1 ----------------------------------------------------------------------------
3 SimGrid (3.27.1) NOT RELEASED YET (v3.28 expected June 21. 2021, 03:32 UTC)
6 - New plugin: Producer-Consumer with monitor. Just requires to include the
7 include/simgrid/plugins/ProducerConsumer.hpp header to be used. See the
8 associated example (examples/cpp/plugin-prodcons).
11 - Fixed a bug where Activity::wait_for() killed the activity on timeout.
12 Explicitly cancel the activity to get back to previous behavior.
15 - The default SMPI compiler flags are no more taken from the environment.
16 They can be explicitly set through cmake parameters SMPI_C_FLAGS,
17 SMPI_CXX_FLAGS, or SMPI_Fortran_FLAGS.
18 - New option: --cfg=smpi/finalization-barrier, which can be used to add
19 a barrier inside MPI_Finalize. This can help for some codes which cleanup
20 data attached to a process, but still used in other SMPI processes.
22 - Lua platform files are deprecated. Their support will be dropped after v3.31.
25 - Legacy functions deprecated in this release: SIMIX_get_clock(), SIMIX_run().
28 - xbt_assert is not disabled anymore, even when built with enable_debug=off.
31 - New tutorial: Model-checking and formal assessment
33 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
34 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
35 - FG#47: Complete and fix tests from teshuite/s4u/activity-lifecycle
36 - FG#64: Configuring smpi/IB-penalty-factors
37 - FG#67: Running computation concurrently with MPI_Iallreduce
38 - FG#69: Tutorial misleading users of pre-v3.26 versions of SimGrid
39 - FG#71: Segmentation fault on invalid gw_src/gw_dst
40 - GH#322: Issue when an actor kills his host vm
42 ----------------------------------------------------------------------------
44 SimGrid (3.27) March 29. 2021
46 The Ragnar Release (anniversary of the siege of Paris by Vikings in 845)
48 Main user-visible changes:
49 - All functions marked as deprecated for one year were removed.
50 - Removed all that was related to the "Storage" type in all APIs and bindings.
51 "Disks" have been introduced more than a year ago to replace them for the better.
52 It's time to finish this replacement.
53 - SMPI ABI breaks from previous version: MPI global handles (predefined datatypes,
54 ops, COMM_WORLD) are now defined differently, and can be used in initializers,
55 avoiding "initializer element is not constant" errors with some codes. Please be
56 careful to recompile your MPI codes if switching to this new SimGrid version.
57 - Network model ns-3 now requires version ns-3.28 or higher.
58 - Removed some option aliases that were deprecated since v3.23 (June 2019)
59 - 'verbose-exit' is now 'debug/verbose-exit'
60 - 'simix/breakpoint' is now 'debug/breakpoint'
61 - 'clean-atexit' is now 'debug/clean-atexit'
64 - New functions: Comm::sendto_{init,async} to initiate a communication
65 on between two (possibly remote) hosts.
66 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
67 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
68 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
69 or Mailbox::get_async<void>() if you really want to play with void*.
70 - Unify the interface of Activity::on_{start/activity}
71 - New function: Comm::get_dst_data()
72 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
75 - ptask should now work better with multicore hosts (FG#37).
76 - Internal cleanups and refactoring around activities, to prepare for SimDAG++.
77 - Internal cleanups and refactoring around models, to prepare for programmatic
79 - Modernize our tests of the performance models
83 - New section: Configuring the logs from the command line.
84 - New section: Programming API of logs.
85 - New section: Getting realistic results
86 - New section: Examples of SMPI Usage
87 - New part: SimGrid's Internals, with a section on Design goals
88 - Move back to the classical Breathe tool chain, and fix many small glitches.
91 - Handle leak checking: SMPI will now display a message if MPI handles are still
92 present at the end of a simulation. Option smpi/list-leaks:n will display the n
93 first leaked pointers and their types.
94 - smpi/display-allocs option : Display Memory allocations analysis, with hints
95 about the largest allocations. This only works for allocations with malloc/calloc
96 and not compiled with SMPI_NO_OVERRIDE_MALLOC
97 - "-analyze" option in smpirun will activate smpi/display-allocs and smpi/display-timing
98 - New implemented MPI calls: MPI_File_set_size, MPI_Aint_add, MPI_Aint_diff,
99 MPI_Type_get_extent_x, MPI_Type_get_true_extent_x, MPI_Status_set_elements_x
100 - Support for MPI C++ datatypes (MPI_CXX_BOOL, MPI_CXX_FLOAT_COMPLEX,
101 MPI_CXX_DOUBLE_COMPLEX, MPI_CXX_LONG_DOUBLE_COMPLEX)
102 - Replay now opens and parses the trace file only once for all ranks when possible.
104 - better handling of MPI_PROC_NULL
105 - MPI_Group_incl was broken in some cases
106 - PAPI tracing now fixed
107 - MPI_Errhandlers are properly handled for MPI_COMM_WORLD
108 - SMPI will now fail properly when receive size < sent size, instead of
109 just setting MPI_ERR_TRUNCATE and keep going.
110 - Proxy Apps ( https://framagit.org/simgrid/SMPI-proxy-apps ):
111 - Remove useless build patches and split build/run patches on the page
112 - Build in parallel when possible
113 - Test on yum-based systems
114 - New nightly tested Proxy-apps:
115 - OpenMPI mpi-test-suite
116 - Coral MCB, LSMS, Qbox, CLOMP
117 - ECP/Mantevo Cloverleaf3D, Ember
118 - Codevault bhtree_mpi
121 - Only initialize the Dwarf parser when needed, saving 10 seconds in most cases.
122 - Start to overhaul the implementation to modernize it (still ongoing).
125 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
126 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
127 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
128 - Deprecate xbt_dynar_to_array().
129 - When an exception occurs in kernel mode, display both the actor and kernel stacks.
131 C binding and interface:
132 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
133 - Constness requirement has been lowered for parameter argv of sg_actor_start()
134 and sg_actor_create() It's now (char* const*). The previous version is still
135 available as sg_actor_start_() and sg_actor_create_().
137 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
138 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
139 - FG#37: Parallel tasks are limited to 1 core per host
140 - FG#62: Running "smpirun -replay" on large networks
141 - FG!46: Fix a few potential memory leaks in SMPI colls
142 - GH#343: Wrong consumed energy when an exec timeouts
144 ----------------------------------------------------------------------------
146 SimGrid (3.26) December 15. 2020
148 The Release release (the French lockdown was eased today).
150 Important user-visible changes:
151 - SimGrid now requires a compiler with C++14 support.
152 Sibling projects should upgrade their FindSimgrid.cmake
153 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
154 several users had difficulties to understand issues when using high bandwidth or
155 small latency events. The new value was already the default for SMPI and
156 should not cause too much performance hit. It can change some simulation
157 timings if simulation was skipping a lot of small events (which was actually
158 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
159 - On some network models (L07, CM02, LV08), an implicit loopback link is created
160 for local communication. Default bandwidth was 500MBps and latency was 15µs,
161 which was rather bad. Default values have been changed to 10GBps and 0 latency.
162 Options network/loopback-lat and network/loopback-bw have been added to adapt
163 these values if needed. Network and routing models already providing separate
164 loopbacks (clusters and such) are not impacted by this change.
165 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
166 metadata that are added to each MPI message by MPI libraries (tag,
167 src/dest, and communicator information). This may change simulation times
168 for very small message sizes, but should be more realistic.
171 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
172 - New plugins: wifi_energy and link_load. See the associated examples.
173 - New examples: wifi networks, and communication suspend.
176 - LTO is now enabled for Intel/clang compilers.
177 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
178 Setting it to "auto" will use all available cores, while setting it to n will
179 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
180 - Remove obsolete runtime option 'contexts/parallel-threshold'.
181 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
182 has been kept for compatibility.
183 - Finally remove obsolete snake_case() aliases for runtime options.
184 - Further improve the documentation.
187 - the on_exit() of each actor is also executed when the simulation deadlocks.
188 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
189 An example is provided for s4u::Comm but it should work with Exec and Io.
192 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
193 https://framagit.org/simgrid/SMPI-proxy-apps
194 - MPI/IO: supports names starting with "./"
195 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
197 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
198 building shared libraries, this may cause issues, so environment variable
199 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
200 - most temporary files should now be created in /tmp dir (or equivalent).
201 If this one does not allow execution of code (noexec flag), this may cause issues.
202 Please use another tmp directory (using TMPDIR or equivalent system variable)
206 - A large refactoring is ongoing, but this should be transparent for users so far.
209 - Drop xbt_str_split().
211 C binding and interface:
212 - The return type of the sg_actor_on_exit() callbacks is now 'void'
213 (int value was previously ignored)
214 - Many C functions were renamed to follow a common naming style
215 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
216 names are kept until version 3.30.
217 - Many MSG tests were converted to the new S4U's interface in C, that
218 was extended for that.
220 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
221 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
222 - FG#41: Add sg_actor_create C interface
223 - FG#43: xbt::random needs some care
224 - FG#48: The Impossible Did Happen (yet again)
225 - FG#50: Suspending an actor executed at the current timestamp fails
226 - FG#52: Zero-seconds timeout : "That's in the past already"
227 - FG#53: Crash while using ns-3 network model
228 - FG#54: How to suspend a comm?
229 - FG!22: Stochastic Profiles
230 - FG!24: Documentation and fix for xbt/random
231 - FG!35: Add a modeling hint for parallel links in doc
232 - FG!36: [xbt/random] Read/Write the state of the RNG
233 - FG!42: Link cumulated load plugin
234 - FG!44: Test condition variable wait_until+notify_all
235 - GF#18137: Allow different stack sizes?
236 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
237 - GH#139: Allow pthread creation in SMPI
238 - GH#336: Packet-level simulation using SMPI?
239 - GH#345: Error in the LMM while migrating a VM
240 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
241 - GH#352: pip install / python setup.py install fail to find pybind11
242 - GH!337: Fix link_energy plugin for wifi platforms
243 - GH!339: Add Mailbox set_receiver method to python binding
244 - GH!344: Cast hugepages macros parameters to int64
246 ----------------------------------------------------------------------------
248 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
250 The Palindrom Day Release.
252 Important user-visible changes:
253 - Improve the Python usability (stability and documentation).
254 - A nasty synchronization bug was ironed out, see also below.
255 - Python's doc was organized as a separate tree, now integrated with C++.
256 - C bindings of S4U were not part of the doc.
257 - The C++ doc was also improved as methods are now split by theme.
258 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
259 - This is mandatory to use the Java bindings.
260 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
261 - SimDAG++: Automatic dependencies on S4U activities (experimental)
262 - Some features are already implemented but not all of them
263 - Cannot block an activity until it's scheduled on a resource
264 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
265 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
268 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
269 - Drop support for Viva/Triva (old visualization tools), hence removed
270 TRACE_get_node_types() and TRACE_get_edge_types() functions.
273 - Require pybind11 version 2.4.
274 - Greatly improve locking during context switches with Python.
275 - Declaring Python callbacks to this_actor::on_exit is now working.
276 - No more tests are failing currently, even on win32.
279 - Actor: Merge signals on_migration_start/end into on_host_change
280 - Actor: Rename migrate() into set_host()
281 - Disk: Allow users to get the read and write nominal bandwidth values
282 - Exec: Implement wait_for(timeout)
283 - Io: Implement wait_for(timeout)
286 - Parse errors now raise a simgrid::ParseError that you may want to catch.
289 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
290 been renamed to fix a spelling error. As usual, the previous definitions are
291 kept with a deprecation warning. However, to avoid runtime errors, it is now
292 forbidden to override these deprecated methods in a derived class. Please use
293 the new names immediately if you need to override them.
296 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
297 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
298 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
299 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
300 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
301 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
302 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
303 - New module: random, providing classical random numbers generators.
306 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
307 allocations' size above which they are considered "shared" by default (as if
308 they were performed through SMPI_SHARED_MALLOC macros).
309 Default = 0 = disabled feature.
310 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
311 This can cause some troubles if codes are already overriding these. If this
312 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
313 help, but will make this feature unusable.
314 - Input check is now performed much more consistently and outputs precise
315 warnings for each error found
316 - New supported calls : MPI_File_set_view, MPI_File_get_view
317 - MPI I/O now only supports disks, as storage is discontinued.
318 - MPI I/O allows opening files without specifying absolute path
319 - Note : SMPI configuration options are now initialized by smpi_init_options()
320 call, which should be called if SMPI is not being used through smpirun.
322 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
323 - FG#9: Python bindings crashing
324 - FG#39: Missing s4u::Comm::wait_any_for example
325 - FG#42: Add support for ThreadSanitizer (TSan)
326 - FG!19: Removing RngStream
327 - FG!20: A module for RNG calls
328 - FG!21: Choice between ad-hoc and standard distributions implementations
330 - FG!25: Fix link in Mutex doc
331 - FG!26: Fix links in SMPI interface doc
332 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
333 - GH#305: Unscheduled tasks are still excuted
334 - GH#313: smpirun: manual outdated w.r.t. --help
335 - GH#321: [S4U] Get task remaining work ratio
336 - GH#323: Crash when an actor turn off his physical host
337 - GH#335: Missing links on dragonfly example svg
338 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
339 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
341 ----------------------------------------------------------------------------
343 SimGrid (3.24) October 9. 2019
345 The Clean Disk Release.
348 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
349 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
350 thus becomes the default resource to execute I/O operations. Examples
351 and the file system plugin have been modified to use disks instead of
352 storage elements. Storage can still be used but is doomed to disappear.
353 Users are thus adviced to update their simulators accordingly.
354 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
355 for consistency with pthread_barrier_wait()
356 - Host::get_englobing_zone() returns the englobing netzone
357 - Actor::on_destruction is now called in the destructor
358 Actor::on_termination new signal called when the actor terminates
360 - Global signals are now part of the Engine:
361 - on_platform_creation: after config settings, before the XML parsing
362 - on_platform_created: right after the XML parsing
363 - on_time_advance: each time the clock advances
364 - on_simulation_end: after simulation, before cleanups
365 - on_deadlock: as the name implies.
367 - sg_{actor,host,link}_{data,data_set}() now all exist.
368 Use them to attach user data to the object and retrieve it.
371 - Introduce an experimental Wifi model. It sounds reasonable
372 according to the state of the art, but it still has to be properly
373 validated, at least against ns-3 if not against reality.
374 - Improved the usability of ns-3. Several bugs were ironed out.
375 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
376 It is now expressed as 'idle:epsilon:allCores' to properly model the
377 consumption of non-whole tasks on mono-core hosts. More info in the doc.
380 - convert a new set of functions to the S4U C interface and move the old MSG
381 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
384 - Fortran bindings for DVFS have been removed.
385 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
386 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
387 - SMPI can now be selected by cmake's find_module(MPI) with
388 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
389 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
390 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
391 - trace-call-location can be used with TI traces, and replayed, and is compatible
392 with smpi/comp-adjustment-file.
393 - sleep events are now correctly traced/replayed.
394 - Default for trace-call-location is now to use file names and not full paths.
395 To revert to previous behavior (in case of collision of filenames), option
396 "smpi/trace-call-use-absolute-path" can be set to yes.
399 - Use the included xxHash as an hash implem when C++14 is usable.
400 - Option model-checker/hash was removed. This is always activated now.
401 - New option smpi/buffering controls the MPI buffering in MC mode.
402 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
403 This is useful to check for MPI compliance.
406 - New section on plugins: how to define a new one, and existing ones.
409 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
410 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
413 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
414 and <mount> tags. The rationale is finish to decouple what should be handle
415 by SimGrid kernel (i.e., resources and activities) from plugins built in
416 user space (e.g., the file system plugin). This new tag currently accepts
417 only three attributes (id, read_bw, and write_bw). All the other information
418 that was declared with the storage related tags now has to be expressed as
419 properties. An example of platform using this new tag is available at
420 examples/platforms/hosts_with_disks.xml
423 - 'expect signal' can now accept more than one potential signal.
425 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
426 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
427 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
428 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
429 - FG#31: per-actor data
430 - FG#34: SG_BARRIER_SERIAL_THREAD?
431 - FG#35: model-checker does not like buster-produced binaries
432 - FG!13: MC: complete workaround in the error msg seen on modern systems
433 - FG!15: execute_flops now logs compute
434 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
435 - FG!17: ns-3: unblock the right number of communications + others issues
436 - FG!18: Improving the performance of the ns-3 bindings
437 - GH#189: Energy consumption of parallel tasks
438 - GH#219: Error in the throughput of TCP transfer
439 - GH!330: Minor polishing of English: Configuring_Simgrid
440 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
441 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
442 - GH!334: Rework energy plugin, again
444 ----------------------------------------------------------------------------
446 SimGrid (3.23.2) July 8. 2019
449 - Nicer introduction page.
450 - Migrate the "Deploy your application" page to the new doc.
451 - Move Java as a subtree of MSG.
454 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
457 - Drop xbt_dynar_sort_strings().
460 - Really fix FG#26: Turning off a link should raise NetworkFailureException
461 - FG#27: Wrong exception thrown to wait_any when link is turned off
462 - GH#328: Java: Canceling multiple tasks in a single vm/host
464 ----------------------------------------------------------------------------
466 SimGrid (3.23) June 25. 2019
468 The Exotic Solstice Release.
471 - SunOS and Haiku OS support. Because exotic platforms are fun.
472 - Stop setting random seed with srand48() at initialization.
473 - Use addr2line as a fallback for stacktraces when backtrace is not available.
474 - Build option -Denable_documentation is now OFF by default.
475 - Network model 'NS3' was renamed into 'ns-3'.
478 - Simgrid can now hopefully be installed with pip.
481 - wait_any can now be used for asynchronous executions too.
484 - New log appenders: stdout and stderr. Use stdout for xbt_help.
485 - Drop xbt_dict_dump.
488 - 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
489 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
490 - MPI interface is now const correct for input parameters
491 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
492 - Fortran bindings for async collectives.
493 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
496 - Remove option 'model-check/record': Paths are recorded in any cases now.
497 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
498 always sparse. That's an immense gain in memory, and can even be
499 faster because of cache effects. So there is no need to clutter the
500 code to allow the user to go for the unefficient mode.
503 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
504 models proved to be more accurate than these old experiments.
506 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
507 - FG#1: Broken link in error messages
508 - FG#2: missing installation documentation
509 - FG#3: missing documentation in smpirun
510 - FG#6: Python bindings not available on PyPI
511 - FG#7: simple cmake call requires doxygen
512 - FG#8: make python bindings an optional dependency
513 - FG#10: Can not use MSG_process_set_data from SMPI any more
514 - FG#11: Auto-restart actors forget their on_exit behavior
515 - FG#12: -Denable_lto=OFF doesn't disable LTO
516 - FG#13: Installs unstripped file 'bin/graphicator'
517 - FG#14: Installs the empty directory 'doc/simgrid/html'
518 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
519 - FG#17: Dead link in doc (pls_ns3)
520 - FG#20: 'tesh --help' should return 0
521 - FG#21: Documentation link on http://simgrid.org/ broken
522 - FG#22: Debian installation instruction are broken
523 - FG#26: Turning off a link should raise NetworkFailureException exceptions
524 - GH#133: Java: a process can run on a VM even if its host is off
525 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
526 - GH#326: Valgrind-detected error for join() when energy plugin is activated
528 ----------------------------------------------------------------------------
530 SimGrid (3.22) April 1. 2019
532 The Easter Christmas Release (this one is somewhat late).
535 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
536 - Not complete yet: asynchronous activities (amongst others) are still missing
537 - Still ongoing: the interface may change in the future. We need more
538 testers! Please report any glitches.
539 - No new project using Java should start now. Please switch to Python.
542 - Some of the internal cleanups may lead to speed improvements:
543 - The hard limitation on the amount of simulated actors with
544 Java+Mac was removed. Now, the available memory is the only limit.
545 - Our refcounting was tidyied, leading to 10% speedups in some cases.
546 - We are still working on making our code robust to the actor kills
547 and hosts' churn. Things are improving, but it's not perfect yet.
548 - Replaced our own code to display a backtrace (that was forking addr2line)
549 with the Boost.Stacktrace library.
550 You won't see your backtraces without this optional dependency.
551 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
552 - Stop setting random seed with srand() at initialization.
555 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
556 XML file version remains 4.2 since old files are still compatible.
559 - Process termination which was broken at version 3.21 has been repaired.
560 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
561 - Hide the examples into examples/deprecated. New users should use Python.
564 - Drop MSG_process_create_from_stdfunc() from the API.
565 This C++ function was a pimple in the C API, made necessary at some
566 point by the Java bindings. This is fixed now.
567 - Hide the examples into examples/deprecated. New users should use S4U.
568 - MSG_process_create and MSG_process_attach now crash if the host used
569 to run this process is off.
570 - Fix the protype of MSG_process_on_exit()
571 Now use (int,void*) callbacks instead of (void*,void*) ones.
572 The implementation was ways too messy to actually work, I guess.
575 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
576 too often and provide more accurate timings.
577 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
578 - MPI_Alltoallw support
579 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
580 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
581 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
582 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
584 - Basic implementation of generalized requests (SMPI doesn't
585 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
588 - Drop sg_cmdline. Please use xbt_cmdline instead.
589 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
590 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
591 OsSemaphore is implemented in a portable way with C++11 threads.
592 This should allow much more threads to be created at the same time,
593 allowing Mac and Java users to simulate many more actors.
594 - Implement the 'thread' factory with std::thread instead of xbt ones.
595 It is not possible to set the stack size with threads anymore, but
596 -fsplit-stack is the way to go nowadays when using threads.
597 - Drop the xbt_os_thread_t module (now unused)
598 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
599 - Drop xbt_str_join_array().
600 - Drop cunit, use Catch2 instead.
603 - Many cleanups in the kernel::activity namespace. This was long
604 overdue, and shall open the path to many future endeavors.
607 - #132: Java: a process can not shut down its own host
608 - #220: S4U: detached send causes exception if sender terminates soon after sending
609 - #261: Document the parameters of parallel execution's constructor
610 - #300: [s4u] BarrierPtr is missing
611 - #314: SMPI args internal cleanup
612 - #316: Fix a bug related to the CPU utilization of multi-core VM
613 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
614 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
615 - #325: Turning off a host has different behavior on sleeping actors and computing actors
617 ----------------------------------------------------------------------------
619 SimGrid (3.21) October 3. 2018
621 The Restarting Documentation (TRD) Release.
624 - Convert the user manual to Sphinx for improved usability.
625 Unstable version now hosted on https://simgrid.frama.io/simgrid
626 Many glitches remain; stay tuned.
627 - Restrict the Doxygen reference API to the installed header files.
628 (documenting all internals this way is overkill + doxygen scales badly)
629 - New tutorial on S4U (highly inspired from the old MSG one, but with
630 a git repository to fork as a starting point, and a docker image)
631 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
632 - The Developper manual is still to be converted (not compiled until then)
633 - Some parts are still missing in this conversion (such as the
634 platform and deployment sections) while others were blindly converted
635 and would need more love (such as the configuration flags).
636 Things will certainly further improve in the future releases.
639 - Fully reimplement auto_restart mechanism that was utterly broken
640 (fix #22, #131 and #281 that were all very related)
641 - Implement semaphores (fix #303 and #291)
642 - When creating an actor from a function and its parameters,
643 move-only parameters are not allowed anymore, as it would prevent
644 the actor to be restartable if its parameters are consumed on run.
645 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
646 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
647 IO operations that can be started, waited for, or canceled as a regular
649 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
653 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
654 - New host variable: 'core_count'
657 - Remove xbt_os_thread_specific features
658 - Remove portability wrapper to condition variables
659 - Remove xbt_os_thread_yield()
662 - MPICH collective selector now mimics MPICH 3.3b
663 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
664 - MPI_Init(NULL,NULL) is now allowed
665 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
666 - Command-line arguments should now be handled properly in Fortran simulations
667 - MPI Attributes and related callbacks should work in Fortran
668 - Apps using C + Fortran should now work
669 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
670 - Support MPI_Error_String
673 - Due to an internal bug, Msg.run() must now be your last line.
674 We hope to fix it in a future release, and we are sorry for the inconvenience.
677 - #22: Process autorestart seem to only work with CAS01 cpus
678 - #56: Feature request: dynamic SMPI replay
679 - #93: simgrid should not eat --help
680 - #111: How to change the loopback link characteristics?
681 - #116: Platform cluster radical
682 - #117: Platform cluster/host order
683 - #129: Replay_multiple on steroid
684 - #131: Java : autostart does not work
685 - #143: Setting a breakpoint at a given time
686 - #235: Network energy models should be integrated
687 - #264: Add ptask L07 resource tracing
688 - #271: Dynamic and manual replay of SMPI traces
689 - #279: Breakpoints option uses time deltas instead of absolute time
690 - #281: Daemonized actor no longer daemonized after an auto-restart
691 - #282: TIT convertor erase traces when using absolute path in the trace list file
692 - #285: segfault when a s4u actor kills itself with msg process tracing activated
693 - #286: Pajé traces are not exposing the number of cores
694 - #287: Command-line parsing should stop when encountering '--'
695 - #288: MPI_Init(NULL, NULL)
696 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
697 - #290: Method to check if a mailbox has a message ready to be consumed
698 - #291: [s4u] semaphores do not seem available yet
699 - #294: Alltoall collective implementation of mpich
700 - #295: Starting a Host as OFF is not supported
701 - #296: DTD too permissive
702 - #299: Add a s4u barrier example
704 ----------------------------------------------------------------------------
706 SimGrid (3.20) Released June 24. 2018
708 The proxy snake_case() release.
711 - The whole API is now using snake_case() instead of mixing it with camelCase.
712 Compatibility wrappers in place for one year.
713 - We took the opportunity to increase the consistency of the API.
714 Please report any remaining glitches.
715 - New function simgrid::s4u::Host::get_actor_count:
716 Returns the number of actors running on a specific host.
719 - SMPI is now tested with ~45 proxy apps from various sources, with none or
720 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
721 - Replay: The replay file has been re-written in C++.
722 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
723 supported. They are stored in the trace and used when replayed.
724 - Basic support of MPI_Cancel. Robustness not guaranteed.
725 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
726 (only for MPI_COMM_TYPE_SHARED).
727 - New option: smpi/privatize-libs, to add external shared libs to be privatized
728 by SMPI. They will be copied locally and loaded separately by each process.
729 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
730 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
731 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
732 with internal index optind with multiple processes. Only works if MPI_Init has
734 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
735 syntax without preprocessing tricks.
738 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
739 behavior tracing is no longer limited to MSG
740 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
741 this means that traces now need to be updated or re-obtained; both Irecv
742 and Isend lines in your traces have now in total 5 mandatory fields:
743 <rankid> <command> <to/from rankid> <tag> <size>
744 To update your traces, it suffices to add a 0 for the tag here.
745 - TIT now also supports waiting for a distinct request via MPI_Wait.
746 Wait/Test now wait for a specific request, not just the last one that was
747 issued. This unfortunately means another update, because we need to
748 identify which request you want to wait for. We do this via the
749 triplet (sender, receiver, tag), which needs to be added:
750 <rankid> <command> <sender> <receiver> <tag>
751 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
752 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
755 - The deprecation of MSG is ongoing (but this should not impact you).
756 Many MSG functions are now simple wrappers on the C API of S4U. If
757 you wish to convert your code to S4U, find the S4U counterparts of
758 your MSG calls in src/msg/msg_legacy.cpp.
759 - MSG can still be used, but won't evolve anymore.
762 - Allow to run the Link energy plugin from the command line with
763 --cfg=plugin:link_energy
764 - Rename Energy plugin into host_energy
765 - Rename Load plugin into host_load
768 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
769 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
770 - Improve performance when handling timeouts of simix synchros.
773 - Config: the C API is now deprecated (will be removed in 3.23), and
774 the C++ API has been slightly improved.
777 - Fix several build issues on OSX.
778 - Move simgrid_config.h to simgrid/config.h (old header still working)
781 - #143: Setting a breakpoint at a given time
782 - #150: Inconsistent event names in SMPI replay
783 - #258: daemonized actors hang after all non-daemonized actors have completed
784 - #267: Linker error on unit_tmgr
785 - #269: SMPI: tracing of MPI_Wait/all/any broken
786 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
788 ----------------------------------------------------------------------------
790 SimGrid (3.19.1) Released March 22 2018
792 The Fixed ABI Release.
794 As you may know, we are currently refactoring SimGrid in deep.
795 Upcoming SimGrid4 will be really different from SimGrid3: modular,
796 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
798 Our goal is to smooth this transition, with backward compatibility
799 and automatic update paths, while still progressing toward SimGrid4.
801 SimGrid remains open during works: The last pure SimGrid3 release was
802 v3.12 while all subsequent versions are usable alpha versions of
803 SimGrid4: Existing interfaces remain unchanged, but the new S4U
804 interface is budding and the internals are deeply reorganized.
806 Since 2015, we work hard to reduce the changes to public APIs. When
807 we need to rename a public library symbol in S4U, we let your
808 compiler issue an explicative warning when you use the deprecated
809 function. These messages remain for four releases, i.e. for one full
810 year, before turning into an error. Starting with v3.15, your can
811 also adapt to API changes with the SIMGRID_VERSION macro, that is
812 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
814 Starting with this v3.19.1, our commitment to reduce the changes to
815 the public interfaces is extended from the API to the ABI: a program
816 using only MSG or SimDag and compiled against a given version of
817 simgrid can probably be used with a later version of SimGrid without
818 recompilation. We will do our best... but don't expect too much of
819 it, that's a really difficult goal during such profund refactoring.
821 The difference between v3.19 and v3.19.1 is that the former was
822 accidentally breaking the ABI of MSG, while the later is restoring
825 S4U and kernel APIs will still evolve until SimGrid4, with one-year
826 deprecation warnings as currently. In fact, cleaning up these
827 interfaces and converting them to snake_case() is one release goal of
828 v3.20. But don't worry, we are working to smooth this upgrade path.
830 In summary, new projects should start with S4U to benefit of the
831 future, but old MSG projects should still be usable with no change.
833 ----------------------------------------------------------------------------
835 SimGrid (3.19) Released March 20 2018
837 The Moscovitly-cold Spring release.
839 SMPI user-visible improvements:
840 - Support of MPI_Type_create_subarray
841 - Unimplemented MPI calls now abort (instead of issuing a warning)
842 Some harmless calls around errhandlers still only issue a warning.
843 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
844 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
845 is now possible (mmap privatization was refactored to this end).
846 Adding new SMPI instances should also be technically doable now.
848 SMPI internal cleanups:
849 - Removed index notion from SMPI -- no more "getPid() - 1"!
850 In tracing files, actors are called rank-<PROCESS-ID> but these are
851 NOT the real ranks (tracing several communicators is not supported yet).
852 - Deprecate smpi_process_index() (will be removed in 3.22)
853 - Better testing on classical proxy apps, through an external project:
854 https://github.com/simgrid/SMPI-proxy-apps/
857 - Use a graphical TOC to make it easier to find the documentation you need
858 - Revamp the MSG tutorial
861 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
862 - Remove parameter of MSG_process_killall().
863 Resetting the PID was bogus anyway (several actors could have the same PID).
866 - Execution->setHost() can be called after start(), to migrate it.
867 - Comm::test_any() is now implemented.
868 - s4u::Actor now has onCreation() and onDestruction() signals.
869 - Install some kernel header files for the users' plugins and more.
870 Warning, their API is really not stable yet.
871 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
872 include/simgrid/kernel/routing/*Zone.hpp
875 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
877 Build System and other cleanups:
878 - Remove unused run-time parameter "tracing/onelink-only".
879 - SimGrid now works with intel compilers (MC still buggy)
880 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
881 - tesh kills subprocesses on timeout (but not on windows)
884 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
887 - #194: Feature request: simgrid::s4u::Comm::test_any()
888 - #245: migrating an actor does not migrate its execution
889 - #253: Feature Request: expose clusters as objects
890 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
891 - #255: Tesh broken on Windows
892 - #256: Modernize FindSimGrid.cmake
893 - #257: Fix (ab)use of CMake install
895 ----------------------------------------------------------------------------
897 SimGrid (3.18) Released December 24 2017
899 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
901 MSG IS NOW DEPRECATED. Please use S4U for new projects.
902 - Support for MSG should not be removed from SimGrid before 2020,
903 but future evolutions will be limited to the integration of
904 user-provided patches.
907 - s4u::onDeadlock() signal, run before stopping the simulation.
908 - s4u::Exec: asynchronous executions (abstraction of background
910 - s4u::Host->getLoad() returns the achieved speed in flops/s
913 - Link::name() is deprecated and will be removed in v3.21, use
914 Link::getCname() instead.
915 - Mailbox::getName() changed to return a std::string, use
916 Mailbox::getCname() instead to get a char*.
917 - Storage::getName() changed to return a std::string, use
918 Storage::getCname() instead to get a char*.
919 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
920 to have both allocation and deallocation of the map in user space.
923 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
924 The amount of flops remaining to do is a vector, not a scalar.
925 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
926 name implies on both sequential and parallel tasks.
927 - Both changes fix GitHub's #223 using PR #237 as a basis.
928 Thanks Michael Mercier.
929 - Most examples were converted to S4U and hidden elsewhere at tests.
930 Rationale: we still want MSG to work; we want newcomers to use S4U.
933 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
934 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
937 - Switch to the faster dlopen privatization mechanism by default
938 - Documentation improvments
941 - Live migration is getting moved to a plugin. Dirty page tracking is
942 the first part of this plugin. This imply that VM migration is now
943 only possible if one this function is called:
944 - C/MSG: MSG_vm_live_migration_plugin_init()
945 - C/C++: sg_vm_live_migration_plugin_init()
946 - Java: Msg.liveMigrationInit()
947 For C and C++, "simgrid/plugins/live_migration.h" also has to be
951 - Define class simgrid::xbt::Path to manage file names.
952 - Removed unused functions:
953 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
954 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
955 xbt_graph_export_graphviz()
956 - xbt/str.h: xbt_str_join()
957 - Remove unused datatypes:
958 - xbt/heap.h: use std::priority_queue or boost::heap instead
959 - xbt/swag.h: use boost::intrusive::list instead
962 - New link_energy plugin for the consumption of the links.
963 - All of the operations on files and storage contents have been
964 packaged into a plugin (src/plugins/file_system). The current
965 public interface can be found in
966 include/simgrid/plugins/file_system.h
967 To use these functions you now have to initialize the plugin by
968 calling MSG_storage_file_system_init() just after calling
969 MSG_init() or sg_storage_file_system_init() just after creating
973 - Remove the undocumented/untested tag <include>
976 - Remove viva specific tracing as the tool is no longer maintained
979 - #248: Kill of finished processes leads to segfault
980 - #240: xbt_cond_wait_timeout should gracefully return for C apps
981 - #239: please implement signal s4u::onDeadlock()
982 - #230: segfaults when exit() without run()
983 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
984 - #223: MSG_task_get_flops_amount() not working with parallel tasks
985 - #222: Actor::kill() doesn't really kill and segfaults
986 - #221: odd LMM warning when killing an actor
987 - #120: Memory leak when the processes are forcefully killed
989 ----------------------------------------------------------------------------
991 SimGrid (3.17) Released October 8 2017
993 The Drained Leaks release: (almost) no known leaks despite the tests.
995 Even very long simulations will preserve your memory: our testsuite
996 stresses SimGrid for over 45mn on fast machines for a coverage of
997 over 80%, and there is only one single known leak, of about 4kb.
1001 - Comm.detach(): start and forget about asynchronous emission. A cleanup
1002 handler may be given to free resources if the comm cannot be completed.
1003 - this_actor::send(mailbox) is now mailbox->put()
1004 - New: simgrid::s4u::Comm::wait_all()
1005 - New: Host.execute() for remote executions.
1008 - Unused option network/sender-gap is removed.
1011 - Deprecate MSG_task_isend_with_matching(): This unused feature
1012 really complicates our internals. Will be removed in v3.20.
1015 - Improved context termination. It is now possible to free resources, even
1016 when a process is forcibly killed.
1017 - ContextBoost: add support for Boost versions above 1.61.
1020 - Bring back run-time option --cfg=exception/cutpath to remove exception
1022 - Removed unused functions:
1023 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
1024 xbt_str_rtrim(), xbt_str_trim().
1025 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
1028 - Removed header files obsolete since SimGrid 3.12:
1029 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
1031 - #3: SD_exit should be made optional
1032 - #120: Memory leak when the processes are forcefully killed
1033 - #159: Threading test regression in Actor refcounting
1034 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
1035 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
1036 - #186: Actor::killAll() segfaults if some process is blocked on wait()
1037 - #191: VM migration and pstate
1038 - #192: Updating the energy consumptions of all hosts crashes with VMs
1039 - #195: All actors have PID=0 in the logs
1040 - #204: Sometimes segfault with thread contexts and mmap privatization
1041 - #222: Actor::kill() doesn't really kill and segfaults
1042 - #225: Actor::kill() doesn't really kill when victims are doing a join()
1044 ----------------------------------------------------------------------------
1046 SimGrid (3.16) Released June 22. 2017.
1048 The Blooming Spring Release: developments are budding.
1051 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
1052 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
1053 - Java JAR file should be finally fully working on Mac OSX too.
1055 XML platforms: Switch to platform v4.1 format.
1056 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
1057 - <zone> can be used as a synonym for the now deprecated <as>
1058 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
1059 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
1060 - <actor> can be used as a synonym for the now deprecated <process>
1061 - state_file and avail_file periodicity is now easier to express
1062 (check the documentation) the old behavior should still work.
1063 * Storage: not backward compatible, but it seems that nobody ever used it
1064 - Remove attribute 'content_type' of <storage_type>: was never used
1065 - Make attribute 'model' of <storage_type> optional: for future usage
1066 - Remove Bconnection model property: was never (in)validated. Replaced by
1067 the maximum of the read and write bandwidth as a resource constraint.
1070 - New and Backwards Compatibility break:
1071 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
1072 When one wants to get the list of tasks whose states have changed during a
1073 simulation round, s/he has to allocate and free a dynar and use it as
1074 argument to this function. The former SD_simulate (double how_long)
1078 - Allow multicore VMs, with the correct sharing (unless you overcommit)
1079 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
1082 - New function to update the consumption of all hosts at once.
1083 - Fix the model for multi-core hosts, linear in the amount of busy
1084 cores with an abnormality for pIdle (see paper).
1085 BUG: Still not satisfactory for ptask on mono-cores.
1088 - The netzone are now available from the MSG API.
1089 The old names still work, but are now deprecated.
1092 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
1093 - New option: smpi/keep-temps to not cleanup temp files
1094 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
1095 mallocs mode is used, allows to change the size of the fake file used
1096 (default 1MB), to potentially limit the number of mappings for large runs.
1097 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
1098 - Fortran ifort and flang compilers support
1099 - New RMA calls supported (experimental):
1100 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1101 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1102 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1103 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1104 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1105 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1108 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1109 - Behind the scene, ActivityImpl are now automatically refcounted.
1110 Removing this nasty bottleneck will greatly simplify our internals.
1113 - Replay: New function xbt_replay_action_get():
1114 Retrieve the function previously associated to an event type.
1115 - DROPPED FUNCTION: xbt_str_varsubst()
1116 - DROPPED MODULE: strbuff. We don't need it anymore.
1117 - DROPPED MODULE: matrix. We don't need it anymore.
1118 - DROPPED MODULE: lib. We don't need it anymore.
1120 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1122 SimGrid (3.15) Released March 22. 2017
1124 The Spring Release: continuous integration servers become green
1125 We fixed even the transient bugs on all target architectures:
1126 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1127 Windows, FreeBSD, NetBSD.
1129 - Rename NetCards to NetPoints.
1130 This was intended to help NS3 users, but that's not a netcard.
1131 That's a point in the routing algorithm, let's avoid wrong simplifications.
1134 - New: MSG_process_yield(). Stop and yield to other processes.
1135 - New: MSG_process_daemon(). Daemon processes are automatically killed
1136 when the last non-daemon process terminates
1137 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1138 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1139 to make room for the upcoming network energy plugin.
1140 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1143 - Ensure that an actor can kill itself with Process::exit()
1144 - Kill the obscure NativeException. Nobody want to survive the issues
1145 it denotes, so use JniException that is a RuntimeException (not to
1146 be caught explicitly).
1147 - Partial bug fix in initialization. SimGrid flags on command line were
1148 consumed at C level but stayed in the original Java String[] args.
1149 This could mess users' args[i] if SG flags were not put at the end of
1151 The SimGrid flags are now removed from the Java arguments. However,
1152 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1153 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1154 It might be if you have --log or --cfg flags in the command line.
1155 - Fix numerous memleaks all around the place. In particular, around VMs.
1159 - s4u::onPlatformCreated: right before the simulation starts
1160 - s4u::onSimulationEnd: right after the main simulation loop
1161 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1162 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1163 event from the availability_file changes the avail speed.
1164 - Links are now usable from s4u
1165 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1166 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1167 - New examples: The conversion of MSG examples to S4U has begun
1168 - Actors: create, kill, migration, and suspend
1169 - Applications: master-worker and token-ring
1170 - Action replay: communications and storage
1171 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1174 - Backwards Compatibility breaks
1175 - The SD_route_get_list and SD_route_get_size functions have been removed.
1176 They are replaced by the unique
1177 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1178 The route from src to dst is built in the links dynar whose size can
1179 be obtained with xbt_dynar_length.
1180 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1181 sg_host_route_bandwidth and sg_host_route_latency.
1182 Macros ensure the backwards compatibility, but you should fix your code
1185 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1186 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1187 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1188 We don't support 100% of it, but it should be enough. Contact us if not.
1189 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1190 naive and probably inaccurate)
1191 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1192 mapped onto a unique small file using some system magic.
1193 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1194 - Fortran: cleanups, fixes, support of user-added operations
1195 - RMA: MPI_Accumulate are applied in correct order.
1196 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1197 - Better support for MPI_IN_PLACE
1198 - Support for MPI_Win attrs and keyvals.
1199 - Support MPI_Comm_group_create, MPI_Type_size_x
1200 - MPI ops should be properly applied to all allowed types
1201 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1202 algorithm by a non failing one from mpich.
1205 - Kill the fifo data container: we don't use it anymore.
1208 - A <cluster> can now be created with different speed values to represent pstates.
1209 see examples/platforms/energy_cluster.xml for an example.
1211 SimGrid (3.14.159) Released December 28. 2016
1213 The previous release was only a crude approximation of a Pi release;
1214 this one should be a bit better.
1216 - Revert a last minute change that broke on Mac OSX.
1217 - Fix the changelog and NEWS files.
1218 - Improve the documentation, in particular of the routing module.
1219 - Integrate some patches flying around in Debian.
1221 SimGrid (3.14) Released December 24. 2016
1224 * The whole documentation was reworked and reorganized. There is
1225 still much room for improvement here, but we're on it.
1228 * We now need python3 for our testing infrastructure.
1229 * Model checking needs libevent
1232 * Now works on FreeBSD
1235 * Privatization now works on FreeBSD
1237 * Privatization is now activated by default in smpirun
1238 You can turn it off with -no-privatize if something goes wrong.
1240 * Call-location tracing for SMPI.
1241 You can add the exact location (filename / linenumber) of an MPI call to
1242 your trace files and slow down or speed up the simulation between two
1243 consecutive calls by using an adjustment file (see the documentation).
1245 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1246 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1247 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1248 values were ignored in some cases. The timings of these functions can now
1249 be significantly different.
1251 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1252 smpi/running-power is renamed to smpi/host-speed
1254 * smpi/grow-injected-times option to enable or disable multiplication of the
1255 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1256 make simulation less precise (but also much faster).
1258 * smpirun script should be (much) faster for large deployments.
1260 * SMPI tracing: fixed issue with poor matching of send/receives.
1262 * Replay: Fix broken waitall
1264 New functions and features
1265 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1267 Dropped / renamed functions and features
1268 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1269 - MSG_mailbox_is_empty() -> Mailbox::empty()
1270 - MSG_mailbox_front() -> Mailbox::front()
1271 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1272 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1273 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1274 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1275 * Don't pass the free_f parameter to property related functions:
1276 - MSG_host_set_property_value()
1277 - MSG_as_router_set_property_value()
1278 - MSG_storage_set_property_value()
1279 * VM properties. Since msg_vm_t are msg_host_t, just use
1280 MSG_host_get_property_value() and friends
1281 * VM I/O related things:
1282 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1283 - Unimplemented save/restore methods
1284 * MSG_as_router_get_property_value() was redundant with
1285 MSG_environment_as_get_property_value().
1286 - Removed MSG_as_router_*propert*() functions
1287 - Added MSG_environment_as_set_property_value() for consistency
1288 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1289 Well, they are still there for now, but deprecated with a warning.
1290 Please switch to xbt_dict_new_homogeneous() before this is removed
1292 * Task affinity. Its intended behavior (that was very badly tested
1293 and probably not really working) was deceiving what most users
1294 would have hoped here.
1295 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1296 * Option network/coordinates is now useless and should be dropped.
1299 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1302 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1303 * Added Dragonfly topology support
1306 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1307 EiBps, EBps, Eibps, Ebps for bandwidth.
1308 They may become useful to some lucky ones.
1311 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1312 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1313 * ex/async_waitAll: new example, on asynchronous communications
1316 * Memory usage should be decreased for simulations with a large number
1317 of processes. This also helps for SMPI.
1319 SimGrid (3.13) stable; urgency=low
1321 The Half Release, a.k.a. the Zealous Easter Trim.
1323 - We removed half of the lines, that were mostly experimental cruft.
1324 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1325 (not counting blanks and comments -- according to openhub.net)
1326 - The internals are now compiled in C++ (and will soon be clean C++)
1327 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1328 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1330 * Backwards Compatibility breaks
1331 - Removed Lua simulation bindings (switch to C or Java for that).
1332 Lua can still be used to describe platforms
1333 - Removed Java kernel plug-ins.
1334 Will be reintroduced after the ongoing major internals reorg.
1336 - the following functions were removed.
1337 They were too specific and should be reimplemented in a generic
1338 way, with filter function.
1339 - MSG_task_listen_from_host
1340 - MSG_mailbox_get_count_host_waiting_tasks
1341 - MSG_mailbox_put_with_timeout was removed.
1342 Please use MSG_task_send_with_timeout instead.
1344 - the SD_application_reinit function was removed. It has been a noop for a while.
1345 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1346 replaced by a more flexible mechanism.
1347 - The following functions thus do not exist anymore
1348 - SD_workstation_get_access_mode
1349 - SD_workstation_set_access_mode
1350 - SD_workstation_get_current_task
1351 - Basic estimation functions have been removed but can easily be replaced
1352 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1353 - SD_workstation_get_computation_time => amount / sg_host_speed()
1355 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1356 Use VM.getSpeed()*load/100 for the legacy behavior.
1358 - option enable_tracing was removed. It was not doing anything for a while.
1359 - In the ModelChecker:
1360 - the model-checker now ptraces the model-checked process which means
1361 you cannot use a debugger on the latter anymore (we might make this
1362 optional in the feature);
1363 - removed soft-dirty page tracking;
1364 - remove model-checked side snapshot management,
1365 MC_snapshot() and MC_compare_snapshot();
1366 - keep the MC_cut() function as a stub (it was not really working
1367 in the previous release).
1370 * All options are consistently in kebab-case. Old names are kept as alias.
1373 * Switch to platform v4 format.
1374 - Rename from 'power' to 'speed' the attributes describing the amount of
1375 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1376 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1377 - In <host> and <link>, attributes availability and state are gone.
1378 It was redundant with state and availability traces, and with peak values.
1379 - In <cluster>, attributes availability_file and state_file are gone.
1380 It was too complex and unused.
1381 - Kill <gpu>. Was not doing anything.
1382 - The DOCTYPE points to the right URL:
1383 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1384 (the file at this address now documents the changelog since its v1)
1385 - A warning is emitted for unit-less values (they are still accepted).
1386 - speed. Default: 'f' or 'flops'. Also defined:
1387 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1388 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1389 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1390 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1391 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1392 - latency. Default: 's' second. Also defined:
1393 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1395 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1396 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1399 * s4u::Host is now the preferred public interface to the Host features.
1400 sg_host_* functions are C bindings to the exact same behavior
1401 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1404 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1405 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1406 They were deprecated since a few years, and probably did not even compile anymore.
1409 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1410 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1411 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1412 encouraged to update their codes. A list of the modified functions can be found at the end of
1413 include/simgrid/simdag.h
1416 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1417 the lambda or closure passed as a parameter will run in kernel mode.
1418 All the callback functions should be rewritten to that interface at some point.
1421 * Reorganizing and cleaning the internals all around the place.
1424 * Remove old default barrier/bcast buggy algorithms (see #18407)
1425 * Various bug fixes to handle more codes
1426 * Remove the need for the --foreground option of smpirun (it is still
1427 accepted for backward compatibility).
1430 * Kill the setset data container: MC don't use it anymore.
1431 * Kill the queue data container: it made more sense with GRAS.
1432 * Kill the xbt_peer_t data type: it's useless without GRAS.
1433 * Kill rm_cb feature of config sets: it was never useful.
1434 * Kill graphxml parsing feature. It was not used.
1435 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1437 - xbt_dynar_sort_strings(), when the content is char*
1438 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1439 They throw exceptions on invalid input;
1440 * C++ support for declaring CLI flags (simgrid::config::Flag);
1441 * class for abstracting different signal backends (simgrid::xbt::signal).
1442 with no external dependencies (we need very simple signals).
1445 * refactoring and cleanup of the code;
1446 * ongoing process to cleanly separate the model-checking algorithms
1447 from the code model-checking support.
1449 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1451 SimGrid (3.12) stable; urgency=low
1453 The Facelift Release.
1456 * Require g++ v4.7 at least to not speak prehistorical C++.
1457 * Require Boost 1.48 (for signal2 component).
1458 * Java must be version 7 at least when activated.
1459 * Builds on Windows again (including Java bindings).
1460 * Tracing is now always enabled (no way to turn it out)
1461 * Remove GTNetS. It was not working anyway.
1462 * Various cleanups in the cmake scripts.
1463 * Move headers around to sort them out on installed systems:
1464 - instr/instr.h -> simgrid/instr.h
1465 - instr/jedule/* -> simgrid/jedule
1466 - simdag/datatypes.h was removed
1467 - simdag/simdag.h -> simgrid/simdag.h
1468 - msg/datatypes.h was removed
1469 - msg/msg.h -> simgrid/msg.h
1472 * Interface improvement:
1473 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1474 - Sanitize the interface in MSG_task_ module:
1475 - Merge two functions that were close enough but misleading:
1476 set_compute_duration(t) -> set_flops_amount(t)
1477 get_remaining_computation(t) -> get_flops_amount(t)
1478 - set_data_size(t) -> set_bytes_amount(t)
1479 get_data_size(t) -> get_bytes_amount(t)
1480 - Massive cleanups in the functions related to the energy
1481 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1482 - New: MSG_host_get_pstate()
1484 - msg/energy/onoff: switching hosts on and off
1486 * Interface improvement:
1487 - Rename Host.isAvail() to Host.isOn()
1488 - Rename Process.currentProcess() to Process.getCurrentProcess()
1489 - Rename Task.setDataSize() to Task.setBytesAmount()
1490 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1492 - #18874: Actually allows the GC to reclaim tasks
1496 - SIMIX_process_throw: raises an exception in a remote process
1497 * Refactoring: Separate sub-modules
1498 - libsmx: the public interface, as libc in a real system
1499 - popping: the strange dance that converts a user request into a kernel handling
1500 - smx_context_*: the virtualization mechanisms that embed the user code
1501 - smx_*: the handling of each simcalls
1502 * Interface cleanup:
1503 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1504 * Rename smx_action_t into smx_synchro_t, making explicit that these
1505 things are used to synchronize processes with their environment.
1506 For example, a communication is a sort of synchronization involving
1507 the communicating processes (that may block until the exchange) and
1508 the platform. The same can be said from computations, etc.
1510 - #18888: segfault when a process ends before its kill_time
1514 - 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.
1515 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1516 - MPI_Type_set_name, MPI_Type_get_name
1517 - MPI_*_c2f and MPI_*_f2c functions
1518 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1519 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1520 - Activate a lot of new tests from the mpich 3 testsuite
1522 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1523 - InfiniBand network model added: Based on the works of Jerome Vienne
1524 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1525 - When smpi/display_timing is set, also display global simulation time and application times
1526 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1527 * Collective communications
1528 - 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
1529 - 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/).
1530 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1532 - Replay now uses algorithms from wanted collective selector
1533 - Replay can be used with SMP-aware algorithms
1534 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1535 - 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.
1537 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1538 - Correctly use loopback on fat-tree clusters
1539 - Asynchronous small messages shouldn't trigger deadlocks anymore
1540 * Energy/DVFS cleanup and improvement
1541 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1542 - new: smpi_get_host_pstate()
1545 - "Full" network optimization flag was broken since Surf++
1546 - Better handling of precision flags in maxmin
1547 - Fix bug causing sometimes "Impossible" errors
1548 - Properly pass cluster properties to included hosts
1549 * Improvement of the Energy plugin.
1550 - Always update the consumption before returning that value
1551 - New property: watt_off to denote the dissipation when the host is off
1552 - New functions getWattMinAt and getWattMaxAt to retrieve the
1553 dissipation of pstates that we are not currently at.
1554 * Java: class NetworkLink renamed to Link
1555 * New function: simcall_process_get_kill_time()
1556 * Massive rename s/workstation/host/
1557 - That's intrusive, but that's good for the project consistency. Sorry.
1558 - Change config option "workstation/model" into "host/model"
1561 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1562 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1563 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1564 - xbt_dirname and xbt_basename for non-POSIX systems
1566 * The model checker now runs as a separate process.
1567 * The model checker runs must now be launched with the new simgrid-mc program.
1568 * Record/Replay: the MC can display a textual representation of a path in the
1569 execution graph. It can then be replayed outside of the model checker.
1571 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1573 SimGrid (3.11) stable; urgency=low
1578 * Normalizing pointers addresses tool for better diff between logs
1581 * Add cloud examples using new VMs
1582 - examples/msg/cloud/two_tasks_vm.tesh
1583 - examples/msg/cloud/simple_vm.tesh
1584 - examples/java/cloud/cloud.tesh
1585 - examples/java/cloud/migration/migration.tesh
1586 * Add java surf examples:
1587 - examples/java/surfPlugin/surf_plugin.tesh
1588 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1589 - examples/java/surfCpuModel/surf_cpu_model.tesh
1590 * Add SMPI+MSG example:
1591 - examples/smpi/smpi_msg_masterslave/
1596 - msg task destroy cancel test
1597 - msg_host on/off test
1598 * Move all tests in testsuite to teshsuite (adding tesh files)
1599 * Restructure teshsuites
1600 - one folder for each kind of test
1601 * Restructure AddTests.cmake
1603 - structure the order of tests (with sections)
1606 * Add virtual machine
1607 - creation of a VM on a PM
1608 - migration of a VM from a PM to another PM
1610 - MSG_process_join(msg_process_t process, double timeout)
1611 - msg_bar_t MSG_barrier_init(unsigned int count)
1612 - int MSG_barrier_wait(msg_bar_t barrier)
1613 - void MSG_barrier_destroy(msg_bar_t barrier)
1614 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1615 * New option "msg/debug_multiple_use" to help debugging when a task is used
1618 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1620 - Increase code coverage in test suites
1624 * Protect context stacks against stack overflow. The number of protected memory
1625 pages allocated on the top of each stack (1 by default) can be configured
1626 with the new command line option --cfg=contexts/guard_size:<value>.
1627 * Simcalls are now generated by a python script that
1628 - generates files included by SimGrid
1629 - checks that all the functions exist, and proposes prototypes
1631 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1632 host_set_data, host_get_data
1634 - simcall_process_join(smx_process_t process, double timeout)
1635 * Fix bug where sleeping processing could not be suspended.
1638 * Translate surf models from C to C++
1639 - Generic classes for all models: Model, Resource, Action
1640 - A generic interface for each kind of model (CPU, Network, Storage
1641 Workstation, WorkstationVM)
1643 * Translate surf routings from C to C++
1644 * Add callbacks using sigc++ or boost::signals2
1645 - Add callback functions for resource creation/destruction
1646 - Add callback functions for action state change
1647 - Handle Energy as a plugin
1648 * Replace swag by boost::intrusive
1649 * Add new routing models for clusters. For documentation, see
1650 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1651 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1652 parameters for cluster tag
1653 - Fat trees, with topology="FAT_TREE" and
1654 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1655 - see examples/platforms/torus_cluster.xml and
1656 examples/platforms/fat_tree_cluster.xml
1657 * More documentation
1660 * Hostfiles support host:nb_processes construct to deploy several processes on
1662 * Collective communication algorithms should not crash if used with
1663 improper number of nodes and report the error.
1664 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1665 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1666 MPI_Dims_create, MPI_Cart_sub are supported.
1667 * New interface to use SMPI programmatically (still depends on MSG for
1668 some parts, see examples/smpi/smpi_msg_masterslave):
1669 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1673 * Global variables privatization in MPI executables is now performed at runtime
1674 with the option smpi/privatize_global_variables (default:no).
1675 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1676 dynamic libraries loaded by the application are not privatized (static
1677 linking with these libraries is advised in this case)
1680 * Options defined in XML work correctly now.
1683 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1684 copied into simgrid.jar or not (ON by default). Use this option if you want
1685 to reduce the size of the installed simgrid.jar, *and* the native libraries
1686 are kept installed elsewhere.
1687 * Surf binding with SWIG (code generated in maintainer mode only):
1688 - plugin to handle callbacks
1689 - CPU model only for the moment
1692 * Supernovae build mode is definitively removed. It was used to improve
1693 inlining and inter-module optimizations. It is nowadays superseded by
1694 link-time optimizations commonly available in compilers.
1695 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1696 * Add boost dependency for surf++
1697 * Add new macro for tests
1698 - ADD_TESH(name <tesh_args>)
1699 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1703 - xbt_bar_t XBT_barrier_init(unsigned int count)
1704 - int XBT_barrier_wait(xbt_bar_t barrier)
1705 - void XBT_barrier_destroy(xbt_bar_t barrier)
1706 * Make the xbt_os_time module public
1708 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1710 SimGrid (3.10) stable; urgency=low
1712 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1715 * Reintegrate Java to the main archive as desynchronizing these
1716 package is not acceptable anymore (Java is now considered stable)
1717 * Add explicit synchronization facilities through semaphores
1718 * Bug fix: Task.setDataSize() only changed the C world, not the value
1719 cached in the Java world
1722 * Dramatically change the way files are handled. API and internals changed, but
1723 this part of MSG was not considered as production grade either.
1724 * Add explicit synchronization facilities through semaphores
1725 * Add a new function MSG_host_get_process_list()
1726 * Preliminary DVFS support (see examples/msg/energy/ for details)
1729 * SMPI is now included directly in the libsimgrid as the windows
1730 linker doesn't force us on splitting it anymore.
1731 * Improvements of the SMPI replay tool:
1732 - Most of the collective communications are now rooted in the same process as
1733 in the original application.
1734 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1735 used until now). Multiple data types can now be used in a trace.
1736 - The replay tool now supports traces produce either by TAU or a modified
1738 - Bug Fix: the compute part of the reduce action is now taken into account.
1739 - Gatherv collective is now supported
1740 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1741 of smpirun outputs time independent traces for the current run. One file
1742 is created per process. If too many processes are simulated, this behavior
1743 can be changed to one file for all processes by using the
1744 tracing/smpi/format/ti_one_file flag
1745 * smpirun generates the host file if needed (with given host count and platform)
1746 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1747 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1748 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1749 --cfg=smpi/coll_selector:(mpich/ompi)
1750 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1752 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1753 happening inside a collective SMPI call.
1754 * Fix the behavior of complex data types handling.
1755 * Make MPI_Wtime another synchronization point to take computations into
1757 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1758 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1759 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1760 (no privatization of global variables yet)
1761 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1764 * Verification of liveness properties is now available for SMPI applications
1765 (in addition to MSG applications)
1766 * Bugged examples using SMPI in examples/smpi/mc/
1767 * Add --cfg=model-check/visited option. Allows the verification of infinite
1768 programs. Detection of loops in the execution thanks to the system state
1769 comparison and reduction of the state space to explore. Can be combined with
1770 DPOR for safety properties.
1773 * Allow to change SimGrid configuration (see --help) within the code
1774 thanks to SD_config() as it can be done in MSG.
1775 * Add a new function SD_task_set_amount() upon user request.
1778 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1779 * Remove rule based routing (no more PCRE dependency)
1780 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1781 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1782 * Add a loopback_bw and loopback_lat options to cluster tag.
1783 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1784 routing. Now peers are encapsulated in an AS and have their own private
1785 router but this is transparent.
1788 * Our own implementation of getline is renamed xbt_getline, and gets
1789 used even if the OS provide a getline(). This should reduce the
1790 configuration complexity by using the same code on all platforms.
1791 * New type: xbt_cfg_elm_boolean.
1792 * Allow to use yes/no for boolean configuration options in the command line.
1793 * Allow to disable SimGrid cleanups at exit from command line option.
1794 There are situations where one may want a simulation to end with an exit.
1795 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1796 annoying when scripting around the simulator. Adding a
1797 --cfg=clean_atexit:no allows to circumvent this issue.
1800 * Lots of memory leaks were corrected in this release.
1801 There is no known memory leaks anymore, in all of our 600+ tests.
1802 * New command line option --version, to get SimGrid version information.
1803 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1805 * Supernovae builds are deprecated, and expected to be removed in the next
1808 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1810 SimGrid (3.9) stable; urgency=low
1812 The Grasgory release: GRAS is really dead now.
1814 * Complete overhaul of the internal host structures scheme.
1817 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1818 was considered as experimental and badly maintained since then.
1819 * Keeping it was thus a trap to our potential users, that could take
1820 it instead of MSG or SMPI by mistake despite is pity state.
1821 * GRAS seems to have very few users (if any), and no one volunteered
1822 to maintain it further. It also induces a lot of XBT code (for
1823 portability sake), that must be maintained too.
1824 * For all these reasons, we killed GRAS. If someone wants to revive it
1825 in the future, don't cry, our git history still remembers of GRAS.
1828 * Major overhaul. Merge our documentation again as time proved that
1829 splitting it was really not helping our users.
1830 * Further improve the developer documentation to help newcomers
1831 hacking on SimGrid itself. The user documentation (and in
1832 particular, the beginner documentation) is still in a sorry state.
1835 * Now works on Windows too!
1836 * Much more extensive test suite, from MPICH
1839 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1840 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1841 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1842 * Bug fix: let task be scheduled when the last dependency to be solved is
1843 a control dependency.
1844 * Remove SD_load_environment_script function.
1845 Use the C sg_platf function if you want to declare a platform
1849 * New function: MSG_process_get_number()
1850 * Old function documented: MSG_config()
1851 * Remove MSG_load_platform_script function
1852 Use the C sg_platf function if you want to declare a platform
1856 * Change the default value of the TCP_gamma constant (maximal size of TCP
1857 congestion window) to a more realistic 4MiB value. If you notice changes in
1858 your simulation results, you can fall back to the previous 20k tiny window
1859 by adding --cfg=network/TCP_gamma:20000 on command line.
1860 * (Hopefully) fix a bug wrt periodic availability and state traces
1861 * Bug fix: use default values at start when first event in availability/state
1862 trace is not at time 0.
1865 * remove the "new_" part of function name sg_platf_new_trace_connect
1866 (resulting in sg_platf_trace_connect), since it does not create
1870 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1871 use case where it's really mandatory, and maintaining it was a pain in
1873 * New: xbt_fifo_search(), search an item with a user-provided
1874 comparison function instead of dumb pointer comparison.
1877 * Fix the lua deployment:
1878 Use `simgrid.init_application()` before deployment instead of
1879 `simgrid.msg_register_application()` after.
1882 * Transfer the tracing files into the corresponding modules.
1884 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1886 SimGrid (3.8.1) stable; urgency=low
1888 The "we are told that some people want to also *install* the simgrid
1891 * Add missing file "tesh.1" to the archive.
1893 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1895 SimGrid (3.8) stable; urgency=low
1897 The Psssshiiiit release: SimGrid jumps into the Cloud.
1900 * Add an experimental interface to manipulate VMs. They are mainly
1901 process groups with very few intrinsic semantic, but they should
1902 allow you to build the semantic you want easily.
1903 * New function: MSG_host_set_property_value()
1904 * New function: MSG_process_on_exit(). To clean memory in all cases.
1905 * Bug fixes that made the host (and link) failures unusable.
1906 * Add a way to auto-restart process when the host in which they are
1907 executing comes back (ON_FAILURE="RESTART" on deployment file,
1908 MSG_process_auto_restart_set).
1909 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1910 please stop using the old ones, they are DEPRECATED.
1912 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1913 Please use MSG_init() instead. (reducing the amount of entry
1914 points in the library helps us).
1915 * Make it impossible to link against the wrong version of the lib
1916 * Deprecate MSG_clean(). No need to call it anymore.
1917 * Function MSG_get_host_number() is not deprecated anymore.
1920 * Split the doc into a user guide and a reference guide.
1921 * Start a developper guide to help people hacking on SimGrid.
1924 * Enable tracing by default. This modules rocks you should use it.
1925 * Remove option custom_flags. Now use environment variables CFLAGS
1927 * Use default cmake things to detect lua instead of home grown ones.
1928 * New option "enable_mallocators" to disable mallocators, for debugging
1929 purpose ("on" by default).
1932 * Bug fixes around the resource failures: don't let the processes
1933 survive the host they are running onto.
1934 * Add an interface to auto-restart processes when the host in which they are
1935 executing comes back.
1936 * Ensures that SIMIX_clean is called automatically. It's not part of
1937 the public interface anymore (bindings should be updated).
1940 * Bug fix for when SD_Simulate is called with a positive value: be careful
1941 when comparing doubles. Sometimes they are different for non significant
1943 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1944 parallel task whose initial work is distributed among host according
1945 to the Amdahl's law. Such tasks are created with a parameter alpha
1946 that corresponds to the non-parallelizable part of the computation.
1947 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1948 between two sets of workstations assuming a 1D block distribution (each
1949 workstation owns a similar share of data) on both sides.
1951 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1952 Data redistribution will be automatically scheduled once parent and child
1953 are both scheduled. The filling of computation_amount and
1954 communication_amount structures is now done seamlessly thanks to the chosen
1956 * New function SD_workstation_dump to display various information
1957 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1958 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1959 bandwidth on the route onto which the task is scheduled and the amount of
1961 To divide the nominal bandwidth by 2, the rate then has to be:
1962 rate = bandwidth/(2*amount)
1963 * Compute tasks that have failed can now be rescheduled and executed again
1964 (from their beginning)
1965 * Increasing source code coverage (src/simdag is now covered at 95.8%
1969 * Re-implement time-independent trace replay using SMPI (at the
1970 smpi_smp_* level) instead of MSG. This should replace
1971 examples/msg/actions/actions.c
1972 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1973 hindexed and structs)
1974 * Implement the exchange of non-contiguous data.
1975 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1976 * Correct behavior of smpi/sender_gap and set its default value to 0
1977 * Add option to asynchronously send small messages to allow better
1978 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1979 specifies the size in bytes under which messages will be asynchronously sent.
1980 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1981 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1984 * New C interface to define a platform: XML is now optional.
1985 For more info, please check include/simgrid/platf.h.
1986 * New interface to define random platforms from the C:
1987 For more info, please check include/simgrid/platf_generator.h and
1988 examples/msg/masterslave/masterslave_platfgen.c
1989 * Export a sg_cmdline dynar containing all the arguments we got from
1993 * Two new tracing options for adding comments to trace file so you
1994 can track your experiments (see --help-tracing for details).
1995 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
1996 * Adding the SimGrid version that generated the trace file as a comment.
1997 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1998 * Fix to avoid key clashes on Paje links
1999 * Other minor fixes related to the Paje specification
2002 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
2003 and renamed to xbt_str_hash() and xbt_str_hash_ext().
2004 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
2005 * Greatly improve the robustness of mmalloc to user errors (such as
2006 using an area after freeing it, or freeing it twice)
2008 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2010 SimGrid-java (3.8.1) stable; urgency=low
2012 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
2013 Warning: all methods are visible, but only some of them are
2014 implemented so far. Check the source (src/jtrace.c)
2015 for further information.
2016 * New module: org.simgrid.msg.File (SimGrid File management functions)
2017 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
2018 * Change the meaning of Process.restart: now restart the process from
2019 the beginning, like MSG_process_restart in C.
2020 * Add Process.setAutoRestart: handling of process restart when failed
2022 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
2023 you to retrieve the properties of the processes/hosts
2024 * Deprecate Msg.clean(): you can just forget about it now.
2025 * New function Process.getCount(), that only works when compiling
2026 with the not yet released version 3.9 of the C library.
2028 * New context factory based on Coroutines. It mandates a modified JVM
2029 but then, the simulations run about five times faster, and there is
2030 no limit to the amount of processes (beside of the available memory).
2032 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2034 SimGrid (3.7.1) stable; urgency=low
2037 * Restore the prototype of MSG_process_create_with_environment() to
2038 the pre-3.7 situation by removing the kill_time argument.
2039 * Add a MSG_process_set_kill_time() function instead.
2042 * Fix weird behaviors when dealing with parallel tasks.
2045 * Simgrid is now built as a dll.
2046 * Simgrid-java now works on Windows.
2047 * Simgrid-Java is now included into Windows package.
2050 * First pre-build package for MacOSX.
2053 * Fix compilation when using MSG_USE_DEPRECATED.
2054 * Fix some compilation issues on Macs and Windows.
2055 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
2056 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
2058 We discovered that the Lua console is broken, but we are missing the
2059 manpower to fix it right now. The problem existed in 3.7 too, so we
2060 are not blocking the release for that. Sorry if you depended on this
2061 feature, any help would be really welcome.
2063 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2066 SimGrid-java (3.7.1) stable; urgency=low
2068 The "Java aint got to be bloated and slow" release
2071 * Various internal cleanups and performance improvement
2072 Simulations are expected to run up to twice faster or so
2073 * Make Process.kill(process) an instance method, not a static one
2074 * User processes are not java.lang.Thread subclasses.
2075 This breaks the compatibility (sorry), but previous API was
2076 brain-dead, making it impossible to have non-trivial
2077 initializations in the process constructor.
2078 * Require a full constructor per Process sub-class.
2079 Kinda breaks the compatibility (sorry), but this allows a much more
2080 efficient way to launch the processes at simulation startup.
2081 * Do not embeed our version of semaphores, java 1.5 can be considered
2082 as sufficiently prevalent for us to not dupplicate its features.
2086 * Add examples for almost every part of the API
2087 We spotted and fixed a lot of bugs in the process
2089 * New module: asynchronous communication API
2090 * New function: Process.sleep()
2091 It takes milliseconds as argument, just as java.lang.Thread.sleep()
2092 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
2093 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
2095 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2097 SimGrid (3.7) stable; urgency=low
2099 The "spring cleanups (before the next Big Project kicks in)" release.
2102 * We can specify the SMPI latency/bandwidth factor with command line
2103 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2104 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2105 You can also use the "config tag" from platform file by adding this line
2106 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2107 (see "example/platforms/tag_config.xml" to use "config tag").
2108 Note that the command line supersedes the platform file configuration.
2109 * Change the correction factors used in LMM model, according to
2110 the latest experiments described in INRIA RR-7821.
2111 Accuracy should be improved this way.
2112 * Use the partial invalidation optimization by default for the
2113 network too. Should produce the exact same results, only faster.
2114 * Major cleanup in surf to merge models and split some optimization
2115 mechanisms from the core of the models. As a result you can now
2116 specify which model to use (e.g., --cfg=network/model:LV08
2117 --cfg=cpu/model:Cas01) and which optimization mode to use
2118 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2119 Incompatible combinations should err at initialization. See
2120 --help-models for the list of all models and optimization modes.
2121 * The CLM03 workstation model was dropped for simplicity because it
2122 used the deprecated CM02 network model. Use default instead.
2123 * Rename the TCP_gamma configuration option to network/TCP_gamma
2124 * Rename the coordinates configuration option to
2125 network/coordinates, and document it
2126 * Use now crosstraffic keyword instead of the terribly misleading
2127 fullduplex keyword. It is activated by default now in the current
2128 default model, use --cfg=network/crosstraffic:0 to turn it off.
2129 * Ongoing refactoring the model parsing to make XML files optional
2130 See include/simgrid/platf.h for details (still to be completed)
2133 * Major overhaul of the documentation. Almost instructive now :/
2134 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2135 functions and friends. This interface was considered as
2136 deprecated since over 2 years, it's time to inform our users that it is.
2137 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2138 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2139 use these (crappy) functions in your code.
2140 These functions will be removed soon. Stop using them now.
2141 * Deprecate MSG_get_host_{table,number}
2142 Implement MSG_hosts_as_dynar() instead.
2143 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2144 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2147 * Stabilize the parallel execution mode of user contexts
2148 * Introduce configuration variables to control parallel execution:
2149 - contexts/synchro: Synchronization mode to use when running
2150 contexts in parallel (either futex, posix or busy_wait)
2151 - contexts/parallel_threshold: Minimal number of user contexts
2152 that must be part of a scheduling round to switch to parallel
2153 execution mode (raw contexts only)
2154 * Fix bugs that prevented to use suspend/resume along with
2155 synchronization structures.
2156 * Fix bugs in process termination that lead to invalid memory access
2157 in very specific conditions.
2160 * Introduce a parallel mode for the models (controlled by surf/nthreads
2161 configuration item). In our tests, running the models in parallel
2162 never lead to any speedups because they are so fast that the gain
2163 of computing each model in parallel does not amortizes the
2164 synchronization costs, even when ultra fast futexes are used.
2165 This is released anyway because YMMV.
2168 * Performance boost by using a swag internally to compute the set of
2169 tasks that are finished and should constitute the return value of
2173 * Enable it by default now that it is considered rather stable.
2176 * Documentation of the tracing functions.
2177 * Performance gains when tracing categorized/uncategorized resource
2178 utilization by avoiding calls to get route when updating resource
2179 variables. LMM constraints are being used instead.
2180 * API changed to set task categories. Use MSG_task_set_category instead
2181 of TRACE_msg_set_task_category, and SD_task_set_category instead
2182 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2183 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2184 * Examples for link user variables added (see at examples/msg/tracing/)
2185 * Deprecated function TRACE_msg_set_process_category completely removed
2186 * Trace header updated according to the latest Paje file format
2187 * Tracing network lazy updates, no longer obligate users to use full updates
2188 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2189 * Experimental: let user code declare/set/push/pop application states for hosts
2190 * API changed to allow the manual creation of graph configuration files
2191 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2194 * Improve the API of Lua MSG bindings, using the Lua spirit.
2195 * Each simulated process now lives in its own Lua world (globals are
2196 automatically duplicated). It helps writing simulators. Will allow
2197 to run Splay programs within SimGrid in the future.
2198 * Add a Chord example in Lua, equivalent to the MSG one.
2201 * Start the implementation of a solution to express temporal
2202 properties, not only local assertions. This is still an
2203 experimental work in progress, stay clear from it to be safe.
2207 - Add new runtime parameters --help-logs and --help-log-categories
2208 to display information about supported logging parameters and
2210 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2211 don't exists anymore.
2212 * Mallocators: allow value NULL for the reset function.
2214 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2215 create homogeneous dictionaries, where all the elements share the
2216 same free function. Non homogeneous dictionaries will be
2217 deprecated in the next release.
2218 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2219 - Multi-level dictionaries are deprecated.
2221 - new function xbt_dynar_search_or_negative() that is useful when
2222 you have less than 2 million elements in your dynar and don't
2223 want of the extra complexity of catching exceptions when the
2224 element is not found.
2226 - Make xbt_os_thread module (for thread portability) public.
2227 Documentation is still to come, sorry.
2229 - Cleanups and simplifications to make it maintainable again.
2230 - Exotic features (such as memalign and valloc) were removed.
2231 - The metadata were extended and improved so that the
2232 model-checker becomes able to explore and inspect the heaps.
2233 - This may induce a performance drop when enable_model-checking is
2234 ON in cmake (even if it's not used in the simulation), but it is
2235 necessary at this point to get MC working.
2237 Turn model-checking OFF if simulation performance matters to you.
2238 Not enabling it at runtime is not enough, disable it in cmake.
2240 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2243 SimGrid (3.6.2) stable; urgency=low
2245 The "Not coding new stuff allows to polish old things" release.
2248 * New bindings to the NS3 packet level simulator (experimental)
2249 * Use the raw (efficient) execution contexts instead of the sysv
2250 (portable) ones when possible.
2251 * libpcre is now mandatory in any cases since not using it led to
2252 severe performance loss and possibly other issues
2253 * Update the XML platforms:
2254 - G5K: include the latest machine in Nancy
2255 - GridPP and LCG: new platforms
2256 * Documentation was partially updated, at least (more to come)
2258 Bug fixes, cosmetics and small improvements
2259 * Free terminated processes before the end of the simulation to avoid
2260 exhausting the memory of users having very dynamic amount of
2262 * Bug fix and cosmetics about canceling non-running tasks
2263 * Bug fix about the dot loader's issues when using libcgraph
2266 * Create an installer for windows with nsis (amd64 and win32)
2267 - Add an hello world project to illustrate simgrid project creation.
2268 - Embed libpcre into the Simgrid installer to avoid
2269 its compilation burden
2270 * The raw execution contexts should work on Apple now
2271 * Port to Windows 64 bits
2272 - Sysv contexts now have an implementation for this arch
2273 - GRAS communication features now support this arch
2274 * Drop support for borland compiler on windows
2275 - this code was not maintained, and we kinda depend on gcc nowadays
2276 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2277 * Fix portability issue on unstable ubuntu: linker became picky on
2280 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2282 SimGrid (3.6.1) stable; urgency=low
2284 The "Oops, we broke Macs too" release
2287 * Fixed contexts detection so that raw ones are used when possible
2288 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2289 leads to a strange error, with user code segfaulting sometimes when
2290 the generated code is not perfectly aligned (which is not
2291 controllable from the user side, depends on the amount of code)
2294 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2296 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2298 SimGrid-java (3.6) unstable; urgency=low
2301 * Split of every thing from simgrid v3.5 into a separate package.
2303 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2305 SimGrid (3.6) stable; urgency=medium
2307 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2310 * Bindings now constitute their own package, separated from the main one.
2311 Rationale: reduce our maintenance nightmare by reducing the module coupling
2312 They will soon be released on their own on gforge.
2314 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2315 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2317 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2318 * It's not quite deprecated for now because we have no replacement,
2319 but it may soon become the case.
2322 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2323 * Fortran: New user-level cache variable to store the rank of the running
2324 process. This improves performance by an order of magnitude.
2325 * C: New coccinelle script to automatically locate and modify global and
2326 local static variables.
2327 * Improved SMPI network model with a sender-side gap to account for multiple
2331 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2332 only say if a communication is finished, no matter whether it succeeded or
2333 failed. You can call MSG_comm_get_status() to know the status of a finished
2335 * New function MSG_task_dsend() to send a task and detach it. When a
2336 communication is detached, you are never notified of its success or failure
2337 and the memory is released automatically once it is finished. This function
2338 is useful when you don't care about the end nor the success of a
2340 * Change the prototypes of action replay. Sorry for inconvenience,
2341 but this is really more efficient this way (and to adapt your code,
2342 you just have to fix the initialization, that shouldn't be too long)
2343 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2344 ever managed to do anything useful with it.
2345 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2346 takes a dynar of communications. It returns immediately and gives the
2347 index of a finished communication (if any).
2348 * New example: a basic implementation of the Chord P2P algorithm.
2351 * New model for multi-core CPUs. You can now use the core attribute to
2352 precise the number of cores of a host. This is a basic model. Every
2353 process running on the host receives at most the power provided in
2354 the DTD (throughput<=power). Total throughput of process cannot exceed
2356 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2357 router linked by an up-link and a down-link (possibly asymmetrical).
2358 This kind of pattern allows to easily build last-mile model style platforms.
2359 Aggregating such patterns in a rule-based AS is thus the technique of
2360 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2361 * New model for Vivaldi routing. We transformed the Vivaldi network model
2362 into a Vivaldi routing model (based on the rule-based model). This allows to
2363 combine Vivaldi based latencies with last-mile platforms.
2366 * Added a check for NaN of IEEE754 infinite in the double entries of
2368 * Introduce a new context factory "raw", highly inspirated from the
2369 ucontext factory, but using manually crafted functions in assembly to
2370 do the work in an efficient manner.
2371 * Allow to change the used context factory at run time, not only at
2372 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2373 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2374 contexts at runtime (only with raw contexts or ucontexts).
2375 * Completely rewrote this module to allow parallel execution of user
2376 processes. Use --cfg=contexts/nthreads:N to execute user processes
2377 with N parallel threads (the default is 1, meaning no parallelism).
2378 * Allow to decide dynamically between sequential and parallel modes.
2379 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2380 processes in parallel only when their number is greater than or equal to P
2382 * Added a check for NaN of IEEE754 infinite in the double entries of
2386 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2387 won't output the state of processes when interrupted with Ctrl-C
2388 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2389 NULL-terminated array. This may solve backward compatibility issues
2390 due to the change to return type of SD_simulate. See also:
2391 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2392 * Add new macros with variable number of arguments.
2393 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2394 - in xbt/asserts.h: xbt_assert
2395 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2396 - in xbt/ex.h: THROWF and RETHROWF.
2397 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2399 * Change xbt_die() to accept a format string with arguments, just like printf.
2400 * New data structure: xbt_lib_t, like a dict but more general and with better
2404 * New configuration options
2405 Options triva/categorized and triva/uncategorized can be used to generate
2406 graph configuration files for Triva visualization tool.
2407 * Configuration option tracing/platform is renamed to tracing/categorized
2408 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2409 * New TRACE_user_link_variable interface:
2410 User provides the name of the link and the tracing variable to attach to it
2411 * the declaration of tracing categories must be done after the environment
2413 * simpler tracing interface, just one way to declare categories
2414 TRACE_category or TRACE_category_with_color, it is up to you
2415 * links in the trace file are again identified by their names
2416 * trace contains the full platform hierarchy exactly as declared using the ASes
2417 * Options tracing/msg/[task|process]:1 groups the process by hosts
2418 for both cases, tasks and processes must have names that are unique during the simulation
2419 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2420 * The experimental option tracing/msg/volume is deprecated
2421 its functionality may be reincorporated if needed
2423 The tracing generates a trace file with unordered timestamped events,
2424 because of the way the core simulator (surf) works. A script available
2425 at the tools directory (fix-paje-trace.sh) can be used to put the events
2426 in order. We have changed the tracing so it can generate ordered timestamped
2427 events in the final trace, but depending on the simulator (and how much time
2428 is simulated) that can lead to a huge memory utilization. It is deactivated
2429 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2431 Build Infrastructure
2432 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2433 - We are trying hard to keep the API stable, but it may happen that
2434 some things change (we're a research project after all, not a
2435 nuclear plant operating system). If such things should happen, you
2436 could rely on that macro to adapt.
2437 - current value: 30600 for 3.06.00, aka 3.6
2438 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2439 a number that can be compared with SIMGRID_VERSION.
2440 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2441 assertions and verbose/debug logging events are disabled at compile time.
2443 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2445 SimGrid (3.5) stable; urgency=medium
2448 * New feature to any SimGrid-based simulator: Model-Checking
2449 Check SIN#1 for more details.
2452 * New Model SMPI (three-interval linear regression for correction factors)
2453 See RR-7426, available at http://hal.inria.fr/inria-00527150
2454 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2455 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2456 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2457 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2458 * See also src/smpi/README
2463 - Tracing API changes: TRACE_start and TRACE_end should not be called
2464 by user-code. They are automatically called by simulators created
2465 with SimDAG, MSG and SMPI if the toolkit is compiled with
2466 tracing_enabled=ON. Categories declaration and utilization remain the
2467 same for MSG and SimDag.
2468 - A function was added to the tracing API to declare categories with
2470 - TRACE_category_with_color (char *category, char *color)
2471 where color must be in the following format
2472 "%f %f %f", red, green, blue
2473 and red, green, blue are float values in the interval [0, 1]
2474 - User can specify NULL as color parameter, or continue calling
2475 TRACE_category (cat)
2476 On that case, the tracing system will define random colors
2477 - The following command-line options are supported:
2478 --cfg=tracing/filename:msg.trace
2479 --cfg=tracing:1 (activate tracing, needed to use others)
2480 --cfg=tracing/platform:1 (categorized resource use)
2481 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2482 --cfg=tracing/msg/task:1 (task creation)
2483 --cfg=tracing/msg/process:1 (process creation, migration)
2484 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2485 --cfg=tracing/smpi:1 (SMPI interface tracing)
2486 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2487 - examples of examples/msg/tracing updated
2489 - DAXLoader and DOTLoader functions can generate tasks with categories
2490 - A new function to attribute a category to SD tasks:
2491 TRACE_sd_set_task_category (SD_task_t task, char *category)
2492 * Tracing the MPI interface implemented by SMPI
2493 - Collective operations are traced with states
2494 - Point-to-Point operations are traced with states/links
2495 - Tracing activated by a parameter "-trace filename" passed
2496 to smpirun during execution (considering that simgrid
2497 is compiled with tracing enabled)
2498 - To run the simulation with gdb, the simulator
2499 accepts --cfg=tracing/smpi:1 to trace SMPI
2500 - tesh files to check if smpi tracing is ok
2501 - See examples/smpi/NAS/DT-trace
2502 * GTNetS tracing re-worked
2503 - adaptation to the tracing system of GTNets to cope
2504 with modifications regarding the fullduplex mode
2505 - new tesh files to check if gtnets tracing is ok
2508 * Asynchronous communications through the functions:
2509 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2510 * New function: MSG_load_platform_script()
2511 to make possible using a lua script instead of XML files to set up platforms
2512 * New function: MSG_set_function
2513 to associate functions to processes, used when bypassing the parser
2514 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2516 Platforms: Add some more examples in examples/platforms
2517 * Grid'5000: see www.grid5000.fr
2518 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2521 * Change the XML format. This is a very important modification. SimGrid
2522 3.5 introduces a new hierarchical format based on the notion of
2523 Autonomous Systems. Compatibility with old format is ensured
2524 through the perl script provided in the install bin directory
2525 bin/simgrid_update_xml.
2526 It is now possible to build platforms with specific routing
2527 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2528 connect several platforms together. We will try to provide soon
2529 set of realistic platforms exploiting these properties (have a
2530 look at examples/platforms/ for the moment).
2531 * Take the opportunity of the XML format change to be a good XML citizen:
2532 rename link:ctn to link_ctn and similar changes (also dealt with by
2534 * Add a new routing scheme (rule-based) using regular expressions. It
2535 enables to have an extremely low memory footprint when the
2536 underlying routing is simple and can be compactly described. You
2537 need to have libpcre4-dev (perl regular expressions) installed if
2538 you want to use this routing scheme.
2539 * Revive the cluster TAG and allow to easily and efficiently (both in
2540 term of memory and speed) connect clusters together. Have a look
2541 at teshsuite/simdag/platforms/ to see how this can be done. With
2542 this tag, you can create clusters with thousands of tasks at no
2543 cost (have a look at examples/platforms/).
2544 Note: clusters are implemented as ASes, so there is no need for an
2545 enclosing AS tag if you have only one cluster in your platform.
2546 * Add new generic functions in the public interface that allows the user
2547 to call SURF 'create_resource' methods from your code (same
2548 functionality as the XML bypass mechanism but with a much lighter
2550 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2551 that uses a piecewise linear approximation to produce better
2552 results when exchanging small messages.
2553 * Add a new parameter to handle correctly full duplex link and account
2554 for interferences between uplink and downlink communications
2555 (activate with --cfg=fullduplex:1).
2558 * Rename the SD_READY (all dependencies are satisfied and task is
2559 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2560 dependencies are satisfied) state.
2561 This prevents a confusion between the notion of "ready to schedule"
2562 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2563 simulated" (SD_RUNNABLE) used by the simulation kernel.
2564 * Change the way a task is considered as ready. Instead of removing
2565 dependencies when a task is done, a counter is decreased. This way,
2566 it is always possible to reach ancestors thanks to the
2567 SD_taks_get_parents function (even after the end of the simulation.)
2568 * Change the return type of SD_Simulate from (SD_task_t*) into
2569 xbt_dynar_t. This function was in handling a dynar internally and
2570 converted it into a NULL terminated array for historical reasons.
2571 * New function SD_dotload(char*) to load a DAG described in dot
2572 format. This loader and the corresponding examples require the
2573 installation of the graphviz library.
2574 * Fix a bug in the management of tasks of size 0 in the surf network
2575 models. This problem was only visible with SIMDAG and you should
2576 thus disregard results produced with earlier versions if you
2577 relied on this feature (some tasks were blocked because of this).
2578 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2579 even though your applications were doing only point-to-point
2580 communications and sequential computations. Now you can really use any
2581 model you want (of course, if you create real parallel tasks, which are
2582 not implemented in most models beside ptaskL07, this will abort).
2583 * Add an example that schedules a DAX on an heterogeneous platform
2584 using a Min-Min strategy.
2585 * New function SD_workstation_get_current_task() that returns the kind
2586 of task currently running on a workstation in the sequential access
2588 * Raise some warnings when unexecuted tasks remains at the end of the
2589 simulation. This is usually caused by cycles in the DAG.
2592 * New function: SIMIX_process_set_function() called by MSG_set_function
2593 * Change the underlying waiting queue in semaphores so that a process
2594 can wait on several of them simultaneously (as in waitany).
2595 * Fix the way to handle tokens in semaphores so that all access patterns
2596 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2597 * kill the dirty pimple SIMIX_message_sizes_output()
2598 Please use (proper) visualization instead
2601 * New data container: setset (set of sets of elements)
2602 * New module: mmalloc (mapped malloc, allowing to have several
2603 independent segments of malloc)
2604 * New function: xbt_dict_cursor_set_data()
2605 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2606 * New function: xbt_dynar_is_empty()
2607 * New function: xbt_fifo_get_last_item()
2608 * Fix xbt_dynar_shrink(): use the right element size.
2609 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2610 uninitialized areas during expand.
2611 * Fix semaphores: previous implementation was severely broken.
2612 * Use library init/fini functions for our initialization.
2613 - you can use logs and other feature as soon as you want in your
2614 code (even before the xbt_init / MSG_init)
2615 - xbt_exit is now a no-op and produce a warning when used.
2618 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2619 benefit from the latest improvement to the simulation kernel.
2620 * Kill measurement sockets for now. If you rely on them, sorry. This
2621 release is not for you. This feature will be reintroduced in the
2622 future, but we cannot delay the release any further.
2623 * New function: gras_msgtype_get_name().
2624 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2625 * Fix (at last) the pmm example: it should not randomly fail anymore.
2627 Build chain: bug fixes and overall polishing
2628 * Cmake is now stable enough. Hence, we killed the autotools.
2629 * Port to windows ( TM :)
2630 * Fix the 'make install' target.
2631 No need to use 'make install-simgrid' anymore
2632 * Introduce a 'make dist' target compiling a *source* archive
2633 'make package' compiles a binary archive
2634 * Compile java files only on need
2635 * Add --cd and --setenv command line options to tesh
2636 * Out of source builds are not fully supported yet, but we are close
2637 * Enable supernovae and optimization flags by default for our users
2640 * Add layer to set up environment directly from lua, without XML.
2641 * The effect of gras_stub_generator can be achieved through
2642 lua too (check examples/gras/console/ping_generator.lua)
2644 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2646 SimGrid (3.4.1) stable; urgency=low
2648 The "Polishing easter eggs is probably a good idea" release.
2649 This is a bug fixes release only.
2653 * Fix a bug preventing the tasks from begin garbage collected.
2656 * Fix a bug occurring when a host involved in a communication fails.
2657 This was not detected properly by the other peer involved in the
2658 communication. Now, it's reported as a network error.
2661 * Warn the user about loop dependencies in data flow of DAX files
2662 * Obey the control-flow dependencies of DAX files
2665 * Add option "enable_smpi" allowing to not compile SMPI.
2666 Probably useful for the (Mac) users experiencing a build error here
2667 * Improve the detection of lua5.1 and ruby1.8
2669 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2671 SimGrid (3.4) stable; urgency=low
2673 The "Easter in Cargese" release. Also known as (major changes):
2675 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2676 ~> bindings were greatly improved
2677 ~> new tracing infrastructure for better visualization introduced
2679 * the "Welcome to configury modernity" release.
2680 ~> we switched from autotools to cmake, and improved our cdash
2683 A more detailled list of changes follow (full detail in svn log).
2685 Java Bindings: Various Cleanups
2686 * (install java-gcj-compat-dev on debian-like to use them)
2687 * Remove put/get: no need to export deprecated interface in Java
2688 Use send/receive instead.
2689 * Cleanup the examples and add a README per directory
2690 * Remove example autoDestination (that's the only way to go now)
2691 * Remove example explicitDestination (was a plain copy of basic)
2692 * Make JniException a runtime exception, so that there is no need to
2693 declare the fact that you may encounter such a beast. I guess that
2694 nobody will ever want to survive such error.
2695 * Create specific errors for each MSG case of failure:
2696 host failure, transfer failure, timeout, task cancelled
2697 * Cleanup the exceptions that may get thrown by each function
2698 * Other internal cleanups in Java bindings. Performance still bad :/
2699 Ruby and Lua Bindings: create them
2700 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2701 * That's new and great, you should try them out.
2702 Same functionalities than Java bindings, only even less polished
2704 * Kill the useless "rate" argument of SD_task_get_execution_time()
2705 Everyone used to provide -1 as a value, it was not used, and the
2706 semantic of a possible use wasn't even clear.
2707 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2708 either as comm costs or compute costs to mean that there is no
2709 such thing for that specific task.
2710 * Add a SD_task_set_name() function
2711 * Fix SD_task_unschedule() on typed tasks
2712 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2713 * In DAX loader, accept useless 'level' attributes to <job> since
2714 LIGO DAGs have them (seem to be to ease graphical representation).
2716 * Add an example masterslave_mailbox.c using send/receive and not
2717 the deprecated put/get interface.
2718 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2719 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2721 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2722 (declare MSG_USE_DEPRECATED to still have the old name)
2723 * Add a MSG_task_set_data() function
2724 * About trace replay (see examples/msg/actions):
2726 - Allow to work with splitted trace files for each process
2727 Give the specific trace file as argument of each process,
2728 and call MSG_action_trace_run(NULL)
2729 You can still have one merged file for all processes.
2730 - Fix implementation of collective operations
2731 * Allow task_execute() on 0-sized tasks (closes #10063)
2733 * This is the first release of SimGrid where SMPI is not considered
2734 beta anymore (even if some corners should still be improved)
2735 * Port over the new SIMIX_network submodule (internal refactoring)
2736 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2737 * Implement more missing elements of the standard:
2739 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2741 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2742 MPI_Scan() MPI_Get_processor_name()
2743 - Added implementation of missing case for Alltoall (warning: it's
2744 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2745 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2746 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2747 SMPI_MPI_Allgatherv()
2748 * Bug fixes include:
2749 - MPI_Waitsome() was broken
2750 - Allow relative includes in smpicc
2751 - Command line cfg argument 'reference_speed' was ignored...
2752 - Some functions did not properly lead to auto-benching of user code
2753 - smpicc passes -O2 by default (just like openmpi one)
2755 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2756 * Bug fixes about timeouts during communications
2757 * add SIMIX_message_sizes_output() as a pimple to write to file the
2758 amount of messages per size. Use gnuplot to get histogram.
2759 Pimple because that's the only user-visible function of simix,
2760 defined directly in xbt.h (irk, sorry)
2762 - Add a SIMIX_sem_get_capacity() function
2763 - Fix interactions with processe resume/suspende
2764 - release_forever() was stupidly broken
2765 - Fix SIMIX_display_process_status() for processes in a semaphore
2766 - Make SIMIX_sem_block_onto() user-visible
2767 * Refactoring context stuff:
2768 - Use pseudo-OOP for better modularity
2769 - reimplement SIMIX_process_kill() without process_schedule() so
2770 that the latter can take as invariant that it is called from
2772 - Merge context_start into context_new for sake of simplicity
2774 * Add a Vivaldi network model, coded live during SUD'10 ;)
2775 * Rename configuration variables to start a hierarchy:
2776 o cpu_model -> cpu/model
2777 o network_model -> network/model
2778 o workstation_model -> workstation/model
2779 * New configuration variables:
2780 o network/bandwidth_factor: correction to bandwidth
2781 o network/latency_factor: correction to latency
2782 o netwotk/weight_S: correction to the weight of competing streams
2783 * Add a long description to the models, that users can see with such
2784 argument on the command line: --cfg=cpu/model:help
2785 * --help-models display the long description of all known models
2787 * config: add the ability to set a default value after registration
2788 Does not override any previously set value (e.g. from cmd line)
2789 * dict: allow to have integer key and data.
2790 When so, you need to use the following functions
2791 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2792 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2793 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2794 In contrary to regular dicts, the key is not malloced before copy.
2795 Mixing scalar and regular elements in the same dict is not tested
2797 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2798 Tracing for Visualization:
2799 * SimGrid is now instrumented in order to generate a trace file for
2800 visualization analysis: to use it, need to compile SimGrid with the
2801 "tracing" option enabled, and instrument the program using SimGrid with
2802 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2803 (among other functions).
2804 * The instrumentation only traces the platform utilization for now
2805 * Documentation to use the tracing functions and how to analyze the
2806 traces with the Triva tool is written.
2807 * More information about: SimGrid FAQ (in the section Tracing Simulations
2810 * We moved to cmake as default build system. Autotools support will
2811 be dropped soon. Check the FAQ for more info about how to use it.
2812 * Greatly improved our cdash/ctest interactions
2813 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2814 * Added memory checking tests with valgrind; lot of memleak fixing.
2815 This may be the first release of simgrid with so few memory issues
2816 * Added code coverage tests.
2817 Our coverage is still improvable, but at least we see it on cdash.
2819 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2821 SimGrid (3.3.4) stable; urgency=low
2823 The "Desktop Grid needs love too" release (also called Xmas release).
2825 Models improvements:
2826 * Major speedup in the maxmin system solving by using lazy evaluation
2827 Instead of solving completely the maxmin system at each iteration,
2828 only invalidate (and recompute) the modified parts.
2829 This new feature is enabled in default models but you can try to
2830 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2831 * Cas01 IMproved as default CPU model
2832 This CPU model is the same Cas01 model, but it uses the
2833 maxmin-selective-update flag and a heap structure to manage
2834 actions on SURF kernel.
2835 It reduces the complexity to find the next action to finish and,
2836 consequently, it's faster than the old Cas01.
2837 This is the new default CPU model (Cas01).
2838 * Rename the old Cas01 model to Cas01_fullupdate
2839 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2840 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2841 * CpuTI (CPU Trace Integration)
2842 A new CPU model whose objective is simulate faster when using
2843 availability trace files.
2844 Instead of using a full featured, over engineered maxmin system for
2845 CPU modeling, this model does the pre-integration of traces files
2846 to calculate the amount of CPU power available, and so, executes
2847 faster than the old CPU models.
2848 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2849 * Use LV08 as default network model since it gives better accuracy
2850 for small messages and shouldn't change things for big ones.
2851 Use --cfg=network_model:CM02 to get the previous behavior.
2854 ******************************************
2855 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2856 ******************************************
2857 * The new CPU model may changes simulations!
2858 The point is that events occurring at the exact same timestamp
2859 are not scheduled in the same order with the old and new
2860 version. This may be enough to completely change the execution
2861 of simulations in some cases.
2862 * The new network model will change simulations!
2863 This new model is more realistic than the previous one, so you
2864 should consider redoing your old experiments with this model.
2865 Sorry for the inconvenience.
2868 * Introduce the supernovae compilation mode
2869 When compiled that way, the whole SimGrid (or almost) is put in a
2870 single compilation unit and compiled in one shoot.
2871 This is to help gcc which has difficulties to inline stuff from one
2873 The speedup seem to be above 15%, althrough more tests are needed on
2874 amd64 to confirm that gain.
2877 * Port of MSG's mailbox on top of SIMIX network
2878 The put/get mechanism was greatly simplified on the way.
2881 * New SIMIX network module. Provides:
2882 - Mailbox: rendez-vous mechanism to find with who you want to speak
2883 - Synchronous send/recv: easier and hopefully faster since the
2884 logic is handled in the maestro process directly now
2885 - Asynchronous send/recv: you dreamt of it? It's here now
2886 Too bad that nobody cared enough to propagate the change to MSG.
2887 * Add semaphores as SIMIX synchronization mechanism.
2890 * new function SD_daxload(char*) to load a DAX file
2891 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2892 * Introduce typed tasks. Specify its kind and cost at creation.
2893 At scheduling, just give where it should be placed, and the cost
2894 for each involved resource is automatically computed.
2895 Existing constructors so far (more to come of course):
2896 - SD_task_create_comm_e2e() for end-to-end communication
2897 - SD_task_create_comp_seq() for sequential computation
2898 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2899 * new function SD_task_dump() for debugging display
2900 * new function SD_task_dotty(task,FILE*) writing to file the info
2901 about the task in dotty format
2902 * SD_task_dependency_exists() can now cope with having one of its
2903 arguments NULL. If so, it tests whether the other argument has any
2905 * Add getters on list of preceding/following tasks:
2906 SD_task_get_parents(task) and SD_task_get_children(task)
2907 * Add getters on amount of workstations and list:
2908 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2909 * Add getter on task kind: SD_task_get_kind(task)
2910 * Update the start_time and finish_time of tasks on completion/failure
2911 * Bugfix: Remove task from state swags when destroyed
2914 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2917 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2918 dynar with one separate thread per value of the dynar.
2919 * Change the prototype of xbt_thread_create(), sorry.
2920 Added a boolean parameter indicating whether we want to join this
2921 thread (used in SG only for now)
2922 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2925 * GTNetS wrappers should now be usable again (and betterly tested too)
2926 * Fix a major regression from 3.2 where the timeout provided to
2927 MSG_task_put_with_timeout() was used as absolute time before which
2928 the comm should be done.
2929 * Start to fix the <cluster> tag.
2930 - Internal links should be good now (beside of the loopback, which
2931 use the private link instead)
2932 - paths to the external world is still rather broken
2933 - the <route:multi> tag is just broken. Actually that's brain-dead.
2934 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2935 to make it less stupid
2936 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2937 * Fix a source-level compatibility glitch from 3.2: after defining
2938 MSG_USE_DEPRECATED, you can use the old name
2939 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2940 * Allow to compile from the SVN with automake 1.11
2941 * Fix some problems when using the "start_time" tag in deployment XMLs.
2942 * Fix #8569: XBT/synchro.h has redundant declarations
2943 * Fix #8563: MSG return values and exceptions
2944 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2945 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2946 Thanks to Robson Peixoto.
2947 * Fix a vicious bug in dictionaries inducing that some elements were
2948 not freed on xbt_dict_free()
2950 Portability report of this version:
2951 * Main portability targets:
2952 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2953 on (amd64/i386/ia64)
2954 - mac leopard on i386
2955 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2956 but nothing critical.
2957 * Other platforms: windows, AIX and others were not tested for this release
2959 Timing report of this version:
2960 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2961 scenario parameters). From 8h to a few seconds in desktop grid settings.
2962 * Supernovae brings about 25% speedup on i386.
2964 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2966 SimGrid (3.3.3) stable; urgency=low
2968 The "Need for Speed" release.
2970 The timings done to validate the 3.3.2 were faulty.
2971 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2973 The problem was a conversion from a manually handled vector to
2974 xbt_dynar_t on the critical path.
2975 xbt_dynar_foreach calls functions, inducing stack management crap.
2977 We inlined these functions and xbt_dynar_foreach is now breath taking.
2978 We also inlined xbt_swag_belong on the way.
2980 Here are some approximate speedup measurements (on master/slaves
2981 simulations lasting between 10s and 20s each):
2982 3.3.1 -> 3.3.2: about same performance
2983 3.3.2 -> 3.3.3: 40% speedup
2984 3.3.1 -> 3.3.3: 40% speedup
2985 3.3.1 with inline patch -> 3.3.3: 30% speedup
2987 Our reading is that the refactoring which occurred in 3.3.2 made us
2988 suffer much more from the xbt_dynar_foreach low performance, but
2989 once we solved this, this refactoring proved to be very performance
2990 effective. From the 40% speedup, somehow, 10% are due to the
2991 inlining and 30% to the refactoring.
2993 That's a pitty that gcc cannot inline functions placed in other files
2994 alone. We have to choose between:
2995 - break the encapsulation (by putting private data structures and
2996 accessors in headers files to help gcc)
2997 - live with low performance
2998 - switch to a decent compiler such as icc (not quite possible).
3000 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
3002 SimGrid (3.3.2) stable; urgency=low
3004 The "Simplicity does not preceed complexity, but follows it" release.
3006 The main contributors of this release were (lexical order):
3007 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
3010 * Extract the routing logic into its own object.
3011 (was dupplicated in network.c and workstation_LV07.c;
3012 Allows to implement other ways of storing that info)
3013 => kill now useless network_card concept
3014 - Use dynar to represent routes (instead of void** + int*)
3015 - kill link_set (use surf_network_model->resource_set instead)
3016 - Add a command-line option to choose the routing schema to use
3017 - Add three new models:
3018 * Floyd (shortest path computed at initialization)
3019 * Dijikstra (shortest path recomputed all the time)
3020 * Cached Dijikstra (shortest path computed on need)
3021 All these models where contributed by Silas De Munck, and are
3022 described in his ICCS09 paper.
3024 * Simplify model declaration
3025 (less redirections, less function to write when defining a model)
3026 - Factorize stuff between models:
3029 surf_model_resource_set(model)
3030 surf_model_resource_by_name(model, name)
3031 - Unify the types of models in s_surf_model_t (using an union)
3032 - Embeed fields of common_public directly into s_surf_model_t
3033 - Rename model methods:
3034 action_free ~> action_unref
3035 action_change_state ~> action_state_set
3036 action_get_state ~> action_state_get
3037 - Change model methods into functions:
3038 (model)->common_public->action_use ~> surf_action_ref
3040 * Implement a generic resource; use it as ancestor to specific ones
3041 (allows to kill duplicated code in models)
3042 Drawback: timer command don't need no name nor properties;
3043 workstation_CLM03 don't need no properties
3044 (but I guess we can live with those few bytes wasted)
3046 * Improve the action object model
3047 - implement a constructor avoiding dupplicated code about field
3048 initialization in generic_action part.
3050 * Kill the SDP model: it has an external dependency, is deprecated
3051 in flavor of modern lmm models, and didn't compile since a while
3054 * Relocation of the context module from XBT to SIMIX.
3055 (the context were decoupled from the simix processes, duplicating a lot of code)
3056 => a lot of code was factorized
3057 - less overhead is introduced during scheduling
3058 - simpler API for the context factory
3059 - the logic for process creation,destruction and manipulation was simplified
3060 * Simplification of the s_smx_process_t data structure.
3061 => accesing the simix level data associated to a process is faster now,
3062 and the code is a lot more readable.
3065 * Implement some more MPI primitives:
3066 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
3067 -implementation: Bcast: flat or 2-ary tree (default),
3068 Barrier: 4-ary tree,
3070 Allreduce: Reduce then Bcast
3071 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
3072 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
3073 Alltoallv: flat tree, like ompi
3075 * Add support for optimized collectives (Bcast is now binomial by default)
3076 * Port smpirun and smpicc to OS X
3079 * Kill SD_link_get_properties: hard to maintain and makes very little sense
3080 Shout out if you used it.
3083 * Display the list of still queued messages in SG mode when existing
3087 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
3088 * Add xbt_graph_node_get_outedges() [Silas De Munck]
3089 * Add xbt_str_from_file(FILE*)
3090 * Add xbt_dict_get_key achieving a linear reverse search
3091 * Remove the context module
3093 Portability report of this version:
3094 * Main portability targets:
3095 - Linux(debian)/x86/context
3096 - Linux(debian)/x86/pthreads
3097 - Linux(debian)/amd64/context
3098 - Linux(debian)/amd64/pthreads
3099 On these, we still have the eratic breakages of gras/pmm and
3100 amok/saturate_sg reported in previous version. We still think
3101 that the tests are the cause of the fault, not the tested code.
3103 - Mac OSX Leopard/x86/context
3104 Still false negative in tesh autotesting.
3105 Smpi still fails, but this time because readlink does not accept -f
3106 Everything seems to work properly beside of that.
3109 - AIX version 5.3 (only tested contexts this time)
3110 Smpi still fails there because mktemp is not installed.
3111 Everything seems to work properly beside of that.
3113 I managed to compile it for the first time, but several breakages.
3114 Won't delay the release for this exotic platform.
3116 * Windows: it's still lagging behind. If you want to help, please
3119 Timing report of this version:
3120 This version seem to be more than 5% faster than 3.3.1 (on linux
3121 64bits with contextes). The gain is less than expected, we are
3122 investigating this for next release.
3124 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3126 SimGrid (3.3.1) stable; urgency=low
3129 * Implement a --cfg-help to show existing configuration variables
3130 * Build chain do not require doxygen in maintainer mode
3133 * fix a bug on struct sizeof computation, which prevented the
3134 exchange of arrays of structs in some conditions
3135 - added a regression test about this in datadesc_usage
3136 * Allow the exchange of 0-long dynamic vectors.
3137 - for that, use -1 as indicator of dynamic size instead of 0
3138 - This implied to change any size from unsigned long to long,
3139 reducing a bit communication abilities, but I guess that with
3140 64bits being quite common, this is more than enough.
3141 - This also induce a protocol change, thus bumping network protocol
3142 version from 0 to 1 (if we have external users, we have to get
3143 clean on that point too ;)
3144 - added two regression tests about this in datadesc_usage
3145 * Be more verbose when propagating local exceptions
3146 This helps debugging.
3147 * Display the status of simulated processes when receiving SIGINT in
3151 * Allow to control the simulation from a trace file.
3152 New functions MSG_action_register() and MSG_action_trace_run()
3153 The first one allows to associate a function execution to each
3154 kind of action while the second one parses a trace file and
3155 triggers the corresponding actions within the system.
3156 For now, only a toy example is provided in examples/msg/actions
3157 * Add an example of process migration in examples/msg/migration
3158 * Fix a bug in task exchange which broke MSG_task_get_sender()
3159 Add a teshsuite regression test for that.
3160 [Bug: if MSG_task_get_sender() is called after sender exit,
3162 * Fix a bug which prevented suspend/resume to work properly
3163 * Display the status of simulated processes when receiving SIGINT
3164 This fixes a regression of v3.3. due to the introduction of SIMIX
3165 * Bug fixing in failure management:
3166 - trace could not start by a failure at time 0
3167 - failure during communications were not working
3170 * Add SIMIX_process_set_name() to change the name of the current
3171 process in the log messages.
3172 * Store smx_hosts in a dict since we only retrieve them by name
3173 * Move the configuration infrastructure to surf
3176 * Move the configuration infrastructure to surf
3179 * Massive internal cleanups:
3180 - Store internal structures on processes instead of hosts (allows
3181 to have more than one process per host, in addition of being more
3183 - Cleanup the initialization/finalization process
3184 - Kill a whole bunch of unneeded synchronization:
3185 processes run in exclusive manner within the simulator
3186 - Move queues from global tables to process data fields
3188 - now accept -platform and -hostfile arguments
3189 - Pass the right rank value to processes according to the hostfile
3190 * Compile the examples by default, and use them as regression tests
3191 * Implement MPI_Wtime()
3192 * Change the reference speed to a command line option
3195 * TCP_gamma can now be specified as command line option using
3196 --cfg=TCP_gamma:10000000.0
3197 * Change the --surf-path cmd line option into --cfg=path:
3200 * Also include strbuff from xbt.h public header
3201 * xbt_ex_display(): do not free the exception after displaying
3202 This allows to do more with the given exception afterward.
3203 Users should call xbt_ex_free() themselves.
3207 Portability report of this version:
3208 * Main portability targets:
3209 - Linux(debian)/x86/context
3210 - Linux(debian)/x86/pthreads
3211 - Linux(debian)/amd64/context
3212 - Linux(debian)/amd64/pthreads
3213 These targets fail about 1/10 of times on gras/pmm, but we believe
3214 that this is because of the test, not because of simgrid.
3215 amok/saturate_sg fails even more rarely, and the test may not be
3218 - Mac OSX Leopard/x86/context
3219 The test suite still spits tons of errors because some obscure
3220 force prevents us from removing the temporary directories
3221 arguing that they still contain some metadata I've never heard of.
3222 Smpi fails because seq is not installed.
3223 Everything seems to work properly beside of that.
3226 - AIX version 5.3 (both contexts and pthread)
3227 Smpi still fails there because mktemp is not installed.
3228 XML inclusions seems rosty on AIX.
3230 * Windows: it's still lagging behind. If you want to help, please
3233 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3235 SimGrid (3.3) stable; urgency=high
3239 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3240 [Malek Cherier & Mt]
3242 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3245 Until now, GRAS were using MSG as an interface to SURF. It was
3246 quite difficult because both interface have several differences
3247 (MSG channels vs GRAS sockets were the most notable point).
3249 This also opens the gate to SMPI (which should occur soon) and speed
3250 up simulations by to 40% (even if it were not the main goal).
3252 **************************************
3253 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3254 **************************************
3255 The point is that events occurring at the exact same timestamp are
3256 not scheduled in the same order with the old and new version. This
3257 may be enough to completely change the execution of simulations in
3258 some cases. Sorry for the inconvenience.
3260 * Cleanup and upgrade the XML format to push further scalability
3261 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3263 * Improve the testing infrastructure with tesh. Now a very large part of
3264 the code is tested not only by being run but also by checking that the
3265 output match an expected output [Mt].
3267 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3268 is really less memory-demanding, which should allow you to use
3269 larger files in SimGrid [AL].
3271 * Inform valgrind about our contextes, so that it becomes usable
3272 with the default (and more effecient) version of SimGrid
3273 [contributed by Sékou Diakite, many thanks]
3276 * Introduce a listener thread in charge of receiving incoming
3277 messages from the network. It allows to overlap communication and
3278 computation but most notably, it removes some stupid deadlocks due
3279 to the fact that so far, a process could not send and receive at
3280 the same time. This made most non trivial communication schema
3282 * Convert the PIDs from long int to int to match the MSG ones (and
3283 linux ones too) [Mt]
3284 * New function: gras_agent_spawn() to launch a new process on
3285 current host. Only working in simulation for now. [Mt]
3286 * New function: gras_os_hostport() returning a constant form (ie,
3287 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3290 * Make the backtrace of exceptions more human readable [Mt]
3291 * New module: xbt/str [Mt]
3292 a ton of string utility functions (split, join, printf to a newly
3293 allocated buffer, trim, etc)
3294 * New module: xbt/hash [Mt]
3295 SHA1 hashing algorithm (more to come if needed)
3296 * New module: xbt/synchro [Mt]
3297 synchronization tools (mutex and conditions) working the same way
3298 in simulation and in real life (mainly useful for GRAS, but not
3300 * New module: xbt/queue [Mt]
3301 classical producer/consumer synchronization scheme
3302 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3303 (using the classical functions will get serialized) [Mt]
3304 * Make dictionary internal table dynamic. No need to specify its size
3305 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3307 * Make sure the log channels are organized as a tree under windows
3308 (because of ANSI C compatibility issue, any channel were child of
3309 root directly) [Mt].
3312 * Cleaned many thing in surf and fixed a few bugs [AL].
3313 * Add a nice command line configuration mechanism to compose models [AL].
3314 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3315 the previous one (KCCFLN05). It relies on something that looks like
3316 a max-min sharing mechanism but cannot be written as such. A new solver
3317 was thus designed [AL].
3318 * Add a new solver to lmm. Based on Lagrange optimization and
3319 gradient-based descent, it enables to efficiently maximise systems s.a
3321 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3323 This solver enables to propose two new network models for TCP Reno and
3324 TCP Vegas based on Low's work. These models still need to be fully
3325 tested though [Pedro Velho].
3328 * Bug fix in SD_simulate. Now the time bound given as argument is
3330 * Use the new parallel task model (ptask_L07) as default.
3331 * Use the SURF command line configuration mechanism.
3332 * 0-size tasks (for synchronization) should now work.
3334 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3336 SimGrid (3.2) stable; urgency=high
3340 We still experience issues on this platform, but we believe that at
3341 least MSG is usable.
3343 GRAS API BREAKAGE (for simplification purpose, sorry):
3344 * the gras_msgtype_by_name is not used anymore. Instead of
3345 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3346 you can write (and must)
3347 gras_msg_send(toserver, "request", &request);
3348 - If you still want to pass a gras_msgtype_t to the function (to cache
3349 the type and avoid the lookup time), use the gras_msg_send_() variant.
3350 - Impacted functions:
3351 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3352 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3353 * The callbacks are now expected to return 0 when everything went well
3354 (just like the main() function)
3356 GRAS new features and improvements:
3357 * New module mechanism where user code can use per process globals [Mt]
3358 This is similar to gras_userdata_*() functions, but for libraries. It
3359 factorize some code developped over and over in the examples and AMOK.
3360 It has still to be documented and used (only amok/peermanagement is
3362 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3363 loss when they were small enough to fit into the buffer and sent quickly
3364 enough so that they can all get received in one shoot.
3365 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3366 if not found. Use the *_or_null() variant for the old semantic.
3367 * In gras_msg_handle, do not discard messages without callback.
3368 They are probably messages to be explicitly awaited later (ie, proofs of
3369 mis-synchronization in userland since they are sent before being awaited)
3370 No big deal usually.
3371 * gras_socket_meas_send/recv: semantic changed!
3372 The numerical arguments used to be (1) the total amount of data to send
3373 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3374 messages. This was need for the fool willing to send more than MAXINT
3375 bytes on quite fat pipes.
3378 * Do really rename the hostmanagement module to peermanagement. [Mt]
3379 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3380 but this is rather new and this was documented in the module
3381 documentation (poor excuses, I admit)
3382 * Bandwidth measurement semantic changed! This follows the changes to
3383 gras_socket_meas_send/recv explained above.
3386 * A sequential mode has been added to the workstations. When a workstation
3387 is in sequential mode, it can execute only one task, and the other tasks
3388 are waiting in a FIFO. [Christophe Thiery]
3391 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3392 model for SIMDAG. [Christophe Thiery]
3393 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3394 equal to 0, which caused some bad behaviors in
3395 saturated_constraint_set_update. I now use a threshold mechanism like in
3399 * When running manually src/testall, you select specific units [Mt]
3400 testall is the result of our cunit mechanism, and should replace all
3401 the scripty thingy around since bash don't run easily on billware.
3403 * A mallocator system has been added. [Christophe Thiery]
3404 Mallocators allow you to recycle your unused objects instead of freeing them
3405 and allocating new ones.
3407 Documentation update:
3408 * FAQ reworking + New FAQs:
3409 - "Valgrind spits tons of errors!" [Mt]
3410 - "How to repport bugs" [Mt]
3411 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3412 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3413 - Communication time measurement within MSG [AL]
3414 - I experience weird communication times when I change the latency [AL]
3415 * GRAS tutorial [Mt]
3417 - an introduction to the framework and to the used communication model
3418 - an initiatic tour introducing the most proheminent features:
3420 . Lesson 0: Installing GRAS
3421 . Lesson 1: Setting up your own project
3422 o Part 2: Message passing
3423 . Lesson 2: Exchanging simple messages
3424 . Lesson 3: Passing arguments to the processes (in SG)
3425 . Lesson 4: Attaching callbacks to messages
3426 . Lesson 5: Using globals in processes
3427 . Lesson 6: Logging information properly
3428 . Lesson 7: Using internal timers
3429 . Lesson 8: Handling errors through exceptions
3430 . Lesson 9: Exchanging simple data
3431 . Lesson 10: Remote Procedure Calling (RPC)
3432 . Lesson 11: Explicitly waiting for messages
3433 . Recapping of message passing features in GRAS
3434 - A HOWTO section containing:
3435 o HOWTO design a GRAS application
3436 More are due, of course. They will come latter. In the meanwhile, you can
3437 check the examples which are still here.
3439 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3441 SimGrid (3.1) stable; urgency=high
3445 There was a stack corruption somewhere, visible only when optimizing
3446 with these versions. [Vince]
3449 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3450 module that enabled to play with Directed Acyclic Graphs. It is built
3451 directly on top of SURF and provides an API rather close to the old
3452 SG. Some old codes using SG are currently under rewrite to check that
3453 all needful functions are provided. [Christophe Thiery]
3456 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3457 extension of the classical CLM03 model that gracefully handles
3458 failures. This is now the default model for MSG and GRAS. It doesn't
3459 handle parallel tasks yet though. [AL]
3460 * Bug fix: Weights were not correctly set in the network part.
3461 WARNING: This may have resulted in incorrect results with simulations
3462 where there are more than one flow on a given link. [AL]
3465 * After a (long ?) discussion on simgrid-devel, we have decided that the
3466 convention we had on units was stupid. That is why it has been decided
3467 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3468 WARNING: This means that all previous platform files will not work as
3469 such with this version! A warning is issued to ask users to update
3471 A conversion script can be found in the contrib module of the CVS, under
3472 the name contrib/platform_generation/surfxml_update.pl [MQ]
3475 * Bug fix: Processes were started in reverse order, wrt deployment file.
3476 WARNING: if your code relies on this bug, please fix it. [AL]
3477 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3478 being executed on two different locations. [AL]
3479 * Bug fix: Failures are now better supported thanks to Derrick's tests
3480 (there was many failure situations I hadn't thought of and that weren't
3481 correctly handled). [AL]
3482 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3486 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3487 Exception occurring on server-side are propagated back to client (!).
3489 API CHANGE: the callback changed their prototype. Change:
3490 int my_handler(gras_socket_t expeditor, void *payload_data) {
3492 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3493 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3495 * New! function: gras_msg_handleall to deal with all messages arriving
3496 within a given period.
3497 * New! function: gras_socket_server_range to get a server socket in a
3498 range of port numbers (ease to avoid port number conflicts) [MQ]
3499 * New! gras processes display their backtrace when they get a SIGUSR1
3500 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3501 Sweet to debug RL processes [MQ]
3505 - Do not force experiment sizes to be expressed in kb, or it becomes
3506 impossible to measure the latency this way (needs one byte-long tests)
3507 WARNING: this changes the amok_bw_* function semantic. [MQ]
3508 - Implements the link saturation stuff. [MQ]
3509 * Peer management module:
3510 New! module factorizing code that we wrote over and over [MQ].
3513 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3514 - Test units are placed directly into the library code, they get extracted
3515 automatically and placed into the src/testall binary.
3516 - Convert most of the XBT tests to this system.
3517 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3518 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3519 live in the same C code. This required to change a little bit the API
3520 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3521 * New module: sparse graph structure with basic algorithms (this is work
3522 in progress and the API is not considered to be frozen yet). [AL]
3523 * Display more information on backtraces: source line & function names are
3524 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3525 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3526 * Reworked a little bit some #include statements to load only required
3527 headers. Some user code that relied on SimGrid to include stdlib or
3528 stdio may need to include it by themselves. [AL]
3529 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3531 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3532 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3533 but I guess that almost nobody used those parts. [MQ]
3535 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3537 SimGrid (3.0.1) stable; urgency=low
3540 * Unfortunately, I had missed 5 misnamed functions:
3541 xbt_fifo_item_t xbt_fifo_newitem(void);
3542 void xbt_fifo_freeitem(xbt_fifo_item_t);
3543 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3544 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3545 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3546 They're now deprecated. Please use their new versions:
3547 xbt_fifo_item_t xbt_fifo_new_item(void);
3548 void xbt_fifo_free_item(xbt_fifo_item_t);
3549 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3550 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3551 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3553 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3554 * Documentation: xbt_log module unmercifully reworked [MQ]
3555 * Bugfix: there was a problem with the ending of contexts with
3556 the pthread backend. It caused some weird deadlock or behavior
3557 depending on the pthread implementation. [AL]
3558 * Bugfix: get the exceptions raised in the simulator repport where
3559 and why they come from when they are not caught in time [AL, MQ]
3562 * Bugfix: Do repport the error when two non-connected hosts try to
3563 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3566 * Add additionnal checkings on communications. Assert that two
3567 communicating hosts are connected by a set of links... [AL]
3570 * Add additionnal checkings on channel values in communication [AL]
3571 * New: MSG_task_get_source to see on which host a task was generated [HC]
3572 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3573 the number of tasks waiting to be received on channel and sent
3575 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3576 waits for the first task coming from a given host.. [AL]
3578 GRAS new functionnalities: [MQ]
3579 * Enhance the parsing macro to allow the size of multidimentional objects
3580 to be given thru annotations.
3581 * New example (and documentation): Matrix Multiplication a la RPC
3582 (as when I was young!) and fix a bunch of bugs found on the way.
3584 GRAS performance improvements: [MQ]
3586 * Reduce the amount of cbps creation/destruction by making it static to
3587 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3590 * Change libdata to a set so that we can search for stuff by ID (and thus
3591 reduce the insane amount of dict lookups)
3594 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3595 RL communication on top of SG since it's so uneffective.
3596 It may also be used for inter-thread communication in RL, one day.
3597 * Use gras_datadesc_copy() to exchange messages on top of SG
3599 - improve message exchange performance on top of SG
3600 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3601 * Don't exchange on the network the size of the used part of buffer,
3602 instead, specify the possible buffer size to read().
3604 - reduces the amount of read/write calls (one pair per exchange)
3605 - reduces the amount of exchanged data (the size)
3606 - allows to retrieve all arrived data on receiver side, if we don't need
3607 it right now (subsequent read will peek the buffer)
3608 - allows the receiver to proceed with the beginning of the stream before
3609 everything is arrived
3610 - make it possible to build an iov transport (using readv/writev)
3612 - take care of the data with non-stable storage (like stacked data),
3614 * If possible, TCP send uses vector I/O (when writev() is here)
3615 - Don't use it for receive since we send data sizes and data on the
3616 same stream, so we wouldn't be able to chain large amount of chunks
3617 before having to flush the stuff to read the size.
3618 * Rework the transport plugin mechanism to simplify it and reduce the
3619 amount of pointer dereferencement when searching for the right function
3622 * I guess that now, we do almost as few system calls as possible while
3623 doing as few data copy as possible.
3625 To improve it further, we could try to send all the sizes first and then
3626 all the data (to use iov on receiving size), but it's only a partial
3627 solution: when you have 2 dimensional data, the sizes of the second
3628 dimension is data of the first dimension, so you need 3 streams.
3630 I'm not sure the potential performance gains justify the coding burden.
3632 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3634 SimGrid (3.00) stable; urgency=high
3637 * New! Give the possibility to hijack the surf parser and thus bypass
3638 MSG_create_environment and MSG_launch_application. Have a look at
3639 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3641 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3643 SimGrid (2.96) unstable; urgency=low
3648 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3649 This deprecates the xbt_error_t mechanisms.
3650 It modifies (simplifies) all XBT and GRAS API.
3651 MSG API keeps unchanged (exceptions raised by XBT are caught from
3652 within MSG and masked with existing error handling facilities)
3655 * New! Add a FATPIPE model. [AL]
3656 * New! Add a parallel task model. [AL]
3657 * New! Add automatically a loopback interface (in the default
3658 network model) if none was precised.
3661 * Bugfix: MSG_process_resume now works with the current running process.
3663 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3664 * Modification of MSG_task_get_compute_duration. Once a task has been
3665 processed, the value returned by this function is now equal to 0. [AL]
3666 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3667 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3668 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3669 of a m_task. For now, MSG_task_cancel only works with computation
3671 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3672 of the processor (in Mflop/s) regardless of the current load on the
3674 * API Change: use proper naming convention for MSG_getClock and
3675 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3677 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3678 This function changes the priority of a computation task. This priority
3679 doesn't affect the transfer rate. A priority of 2 will make a task
3680 receive two times more cpu power than the other ones. This function
3681 has been added to suit the needs of Nguyen The Loc and hasn't been that
3682 much tested yet. So if it fails, please report it and send me your code.
3684 * API Change: removed all functions and types that were marked "deprecated"
3685 since many months. Renamed MSG_global_init_args to MSG_global_init.
3687 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3689 SimGrid (2.95) unstable; urgency=low
3692 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3693 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3694 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3697 * Bugfix: Allow absolute paths for platform description files [MQ]
3698 * Bugfix: do free the variables after use. Leads to drastic performance
3700 * Implement max_duration (ie, timeouts) on resources [AL]
3703 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3705 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3706 the needed support (timeouts) [AL]
3709 * Implement measurement sockets. You can now get the bandwidth between two
3710 hosts thanks to AMOK (see below). [MQ]
3711 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3712 dynar over the network (yeah) [MQ]
3713 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3714 * simple/static token-ring example. [Alexandre Colucci and MQ]
3715 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3716 in place before [MQ]
3719 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3720 type parsing. "short" and "long" modifiers were also ignored; other
3721 modifier (such as reference level) are still ignored. [MQ]
3722 * Embeed the buffer size within the buffer itself on SG. [MQ]
3723 That way, send() are atomic and cannot get intermixed anymore (at least
3724 the ones which are less than 100k; bigger messages still have the issue)
3725 * Array size pushed by the field, not by the field type (or each
3726 and every long int will push stuff to the cbps) [MQ]
3727 * use select() to sleep since it allows to portably sleep less than one
3730 GRAS (minor cleanups)
3731 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3734 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3735 * type category 'ignored' killed as it was never used and were difficult
3737 * whether a type can cycle or not is now a flag, leaving room for more
3738 flags (as "ignored", if we feel the need one day ;)
3739 * Rename raw sockets to measurement sockets since "raw" has another
3740 meaning in networking community.
3743 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3744 and offers features not belonging to GRAS but that most applications
3745 need. One day, it may be a set of plugins loadable at runtime.
3746 * New module: bandwidth
3747 bandwidth measurement between arbitrary nodes running this module. [MQ]
3749 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3751 SimGrid (2.94) unstable; urgency=low
3753 The first beta release of SimGrid 3 !
3757 * Update the main page and the FAQ. Adding references to gforge.
3760 * Add a gras_os_getpid function.
3763 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3764 * Extend the logs so that they also print PID, hostname, date, ... if
3766 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3767 and kill the old version which were in testsuite/
3768 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3772 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3773 for users. Instead, each programming environment (GRAS, MSG,...) use it
3774 when needed to provide such a feature to users.
3775 Don't shortcut the mechanism or you will also shortcut the virtualization
3776 you need on the simulator.
3780 * Cleanups in configury with regard to compile optimization/warning flags.
3781 Also add -fno-loop-optimize to any powerpc since it's the optimization
3782 killing gcc (< 3.4.0).
3783 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3784 needed by MSG examples complications
3785 * Borrow configury beautifications from PHP
3788 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3789 warning (thanks loris for stumbling into this one).
3790 * Bugfix: stop loading private headers (gras_config.h) from the public
3794 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3795 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3796 RL would imply a DNS resolver, which is *hard* to do in a portable way
3797 (and therefore delayed).
3798 * Implement a real timer mechanism and use it in timing macros. This allows
3799 to avoid rounding errors and get a 0.000005 sec precision in timing
3800 macros. While I was at it, various cleanups:
3801 - allow to declare more than one timed section per file (fix a stupid bug)
3802 - move some private declaration to the right place
3803 - merge conditional execution and timing macros into emulation module
3804 - document the module
3805 - make sure the module cleanups its mess on gras_exit
3806 * Documentation improvements:
3807 - (new) how to compile applications using GRAS
3808 - (new) emulation support (timing macros)
3810 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3812 SimGrid (2.93) unstable; urgency=low
3814 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3817 - Use Paje properly where used. Still to be sanitized properly.
3818 - Portability fix: Add an implementation of the contexts using pthread
3822 - Add xbt_procname(): returns the name of the current process.
3823 Use it to show the current process's name in all logging.
3825 - fix detection of older flex version and the reaction, since we do
3826 depend on modern ones (we use lex_destroy)
3827 - Better separation of SG and RL in the libs: remove all simulation code
3828 from libgras. As a result, this lib is now only 200k when stripped.
3829 Some of the xbt modules may also be duplicated (two sets and such) and
3830 should be cleaned/killed before SG3.
3831 - Insist on using xlC on AIX because of weird problems involving gcc there.
3832 - Cleanup the make remote stuff. This is now done by scripts
3833 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3834 mainly for our private use, but we're working on changing them to user
3837 - Bugfix: flush the socket on close only if there is some *output*.
3838 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3840 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3843 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3845 SimGrid (2.92) unstable; urgency=low
3847 Alpha 3 on the path to SimGrid 3
3851 - New! First try of benchmarking macros.
3852 - New! First try so that gras_stub_generator generate deployment and
3853 remote compilation helpers.
3855 - Bugfix: Initialization fix in msg_test.
3859 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3861 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3863 - API Change: gras_os_sleep to take the amount of seconds as a double.
3864 Accepting an int was error prone since it was the only location where
3865 seconds were coded as such. It leaded to damn rounding errors.
3866 - Bugfix: Hard to belive that timers ever worked before this.
3868 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3870 SimGrid (2.91) unstable; urgency=low
3872 Alpha 2 on the path to SimGrid 3
3876 - Bug fix in the lmm_solver.
3878 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3879 through the function MSG_paje_output.
3880 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3881 - It is possible to bound the rate of a communication in MSG with
3882 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3883 to write it in the changelog).
3884 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3889 - Various cleanups to the autotools stuff
3890 - Begin to move Gras examples to examples/gras/
3891 - Let make distcheck work again (yeah!)
3893 - documentation overhauled using doxygen.
3894 gtk-doc-tools is dead in SimGrid now.
3895 - Automatically extract all existing logging categories, and add the list
3896 to the documentation (long standing one, to say the less)
3898 - Cleanup the known architecture table. Reorder the entries to group what
3899 should be, and use a more consistent naming scheme.
3900 (some of the test dataset are still to be regenerated)
3901 - New! Allow library to register globals on each process just as userdata
3903 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3904 lookup time (for now).
3905 Use it in msg and trp.
3906 This cleans a lot the internals and helps enforcing privacy of the
3907 headers between the gras components.
3908 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3909 - Bugfix: gras_os_time was delirious in RL.
3910 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3912 - Reenable GRAS now that it works.
3914 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3916 SimGrid (2.90) unstable; urgency=low
3918 Alpha 1 on the path to SimGrid 3
3920 * It is a long time since the last release of SimGrid. I'm sorry about
3921 that but as I had told you, I was rewriting a lot of things. I apologize
3922 to those who had been reporting bugs to me and that I had not answered.
3923 If your bug is still in the new version, please tell me. Here is a
3924 summary of the main changes.
3926 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3927 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3928 portability, speed, and a lot more but you'll figure it out later.
3929 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3930 I wanted to release the new MSG as soon as possible and I have
3931 broken GRAS, which is the reason why, for now, only MSG is fully
3932 functional. A laconic description of these projects is available
3933 in the documentation.
3935 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3936 kernel optimized for our needs (called SURF but only the developers
3937 should use it). Hence, MSG is now roughly 30 times faster and I think
3938 that by rewriting a little bit MSG, I could event speed it up a little
3939 bit more. Beside the gain in speed, it is also much easier to encode a
3940 new platform model with SURF than it was with SG. More to come...
3942 * REVOLUTION 3: I have tried to change a little as possible the API of
3943 MSG but a few things really had to disappear. The main differences
3944 with the previous version are:
3945 1) no more m_links_t and the corresponding functions. Platforms are
3946 directly read from an XML description and cannot be hard-coded
3947 anymore. The same format is used for application deployment
3948 description. The new format is described in the documentation.
3949 Have a look in tools/platform_generation. There is a tiny script
3950 that converts from the old platform format to the new one. Concerning
3951 the application deployment format, parsing the old one is tricky.
3952 I think most of you should however be able to convert your files. If
3953 it is really an issue, I can write a C code that does the conversion.
3955 2) the toolbox tbx does not exist anymore. We now have a library
3956 with much more data-structures but without the hash-tables (we have
3957 dictionaries that are much faster).
3959 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3961 *****************************************************************************
3962 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3963 * don't feel like converting the oldies. (Mt) *
3964 *****************************************************************************
3967 Version 2.90: "the long awaited one"
3968 - Finished rewriting and debugging MSG. Rewrote the documentation.
3969 - disable GRAS for now since it needs to be ported to the newest SG
3972 - Finish the port to windows (using mingw32 for cross-compile)
3975 - Main loop and datastructures of SURF. A cpu resource object is
3976 functional. Surf can thus be used to create cpu's with variable
3977 performance on which you can execute some actions.
3979 2004-11-15 Martin Quinson
3980 - Port to ARM. Simply added the alignment and size descriptions. Should
3981 work, but the ARM machines are so slow that I didn't had the opportunity
3982 to 'make check' over there yet.
3984 2004-11-15 Arnaud Legrand
3985 - Trace manager now written. It uses a heap structure and is therefore
3986 expected to be efficient. It may however be speeded up (particularly
3987 when many events occur at the same date) by using red and black
3988 trees. One day maybe...
3989 - Max-min linear system solver written. It uses a sparse matrix
3990 structure taking advantage of its expected use. Most operations are
3991 O(1) and free/calloc are called as few as possible. The computation of
3992 the minimum could however be improved by using a red and black tree
3995 2004-11-03 Arnaud Legrand
3996 - Rename every gras_* function that was in xbt/ to its xbt_
3998 - Add a heap and a doubly-linked list to xbt
3999 - Added a dichotomy to the dictionaries. make check works as well before
4000 so I assume that the patch is correct. I do not know however if things
4001 run effectively faster than before now. :)
4003 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
4004 SimGrid, and the version number is bumped to 2.x
4006 2004-10-29 Martin Quinson
4007 - Introduction of the remote errors.
4008 They are the result of a RMI/RPC on the remote machine.
4009 ErrCodes being scalar values, you can't get the host on which those
4010 errors did happen. Extending the error mechanism as in Gnome is possible.
4011 No idea yet whether it is a good idea.
4013 2004-10-28 Martin Quinson
4014 - Interface revolution: the Starred Structure Eradication.
4015 I used to do typedef struct {} toto_t; and then handle *toto_t.
4016 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
4017 - ???_t is a valid type (builded with typedef)
4018 - s_toto_t is a structure (access to fields with .)
4019 - s_toto is a structure needing 'struct' keyword to be used
4020 - e_toto_t is an enum
4021 - toto_t is an 'object' (struct*)
4023 typedef struct s_toto {} s_toto_t, *toto_t;
4024 typedef enum {} e_toto_t;
4025 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
4026 s_toto_t) is private.
4028 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
4029 so that it can set the variable to NULL. It was so for dicts and sets,
4030 it changed for dynars.
4032 - Fix a bunch of memleaks in dict_remove
4033 - Fix a bug in sg/server_socket opening: it failed all the time.
4035 2004-10-07 Martin Quinson
4036 - Speed up dynar lookup operation a bit.
4038 gras_dynar_get is dead.
4040 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
4041 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
4042 (which gives you the address of the stored data).
4044 gras_dynar_get_as is an helpful macro which allows you to retrieve a
4045 copy of the data using an affectation to do the job and not a memcpy.
4047 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
4048 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
4050 It does not really speedup the dynar test because they are
4051 setting elements all the time (and look them seldom). But the dict does
4052 far more lookup than setting.
4054 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
4056 2004-10-05 Martin Quinson
4057 - Allow to (en/dis)able the cycle detection at run time.
4059 Whether we should check for cycle or not is now a property of each
4060 datatype. When you think there may be some cycle, use datadesc_cycle_set.
4061 datadesc_cycle_unset allow to remove this property when previously set.
4063 Note that the cycle detection is off by default since it impacts the
4064 performance. Watch the data you feed GRAS with ;)
4066 This property is hereditary. Any element embedded in a structure having it
4067 set have it set for the time of this data exchange.
4069 You should set it both on sender and receiver side. If you don't set it on
4070 sender side, it will enter an endless loop. If you forget on receiver
4071 side, the cycles won't be recreated after communication.
4073 - Header reorganization.
4074 Kill gras_private.h, each submodule must load the headers it needs.
4076 2004-10-04 Martin Quinson
4077 - Interface revolution: do not try to survive to malloc failure.
4079 Now, gras_malloc and friends call gras_abort() on failure.
4080 As a conclusion, malloc_error is not a valid error anymore, and all
4081 functions for which it was the only gras_error_t return value are
4082 changed. They now return void, or there result directly.
4083 This simplify the API a lot.
4085 2004-09-29 Martin Quinson
4086 - Re-enable raw sockets.
4087 Created by gras_socket_{client,server}_ext;
4088 Used with gras_raw_{send,recv}
4091 It should allow to kill the last bits of gras first version soon.
4093 This is not completely satisfactory yet (duplicate code with
4094 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
4097 - Simplify transport plugin (internal) interface by not passing any
4098 argument to _server and _client, but embedding them in the socket
4101 2004-09-28 Martin Quinson
4102 - Finish the port to AIX.
4103 autoconf was my problem (segfault within the malloc replacement
4104 function. No idea why)
4106 2004-09-16 Martin Quinson
4107 - Fix some size_t madness on 64bit architectures.
4109 2004-09-08 Martin Quinson
4110 - Reduce the number of system headers loaded, overload some more system
4111 calls (such as malloc to cast the result of the system one, and work
4113 - Fix and reintroduce the config support
4115 2004-09-07 Martin Quinson
4116 - Source code reorganization to allow Arnaud to surf all over there.
4117 - Allow to document the logging categories.
4118 - Remove all uppercase from logging categories and useless cleanup in names.
4120 2004-08-18 Martin Quinson
4121 Version 0.6.2 (protocol not changed; API changed)
4122 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4123 gras_error_t), and NULL when not found. Functions expecting a msgtype
4124 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4125 hopefully usefull message.
4126 - Portability to prehistoric sparcs again
4128 2004-08-17 Martin Quinson
4129 Version 0.6.1 (protocol not changed; ABI not changed)
4130 - prealloc some buffers to speed things up
4132 2004-08-11 Martin Quinson
4133 Version 0.6 (protocol not changed; ABI expended)
4134 - The parsing macro can deal with the references, provided that you add
4135 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4137 2004-08-09 Martin Quinson
4138 Version 0.5 (protocol not changed; ABI changed)
4139 - Allow to off turn the cycle detection code in data exchange at
4140 compilation time. It should be at run time, but I'm short of time (and
4141 the config stuff is still broken). That way, we keep dict out of the
4142 critical path, which is good because the performance is poor:
4143 - search not dichotomial yet
4144 - dynar give no way to access their content and memcpy everytime
4145 - In composed data description (struct, ref and so on), stop foolness of
4146 keeping the subtype's ID, but store the type itself. This keeps sets out
4147 of the critical path, which is good since they rely on dynar and
4148 dictionaries. The only loose of that is that we cannot detect the
4149 redeclaration of a structure/union with another content (but I'm not sure
4150 the code detected well this error before anyway). We still can detect
4151 the redefinition discrepancy for the other types.
4152 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4153 it breaks the code because of type-punning used all over the place).
4154 This breaks on all non-gcc architectures (for now).
4156 All those changes (plus the buffer of last time) allow me to gain 2 order
4157 of magnitude on cruel tests consisting of 800000 array of integers on two
4158 level of a hierarchical structure (200 secondes -> 4 secondes)
4161 - the selector of reference must now return the type it points to, not
4162 the ID of this type.
4164 2004-08-06 Martin Quinson
4165 Version 0.4 (protocol changed; ABI not changed)
4166 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4167 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4168 - Implement a buffer transport, and use it by default (it relies on tcp in
4169 real life and on sg in simulation).
4170 That's a bit hackish since I had a new field to the structure to store
4171 its data without interfering with the subtype ones. Inheritance
4172 is tricky in C. And that's a kind of reverse inheritance with one class
4173 derivating two classes. Or maybe a game with java interfaces. Anyway,
4174 that's damn hard in C (at least).
4175 Moreover, I got tired while trying to ensure plugin separation and
4176 genericity in SG mode. MSG wants me to do weird things, so let's go for
4177 cruel hacks (temporarily of course ;).
4178 See comment in transport_private.h:71
4179 - do not include all the _interface headers in private but in the files
4180 which really need them (to cut the compilation time when they are
4183 2004-07-26 Martin Quinson
4184 Version 0.3 (protocol not changed; ABI changed)
4185 - Major overhault of the datadesc interface to simplify it:
4186 - shorted the function names:
4187 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4188 - add a trivial way to push/pop integers into the cbps without malloc.
4189 This allows to make really generic sub_type description, which simply
4190 pop their size of the stack.
4191 - add a function gras_datadesc_ref_pop_arr() which does what users want
4192 most of the time: Declare a dynamic array (which pops its size of the
4193 stack) and declare a reference to it. Poor name, but anyway.
4194 - kill the post-send callback, add a post-receive one
4196 2004-07-23 Martin Quinson
4197 Version 0.2 (protocol changed; ABI changed)
4198 - add some testing for cpbs in the test cases, and fix some more bugs.
4199 This invalidate again the little64 data file, since I cannot regenerate
4201 - remove an awfull optimization in the logging stuff, allowing me to:
4202 - understand it again
4203 - learn gcc how to check that the argument match the provided format
4204 - fix all errors revealed by gcc after that
4205 - internal keys of dict are not \0 terminated. Deal with it properly in
4206 loggings instead of segfaulting when the user want to see the logs :-/
4208 2004-07-22 Martin Quinson
4209 - Fix some stupid bug preventing cbps (callback postit) from working
4211 2004-07-21 Martin Quinson
4212 - Some documentation cleanups
4213 - remove the useless last argument of msgtype_declare
4214 - rename the Virtu functions to fit into the 'os' namespace
4215 - move headers src/include -> src/include/gras/ and stop fooling with
4216 gras -> . symbolic link
4217 - make distcheck is now successful
4219 2004-07-19 Martin Quinson
4221 - Build shared library also
4222 - Install html doc to the right location
4223 - stop removing maintainer files in make clean
4224 - build tests only on make check
4226 2004-07-13 Martin Quinson
4228 - No major issue in previous version => change versioning schema
4229 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4230 corresponding dataset.
4232 2004-07-11 Martin Quinson
4234 - Get it working with any kind of structure (we can compute the padding
4235 bytes remotely for all the architectures I have access to)
4236 - Implement the structure parsing macro (still not quite robust/complete)
4237 - Improvement to the remote testing toysuite
4239 2004-07-10 Martin Quinson
4240 [autoconf mechanism]
4241 - get ride of a bunch of deprecated macros
4242 - actually run the test for two-compliment, not only compile it :-/
4243 - test whether the structures get packed (and bail out if yes. Damn.
4244 Alignment is a serious matter)
4245 - test whether the structures get compacted (but respecting the alignment
4246 constraints of each types)
4247 - test whether the array fields of structures can straddle alignment boundaries
4249 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4250 (took me 2 hours to find that bug, looking at the wrong place)
4251 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4252 before sending/receiving and are used to compute the offsets of fields
4253 - Given that padding size depend even on compiler options, keep track of
4254 alignment and aligned_size only for the current architecture. Not a big
4255 deal since we send structure fields one after the other (seems
4257 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4258 program, let it work on linux/gcc/little32. portability todo.
4260 2004-07-08 Martin Quinson
4261 - import and improve remote compilation support from FAST
4262 - make sure make check works on half a dozen of machines out there
4264 2004-07-07 Martin Quinson
4265 Let's say it's version 0.0.3 ;)
4266 - Implement conversions (yuhu!)
4267 - Let it work on solaris (beside conversion, of course)
4268 - Stupid me, using rand() to generate the conversion datatests in not wise.
4270 2004-07-06 Martin Quinson
4271 - Let make dist work, since I'm gonna need it to compile on remote hosts
4272 - Let Tests/datadesc_usage write the architecture on which the file was
4273 generated as first byte.
4274 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4275 architecture descriptions.
4276 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4277 on those architectures.
4278 - Optimization: send/recv array of scalar in one shoot
4280 2004-07-05 Martin Quinson
4281 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4283 - Plug a whole bunch of memleaks
4284 - each process now have to call gras_{init,exit}. One day, their log
4285 settings will be separated
4286 - Continue the code factorisation between SG, RL and common in Transport.
4288 2004-07-04 Martin Quinson
4290 - Redistribution between SG and RL.
4291 We wanna have to accept in SG, so move accepted related parts of RL in
4292 the common part. (more precisely, the dynar of all known sockets is no
4293 more a static in transport.c, but part of the process_data)
4295 [gras_stub_generator]
4296 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4298 2004-07-03 Martin Quinson
4299 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4300 them all up in one shot)
4302 - Fix the ugly case of reference to dynamic array.
4303 - New (semi-public) function gras_datadesc_size to allow the messaging
4304 layer to malloc the needed space for the buffer.
4306 - gras_socket_close now expect the socket to close (and not its address to
4307 put NULL in it after it). This is because the socket passed to handlers
4308 is one of their argument (=> not writable).
4310 - propagate the interface cleanup from last week in datadesc, ie remove a
4311 superfluous level of indirection. User pass adress of variable
4312 containing data (both when sending and receiving), and not of a variable
4313 being a pointer to the data. Let's say that I like it better ;)
4314 The price for that is constructs like "int msg=*(int*)payload" in
4315 handlers, but it's a fine price, IMHO.
4317 - Let it work in RL (yuhu)
4319 2004-06-21 Martin Quinson
4321 - porting SG plugin and SG select to new standards (works almost).
4322 - plug memleaks and fix bugs around.
4325 - cleanup the prototype of data recv and force users to specify when they
4326 want to handle references to objects. Test case working even for cycles.
4327 - plug memleaks. Valgrind is perfectly ok with this.
4329 2004-06-12 Martin Quinson
4331 - cleanup the separation between plugin and main code in plugin creation
4333 2004-06-11 Martin Quinson
4335 - Reput hook for raw sockets, needed for BW experiments
4336 - kill a few lines of dead code
4337 [Data description] Interface cleanup
4338 - gras_datadesc_by_name returns the searched type or NULL.
4339 That way, no variable is needed to use a type desc once, which makes
4341 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4342 two parameters were strings (field name, type name), leading to
4344 [Dicos] Interface cleanup
4345 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4346 This is consistant with the dynar API.
4348 2004-04-21 Martin Quinson
4350 - Porting to new standards.
4352 - interface cleanup.
4353 There is no bag anymore, no need to take extra provision to mask the
4354 pointers behind "ID".
4355 Better splitup of functions between files create/exchange/convert.
4356 This is still a bit artificial since convert and receive are so
4357 interleaved, but anyway.
4359 - add a queued message list to procdata (the ones not matching criteria
4361 - factorize some more code between SG and RL wrt procdata
4363 - use gras_exit in example to track memleaks
4364 - get rid of gs_example now that GS is properly integrated into gras
4365 - update run_test to integrate the latest tests (datadesc)
4367 - rename WARNINGn macros to WARNn since it prooved error-prone
4369 2004-04-19 Martin Quinson
4371 - register init/exit functions within gras module mechanism
4372 - send/receive function.
4373 Convertion is not implemented, but short-cutted if not needed.
4374 struct/array elements are sent one by one (instead of block-wise), but
4375 nobody really cares (yet). Get a prototype before optimizing.
4376 - tests (using a file socket) for DD send/receive on:
4377 - base types: int, float
4378 - array: fixed size, string (ie ref to dynamic string)
4379 - structure: homogeneous, heterogeneous
4380 - chained list, graph with cycle
4381 Believe it or not, valgrind is not too unhappy with the results. The
4382 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4386 - Bugfix when using a filename explicitly (instead of '-')
4388 2004-04-09 Martin Quinson
4390 - factorize more code between RL and SG in socket creation
4391 - Complete the implementation and tests of:
4393 o file (only in RL, and mainly for debugging)
4395 I lost 3 days to design a portable address resolver, and then decided
4396 that the prototype mainly have to run on my box.
4397 Addressing portability too early may be like optimizing too early :-/
4399 - use gras_init in the Tests instead of the crappy parse_log_opt
4400 (the latter function is removed)
4401 [Conditional execution]
4402 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4403 [Code reorganisation]
4404 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4405 Build examples against the RL library, since there is no way to disable
4406 its creation for now.
4408 For information, the beginning of coding on GRAS was back in june
4409 2003. I guess that every line has been rewritten at least twice since