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
22 - New: s4u::Disk::set_sharing_policy() and s4u::Host::set_sharing_policy().
23 Allows the configuration of non-linear resource sharing for hosts and
27 * New section "Release Notes" documenting recent and current developments.
29 ----------------------------------------------------------------------------
31 SimGrid (3.28) July 14. 2021
33 The Victoriadagarna Release.
36 - C++ platform interface: Users can now describe their platform directly in C++.
37 This provides greatly flexibility and performance improvement for complex
38 platforms. Main features:
39 - Fat-Tree/DragonFly/Torus composing: allows you to create clusters of
40 "zones", instead of single hosts. This feature enables the description
41 of clusters with complex hosts, composed of several CPUs, GPUs, etc.
42 - StarZone: new zone with a Star-like topology. The routes are defined
43 as a set of links used to communicate from node to everybody (node<->ALL).
44 - Split-Duplex links: auxiliary method to create split-duplex links in
45 the platform, easing its utilisation. It automatically creates both UP
46 and DOWN links (similarly as done in XML).
47 - Please refer to the documentation and the examples included:
48 e.g. examples/cpp/clusters-multicpu/ and examples/platforms/*.cpp.
49 - New plugin: Producer-Consumer with monitor. Just requires to include the
50 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
51 associated example (examples/cpp/plugin-prodcons).
54 - New: s4u::Comm::wait_all_for() (like s4u::Comm::wait_all, but with a timeout),
55 s4u::Io::wait_any(), s4u::Io::wait_any_for().
56 - Methods test_all/test_any/wait_all/wait_any in s4u now take their vector
57 parameter by reference, instead of a pointer.
58 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
59 Explicitly cancel the activity to get back to previous behavior.
60 - New: Link::set_concurrency_limit() to limit the amount of concurrent flows.
63 - The default SMPI compiler flags are no more taken from the environment.
64 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
65 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
67 --cfg=smpi/finalization-barrier: which can be used to add
68 a barrier inside MPI_Finalize. This can help for some codes which cleanup
69 data attached to a process, but still used in other SMPI processes.
70 --cfg=smpi/errors-are-fatal: True by default, behaves like if MPI_ERRORS_RETURN
71 is active when set to false, to keep going after a small error
72 --cfg=smpi/pedantic: True by default. Do not report some harmless MPI errors
73 which may or may not be problematic in the end.
75 - fix behaviour, as maximum iteration count could be ignored
76 - add SMPI_SAMPLE_LOCAL_TAG and SMPI_SAMPLE_GLOBAL_TAG macros, to allow user to
77 use sampling when the same kernel is called with a different set of parameters
78 which have an impact on the timing.
79 - realloc is now intercepted, to be coherent, as malloc/calloc/free were already.
80 It should now work with smpi/auto-shared-malloc-thresh.
81 - Improve error handling and reporting in multiple places
82 - Improve correctness checks on the MPI code.(MPI_Op and MPI_Datatype
83 validity checks, truncated messages are now an error, return errors
84 when explicitely deleted handles are reused, ...)
85 - RMA: multiple fixes and stability improvements.
86 - analysis (-analyze flag in smpirun):
87 - SMPI can now report buffer leaks as well as MPI handles leaks,
88 if code was compiled without SMPI_NO_OVERRIDE_MALLOC.
89 - if -trace-call-location is used when compiling, SMPI can report
90 origin of leaked handles/buffers
91 - group leaks by type/origin in output message if possible
92 - New implemented MPI calls: MPI_Comm_test_inter
95 - Changed internal implementation of bandwidth factors in network models.
96 Models affected: CM02, LV08 (default), SMPI, IB.
97 Configuration affected: "network/bandwidth-factors" and "smpi/bw-factors".
98 Bandwidth factors are applied to communications to describe that users
99 cannot use 100% of the available bandwidth. For example, the default network model,
100 LV08, applies a factor of 0.97 to the bandwidth. In older versions, this
101 behavior was implemented by limiting the bandwidth available in the LMM
102 system for this flow. This may give the false impression that there is
103 bandwidth available for other flows due to its underutilization, especially
104 for the dynamic bandwidth factors used in SMPI models.
105 To avoid this, we have modified the implementation so that each flow uses the
106 maximum physical bandwidth according to the LMM system.
107 However, the actual throughput of the flow seen by the user is defined by
108 the physical bandwidth multiplied by the bandwidth factor.
109 This change impacts on the simulation results for all network models on
110 which we have bandwidth factors configured.
111 *****************************************
112 *DO NOT MIX 3.28 RESULTS WITH OLDER ONES*
113 *****************************************
114 This change may impact on the timing of your simulation results.
115 Take care when comparing simulations from different SimGrid's
116 versions. Sorry for the inconvenience.
117 - Dynamic network factors: users can configure a callback to define
118 the network factors dynamically. This API is available at
119 simgrid::kernel::resource::NetworkModelIntf.
120 - Users have access to complete information about the current communication
121 to decide which factor to apply. This includes: message size, source and
122 destination hosts, links and zones traversed.
123 - Dynamic factors for both latency and bandwidth.
124 - For more details, see the example in (examples/cpp/network-factors).
125 - Plugin host_energy: the "watt_off" and "watt_per_state" host properties,
126 deprecated since version 3.24, are no longer supported. Instead, use
127 "wattage_off" and "wattage_per_state".
130 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
133 - New tutorial: Model-checking and formal assessment
134 - New sections: "Demystifying the routing" and "C++ platforms"
135 - Update and improve the part on visualization in MPI and Algo tutorials.
136 - Improve the section on routing: how to define it, how it's used internally
137 - Fix many issues, broken links and missing references in doxygen and Sphinx
140 - Lua platform files are deprecated. Their support will be dropped after v3.31.
143 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
145 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
146 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
147 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
148 - FG#64: Configuring smpi/IB-penalty-factors
149 - FG#67: Running computation concurrently with MPI_Iallreduce
150 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
151 - FG#71: Segmentation fault on invalid gw_src/gw_dst
152 - GH#322: Issue when an actor kills his host vm
154 ----------------------------------------------------------------------------
156 SimGrid (3.27) March 29. 2021
158 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
160 Main user-visible changes:
161 - All functions marked as deprecated for one year were removed.
162 - Removed all that was related to the "Storage" type in all APIs and bindings.
163 "Disks" have been introduced more than a year ago to replace them for the better.
164 It's time to finish this replacement.
165 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
166 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
167 avoiding "initializer element is not constant" errors with some codes. Please be
168 careful to recompile your MPI codes if switching to this new SimGrid version.
169 - Network model ns-3 now requires version ns-3.28 or higher.
170 - Removed some option aliases that were deprecated since v3.23 (June 2019)
171 - 'verbose-exit' is now 'debug/verbose-exit'
172 - 'simix/breakpoint' is now 'debug/breakpoint'
173 - 'clean-atexit' is now 'debug/clean-atexit'
176 - New functions: Comm::sendto_{init,async} to initiate a communication
177 on between two (possibly remote) hosts.
178 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
179 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
180 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
181 or Mailbox::get_async<void>() if you really want to play with void*.
182 - Unify the interface of Activity::on_{start/activity}
183 - New function: Comm::get_dst_data()
184 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
187 - ptask should now work better with multicore hosts (FG#37).
188 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
189 - Internal cleanups and refactoring around models, to prepare for programmatic
191 - Modernize our tests of the performance models
195 - New section: Configuring the logs from the command line.
196 - New section: Programming API of logs.
197 - New section: Getting realistic results
198 - New section: Examples of SMPI Usage
199 - New part: SimGrid's Internals, with a section on Design goals
200 - Move back to the classical Breathe tool chain, and fix many small glitches.
203 - Handle leak checking: SMPI will now display a message if MPI handles are still
204 present at the end of a simulation. Option smpi/list-leaks:n will display the n
205 first leaked pointers and their types.
206 - smpi/display-allocs option : Display Memory allocations analysis, with hints
207 about the largest allocations. This only works for allocations with malloc/calloc
208 and not compiled with SMPI_NO_OVERRIDE_MALLOC
209 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
210 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
211 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
212 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
213 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
214 - Replay now opens and parses the trace file only once for all ranks when possible.
216 - better handling of MPI_PROC_NULL
217 - MPI_Group_incl was broken in some cases
218 - PAPI tracing now fixed
219 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
220 - SMPI will now fail properly when receive size < sent size, instead of
221 just setting MPI_ERR_TRUNCATE and keep going.
222 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
223 - Remove useless build patches and split build/run patches on the page
224 - Build in parallel when possible
225 - Test on yum-based systems
226 - New nightly tested Proxy-apps:
227 - OpenMPI mpi-test-suite
228 - Coral MCB, LSMS, Qbox, CLOMP
229 - ECP/Mantevo Cloverleaf3D, Ember
230 - Codevault bhtree_mpi
233 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
234 - Start to overhaul the implementation to modernize it (still ongoing).
237 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
238 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
239 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
240 - Deprecate xbt_dynar_to_array().
241 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
243 C binding and interface:
244 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
245 - Constness requirement has been lowered for parameter argv of sg_actor_start()
246 and sg_actor_create() It's now (char* const*). The previous version is still
247 available as sg_actor_start_() and sg_actor_create_().
249 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
250 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
251 - FG#37: Parallel tasks are limited to 1 core per host
252 - FG#62: Running "smpirun -replay" on large networks
253 - FG!46: Fix a few potential memory leaks in SMPI colls
254 - GH#343: Wrong consumed energy when an exec timeouts
256 ----------------------------------------------------------------------------
258 SimGrid (3.26) December 15. 2020
260 The Release release (the French lockdown was eased today).
262 Important user-visible changes:
263 - SimGrid now requires a compiler with C++14 support.
264 Sibling projects should upgrade their FindSimgrid.cmake
265 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
266 several users had difficulties to understand issues when using high bandwidth or
267 small latency events. The new value was already the default for SMPI and
268 should not cause too much performance hit. It can change some simulation
269 timings if simulation was skipping a lot of small events (which was actually
270 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
271 - On some network models (L07, CM02, LV08), an implicit loopback link is created
272 for local communication. Default bandwidth was 500MBps and latency was 15µs,
273 which was rather bad. Default values have been changed to 10GBps and 0 latency.
274 Options network/loopback-lat and network/loopback-bw have been added to adapt
275 these values if needed. Network and routing models already providing separate
276 loopbacks (clusters and such) are not impacted by this change.
277 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
278 metadata that are added to each MPI message by MPI libraries (tag,
279 src/dest, and communicator information). This may change simulation times
280 for very small message sizes, but should be more realistic.
283 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
284 - New plugins: wifi_energy and link_load. See the associated examples.
285 - New examples: wifi networks, and communication suspend.
288 - LTO is now enabled for Intel/clang compilers.
289 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
290 Setting it to "auto" will use all available cores, while setting it to n will
291 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
292 - Remove obsolete runtime option 'contexts/parallel-threshold'.
293 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
294 has been kept for compatibility.
295 - Finally remove obsolete snake_case() aliases for runtime options.
296 - Further improve the documentation.
299 - the on_exit() of each actor is also executed when the simulation deadlocks.
300 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
301 An example is provided for s4u::Comm but it should work with Exec and Io.
304 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
305 https://framagit.org/simgrid/SMPI-proxy-apps
306 - MPI/IO: supports names starting with "./"
307 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
309 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
310 building shared libraries, this may cause issues, so environment variable
311 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
312 - most temporary files should now be created in /tmp dir (or equivalent).
313 If this one does not allow execution of code (noexec flag), this may cause issues.
314 Please use another tmp directory (using TMPDIR or equivalent system variable)
318 - A large refactoring is ongoing, but this should be transparent for users so far.
321 - Drop xbt_str_split().
323 C binding and interface:
324 - The return type of the sg_actor_on_exit() callbacks is now 'void'
325 (int value was previously ignored)
326 - Many C functions were renamed to follow a common naming style
327 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
328 names are kept until version 3.30.
329 - Many MSG tests were converted to the new S4U's interface in C, that
330 was extended for that.
332 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
333 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
334 - FG#41: Add sg_actor_create C interface
335 - FG#43: xbt::random needs some care
336 - FG#48: The Impossible Did Happen (yet again)
337 - FG#50: Suspending an actor executed at the current timestamp fails
338 - FG#52: Zero-seconds timeout : "That's in the past already"
339 - FG#53: Crash while using ns-3 network model
340 - FG#54: How to suspend a comm?
341 - FG!22: Stochastic Profiles
342 - FG!24: Documentation and fix for xbt/random
343 - FG!35: Add a modeling hint for parallel links in doc
344 - FG!36: [xbt/random] Read/Write the state of the RNG
345 - FG!42: Link cumulated load plugin
346 - FG!44: Test condition variable wait_until+notify_all
347 - GF#18137: Allow different stack sizes?
348 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
349 - GH#139: Allow pthread creation in SMPI
350 - GH#336: Packet-level simulation using SMPI?
351 - GH#345: Error in the LMM while migrating a VM
352 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
353 - GH#352: pip install / python setup.py install fail to find pybind11
354 - GH!337: Fix link_energy plugin for wifi platforms
355 - GH!339: Add Mailbox set_receiver method to python binding
356 - GH!344: Cast hugepages macros parameters to int64
358 ----------------------------------------------------------------------------
360 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
362 The Palindrom Day Release.
364 Important user-visible changes:
365 - Improve the Python usability (stability and documentation).
366 - A nasty synchronization bug was ironed out, see also below.
367 - Python's doc was organized as a separate tree, now integrated with C++.
368 - C bindings of S4U were not part of the doc.
369 - The C++ doc was also improved as methods are now split by theme.
370 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
371 - This is mandatory to use the Java bindings.
372 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
373 - SimDAG++: Automatic dependencies on S4U activities (experimental)
374 - Some features are already implemented but not all of them
375 - Cannot block an activity until it's scheduled on a resource
376 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
377 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
380 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
381 - Drop support for Viva/Triva (old visualization tools), hence removed
382 TRACE_get_node_types() and TRACE_get_edge_types() functions.
385 - Require pybind11 version 2.4.
386 - Greatly improve locking during context switches with Python.
387 - Declaring Python callbacks to this_actor::on_exit is now working.
388 - No more tests are failing currently, even on win32.
391 - Actor: Merge signals on_migration_start/end into on_host_change
392 - Actor: Rename migrate() into set_host()
393 - Disk: Allow users to get the read and write nominal bandwidth values
394 - Exec: Implement wait_for(timeout)
395 - Io: Implement wait_for(timeout)
398 - Parse errors now raise a simgrid::ParseError that you may want to catch.
401 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
402 been renamed to fix a spelling error. As usual, the previous definitions are
403 kept with a deprecation warning. However, to avoid runtime errors, it is now
404 forbidden to override these deprecated methods in a derived class. Please use
405 the new names immediately if you need to override them.
408 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
409 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
410 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
411 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
412 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
413 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
414 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
415 - New module: random, providing classical random numbers generators.
418 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
419 allocations' size above which they are considered "shared" by default (as if
420 they were performed through SMPI_SHARED_MALLOC macros).
421 Default = 0 = disabled feature.
422 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
423 This can cause some troubles if codes are already overriding these. If this
424 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
425 help, but will make this feature unusable.
426 - Input check is now performed much more consistently and outputs precise
427 warnings for each error found
428 - New supported calls : MPI_File_set_view, MPI_File_get_view
429 - MPI I/O now only supports disks, as storage is discontinued.
430 - MPI I/O allows opening files without specifying absolute path
431 - Note : SMPI configuration options are now initialized by smpi_init_options()
432 call, which should be called if SMPI is not being used through smpirun.
434 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
435 - FG#9: Python bindings crashing
436 - FG#39: Missing s4u::Comm::wait_any_for example
437 - FG#42: Add support for ThreadSanitizer (TSan)
438 - FG!19: Removing RngStream
439 - FG!20: A module for RNG calls
440 - FG!21: Choice between ad-hoc and standard distributions implementations
442 - FG!25: Fix link in Mutex doc
443 - FG!26: Fix links in SMPI interface doc
444 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
445 - GH#305: Unscheduled tasks are still excuted
446 - GH#313: smpirun: manual outdated w.r.t. --help
447 - GH#321: [S4U] Get task remaining work ratio
448 - GH#323: Crash when an actor turn off his physical host
449 - GH#335: Missing links on dragonfly example svg
450 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
451 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
453 ----------------------------------------------------------------------------
455 SimGrid (3.24) October 9. 2019
457 The Clean Disk Release.
460 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
461 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
462 thus becomes the default resource to execute I/O operations. Examples
463 and the file system plugin have been modified to use disks instead of
464 storage elements. Storage can still be used but is doomed to disappear.
465 Users are thus adviced to update their simulators accordingly.
466 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
467 for consistency with pthread_barrier_wait()
468 - Host::get_englobing_zone() returns the englobing netzone
469 - Actor::on_destruction is now called in the destructor
470 Actor::on_termination new signal called when the actor terminates
472 - Global signals are now part of the Engine:
473 - on_platform_creation: after config settings, before the XML parsing
474 - on_platform_created: right after the XML parsing
475 - on_time_advance: each time the clock advances
476 - on_simulation_end: after simulation, before cleanups
477 - on_deadlock: as the name implies.
479 - sg_{actor,host,link}_{data,data_set}() now all exist.
480 Use them to attach user data to the object and retrieve it.
483 - Introduce an experimental Wifi model. It sounds reasonable
484 according to the state of the art, but it still has to be properly
485 validated, at least against ns-3 if not against reality.
486 - Improved the usability of ns-3. Several bugs were ironed out.
487 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
488 It is now expressed as 'idle:epsilon:allCores' to properly model the
489 consumption of non-whole tasks on mono-core hosts. More info in the doc.
492 - convert a new set of functions to the S4U C interface and move the old MSG
493 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
496 - Fortran bindings for DVFS have been removed.
497 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
498 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
499 - SMPI can now be selected by cmake's find_module(MPI) with
500 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
501 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
502 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
503 - trace-call-location can be used with TI traces, and replayed, and is compatible
504 with smpi/comp-adjustment-file.
505 - sleep events are now correctly traced/replayed.
506 - Default for trace-call-location is now to use file names and not full paths.
507 To revert to previous behavior (in case of collision of filenames), option
508 "smpi/trace-call-use-absolute-path" can be set to yes.
511 - Use the included xxHash as an hash implem when C++14 is usable.
512 - Option model-checker/hash was removed. This is always activated now.
513 - New option smpi/buffering controls the MPI buffering in MC mode.
514 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
515 This is useful to check for MPI compliance.
518 - New section on plugins: how to define a new one, and existing ones.
521 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
522 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
525 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
526 and <mount> tags. The rationale is finish to decouple what should be handle
527 by SimGrid kernel (i.e., resources and activities) from plugins built in
528 user space (e.g., the file system plugin). This new tag currently accepts
529 only three attributes (id, read_bw, and write_bw). All the other information
530 that was declared with the storage related tags now has to be expressed as
531 properties. An example of platform using this new tag is available at
532 examples/platforms/hosts_with_disks.xml
535 - 'expect signal' can now accept more than one potential signal.
537 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
538 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
539 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
540 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
541 - FG#31: per-actor data
542 - FG#34: SG_BARRIER_SERIAL_THREAD?
543 - FG#35: model-checker does not like buster-produced binaries
544 - FG!13: MC: complete workaround in the error msg seen on modern systems
545 - FG!15: execute_flops now logs compute
546 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
547 - FG!17: ns-3: unblock the right number of communications + others issues
548 - FG!18: Improving the performance of the ns-3 bindings
549 - GH#189: Energy consumption of parallel tasks
550 - GH#219: Error in the throughput of TCP transfer
551 - GH!330: Minor polishing of English: Configuring_Simgrid
552 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
553 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
554 - GH!334: Rework energy plugin, again
556 ----------------------------------------------------------------------------
558 SimGrid (3.23.2) July 8. 2019
561 - Nicer introduction page.
562 - Migrate the "Deploy your application" page to the new doc.
563 - Move Java as a subtree of MSG.
566 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
569 - Drop xbt_dynar_sort_strings().
572 - Really fix FG#26: Turning off a link should raise NetworkFailureException
573 - FG#27: Wrong exception thrown to wait_any when link is turned off
574 - GH#328: Java: Canceling multiple tasks in a single vm/host
576 ----------------------------------------------------------------------------
578 SimGrid (3.23) June 25. 2019
580 The Exotic Solstice Release.
583 - SunOS and Haiku OS support. Because exotic platforms are fun.
584 - Stop setting random seed with srand48() at initialization.
585 - Use addr2line as a fallback for stacktraces when backtrace is not available.
586 - Build option -Denable_documentation is now OFF by default.
587 - Network model 'NS3' was renamed into 'ns-3'.
590 - Simgrid can now hopefully be installed with pip.
593 - wait_any can now be used for asynchronous executions too.
596 - New log appenders: stdout and stderr. Use stdout for xbt_help.
597 - Drop xbt_dict_dump.
600 - 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
601 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
602 - MPI interface is now const correct for input parameters
603 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
604 - Fortran bindings for async collectives.
605 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
608 - Remove option 'model-check/record': Paths are recorded in any cases now.
609 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
610 always sparse. That's an immense gain in memory, and can even be
611 faster because of cache effects. So there is no need to clutter the
612 code to allow the user to go for the unefficient mode.
615 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
616 models proved to be more accurate than these old experiments.
618 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
619 - FG#1: Broken link in error messages
620 - FG#2: missing installation documentation
621 - FG#3: missing documentation in smpirun
622 - FG#6: Python bindings not available on PyPI
623 - FG#7: simple cmake call requires doxygen
624 - FG#8: make python bindings an optional dependency
625 - FG#10: Can not use MSG_process_set_data from SMPI any more
626 - FG#11: Auto-restart actors forget their on_exit behavior
627 - FG#12: -Denable_lto=OFF doesn't disable LTO
628 - FG#13: Installs unstripped file 'bin/graphicator'
629 - FG#14: Installs the empty directory 'doc/simgrid/html'
630 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
631 - FG#17: Dead link in doc (pls_ns3)
632 - FG#20: 'tesh --help' should return 0
633 - FG#21: Documentation link on http://simgrid.org/ broken
634 - FG#22: Debian installation instruction are broken
635 - FG#26: Turning off a link should raise NetworkFailureException exceptions
636 - GH#133: Java: a process can run on a VM even if its host is off
637 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
638 - GH#326: Valgrind-detected error for join() when energy plugin is activated
640 ----------------------------------------------------------------------------
642 SimGrid (3.22) April 1. 2019
644 The Easter Christmas Release (this one is somewhat late).
647 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
648 - Not complete yet: asynchronous activities (amongst others) are still missing
649 - Still ongoing: the interface may change in the future. We need more
650 testers! Please report any glitches.
651 - No new project using Java should start now. Please switch to Python.
654 - Some of the internal cleanups may lead to speed improvements:
655 - The hard limitation on the amount of simulated actors with
656 Java+Mac was removed. Now, the available memory is the only limit.
657 - Our refcounting was tidyied, leading to 10% speedups in some cases.
658 - We are still working on making our code robust to the actor kills
659 and hosts' churn. Things are improving, but it's not perfect yet.
660 - Replaced our own code to display a backtrace (that was forking addr2line)
661 with the Boost.Stacktrace library.
662 You won't see your backtraces without this optional dependency.
663 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
664 - Stop setting random seed with srand() at initialization.
667 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
668 XML file version remains 4.2 since old files are still compatible.
671 - Process termination which was broken at version 3.21 has been repaired.
672 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
673 - Hide the examples into examples/deprecated. New users should use Python.
676 - Drop MSG_process_create_from_stdfunc() from the API.
677 This C++ function was a pimple in the C API, made necessary at some
678 point by the Java bindings. This is fixed now.
679 - Hide the examples into examples/deprecated. New users should use S4U.
680 - MSG_process_create and MSG_process_attach now crash if the host used
681 to run this process is off.
682 - Fix the protype of MSG_process_on_exit()
683 Now use (int,void*) callbacks instead of (void*,void*) ones.
684 The implementation was ways too messy to actually work, I guess.
687 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
688 too often and provide more accurate timings.
689 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
690 - MPI_Alltoallw support
691 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
692 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
693 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
694 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
696 - Basic implementation of generalized requests (SMPI doesn't
697 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
700 - Drop sg_cmdline. Please use xbt_cmdline instead.
701 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
702 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
703 OsSemaphore is implemented in a portable way with C++11 threads.
704 This should allow much more threads to be created at the same time,
705 allowing Mac and Java users to simulate many more actors.
706 - Implement the 'thread' factory with std::thread instead of xbt ones.
707 It is not possible to set the stack size with threads anymore, but
708 -fsplit-stack is the way to go nowadays when using threads.
709 - Drop the xbt_os_thread_t module (now unused)
710 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
711 - Drop xbt_str_join_array().
712 - Drop cunit, use Catch2 instead.
715 - Many cleanups in the kernel::activity namespace. This was long
716 overdue, and shall open the path to many future endeavors.
719 - #132: Java: a process can not shut down its own host
720 - #220: S4U: detached send causes exception if sender terminates soon after sending
721 - #261: Document the parameters of parallel execution's constructor
722 - #300: [s4u] BarrierPtr is missing
723 - #314: SMPI args internal cleanup
724 - #316: Fix a bug related to the CPU utilization of multi-core VM
725 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
726 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
727 - #325: Turning off a host has different behavior on sleeping actors and computing actors
729 ----------------------------------------------------------------------------
731 SimGrid (3.21) October 3. 2018
733 The Restarting Documentation (TRD) Release.
736 - Convert the user manual to Sphinx for improved usability.
737 Unstable version now hosted on https://simgrid.frama.io/simgrid
738 Many glitches remain; stay tuned.
739 - Restrict the Doxygen reference API to the installed header files.
740 (documenting all internals this way is overkill + doxygen scales badly)
741 - New tutorial on S4U (highly inspired from the old MSG one, but with
742 a git repository to fork as a starting point, and a docker image)
743 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
744 - The Developper manual is still to be converted (not compiled until then)
745 - Some parts are still missing in this conversion (such as the
746 platform and deployment sections) while others were blindly converted
747 and would need more love (such as the configuration flags).
748 Things will certainly further improve in the future releases.
751 - Fully reimplement auto_restart mechanism that was utterly broken
752 (fix #22, #131 and #281 that were all very related)
753 - Implement semaphores (fix #303 and #291)
754 - When creating an actor from a function and its parameters,
755 move-only parameters are not allowed anymore, as it would prevent
756 the actor to be restartable if its parameters are consumed on run.
757 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
758 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
759 IO operations that can be started, waited for, or canceled as a regular
761 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
765 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
766 - New host variable: 'core_count'
769 - Remove xbt_os_thread_specific features
770 - Remove portability wrapper to condition variables
771 - Remove xbt_os_thread_yield()
774 - MPICH collective selector now mimics MPICH 3.3b
775 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
776 - MPI_Init(NULL,NULL) is now allowed
777 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
778 - Command-line arguments should now be handled properly in Fortran simulations
779 - MPI Attributes and related callbacks should work in Fortran
780 - Apps using C + Fortran should now work
781 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
782 - Support MPI_Error_String
785 - Due to an internal bug, Msg.run() must now be your last line.
786 We hope to fix it in a future release, and we are sorry for the inconvenience.
789 - #22: Process autorestart seem to only work with CAS01 cpus
790 - #56: Feature request: dynamic SMPI replay
791 - #93: simgrid should not eat --help
792 - #111: How to change the loopback link characteristics?
793 - #116: Platform cluster radical
794 - #117: Platform cluster/host order
795 - #129: Replay_multiple on steroid
796 - #131: Java : autostart does not work
797 - #143: Setting a breakpoint at a given time
798 - #235: Network energy models should be integrated
799 - #264: Add ptask L07 resource tracing
800 - #271: Dynamic and manual replay of SMPI traces
801 - #279: Breakpoints option uses time deltas instead of absolute time
802 - #281: Daemonized actor no longer daemonized after an auto-restart
803 - #282: TIT convertor erase traces when using absolute path in the trace list file
804 - #285: segfault when a s4u actor kills itself with msg process tracing activated
805 - #286: Pajé traces are not exposing the number of cores
806 - #287: Command-line parsing should stop when encountering '--'
807 - #288: MPI_Init(NULL, NULL)
808 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
809 - #290: Method to check if a mailbox has a message ready to be consumed
810 - #291: [s4u] semaphores do not seem available yet
811 - #294: Alltoall collective implementation of mpich
812 - #295: Starting a Host as OFF is not supported
813 - #296: DTD too permissive
814 - #299: Add a s4u barrier example
816 ----------------------------------------------------------------------------
818 SimGrid (3.20) Released June 24. 2018
820 The proxy snake_case() release.
823 - The whole API is now using snake_case() instead of mixing it with camelCase.
824 Compatibility wrappers in place for one year.
825 - We took the opportunity to increase the consistency of the API.
826 Please report any remaining glitches.
827 - New function simgrid::s4u::Host::get_actor_count:
828 Returns the number of actors running on a specific host.
831 - SMPI is now tested with ~45 proxy apps from various sources, with none or
832 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
833 - Replay: The replay file has been re-written in C++.
834 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
835 supported. They are stored in the trace and used when replayed.
836 - Basic support of MPI_Cancel. Robustness not guaranteed.
837 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
838 (only for MPI_COMM_TYPE_SHARED).
839 - New option: smpi/privatize-libs, to add external shared libs to be privatized
840 by SMPI. They will be copied locally and loaded separately by each process.
841 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
842 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
843 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
844 with internal index optind with multiple processes. Only works if MPI_Init has
846 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
847 syntax without preprocessing tricks.
850 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
851 behavior tracing is no longer limited to MSG
852 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
853 this means that traces now need to be updated or re-obtained; both Irecv
854 and Isend lines in your traces have now in total 5 mandatory fields:
855 <rankid> <command> <to/from rankid> <tag> <size>
856 To update your traces, it suffices to add a 0 for the tag here.
857 - TIT now also supports waiting for a distinct request via MPI_Wait.
858 Wait/Test now wait for a specific request, not just the last one that was
859 issued. This unfortunately means another update, because we need to
860 identify which request you want to wait for. We do this via the
861 triplet (sender, receiver, tag), which needs to be added:
862 <rankid> <command> <sender> <receiver> <tag>
863 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
864 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
867 - The deprecation of MSG is ongoing (but this should not impact you).
868 Many MSG functions are now simple wrappers on the C API of S4U. If
869 you wish to convert your code to S4U, find the S4U counterparts of
870 your MSG calls in src/msg/msg_legacy.cpp.
871 - MSG can still be used, but won't evolve anymore.
874 - Allow to run the Link energy plugin from the command line with
875 --cfg=plugin:link_energy
876 - Rename Energy plugin into host_energy
877 - Rename Load plugin into host_load
880 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
881 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
882 - Improve performance when handling timeouts of simix synchros.
885 - Config: the C API is now deprecated (will be removed in 3.23), and
886 the C++ API has been slightly improved.
889 - Fix several build issues on OSX.
890 - Move simgrid_config.h to simgrid/config.h (old header still working)
893 - #143: Setting a breakpoint at a given time
894 - #150: Inconsistent event names in SMPI replay
895 - #258: daemonized actors hang after all non-daemonized actors have completed
896 - #267: Linker error on unit_tmgr
897 - #269: SMPI: tracing of MPI_Wait/all/any broken
898 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
900 ----------------------------------------------------------------------------
902 SimGrid (3.19.1) Released March 22 2018
904 The Fixed ABI Release.
906 As you may know, we are currently refactoring SimGrid in deep.
907 Upcoming SimGrid4 will be really different from SimGrid3: modular,
908 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
910 Our goal is to smooth this transition, with backward compatibility
911 and automatic update paths, while still progressing toward SimGrid4.
913 SimGrid remains open during works: The last pure SimGrid3 release was
914 v3.12 while all subsequent versions are usable alpha versions of
915 SimGrid4: Existing interfaces remain unchanged, but the new S4U
916 interface is budding and the internals are deeply reorganized.
918 Since 2015, we work hard to reduce the changes to public APIs. When
919 we need to rename a public library symbol in S4U, we let your
920 compiler issue an explicative warning when you use the deprecated
921 function. These messages remain for four releases, i.e. for one full
922 year, before turning into an error. Starting with v3.15, your can
923 also adapt to API changes with the SIMGRID_VERSION macro, that is
924 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
926 Starting with this v3.19.1, our commitment to reduce the changes to
927 the public interfaces is extended from the API to the ABI: a program
928 using only MSG or SimDag and compiled against a given version of
929 simgrid can probably be used with a later version of SimGrid without
930 recompilation. We will do our best... but don't expect too much of
931 it, that's a really difficult goal during such profund refactoring.
933 The difference between v3.19 and v3.19.1 is that the former was
934 accidentally breaking the ABI of MSG, while the later is restoring
937 S4U and kernel APIs will still evolve until SimGrid4, with one-year
938 deprecation warnings as currently. In fact, cleaning up these
939 interfaces and converting them to snake_case() is one release goal of
940 v3.20. But don't worry, we are working to smooth this upgrade path.
942 In summary, new projects should start with S4U to benefit of the
943 future, but old MSG projects should still be usable with no change.
945 ----------------------------------------------------------------------------
947 SimGrid (3.19) Released March 20 2018
949 The Moscovitly-cold Spring release.
951 SMPI user-visible improvements:
952 - Support of MPI_Type_create_subarray
953 - Unimplemented MPI calls now abort (instead of issuing a warning)
954 Some harmless calls around errhandlers still only issue a warning.
955 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
956 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
957 is now possible (mmap privatization was refactored to this end).
958 Adding new SMPI instances should also be technically doable now.
960 SMPI internal cleanups:
961 - Removed index notion from SMPI -- no more "getPid() - 1"!
962 In tracing files, actors are called rank-<PROCESS-ID> but these are
963 NOT the real ranks (tracing several communicators is not supported yet).
964 - Deprecate smpi_process_index() (will be removed in 3.22)
965 - Better testing on classical proxy apps, through an external project:
966 https://github.com/simgrid/SMPI-proxy-apps/
969 - Use a graphical TOC to make it easier to find the documentation you need
970 - Revamp the MSG tutorial
973 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
974 - Remove parameter of MSG_process_killall().
975 Resetting the PID was bogus anyway (several actors could have the same PID).
978 - Execution->setHost() can be called after start(), to migrate it.
979 - Comm::test_any() is now implemented.
980 - s4u::Actor now has onCreation() and onDestruction() signals.
981 - Install some kernel header files for the users' plugins and more.
982 Warning, their API is really not stable yet.
983 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
984 include/simgrid/kernel/routing/*Zone.hpp
987 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
989 Build System and other cleanups:
990 - Remove unused run-time parameter "tracing/onelink-only".
991 - SimGrid now works with intel compilers (MC still buggy)
992 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
993 - tesh kills subprocesses on timeout (but not on windows)
996 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
999 - #194: Feature request: simgrid::s4u::Comm::test_any()
1000 - #245: migrating an actor does not migrate its execution
1001 - #253: Feature Request: expose clusters as objects
1002 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
1003 - #255: Tesh broken on Windows
1004 - #256: Modernize FindSimGrid.cmake
1005 - #257: Fix (ab)use of CMake install
1007 ----------------------------------------------------------------------------
1009 SimGrid (3.18) Released December 24 2017
1011 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
1013 MSG IS NOW DEPRECATED. Please use S4U for new projects.
1014 - Support for MSG should not be removed from SimGrid before 2020,
1015 but future evolutions will be limited to the integration of
1016 user-provided patches.
1019 - s4u::onDeadlock() signal, run before stopping the simulation.
1020 - s4u::Exec: asynchronous executions (abstraction of background
1022 - s4u::Host->getLoad() returns the achieved speed in flops/s
1025 - Link::name() is deprecated and will be removed in v3.21, use
1026 Link::getCname() instead.
1027 - Mailbox::getName() changed to return a std::string, use
1028 Mailbox::getCname() instead to get a char*.
1029 - Storage::getName() changed to return a std::string, use
1030 Storage::getCname() instead to get a char*.
1031 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
1032 to have both allocation and deallocation of the map in user space.
1035 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
1036 The amount of flops remaining to do is a vector, not a scalar.
1037 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
1038 name implies on both sequential and parallel tasks.
1039 - Both changes fix GitHub's #223 using PR #237 as a basis.
1040 Thanks Michael Mercier.
1041 - Most examples were converted to S4U and hidden elsewhere at tests.
1042 Rationale: we still want MSG to work; we want newcomers to use S4U.
1045 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
1046 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
1049 - Switch to the faster dlopen privatization mechanism by default
1050 - Documentation improvments
1053 - Live migration is getting moved to a plugin. Dirty page tracking is
1054 the first part of this plugin. This imply that VM migration is now
1055 only possible if one this function is called:
1056 - C/MSG: MSG_vm_live_migration_plugin_init()
1057 - C/C++: sg_vm_live_migration_plugin_init()
1058 - Java: Msg.liveMigrationInit()
1059 For C and C++, "simgrid/plugins/live_migration.h" also has to be
1063 - Define class simgrid::xbt::Path to manage file names.
1064 - Removed unused functions:
1065 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
1066 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
1067 xbt_graph_export_graphviz()
1068 - xbt/str.h: xbt_str_join()
1069 - Remove unused datatypes:
1070 - xbt/heap.h: use std::priority_queue or boost::heap instead
1071 - xbt/swag.h: use boost::intrusive::list instead
1074 - New link_energy plugin for the consumption of the links.
1075 - All of the operations on files and storage contents have been
1076 packaged into a plugin (src/plugins/file_system). The current
1077 public interface can be found in
1078 include/simgrid/plugins/file_system.h
1079 To use these functions you now have to initialize the plugin by
1080 calling MSG_storage_file_system_init() just after calling
1081 MSG_init() or sg_storage_file_system_init() just after creating
1085 - Remove the undocumented/untested tag <include>
1088 - Remove viva specific tracing as the tool is no longer maintained
1091 - #248: Kill of finished processes leads to segfault
1092 - #240: xbt_cond_wait_timeout should gracefully return for C apps
1093 - #239: please implement signal s4u::onDeadlock()
1094 - #230: segfaults when exit() without run()
1095 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
1096 - #223: MSG_task_get_flops_amount() not working with parallel tasks
1097 - #222: Actor::kill() doesn't really kill and segfaults
1098 - #221: odd LMM warning when killing an actor
1099 - #120: Memory leak when the processes are forcefully killed
1101 ----------------------------------------------------------------------------
1103 SimGrid (3.17) Released October 8 2017
1105 The Drained Leaks release: (almost) no known leaks despite the tests.
1107 Even very long simulations will preserve your memory: our testsuite
1108 stresses SimGrid for over 45mn on fast machines for a coverage of
1109 over 80%, and there is only one single known leak, of about 4kb.
1113 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1114 handler may be given to free resources if the comm cannot be completed.
1115 - this_actor::send(mailbox) is now mailbox->put()
1116 - New: simgrid::s4u::Comm::wait_all()
1117 - New: Host.execute() for remote executions.
1120 - Unused option network/sender-gap is removed.
1123 - Deprecate MSG_task_isend_with_matching(): This unused feature
1124 really complicates our internals. Will be removed in v3.20.
1127 - Improved context termination. It is now possible to free resources, even
1128 when a process is forcibly killed.
1129 - ContextBoost: add support for Boost versions above 1.61.
1132 - Bring back run-time option --cfg=exception/cutpath to remove exception
1134 - Removed unused functions:
1135 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1136 xbt_str_rtrim(), xbt_str_trim().
1137 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1140 - Removed header files obsolete since SimGrid 3.12:
1141 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1143 - #3: SD_exit should be made optional
1144 - #120: Memory leak when the processes are forcefully killed
1145 - #159: Threading test regression in Actor refcounting
1146 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1147 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1148 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1149 - #191: VM migration and pstate
1150 - #192: Updating the energy consumptions of all hosts crashes with VMs
1151 - #195: All actors have PID=0 in the logs
1152 - #204: Sometimes segfault with thread contexts and mmap privatization
1153 - #222: Actor::kill() doesn't really kill and segfaults
1154 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1156 ----------------------------------------------------------------------------
1158 SimGrid (3.16) Released June 22. 2017.
1160 The Blooming Spring Release: developments are budding.
1163 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1164 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1165 - Java JAR file should be finally fully working on Mac OSX too.
1167 XML platforms: Switch to platform v4.1 format.
1168 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1169 - <zone> can be used as a synonym for the now deprecated <as>
1170 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1171 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1172 - <actor> can be used as a synonym for the now deprecated <process>
1173 - state_file and avail_file periodicity is now easier to express
1174 (check the documentation) the old behavior should still work.
1175 * Storage: not backward compatible, but it seems that nobody ever used it
1176 - Remove attribute 'content_type' of <storage_type>: was never used
1177 - Make attribute 'model' of <storage_type> optional: for future usage
1178 - Remove Bconnection model property: was never (in)validated. Replaced by
1179 the maximum of the read and write bandwidth as a resource constraint.
1182 - New and Backwards Compatibility break:
1183 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1184 When one wants to get the list of tasks whose states have changed during a
1185 simulation round, s/he has to allocate and free a dynar and use it as
1186 argument to this function. The former SD_simulate (double how_long)
1190 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1191 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1194 - New function to update the consumption of all hosts at once.
1195 - Fix the model for multi-core hosts, linear in the amount of busy
1196 cores with an abnormality for pIdle (see paper).
1197 BUG: Still not satisfactory for ptask on mono-cores.
1200 - The netzone are now available from the MSG API.
1201 The old names still work, but are now deprecated.
1204 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1205 - New option: smpi/keep-temps to not cleanup temp files
1206 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1207 mallocs mode is used, allows to change the size of the fake file used
1208 (default 1MB), to potentially limit the number of mappings for large runs.
1209 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1210 - Fortran ifort and flang compilers support
1211 - New RMA calls supported (experimental):
1212 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1213 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1214 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1215 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1216 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1217 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1220 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1221 - Behind the scene, ActivityImpl are now automatically refcounted.
1222 Removing this nasty bottleneck will greatly simplify our internals.
1225 - Replay: New function xbt_replay_action_get():
1226 Retrieve the function previously associated to an event type.
1227 - DROPPED FUNCTION: xbt_str_varsubst()
1228 - DROPPED MODULE: strbuff. We don't need it anymore.
1229 - DROPPED MODULE: matrix. We don't need it anymore.
1230 - DROPPED MODULE: lib. We don't need it anymore.
1232 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1234 SimGrid (3.15) Released March 22. 2017
1236 The Spring Release: continuous integration servers become green
1237 We fixed even the transient bugs on all target architectures:
1238 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1239 Windows, FreeBSD, NetBSD.
1241 - Rename NetCards to NetPoints.
1242 This was intended to help NS3 users, but that's not a netcard.
1243 That's a point in the routing algorithm, let's avoid wrong simplifications.
1246 - New: MSG_process_yield(). Stop and yield to other processes.
1247 - New: MSG_process_daemon(). Daemon processes are automatically killed
1248 when the last non-daemon process terminates
1249 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1250 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1251 to make room for the upcoming network energy plugin.
1252 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1255 - Ensure that an actor can kill itself with Process::exit()
1256 - Kill the obscure NativeException. Nobody want to survive the issues
1257 it denotes, so use JniException that is a RuntimeException (not to
1258 be caught explicitly).
1259 - Partial bug fix in initialization. SimGrid flags on command line were
1260 consumed at C level but stayed in the original Java String[] args.
1261 This could mess users' args[i] if SG flags were not put at the end of
1263 The SimGrid flags are now removed from the Java arguments. However,
1264 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1265 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1266 It might be if you have --log or --cfg flags in the command line.
1267 - Fix numerous memleaks all around the place. In particular, around VMs.
1271 - s4u::onPlatformCreated: right before the simulation starts
1272 - s4u::onSimulationEnd: right after the main simulation loop
1273 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1274 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1275 event from the availability_file changes the avail speed.
1276 - Links are now usable from s4u
1277 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1278 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1279 - New examples: The conversion of MSG examples to S4U has begun
1280 - Actors: create, kill, migration, and suspend
1281 - Applications: master-worker and token-ring
1282 - Action replay: communications and storage
1283 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1286 - Backwards Compatibility breaks
1287 - The SD_route_get_list and SD_route_get_size functions have been removed.
1288 They are replaced by the unique
1289 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1290 The route from src to dst is built in the links dynar whose size can
1291 be obtained with xbt_dynar_length.
1292 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1293 sg_host_route_bandwidth and sg_host_route_latency.
1294 Macros ensure the backwards compatibility, but you should fix your code
1297 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1298 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1299 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1300 We don't support 100% of it, but it should be enough. Contact us if not.
1301 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1302 naive and probably inaccurate)
1303 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1304 mapped onto a unique small file using some system magic.
1305 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1306 - Fortran: cleanups, fixes, support of user-added operations
1307 - RMA: MPI_Accumulate are applied in correct order.
1308 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1309 - Better support for MPI_IN_PLACE
1310 - Support for MPI_Win attrs and keyvals.
1311 - Support MPI_Comm_group_create, MPI_Type_size_x
1312 - MPI ops should be properly applied to all allowed types
1313 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1314 algorithm by a non failing one from mpich.
1317 - Kill the fifo data container: we don't use it anymore.
1320 - A <cluster> can now be created with different speed values to represent pstates.
1321 see examples/platforms/energy_cluster.xml for an example.
1323 SimGrid (3.14.159) Released December 28. 2016
1325 The previous release was only a crude approximation of a Pi release;
1326 this one should be a bit better.
1328 - Revert a last minute change that broke on Mac OSX.
1329 - Fix the changelog and NEWS files.
1330 - Improve the documentation, in particular of the routing module.
1331 - Integrate some patches flying around in Debian.
1333 SimGrid (3.14) Released December 24. 2016
1336 * The whole documentation was reworked and reorganized. There is
1337 still much room for improvement here, but we're on it.
1340 * We now need python3 for our testing infrastructure.
1341 * Model checking needs libevent
1344 * Now works on FreeBSD
1347 * Privatization now works on FreeBSD
1349 * Privatization is now activated by default in smpirun
1350 You can turn it off with -no-privatize if something goes wrong.
1352 * Call-location tracing for SMPI.
1353 You can add the exact location (filename / linenumber) of an MPI call to
1354 your trace files and slow down or speed up the simulation between two
1355 consecutive calls by using an adjustment file (see the documentation).
1357 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1358 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1359 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1360 values were ignored in some cases. The timings of these functions can now
1361 be significantly different.
1363 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1364 smpi/running-power is renamed to smpi/host-speed
1366 * smpi/grow-injected-times option to enable or disable multiplication of the
1367 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1368 make simulation less precise (but also much faster).
1370 * smpirun script should be (much) faster for large deployments.
1372 * SMPI tracing: fixed issue with poor matching of send/receives.
1374 * Replay: Fix broken waitall
1376 New functions and features
1377 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1379 Dropped / renamed functions and features
1380 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1381 - MSG_mailbox_is_empty() -> Mailbox::empty()
1382 - MSG_mailbox_front() -> Mailbox::front()
1383 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1384 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1385 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1386 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1387 * Don't pass the free_f parameter to property related functions:
1388 - MSG_host_set_property_value()
1389 - MSG_as_router_set_property_value()
1390 - MSG_storage_set_property_value()
1391 * VM properties. Since msg_vm_t are msg_host_t, just use
1392 MSG_host_get_property_value() and friends
1393 * VM I/O related things:
1394 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1395 - Unimplemented save/restore methods
1396 * MSG_as_router_get_property_value() was redundant with
1397 MSG_environment_as_get_property_value().
1398 - Removed MSG_as_router_*propert*() functions
1399 - Added MSG_environment_as_set_property_value() for consistency
1400 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1401 Well, they are still there for now, but deprecated with a warning.
1402 Please switch to xbt_dict_new_homogeneous() before this is removed
1404 * Task affinity. Its intended behavior (that was very badly tested
1405 and probably not really working) was deceiving what most users
1406 would have hoped here.
1407 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1408 * Option network/coordinates is now useless and should be dropped.
1411 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1414 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1415 * Added Dragonfly topology support
1418 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1419 EiBps, EBps, Eibps, Ebps for bandwidth.
1420 They may become useful to some lucky ones.
1423 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1424 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1425 * ex/async_waitAll: new example, on asynchronous communications
1428 * Memory usage should be decreased for simulations with a large number
1429 of processes. This also helps for SMPI.
1431 SimGrid (3.13) stable; urgency=low
1433 The Half Release, a.k.a. the Zealous Easter Trim.
1435 - We removed half of the lines, that were mostly experimental cruft.
1436 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1437 (not counting blanks and comments -- according to openhub.net)
1438 - The internals are now compiled in C++ (and will soon be clean C++)
1439 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1440 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1442 * Backwards Compatibility breaks
1443 - Removed Lua simulation bindings (switch to C or Java for that).
1444 Lua can still be used to describe platforms
1445 - Removed Java kernel plug-ins.
1446 Will be reintroduced after the ongoing major internals reorg.
1448 - the following functions were removed.
1449 They were too specific and should be reimplemented in a generic
1450 way, with filter function.
1451 - MSG_task_listen_from_host
1452 - MSG_mailbox_get_count_host_waiting_tasks
1453 - MSG_mailbox_put_with_timeout was removed.
1454 Please use MSG_task_send_with_timeout instead.
1456 - the SD_application_reinit function was removed. It has been a noop for a while.
1457 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1458 replaced by a more flexible mechanism.
1459 - The following functions thus do not exist anymore
1460 - SD_workstation_get_access_mode
1461 - SD_workstation_set_access_mode
1462 - SD_workstation_get_current_task
1463 - Basic estimation functions have been removed but can easily be replaced
1464 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1465 - SD_workstation_get_computation_time => amount / sg_host_speed()
1467 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1468 Use VM.getSpeed()*load/100 for the legacy behavior.
1470 - option enable_tracing was removed. It was not doing anything for a while.
1471 - In the ModelChecker:
1472 - the model-checker now ptraces the model-checked process which means
1473 you cannot use a debugger on the latter anymore (we might make this
1474 optional in the feature);
1475 - removed soft-dirty page tracking;
1476 - remove model-checked side snapshot management,
1477 MC_snapshot() and MC_compare_snapshot();
1478 - keep the MC_cut() function as a stub (it was not really working
1479 in the previous release).
1482 * All options are consistently in kebab-case. Old names are kept as alias.
1485 * Switch to platform v4 format.
1486 - Rename from 'power' to 'speed' the attributes describing the amount of
1487 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1488 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1489 - In <host> and <link>, attributes availability and state are gone.
1490 It was redundant with state and availability traces, and with peak values.
1491 - In <cluster>, attributes availability_file and state_file are gone.
1492 It was too complex and unused.
1493 - Kill <gpu>. Was not doing anything.
1494 - The DOCTYPE points to the right URL:
1495 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1496 (the file at this address now documents the changelog since its v1)
1497 - A warning is emitted for unit-less values (they are still accepted).
1498 - speed. Default: 'f' or 'flops'. Also defined:
1499 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1500 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1501 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1502 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1503 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1504 - latency. Default: 's' second. Also defined:
1505 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1507 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1508 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1511 * s4u::Host is now the preferred public interface to the Host features.
1512 sg_host_* functions are C bindings to the exact same behavior
1513 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1516 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1517 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1518 They were deprecated since a few years, and probably did not even compile anymore.
1521 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1522 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1523 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1524 encouraged to update their codes. A list of the modified functions can be found at the end of
1525 include/simgrid/simdag.h
1528 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1529 the lambda or closure passed as a parameter will run in kernel mode.
1530 All the callback functions should be rewritten to that interface at some point.
1533 * Reorganizing and cleaning the internals all around the place.
1536 * Remove old default barrier/bcast buggy algorithms (see #18407)
1537 * Various bug fixes to handle more codes
1538 * Remove the need for the --foreground option of smpirun (it is still
1539 accepted for backward compatibility).
1542 * Kill the setset data container: MC don't use it anymore.
1543 * Kill the queue data container: it made more sense with GRAS.
1544 * Kill the xbt_peer_t data type: it's useless without GRAS.
1545 * Kill rm_cb feature of config sets: it was never useful.
1546 * Kill graphxml parsing feature. It was not used.
1547 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1549 - xbt_dynar_sort_strings(), when the content is char*
1550 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1551 They throw exceptions on invalid input;
1552 * C++ support for declaring CLI flags (simgrid::config::Flag);
1553 * class for abstracting different signal backends (simgrid::xbt::signal).
1554 with no external dependencies (we need very simple signals).
1557 * refactoring and cleanup of the code;
1558 * ongoing process to cleanly separate the model-checking algorithms
1559 from the code model-checking support.
1561 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1563 SimGrid (3.12) stable; urgency=low
1565 The Facelift Release.
1568 * Require g++ v4.7 at least to not speak prehistorical C++.
1569 * Require Boost 1.48 (for signal2 component).
1570 * Java must be version 7 at least when activated.
1571 * Builds on Windows again (including Java bindings).
1572 * Tracing is now always enabled (no way to turn it out)
1573 * Remove GTNetS. It was not working anyway.
1574 * Various cleanups in the cmake scripts.
1575 * Move headers around to sort them out on installed systems:
1576 - instr/instr.h -> simgrid/instr.h
1577 - instr/jedule/* -> simgrid/jedule
1578 - simdag/datatypes.h was removed
1579 - simdag/simdag.h -> simgrid/simdag.h
1580 - msg/datatypes.h was removed
1581 - msg/msg.h -> simgrid/msg.h
1584 * Interface improvement:
1585 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1586 - Sanitize the interface in MSG_task_ module:
1587 - Merge two functions that were close enough but misleading:
1588 set_compute_duration(t) -> set_flops_amount(t)
1589 get_remaining_computation(t) -> get_flops_amount(t)
1590 - set_data_size(t) -> set_bytes_amount(t)
1591 get_data_size(t) -> get_bytes_amount(t)
1592 - Massive cleanups in the functions related to the energy
1593 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1594 - New: MSG_host_get_pstate()
1596 - msg/energy/onoff: switching hosts on and off
1598 * Interface improvement:
1599 - Rename Host.isAvail() to Host.isOn()
1600 - Rename Process.currentProcess() to Process.getCurrentProcess()
1601 - Rename Task.setDataSize() to Task.setBytesAmount()
1602 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1604 - #18874: Actually allows the GC to reclaim tasks
1608 - SIMIX_process_throw: raises an exception in a remote process
1609 * Refactoring: Separate sub-modules
1610 - libsmx: the public interface, as libc in a real system
1611 - popping: the strange dance that converts a user request into a kernel handling
1612 - smx_context_*: the virtualization mechanisms that embed the user code
1613 - smx_*: the handling of each simcalls
1614 * Interface cleanup:
1615 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1616 * Rename smx_action_t into smx_synchro_t, making explicit that these
1617 things are used to synchronize processes with their environment.
1618 For example, a communication is a sort of synchronization involving
1619 the communicating processes (that may block until the exchange) and
1620 the platform. The same can be said from computations, etc.
1622 - #18888: segfault when a process ends before its kill_time
1626 - 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.
1627 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1628 - MPI_Type_set_name, MPI_Type_get_name
1629 - MPI_*_c2f and MPI_*_f2c functions
1630 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1631 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1632 - Activate a lot of new tests from the mpich 3 testsuite
1634 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1635 - InfiniBand network model added: Based on the works of Jerome Vienne
1636 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1637 - When smpi/display_timing is set, also display global simulation time and application times
1638 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1639 * Collective communications
1640 - 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
1641 - 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/).
1642 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1644 - Replay now uses algorithms from wanted collective selector
1645 - Replay can be used with SMP-aware algorithms
1646 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1647 - 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.
1649 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1650 - Correctly use loopback on fat-tree clusters
1651 - Asynchronous small messages shouldn't trigger deadlocks anymore
1652 * Energy/DVFS cleanup and improvement
1653 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1654 - new: smpi_get_host_pstate()
1657 - "Full" network optimization flag was broken since Surf++
1658 - Better handling of precision flags in maxmin
1659 - Fix bug causing sometimes "Impossible" errors
1660 - Properly pass cluster properties to included hosts
1661 * Improvement of the Energy plugin.
1662 - Always update the consumption before returning that value
1663 - New property: watt_off to denote the dissipation when the host is off
1664 - New functions getWattMinAt and getWattMaxAt to retrieve the
1665 dissipation of pstates that we are not currently at.
1666 * Java: class NetworkLink renamed to Link
1667 * New function: simcall_process_get_kill_time()
1668 * Massive rename s/workstation/host/
1669 - That's intrusive, but that's good for the project consistency. Sorry.
1670 - Change config option "workstation/model" into "host/model"
1673 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1674 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1675 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1676 - xbt_dirname and xbt_basename for non-POSIX systems
1678 * The model checker now runs as a separate process.
1679 * The model checker runs must now be launched with the new simgrid-mc program.
1680 * Record/Replay: the MC can display a textual representation of a path in the
1681 execution graph. It can then be replayed outside of the model checker.
1683 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1685 SimGrid (3.11) stable; urgency=low
1690 * Normalizing pointers addresses tool for better diff between logs
1693 * Add cloud examples using new VMs
1694 - examples/msg/cloud/two_tasks_vm.tesh
1695 - examples/msg/cloud/simple_vm.tesh
1696 - examples/java/cloud/cloud.tesh
1697 - examples/java/cloud/migration/migration.tesh
1698 * Add java surf examples:
1699 - examples/java/surfPlugin/surf_plugin.tesh
1700 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1701 - examples/java/surfCpuModel/surf_cpu_model.tesh
1702 * Add SMPI+MSG example:
1703 - examples/smpi/smpi_msg_masterslave/
1708 - msg task destroy cancel test
1709 - msg_host on/off test
1710 * Move all tests in testsuite to teshsuite (adding tesh files)
1711 * Restructure teshsuites
1712 - one folder for each kind of test
1713 * Restructure AddTests.cmake
1715 - structure the order of tests (with sections)
1718 * Add virtual machine
1719 - creation of a VM on a PM
1720 - migration of a VM from a PM to another PM
1722 - MSG_process_join(msg_process_t process, double timeout)
1723 - msg_bar_t MSG_barrier_init(unsigned int count)
1724 - int MSG_barrier_wait(msg_bar_t barrier)
1725 - void MSG_barrier_destroy(msg_bar_t barrier)
1726 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1727 * New option "msg/debug_multiple_use" to help debugging when a task is used
1730 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1732 - Increase code coverage in test suites
1736 * Protect context stacks against stack overflow. The number of protected memory
1737 pages allocated on the top of each stack (1 by default) can be configured
1738 with the new command line option --cfg=contexts/guard_size:<value>.
1739 * Simcalls are now generated by a python script that
1740 - generates files included by SimGrid
1741 - checks that all the functions exist, and proposes prototypes
1743 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1744 host_set_data, host_get_data
1746 - simcall_process_join(smx_process_t process, double timeout)
1747 * Fix bug where sleeping processing could not be suspended.
1750 * Translate surf models from C to C++
1751 - Generic classes for all models: Model, Resource, Action
1752 - A generic interface for each kind of model (CPU, Network, Storage
1753 Workstation, WorkstationVM)
1755 * Translate surf routings from C to C++
1756 * Add callbacks using sigc++ or boost::signals2
1757 - Add callback functions for resource creation/destruction
1758 - Add callback functions for action state change
1759 - Handle Energy as a plugin
1760 * Replace swag by boost::intrusive
1761 * Add new routing models for clusters. For documentation, see
1762 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1763 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1764 parameters for cluster tag
1765 - Fat trees, with topology="FAT_TREE" and
1766 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1767 - see examples/platforms/torus_cluster.xml and
1768 examples/platforms/fat_tree_cluster.xml
1769 * More documentation
1772 * Hostfiles support host:nb_processes construct to deploy several processes on
1774 * Collective communication algorithms should not crash if used with
1775 improper number of nodes and report the error.
1776 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1777 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1778 MPI_Dims_create, MPI_Cart_sub are supported.
1779 * New interface to use SMPI programmatically (still depends on MSG for
1780 some parts, see examples/smpi/smpi_msg_masterslave):
1781 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1785 * Global variables privatization in MPI executables is now performed at runtime
1786 with the option smpi/privatize_global_variables (default:no).
1787 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1788 dynamic libraries loaded by the application are not privatized (static
1789 linking with these libraries is advised in this case)
1792 * Options defined in XML work correctly now.
1795 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1796 copied into simgrid.jar or not (ON by default). Use this option if you want
1797 to reduce the size of the installed simgrid.jar, *and* the native libraries
1798 are kept installed elsewhere.
1799 * Surf binding with SWIG (code generated in maintainer mode only):
1800 - plugin to handle callbacks
1801 - CPU model only for the moment
1804 * Supernovae build mode is definitively removed. It was used to improve
1805 inlining and inter-module optimizations. It is nowadays superseded by
1806 link-time optimizations commonly available in compilers.
1807 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1808 * Add boost dependency for surf++
1809 * Add new macro for tests
1810 - ADD_TESH(name <tesh_args>)
1811 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1815 - xbt_bar_t XBT_barrier_init(unsigned int count)
1816 - int XBT_barrier_wait(xbt_bar_t barrier)
1817 - void XBT_barrier_destroy(xbt_bar_t barrier)
1818 * Make the xbt_os_time module public
1820 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1822 SimGrid (3.10) stable; urgency=low
1824 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1827 * Reintegrate Java to the main archive as desynchronizing these
1828 package is not acceptable anymore (Java is now considered stable)
1829 * Add explicit synchronization facilities through semaphores
1830 * Bug fix: Task.setDataSize() only changed the C world, not the value
1831 cached in the Java world
1834 * Dramatically change the way files are handled. API and internals changed, but
1835 this part of MSG was not considered as production grade either.
1836 * Add explicit synchronization facilities through semaphores
1837 * Add a new function MSG_host_get_process_list()
1838 * Preliminary DVFS support (see examples/msg/energy/ for details)
1841 * SMPI is now included directly in the libsimgrid as the windows
1842 linker doesn't force us on splitting it anymore.
1843 * Improvements of the SMPI replay tool:
1844 - Most of the collective communications are now rooted in the same process as
1845 in the original application.
1846 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1847 used until now). Multiple data types can now be used in a trace.
1848 - The replay tool now supports traces produce either by TAU or a modified
1850 - Bug Fix: the compute part of the reduce action is now taken into account.
1851 - Gatherv collective is now supported
1852 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1853 of smpirun outputs time independent traces for the current run. One file
1854 is created per process. If too many processes are simulated, this behavior
1855 can be changed to one file for all processes by using the
1856 tracing/smpi/format/ti_one_file flag
1857 * smpirun generates the host file if needed (with given host count and platform)
1858 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1859 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1860 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1861 --cfg=smpi/coll_selector:(mpich/ompi)
1862 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1864 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1865 happening inside a collective SMPI call.
1866 * Fix the behavior of complex data types handling.
1867 * Make MPI_Wtime another synchronization point to take computations into
1869 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1870 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1871 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1872 (no privatization of global variables yet)
1873 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1876 * Verification of liveness properties is now available for SMPI applications
1877 (in addition to MSG applications)
1878 * Bugged examples using SMPI in examples/smpi/mc/
1879 * Add --cfg=model-check/visited option. Allows the verification of infinite
1880 programs. Detection of loops in the execution thanks to the system state
1881 comparison and reduction of the state space to explore. Can be combined with
1882 DPOR for safety properties.
1885 * Allow to change SimGrid configuration (see --help) within the code
1886 thanks to SD_config() as it can be done in MSG.
1887 * Add a new function SD_task_set_amount() upon user request.
1890 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1891 * Remove rule based routing (no more PCRE dependency)
1892 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1893 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1894 * Add a loopback_bw and loopback_lat options to cluster tag.
1895 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1896 routing. Now peers are encapsulated in an AS and have their own private
1897 router but this is transparent.
1900 * Our own implementation of getline is renamed xbt_getline, and gets
1901 used even if the OS provide a getline(). This should reduce the
1902 configuration complexity by using the same code on all platforms.
1903 * New type: xbt_cfg_elm_boolean.
1904 * Allow to use yes/no for boolean configuration options in the command line.
1905 * Allow to disable SimGrid cleanups at exit from command line option.
1906 There are situations where one may want a simulation to end with an exit.
1907 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1908 annoying when scripting around the simulator. Adding a
1909 --cfg=clean_atexit:no allows to circumvent this issue.
1912 * Lots of memory leaks were corrected in this release.
1913 There is no known memory leaks anymore, in all of our 600+ tests.
1914 * New command line option --version, to get SimGrid version information.
1915 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1917 * Supernovae builds are deprecated, and expected to be removed in the next
1920 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1922 SimGrid (3.9) stable; urgency=low
1924 The Grasgory release: GRAS is really dead now.
1926 * Complete overhaul of the internal host structures scheme.
1929 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1930 was considered as experimental and badly maintained since then.
1931 * Keeping it was thus a trap to our potential users, that could take
1932 it instead of MSG or SMPI by mistake despite is pity state.
1933 * GRAS seems to have very few users (if any), and no one volunteered
1934 to maintain it further. It also induces a lot of XBT code (for
1935 portability sake), that must be maintained too.
1936 * For all these reasons, we killed GRAS. If someone wants to revive it
1937 in the future, don't cry, our git history still remembers of GRAS.
1940 * Major overhaul. Merge our documentation again as time proved that
1941 splitting it was really not helping our users.
1942 * Further improve the developer documentation to help newcomers
1943 hacking on SimGrid itself. The user documentation (and in
1944 particular, the beginner documentation) is still in a sorry state.
1947 * Now works on Windows too!
1948 * Much more extensive test suite, from MPICH
1951 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1952 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1953 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1954 * Bug fix: let task be scheduled when the last dependency to be solved is
1955 a control dependency.
1956 * Remove SD_load_environment_script function.
1957 Use the C sg_platf function if you want to declare a platform
1961 * New function: MSG_process_get_number()
1962 * Old function documented: MSG_config()
1963 * Remove MSG_load_platform_script function
1964 Use the C sg_platf function if you want to declare a platform
1968 * Change the default value of the TCP_gamma constant (maximal size of TCP
1969 congestion window) to a more realistic 4MiB value. If you notice changes in
1970 your simulation results, you can fall back to the previous 20k tiny window
1971 by adding --cfg=network/TCP_gamma:20000 on command line.
1972 * (Hopefully) fix a bug wrt periodic availability and state traces
1973 * Bug fix: use default values at start when first event in availability/state
1974 trace is not at time 0.
1977 * remove the "new_" part of function name sg_platf_new_trace_connect
1978 (resulting in sg_platf_trace_connect), since it does not create
1982 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1983 use case where it's really mandatory, and maintaining it was a pain in
1985 * New: xbt_fifo_search(), search an item with a user-provided
1986 comparison function instead of dumb pointer comparison.
1989 * Fix the lua deployment:
1990 Use `simgrid.init_application()` before deployment instead of
1991 `simgrid.msg_register_application()` after.
1994 * Transfer the tracing files into the corresponding modules.
1996 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1998 SimGrid (3.8.1) stable; urgency=low
2000 The "we are told that some people want to also *install* the simgrid
2003 * Add missing file "tesh.1" to the archive.
2005 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2007 SimGrid (3.8) stable; urgency=low
2009 The Psssshiiiit release: SimGrid jumps into the Cloud.
2012 * Add an experimental interface to manipulate VMs. They are mainly
2013 process groups with very few intrinsic semantic, but they should
2014 allow you to build the semantic you want easily.
2015 * New function: MSG_host_set_property_value()
2016 * New function: MSG_process_on_exit(). To clean memory in all cases.
2017 * Bug fixes that made the host (and link) failures unusable.
2018 * Add a way to auto-restart process when the host in which they are
2019 executing comes back (ON_FAILURE="RESTART" on deployment file,
2020 MSG_process_auto_restart_set).
2021 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
2022 please stop using the old ones, they are DEPRECATED.
2024 * Deprecate functions MSG_global_init() / MSG_global_init_args()
2025 Please use MSG_init() instead. (reducing the amount of entry
2026 points in the library helps us).
2027 * Make it impossible to link against the wrong version of the lib
2028 * Deprecate MSG_clean(). No need to call it anymore.
2029 * Function MSG_get_host_number() is not deprecated anymore.
2032 * Split the doc into a user guide and a reference guide.
2033 * Start a developper guide to help people hacking on SimGrid.
2036 * Enable tracing by default. This modules rocks you should use it.
2037 * Remove option custom_flags. Now use environment variables CFLAGS
2039 * Use default cmake things to detect lua instead of home grown ones.
2040 * New option "enable_mallocators" to disable mallocators, for debugging
2041 purpose ("on" by default).
2044 * Bug fixes around the resource failures: don't let the processes
2045 survive the host they are running onto.
2046 * Add an interface to auto-restart processes when the host in which they are
2047 executing comes back.
2048 * Ensures that SIMIX_clean is called automatically. It's not part of
2049 the public interface anymore (bindings should be updated).
2052 * Bug fix for when SD_Simulate is called with a positive value: be careful
2053 when comparing doubles. Sometimes they are different for non significant
2055 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
2056 parallel task whose initial work is distributed among host according
2057 to the Amdahl's law. Such tasks are created with a parameter alpha
2058 that corresponds to the non-parallelizable part of the computation.
2059 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
2060 between two sets of workstations assuming a 1D block distribution (each
2061 workstation owns a similar share of data) on both sides.
2063 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
2064 Data redistribution will be automatically scheduled once parent and child
2065 are both scheduled. The filling of computation_amount and
2066 communication_amount structures is now done seamlessly thanks to the chosen
2068 * New function SD_workstation_dump to display various information
2069 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
2070 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
2071 bandwidth on the route onto which the task is scheduled and the amount of
2073 To divide the nominal bandwidth by 2, the rate then has to be:
2074 rate = bandwidth/(2*amount)
2075 * Compute tasks that have failed can now be rescheduled and executed again
2076 (from their beginning)
2077 * Increasing source code coverage (src/simdag is now covered at 95.8%
2081 * Re-implement time-independent trace replay using SMPI (at the
2082 smpi_smp_* level) instead of MSG. This should replace
2083 examples/msg/actions/actions.c
2084 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
2085 hindexed and structs)
2086 * Implement the exchange of non-contiguous data.
2087 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
2088 * Correct behavior of smpi/sender_gap and set its default value to 0
2089 * Add option to asynchronously send small messages to allow better
2090 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
2091 specifies the size in bytes under which messages will be asynchronously sent.
2092 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
2093 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
2096 * New C interface to define a platform: XML is now optional.
2097 For more info, please check include/simgrid/platf.h.
2098 * New interface to define random platforms from the C:
2099 For more info, please check include/simgrid/platf_generator.h and
2100 examples/msg/masterslave/masterslave_platfgen.c
2101 * Export a sg_cmdline dynar containing all the arguments we got from
2105 * Two new tracing options for adding comments to trace file so you
2106 can track your experiments (see --help-tracing for details).
2107 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
2108 * Adding the SimGrid version that generated the trace file as a comment.
2109 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
2110 * Fix to avoid key clashes on Paje links
2111 * Other minor fixes related to the Paje specification
2114 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2115 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2116 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2117 * Greatly improve the robustness of mmalloc to user errors (such as
2118 using an area after freeing it, or freeing it twice)
2120 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2122 SimGrid-java (3.8.1) stable; urgency=low
2124 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2125 Warning: all methods are visible, but only some of them are
2126 implemented so far. Check the source (src/jtrace.c)
2127 for further information.
2128 * New module: org.simgrid.msg.File (SimGrid File management functions)
2129 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2130 * Change the meaning of Process.restart: now restart the process from
2131 the beginning, like MSG_process_restart in C.
2132 * Add Process.setAutoRestart: handling of process restart when failed
2134 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2135 you to retrieve the properties of the processes/hosts
2136 * Deprecate Msg.clean(): you can just forget about it now.
2137 * New function Process.getCount(), that only works when compiling
2138 with the not yet released version 3.9 of the C library.
2140 * New context factory based on Coroutines. It mandates a modified JVM
2141 but then, the simulations run about five times faster, and there is
2142 no limit to the amount of processes (beside of the available memory).
2144 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2146 SimGrid (3.7.1) stable; urgency=low
2149 * Restore the prototype of MSG_process_create_with_environment() to
2150 the pre-3.7 situation by removing the kill_time argument.
2151 * Add a MSG_process_set_kill_time() function instead.
2154 * Fix weird behaviors when dealing with parallel tasks.
2157 * Simgrid is now built as a dll.
2158 * Simgrid-java now works on Windows.
2159 * Simgrid-Java is now included into Windows package.
2162 * First pre-build package for MacOSX.
2165 * Fix compilation when using MSG_USE_DEPRECATED.
2166 * Fix some compilation issues on Macs and Windows.
2167 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2168 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2170 We discovered that the Lua console is broken, but we are missing the
2171 manpower to fix it right now. The problem existed in 3.7 too, so we
2172 are not blocking the release for that. Sorry if you depended on this
2173 feature, any help would be really welcome.
2175 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2178 SimGrid-java (3.7.1) stable; urgency=low
2180 The "Java aint got to be bloated and slow" release
2183 * Various internal cleanups and performance improvement
2184 Simulations are expected to run up to twice faster or so
2185 * Make Process.kill(process) an instance method, not a static one
2186 * User processes are not java.lang.Thread subclasses.
2187 This breaks the compatibility (sorry), but previous API was
2188 brain-dead, making it impossible to have non-trivial
2189 initializations in the process constructor.
2190 * Require a full constructor per Process sub-class.
2191 Kinda breaks the compatibility (sorry), but this allows a much more
2192 efficient way to launch the processes at simulation startup.
2193 * Do not embeed our version of semaphores, java 1.5 can be considered
2194 as sufficiently prevalent for us to not dupplicate its features.
2198 * Add examples for almost every part of the API
2199 We spotted and fixed a lot of bugs in the process
2201 * New module: asynchronous communication API
2202 * New function: Process.sleep()
2203 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2204 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2205 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2207 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2209 SimGrid (3.7) stable; urgency=low
2211 The "spring cleanups (before the next Big Project kicks in)" release.
2214 * We can specify the SMPI latency/bandwidth factor with command line
2215 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2216 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2217 You can also use the "config tag" from platform file by adding this line
2218 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2219 (see "example/platforms/tag_config.xml" to use "config tag").
2220 Note that the command line supersedes the platform file configuration.
2221 * Change the correction factors used in LMM model, according to
2222 the latest experiments described in INRIA RR-7821.
2223 Accuracy should be improved this way.
2224 * Use the partial invalidation optimization by default for the
2225 network too. Should produce the exact same results, only faster.
2226 * Major cleanup in surf to merge models and split some optimization
2227 mechanisms from the core of the models. As a result you can now
2228 specify which model to use (e.g., --cfg=network/model:LV08
2229 --cfg=cpu/model:Cas01) and which optimization mode to use
2230 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2231 Incompatible combinations should err at initialization. See
2232 --help-models for the list of all models and optimization modes.
2233 * The CLM03 workstation model was dropped for simplicity because it
2234 used the deprecated CM02 network model. Use default instead.
2235 * Rename the TCP_gamma configuration option to network/TCP_gamma
2236 * Rename the coordinates configuration option to
2237 network/coordinates, and document it
2238 * Use now crosstraffic keyword instead of the terribly misleading
2239 fullduplex keyword. It is activated by default now in the current
2240 default model, use --cfg=network/crosstraffic:0 to turn it off.
2241 * Ongoing refactoring the model parsing to make XML files optional
2242 See include/simgrid/platf.h for details (still to be completed)
2245 * Major overhaul of the documentation. Almost instructive now :/
2246 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2247 functions and friends. This interface was considered as
2248 deprecated since over 2 years, it's time to inform our users that it is.
2249 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2250 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2251 use these (crappy) functions in your code.
2252 These functions will be removed soon. Stop using them now.
2253 * Deprecate MSG_get_host_{table,number}
2254 Implement MSG_hosts_as_dynar() instead.
2255 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2256 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2259 * Stabilize the parallel execution mode of user contexts
2260 * Introduce configuration variables to control parallel execution:
2261 - contexts/synchro: Synchronization mode to use when running
2262 contexts in parallel (either futex, posix or busy_wait)
2263 - contexts/parallel_threshold: Minimal number of user contexts
2264 that must be part of a scheduling round to switch to parallel
2265 execution mode (raw contexts only)
2266 * Fix bugs that prevented to use suspend/resume along with
2267 synchronization structures.
2268 * Fix bugs in process termination that lead to invalid memory access
2269 in very specific conditions.
2272 * Introduce a parallel mode for the models (controlled by surf/nthreads
2273 configuration item). In our tests, running the models in parallel
2274 never lead to any speedups because they are so fast that the gain
2275 of computing each model in parallel does not amortizes the
2276 synchronization costs, even when ultra fast futexes are used.
2277 This is released anyway because YMMV.
2280 * Performance boost by using a swag internally to compute the set of
2281 tasks that are finished and should constitute the return value of
2285 * Enable it by default now that it is considered rather stable.
2288 * Documentation of the tracing functions.
2289 * Performance gains when tracing categorized/uncategorized resource
2290 utilization by avoiding calls to get route when updating resource
2291 variables. LMM constraints are being used instead.
2292 * API changed to set task categories. Use MSG_task_set_category instead
2293 of TRACE_msg_set_task_category, and SD_task_set_category instead
2294 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2295 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2296 * Examples for link user variables added (see at examples/msg/tracing/)
2297 * Deprecated function TRACE_msg_set_process_category completely removed
2298 * Trace header updated according to the latest Paje file format
2299 * Tracing network lazy updates, no longer obligate users to use full updates
2300 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2301 * Experimental: let user code declare/set/push/pop application states for hosts
2302 * API changed to allow the manual creation of graph configuration files
2303 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2306 * Improve the API of Lua MSG bindings, using the Lua spirit.
2307 * Each simulated process now lives in its own Lua world (globals are
2308 automatically duplicated). It helps writing simulators. Will allow
2309 to run Splay programs within SimGrid in the future.
2310 * Add a Chord example in Lua, equivalent to the MSG one.
2313 * Start the implementation of a solution to express temporal
2314 properties, not only local assertions. This is still an
2315 experimental work in progress, stay clear from it to be safe.
2319 - Add new runtime parameters --help-logs and --help-log-categories
2320 to display information about supported logging parameters and
2322 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2323 don't exists anymore.
2324 * Mallocators: allow value NULL for the reset function.
2326 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2327 create homogeneous dictionaries, where all the elements share the
2328 same free function. Non homogeneous dictionaries will be
2329 deprecated in the next release.
2330 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2331 - Multi-level dictionaries are deprecated.
2333 - new function xbt_dynar_search_or_negative() that is useful when
2334 you have less than 2 million elements in your dynar and don't
2335 want of the extra complexity of catching exceptions when the
2336 element is not found.
2338 - Make xbt_os_thread module (for thread portability) public.
2339 Documentation is still to come, sorry.
2341 - Cleanups and simplifications to make it maintainable again.
2342 - Exotic features (such as memalign and valloc) were removed.
2343 - The metadata were extended and improved so that the
2344 model-checker becomes able to explore and inspect the heaps.
2345 - This may induce a performance drop when enable_model-checking is
2346 ON in cmake (even if it's not used in the simulation), but it is
2347 necessary at this point to get MC working.
2349 Turn model-checking OFF if simulation performance matters to you.
2350 Not enabling it at runtime is not enough, disable it in cmake.
2352 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2355 SimGrid (3.6.2) stable; urgency=low
2357 The "Not coding new stuff allows to polish old things" release.
2360 * New bindings to the NS3 packet level simulator (experimental)
2361 * Use the raw (efficient) execution contexts instead of the sysv
2362 (portable) ones when possible.
2363 * libpcre is now mandatory in any cases since not using it led to
2364 severe performance loss and possibly other issues
2365 * Update the XML platforms:
2366 - G5K: include the latest machine in Nancy
2367 - GridPP and LCG: new platforms
2368 * Documentation was partially updated, at least (more to come)
2370 Bug fixes, cosmetics and small improvements
2371 * Free terminated processes before the end of the simulation to avoid
2372 exhausting the memory of users having very dynamic amount of
2374 * Bug fix and cosmetics about canceling non-running tasks
2375 * Bug fix about the dot loader's issues when using libcgraph
2378 * Create an installer for windows with nsis (amd64 and win32)
2379 - Add an hello world project to illustrate simgrid project creation.
2380 - Embed libpcre into the Simgrid installer to avoid
2381 its compilation burden
2382 * The raw execution contexts should work on Apple now
2383 * Port to Windows 64 bits
2384 - Sysv contexts now have an implementation for this arch
2385 - GRAS communication features now support this arch
2386 * Drop support for borland compiler on windows
2387 - this code was not maintained, and we kinda depend on gcc nowadays
2388 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2389 * Fix portability issue on unstable ubuntu: linker became picky on
2392 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2394 SimGrid (3.6.1) stable; urgency=low
2396 The "Oops, we broke Macs too" release
2399 * Fixed contexts detection so that raw ones are used when possible
2400 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2401 leads to a strange error, with user code segfaulting sometimes when
2402 the generated code is not perfectly aligned (which is not
2403 controllable from the user side, depends on the amount of code)
2406 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2408 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2410 SimGrid-java (3.6) unstable; urgency=low
2413 * Split of every thing from simgrid v3.5 into a separate package.
2415 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2417 SimGrid (3.6) stable; urgency=medium
2419 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2422 * Bindings now constitute their own package, separated from the main one.
2423 Rationale: reduce our maintenance nightmare by reducing the module coupling
2424 They will soon be released on their own on gforge.
2426 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2427 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2429 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2430 * It's not quite deprecated for now because we have no replacement,
2431 but it may soon become the case.
2434 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2435 * Fortran: New user-level cache variable to store the rank of the running
2436 process. This improves performance by an order of magnitude.
2437 * C: New coccinelle script to automatically locate and modify global and
2438 local static variables.
2439 * Improved SMPI network model with a sender-side gap to account for multiple
2443 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2444 only say if a communication is finished, no matter whether it succeeded or
2445 failed. You can call MSG_comm_get_status() to know the status of a finished
2447 * New function MSG_task_dsend() to send a task and detach it. When a
2448 communication is detached, you are never notified of its success or failure
2449 and the memory is released automatically once it is finished. This function
2450 is useful when you don't care about the end nor the success of a
2452 * Change the prototypes of action replay. Sorry for inconvenience,
2453 but this is really more efficient this way (and to adapt your code,
2454 you just have to fix the initialization, that shouldn't be too long)
2455 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2456 ever managed to do anything useful with it.
2457 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2458 takes a dynar of communications. It returns immediately and gives the
2459 index of a finished communication (if any).
2460 * New example: a basic implementation of the Chord P2P algorithm.
2463 * New model for multi-core CPUs. You can now use the core attribute to
2464 precise the number of cores of a host. This is a basic model. Every
2465 process running on the host receives at most the power provided in
2466 the DTD (throughput<=power). Total throughput of process cannot exceed
2468 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2469 router linked by an up-link and a down-link (possibly asymmetrical).
2470 This kind of pattern allows to easily build last-mile model style platforms.
2471 Aggregating such patterns in a rule-based AS is thus the technique of
2472 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2473 * New model for Vivaldi routing. We transformed the Vivaldi network model
2474 into a Vivaldi routing model (based on the rule-based model). This allows to
2475 combine Vivaldi based latencies with last-mile platforms.
2478 * Added a check for NaN of IEEE754 infinite in the double entries of
2480 * Introduce a new context factory "raw", highly inspirated from the
2481 ucontext factory, but using manually crafted functions in assembly to
2482 do the work in an efficient manner.
2483 * Allow to change the used context factory at run time, not only at
2484 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2485 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2486 contexts at runtime (only with raw contexts or ucontexts).
2487 * Completely rewrote this module to allow parallel execution of user
2488 processes. Use --cfg=contexts/nthreads:N to execute user processes
2489 with N parallel threads (the default is 1, meaning no parallelism).
2490 * Allow to decide dynamically between sequential and parallel modes.
2491 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2492 processes in parallel only when their number is greater than or equal to P
2494 * Added a check for NaN of IEEE754 infinite in the double entries of
2498 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2499 won't output the state of processes when interrupted with Ctrl-C
2500 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2501 NULL-terminated array. This may solve backward compatibility issues
2502 due to the change to return type of SD_simulate. See also:
2503 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2504 * Add new macros with variable number of arguments.
2505 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2506 - in xbt/asserts.h: xbt_assert
2507 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2508 - in xbt/ex.h: THROWF and RETHROWF.
2509 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2511 * Change xbt_die() to accept a format string with arguments, just like printf.
2512 * New data structure: xbt_lib_t, like a dict but more general and with better
2516 * New configuration options
2517 Options triva/categorized and triva/uncategorized can be used to generate
2518 graph configuration files for Triva visualization tool.
2519 * Configuration option tracing/platform is renamed to tracing/categorized
2520 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2521 * New TRACE_user_link_variable interface:
2522 User provides the name of the link and the tracing variable to attach to it
2523 * the declaration of tracing categories must be done after the environment
2525 * simpler tracing interface, just one way to declare categories
2526 TRACE_category or TRACE_category_with_color, it is up to you
2527 * links in the trace file are again identified by their names
2528 * trace contains the full platform hierarchy exactly as declared using the ASes
2529 * Options tracing/msg/[task|process]:1 groups the process by hosts
2530 for both cases, tasks and processes must have names that are unique during the simulation
2531 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2532 * The experimental option tracing/msg/volume is deprecated
2533 its functionality may be reincorporated if needed
2535 The tracing generates a trace file with unordered timestamped events,
2536 because of the way the core simulator (surf) works. A script available
2537 at the tools directory (fix-paje-trace.sh) can be used to put the events
2538 in order. We have changed the tracing so it can generate ordered timestamped
2539 events in the final trace, but depending on the simulator (and how much time
2540 is simulated) that can lead to a huge memory utilization. It is deactivated
2541 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2543 Build Infrastructure
2544 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2545 - We are trying hard to keep the API stable, but it may happen that
2546 some things change (we're a research project after all, not a
2547 nuclear plant operating system). If such things should happen, you
2548 could rely on that macro to adapt.
2549 - current value: 30600 for 3.06.00, aka 3.6
2550 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2551 a number that can be compared with SIMGRID_VERSION.
2552 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2553 assertions and verbose/debug logging events are disabled at compile time.
2555 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2557 SimGrid (3.5) stable; urgency=medium
2560 * New feature to any SimGrid-based simulator: Model-Checking
2561 Check SIN#1 for more details.
2564 * New Model SMPI (three-interval linear regression for correction factors)
2565 See RR-7426, available at http://hal.inria.fr/inria-00527150
2566 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2567 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2568 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2569 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2570 * See also src/smpi/README
2575 - Tracing API changes: TRACE_start and TRACE_end should not be called
2576 by user-code. They are automatically called by simulators created
2577 with SimDAG, MSG and SMPI if the toolkit is compiled with
2578 tracing_enabled=ON. Categories declaration and utilization remain the
2579 same for MSG and SimDag.
2580 - A function was added to the tracing API to declare categories with
2582 - TRACE_category_with_color (char *category, char *color)
2583 where color must be in the following format
2584 "%f %f %f", red, green, blue
2585 and red, green, blue are float values in the interval [0, 1]
2586 - User can specify NULL as color parameter, or continue calling
2587 TRACE_category (cat)
2588 On that case, the tracing system will define random colors
2589 - The following command-line options are supported:
2590 --cfg=tracing/filename:msg.trace
2591 --cfg=tracing:1 (activate tracing, needed to use others)
2592 --cfg=tracing/platform:1 (categorized resource use)
2593 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2594 --cfg=tracing/msg/task:1 (task creation)
2595 --cfg=tracing/msg/process:1 (process creation, migration)
2596 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2597 --cfg=tracing/smpi:1 (SMPI interface tracing)
2598 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2599 - examples of examples/msg/tracing updated
2601 - DAXLoader and DOTLoader functions can generate tasks with categories
2602 - A new function to attribute a category to SD tasks:
2603 TRACE_sd_set_task_category (SD_task_t task, char *category)
2604 * Tracing the MPI interface implemented by SMPI
2605 - Collective operations are traced with states
2606 - Point-to-Point operations are traced with states/links
2607 - Tracing activated by a parameter "-trace filename" passed
2608 to smpirun during execution (considering that simgrid
2609 is compiled with tracing enabled)
2610 - To run the simulation with gdb, the simulator
2611 accepts --cfg=tracing/smpi:1 to trace SMPI
2612 - tesh files to check if smpi tracing is ok
2613 - See examples/smpi/NAS/DT-trace
2614 * GTNetS tracing re-worked
2615 - adaptation to the tracing system of GTNets to cope
2616 with modifications regarding the fullduplex mode
2617 - new tesh files to check if gtnets tracing is ok
2620 * Asynchronous communications through the functions:
2621 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2622 * New function: MSG_load_platform_script()
2623 to make possible using a lua script instead of XML files to set up platforms
2624 * New function: MSG_set_function
2625 to associate functions to processes, used when bypassing the parser
2626 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2628 Platforms: Add some more examples in examples/platforms
2629 * Grid'5000: see www.grid5000.fr
2630 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2633 * Change the XML format. This is a very important modification. SimGrid
2634 3.5 introduces a new hierarchical format based on the notion of
2635 Autonomous Systems. Compatibility with old format is ensured
2636 through the perl script provided in the install bin directory
2637 bin/simgrid_update_xml.
2638 It is now possible to build platforms with specific routing
2639 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2640 connect several platforms together. We will try to provide soon
2641 set of realistic platforms exploiting these properties (have a
2642 look at examples/platforms/ for the moment).
2643 * Take the opportunity of the XML format change to be a good XML citizen:
2644 rename link:ctn to link_ctn and similar changes (also dealt with by
2646 * Add a new routing scheme (rule-based) using regular expressions. It
2647 enables to have an extremely low memory footprint when the
2648 underlying routing is simple and can be compactly described. You
2649 need to have libpcre4-dev (perl regular expressions) installed if
2650 you want to use this routing scheme.
2651 * Revive the cluster TAG and allow to easily and efficiently (both in
2652 term of memory and speed) connect clusters together. Have a look
2653 at teshsuite/simdag/platforms/ to see how this can be done. With
2654 this tag, you can create clusters with thousands of tasks at no
2655 cost (have a look at examples/platforms/).
2656 Note: clusters are implemented as ASes, so there is no need for an
2657 enclosing AS tag if you have only one cluster in your platform.
2658 * Add new generic functions in the public interface that allows the user
2659 to call SURF 'create_resource' methods from your code (same
2660 functionality as the XML bypass mechanism but with a much lighter
2662 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2663 that uses a piecewise linear approximation to produce better
2664 results when exchanging small messages.
2665 * Add a new parameter to handle correctly full duplex link and account
2666 for interferences between uplink and downlink communications
2667 (activate with --cfg=fullduplex:1).
2670 * Rename the SD_READY (all dependencies are satisfied and task is
2671 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2672 dependencies are satisfied) state.
2673 This prevents a confusion between the notion of "ready to schedule"
2674 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2675 simulated" (SD_RUNNABLE) used by the simulation kernel.
2676 * Change the way a task is considered as ready. Instead of removing
2677 dependencies when a task is done, a counter is decreased. This way,
2678 it is always possible to reach ancestors thanks to the
2679 SD_taks_get_parents function (even after the end of the simulation.)
2680 * Change the return type of SD_Simulate from (SD_task_t*) into
2681 xbt_dynar_t. This function was in handling a dynar internally and
2682 converted it into a NULL terminated array for historical reasons.
2683 * New function SD_dotload(char*) to load a DAG described in dot
2684 format. This loader and the corresponding examples require the
2685 installation of the graphviz library.
2686 * Fix a bug in the management of tasks of size 0 in the surf network
2687 models. This problem was only visible with SIMDAG and you should
2688 thus disregard results produced with earlier versions if you
2689 relied on this feature (some tasks were blocked because of this).
2690 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2691 even though your applications were doing only point-to-point
2692 communications and sequential computations. Now you can really use any
2693 model you want (of course, if you create real parallel tasks, which are
2694 not implemented in most models beside ptaskL07, this will abort).
2695 * Add an example that schedules a DAX on an heterogeneous platform
2696 using a Min-Min strategy.
2697 * New function SD_workstation_get_current_task() that returns the kind
2698 of task currently running on a workstation in the sequential access
2700 * Raise some warnings when unexecuted tasks remains at the end of the
2701 simulation. This is usually caused by cycles in the DAG.
2704 * New function: SIMIX_process_set_function() called by MSG_set_function
2705 * Change the underlying waiting queue in semaphores so that a process
2706 can wait on several of them simultaneously (as in waitany).
2707 * Fix the way to handle tokens in semaphores so that all access patterns
2708 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2709 * kill the dirty pimple SIMIX_message_sizes_output()
2710 Please use (proper) visualization instead
2713 * New data container: setset (set of sets of elements)
2714 * New module: mmalloc (mapped malloc, allowing to have several
2715 independent segments of malloc)
2716 * New function: xbt_dict_cursor_set_data()
2717 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2718 * New function: xbt_dynar_is_empty()
2719 * New function: xbt_fifo_get_last_item()
2720 * Fix xbt_dynar_shrink(): use the right element size.
2721 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2722 uninitialized areas during expand.
2723 * Fix semaphores: previous implementation was severely broken.
2724 * Use library init/fini functions for our initialization.
2725 - you can use logs and other feature as soon as you want in your
2726 code (even before the xbt_init / MSG_init)
2727 - xbt_exit is now a no-op and produce a warning when used.
2730 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2731 benefit from the latest improvement to the simulation kernel.
2732 * Kill measurement sockets for now. If you rely on them, sorry. This
2733 release is not for you. This feature will be reintroduced in the
2734 future, but we cannot delay the release any further.
2735 * New function: gras_msgtype_get_name().
2736 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2737 * Fix (at last) the pmm example: it should not randomly fail anymore.
2739 Build chain: bug fixes and overall polishing
2740 * Cmake is now stable enough. Hence, we killed the autotools.
2741 * Port to windows ( TM :)
2742 * Fix the 'make install' target.
2743 No need to use 'make install-simgrid' anymore
2744 * Introduce a 'make dist' target compiling a *source* archive
2745 'make package' compiles a binary archive
2746 * Compile java files only on need
2747 * Add --cd and --setenv command line options to tesh
2748 * Out of source builds are not fully supported yet, but we are close
2749 * Enable supernovae and optimization flags by default for our users
2752 * Add layer to set up environment directly from lua, without XML.
2753 * The effect of gras_stub_generator can be achieved through
2754 lua too (check examples/gras/console/ping_generator.lua)
2756 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2758 SimGrid (3.4.1) stable; urgency=low
2760 The "Polishing easter eggs is probably a good idea" release.
2761 This is a bug fixes release only.
2765 * Fix a bug preventing the tasks from begin garbage collected.
2768 * Fix a bug occurring when a host involved in a communication fails.
2769 This was not detected properly by the other peer involved in the
2770 communication. Now, it's reported as a network error.
2773 * Warn the user about loop dependencies in data flow of DAX files
2774 * Obey the control-flow dependencies of DAX files
2777 * Add option "enable_smpi" allowing to not compile SMPI.
2778 Probably useful for the (Mac) users experiencing a build error here
2779 * Improve the detection of lua5.1 and ruby1.8
2781 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2783 SimGrid (3.4) stable; urgency=low
2785 The "Easter in Cargese" release. Also known as (major changes):
2787 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2788 ~> bindings were greatly improved
2789 ~> new tracing infrastructure for better visualization introduced
2791 * the "Welcome to configury modernity" release.
2792 ~> we switched from autotools to cmake, and improved our cdash
2795 A more detailled list of changes follow (full detail in svn log).
2797 Java Bindings: Various Cleanups
2798 * (install java-gcj-compat-dev on debian-like to use them)
2799 * Remove put/get: no need to export deprecated interface in Java
2800 Use send/receive instead.
2801 * Cleanup the examples and add a README per directory
2802 * Remove example autoDestination (that's the only way to go now)
2803 * Remove example explicitDestination (was a plain copy of basic)
2804 * Make JniException a runtime exception, so that there is no need to
2805 declare the fact that you may encounter such a beast. I guess that
2806 nobody will ever want to survive such error.
2807 * Create specific errors for each MSG case of failure:
2808 host failure, transfer failure, timeout, task cancelled
2809 * Cleanup the exceptions that may get thrown by each function
2810 * Other internal cleanups in Java bindings. Performance still bad :/
2811 Ruby and Lua Bindings: create them
2812 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2813 * That's new and great, you should try them out.
2814 Same functionalities than Java bindings, only even less polished
2816 * Kill the useless "rate" argument of SD_task_get_execution_time()
2817 Everyone used to provide -1 as a value, it was not used, and the
2818 semantic of a possible use wasn't even clear.
2819 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2820 either as comm costs or compute costs to mean that there is no
2821 such thing for that specific task.
2822 * Add a SD_task_set_name() function
2823 * Fix SD_task_unschedule() on typed tasks
2824 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2825 * In DAX loader, accept useless 'level' attributes to <job> since
2826 LIGO DAGs have them (seem to be to ease graphical representation).
2828 * Add an example masterslave_mailbox.c using send/receive and not
2829 the deprecated put/get interface.
2830 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2831 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2833 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2834 (declare MSG_USE_DEPRECATED to still have the old name)
2835 * Add a MSG_task_set_data() function
2836 * About trace replay (see examples/msg/actions):
2838 - Allow to work with splitted trace files for each process
2839 Give the specific trace file as argument of each process,
2840 and call MSG_action_trace_run(NULL)
2841 You can still have one merged file for all processes.
2842 - Fix implementation of collective operations
2843 * Allow task_execute() on 0-sized tasks (closes #10063)
2845 * This is the first release of SimGrid where SMPI is not considered
2846 beta anymore (even if some corners should still be improved)
2847 * Port over the new SIMIX_network submodule (internal refactoring)
2848 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2849 * Implement more missing elements of the standard:
2851 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2853 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2854 MPI_Scan() MPI_Get_processor_name()
2855 - Added implementation of missing case for Alltoall (warning: it's
2856 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2857 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2858 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2859 SMPI_MPI_Allgatherv()
2860 * Bug fixes include:
2861 - MPI_Waitsome() was broken
2862 - Allow relative includes in smpicc
2863 - Command line cfg argument 'reference_speed' was ignored...
2864 - Some functions did not properly lead to auto-benching of user code
2865 - smpicc passes -O2 by default (just like openmpi one)
2867 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2868 * Bug fixes about timeouts during communications
2869 * add SIMIX_message_sizes_output() as a pimple to write to file the
2870 amount of messages per size. Use gnuplot to get histogram.
2871 Pimple because that's the only user-visible function of simix,
2872 defined directly in xbt.h (irk, sorry)
2874 - Add a SIMIX_sem_get_capacity() function
2875 - Fix interactions with processe resume/suspende
2876 - release_forever() was stupidly broken
2877 - Fix SIMIX_display_process_status() for processes in a semaphore
2878 - Make SIMIX_sem_block_onto() user-visible
2879 * Refactoring context stuff:
2880 - Use pseudo-OOP for better modularity
2881 - reimplement SIMIX_process_kill() without process_schedule() so
2882 that the latter can take as invariant that it is called from
2884 - Merge context_start into context_new for sake of simplicity
2886 * Add a Vivaldi network model, coded live during SUD'10 ;)
2887 * Rename configuration variables to start a hierarchy:
2888 o cpu_model -> cpu/model
2889 o network_model -> network/model
2890 o workstation_model -> workstation/model
2891 * New configuration variables:
2892 o network/bandwidth_factor: correction to bandwidth
2893 o network/latency_factor: correction to latency
2894 o netwotk/weight_S: correction to the weight of competing streams
2895 * Add a long description to the models, that users can see with such
2896 argument on the command line: --cfg=cpu/model:help
2897 * --help-models display the long description of all known models
2899 * config: add the ability to set a default value after registration
2900 Does not override any previously set value (e.g. from cmd line)
2901 * dict: allow to have integer key and data.
2902 When so, you need to use the following functions
2903 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2904 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2905 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2906 In contrary to regular dicts, the key is not malloced before copy.
2907 Mixing scalar and regular elements in the same dict is not tested
2909 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2910 Tracing for Visualization:
2911 * SimGrid is now instrumented in order to generate a trace file for
2912 visualization analysis: to use it, need to compile SimGrid with the
2913 "tracing" option enabled, and instrument the program using SimGrid with
2914 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2915 (among other functions).
2916 * The instrumentation only traces the platform utilization for now
2917 * Documentation to use the tracing functions and how to analyze the
2918 traces with the Triva tool is written.
2919 * More information about: SimGrid FAQ (in the section Tracing Simulations
2922 * We moved to cmake as default build system. Autotools support will
2923 be dropped soon. Check the FAQ for more info about how to use it.
2924 * Greatly improved our cdash/ctest interactions
2925 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2926 * Added memory checking tests with valgrind; lot of memleak fixing.
2927 This may be the first release of simgrid with so few memory issues
2928 * Added code coverage tests.
2929 Our coverage is still improvable, but at least we see it on cdash.
2931 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2933 SimGrid (3.3.4) stable; urgency=low
2935 The "Desktop Grid needs love too" release (also called Xmas release).
2937 Models improvements:
2938 * Major speedup in the maxmin system solving by using lazy evaluation
2939 Instead of solving completely the maxmin system at each iteration,
2940 only invalidate (and recompute) the modified parts.
2941 This new feature is enabled in default models but you can try to
2942 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2943 * Cas01 IMproved as default CPU model
2944 This CPU model is the same Cas01 model, but it uses the
2945 maxmin-selective-update flag and a heap structure to manage
2946 actions on SURF kernel.
2947 It reduces the complexity to find the next action to finish and,
2948 consequently, it's faster than the old Cas01.
2949 This is the new default CPU model (Cas01).
2950 * Rename the old Cas01 model to Cas01_fullupdate
2951 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2952 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2953 * CpuTI (CPU Trace Integration)
2954 A new CPU model whose objective is simulate faster when using
2955 availability trace files.
2956 Instead of using a full featured, over engineered maxmin system for
2957 CPU modeling, this model does the pre-integration of traces files
2958 to calculate the amount of CPU power available, and so, executes
2959 faster than the old CPU models.
2960 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2961 * Use LV08 as default network model since it gives better accuracy
2962 for small messages and shouldn't change things for big ones.
2963 Use --cfg=network_model:CM02 to get the previous behavior.
2966 ******************************************
2967 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2968 ******************************************
2969 * The new CPU model may changes simulations!
2970 The point is that events occurring at the exact same timestamp
2971 are not scheduled in the same order with the old and new
2972 version. This may be enough to completely change the execution
2973 of simulations in some cases.
2974 * The new network model will change simulations!
2975 This new model is more realistic than the previous one, so you
2976 should consider redoing your old experiments with this model.
2977 Sorry for the inconvenience.
2980 * Introduce the supernovae compilation mode
2981 When compiled that way, the whole SimGrid (or almost) is put in a
2982 single compilation unit and compiled in one shoot.
2983 This is to help gcc which has difficulties to inline stuff from one
2985 The speedup seem to be above 15%, althrough more tests are needed on
2986 amd64 to confirm that gain.
2989 * Port of MSG's mailbox on top of SIMIX network
2990 The put/get mechanism was greatly simplified on the way.
2993 * New SIMIX network module. Provides:
2994 - Mailbox: rendez-vous mechanism to find with who you want to speak
2995 - Synchronous send/recv: easier and hopefully faster since the
2996 logic is handled in the maestro process directly now
2997 - Asynchronous send/recv: you dreamt of it? It's here now
2998 Too bad that nobody cared enough to propagate the change to MSG.
2999 * Add semaphores as SIMIX synchronization mechanism.
3002 * new function SD_daxload(char*) to load a DAX file
3003 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
3004 * Introduce typed tasks. Specify its kind and cost at creation.
3005 At scheduling, just give where it should be placed, and the cost
3006 for each involved resource is automatically computed.
3007 Existing constructors so far (more to come of course):
3008 - SD_task_create_comm_e2e() for end-to-end communication
3009 - SD_task_create_comp_seq() for sequential computation
3010 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
3011 * new function SD_task_dump() for debugging display
3012 * new function SD_task_dotty(task,FILE*) writing to file the info
3013 about the task in dotty format
3014 * SD_task_dependency_exists() can now cope with having one of its
3015 arguments NULL. If so, it tests whether the other argument has any
3017 * Add getters on list of preceding/following tasks:
3018 SD_task_get_parents(task) and SD_task_get_children(task)
3019 * Add getters on amount of workstations and list:
3020 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
3021 * Add getter on task kind: SD_task_get_kind(task)
3022 * Update the start_time and finish_time of tasks on completion/failure
3023 * Bugfix: Remove task from state swags when destroyed
3026 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
3029 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
3030 dynar with one separate thread per value of the dynar.
3031 * Change the prototype of xbt_thread_create(), sorry.
3032 Added a boolean parameter indicating whether we want to join this
3033 thread (used in SG only for now)
3034 * Implement xbt_thread_join and xbt_thread_yield in SG also.
3037 * GTNetS wrappers should now be usable again (and betterly tested too)
3038 * Fix a major regression from 3.2 where the timeout provided to
3039 MSG_task_put_with_timeout() was used as absolute time before which
3040 the comm should be done.
3041 * Start to fix the <cluster> tag.
3042 - Internal links should be good now (beside of the loopback, which
3043 use the private link instead)
3044 - paths to the external world is still rather broken
3045 - the <route:multi> tag is just broken. Actually that's brain-dead.
3046 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
3047 to make it less stupid
3048 ** Check your platform with teshsuite/simdag/platforms/flatifier **
3049 * Fix a source-level compatibility glitch from 3.2: after defining
3050 MSG_USE_DEPRECATED, you can use the old name
3051 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
3052 * Allow to compile from the SVN with automake 1.11
3053 * Fix some problems when using the "start_time" tag in deployment XMLs.
3054 * Fix #8569: XBT/synchro.h has redundant declarations
3055 * Fix #8563: MSG return values and exceptions
3056 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
3057 * Integrate patch #8636: Obey DESTDIR when installing documentation.
3058 Thanks to Robson Peixoto.
3059 * Fix a vicious bug in dictionaries inducing that some elements were
3060 not freed on xbt_dict_free()
3062 Portability report of this version:
3063 * Main portability targets:
3064 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
3065 on (amd64/i386/ia64)
3066 - mac leopard on i386
3067 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
3068 but nothing critical.
3069 * Other platforms: windows, AIX and others were not tested for this release
3071 Timing report of this version:
3072 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
3073 scenario parameters). From 8h to a few seconds in desktop grid settings.
3074 * Supernovae brings about 25% speedup on i386.
3076 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
3078 SimGrid (3.3.3) stable; urgency=low
3080 The "Need for Speed" release.
3082 The timings done to validate the 3.3.2 were faulty.
3083 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
3085 The problem was a conversion from a manually handled vector to
3086 xbt_dynar_t on the critical path.
3087 xbt_dynar_foreach calls functions, inducing stack management crap.
3089 We inlined these functions and xbt_dynar_foreach is now breath taking.
3090 We also inlined xbt_swag_belong on the way.
3092 Here are some approximate speedup measurements (on master/slaves
3093 simulations lasting between 10s and 20s each):
3094 3.3.1 -> 3.3.2: about same performance
3095 3.3.2 -> 3.3.3: 40% speedup
3096 3.3.1 -> 3.3.3: 40% speedup
3097 3.3.1 with inline patch -> 3.3.3: 30% speedup
3099 Our reading is that the refactoring which occurred in 3.3.2 made us
3100 suffer much more from the xbt_dynar_foreach low performance, but
3101 once we solved this, this refactoring proved to be very performance
3102 effective. From the 40% speedup, somehow, 10% are due to the
3103 inlining and 30% to the refactoring.
3105 That's a pitty that gcc cannot inline functions placed in other files
3106 alone. We have to choose between:
3107 - break the encapsulation (by putting private data structures and
3108 accessors in headers files to help gcc)
3109 - live with low performance
3110 - switch to a decent compiler such as icc (not quite possible).
3112 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3114 SimGrid (3.3.2) stable; urgency=low
3116 The "Simplicity does not preceed complexity, but follows it" release.
3118 The main contributors of this release were (lexical order):
3119 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3122 * Extract the routing logic into its own object.
3123 (was dupplicated in network.c and workstation_LV07.c;
3124 Allows to implement other ways of storing that info)
3125 => kill now useless network_card concept
3126 - Use dynar to represent routes (instead of void** + int*)
3127 - kill link_set (use surf_network_model->resource_set instead)
3128 - Add a command-line option to choose the routing schema to use
3129 - Add three new models:
3130 * Floyd (shortest path computed at initialization)
3131 * Dijikstra (shortest path recomputed all the time)
3132 * Cached Dijikstra (shortest path computed on need)
3133 All these models where contributed by Silas De Munck, and are
3134 described in his ICCS09 paper.
3136 * Simplify model declaration
3137 (less redirections, less function to write when defining a model)
3138 - Factorize stuff between models:
3141 surf_model_resource_set(model)
3142 surf_model_resource_by_name(model, name)
3143 - Unify the types of models in s_surf_model_t (using an union)
3144 - Embeed fields of common_public directly into s_surf_model_t
3145 - Rename model methods:
3146 action_free ~> action_unref
3147 action_change_state ~> action_state_set
3148 action_get_state ~> action_state_get
3149 - Change model methods into functions:
3150 (model)->common_public->action_use ~> surf_action_ref
3152 * Implement a generic resource; use it as ancestor to specific ones
3153 (allows to kill duplicated code in models)
3154 Drawback: timer command don't need no name nor properties;
3155 workstation_CLM03 don't need no properties
3156 (but I guess we can live with those few bytes wasted)
3158 * Improve the action object model
3159 - implement a constructor avoiding dupplicated code about field
3160 initialization in generic_action part.
3162 * Kill the SDP model: it has an external dependency, is deprecated
3163 in flavor of modern lmm models, and didn't compile since a while
3166 * Relocation of the context module from XBT to SIMIX.
3167 (the context were decoupled from the simix processes, duplicating a lot of code)
3168 => a lot of code was factorized
3169 - less overhead is introduced during scheduling
3170 - simpler API for the context factory
3171 - the logic for process creation,destruction and manipulation was simplified
3172 * Simplification of the s_smx_process_t data structure.
3173 => accesing the simix level data associated to a process is faster now,
3174 and the code is a lot more readable.
3177 * Implement some more MPI primitives:
3178 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3179 -implementation: Bcast: flat or 2-ary tree (default),
3180 Barrier: 4-ary tree,
3182 Allreduce: Reduce then Bcast
3183 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3184 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3185 Alltoallv: flat tree, like ompi
3187 * Add support for optimized collectives (Bcast is now binomial by default)
3188 * Port smpirun and smpicc to OS X
3191 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3192 Shout out if you used it.
3195 * Display the list of still queued messages in SG mode when existing
3199 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3200 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3201 * Add xbt_str_from_file(FILE*)
3202 * Add xbt_dict_get_key achieving a linear reverse search
3203 * Remove the context module
3205 Portability report of this version:
3206 * Main portability targets:
3207 - Linux(debian)/x86/context
3208 - Linux(debian)/x86/pthreads
3209 - Linux(debian)/amd64/context
3210 - Linux(debian)/amd64/pthreads
3211 On these, we still have the eratic breakages of gras/pmm and
3212 amok/saturate_sg reported in previous version. We still think
3213 that the tests are the cause of the fault, not the tested code.
3215 - Mac OSX Leopard/x86/context
3216 Still false negative in tesh autotesting.
3217 Smpi still fails, but this time because readlink does not accept -f
3218 Everything seems to work properly beside of that.
3221 - AIX version 5.3 (only tested contexts this time)
3222 Smpi still fails there because mktemp is not installed.
3223 Everything seems to work properly beside of that.
3225 I managed to compile it for the first time, but several breakages.
3226 Won't delay the release for this exotic platform.
3228 * Windows: it's still lagging behind. If you want to help, please
3231 Timing report of this version:
3232 This version seem to be more than 5% faster than 3.3.1 (on linux
3233 64bits with contextes). The gain is less than expected, we are
3234 investigating this for next release.
3236 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3238 SimGrid (3.3.1) stable; urgency=low
3241 * Implement a --cfg-help to show existing configuration variables
3242 * Build chain do not require doxygen in maintainer mode
3245 * fix a bug on struct sizeof computation, which prevented the
3246 exchange of arrays of structs in some conditions
3247 - added a regression test about this in datadesc_usage
3248 * Allow the exchange of 0-long dynamic vectors.
3249 - for that, use -1 as indicator of dynamic size instead of 0
3250 - This implied to change any size from unsigned long to long,
3251 reducing a bit communication abilities, but I guess that with
3252 64bits being quite common, this is more than enough.
3253 - This also induce a protocol change, thus bumping network protocol
3254 version from 0 to 1 (if we have external users, we have to get
3255 clean on that point too ;)
3256 - added two regression tests about this in datadesc_usage
3257 * Be more verbose when propagating local exceptions
3258 This helps debugging.
3259 * Display the status of simulated processes when receiving SIGINT in
3263 * Allow to control the simulation from a trace file.
3264 New functions MSG_action_register() and MSG_action_trace_run()
3265 The first one allows to associate a function execution to each
3266 kind of action while the second one parses a trace file and
3267 triggers the corresponding actions within the system.
3268 For now, only a toy example is provided in examples/msg/actions
3269 * Add an example of process migration in examples/msg/migration
3270 * Fix a bug in task exchange which broke MSG_task_get_sender()
3271 Add a teshsuite regression test for that.
3272 [Bug: if MSG_task_get_sender() is called after sender exit,
3274 * Fix a bug which prevented suspend/resume to work properly
3275 * Display the status of simulated processes when receiving SIGINT
3276 This fixes a regression of v3.3. due to the introduction of SIMIX
3277 * Bug fixing in failure management:
3278 - trace could not start by a failure at time 0
3279 - failure during communications were not working
3282 * Add SIMIX_process_set_name() to change the name of the current
3283 process in the log messages.
3284 * Store smx_hosts in a dict since we only retrieve them by name
3285 * Move the configuration infrastructure to surf
3288 * Move the configuration infrastructure to surf
3291 * Massive internal cleanups:
3292 - Store internal structures on processes instead of hosts (allows
3293 to have more than one process per host, in addition of being more
3295 - Cleanup the initialization/finalization process
3296 - Kill a whole bunch of unneeded synchronization:
3297 processes run in exclusive manner within the simulator
3298 - Move queues from global tables to process data fields
3300 - now accept -platform and -hostfile arguments
3301 - Pass the right rank value to processes according to the hostfile
3302 * Compile the examples by default, and use them as regression tests
3303 * Implement MPI_Wtime()
3304 * Change the reference speed to a command line option
3307 * TCP_gamma can now be specified as command line option using
3308 --cfg=TCP_gamma:10000000.0
3309 * Change the --surf-path cmd line option into --cfg=path:
3312 * Also include strbuff from xbt.h public header
3313 * xbt_ex_display(): do not free the exception after displaying
3314 This allows to do more with the given exception afterward.
3315 Users should call xbt_ex_free() themselves.
3319 Portability report of this version:
3320 * Main portability targets:
3321 - Linux(debian)/x86/context
3322 - Linux(debian)/x86/pthreads
3323 - Linux(debian)/amd64/context
3324 - Linux(debian)/amd64/pthreads
3325 These targets fail about 1/10 of times on gras/pmm, but we believe
3326 that this is because of the test, not because of simgrid.
3327 amok/saturate_sg fails even more rarely, and the test may not be
3330 - Mac OSX Leopard/x86/context
3331 The test suite still spits tons of errors because some obscure
3332 force prevents us from removing the temporary directories
3333 arguing that they still contain some metadata I've never heard of.
3334 Smpi fails because seq is not installed.
3335 Everything seems to work properly beside of that.
3338 - AIX version 5.3 (both contexts and pthread)
3339 Smpi still fails there because mktemp is not installed.
3340 XML inclusions seems rosty on AIX.
3342 * Windows: it's still lagging behind. If you want to help, please
3345 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3347 SimGrid (3.3) stable; urgency=high
3351 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3352 [Malek Cherier & Mt]
3354 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3357 Until now, GRAS were using MSG as an interface to SURF. It was
3358 quite difficult because both interface have several differences
3359 (MSG channels vs GRAS sockets were the most notable point).
3361 This also opens the gate to SMPI (which should occur soon) and speed
3362 up simulations by to 40% (even if it were not the main goal).
3364 **************************************
3365 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3366 **************************************
3367 The point is that events occurring at the exact same timestamp are
3368 not scheduled in the same order with the old and new version. This
3369 may be enough to completely change the execution of simulations in
3370 some cases. Sorry for the inconvenience.
3372 * Cleanup and upgrade the XML format to push further scalability
3373 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3375 * Improve the testing infrastructure with tesh. Now a very large part of
3376 the code is tested not only by being run but also by checking that the
3377 output match an expected output [Mt].
3379 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3380 is really less memory-demanding, which should allow you to use
3381 larger files in SimGrid [AL].
3383 * Inform valgrind about our contextes, so that it becomes usable
3384 with the default (and more effecient) version of SimGrid
3385 [contributed by Sékou Diakite, many thanks]
3388 * Introduce a listener thread in charge of receiving incoming
3389 messages from the network. It allows to overlap communication and
3390 computation but most notably, it removes some stupid deadlocks due
3391 to the fact that so far, a process could not send and receive at
3392 the same time. This made most non trivial communication schema
3394 * Convert the PIDs from long int to int to match the MSG ones (and
3395 linux ones too) [Mt]
3396 * New function: gras_agent_spawn() to launch a new process on
3397 current host. Only working in simulation for now. [Mt]
3398 * New function: gras_os_hostport() returning a constant form (ie,
3399 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3402 * Make the backtrace of exceptions more human readable [Mt]
3403 * New module: xbt/str [Mt]
3404 a ton of string utility functions (split, join, printf to a newly
3405 allocated buffer, trim, etc)
3406 * New module: xbt/hash [Mt]
3407 SHA1 hashing algorithm (more to come if needed)
3408 * New module: xbt/synchro [Mt]
3409 synchronization tools (mutex and conditions) working the same way
3410 in simulation and in real life (mainly useful for GRAS, but not
3412 * New module: xbt/queue [Mt]
3413 classical producer/consumer synchronization scheme
3414 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3415 (using the classical functions will get serialized) [Mt]
3416 * Make dictionary internal table dynamic. No need to specify its size
3417 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3419 * Make sure the log channels are organized as a tree under windows
3420 (because of ANSI C compatibility issue, any channel were child of
3421 root directly) [Mt].
3424 * Cleaned many thing in surf and fixed a few bugs [AL].
3425 * Add a nice command line configuration mechanism to compose models [AL].
3426 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3427 the previous one (KCCFLN05). It relies on something that looks like
3428 a max-min sharing mechanism but cannot be written as such. A new solver
3429 was thus designed [AL].
3430 * Add a new solver to lmm. Based on Lagrange optimization and
3431 gradient-based descent, it enables to efficiently maximise systems s.a
3433 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3435 This solver enables to propose two new network models for TCP Reno and
3436 TCP Vegas based on Low's work. These models still need to be fully
3437 tested though [Pedro Velho].
3440 * Bug fix in SD_simulate. Now the time bound given as argument is
3442 * Use the new parallel task model (ptask_L07) as default.
3443 * Use the SURF command line configuration mechanism.
3444 * 0-size tasks (for synchronization) should now work.
3446 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3448 SimGrid (3.2) stable; urgency=high
3452 We still experience issues on this platform, but we believe that at
3453 least MSG is usable.
3455 GRAS API BREAKAGE (for simplification purpose, sorry):
3456 * the gras_msgtype_by_name is not used anymore. Instead of
3457 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3458 you can write (and must)
3459 gras_msg_send(toserver, "request", &request);
3460 - If you still want to pass a gras_msgtype_t to the function (to cache
3461 the type and avoid the lookup time), use the gras_msg_send_() variant.
3462 - Impacted functions:
3463 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3464 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3465 * The callbacks are now expected to return 0 when everything went well
3466 (just like the main() function)
3468 GRAS new features and improvements:
3469 * New module mechanism where user code can use per process globals [Mt]
3470 This is similar to gras_userdata_*() functions, but for libraries. It
3471 factorize some code developped over and over in the examples and AMOK.
3472 It has still to be documented and used (only amok/peermanagement is
3474 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3475 loss when they were small enough to fit into the buffer and sent quickly
3476 enough so that they can all get received in one shoot.
3477 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3478 if not found. Use the *_or_null() variant for the old semantic.
3479 * In gras_msg_handle, do not discard messages without callback.
3480 They are probably messages to be explicitly awaited later (ie, proofs of
3481 mis-synchronization in userland since they are sent before being awaited)
3482 No big deal usually.
3483 * gras_socket_meas_send/recv: semantic changed!
3484 The numerical arguments used to be (1) the total amount of data to send
3485 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3486 messages. This was need for the fool willing to send more than MAXINT
3487 bytes on quite fat pipes.
3490 * Do really rename the hostmanagement module to peermanagement. [Mt]
3491 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3492 but this is rather new and this was documented in the module
3493 documentation (poor excuses, I admit)
3494 * Bandwidth measurement semantic changed! This follows the changes to
3495 gras_socket_meas_send/recv explained above.
3498 * A sequential mode has been added to the workstations. When a workstation
3499 is in sequential mode, it can execute only one task, and the other tasks
3500 are waiting in a FIFO. [Christophe Thiery]
3503 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3504 model for SIMDAG. [Christophe Thiery]
3505 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3506 equal to 0, which caused some bad behaviors in
3507 saturated_constraint_set_update. I now use a threshold mechanism like in
3511 * When running manually src/testall, you select specific units [Mt]
3512 testall is the result of our cunit mechanism, and should replace all
3513 the scripty thingy around since bash don't run easily on billware.
3515 * A mallocator system has been added. [Christophe Thiery]
3516 Mallocators allow you to recycle your unused objects instead of freeing them
3517 and allocating new ones.
3519 Documentation update:
3520 * FAQ reworking + New FAQs:
3521 - "Valgrind spits tons of errors!" [Mt]
3522 - "How to repport bugs" [Mt]
3523 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3524 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3525 - Communication time measurement within MSG [AL]
3526 - I experience weird communication times when I change the latency [AL]
3527 * GRAS tutorial [Mt]
3529 - an introduction to the framework and to the used communication model
3530 - an initiatic tour introducing the most proheminent features:
3532 . Lesson 0: Installing GRAS
3533 . Lesson 1: Setting up your own project
3534 o Part 2: Message passing
3535 . Lesson 2: Exchanging simple messages
3536 . Lesson 3: Passing arguments to the processes (in SG)
3537 . Lesson 4: Attaching callbacks to messages
3538 . Lesson 5: Using globals in processes
3539 . Lesson 6: Logging information properly
3540 . Lesson 7: Using internal timers
3541 . Lesson 8: Handling errors through exceptions
3542 . Lesson 9: Exchanging simple data
3543 . Lesson 10: Remote Procedure Calling (RPC)
3544 . Lesson 11: Explicitly waiting for messages
3545 . Recapping of message passing features in GRAS
3546 - A HOWTO section containing:
3547 o HOWTO design a GRAS application
3548 More are due, of course. They will come latter. In the meanwhile, you can
3549 check the examples which are still here.
3551 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3553 SimGrid (3.1) stable; urgency=high
3557 There was a stack corruption somewhere, visible only when optimizing
3558 with these versions. [Vince]
3561 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3562 module that enabled to play with Directed Acyclic Graphs. It is built
3563 directly on top of SURF and provides an API rather close to the old
3564 SG. Some old codes using SG are currently under rewrite to check that
3565 all needful functions are provided. [Christophe Thiery]
3568 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3569 extension of the classical CLM03 model that gracefully handles
3570 failures. This is now the default model for MSG and GRAS. It doesn't
3571 handle parallel tasks yet though. [AL]
3572 * Bug fix: Weights were not correctly set in the network part.
3573 WARNING: This may have resulted in incorrect results with simulations
3574 where there are more than one flow on a given link. [AL]
3577 * After a (long ?) discussion on simgrid-devel, we have decided that the
3578 convention we had on units was stupid. That is why it has been decided
3579 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3580 WARNING: This means that all previous platform files will not work as
3581 such with this version! A warning is issued to ask users to update
3583 A conversion script can be found in the contrib module of the CVS, under
3584 the name contrib/platform_generation/surfxml_update.pl [MQ]
3587 * Bug fix: Processes were started in reverse order, wrt deployment file.
3588 WARNING: if your code relies on this bug, please fix it. [AL]
3589 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3590 being executed on two different locations. [AL]
3591 * Bug fix: Failures are now better supported thanks to Derrick's tests
3592 (there was many failure situations I hadn't thought of and that weren't
3593 correctly handled). [AL]
3594 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3598 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3599 Exception occurring on server-side are propagated back to client (!).
3601 API CHANGE: the callback changed their prototype. Change:
3602 int my_handler(gras_socket_t expeditor, void *payload_data) {
3604 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3605 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3607 * New! function: gras_msg_handleall to deal with all messages arriving
3608 within a given period.
3609 * New! function: gras_socket_server_range to get a server socket in a
3610 range of port numbers (ease to avoid port number conflicts) [MQ]
3611 * New! gras processes display their backtrace when they get a SIGUSR1
3612 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3613 Sweet to debug RL processes [MQ]
3617 - Do not force experiment sizes to be expressed in kb, or it becomes
3618 impossible to measure the latency this way (needs one byte-long tests)
3619 WARNING: this changes the amok_bw_* function semantic. [MQ]
3620 - Implements the link saturation stuff. [MQ]
3621 * Peer management module:
3622 New! module factorizing code that we wrote over and over [MQ].
3625 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3626 - Test units are placed directly into the library code, they get extracted
3627 automatically and placed into the src/testall binary.
3628 - Convert most of the XBT tests to this system.
3629 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3630 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3631 live in the same C code. This required to change a little bit the API
3632 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3633 * New module: sparse graph structure with basic algorithms (this is work
3634 in progress and the API is not considered to be frozen yet). [AL]
3635 * Display more information on backtraces: source line & function names are
3636 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3637 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3638 * Reworked a little bit some #include statements to load only required
3639 headers. Some user code that relied on SimGrid to include stdlib or
3640 stdio may need to include it by themselves. [AL]
3641 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3643 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3644 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3645 but I guess that almost nobody used those parts. [MQ]
3647 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3649 SimGrid (3.0.1) stable; urgency=low
3652 * Unfortunately, I had missed 5 misnamed functions:
3653 xbt_fifo_item_t xbt_fifo_newitem(void);
3654 void xbt_fifo_freeitem(xbt_fifo_item_t);
3655 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3656 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3657 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3658 They're now deprecated. Please use their new versions:
3659 xbt_fifo_item_t xbt_fifo_new_item(void);
3660 void xbt_fifo_free_item(xbt_fifo_item_t);
3661 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3662 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3663 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3665 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3666 * Documentation: xbt_log module unmercifully reworked [MQ]
3667 * Bugfix: there was a problem with the ending of contexts with
3668 the pthread backend. It caused some weird deadlock or behavior
3669 depending on the pthread implementation. [AL]
3670 * Bugfix: get the exceptions raised in the simulator repport where
3671 and why they come from when they are not caught in time [AL, MQ]
3674 * Bugfix: Do repport the error when two non-connected hosts try to
3675 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3678 * Add additionnal checkings on communications. Assert that two
3679 communicating hosts are connected by a set of links... [AL]
3682 * Add additionnal checkings on channel values in communication [AL]
3683 * New: MSG_task_get_source to see on which host a task was generated [HC]
3684 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3685 the number of tasks waiting to be received on channel and sent
3687 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3688 waits for the first task coming from a given host.. [AL]
3690 GRAS new functionnalities: [MQ]
3691 * Enhance the parsing macro to allow the size of multidimentional objects
3692 to be given thru annotations.
3693 * New example (and documentation): Matrix Multiplication a la RPC
3694 (as when I was young!) and fix a bunch of bugs found on the way.
3696 GRAS performance improvements: [MQ]
3698 * Reduce the amount of cbps creation/destruction by making it static to
3699 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3702 * Change libdata to a set so that we can search for stuff by ID (and thus
3703 reduce the insane amount of dict lookups)
3706 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3707 RL communication on top of SG since it's so uneffective.
3708 It may also be used for inter-thread communication in RL, one day.
3709 * Use gras_datadesc_copy() to exchange messages on top of SG
3711 - improve message exchange performance on top of SG
3712 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3713 * Don't exchange on the network the size of the used part of buffer,
3714 instead, specify the possible buffer size to read().
3716 - reduces the amount of read/write calls (one pair per exchange)
3717 - reduces the amount of exchanged data (the size)
3718 - allows to retrieve all arrived data on receiver side, if we don't need
3719 it right now (subsequent read will peek the buffer)
3720 - allows the receiver to proceed with the beginning of the stream before
3721 everything is arrived
3722 - make it possible to build an iov transport (using readv/writev)
3724 - take care of the data with non-stable storage (like stacked data),
3726 * If possible, TCP send uses vector I/O (when writev() is here)
3727 - Don't use it for receive since we send data sizes and data on the
3728 same stream, so we wouldn't be able to chain large amount of chunks
3729 before having to flush the stuff to read the size.
3730 * Rework the transport plugin mechanism to simplify it and reduce the
3731 amount of pointer dereferencement when searching for the right function
3734 * I guess that now, we do almost as few system calls as possible while
3735 doing as few data copy as possible.
3737 To improve it further, we could try to send all the sizes first and then
3738 all the data (to use iov on receiving size), but it's only a partial
3739 solution: when you have 2 dimensional data, the sizes of the second
3740 dimension is data of the first dimension, so you need 3 streams.
3742 I'm not sure the potential performance gains justify the coding burden.
3744 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3746 SimGrid (3.00) stable; urgency=high
3749 * New! Give the possibility to hijack the surf parser and thus bypass
3750 MSG_create_environment and MSG_launch_application. Have a look at
3751 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3753 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3755 SimGrid (2.96) unstable; urgency=low
3760 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3761 This deprecates the xbt_error_t mechanisms.
3762 It modifies (simplifies) all XBT and GRAS API.
3763 MSG API keeps unchanged (exceptions raised by XBT are caught from
3764 within MSG and masked with existing error handling facilities)
3767 * New! Add a FATPIPE model. [AL]
3768 * New! Add a parallel task model. [AL]
3769 * New! Add automatically a loopback interface (in the default
3770 network model) if none was precised.
3773 * Bugfix: MSG_process_resume now works with the current running process.
3775 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3776 * Modification of MSG_task_get_compute_duration. Once a task has been
3777 processed, the value returned by this function is now equal to 0. [AL]
3778 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3779 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3780 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3781 of a m_task. For now, MSG_task_cancel only works with computation
3783 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3784 of the processor (in Mflop/s) regardless of the current load on the
3786 * API Change: use proper naming convention for MSG_getClock and
3787 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3789 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3790 This function changes the priority of a computation task. This priority
3791 doesn't affect the transfer rate. A priority of 2 will make a task
3792 receive two times more cpu power than the other ones. This function
3793 has been added to suit the needs of Nguyen The Loc and hasn't been that
3794 much tested yet. So if it fails, please report it and send me your code.
3796 * API Change: removed all functions and types that were marked "deprecated"
3797 since many months. Renamed MSG_global_init_args to MSG_global_init.
3799 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3801 SimGrid (2.95) unstable; urgency=low
3804 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3805 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3806 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3809 * Bugfix: Allow absolute paths for platform description files [MQ]
3810 * Bugfix: do free the variables after use. Leads to drastic performance
3812 * Implement max_duration (ie, timeouts) on resources [AL]
3815 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3817 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3818 the needed support (timeouts) [AL]
3821 * Implement measurement sockets. You can now get the bandwidth between two
3822 hosts thanks to AMOK (see below). [MQ]
3823 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3824 dynar over the network (yeah) [MQ]
3825 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3826 * simple/static token-ring example. [Alexandre Colucci and MQ]
3827 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3828 in place before [MQ]
3831 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3832 type parsing. "short" and "long" modifiers were also ignored; other
3833 modifier (such as reference level) are still ignored. [MQ]
3834 * Embeed the buffer size within the buffer itself on SG. [MQ]
3835 That way, send() are atomic and cannot get intermixed anymore (at least
3836 the ones which are less than 100k; bigger messages still have the issue)
3837 * Array size pushed by the field, not by the field type (or each
3838 and every long int will push stuff to the cbps) [MQ]
3839 * use select() to sleep since it allows to portably sleep less than one
3842 GRAS (minor cleanups)
3843 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3846 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3847 * type category 'ignored' killed as it was never used and were difficult
3849 * whether a type can cycle or not is now a flag, leaving room for more
3850 flags (as "ignored", if we feel the need one day ;)
3851 * Rename raw sockets to measurement sockets since "raw" has another
3852 meaning in networking community.
3855 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3856 and offers features not belonging to GRAS but that most applications
3857 need. One day, it may be a set of plugins loadable at runtime.
3858 * New module: bandwidth
3859 bandwidth measurement between arbitrary nodes running this module. [MQ]
3861 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3863 SimGrid (2.94) unstable; urgency=low
3865 The first beta release of SimGrid 3 !
3869 * Update the main page and the FAQ. Adding references to gforge.
3872 * Add a gras_os_getpid function.
3875 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3876 * Extend the logs so that they also print PID, hostname, date, ... if
3878 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3879 and kill the old version which were in testsuite/
3880 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3884 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3885 for users. Instead, each programming environment (GRAS, MSG,...) use it
3886 when needed to provide such a feature to users.
3887 Don't shortcut the mechanism or you will also shortcut the virtualization
3888 you need on the simulator.
3892 * Cleanups in configury with regard to compile optimization/warning flags.
3893 Also add -fno-loop-optimize to any powerpc since it's the optimization
3894 killing gcc (< 3.4.0).
3895 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3896 needed by MSG examples complications
3897 * Borrow configury beautifications from PHP
3900 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3901 warning (thanks loris for stumbling into this one).
3902 * Bugfix: stop loading private headers (gras_config.h) from the public
3906 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3907 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3908 RL would imply a DNS resolver, which is *hard* to do in a portable way
3909 (and therefore delayed).
3910 * Implement a real timer mechanism and use it in timing macros. This allows
3911 to avoid rounding errors and get a 0.000005 sec precision in timing
3912 macros. While I was at it, various cleanups:
3913 - allow to declare more than one timed section per file (fix a stupid bug)
3914 - move some private declaration to the right place
3915 - merge conditional execution and timing macros into emulation module
3916 - document the module
3917 - make sure the module cleanups its mess on gras_exit
3918 * Documentation improvements:
3919 - (new) how to compile applications using GRAS
3920 - (new) emulation support (timing macros)
3922 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3924 SimGrid (2.93) unstable; urgency=low
3926 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3929 - Use Paje properly where used. Still to be sanitized properly.
3930 - Portability fix: Add an implementation of the contexts using pthread
3934 - Add xbt_procname(): returns the name of the current process.
3935 Use it to show the current process's name in all logging.
3937 - fix detection of older flex version and the reaction, since we do
3938 depend on modern ones (we use lex_destroy)
3939 - Better separation of SG and RL in the libs: remove all simulation code
3940 from libgras. As a result, this lib is now only 200k when stripped.
3941 Some of the xbt modules may also be duplicated (two sets and such) and
3942 should be cleaned/killed before SG3.
3943 - Insist on using xlC on AIX because of weird problems involving gcc there.
3944 - Cleanup the make remote stuff. This is now done by scripts
3945 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3946 mainly for our private use, but we're working on changing them to user
3949 - Bugfix: flush the socket on close only if there is some *output*.
3950 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3952 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3955 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3957 SimGrid (2.92) unstable; urgency=low
3959 Alpha 3 on the path to SimGrid 3
3963 - New! First try of benchmarking macros.
3964 - New! First try so that gras_stub_generator generate deployment and
3965 remote compilation helpers.
3967 - Bugfix: Initialization fix in msg_test.
3971 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3973 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3975 - API Change: gras_os_sleep to take the amount of seconds as a double.
3976 Accepting an int was error prone since it was the only location where
3977 seconds were coded as such. It leaded to damn rounding errors.
3978 - Bugfix: Hard to belive that timers ever worked before this.
3980 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3982 SimGrid (2.91) unstable; urgency=low
3984 Alpha 2 on the path to SimGrid 3
3988 - Bug fix in the lmm_solver.
3990 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3991 through the function MSG_paje_output.
3992 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3993 - It is possible to bound the rate of a communication in MSG with
3994 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3995 to write it in the changelog).
3996 - Bug fix to let GRAS run on top of MSG until we move it directly on top
4001 - Various cleanups to the autotools stuff
4002 - Begin to move Gras examples to examples/gras/
4003 - Let make distcheck work again (yeah!)
4005 - documentation overhauled using doxygen.
4006 gtk-doc-tools is dead in SimGrid now.
4007 - Automatically extract all existing logging categories, and add the list
4008 to the documentation (long standing one, to say the less)
4010 - Cleanup the known architecture table. Reorder the entries to group what
4011 should be, and use a more consistent naming scheme.
4012 (some of the test dataset are still to be regenerated)
4013 - New! Allow library to register globals on each process just as userdata
4015 This is implemented using a xbt_dict and not a xbt_set, so we loose the
4016 lookup time (for now).
4017 Use it in msg and trp.
4018 This cleans a lot the internals and helps enforcing privacy of the
4019 headers between the gras components.
4020 - New! Add a timer mechanism, not unlike cron(8) and at(1).
4021 - Bugfix: gras_os_time was delirious in RL.
4022 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
4024 - Reenable GRAS now that it works.
4026 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
4028 SimGrid (2.90) unstable; urgency=low
4030 Alpha 1 on the path to SimGrid 3
4032 * It is a long time since the last release of SimGrid. I'm sorry about
4033 that but as I had told you, I was rewriting a lot of things. I apologize
4034 to those who had been reporting bugs to me and that I had not answered.
4035 If your bug is still in the new version, please tell me. Here is a
4036 summary of the main changes.
4038 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
4039 lead by Martin Quinson. As a consequence SimGrid gains a lot in
4040 portability, speed, and a lot more but you'll figure it out later.
4041 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
4042 I wanted to release the new MSG as soon as possible and I have
4043 broken GRAS, which is the reason why, for now, only MSG is fully
4044 functional. A laconic description of these projects is available
4045 in the documentation.
4047 * REVOLUTION 2: I have removed SG and I am now using a new simulation
4048 kernel optimized for our needs (called SURF but only the developers
4049 should use it). Hence, MSG is now roughly 30 times faster and I think
4050 that by rewriting a little bit MSG, I could event speed it up a little
4051 bit more. Beside the gain in speed, it is also much easier to encode a
4052 new platform model with SURF than it was with SG. More to come...
4054 * REVOLUTION 3: I have tried to change a little as possible the API of
4055 MSG but a few things really had to disappear. The main differences
4056 with the previous version are:
4057 1) no more m_links_t and the corresponding functions. Platforms are
4058 directly read from an XML description and cannot be hard-coded
4059 anymore. The same format is used for application deployment
4060 description. The new format is described in the documentation.
4061 Have a look in tools/platform_generation. There is a tiny script
4062 that converts from the old platform format to the new one. Concerning
4063 the application deployment format, parsing the old one is tricky.
4064 I think most of you should however be able to convert your files. If
4065 it is really an issue, I can write a C code that does the conversion.
4067 2) the toolbox tbx does not exist anymore. We now have a library
4068 with much more data-structures but without the hash-tables (we have
4069 dictionaries that are much faster).
4071 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
4073 *****************************************************************************
4074 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
4075 * don't feel like converting the oldies. (Mt) *
4076 *****************************************************************************
4079 Version 2.90: "the long awaited one"
4080 - Finished rewriting and debugging MSG. Rewrote the documentation.
4081 - disable GRAS for now since it needs to be ported to the newest SG
4084 - Finish the port to windows (using mingw32 for cross-compile)
4087 - Main loop and datastructures of SURF. A cpu resource object is
4088 functional. Surf can thus be used to create cpu's with variable
4089 performance on which you can execute some actions.
4091 2004-11-15 Martin Quinson
4092 - Port to ARM. Simply added the alignment and size descriptions. Should
4093 work, but the ARM machines are so slow that I didn't had the opportunity
4094 to 'make check' over there yet.
4096 2004-11-15 Arnaud Legrand
4097 - Trace manager now written. It uses a heap structure and is therefore
4098 expected to be efficient. It may however be speeded up (particularly
4099 when many events occur at the same date) by using red and black
4100 trees. One day maybe...
4101 - Max-min linear system solver written. It uses a sparse matrix
4102 structure taking advantage of its expected use. Most operations are
4103 O(1) and free/calloc are called as few as possible. The computation of
4104 the minimum could however be improved by using a red and black tree
4107 2004-11-03 Arnaud Legrand
4108 - Rename every gras_* function that was in xbt/ to its xbt_
4110 - Add a heap and a doubly-linked list to xbt
4111 - Added a dichotomy to the dictionaries. make check works as well before
4112 so I assume that the patch is correct. I do not know however if things
4113 run effectively faster than before now. :)
4115 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4116 SimGrid, and the version number is bumped to 2.x
4118 2004-10-29 Martin Quinson
4119 - Introduction of the remote errors.
4120 They are the result of a RMI/RPC on the remote machine.
4121 ErrCodes being scalar values, you can't get the host on which those
4122 errors did happen. Extending the error mechanism as in Gnome is possible.
4123 No idea yet whether it is a good idea.
4125 2004-10-28 Martin Quinson
4126 - Interface revolution: the Starred Structure Eradication.
4127 I used to do typedef struct {} toto_t; and then handle *toto_t.
4128 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4129 - ???_t is a valid type (builded with typedef)
4130 - s_toto_t is a structure (access to fields with .)
4131 - s_toto is a structure needing 'struct' keyword to be used
4132 - e_toto_t is an enum
4133 - toto_t is an 'object' (struct*)
4135 typedef struct s_toto {} s_toto_t, *toto_t;
4136 typedef enum {} e_toto_t;
4137 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4138 s_toto_t) is private.
4140 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4141 so that it can set the variable to NULL. It was so for dicts and sets,
4142 it changed for dynars.
4144 - Fix a bunch of memleaks in dict_remove
4145 - Fix a bug in sg/server_socket opening: it failed all the time.
4147 2004-10-07 Martin Quinson
4148 - Speed up dynar lookup operation a bit.
4150 gras_dynar_get is dead.
4152 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4153 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4154 (which gives you the address of the stored data).
4156 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4157 copy of the data using an affectation to do the job and not a memcpy.
4159 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4160 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4162 It does not really speedup the dynar test because they are
4163 setting elements all the time (and look them seldom). But the dict does
4164 far more lookup than setting.
4166 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4168 2004-10-05 Martin Quinson
4169 - Allow to (en/dis)able the cycle detection at run time.
4171 Whether we should check for cycle or not is now a property of each
4172 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4173 datadesc_cycle_unset allow to remove this property when previously set.
4175 Note that the cycle detection is off by default since it impacts the
4176 performance. Watch the data you feed GRAS with ;)
4178 This property is hereditary. Any element embedded in a structure having it
4179 set have it set for the time of this data exchange.
4181 You should set it both on sender and receiver side. If you don't set it on
4182 sender side, it will enter an endless loop. If you forget on receiver
4183 side, the cycles won't be recreated after communication.
4185 - Header reorganization.
4186 Kill gras_private.h, each submodule must load the headers it needs.
4188 2004-10-04 Martin Quinson
4189 - Interface revolution: do not try to survive to malloc failure.
4191 Now, gras_malloc and friends call gras_abort() on failure.
4192 As a conclusion, malloc_error is not a valid error anymore, and all
4193 functions for which it was the only gras_error_t return value are
4194 changed. They now return void, or there result directly.
4195 This simplify the API a lot.
4197 2004-09-29 Martin Quinson
4198 - Re-enable raw sockets.
4199 Created by gras_socket_{client,server}_ext;
4200 Used with gras_raw_{send,recv}
4203 It should allow to kill the last bits of gras first version soon.
4205 This is not completely satisfactory yet (duplicate code with
4206 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4209 - Simplify transport plugin (internal) interface by not passing any
4210 argument to _server and _client, but embedding them in the socket
4213 2004-09-28 Martin Quinson
4214 - Finish the port to AIX.
4215 autoconf was my problem (segfault within the malloc replacement
4216 function. No idea why)
4218 2004-09-16 Martin Quinson
4219 - Fix some size_t madness on 64bit architectures.
4221 2004-09-08 Martin Quinson
4222 - Reduce the number of system headers loaded, overload some more system
4223 calls (such as malloc to cast the result of the system one, and work
4225 - Fix and reintroduce the config support
4227 2004-09-07 Martin Quinson
4228 - Source code reorganization to allow Arnaud to surf all over there.
4229 - Allow to document the logging categories.
4230 - Remove all uppercase from logging categories and useless cleanup in names.
4232 2004-08-18 Martin Quinson
4233 Version 0.6.2 (protocol not changed; API changed)
4234 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4235 gras_error_t), and NULL when not found. Functions expecting a msgtype
4236 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4237 hopefully usefull message.
4238 - Portability to prehistoric sparcs again
4240 2004-08-17 Martin Quinson
4241 Version 0.6.1 (protocol not changed; ABI not changed)
4242 - prealloc some buffers to speed things up
4244 2004-08-11 Martin Quinson
4245 Version 0.6 (protocol not changed; ABI expended)
4246 - The parsing macro can deal with the references, provided that you add
4247 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4249 2004-08-09 Martin Quinson
4250 Version 0.5 (protocol not changed; ABI changed)
4251 - Allow to off turn the cycle detection code in data exchange at
4252 compilation time. It should be at run time, but I'm short of time (and
4253 the config stuff is still broken). That way, we keep dict out of the
4254 critical path, which is good because the performance is poor:
4255 - search not dichotomial yet
4256 - dynar give no way to access their content and memcpy everytime
4257 - In composed data description (struct, ref and so on), stop foolness of
4258 keeping the subtype's ID, but store the type itself. This keeps sets out
4259 of the critical path, which is good since they rely on dynar and
4260 dictionaries. The only loose of that is that we cannot detect the
4261 redeclaration of a structure/union with another content (but I'm not sure
4262 the code detected well this error before anyway). We still can detect
4263 the redefinition discrepancy for the other types.
4264 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4265 it breaks the code because of type-punning used all over the place).
4266 This breaks on all non-gcc architectures (for now).
4268 All those changes (plus the buffer of last time) allow me to gain 2 order
4269 of magnitude on cruel tests consisting of 800000 array of integers on two
4270 level of a hierarchical structure (200 secondes -> 4 secondes)
4273 - the selector of reference must now return the type it points to, not
4274 the ID of this type.
4276 2004-08-06 Martin Quinson
4277 Version 0.4 (protocol changed; ABI not changed)
4278 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4279 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4280 - Implement a buffer transport, and use it by default (it relies on tcp in
4281 real life and on sg in simulation).
4282 That's a bit hackish since I had a new field to the structure to store
4283 its data without interfering with the subtype ones. Inheritance
4284 is tricky in C. And that's a kind of reverse inheritance with one class
4285 derivating two classes. Or maybe a game with java interfaces. Anyway,
4286 that's damn hard in C (at least).
4287 Moreover, I got tired while trying to ensure plugin separation and
4288 genericity in SG mode. MSG wants me to do weird things, so let's go for
4289 cruel hacks (temporarily of course ;).
4290 See comment in transport_private.h:71
4291 - do not include all the _interface headers in private but in the files
4292 which really need them (to cut the compilation time when they are
4295 2004-07-26 Martin Quinson
4296 Version 0.3 (protocol not changed; ABI changed)
4297 - Major overhault of the datadesc interface to simplify it:
4298 - shorted the function names:
4299 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4300 - add a trivial way to push/pop integers into the cbps without malloc.
4301 This allows to make really generic sub_type description, which simply
4302 pop their size of the stack.
4303 - add a function gras_datadesc_ref_pop_arr() which does what users want
4304 most of the time: Declare a dynamic array (which pops its size of the
4305 stack) and declare a reference to it. Poor name, but anyway.
4306 - kill the post-send callback, add a post-receive one
4308 2004-07-23 Martin Quinson
4309 Version 0.2 (protocol changed; ABI changed)
4310 - add some testing for cpbs in the test cases, and fix some more bugs.
4311 This invalidate again the little64 data file, since I cannot regenerate
4313 - remove an awfull optimization in the logging stuff, allowing me to:
4314 - understand it again
4315 - learn gcc how to check that the argument match the provided format
4316 - fix all errors revealed by gcc after that
4317 - internal keys of dict are not \0 terminated. Deal with it properly in
4318 loggings instead of segfaulting when the user want to see the logs :-/
4320 2004-07-22 Martin Quinson
4321 - Fix some stupid bug preventing cbps (callback postit) from working
4323 2004-07-21 Martin Quinson
4324 - Some documentation cleanups
4325 - remove the useless last argument of msgtype_declare
4326 - rename the Virtu functions to fit into the 'os' namespace
4327 - move headers src/include -> src/include/gras/ and stop fooling with
4328 gras -> . symbolic link
4329 - make distcheck is now successful
4331 2004-07-19 Martin Quinson
4333 - Build shared library also
4334 - Install html doc to the right location
4335 - stop removing maintainer files in make clean
4336 - build tests only on make check
4338 2004-07-13 Martin Quinson
4340 - No major issue in previous version => change versioning schema
4341 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4342 corresponding dataset.
4344 2004-07-11 Martin Quinson
4346 - Get it working with any kind of structure (we can compute the padding
4347 bytes remotely for all the architectures I have access to)
4348 - Implement the structure parsing macro (still not quite robust/complete)
4349 - Improvement to the remote testing toysuite
4351 2004-07-10 Martin Quinson
4352 [autoconf mechanism]
4353 - get ride of a bunch of deprecated macros
4354 - actually run the test for two-compliment, not only compile it :-/
4355 - test whether the structures get packed (and bail out if yes. Damn.
4356 Alignment is a serious matter)
4357 - test whether the structures get compacted (but respecting the alignment
4358 constraints of each types)
4359 - test whether the array fields of structures can straddle alignment boundaries
4361 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4362 (took me 2 hours to find that bug, looking at the wrong place)
4363 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4364 before sending/receiving and are used to compute the offsets of fields
4365 - Given that padding size depend even on compiler options, keep track of
4366 alignment and aligned_size only for the current architecture. Not a big
4367 deal since we send structure fields one after the other (seems
4369 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4370 program, let it work on linux/gcc/little32. portability todo.
4372 2004-07-08 Martin Quinson
4373 - import and improve remote compilation support from FAST
4374 - make sure make check works on half a dozen of machines out there
4376 2004-07-07 Martin Quinson
4377 Let's say it's version 0.0.3 ;)
4378 - Implement conversions (yuhu!)
4379 - Let it work on solaris (beside conversion, of course)
4380 - Stupid me, using rand() to generate the conversion datatests in not wise.
4382 2004-07-06 Martin Quinson
4383 - Let make dist work, since I'm gonna need it to compile on remote hosts
4384 - Let Tests/datadesc_usage write the architecture on which the file was
4385 generated as first byte.
4386 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4387 architecture descriptions.
4388 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4389 on those architectures.
4390 - Optimization: send/recv array of scalar in one shoot
4392 2004-07-05 Martin Quinson
4393 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4395 - Plug a whole bunch of memleaks
4396 - each process now have to call gras_{init,exit}. One day, their log
4397 settings will be separated
4398 - Continue the code factorisation between SG, RL and common in Transport.
4400 2004-07-04 Martin Quinson
4402 - Redistribution between SG and RL.
4403 We wanna have to accept in SG, so move accepted related parts of RL in
4404 the common part. (more precisely, the dynar of all known sockets is no
4405 more a static in transport.c, but part of the process_data)
4407 [gras_stub_generator]
4408 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4410 2004-07-03 Martin Quinson
4411 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4412 them all up in one shot)
4414 - Fix the ugly case of reference to dynamic array.
4415 - New (semi-public) function gras_datadesc_size to allow the messaging
4416 layer to malloc the needed space for the buffer.
4418 - gras_socket_close now expect the socket to close (and not its address to
4419 put NULL in it after it). This is because the socket passed to handlers
4420 is one of their argument (=> not writable).
4422 - propagate the interface cleanup from last week in datadesc, ie remove a
4423 superfluous level of indirection. User pass adress of variable
4424 containing data (both when sending and receiving), and not of a variable
4425 being a pointer to the data. Let's say that I like it better ;)
4426 The price for that is constructs like "int msg=*(int*)payload" in
4427 handlers, but it's a fine price, IMHO.
4429 - Let it work in RL (yuhu)
4431 2004-06-21 Martin Quinson
4433 - porting SG plugin and SG select to new standards (works almost).
4434 - plug memleaks and fix bugs around.
4437 - cleanup the prototype of data recv and force users to specify when they
4438 want to handle references to objects. Test case working even for cycles.
4439 - plug memleaks. Valgrind is perfectly ok with this.
4441 2004-06-12 Martin Quinson
4443 - cleanup the separation between plugin and main code in plugin creation
4445 2004-06-11 Martin Quinson
4447 - Reput hook for raw sockets, needed for BW experiments
4448 - kill a few lines of dead code
4449 [Data description] Interface cleanup
4450 - gras_datadesc_by_name returns the searched type or NULL.
4451 That way, no variable is needed to use a type desc once, which makes
4453 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4454 two parameters were strings (field name, type name), leading to
4456 [Dicos] Interface cleanup
4457 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4458 This is consistant with the dynar API.
4460 2004-04-21 Martin Quinson
4462 - Porting to new standards.
4464 - interface cleanup.
4465 There is no bag anymore, no need to take extra provision to mask the
4466 pointers behind "ID".
4467 Better splitup of functions between files create/exchange/convert.
4468 This is still a bit artificial since convert and receive are so
4469 interleaved, but anyway.
4471 - add a queued message list to procdata (the ones not matching criteria
4473 - factorize some more code between SG and RL wrt procdata
4475 - use gras_exit in example to track memleaks
4476 - get rid of gs_example now that GS is properly integrated into gras
4477 - update run_test to integrate the latest tests (datadesc)
4479 - rename WARNINGn macros to WARNn since it prooved error-prone
4481 2004-04-19 Martin Quinson
4483 - register init/exit functions within gras module mechanism
4484 - send/receive function.
4485 Convertion is not implemented, but short-cutted if not needed.
4486 struct/array elements are sent one by one (instead of block-wise), but
4487 nobody really cares (yet). Get a prototype before optimizing.
4488 - tests (using a file socket) for DD send/receive on:
4489 - base types: int, float
4490 - array: fixed size, string (ie ref to dynamic string)
4491 - structure: homogeneous, heterogeneous
4492 - chained list, graph with cycle
4493 Believe it or not, valgrind is not too unhappy with the results. The
4494 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4498 - Bugfix when using a filename explicitly (instead of '-')
4500 2004-04-09 Martin Quinson
4502 - factorize more code between RL and SG in socket creation
4503 - Complete the implementation and tests of:
4505 o file (only in RL, and mainly for debugging)
4507 I lost 3 days to design a portable address resolver, and then decided
4508 that the prototype mainly have to run on my box.
4509 Addressing portability too early may be like optimizing too early :-/
4511 - use gras_init in the Tests instead of the crappy parse_log_opt
4512 (the latter function is removed)
4513 [Conditional execution]
4514 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4515 [Code reorganisation]
4516 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4517 Build examples against the RL library, since there is no way to disable
4518 its creation for now.
4520 For information, the beginning of coding on GRAS was back in june
4521 2003. I guess that every line has been rewritten at least twice since