virtualization
3894 you need on the simulator.
3898 * Cleanups in configury with regard to compile optimization/warning flags.
3899 Also add -fno-loop-optimize to any powerpc since it's the optimization
3900 killing gcc (< 3.4.0).
3901 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3902 needed by MSG examples complications
3903 * Borrow configury beautifications from PHP
3906 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3907 warning (thanks loris for stumbling into this one).
3908 * Bugfix: stop loading private headers (gras_config.h) from the public
3912 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3913 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3914 RL would imply a DNS resolver, which is *hard* to do in a portable way
3915 (and therefore delayed).
3916 * Implement a real timer mechanism and use it in timing macros. This allows
3917 to avoid rounding errors and get a 0.000005 sec precision in timing
3918 macros. While I was at it, various cleanups:
3919 - allow to declare more than one timed section per file (fix a stupid bug)
3920 - move some private declaration to the right place
3921 - merge conditional execution and timing macros into emulation module
3922 - document the module
3923 - make sure the module cleanups its mess on gras_exit
3924 * Documentation improvements:
3925 - (new) how to compile applications using GRAS
3926 - (new) emulation support (timing macros)
3928 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3930 SimGrid (2.93) unstable; urgency=low
3932 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3935 - Use Paje properly where used. Still to be sanitized properly.
3936 - Portability fix: Add an implementation of the contexts using pthread
3940 - Add xbt_procname(): returns the name of the current process.
3941 Use it to show the current process's name in all logging.
3943 - fix detection of older flex version and the reaction, since we do
3944 depend on modern ones (we use lex_destroy)
3945 - Better separation of SG and RL in the libs: remove all simulation code
3946 from libgras. As a result, this lib is now only 200k when stripped.
3947 Some of the xbt modules may also be duplicated (two sets and such) and
3948 should be cleaned/killed before SG3.
3949 - Insist on using xlC on AIX because of weird problems involving gcc there.
3950 - Cleanup the make remote stuff. This is now done by scripts
3951 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3952 mainly for our private use, but we're working on changing them to user
3955 - Bugfix: flush the socket on close only if there is some *output*.
3956 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3958 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3961 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3963 SimGrid (2.92) unstable; urgency=low
3965 Alpha 3 on the path to SimGrid 3
3969 - New! First try of benchmarking macros.
3970 - New! First try so that gras_stub_generator generate deployment and
3971 remote compilation helpers.
3973 - Bugfix: Initialization fix in msg_test.
3977 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3979 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3981 - API Change: gras_os_sleep to take the amount of seconds as a double.
3982 Accepting an int was error prone since it was the only location where
3983 seconds were coded as such. It leaded to damn rounding errors.
3984 - Bugfix: Hard to belive that timers ever worked before this.
3986 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3988 SimGrid (2.91) unstable; urgency=low
3990 Alpha 2 on the path to SimGrid 3
3994 - Bug fix in the lmm_solver.
3996 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3997 through the function MSG_paje_output.
3998 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3999 - It is possible to bound the rate of a communication in MSG with
4000 MSG_task_put_bounded() (was already in the previous version but I had forgotten
4001 to write it in the changelog).
4002 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4007 - Various cleanups to the autotools stuff
4008 - Begin to move Gras examples to examples/gras/
4009 - Let make distcheck work again (yeah!)
4011 - documentation overhauled using doxygen.
4012 gtk-doc-tools is dead in SimGrid now.
4013 - Automatically extract all existing logging categories, and add the list
4014 to the documentation (long standing one, to say the less)
4016 - Cleanup the known architecture table. Reorder the entries to group what
4017 should be, and use a more consistent naming scheme.
4018 (some of the test dataset are still to be regenerated)
4019 - New! Allow library to register globals on each process just as userdata
4021 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4022 lookup time (for now).
4023 Use it in msg and trp.
4024 This cleans a lot the internals and helps enforcing privacy of the
4025 headers between the gras components.
4026 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4027 - Bugfix: gras_os_time was delirious in RL.
4028 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4030 - Reenable GRAS now that it works.
4032 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4034 SimGrid (2.90) unstable; urgency=low
4036 Alpha 1 on the path to SimGrid 3
4038 * It is a long time since the last release of SimGrid. I'm sorry about
4039 that but as I had told you, I was rewriting a lot of things. I apologize
4040 to those who had been reporting bugs to me and that I had not answered.
4041 If your bug is still in the new version, please tell me. Here is a
4042 summary of the main changes.
4044 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4045 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4046 portability, speed, and a lot more but you'll figure it out later.
4047 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4048 I wanted to release the new MSG as soon as possible and I have
4049 broken GRAS, which is the reason why, for now, only MSG is fully
4050 functional. A laconic description of these projects is available
4051 in the documentation.
4053 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4054 kernel optimized for our needs (called SURF but only the developers
4055 should use it). Hence, MSG is now roughly 30 times faster and I think
4056 that by rewriting a little bit MSG, I could event speed it up a little
4057 bit more. Beside the gain in speed, it is also much easier to encode a
4058 new platform model with SURF than it was with SG. More to come...
4060 * REVOLUTION 3: I have tried to change a little as possible the API of
4061 MSG but a few things really had to disappear. The main differences
4062 with the previous version are:
4063 1) no more m_links_t and the corresponding functions. Platforms are
4064 directly read from an XML description and cannot be hard-coded
4065 anymore. The same format is used for application deployment
4066 description. The new format is described in the documentation.
4067 Have a look in tools/platform_generation. There is a tiny script
4068 that converts from the old platform format to the new one. Concerning
4069 the application deployment format, parsing the old one is tricky.
4070 I think most of you should however be able to convert your files. If
4071 it is really an issue, I can write a C code that does the conversion.
4073 2) the toolbox tbx does not exist anymore. We now have a library
4074 with much more data-structures but without the hash-tables (we have
4075 dictionaries that are much faster).
4077 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4079 *****************************************************************************
4080 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4081 * don't feel like converting the oldies. (Mt) *
4082 *****************************************************************************
4085 Version 2.90: "the long awaited one"
4086 - Finished rewriting and debugging MSG. Rewrote the documentation.
4087 - disable GRAS for now since it needs to be ported to the newest SG
4090 - Finish the port to windows (using mingw32 for cross-compile)
4093 - Main loop and datastructures of SURF. A cpu resource object is
4094 functional. Surf can thus be used to create cpu's with variable
4095 performance on which you can execute some actions.
4097 2004-11-15 Martin Quinson
4098 - Port to ARM. Simply added the alignment and size descriptions. Should
4099 work, but the ARM machines are so slow that I didn't had the opportunity
4100 to 'make check' over there yet.
4102 2004-11-15 Arnaud Legrand
4103 - Trace manager now written. It uses a heap structure and is therefore
4104 expected to be efficient. It may however be speeded up (particularly
4105 when many events occur at the same date) by using red and black
4106 trees. One day maybe...
4107 - Max-min linear system solver written. It uses a sparse matrix
4108 structure taking advantage of its expected use. Most operations are
4109 O(1) and free/calloc are called as few as possible. The computation of
4110 the minimum could however be improved by using a red and black tree
4113 2004-11-03 Arnaud Legrand
4114 - Rename every gras_* function that was in xbt/ to its xbt_
4116 - Add a heap and a doubly-linked list to xbt
4117 - Added a dichotomy to the dictionaries. make check works as well before
4118 so I assume that the patch is correct. I do not know however if things
4119 run effectively faster than before now. :)
4121 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4122 SimGrid, and the version number is bumped to 2.x
4124 2004-10-29 Martin Quinson
4125 - Introduction of the remote errors.
4126 They are the result of a RMI/RPC on the remote machine.
4127 ErrCodes being scalar values, you can't get the host on which those
4128 errors did happen. Extending the error mechanism as in Gnome is possible.
4129 No idea yet whether it is a good idea.
4131 2004-10-28 Martin Quinson
4132 - Interface revolution: the Starred Structure Eradication.
4133 I used to do typedef struct {} toto_t; and then handle *toto_t.
4134 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4135 - ???_t is a valid type (builded with typedef)
4136 - s_toto_t is a structure (access to fields with .)
4137 - s_toto is a structure needing 'struct' keyword to be used
4138 - e_toto_t is an enum
4139 - toto_t is an 'object' (struct*)
4141 typedef struct s_toto {} s_toto_t, *toto_t;
4142 typedef enum {} e_toto_t;
4143 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4144 s_toto_t) is private.
4146 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4147 so that it can set the variable to NULL. It was so for dicts and sets,
4148 it changed for dynars.
4150 - Fix a bunch of memleaks in dict_remove
4151 - Fix a bug in sg/server_socket opening: it failed all the time.
4153 2004-10-07 Martin Quinson
4154 - Speed up dynar lookup operation a bit.
4156 gras_dynar_get is dead.
4158 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4159 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4160 (which gives you the address of the stored data).
4162 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4163 copy of the data using an affectation to do the job and not a memcpy.
4165 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4166 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4168 It does not really speedup the dynar test because they are
4169 setting elements all the time (and look them seldom). But the dict does
4170 far more lookup than setting.
4172 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4174 2004-10-05 Martin Quinson
4175 - Allow to (en/dis)able the cycle detection at run time.
4177 Whether we should check for cycle or not is now a property of each
4178 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4179 datadesc_cycle_unset allow to remove this property when previously set.
4181 Note that the cycle detection is off by default since it impacts the
4182 performance. Watch the data you feed GRAS with ;)
4184 This property is hereditary. Any element embedded in a structure having it
4185 set have it set for the time of this data exchange.
4187 You should set it both on sender and receiver side. If you don't set it on
4188 sender side, it will enter an endless loop. If you forget on receiver
4189 side, the cycles won't be recreated after communication.
4191 - Header reorganization.
4192 Kill gras_private.h, each submodule must load the headers it needs.
4194 2004-10-04 Martin Quinson
4195 - Interface revolution: do not try to survive to malloc failure.
4197 Now, gras_malloc and friends call gras_abort() on failure.
4198 As a conclusion, malloc_error is not a valid error anymore, and all
4199 functions for which it was the only gras_error_t return value are
4200 changed. They now return void, or there result directly.
4201 This simplify the API a lot.
4203 2004-09-29 Martin Quinson
4204 - Re-enable raw sockets.
4205 Created by gras_socket_{client,server}_ext;
4206 Used with gras_raw_{send,recv}
4209 It should allow to kill the last bits of gras first version soon.
4211 This is not completely satisfactory yet (duplicate code with
4212 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4215 - Simplify transport plugin (internal) interface by not passing any
4216 argument to _server and _client, but embedding them in the socket
4219 2004-09-28 Martin Quinson
4220 - Finish the port to AIX.
4221 autoconf was my problem (segfault within the malloc replacement
4222 function. No idea why)
4224 2004-09-16 Martin Quinson
4225 - Fix some size_t madness on 64bit architectures.
4227 2004-09-08 Martin Quinson
4228 - Reduce the number of system headers loaded, overload some more system
4229 calls (such as malloc to cast the result of the system one, and work
4231 - Fix and reintroduce the config support
4233 2004-09-07 Martin Quinson
4234 - Source code reorganization to allow Arnaud to surf all over there.
4235 - Allow to document the logging categories.
4236 - Remove all uppercase from logging categories and useless cleanup in names.
4238 2004-08-18 Martin Quinson
4239 Version 0.6.2 (protocol not changed; API changed)
4240 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4241 gras_error_t), and NULL when not found. Functions expecting a msgtype
4242 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4243 hopefully usefull message.
4244 - Portability to prehistoric sparcs again
4246 2004-08-17 Martin Quinson
4247 Version 0.6.1 (protocol not changed; ABI not changed)
4248 - prealloc some buffers to speed things up
4250 2004-08-11 Martin Quinson
4251 Version 0.6 (protocol not changed; ABI expended)
4252 - The parsing macro can deal with the references, provided that you add
4253 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4255 2004-08-09 Martin Quinson
4256 Version 0.5 (protocol not changed; ABI changed)
4257 - Allow to off turn the cycle detection code in data exchange at
4258 compilation time. It should be at run time, but I'm short of time (and
4259 the config stuff is still broken). That way, we keep dict out of the
4260 critical path, which is good because the performance is poor:
4261 - search not dichotomial yet
4262 - dynar give no way to access their content and memcpy everytime
4263 - In composed data description (struct, ref and so on), stop foolness of
4264 keeping the subtype's ID, but store the type itself. This keeps sets out
4265 of the critical path, which is good since they rely on dynar and
4266 dictionaries. The only loose of that is that we cannot detect the
4267 redeclaration of a structure/union with another content (but I'm not sure
4268 the code detected well this error before anyway). We still can detect
4269 the redefinition discrepancy for the other types.
4270 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4271 it breaks the code because of type-punning used all over the place).
4272 This breaks on all non-gcc architectures (for now).
4274 All those changes (plus the buffer of last time) allow me to gain 2 order
4275 of magnitude on cruel tests consisting of 800000 array of integers on two
4276 level of a hierarchical structure (200 secondes -> 4 secondes)
4279 - the selector of reference must now return the type it points to, not
4280 the ID of this type.
4282 2004-08-06 Martin Quinson
4283 Version 0.4 (protocol changed; ABI not changed)
4284 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4285 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4286 - Implement a buffer transport, and use it by default (it relies on tcp in
4287 real life and on sg in simulation).
4288 That's a bit hackish since I had a new field to the structure to store
4289 its data without interfering with the subtype ones. Inheritance
4290 is tricky in C. And that's a kind of reverse inheritance with one class
4291 derivating two classes. Or maybe a game with java interfaces. Anyway,
4292 that's damn hard in C (at least).
4293 Moreover, I got tired while trying to ensure plugin separation and
4294 genericity in SG mode. MSG wants me to do weird things, so let's go for
4295 cruel hacks (temporarily of course ;).
4296 See comment in transport_private.h:71
4297 - do not include all the _interface headers in private but in the files
4298 which really need them (to cut the compilation time when they are
4301 2004-07-26 Martin Quinson
4302 Version 0.3 (protocol not changed; ABI changed)
4303 - Major overhault of the datadesc interface to simplify it:
4304 - shorted the function names:
4305 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4306 - add a trivial way to push/pop integers into the cbps without malloc.
4307 This allows to make really generic sub_type description, which simply
4308 pop their size of the stack.
4309 - add a function gras_datadesc_ref_pop_arr() which does what users want
4310 most of the time: Declare a dynamic array (which pops its size of the
4311 stack) and declare a reference to it. Poor name, but anyway.
4312 - kill the post-send callback, add a post-receive one
4314 2004-07-23 Martin Quinson
4315 Version 0.2 (protocol changed; ABI changed)
4316 - add some testing for cpbs in the test cases, and fix some more bugs.
4317 This invalidate again the little64 data file, since I cannot regenerate
4319 - remove an awfull optimization in the logging stuff, allowing me to:
4320 - understand it again
4321 - learn gcc how to check that the argument match the provided format
4322 - fix all errors revealed by gcc after that
4323 - internal keys of dict are not \0 terminated. Deal with it properly in
4324 loggings instead of segfaulting when the user want to see the logs :-/
4326 2004-07-22 Martin Quinson
4327 - Fix some stupid bug preventing cbps (callback postit) from working
4329 2004-07-21 Martin Quinson
4330 - Some documentation cleanups
4331 - remove the useless last argument of msgtype_declare
4332 - rename the Virtu functions to fit into the 'os' namespace
4333 - move headers src/include -> src/include/gras/ and stop fooling with
4334 gras -> . symbolic link
4335 - make distcheck is now successful
4337 2004-07-19 Martin Quinson
4339 - Build shared library also
4340 - Install html doc to the right location
4341 - stop removing maintainer files in make clean
4342 - build tests only on make check
4344 2004-07-13 Martin Quinson
4346 - No major issue in previous version => change versioning schema
4347 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4348 corresponding dataset.
4350 2004-07-11 Martin Quinson
4352 - Get it working with any kind of structure (we can compute the padding
4353 bytes remotely for all the architectures I have access to)
4354 - Implement the structure parsing macro (still not quite robust/complete)
4355 - Improvement to the remote testing toysuite
4357 2004-07-10 Martin Quinson
4358 [autoconf mechanism]
4359 - get ride of a bunch of deprecated macros
4360 - actually run the test for two-compliment, not only compile it :-/
4361 - test whether the structures get packed (and bail out if yes. Damn.
4362 Alignment is a serious matter)
4363 - test whether the structures get compacted (but respecting the alignment
4364 constraints of each types)
4365 - test whether the array fields of structures can straddle alignment boundaries
4367 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4368 (took me 2 hours to find that bug, looking at the wrong place)
4369 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4370 before sending/receiving and are used to compute the offsets of fields
4371 - Given that padding size depend even on compiler options, keep track of
4372 alignment and aligned_size only for the current architecture. Not a big
4373 deal since we send structure fields one after the other (seems
4375 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4376 program, let it work on linux/gcc/little32. portability todo.
4378 2004-07-08 Martin Quinson
4379 - import and improve remote compilation support from FAST
4380 - make sure make check works on half a dozen of machines out there
4382 2004-07-07 Martin Quinson
4383 Let's say it's version 0.0.3 ;)
4384 - Implement conversions (yuhu!)
4385 - Let it work on solaris (beside conversion, of course)
4386 - Stupid me, using rand() to generate the conversion datatests in not wise.
4388 2004-07-06 Martin Quinson
4389 - Let make dist work, since I'm gonna need it to compile on remote hosts
4390 - Let Tests/datadesc_usage write the architecture on which the file was
4391 generated as first byte.
4392 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4393 architecture descriptions.
4394 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4395 on those architectures.
4396 - Optimization: send/recv array of scalar in one shoot
4398 2004-07-05 Martin Quinson
4399 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4401 - Plug a whole bunch of memleaks
4402 - each process now have to call gras_{init,exit}. One day, their log
4403 settings will be separated
4404 - Continue the code factorisation between SG, RL and common in Transport.
4406 2004-07-04 Martin Quinson
4408 - Redistribution between SG and RL.
4409 We wanna have to accept in SG, so move accepted related parts of RL in
4410 the common part. (more precisely, the dynar of all known sockets is no
4411 more a static in transport.c, but part of the process_data)
4413 [gras_stub_generator]
4414 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4416 2004-07-03 Martin Quinson
4417 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4418 them all up in one shot)
4420 - Fix the ugly case of reference to dynamic array.
4421 - New (semi-public) function gras_datadesc_size to allow the messaging
4422 layer to malloc the needed space for the buffer.
4424 - gras_socket_close now expect the socket to close (and not its address to
4425 put NULL in it after it). This is because the socket passed to handlers
4426 is one of their argument (=> not writable).
4428 - propagate the interface cleanup from last week in datadesc, ie remove a
4429 superfluous level of indirection. User pass adress of variable
4430 containing data (both when sending and receiving), and not of a variable
4431 being a pointer to the data. Let's say that I like it better ;)
4432 The price for that is constructs like "int msg=*(int*)payload" in
4433 handlers, but it's a fine price, IMHO.
4435 - Let it work in RL (yuhu)
4437 2004-06-21 Martin Quinson
4439 - porting SG plugin and SG select to new standards (works almost).
4440 - plug memleaks and fix bugs around.
4443 - cleanup the prototype of data recv and force users to specify when they
4444 want to handle references to objects. Test case working even for cycles.
4445 - plug memleaks. Valgrind is perfectly ok with this.
4447 2004-06-12 Martin Quinson
4449 - cleanup the separation between plugin and main code in plugin creation
4451 2004-06-11 Martin Quinson
4453 - Reput hook for raw sockets, needed for BW experiments
4454 - kill a few lines of dead code
4455 [Data description] Interface cleanup
4456 - gras_datadesc_by_name returns the searched type or NULL.
4457 That way, no variable is needed to use a type desc once, which makes
4459 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4460 two parameters were strings (field name, type name), leading to
4462 [Dicos] Interface cleanup
4463 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4464 This is consistant with the dynar API.
4466 2004-04-21 Martin Quinson
4468 - Porting to new standards.
4470 - interface cleanup.
4471 There is no bag anymore, no need to take extra provision to mask the
4472 pointers behind "ID".
4473 Better splitup of functions between files create/exchange/convert.
4474 This is still a bit artificial since convert and receive are so
4475 interleaved, but anyway.
4477 - add a queued message list to procdata (the ones not matching criteria
4479 - factorize some more code between SG and RL wrt procdata
4481 - use gras_exit in example to track memleaks
4482 - get rid of gs_example now that GS is properly integrated into gras
4483 - update run_test to integrate the latest tests (datadesc)
4485 - rename WARNINGn macros to WARNn since it prooved error-prone
4487 2004-04-19 Martin Quinson
4489 - register init/exit functions within gras module mechanism
4490 - send/receive function.
4491 Convertion is not implemented, but short-cutted if not needed.
4492 struct/array elements are sent one by one (instead of block-wise), but
4493 nobody really cares (yet). Get a prototype before optimizing.
4494 - tests (using a file socket) for DD send/receive on:
4495 - base types: int, float
4496 - array: fixed size, string (ie ref to dynamic string)
4497 - structure: homogeneous, heterogeneous
4498 - chained list, graph with cycle
4499 Believe it or not, valgrind is not too unhappy with the results. The
4500 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4504 - Bugfix when using a filename explicitly (instead of '-')
4506 2004-04-09 Martin Quinson
4508 - factorize more code between RL and SG in socket creation
4509 - Complete the implementation and tests of:
4511 o file (only in RL, and mainly for debugging)
4513 I lost 3 days to design a portable address resolver, and then decided
4514 that the prototype mainly have to run on my box.
4515 Addressing portability too early may be like optimizing too early :-/
4517 - use gras_init in the Tests instead of the crappy parse_log_opt
4518 (the latter function is removed)
4519 [Conditional execution]
4520 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4521 [Code reorganisation]
4522 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4523 Build examples against the RL library, since there is no way to disable
4524 its creation for now.
4526 For information, the beginning of coding on GRAS was back in june
4527 2003. I guess that every line has been rewritten at least twice since